Handbook of Anatomical Models for Radiation Dosimetry (Series in Medical Physics and Biomedical Engineering)

Handbook of Anatomical Models for Radiation Dosimetry Series in Medical Physics and Biomedical Engineering Series Edi...

Author: Xie George Xu | Keith F. Eckerman

86 downloads 1291 Views 16MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Handbook of Anatomical Models for Radiation Dosimetry

Series in Medical Physics and Biomedical Engineering Series Editors: John G Webster, E Russell Ritenour, Slavik Tabakov, and Kwan-Hoong Ng Other recent books in the series: Fundamentals of MRI: An Interactive Learning Approach Elizabeth Berry and Andrew J Bulpitt Handbook of Optical Sensing of Glucose in Biological Fluids and Tissues Valery V Tuchin (Ed) Intelligent and Adaptive Systems in Medicine Oliver C L Haas and Keith J Burnham A Introduction to Radiation Protection in Medicine Jamie V Trapp and Tomas Kron (Eds) A Practical Approach to Medical Image Processing Elizabeth Berry Biomolecular Action of Ionizing Radiation Shirley Lehnert An Introduction to Rehabilitation Engineering R A Cooper, H Ohnabe, and D A Hobson The Physics of Modern Brachytherapy for Oncology D Baltas, N Zamboglou, and L Sakelliou Electrical Impedance Tomography D Holder (Ed) Contemporary IMRT S Webb The Physical Measurement of Bone C M Langton and C F Njeh (Eds) Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine H Zaidi and G Sgouros (Eds) Minimally Invasive Medical Technology J G Webster (Ed) Intensity-Modulated Radiation Therapy S Webb

Physics for Diagnostic Radiology, Second Edition P Dendy and B Heaton

Series in Medical Physics and Biomedical Engineering

Handbook of Anatomical Models for Radiation Dosimetry

Edited by

Xie George Xu Rensselaer Polytechnic Institute Troy, New York, USA

Keith F. Eckerman Oak Ridge National Laboratory Tennessee, USA

Boca Raton London New York

CRC Press is an imprint of the Taylor & Francis Group, an informa business

A TA Y L O R & F R A N C I S B O O K

Taylor & Francis 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 © 2010 by Taylor and Francis Group, LLC Taylor & Francis is an Informa business No claim to original U.S. Government works Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number: 978-1-4200-5979-3 (Hardback) This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint. Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright.com (http:// www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging-in-Publication Data Handbook of anatomical models for radiation dosimetry / editors, Xie George Xu and Keith F. Eckerman. p. ; cm. -- (Series in medical physics and biomedical engineering) Includes bibliographical references and index. ISBN 978-1-4200-5979-3 (hardcover : alk. paper) 1. Radiation dosimetry--Mathematical models. I. Xu, Xie George. II. Eckerman, K. F. III. Series: Series in medical physics and biomedical engineering. [DNLM: 1. Phantoms, Imaging. 2. Computer Simulation. 3. Models, Anatomic. 4. Radiation Dosage. 5. Radiometry--instrumentation. WN 150 H236 2010] R905.H36 2010 612’.014480287--dc22 Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com and the CRC Press Web site at http://www.crcpress.com

2009012396

Contents Series Preface ..................................................................................................................................ix Preface ..............................................................................................................................................xi Editors .............................................................................................................................................xv Contributors ................................................................................................................................ xvii

Part I

Phantoms

1. Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution ............................................................................................3 X. George Xu 2. Stylized Computational Phantoms Developed at ORNL and Elsewhere .......................................................................................................................43 Keith F. Eckerman, John W. Poston, Sr., Wesley E. Bolch, and X. George Xu 3. The GSF Voxel Computational Phantom Family ...........................................................65 Maria Zankl 4. The ADELAIDE Teenage Female Voxel Computational Phantom ............................87 Martin Caon, Giovanni Bibbo, and John E. Pattison 5. The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms ........................................................................................................105 W. Paul Segars and Benjamin M.W. Tsui 6. The 3D and 4D VIP-Man Computational Phantoms ..................................................135 X. George Xu, Tsi-Chian Ephraim Chao, Ahmet Bozkurt, Chengyu Shi, and Juying Zhang 7. The FAX06 and the MAX06 Computational Voxel Phantoms ...................................163 Richard Kramer, Helen Jamil Khoury, José Wilson Vieira, Vanildo Júnior de Melo Lima, Eduardo César de Miranda Loureiro, Gabriela Hoff, and Iwan Kawrakow 8. The University of Florida Pediatric Phantom Series ..................................................199 Choonsik Lee, Daniel L. Lodwick, Deanna Hasenauer Pafundi, Scott R. Whalen, Jonathan L. Williams, and Wesley E. Bolch 9. Japanese Computational Phantoms: Otoko, Onago, JM, JM2, JF, TARO, HANAKO, Pregnant Woman, and Deformable Child ..................................221 Kimiaki Saito, Kaoru Sato, Sakae Kinase, and Tomoaki Nagaoka

v

Contents

vi

10. Korean Computational Phantoms: KMIRD, KORMAN, KORWOMAN, KTMAN-1, KTMAN-2, and HDRK-Man .........................................255 Choonsik Lee and Chan Hyeong Kim 11. Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP ....................................................................................................................279 Binquan Zhang, Jizeng Ma, Guozhi Zhang, Qian Liu, Rui Qiu, and Junli Li 12. Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series Representing 3, 6, and 9 Month Gestational Periods ......................................305 X. George Xu, Chengyu Shi, Michael G. Stabin, and Valery Taranenko 13. The Vanderbilt University Reference Adult and Pediatric Phantom Series ...................................................................................................................337 Michael G. Stabin, Mary Ann Emmons-Keenan, W. Paul Segars, and Michael J. Fernald 14. Mesh-Based and Anatomically Adjustable Adult Phantoms and a Case Study in Virtual Calibration of a Lung Counter for Female Workers ........347 Yong Hum Na, Juying Zhang, Aiping Ding, and X. George Xu 15. The ICRP Reference Computational Phantoms ...........................................................377 Maria Zankl, Keith F. Eckerman, and Wesley E. Bolch 16. Physical Phantoms for Experimental Radiation Dosimetry .....................................389 David E. Hintenlang, William E. Moloney, and James Winslow

Part II

Applications

17. Applications to Environmental Exposures ...................................................................413 Nina Petoussi-Henss and Kimiaki Saito 18. Applications to External Radiation Exposures in Nuclear Power Plants ...............425 Warren Dan Reece, Chan Hyeong Kim, and X. George Xu 19. Applications to Bioassay for Internal Radiation Contamination .............................449 Gary H. Kramer 20. Applications to Nuclear Medicine ..................................................................................471 Michael G. Stabin and Manuel Bardiès 21. Applications to Computed Tomography for Pediatric Patients ................................487 Wesley E. Bolch, Choonsik Lee, Choonik Lee, Jorge Hurtado, and Jonathan L. Williams 22. Applications to Computed Tomography for Adult Patients ......................................511 John J. DeMarco and Michael McNitt-Gray

Contents

vii

23. Applications to Optimization of X-Ray Radiographic Imaging ...............................525 Birsen Yazıcı, Il-Young Son, An Jin, and X. George Xu 24. Applications to Nuclear Medicine Imaging and Dosimetry Involving MCAT, NCAT, and MOBY Phantoms .........................................................549 Benjamin M.W. Tsui and W. Paul Segars 25. Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy ..................................................................................................567 Harald Paganetti 26. Applications to Image-Guided Radiation Treatment Planning ...............................591 Chengyu Shi, Martin Fuss, Niko Papanikolaou, and X. George Xu 27. Dose Calculations in Radiation Therapy Based on Patient Models Using the Geant4 Monte Carlo Code ..............................................................................607 Harald Paganetti 28. Applications to Patient-Specific Voxel Computational Phantoms in EGS Monte Carlo Codes for Radiation Treatment Involving Photons and Electrons .......................................................................................................633 C.-M. Charlie Ma 29. Applications to Nonionizing Radiation Protection ....................................................655 Ji Chen, Wolfgang Kainz, and Dagang Wu 30. Summary and Future Needs Related to Computational Phantoms.........................679 X. George Xu, Michael G. Stabin, Wesley E. Bolch, and W. Paul Segars About the Contributors ............................................................................................................685 Index .............................................................................................................................................705

Series Preface The International Organization for Medical Physics The International Organization for Medical Physics (IOMP), founded in 1963, is a scientific, educational, and professional organization of 76 national adhering organizations, more than 16,500 individual members, several corporate members, and four international regional organizations. The IOMP is administered by a council, which includes delegates from each of the adhering national organizations. Regular meetings of the council are held electronically as well as every three years at the World Congress on Medical Physics and Biomedical Engineering. The president and other officers form the executive committee, and there are also committees covering the main areas of activity, including education and training; scientific, professional relations; and publications. The objectives of the IOMP are • To contribute to the advancement of medical physics in all its aspects • To organize international cooperation in medical physics, especially in developing countries • To encourage and advise on the formation of national organizations of medical physics in those countries that lack such organizations Activities The official journals of the IOMP are Physics in Medicine and Biology, Medical Physics, and Physiological Measurement. The IOMP publishes a bulletin, Medical Physics World, twice a year, which is distributed to all members. A World Congress on Medical Physics and Biomedical Engineering is held every three years in cooperation with the International Federation for Medical and Biological Engineering (IFMBE) through the International Union for Physics and Engineering Sciences in Medicine. A regionally based International Conference on Medical Physics is held between the World Congresses. The IOMP also sponsors international conferences, workshops, and courses. IOMP representatives contribute to various international committees and working groups. The IOMP has several programs to assist medical physicists in developing countries. The joint IOMP Library program supports 69 active libraries in 42 developing countries and the Used Equipment Program coordinates equipment donations. The Travel Assistance Program provides a limited number of grants to enable physicists to attend the World Congresses. The IOMP Web site is being developed to include a scientific database of international standards in medical physics and a virtual education and resource center. Information on the activities of the IOMP can be found on its Web site at www.iomp.org.

ix

Preface Since the 1960s, the radiological science community has developed and applied more than 90 computational models of the human body—often referred to as “phantoms”—for ionizing radiation dosimetry studies. Each of the models not only defines the exterior features of the entire human body, but also includes extensive details on internal organs such as volume, mass, shape, and tissue composition. These computational phantoms are combined with Monte Carlo methods to precisely simulate complex radiation interactions and energy depositions in the human body involving various particles such as photons (x-rays and gamma rays), electrons, neutrons, and protons. Organ dose estimates, often normalized by measurable parameters, have been calculated for different irradiation scenarios found in occupational radiation protection, nuclear medicine, diagnostic imaging, and radiotherapy. Over the years, data derived from these computational phantoms have been adopted into the recommendations of the International Commission on Radiological Protection (ICRP) and other international and national bodies. Anthropomorphic phantoms of adults and children of various ages, as well as pregnant women, were depicted using increasingly sophisticated solid-geometry modeling techniques over the past 40 years. Early computational phantoms were based on simple quadric equations. Voxelized phantoms, which took advantage of medical imaging advances, started to emerge 20 years ago. These image-based phantoms brought an excitement to the research community because of their anatomical realism. In recent years, phantoms involving advanced geometries, such as the nonuniform rational B-splines (NURBS) and polygonal meshes, were reported with unprecedented capabilities such as deformable anatomy and real-time cardiac and respiratory motion simulations. In addition, a number of computational phantoms have been developed for studies involving nonionizing radiation—radio frequencies such as those emitted by electric power lines and wireless cellular phone technologies. These nonionizing radiation phantoms have similar anatomical features, as well as technical challenges, as those used for ionizing radiation dosimetry. Many such computational phantoms have been used for both fields of studies. For the first time, this book provides a comprehensive review of the historical development and application of a large number of important computational phantoms that have been widely reported in the literature. The history of computational phantoms is clearly in parallel with, and thus offers a unique perspective about, advances in computer technologies and medical imaging such as computed tomography and magnetic resonance imaging. By reading this book, the reader can obtain a unique sense of the scientific process in computational phantom development: the conception of an idea, the identification of original anatomical data, solutions of various computing problems, ownership and sharing of results, as well as the satisfaction and frustration associated with any scientific endeavor. This handbook contains 30 chapters and is the result of several years of planning and preparation involving ultimately 64 authors from 13 countries and regions. The idea of this book was first conceived during the Monte Carlo 2005 Topical Meeting in Chattanooga, Tennessee, April 17–21, 2005. A special session on “Tomographic Models for Radiation Protection Dosimetry” was attended by more than a dozen invited speakers worldwide. Recognizing the needs for research collaboration and dissemination, the session attendees strongly suggested two actions. The first was to form the Consortium of Computational Human Phantoms (CCHP) and a portal site for information related to xi

xii

Preface

computational human phantoms (www.virtualphantoms.org). The second was to initiate this collaborative book project to document a 40-year history of research and development. Previously, two workshops on “voxelized phantoms” were held: one at the National Board Radiological Protection, U.K., by Peter Dimbylow in 1995 that resulted in a compilation of presentations and the other at the Oak Ridge National Laboratory by Keith Eckerman in 2000. Despite the large number of phantoms that have been developed over the past four decades, information has been scattered and often lacking in detail, and certain data from the early work were difficult to locate. In the early 2000s, the ICRP decided to adopt voxelized computational phantoms as standards for radiation protection purposes. As a result, many colleagues had been actively involved for several years in activities associated with the ICRP Reference Computational Phantoms. These experiences accelerated our plan for this book, leading to the acceptance of the proposal by the publisher in 2007. It was clear to us that this book should possess the necessary depth and breadth by considering as broadly as possible phantoms that were historically important. At the same time, this book should also include applications of these phantoms in diverse radiological studies. We are extremely pleased that nearly all major phantom developers around the world accepted our invitation to contribute, except for a few colleagues whose busy schedules prevented them from participating—particularly Peter Dimbylow (for the NORMAN phantom) and George Zubal (for the “Zubal” phantom). It is obvious that we were unable to extend our invitation to many other researchers whose works are included in the tables of Chapter 1. Furthermore, there is a chance that we may have missed a few phantoms despite an exhaustive literature search. If your work is not mentioned anywhere in this book, please contact us so we can update our database that is maintained from the CCHP Web site. This book is divided into two main parts: the phantoms and their applications. Part I starts with Chapter 1, which provides a review of 40 years of research and development in computational phantoms. This chapter includes a discussion on the classification of phantoms and a comprehensive listing of computational and physical phantoms used for a variety of ionizing and nonionizing radiation applications. Several rare phantoms were included, such as the CAM phantom developed in 1973 for space radiation dosimetry. Chapters 2 through 15 provide detailed accounts for each of the well-known phantoms, such as the MIRD-5, GSF Voxel Family Phantoms, NCAT, the UF Hybrid Pediatric Phantoms, VIP-Man, as well as the latest ICRP Reference Phantoms. Chapter 11, which was contributed by three Chinese groups, details several phantoms, including one that, at the time of writing, has the smallest voxel size (0.2 mm), developed from the Chinese Visible Human Project. Chapter 16 is the final chapter in Part I, and it summarizes physical phantoms for experimental radiation dosimetry. In Part II, Chapters 17, 18, and 19 cover applications for radiation protection dosimetry involving environmental, nuclear power plant, and internal contamination exposures, respectively. These are followed by medical applications in Chapters 20 through 28, covering topics such as nuclear medicine therapy, CT examinations of pediatric and adult patients, x-ray radiological image optimization, nuclear medicine imaging, external photon and proton treatments, and management of respiration in modern image-guided radiation treatment. Chapters 27 and 28 deal with patient-specific phantoms used for radiation treatment planning involving two Monte Carlo code systems: GEANT4 and EGS, respectively. Applications for nonionizing radiation are described in Chapter 29. Finally, Chapter 30 discusses future needs for research and development. To supplement the information in this book, the following related data sets can be downloaded from http://files. virtualphantoms.org/public/FreeDownloads: (1) the VIP-Man image data; (2) Monte Carlo

Preface

xiii

N-Particle (MCNP) input file for external photon beams source geometry using the VIPMan phantom; and (3) presentation slides. An ambitious project like this is impossible without the generous support from the leading contributors for each chapter. Their time, effort, and patience are greatly appreciated. The field of computational phantoms has a long and rich history, and the research community is deeply indebted to the visionary work by many pioneers. As editors, we would like to thank Binquan Zhang, Juying Zhang, Matthew Mille, and Paul Booth from Rensselaer Polytechnic Institute, who provided valuable assistance in handling, reviewing, and formatting the electronically submitted manuscripts. X. George Xu Keith F. Eckerman

Editors

Dr. X. George Xu is currently a professor of nuclear engineering and biomedical engineering at Rensselaer Polytechnic Institute (RPI), Troy, New York. He received his PhD in nuclear engineering/health physics from Texas A&M University in 1994. His current research interests include radiation dosimetry and anatomical modeling for various applications in health physics, diagnostic imaging, and radiotherapy. Dr. Xu directs the Rensselaer Radiation Measurements and Dosimetry Group (http://RRMDG.rpi. edu) and is also the founding director of the Center for Engineering-Based Patient Modeling (http://CEPM.rpi.edu) at RPI. In 2005, Dr. Xu cofounded the Consortium of Computational Human Phantoms (http://www.virtualphantoms.org) which aims to promote international research collaboration. He serves on several editorial boards and is involved in the technical committees of various associations such as the American Association of Physicists in Medicine, the American Nuclear Society, the Health Physics Society, the International Commission on Radiological Protection, and the National Council on Radiation & Measurements; he is also the past president of the Council on Ionizing Radiation Measurements and Standards. Dr. Keith F. Eckerman received his PhD in environmental health engineering from Northwestern University in 1972. He has been the leader of the Dosimetry Research Team, Environmental Science Division, Oak Ridge National Laboratory (ORNL) since 1979. He is a member of the International Commission on Radiological Protection and the National Council on Radiation Protection and Measurements. An internationally recognized authority on internal dosimetry and biokinetic modeling, Dr. Eckerman’s research covers radiation dosimetry; radiological assessments; and applications of mathematical models to radiation dosimetry, physiology, and metabolism. He has authored and coauthored more than 200 journal publications, book chapters, standards, and proceedings. Dr. Eckerman has received numerous awards, including the DOE Award-Operation Ivory Purpose in 1980, the Health Physics Society Distinguished Scientific Achievements Award in 1995, the NRC Special Achievement Award in 1997, and the Society of Nuclear Medicine Loevinger–Berman Award in 2001.

xv

Contributors

Manuel Bardiès Oncology Research Department INSERM Nantes, France

Aiping Ding Department of Nuclear Engineering Rensselaer Polytechnique Institute Troy, New York

Giovanni Bibbo Division of Medical Imaging Women’s and Children’s Hospital Adelaide, South Australia, Australia

Keith F. Eckerman Oak Ridge National Laboratory Oak Ridge, Tennessee

Wesley E. Bolch Department of Nuclear and Radiological Engineering University of Florida Gainesville, Florida

Mary Ann Emmons-Keenan Department of Radiology Vanderbilt University Nashville, Tennessee

Ahmet Bozkurt Department of Physics Harran University Sanliurfa, Turkey Martin Caon School of Nursing and Midwifery Flinders University Adelaide, South Australia, Australia Tsi-Chian Ephraim Chao Department of Medical Imaging and Radiological Sciences Chang Gung University Taipei, Taiwan Ji Chen Department of Electrical and Computer Engineering University of Houston Houston, Texas John J. DeMarco Department of Radiation Oncology University of California Los Angeles, California

Michael J. Fernald RADAR Inc. Nashville, Tennessee Martin Fuss Department of Radiation Medicine Oregon Health & Science University Portland, Oregon David E. Hintenlang Department of Nuclear and Radiological Engineering University of Florida Gainesville, Florida Gabriela Hoff Department of Physics Pontifical Catholic University of Rio Grande do Sul Porto Alegre, Brazil Jorge Hurtado Department of Nuclear and Radiological Engineering University of Florida Gainesville, Florida xvii

xviii

An Jin Department of Biomedical Engineering Rensselaer Polytechnique Institute Troy, New York

Contributors

Choonsik Lee Department of Nuclear and Radiological Engineering University of Florida Gainesville, Florida

Wolfgang Kainz Center for Devices and Radiological Health Food and Drug Administration Silver Spring, Maryland

Junli Li Department of Engineering Physics Tsinghua University Beijing, China

Iwan Kawrakow Ionizing Radiation Standards Group National Research Council of Canada Ottawa, Ontario, Canada

Vanildo Júnior de Melo Lima Department of Anatomy Federal University of Pernambuco Recife, Brazil

Helen Jamil Khoury Department of Nuclear Energy Federal University of Pernambuco Recife, Brazil Chan Hyeong Kim Department of Nuclear Engineering Hanyang University Seoul, Korea Sakae Kinase Radiation Effect Analysis Group Japan Atomic Energy Agency Tokaimura, Japan Gary H. Kramer National Internal Radiation Assessment Section Health Canada’s Radiation Protection Bureau Ottawa, Ontario, Canada Richard Kramer Department of Nuclear Energy Federal University of Pernambuco Recife, Brazil Choonik Lee Department of Radiation Oncology Anderson Cancer Center Orlando, Florida

Qian Liu Britton Chance Center for Biomedical Photonics Huazhong University of Science and Technology Wuhan, China Daniel L. Lodwick Department of Nuclear and Radiological Engineering University of Florida Gainesville, Florida Eduardo César de Miranda Loureiro Department of Mechanical Engineering University of Pernambuco Recife, Brazil C.-M. Charlie Ma Department of Radiation Oncology Fox Chase Cancer Center Philadelphia, Pennsylvania Jizeng Ma China Institute of Atomic Energy Beijing, China Michael McNitt-Gray Department of Radiological Sciences David Geffen School of Medicine University of California Los Angeles, California

Contributors

William E. Moloney Petrone Associates, LLC New York, New York Yong Hum Na Department of Biomedical Engineering Rensselaer Polytechnic Institute Troy, New York Tomoaki Nagaoka Electromagnetic Compatability Group National Institute of Information and Communications Technology Tokyo, Japan Deanna Hasenauer Pafundi Department of Nuclear and Radiological Engineering University of Florida Gainesville, Florida Harald Paganetti Francis H. Burr Proton Therapy Center Massachusetts General Hospital Boston, Massachusetts

xix

Rui Qiu Department of Engineering Physics Tsinghua University Beijing, China Warren Dan Reece Department of Nuclear Engineering Texas A&M University College Station, Texas Kimiaki Saito Division of Environment and Radiation Japan Atomic Energy Agency Tokaimura, Japan Kaoru Sato Research Group for Radiation Protection Japan Atomic Energy Agency Tokaimura, Japan W. Paul Segars Duke Advanced Imaging Laboratories Duke University Medical Center Durham, North Carolina

Niko Papanikolaou Department of Radiation Oncology The University of Texas Health Science Center at San Antonio San Antonio, Texas

Chengyu Shi Department of Radiation Oncology The University of Texas Health Science Center at San Antonio San Antonio, Texas

John E. Pattison Department of Applied Physics University of South Australia Adelaide, South Australia, Australia

Il-Young Son Department of Electrical Engineering University of California San Diego, California

Nina Petoussi-Henss Institute of Radiation Protection German Research Center for Environmental Health Neuherberg, Germany

Michael G. Stabin Department of Radiology and Radiological Sciences Vanderbilt University Nashville, Tennessee

John W. Poston, Sr. Department of Nuclear Engineering Texas A&M University College Station, Texas

Valery Taranenko Department of Health Physics University of California San Francisco, California

xx

Benjamin M.W. Tsui Department of Radiology and Radiological Science Johns Hopkins University Baltimore, Maryland José Wilson Vieira Basic Department Polytechnic School of Pernambuco Recife, Brazil and Federal Institute of Education Science and Technology of Pernambuco Recife, Brazil Scott R. Whalen Department of Nuclear and Radiological Engineering University of Florida Gainesville, Florida Jonathan L. Williams Department of Radiology University of Florida Gainesville, Florida James Winslow Department of Nuclear and Radiological Engineering University of Florida Gainesville, Florida Dagang Wu Halliburton Houston, Texas

Contributors

X. George Xu Department of Mechanical, Aerospace and Nuclear Engineering and Department of Biomedical Engineering Rensselaer Polytechnic Institute Troy, New York Birsen Yazıcı Department of Electrical, Computer and Systems Engineering and Biomedical Engineering Rensselaer Polytechnic Institute Troy, New York Maria Zankl Institute for Radiation Protection German Research Center for Environmental Health Neuherberg, Germany Binquan Zhang China Institute for Radiation Protection Taiyuan, China Guozhi Zhang Wuhan National Laboratory for Optoelectronics Huazhong University of Science and Technology Wuhan, China Juying Zhang Department of Mechanical, Aerospace and Nuclear Engineering and Department of Biomedical Engineering Rensselaer Polytechnic Institute Troy, New York

Part I

Phantoms

1 Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution X. George Xu

CONTENTS 1.1 Introduction ...................................................................................................................3 1.2 Solid-Geometry Modeling Techniques: Constructive Solid Geometry and Boundary Representation .................................................................................... 5 1.3 Historical Developments..............................................................................................7 1.3.1 Stylized Phantoms from 1960s to 2000s .......................................................7 1.3.2 Voxel Phantoms from 1980s to 2000s .......................................................... 12 1.3.3 BREP Phantoms from 2000s to Present ...................................................... 23 1.4 Physical Phantoms ...................................................................................................... 26 1.5 Monte Carlo Codes Used with Computational Phantoms.................................... 27 1.5.1 MCNP and MCNPX ...................................................................................... 31 1.5.2 EGS ................................................................................................................... 31 1.5.3 GEANT4 .......................................................................................................... 32 1.5.4 PENELOPE...................................................................................................... 32 1.5.5 FLUKA............................................................................................................. 32 1.6 Discussions .................................................................................................................. 32 1.7 Summary ......................................................................................................................34 Acknowledgments ................................................................................................................. 35 References ............................................................................................................................... 35

1.1 Introduction One of the most dynamic areas of research in radiation protection, radiological imaging, and radiotherapy is the modeling of human anatomy for Monte Carlo-based radiation transport and dose simulations. Radiation dosimetry aims to determine the amount and distribution pattern of energy deposited in various parts of the human body by internal or external radiation sources. To protect against occupational exposures, regulatory limits are set for radiation doses associated with radiosensitive organs. In both diagnostic radiology and nuclear medicine, internal and external photons traverse through the body to form an image, depositing radiation energy along the way. Radiotherapy, on the other hand, attempts to deliver a lethal dose to the target while sparing the adjacent healthy tissues from the adverse effects of radiation. Accurate radiation dosimetry is essential but also 3

4

Handbook of Anatomical Models for Radiation Dosimetry

quite challenging for three reasons: (1) there are many diverse exposure scenarios resulting in unique spatial and temporal relationships between the source and human body; (2) an exposure can involve multiple radiation types which are governed by rather different radiation physics principles including photons (and gamma rays), electrons, positrons, alpha particles, neutrons, and protons; (3) the human body consists of three-dimensional (3D) inhomogeneous tissues of various geometric shapes and densities, leading to extremely complex radiation interaction patterns. It is not practical to make direct measurement of radiation doses using physical detectors inside the human body. Consequently, dose estimates for select organs of interest have always depended on physical or computational “anthropomorphic models” that mimic the interior and exterior anatomical features of the human body. Historically, the term “phantom” was used in most radiological science literature to mean a physical model of the human body. In the radiation protection community, however, the term has also been used to refer to a mathematically defined “anatomical” model that is distinctly different from a physiologically based computational model such as that related to respiration or blood flow. Throughout this book, we use the phrases “computational phantom” and “physical phantom” to avoid confusion, but the reader may occasionally encounter “phantom” or “model” in places where a computational model of the human or animal anatomy is discussed. A physical phantom is made of solid materials equivalent to bones and soft-tissues that can be molded to the shape of the human anatomy and then cut into slices containing cavities for tiny radiation dosimeters. The approach of using such physical phantoms was known to be expensive and time-consuming due to tedious experimental and radiation safety procedures. Luckily, the advent of the first-generation of computers and Monte Carlo simulation methods for nuclear weapons research in the 1940s made it gradually possible to calculate organ doses in a computational phantom. Each computational phantom defines not only the exterior features of the entire human body, but also includes details on internal organs such as their volume, mass, and shape. Coupled with information on tissue densities and chemical compositions, a computational phantom allows the Monte Carlo codes to simulate interactions and energy deposition in the body for various types of radiation. Although additional work is needed to specify a radiation source, the computational approach is, in general, quite advantageous in terms of its versatility, efficiency, precision, and radiation safety. In the case of internally distributed radionuclides, it is often necessary to obtain dose estimates via calculations involving a computational phantom rather than through experimental measurement. Therefore, since their advent in the 1960s, the use of computational human phantoms has become increasingly popular in the fields of radiation protection, imaging, and radiotherapy. Today, physical phantoms are used only in radiation protection dosimetry as benchmarks to computational results for external exposures. For nonionizing radiation, similar computational phantoms have been developed over the years to study the biological effects caused by radiofrequencies emitted by devices such as electric power lines and wireless cellular phones. Since the 1960s, approximately 121 computational phantoms,1–93 plus 27 physical phantoms,94–103 have been reported in the literature for studies involving ionizing and nonionizing radiation. A significant portion of the literature on radiation protection dosimetry is related to the development and application of these phantoms. The organs and body surfaces of computational phantoms have been defined in terms of a variety of solid geometry modeling techniques: quadric equations, voxels, and advanced primitives such as B-splines or nonuniform rational B-splines (NURBS) or polygon meshes. Each of these techniques was adopted at specific times in the last 40 years, exhibiting an interesting

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

5

scientific journey that reflects the advances in computer and medical imaging technologies. Like other research fields, nontechnical considerations also had their share in shaping the world of computational phantoms that we know today. As the evolution continues, particularly at the accelerated speed witnessed in the past decade, it is vitally important to know where we are going. There are many interesting questions facing researchers today. Why did the computational phantoms evolve the way they did? What will be the future directions in this research field? The answers to these questions and many others require an understanding and evaluation of the rationales and processes responsible for some of the most widely used phantoms. This chapter serves as an introduction to the following chapters, which detail specific phantoms and their applications. The different modeling techniques are defined and a summary of historical milestones in the development of phantoms for ionizing and nonionizing radiation applications is provided.

1.2 Solid-Geometry Modeling Techniques: Constructive Solid Geometry and Boundary Representation Computational human phantoms are basically solid-geometry models that depict exterior and interior anatomical features of a human body. For radiation dosimetry, a phantom must define the surface of an organ in which radiation interactions and energy depositions are to be calculated by tracing individual particles. Clearly, the construction of such phantoms must consider multiple factors such as anatomy, radiosensitivity, computational efficiency, and geometrical compatibility with a Monte Carlo code. The computer graphics community has dealt extensively with solid-geometry modeling for computer-aided design (CAD). Two general methods of solid-geometry modeling have been widely developed: constructive solid geometry (CSG) and boundary representation (BREP).104–108 The topology—spatial location and relationship of the surfaces—is fundamentally different for these two methods. CSG allows a modeler to create a solid object using Boolean operators (or the equivalent) to combine very simple objects called primitives. Examples of these primitives include cuboids, cylinders, prisms, pyramids, spheres, cones, and ellipsoids—surfaces that are easily described by quadric equations. CSG representations are easy to adopt and can yield good results when the objects are relatively simple in shape. Modern CAD software systems, however, are based on the more powerful BREP methods. There are two types of information in the BREP: topological and geometric. Topological information provides the relationships among vertices, edges, and faces. In addition to connectivity, topological information also includes orientation of edges and faces. In advanced BREP-based CAD, the exterior of an object is defined as NURBS which afford very smooth surfaces. The faces can alternatively be represented as polygons whose vertices are defined by a set of coordinate values x, y, and z. A polygon mesh or unstructured grid is a collection of vertices and polygons that define the geometric shape of a polyhedral object in CAD. In principle, NURBS and polygonal meshes are interchangeable BREP data structures. Unlike the CSG representation, however, BREP is much more flexible because a richer set of operation tools are available (e.g., extrusion, chamfering, blending, drafting, shelling, and tweaking). These features allow BREP-based phantoms to include very complex anatomical features. Furthermore, the BREP technique is ideally suited for surface deformation—an operation necessary for the adjustment of organ size and for organ motion simulations as described later.

Handbook of Anatomical Models for Radiation Dosimetry

6

As an example, the left lung can be represented in the CSG method by “half an ellipsoid with a section removed.”12 The cut-out section, which is not specified by the original authors, can be defined by a Boolean operation of subtracting one ellipsoid (B) from the other (A) to create the left lung, as described in 2

2

2

2

2

2

⎛ X − 8.5 ⎞ ⎛ Y ⎞ ⎛ Z − 43.5 ⎞ A: ⎜ + ⎜ + ⎜ ≤ 1, Z ≥ 43.5 ⎝ 5 ⎟⎠ ⎝ 7.5 ⎟⎠ ⎝ 24 ⎟⎠ ⎛ X − 2.5 ⎞ ⎛ Y ⎞ ⎛ Z − 43.5 ⎞ B: ⎜ + ⎜ + ⎜ ≥ 1, if y < 0 ⎝ 5 ⎟⎠ ⎝ 7.5 ⎟⎠ ⎝ 24 ⎟⎠

(1.1)

In Figure 1.1a, the 3D shapes of the left lung before and after the Boolean operation are illustrated. These surface equations are computationally efficient and are accepted by nearly all Monte Carlo codes. However, even with complicated and carefully designed Boolean operations like this, phantoms based on quadric surfaces are not anatomically realistic in terms of their geometry. When using a Monte Carlo code, the geometry of the left lung is often further simplified by replacing the ellipsoid B with several planes. This type of phantoms is commonly referred to as “stylized” or “mathematical” phantoms. Using voxels as a CSG modeling technique, Figure 1.1b defines the left lung as an assembly of 3D cuboids. Medical image data can be converted to voxel geometry that provides a direct way to realistically describe the human anatomy. The geometry of a voxel is very simple for existing Monte Carlo codes to handle, although the large number of voxels may require the use of enhanced computer hardware or special Monte Carlo software preparation. On the other hand, however, each tomographic image slice needs to be treated by a “segmentation” process, which assigns each pixel to an organ or tissue of interest such as the lung, bone, or skin using a unique identification number. It can take a significant amount of time to prepare a voxel-based phantom because there is no automatic segmentation algorithm that works on all organs. Furthermore, a voxel phantom is based on images for one subject, thus lacking anatomical variability associated with organ size, shape, and

A

B

(a)

A

A

(b)

(c)

FIGURE 1.1 The left lung defined by different modeling techniques. (a) A stylized lung model using the CSG-type method before and after a Boolean operation is performed to remove a section of the ellipsoid B from A. (b) A CSG-type method involving a group of rigid voxels with an anatomical detail dependent on the voxel size. (c) A BREPtype of method involving a polygonal mesh that is easy to deform and anatomically accurate.

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

7

location that are important in the current paradigm for radiation protection dosimetry. Finally, the boundary of the lung in a voxel phantom is defined by uneven steps instead of a smooth surface, as show in Figure 1.1b. As a result, the anatomical fidelity depends on the voxel size, especially for thin and small tissues such as the skin, eye lens, ribs, and bone marrow. An adjustment to the organ shape will likely involve all underlying voxels, which is computationally inefficient. These types of computational human body models are commonly referred to as “voxel” or “tomographic” phantoms. The lung can also be defi ned by the advanced BREP modeling techniques involving NURBS or polygon mesh surfaces. The most common technique to create a BREP-based phantom involves the surface contour extraction of each organ from a tomographic image data set using a commercial software package, followed by the integration of individual organs into a whole body assembly. In essence, the contours convert the voxels into NURBS or mesh surfaces that are smooth and anatomically realistic. These phantoms are commonly referred to as “NURBS,” “mesh,” or “BREP” phantoms. Figure 1.1c shows the triangular meshes of a left lung, which was derived from high-resolution tomographic images.

1.3 Historical Developments Previously published reviews of the historical development of computational phantoms have focused on a certain time period or a particular phantom type.109–111 These reviews did not explicitly classify phantom modeling techniques and since the time of their publication a number of phantoms have been developed using the new BREP methods. An understanding of the modeling techniques is important because each one of them was predominantly adopted by the research community during a specific time of the development. Based on chronological and technical information in the literature, existing computational phantoms can be divided into three classes: (1) stylized phantoms (1960s to 2000s); (2) voxel phantoms (1980s to 2000s); and (3) BREP phantoms (2000s to present). 1.3.1 Stylized Phantoms from 1960s to 2000s Table 1.1 summarizes some of the most important and unique stylized phantoms developed from the 1960s to the 2000s.1–14 This generation of phantoms originated from work performed at Oak Ridge National Laboratory (ORNL) and are covered in detail in Chapter 2. The first attempts at developing a computational anthropomorphic phantom were reported by Fisher and Snyder at ORNL in the 1960s.7,8 Using CSG modeling techniques involving shapes such as elliptical cylinders and cones, they developed the so-called Fisher–Snyer adult phantom. Fisher and Snyder also developed the “similitude” children phantoms, which were scaled-down versions of the adult with added assumption that the entire body was a homogeneous tissue (i.e., the lungs and skeleton were ignored). In 1969, Snyder and his colleagues reported the first heterogeneous phantom that became known as the “MIRD-5 Phantom,” a named derived from the Medical Internal Radiation Dosimetry (MIRD) Committee of the Society of Nuclear Medicine which adopted the phantom.9 This phantom was composed of a skeleton, a pair of lungs, and the remainder (soft tissue). The representation of internal organs in this mathematical phantom was crude, as the simple equations captured only the most general description of the position and geometry of each

KMIRD

MCAT

Hanyang University, Korea

Johns Hopkins University, USA (formerly with the University of North Carolina) NASA, USA

CAM

ADAM and EVA

GSF, Germany

Developers

Phantom Names

5

10

3

Chapters in This Book

Quadric equations

Quadric equations

Quadric equations

Quadric equations

Data Types

A standing U.S. air force adult male representing 50th percentile height and weight. More than 1000 geometric surfaces and 2450 solid regions.

Gender-specific phantoms revised from the ORNL MIRD-5 phantom for external dose assessment. Several minor anatomical changes including the breast size. Outer body and internal organs of the ORNL adult male phantom modified according to Korean anthropometric data. 3D and 4D cardiac torso phantom with gated patient organ motion information for imaging applications.

Anatomical Features

Caucasian adult male

Caucasian adult male

Korean adult male

Caucasian adult male and female

Human Subjects

I

I

I

I

Ionizing (I) or Nonionizing Radiation (N)

List of Developers, in Alphabetical Order, of Stylized Computational Phantoms Including Information on Phantom Names, Chapters in This Book, Phantom Data Types, Anatomical Features, Human Subjects, Ionizing versus Nonionizing Radiation, and References

TABLE 1.1

[6]

[3–5]

[2]

[1]

References

8 Handbook of Anatomical Models for Radiation Dosimetry

Radiation Protection Bureau, Canada

ORNL, USA

2

12

Cristy–Eckerman family phantoms

Pregnant women

Mathematical models of the embryo and fetus

2

Fisher–Snyder phantom (MIRD-5) and others

Quadric equations

Quadric equations

Quadric equations

Quadric equations

The first anthropomorphic phantom representing a hermaphrodite adult for internal dosimetry. Organ masses, body weight, and body height correspond to 50th percentile data recommended in ICRP 23. Later, age-specific phantoms were developed by others. Based on MIRD-5 phantom and others from ORNL. The age 15 phantom represents a 15 year old male and an adult female. Adult female in the family phantom was used to add uterine contents including the fetus, fetal skeleton, and placenta at three different gestational stages. Phantoms of the embryo and fetus representing additional gestational periods not included in the ORNL pregnant women phantoms for dosimetry studies involving commercial flights. Caucasian pregnant women at 8, 13, 26, 38 weeks of gestation

Caucasian pregnant women at 3, 6, and 9 months of gestation

Caucasian adult

Caucasian newborn, 1 year old, 5 years old, 10 years old, 15 years old, and the adult

I

I

I

[14]

[13]

[7–12]

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution 9

10

Handbook of Anatomical Models for Radiation Dosimetry

organ. The original model was intended to represent a healthy “average” adult male, the Reference Man, as defined by the International Commission on Radiological Protection (ICRP) from an extensive review of medical and other scientific literature on the European and North American populations.112 The Reference Man was a 20–30 year old Caucasian, 70 kg in weight and 170 cm in height (the height was later changed to 174 cm). In 1978, Snyder et al. published an elaborative set of specific absorbed fractions using an improved version of their heterogeneous phantom which contained more than 20 organs and more detailed anatomical features.10 The limitations associated with the approach of applying a set of scaling factors to the adult phantom to create age-dependent similitude phantoms were clear. Significant efforts were undertaken at ORNL during the mid-1970s by Poston and others to develop individualized pediatric phantoms of the newborn, 1 year old, 5 years old, 10 years old, and 15 years old.113–115 However, these phantoms were not widely adopted because the geometric shapes were very complex and, after Poston left ORNL, alternative approaches were developed. Building upon previous work, Cristy reported the development of a new series of stylized phantoms in 1980 and then with Eckerman in 1987 in ORNL/TM-8381.11,12 This series or “family” of phantoms consisted of an adult male, a newborn, and individuals of ages 1, 5, 10, and 15 (also representing an adult female with additional anatomical features). Each phantom is composed of three tissue types with distinct densities: bone, soft tissue, and lung. They were analytically defined in three principal geometric sections as illustrated in Figure 1.2a: an elliptical cylinder representing the arms, torso, and hips; a truncated elliptical cone representing the legs and feet; and an elliptical cylinder representing the head and neck. Figure 1.2b shows the skeleton and internal organs and tissues. A picture of the “family” is shown in Figure 1.2c. In 1995, Stabin and his colleagues at ORNL adapted the adult female phantom in this family to represent a pregnant woman at the end of each trimester of pregnancy.13 This set of three stylized pregnant female phantoms was used for various internal nuclear medicine applications. The 9 month pregnant female phantom is shown in Figure 1.2d. Since the 1980s, a number of revised MIRD-5 phantoms were reported, which incrementally improved upon the original Fisher–Snyder and Cristy–Eckerman phantoms using the same stylized modeling techniques; however, they are not explicitly listed in Table 1.1 (for example, Bouchet et al. on a revised head and brain model116). With the availability of general-purpose Monte Carlo codes and affordable computers in the 1980s, this latest series of phantoms, referred to as the “Cristy–Eckerman Phantoms,” were quickly adopted by many users for a wide variety of internal dosimetry applications. Later, this set of phantoms was also used for external dosimetry studies including the one at Texas A&M University discussed in Chapter 18. In parallel with the efforts at ORNL by Cristy and Eckerman to revise the MIRD-5 Phantom, Kramer et al. from the GSF, Germany used the anatomical descriptions of the hermaphrodite MIRD-5 phantom to develop a pair of gender-specific adult phantoms known as the ADAM and EVA for external dosimetry studies.1 The female EVA phantom was chosen to have a weight 83% of the MIRD-5 adult phantom. There are a number of minor anatomical differences, such as breast sizes, from those reported by Cristy and Eckerman.1,12 Table 1.1 also lists several additional efforts related to stylized phantoms. The stylized modeling technique was also adopted by one group for medical applications. The Mathematical Cardiac Torso (MCAT) phantom which includes the major thoracic structures and organs was developed by a research group led by Tsui at the University of North Carolina (currently with Johns Hopkins University) for use in nuclear medicine imaging research, specifically single-photon emission computed tomography (SPECT) and positron

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

(a)

11

(b)

ORNL–DWG 79–19955

0 year (c)

1 year

5 year

10 year

15 year

Adult (d)

FIGURE 1.2 Stylized phantoms. (a) Exterior view of the adult male. (b) Skeleton and internal organs. (c) “Family” phantoms representing individuals of both genders and various ages. (d) Cross-sectional view of the phantom for a 9 month pregnant female.

emission tomography (PET).3–5 The same group later developed the first NURBS-based motion-simulating phantom to be described in Chapter 5. The Computational Anatomical Man (CAM) phantom developed by Billings and Yucker in 1973 for the National Aeronautics and Space Administration (NASA) demonstrated a very different and aggressive approach in stylized modeling because the phantom reportedly consisted of 1100 unique geometric surfaces and 2450 solid regions.6 According to the authors, internal body geometries such as

12

Handbook of Anatomical Models for Radiation Dosimetry

organs, voids, bones, and bone marrow were explicitly modeled using the CSG modeling techniques. A computer program called CAMERA was also developed for performing analyses with the CAM phantom. The authors state that “extremely detailed geometrical model of the human anatomy, the most detailed yet prepared, has been developed for use in investigations dealing with exposure of astronauts to the natural space radiation environment. The model is equally applicable to investigations dealing with exposure of humans to radiation associated with nuclear weapon and nuclear power system environments as well as medical applications such as radiotherapy and radiography.”6 Indeed the surface geometry was so detailed that one may wonder how this was possible in the 1970s with much less capable computers. Unfortunately, the CAM phantom was never adopted for applications outside the aerospace industry and very little information about the work was accessible by the phantom research community until Jordan, a contracted phantom developer and user, recently released some of the images (http://cmpwg.ans.org/phantoms/ camera.pdf). In the early 1990s, it was clear that the research community no longer favored stylized phantom modeling methods. However, at least two groups continued to use these methods to develop computational phantoms of an embryo and fetus for space radiation dosimetry14 and an adult representing the Korean population.2 For 40 years since the first anthropomorphic phantom was reported, these anatomically simplified phantoms have been used as the de facto “standard” representations of the ICRP “Reference Man” methodology, which is based on “population-average” 50th percentile anatomical parameters.112,117 Applications of stylized phantoms have eventually included many aspects of radiation protection, radionuclide therapy, and medical imaging.118 In addition, national and international bodies have adopted organ dose estimates derived from these stylized phantoms in guidelines and regulations related to industrial and medical uses of ionizing radiation. Although stylized phantoms made it possible to carry out Monte Carlo computations during times when computers were much less powerful, the original developers, as discussed in Chapter 2, recognized the obvious shortcomings. Human anatomy is too complex to be realistically modeled with a limited set of surface equations. Many anatomical details in these models were compromised that sometimes led to inaccurate results. For example, when such phantoms were applied to nuclear medicine procedures where precise dosimetry is necessary, the calculated average organ and marrow doses did not produce strong correlations with observed marrow toxicity. Most nuclear medicine physicians consequently tend to administer lower-than-optimal amounts of radioactivity to avoid toxicity. For computed tomography (CT) dose reporting, all existing commercial software systems are based on the stylized patient models that are known to cause very large errors for low-energy x-rays.119 Similar stylized models have also been used to derive dose–response relationships for Japanese atomic bomb survivors and for medical patients in epidemiological studies. In the external-beam radiotherapy community, an early stylized homogenous phantom was used by the Radiation Epidemiology Branch of the National Cancer Institute (NCI) for nearly 30 years in studies related to organ doses of therapeutically irradiated patients.120 By the 1980s, a few groups of researchers began to seek new ways to develop anatomically realistic phantoms. 1.3.2 Voxel Phantoms from 1980s to 2000s The development of anatomically realistic models was desirable but impossible until early 1980s when powerful computer and tomographic imaging technologies became available. With the advent of CT and magnetic resonance (MR) imaging techniques, researchers could

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

13

for the first time visualize the internal structures of the body in three dimensions and store the images in versatile digital formats. These advantages brought about the exciting and prolific era of the so-called voxel or tomographic phantoms. Table 1.2 summarizes a total of 74 phantoms that were constructed from three types of tomographic images: CT and MR images from live subjects, as well as cross sectional photographs of cadavers.15–80 The ionizing and nonionizing radiation applications of these phantoms are clearly labeled in the table. In two previously published review articles, the number of voxel phantoms was reported to be 21 by Caon110 and 38 by Zaidi and Xu.111 The increase in phantom number is due to a more exhaustive literature search, recent developments, and the inclusion of phantoms developed for use solely in nonionizing radiation applications. In terms of solid-geometry modeling techniques, a voxel—one of the basic CSG primitives—is simply a 3D representation of a pixel. However, compared with the medical applications such as radiation treatment planning, the task of developing reference human phantoms presented some unique and intractable challenges: (1) to construct a whole-body phantom, image slices should ideally cover the entire body—a process not normally carried out in routine medical examinations because of x-ray exposures or the lengthy time required for MR procedures. (2) A large amount of internal organs/tissues must be identified and segmented for organ dose calculations, whereas, in radiotherapy, for example, only the tumor volume and adjacent regions are routinely outlined. (3) The image data size of a whole-body model, especially when high-resolution images are used, can be potentially too great for a computer to handle. (4) A standardized patient phantom is often used to study diverse radiation types such as photons, electrons, neutrons, and protons, thus requiring considerable Monte Carlo simulation capabilities. In terms of the developmental process, tomographic phantoms are fundamentally different from the stylized ones. A tomographic image data set is composed of many slices, each displaying a two-dimensional (2D) pixel map of the anatomy. The 3D volume of a voxel is measured by multiplying the pixel size by the thickness of an image slice. Unlike stylized phantoms which are based on quadric surface equations, a voxel phantom contains a huge number of tiny cubes grouped to represent various anatomical structures. However, both quadric surface equations and voxels (cuboids) belong to the same class of CGS geometries. The creation of a tomographic phantom involves four general steps: (1) acquire a set of tomographic images (e.g., CT, MR, or anatomical photography) that cover the entire volume of the body; (2) identify organs or tissues of interest (e.g., lungs, liver, skin, etc.) from the original image slice by assigning every pixel with an identification number; (3) specify the density (e.g., soft tissue, hard bone, air, etc.) and chemical composition of organs or tissues; and (4) register the segmented image slices into a 3D volume that can be used for 3D visualization (for checking anatomical structures) and for Monte Carlo calculations. Figure 1.3 illustrates these steps using the National Library of Medicine’s Visible Human image data set. The earliest effort to create image-based phantoms may have been reported in the work by Gibbs of Vanderbilt University published from 1982 to 1987.75–77 Forgotten by most latecomers, Gibbs and her coworkers explored the use of 2D x-ray images as the basis to form an anatomically realistic phantom. They then used this information in Monte Carlo calculations to assess the doses received by patients who underwent various radiological procedures. At nearly the same time, Zankl and her colleagues at GSF—National Research Center for Environment and Health in Germany decided to use CT imaging on healthy volunteers to develop what eventually became a family of 12 voxel phantoms: BABY, CHILD, DONNA, FRANK, HELGA, IRENE, GOLEM, GODWIN, VISIBLE HUMAN, LAURA, KLARA, and KATJA.27–33 As detailed in Chapter 3, the adult male phantoms were developed first.

Flinders University, Australia

Federal University of Pernambuco, Brazil

Brooks Air Force Base, USA China Institute for Radiation Protection, China Darmastadt University of Technology, Germany FCS Department, Italy

Developers

4

7

MAX06 and FAX06

ADELAIDE

7

CT

CT

CT

CT

MAX

FAX

MRI

DAM

7

Color photos

HUGO

Color photos

CNMAN

11

Color photos

Data Types

Visible Man

Phantom Names

Chapters in This Book

Based VOXTISS8 phantom and adjusted to ICRP 89. Images of the trunk, the neck, and the lower part of the head were from CT scan of a 37 year old female. Images of the legs and feet were from CT scan of a 62 year old woman. The head and arms were from MAX phantom. Extension of MAX and FAX phantoms by adding more details in the skeleton that were further adjusted to match the values by ICRP 89. Torso phantom, without head and arms.

Dielectric anatomical phantom.

VHP. A total of 32 tissues were identified.

VHP. More than 40 tissues were identified. Chinese VHP.

Anatomical Features

Caucasian 14 year old female patient

Caucasian adult male and female patient

34 year old male volunteer Caucasian adult male patient Caucasian adult female patient

Caucasian 39 year old male cadaver

Caucasian 39 year old male cadaver Chinese adult male cadaver

Human Subjects

I

I

I

I

N

N

I

N

[23,24]

[22]

[21]

[20]

[19]

[18]

[17]

[15,16]

Ionizing (I) or Nonionizing Radiation (N) References

List of Developers, in Alphabetical Order, of Voxel Computational Phantoms Including Information on Phantom Names, Chapters in This Book, Phantom Data Types, Anatomical Features, Human Subjects, Ionizing versus Nonionizing Radiation, and References

TABLE 1.2

14 Handbook of Anatomical Models for Radiation Dosimetry

The weight and height are similar to those of ICRP 23 reference man. Modification of Golem phantom to agree with the ICRP 89 anatomical data. Head to knee. CT data from the VHP. 167 cm height and a weight of 59 kg.

CT CT CT CT CT

CT CT CT MRI

CT

FRANK

HELGA

IRENE

GOLEM

GODWIN

VISIBLE HUMAN LAURA

KLARA

KATJA

REGINA (ICRP Reference Phantom)

The modification of Laura to agree with ICRP 89 anatomical data. A woman in her 24th week of pregnancy. Based on the modified REGINA phantom and patient MRIs of the abdominal and pelvic regions. An adjusted LAURA phantom according to ICRP 89.

Whole body phantom (163 cm, 51 kg).

From mid thigh upwards.

Head and torso.

Whole body phantom (176 cm, 79 kg)

115 cm in height and weighed 21.7 kg.

CT

CT

DONNA

3

Trunk was based on MR images of a pregnant women and modified on CT images of a woman in the 30th week of pregnancy developed by RPI. The brain and spinal cord were from NORMAN and fitted into SILVY. 57 cm in height and weighed 4.2 kg.

CT

BABY

GSF, Germany

MRI, CT

CHILD

SILVY

Graz University of Technology, Austria

Caucasian 39 year old male cadaver Caucasian 43 year old female patient Caucasian 43 year old female patient Caucasian pregnant woman patient in her 24th week of pregnancy Caucasian 43 year old female patient

Caucasian 8 week old female cadaver Caucasian 7 year old female leukemia patient Caucasian 40 year old female patient Caucasian 48 year old male patient Caucasian 26 year old female patient Caucasian 32 year old female patient Caucasian 38 year old male patient Caucasian 38 year old male patient

Caucasian 30th week pregnant woman patient

I

I

I

I

I

I

I

I

I

I

I

(continued)

[34,35]

[33]

[32]

[32]

[31]

[32]

[31]

[29,31]

[29,30]

[30]

[29,30]

[27,28]

[27,28]

I I

[25,26]

N

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution 15

Color photos

HDRK-Man

Pregnant female, hybrid phantoms (four phantoms)

NORMAN-05

High-definition reference Korean male phantom from the visible Korean human data. MRI Normalized man. Only 10 ribs. MRI Weight and height were scaled to the values by ICRP 89. MRI Based on NORMAN with new tissues recommended by ICRP. Quadric Based on NAOMI and Chen’s stylized equations and fetal phantoms. MRI

PET and CT

KTMAN-2

MRI

MRI

KORWOMAN

10

Korean male of average height and weight. Korean female of average height and weigh. Legs were modeled from the VHP data. Korean Typical Man (172 cm, 65 kg, without arms). Korean Typical Man-2 (172 cm, 68 kg).

MRI

KTMAN-1

An adjusted GOLEM phantom according ICRP 89

Anatomical Features

CT

Data Types

REX (ICRP Reference Phantom) KORMAN

HPA, U.K. (formerly NORMAN NRPB) NAOMI

Hanyang University, Korea

Developers

Chapters in This Phantom Names Book

Pregnant woman at 8, 13, 26, 38 weeks of gestation

Caucasian adult male Caucasian healthy adult female volunteer Caucasian adult male

Korean 25 year old male volunteer Korean 35 year old male volunteer Korean 33 year old adult male cadaver

Korean 30 year old healthy male Korean 35 year old female

Caucasian 38 year old male leukemia patient

Human Subjects

N

I

N, I N

I

I

I

I

I

I

[47]

[46]

[41–43] [44,45]

[39,40]

[38]

[38]

[37]

[36]

[34,35]

Ionizing (I) or Nonionizing Radiation (N) References

List of Developers, in Alphabetical Order, of Voxel Computational Phantoms Including Information on Phantom Names, Chapters in This Book, Phantom Data Types, Anatomical Features, Human Subjects, Ionizing versus Nonionizing Radiation, and References

TABLE 1.2 (continued)

16 Handbook of Anatomical Models for Radiation Dosimetry

NIICT, Japan

Duke: 34 year old male (174 cm, 70 kg) Ella: 26 year old female (160 cm, 58 kg) Billie: 11 year old female (148 cm, 34 kg) Thelonious: 6 year old male (107 cm, 17 kg). Japanese adult male voxel phantom (170 cm, 65 kg). Japanese adult male voxel phantom: CT scan in supine posture. The male subject recruited for the construction of JM was selected to obtain CT scan in upright posture. Japanese adult female phantom (162 cm, 57 kg). Japanese adult female phantom (152 cm, 44 kg). Adult male phantom (171.4 cm, 65.0 kg) representing average anatomical values of Japanese 18 year old male. Adult female phantom (159.1 cm, 52.6 kg) representing average anatomical values of Japanese 30 year old female. Based on the HANAKO phantom and the abdominal phantom of a 26 week pregnant woman. Transformed from the TARO phantom into children models using the FFD algorithm.

MRI

CT CT MRI

ONAGO

JF

TARO

MRI

MRI/FFD

Pregnant woman

Deformed Children

MRI

CT

JM2

HANAKO

CT

CT

Visible Chinese human project.

Color photos

JM

9

9

JAEA, Japan

OTOKO

11

Huazhong VCH University of Science and Technology, China IT’IS, Switzerland The virtual family (four phantoms)

Japanese 26 week pregnant woman volunteer Japanese 3, 5, and 7 year children

Japanese 22 year old female volunteer

Japanese adult female volunteer Japanese adult female volunteer Japanese 22 year old male volunteer

N

N

N, I

N, I

I

I

I

I

I

N

Caucasian volunteers of different gender and ages

Japanese adult male volunteer Japanese 54 year old male volunteer Japanese 54 year old male volunteer

I

Chinese adult male cadaver

(continued)

[61]

[59,60]

[57,58]

[57]

[56]

[55]

[54]

[53]

[52]

[51]

[48–50]

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution 17

University of Florida, USA

Tsinghua University, China University Hospitals of Leuven, Belgium

RPI, USA

ORNL, USA

Developers

CT

CT

UF 2 months

UF newborn

CT

Phantom 2

MRI

CT

CT

MRI

8

Anatomical Features

A voxel phantom equivalent to a 2 month old male newborn, representing a critically ill child. A normal 6 day female newborn phantom; the lungs were created using CT images of a 1 month old patient and the adrenal glands created using CT images of a 2 month old male patient.

Baby phantom (30.4 cm, 0.59 kg).

Baby phantom (50 cm, 1.91 kg).

A pregnant woman phantom covering the abdominal region. A whole-body model of the RANDO physical phantom. Chinese Voxel phantom (170 cm, 70 kg).

CT and quadric Voxelized head and torso phantom equations with stylized arms and legs. Color photos High resolution images from VHP.

Data Types

Phantom 1

11

12

6

3D VIP-Man

Pregnant woman RANDO CT phantom CVP

21

VOXMAT

Chapters in This Phantom Names Book

Caucasian 6 day old female newborn cadaver

22 week old stillborn male baby cadavers Caucasian 6 month old male cadaver

Chinese adult male volunteer 33 week old stillborn male baby cadaver

Caucasian 39 year old male cadaver 30 week pregnant woman patient Adult male

Caucasian adult male

Human Subjects

I

I

I

I

I

I

I

I

I

[69]

[69]

[68]

[68]

[66,67]

[65]

[64]

[63]

[62]

Ionizing (I) or Nonionizing Radiation (N) References

List of Developers, in Alphabetical Order, of Voxel Computational Phantoms Including Information on Phantom Names, Chapters in This Book, Phantom Data Types, Anatomical Features, Human Subjects, Ionizing versus Nonionizing Radiation, and References

TABLE 1.2 (continued)

18 Handbook of Anatomical Models for Radiation Dosimetry

VOXTISS8

CT

MANTISSUE3–6

Yale University, USA

CT

CT

Radiography

MRI

Anatomically based model Gibbs phantoms Zubal

University of Utah, USA Vanderbilt University, USA

Color photos

CT

MEET Man

21

CT

University of Karlsruhe, Germany

UF series B (five phantoms)

UF series A (five phantoms)

Arms and legs from the VHP were attached to the Zubal phantom. Arms and legs were attached to the Zubal phantom and the arms were straightened along the phantom side.

Based on the UF series A phantoms with arms and legs from CT images of a healthy Korean adult attached. The organ masses were adjusted to ICRP 89 reference data. Models for simulation of electromagnetic, elastomechanic and thermic behavior of man, developed from the VHP. Anatomic phantom. The outer parts of the arms are missing. Head, trunk and proximal extremities from x-ray images. Head and torso.

UF pediatric phantom series without arms and legs.

Caucasian adult male volunteer Caucasian representative female cadaver Caucasian adult male patient Caucasian adult male patient Caucasian adult male patient

Caucasian 38 year old adult male cadaver

9 month, 11 and 14 year old males; 4 and 8 year old females patients 9 month, 11 and 14 year males; 4 and 8 year female patients

I

N

I

I

N

N, I

I

I

[80]

[79]

[78]

[75–77]

[74]

[72,73]

[71]

[70]

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution 19

Handbook of Anatomical Models for Radiation Dosimetry

20

Identification of organs in each slice of a 2D pixel map

Registration of all slices

Finished 3D voxel phantom

FIGURE 1.3 (See color insert following page 524.) Steps to create a voxel phantom illustrated using the Visible Human cadaver image data set.

These were soon followed by the adult female, pediatric, and pregnant woman phantoms. The GOLEM and LAURA phantoms have recently undergone significant revision, as described in Chapter 15, to yield the REGINA and REX phantoms which are being released to the public as the ICRP Reference Computational Phantoms.34,35 Several processes were considered at the time when this pair of reference phantoms was developed: (1) CT image data sets of individuals close to the Reference Man and Reference Woman (height and weight) were needed; (2) the data sets were segmented; (3) the body heights were adjusted to reference values by scaling the voxels; (4) the skeletal masses were adjusted to the reference values; (5) individual organs were adjusted to reference values by adding and subtracting voxels. In 1994, Zubal et al. from Yale University published a head-torso model named VoxelMan which was developed from CT images.78 The original phantom was used for optimizing nuclear medicine imaging. Improvements to the original phantom were made with a magnetic resonance imaging (MRI) scan data of a human brain. This phantom is commonly referred to as the “Zubal Phantom” by users who are allowed to freely download the original data through the Internet. Two early users later revised the original data to report what are known as the MANTISSUE3-6 and VOXTISS8 phantoms by attaching arms and legs in two different positions to the original torso phantom.79,80 Adopting this publically available data, Kramer et al. from Brazil developed an adult male phantom named MAX (Male Adult voXel) in 200320 and later an adult female phantom named FAX in 200421 both adjusted according to ICRP-89 reference body heights and organ masses. As described in Chapter 7, Kramer et al. revised the skeletons (cortical bone, spongiosa, medullary yellow bone marrow, and cartilage) of MAX and FAX in 2006 to improve their compatibility with the latest ICRP-103 recommendations. These revised phantoms are known as MAX06 and FAX06. The work by Kramer et al. is one of the earliest efforts to create ICRP-89 compatible voxel phantoms. In 1996, Dimbylow from the National Radiological Protection Board (NRPB) (which has been recently renamed to Health Protection Agency) in the United Kingdom reported the development of an adult male phantom known as NORMAN from MR images.41 NORMAN, which has a body height similar to the ICRP Reference Man, was first used by Dimbylow in a finite-element simulation code to determine the specific energy absorption

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

21

rate from exposures to nonionizing electromagnetic fields.42 In 1997, his colleague Jones adopted NORMAN to estimate organ doses from external and internal photon sources.43 In 2005, Dimbylow developed an adult female phantom, NAOMI, also from MRI scans.44 The phantom was rescaled to a height of 1.63 m and a mass of 60 kg, the dimensions of the ICRP Reference Woman. However, to date, the NAOMI phantom has been used only in nonionizing radiation calculations. In 2005, a revised version of the NORMAN phantom, called NORMAN-5, was created by Ferrari & Gualdrini from ENEA-ION Istituto di Radioprotezione in Italy to derive external photon dose data.46 On year later, Dimbylow merged the NAOMI with the stylized fetal phantoms developed by Chen to create a series of hybrid phantoms of pregnant women.47 The process of adjusting two types of geometrical information was reported to be cumbersome. In 1999, Caon et al. from Flinders University, Australia, reported a torso phantom named ADELAIDE created from CT images of a 14 year old girl.23,24 As described in more detail in Chapter 4, this phantom was interesting because, for some time, it was the only set of data for a teenage girl and their studies provided CT dose estimates for this patient group. Chapter 4 also includes discussion on how to define the voxel geometry for Monte Carlo calculations. Caon later summarized his and other researchers’ experience on voxel phantoms.110 Chapter 6 provides a detailed review of the work related to the VIP-Man voxel phantom reported in 2000 by Xu and two of his students at Rensselaer Polytechnic Institute (RPI) in the United States.63 VIP-Man was the first phantom that was based on cross-sectional color photographic images of a cadaver. The original photographs were of a 39 year old male, which were made available through the National Library of Medicine’s famous Visible Human Project (VHP). VIP-Man is unique because the digitally captured color transversal photos had the best resolutions at that time: the pixel resolution is 0.33 mm × 0.33 mm and each photograph was taken after removal (by shaving) of each successive 1 mm layer by a cryomacrotome.63 The VIP-Man phantom consists of more than 3.7 billion voxels, and the original images were segmented to yield more than 1400 organs and tissues, although only approximately 80 were ultimately adopted for radiation dosimetry purposes. With ultrafine and color images, attempts were made to segment and label a number of small and radiosensitive tissues: stomach mucosa, skin, and red bone marrow. The finalized VIP-Man phantom had a heavy body mass of 103 kg, which served as an interesting variation from the ICRP reference value. The VIP-Man was used for a large number of studies in health and medical physics, as described in Chapter 6. Later, this group would extend the 3D phantom into 4D using the NURBS technique to be discussed in the section below. It should be noted here that several other groups in Table 1.2 also used the VHP images, but they primarily considered the CT data set without the arms. In 2004, Shi and Xu from RPI also reported the Pregnant Woman phantom developed from rare partial-body CT images of an 8 month pregnant patient.64 Chapter 12 is dedicated to phantoms for pregnant women and fetuses, including the earlier stylized phantoms by Stabin et al. as well as a series of BREP phantoms. Using CT scans of a physical phantom called RANDO, the RPI group created a voxel RANDO phantom for studies involving external beam treatment. Realizing the need for additional phantoms representing children of various ages, Bolch and colleagues from the University of Florida (UF) developed a series of pediatric voxel phantoms that appeared between 2002 and 2006 which represented children with ages ranging from newborn to 15 years old.69–71 Chapter 8 describes an early procedure they used for developing the newborn voxel phantoms. This approach was later extended to two groups (Groups A and B) of phantoms. Group A is composed of male and female

22

Handbook of Anatomical Models for Radiation Dosimetry

voxel phantoms of a newborn, a 1 year old, a 5 years old, a 10 years old, and a 15 years old for whom the phantom stature, total weight, and individual organ masses are targeted to within 1% of ICRP Publication 89 reference values. Group B phantoms are constructed by scaling the Group A phantoms up and down to yield phantom at each 1 year age interval, from newborn to 15 years old. The intent of the UF pediatric series was to provide a reference library of phantoms that could be matched to individual patients for age-specific organ dose assessment. Chapter 8 also describes their more recent efforts to develop BREP phantoms using the NURBS techniques that will be reviewed in the section below. Two Japanese groups were noted in Table 1.2 for their independent efforts to develop voxel phantoms since 2001.52–61 As described in Chapter 9, Saito et al. from the Japanese Atomic Energy Research Institute (JAERI) developed an adult male model named Otoko (the first Asian phantom) and an adult female phantom named Onago. More recently, Saito et al. has developed the JM, JM2, and JF phantoms which have a refined vertical slice thickness.52–56 These phantoms were used mainly for radiation dosimetry applications in Japan. The work of Saito et al. was influenced by earlier projects at the GSF, Germany. The other group, Nagaoka et al., from the National Institute of Information and Communications Technology (NIICT), Japan reported an adult male model, named TARO, and an adult female model, named HANAKO, developed from MR images for radiofrequency electromagneticfield studies.57–60 Later Nagaoka et al. would use a free-form deformation (FFD) to change the exterior features of the adult male phantom to develop Deformed Children phantoms of 3 years old, 5 years old, and 7 years old.61 The authors reported that it was difficult to develop these phantoms with the FFD algorithm and the internal organs are not adjusted to age-dependent values. As described in Chapter 10 by Lee and Kim, since 2004, several Korean phantoms have been developed by researchers at Hanyang University in Korea from various image sources: Korean Man (KORMAN), Korean Typical MAN-1 (KTMAN-1), Korean Typical Man-2 (KTMAN-2), and High-Definition Reference Korean (HDRK), and Korean WOMAN (KORWOMAN). The HDRK phantom was based on sectioned color photographs of an adult male cadaver that has high image resolution.39,40 The early work on these phantoms was carried out by the two Lee brothers who moved in the early 2000s to the University of Florida, where they gradually published work on the Korean phantoms and also made important contributions to the NURBS-based phantoms later. Kim spent several years in the United States to complete his PhD from Texas A&M University and then to serve as a research professor at RPI before returning to Hanyang University as a faculty member in the early 2000s. Their separate involvements in the area of voxel phantom development clearly originated from their experiences in the United States. Three voxel phantoms representing an adult Chinese male have been reported since 2007: CNMAN produced from color photographs of a cadaver by the China Institute for Radiation Protection,17 VCH produced from a different set of cadaver color photographs by the Huazhong University of Science and Technology,48–50 and CVP produced from MR images by Tsinghua University.66,67 The Chinese government undertook the Chinese version of the VHP that resulted in multiple cadaver image data sets, some with slice thickness as fine as 0.2 mm. These phantoms and their applications in ionizing radiation dosimetry are described and compared in Chapter 11. The lead developer of the CNMAN phantom, B Zhang, served as a research associate at RPI in 2007–2008. ORNL has not been actively involved in the development of voxel phantoms, although Eckerman was instrumental in the work at GSF related to the ICRP Reference Computational Phantoms and a number of voxel phantom projects at several universities in the United States. The only reported effort was that of Akkurt et al. in 2008 that involves a hybrid of voxel and stylized geometries.62

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

23

People who have been interested in nonionizing radiation applications form a different group of voxel phantom developers listed in Table 1.2. Most of this work was neglected in the previous review articles by Caon110 and by Zaidi and Xu.111 Interestingly, the phantoms used for studies of temperature rise in the human body from the interactions of radiofrequency energy were constructed through nearly identical steps and some of these phantoms, such as the NORMAN phantom, have been used for both ionizing and nonionizing radiation applications. However, the majority of voxel phantoms were developed with only one application in mind. Table 1.2 clearly labels voxel phantoms that have been used for nonionizing applications: the Visible Man from the VHP color photographs by the Brooks Air Force,15,16 the DAM adult male phantom from MR images by a group in Italy,19 the SILVY 30 week pregnant woman phantom from hybrid CT (originally obtained by RPI) and MR images by the Graz University of Technology, Austria,25,26 the Virtual Family for two adults and two children from MR images by IT’IS, Switzerland,51 the MEET Man from VHP color photographs by the University of Karsrule, Germany,72,73 and the Anatomically Based Model from MR images by the University of Utah.74 The redundancy in developing voxel phantoms from similar image sets such as the VHP is obvious. 1.3.3 BREP Phantoms from 2000s to Present Five groups reported a total of 28 BREP-based phantoms, which are summarized in Table 1.3. Segars’s PhD thesis at the University of North Carolina was the first publication that systematically described the NURBS-based modeling techniques.81 As detailed in Chapter 5, the Cardiac-Torso (NCAT) phantom was developed from the Visible Human CT image data set and the 3D anatomy was later extended into the fourth dimension to model cardiac and respiratory motions. The beating heart model of the 4D NCAT was based on 4D tagged MRI data from a real patient. The 4D NCAT phantom offers a vast improvement over the stylized MCAT phantom, with more realistic models of the anatomy and the cardiac system, and the respiratory motions. The 4D NCAT has gained a widespread use particularly in nuclear medicine imaging research for evaluating and improving myocardial SPECT imaging. The conceptual design of the NCAT phantom also served as basis for the development of a 4D digital mouse phantom named MOBY.83,84 The 4D Extended Cardiac-torso (XCAT) phantom was recently developed as the next version of the 4D NCAT. It includes more detailed and realistic anatomy and physiology, suitable for use in higher-resolution imaging applications. The XCAT phantom includes whole-body male and female anatomies based on the high-resolution Visible Male and Female anatomical data sets.82 In addition to the basic anatomy, the cardiac and respiratory motions were also updated in the XCAT phantom. Anatomically variable XCAT phantoms are currently under development. In 2005, the research group led by Xu at RPI used the 3D VIP-Man phantom to simulate respiratory motions by adopting the gated respiratory motion data of the NCAT phantom.85 The 4D VIP-Man Chest phantom was used to study external-beam treatment planning for a lung cancer patient.86 The group later decided to apply the BREP techniques to a more challenging problem and, in 2007, reported the development of a series of phantoms representing a pregnant woman and her fetus at the end of 3, 6, and 9 month gestations.87 These phantoms, referred to as the RPI Pregnant Females, were defined by polygonal meshes which were derived from separate anatomical information of a nonpregnant female, a 7 month pregnant woman CT data set, and a mesh model of the fetus. The organ volumes were adjusted in the mesh format using a commercial software package.87 The paper by Xu et al.87 was rated one of the 10 best papers in 2007 by Physics in Medicine and Biology. Chapter 12 presents details on this set of phantoms. Continuing their triangular mesh approach,

RPI, USA

Duke University, USA (formerly with the University of North Carolina)

Developers

RPI-Pregnant females (3, 6, and 9 month)

12

Polygon meshes

NURBS

4D VIP-Man chest

6

NURBS

MOBY

NURBS

NURBS

5

Data Types

XCAT

NCAT

Chapters in This Phantom Names Book

Organ surfaces were extracted from the 3D VIP-Man phantom and then extended to 4D by adding the respiration of the NCAT phantom. Based on a mixture of anatomical data. Organs of the mother and fetus were adjusted to match ICRP-89 references.

NURBS-based cardiac torso phantom including organs from the VHP ct data of the male and female. Gated MRI data set of a normal patient and 3D angiogram data are used for motion modeling. 4D eXtended cardiac torso phantom based on the NCAT, including more detailed and realistic anatomy and physiology. Mouse phantom from MR images.

Anatomical Features

3, 6, and 9 month pregnant female

16 week old male C57bl/ 6 mouse Caucasian 39 year old male cadaver

Caucasian 39 year old male and 59 year old female

Caucasian 39 year old male and 59 year old female

Human Subjects

I

I

I

I

I

[87]

[85,86]

[83,84]

[82]

[81]

Ionizing (I) or Nonionizing Radiation (N) References

List of Developers, in Alphabetical Order, of BREP Computational Phantoms Including Information on Phantom Names, Chapters in This Book, Phantom Data Types, Anatomical Features, Human Subjects, Ionizing versus Nonionizing Radiation, and References

TABLE 1.3

24 Handbook of Anatomical Models for Radiation Dosimetry

UFH-NURBS phantoms

Pregnant female (nine phantoms)

Adult and pediatric phantom series (seven phantoms)

University of Florida, USA

University of Houston, USA

Vanderbilt University, USA

Adult Male and Female

13

29

8

14

Based on mesh anatomical models that Adult male and female are adjusted to match with ICRP-89 references. Software supports deformation and posture change. NURBS UF Hybrid NURBS based on previous Caucasian 6 day old female voxel phantoms. newborn cadaver, 14 year male patient and two 14 year female patients. BREP/stl Nine phantoms of limited organs A pregnant woman in the formatted covering 1–9 month pregnant females 34th gestational week and a CAD from MRI of a nonpregnant female and nonpregnant female pregnant woman. NURBS Derived from the NCAT phantom Caucasian adult male and with organ and body masses adjusted to female, newborn, 1 year old, match ICRP-89 references. 5 years old, 10 years old, and 15 years old

Polygon meshes

I

[93]

[92]

[90,91]

I

N

[88,89]

I

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution 25

26

Handbook of Anatomical Models for Radiation Dosimetry

this group reported in 2008 the development of a pair of adult male and female phantoms, the so-called RPI Adult Male and Female.88,89 As described in Chapter 14, this pair of adult phantoms was carefully adjusted to match the ICRP-89 reference values for more than 70 organs and 45 bones (including cortical bone, spongiosa, and cavities) as well as muscles. Several software algorithms were systematically developed to automate the deformation and organ overlap detection that were based entirely on about 126 sets of triangle meshes. The RPI Adult Male and Female phantoms are mesh-based BREP phantoms. As an application, the female phantom was recently used to create phantoms of female workers with different breast size for the purpose of studying the effect of this parameter on the lung counting of internally deposited radionuclides.89 The mesh models had to be converted to voxels to work with Monte Carlo codes that only handle CSG shapes. In 2007 and 2008, the UF group led by Bolch reported their work on “hybrid” male and female phantoms of newborn and 15 year old patients.90,91 They created the BREP phantom series, called UFH-NURBS phantoms using the following steps. First, they segmented patient-specific CT image data from which they then generated polygonal meshes. These meshes were then converted to the NURBS format using commercial software. In this last process, they extracted several contours from the polygonal meshes, and then generated the NURBS surfaces by a software tool called “lofting.” It was then in the NURBS geometrical domain they carried out organ adjustment to match the ICRP89 reference. Therefore, their phantoms are in fact NURBS-based BREP phantoms, like those developed by Segars et al.81 In the fi nal step, the NURBS-based phantoms were voxelized so that they could be implemented in Monte Carlo calculations. However, in order to voxelize the smooth NURBS models, they transferred the NURBS surfaces back the polygonal meshes. More details are available in Chapter 8. The paper by Lee et al.90 was also rated one of the 10 best papers in 2007 by Physics in Medicine and Biology. In 2008, the Vanderbilt group led by Stabin, in collaboration with Segars from Duke University, reported a “family” of adult and pediatric phantoms by adapting the NURBSbased NCAT adult male and female phantoms.93 ICRP-89 reference body and organ values were used to adjust NURBS surfaces. The authors state in Chapter 13 several advantages of this approach: (1) NURBS-based phantoms can be developed much more quickly than working with voxels and manually segmenting individual patient image data sets; (2) the phantoms have a higher level of internal consistency; and (3) the phantoms are complete from head to toe, thus avoiding the problem of missing organs in some of the medical images. It is noted that the groups at RPI, UF, and Vanderbilt (and Duke) developed these BREP phantoms as part of the joint Virtual Patients Project funded by the NCI as well as other individual projects. The last item in Table 1.3 is a series of nine phantoms representing a pregnant female in each gestational month developed by a group from the University of Houston and the U.S. Food Drug Administration (FDA) for studying the effects of radiofrequencies emitted from various electronic devices.92 These phantoms only include a limited number of organs such as the body, placenta, embryonic fluid, bladder, bone, fetus, and the uterus. They used patient-specific MR images and CAD software to model the organ shapes. The applications of these phantoms for nonionizing radiation are covered in detail in Chapter 29.

1.4 Physical Phantoms Table 1.4 summarizes selected physical phantoms that are often used to benchmark calculations performed on computational phantoms.94–103 These phantoms are typically used

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

27

for three different applications: external radiation dosimetry, internal radiation dosimetry, and imaging quality assurance. For external radiation dosimetry, a physical phantom is designed so that small radiation dosimeters can be inserted in different locations of the phantom to measure doses from external irradiation. Examples of this type of phantom include the RANDO phantom by the Phantom Laboratory and the ATOM phantom by the CIRS, which contain tissue equivalent slices that have anatomical maps and cavities for organ dose measurements.94–96 Internal dosimetry phantoms are designed to either contain removable organs that are doped with long-lived radioactive materials or hollow body regions that are filled with short-lived radioactive liquids to mimic an internally exposed individual for the purposes of calibrating bio-assay devices (e.g., a lung counting or nuclear medicine imaging devices). The Physical Torso Phantom by Lawrence Livermore National Laboratory (LLNL) and the Bottle Manikin Absorption (BOMAB) phantom family by the Radiation Protection Bureau, Canada are examples of bioassay calibration phantoms. There are many phantoms that are used for image quality assurance purposes. Most of these phantoms cover only partial body and some are anatomically very simple. Table 1.4 lists examples of several such phantoms by the CIRS and Kyoto Kagaku Co. that are used for image analysis. With anatomically realistic computational phantoms discussed earlier, the UF group led by Hintenlang has fabricated several physical phantoms representing a newborn, 1 year old, and adult male. Details about this effort are provided in Chapter 16. Rapid prototyping processes offer great promise to quickly produce physical phantoms from patient-specific data.

1.5 Monte Carlo Codes Used with Computational Phantoms Computational phantoms are integrated with Monte Carlo codes that simulate radiation transport inside the human body for the purposes of determining the energy and pattern of radiation interactions. Most medical physics applications employ photons and electrons with energies up to 20 MeV and protons up to 250 MeV. But health physics dosimetry often also considers neutron sources found in nuclear reactors. For radiation protection in a high-energy physics accelerator facility or space radiation environment additional particles with energies in the TeV range must be taken into account. Each type of radiation interacts with matter differently. For example, photons (such as x-rays or gamma-rays) interact primarily via the photoelectric effect, Compton scattering, and pair production processes.121 The probability of an interaction occurring within an organ or tissue is determined by “nuclear cross sections” that are associated with the specific radiation energy, the tissue electron density, and the tissue chemical composition. Mathematically, the differential cross section per electron for a photon undergoing the Compton scattering at angle ϕ per unit solid angle Ω is analytically determined by 2

r 2 ⎛ h ν′ ⎞ ⎛ h ν h ν′ dσ ⎞ = 0 ⎜ + − sin 2ϕ⎟ ⎟ ⎜ ⎠ dΩϕ 2 ⎝ h ν ⎠ ⎝ h ν′ h ν

(1.2)

where r0 is the classical electron radius, hν and hν¢ are the photon energies before and after the scattering, respectively. Extensive photon cross-section libraries have been developed.122,123

RANDO

Alderson Research Laboratories (acquired by the Phantom Laboratory), USA

LLNL, USA

CTU-41

Kyoto Kagaku, co. LTD, Japan

Chest phantom N1 “lungman” LLNL

3D sectional torso phantom PBU-50

CIRS, USA

ATOM

Phantom Names

Developers

16

16

16

Chapters in This Book

Bone, lung, and soft tissue are included. Standard phantom includes head, torso, upper femur, and genitalia. Legs and arms are included with the newborn and 1 year pediatric phantoms. Breasts can be added. Removable lungs, heart, liver, pancreas, kidney, and spleen are included. The skeleton, lungs, liver, mediastinum, and kidney models are embedded in soft tissue substitute. One-piece anthropomorphic torso phantom with anatomical structures. The inner components consists of mediastinum, pulmonary vasculature, and an abdomen block. Removable organs such as the lungs, heart, liver, kidneys, spleen, tracheobronchial lymph nodes are included. Chest plates simulate different chest wall thicknesses. The first generation of LLNL phantom contains a real human male rib cage.

Lungs, soft tissue, and breasts are included; natural human skeletons were used.

Anatomical Features

Caucasian adult male

Japanese adult male Japanese

Caucasian adult male torso Japanese adult male

Caucasian newborn, 1, 5, and 10 year old children, adult male and female

Caucasian adult male and female

Human Subjects

I

I

[98]

[97]

[97]

[97]

I

I

[96]

[96]

[94,95]

I

I

I

Ionizing (I) or Nonionizing Radiation (N) References

List of Developers, in Alphabetical Order, of Physical Phantoms Including Information on Phantom Names, Chapters in This Book, Anatomical Features, Human Subjects, Ionizing versus Nonionizing Radiation, and References

TABLE 1.4

28 Handbook of Anatomical Models for Radiation Dosimetry

University of Florida, USA

Radiation Health Research Institute of Korea Hydro and Nuclear Power, Korea Radiation Protection Bureau, Canada

UF newborn physical phantom

16

20

BOMAB phantom family

MIRD stylized newborn UF 1 year old UF adult male

16

Typical Korean male

Based on UF newborn voxel phantom including soft tissue, bone, and lungs.

CT images. CT images. Newborn cadaver

1 year old child Adult male

Based on CT images of a human subject. The Korean adult male phantom contains bone, lungs, and soft tissue without arms and legs. The rapid prototyping and manufacturing technique were used. Each phantom is composed of 10 elliptical Canadian 5 and containers representing head, neck, chest, 10 years old, gut, arm, thigh, and calf. reference man, reference woman, 5 and 95 percentile men. Based on stylized computational phantom. Newborn

I

I

[102,103]

[101] [101]

[101]

[100]

I

I

[99]

I

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution 29

30

Handbook of Anatomical Models for Radiation Dosimetry

In general, Boltzmann radiation transport problems described by various differential, integral, and integro-differential equations can be solved by numerical computational methods including finite difference, finite element, discrete ordinates, and Monte Carlo. However, only the Monte Carlo methods are currently able to account for all aspects of particle interactions within 3D heterogeneous media such as the human body. Monte Carlo methods, which are based on statistical simulations, have a long history, but the real application to radiation transport simulations and the associated software development came from the need for nuclear weapons research at Los Alamos National Laboratory during the World War II.124 In a Monte Carlo code, random numbers are used to determine the distance and fate of a particle by comparing interaction probabilities for every geometrical region of interest. This rather tedious process is repeated for an extremely large number of particles, and each particle is tracked in the 3D anatomical model until all its energy is absorbed. The inherent statistical uncertainty can be controlled to be less than 1%, which is often more precise than an experimental result performed in a physical phantom using a dosimeter (for quantities such as the absorbed dose). Computers have significantly improved in affordability and computing power in the last 20 years. Public-domain Monte Carlo code packages are well supported by scientists at national labs and are updated constantly with the help from a very large pool of users. As a result, there has been widespread interest in recent decades in the use of Monte Carlo techniques in all aspects of nuclear engineering, health physics, and medical physics. First of all, simulations involving a computational phantom need to defi ne the anatomical geometries in terms of shapes that are accepted by a Monte Carlo code. In addition, information on the density and chemical composition for each of the identified organs and tissues of interest must be specified. Then, the radiation source terms are modeled according to a specific source-to-human-body irradiation condition. In radiation protection dosimetry involving external radiation, the most common irradiation geometries are external parallel beams impinging on the entire phantom that stands vertically in vacuum.125 The following standard irradiation geometries are often used: anterior–posterior (AP), posterior–anterior (PA), left lateral (LLAT), right lateral (RLAT), rotational (ROT), and isotropic (ISO), as discussed in detail in Chapter 18. The organ doses are often normalized by the particle fluence or air-kerma that can be measured using a radiation detector. For internal dosimetry, a quantity called the specific absorption fraction is calculated for each of the source-to-target organ pairs, as discussed in Chapter 20. A medical procedure, on the other hand, often involves partial-body irradiation and more complex source terms, such as in the case of x-ray CT examination discussed in Chapters 21 and 22. The benchmarking of these Monte Carlo calculations is often performed using a physical phantom and a radiation-delivering device in identical irradiation conditions. Most Monte Carlo codes were originally developed for nuclear engineering and high energy physics research. Although these codes have been vigorously validated for radiation physics, the software packages are often difficult to use without extensive experience. Some of the codes are not optimized for dealing with large numbers of voxels in the anatomical models. Nearly all existing Monte Carlo codes use the CSG-type of shapes. There are a few comprehensive reviews or introductory articles about the Monte Carlo methods for applications in health physics and medical physics.128–130 In the section below, several public-domain and popular Monte Carlo code systems are briefly summarized:

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

31

Monte Carlo N-Particle (MCNP), Monte Carlo N-Particle eXtended (MCNPX), Electron Gamma Shower (EGS), GEANT4, PENELOPE, and FLUKA. 1.5.1 MCNP and MCNPX The MCNP transport code system is a general-purpose Monte Carlo code that deals with neutron, photon, electron, or coupled neutron/photon/electron transport.131 MCNPX combines MCNP Version 5 with the LAHET intranuclear cascade code.132 MCNPX extends MCNP by supporting additional particle types, new cross-section libraries, and the ability to use physics models for energies where tabular data are not available. For years, both MCNP and MCNPX have been actively maintained by support individuals located at the Los Alamos National Laboratory. Both codes treat an arbitrary 3D configuration of materials in the CSG-type of geometric cells. Pointwise cross-section data typically are used, although group-wise data also are available. Important standard source defi nition features that make these codes very versatile and easy to use include: a powerful general source, criticality source, and surface source; both geometry and output tally plotters; a rich collection of variance reduction techniques; a flexible tally structure; and an extensive collection of cross-section data. The codes work very efficiently with the stylized and voxelized computational phantoms. However, there is a limitation on the maximum number of voxels in a computational phantom and this problem led to the increase in voxel size from 0.33 × 0.33 × 1 mm3 to 4 × 4 × 4 mm3 for the VIP-Man phantom that has been implemented in both codes (see Chapter 6). Recently, there has been an increased interest by the nuclear engineering community in developing algorithms to automatically convert CAD-based geometries to those accepted by MCNP (for examples, see an MS thesis which summarized the status126). Among several commercial software packages, the Monte Carlo automatic modeling system (MCAM) has many interesting features.127 However, none of them are able to run the BREP geometries directly. MCNP and MCNPX are widely used in the nuclear engineering and health physics communities, as well as medical physics. The future of these codes depend on the ability to meet various application needs such as those associated with advanced computational phantoms. The latest versions, MCNP 5.1.40 and MCNPX 2.6.0, can be requested from the Radiation Safety Information Computational Center (RSICC) (http:// www-rsicc.ornl.gov/). 1.5.2 EGS The EGS code system is a general-purpose package for the Monte Carlo simulations of the coupled transport of electrons and photons in an arbitrary geometry for particles with energies from a few keV up to several TeV. Some have referred to the EGS code as the de facto gold standard for medical physics radiation dosimetry. The EGSnrc system, developed and maintained by the National Research Council (NRC) of Canada, is an extended and improved version of the EGS4 package originally developed at Stanford Linear Accelerator (SLAC).133 Its current energy range of applicability is considered to be 1 keV to 10 GeV. Both NRC and SLAC have rights associated with EGS4 and EGSnrc. The VIP-Man phantom was also implemented in the EGS4 version without a problem in terms of the number of voxels. In Chapter 28, detailed information is given on the use of EGS code system for radiation treatment dosimetry using patient-specific phantoms.

32

Handbook of Anatomical Models for Radiation Dosimetry

1.5.3 GEANT4 GEANT4 is a toolkit for Monte Carlo simulations of electromagnetic, hadronic and optical processes, and a large set of long-lived particles, materials, and elements.107,134 The energy range covers from 250 eV to TeV. Its functionality, modeling capabilities, and performance are continuously extended and enhanced.135 The object-oriented design of GEANT4 allows the user to understand, customize, or extend the toolkit in all domains. At the same time, its modular architecture allows the user to load and use only the components needed. In terms of geometry definition, like other codes, GEANT4 cannot directly process BREP geometries such as NURBS and meshes. The software was originally developed and supported by RD44, a worldwide consortium of more than 100 scientists from different countries in Europe and elsewhere.136 GEANT4 is widely used in Europe and the high-energy physics accelerator community, and its open-source developmental approach will likely attract more users in the future. This code was used to implement several adult and pediatric voxel phantoms to study second cancer effects in patient treated with external protons (Chapter 25). The use of GEANT4 code for treatment planning with patient specific phantoms is described in Chapter 27. This chapter also includes a discussion of an effort to model organ motions. 1.5.4 PENELOPE PENELOPE is a Monte Carlo code system for photon–electron transport simulations, which was developed and supported in Spain.137,138 The code covers a wide energy range from a few hundred eV to about 1 GeV. A mixed procedure is used for the simulation of electron and positron interactions. Photon interactions (Rayleigh scattering, Compton scattering, photoelectric effect, and electron–positron pair production) and positron annihilation are simulated in a detailed way. The cross-section data in the very low-energy region of the code allows the calculation of radiation interactions at the cellular levels. Recently, there have been several attempts to develop direct simulations with BREP-type of geometries. The latest version is PENELOPE2006.139 1.5.5 FLUKA FLUKA is a general-purpose Monte Carlo code system for an extended range of 60 different particles: photons and electrons from 1 keV to thousands of TeV, neutrinos, muons of any energy, hadrons of energies up to 20 TeV and all the corresponding antiparticles, neutrons, and heavy ions.140–142 FLUKA can handle CSG geometries. The latest version of this package is FLUKA 2008.3.5. This software was developed and supported by an Italian group. A number of stylized and voxel phantoms were implemented in the FLUKA code but, despite its excellent features, the application of this code has been relatively limited.

1.6 Discussions The shift from stylized phantoms to voxel phantoms in the late 1980s was motivated by the desire to improve upon anatomical realism. The advent of modern computers and medical imaging fueled the research efforts by many researchers whose work is listed in Table 1.2. For a long period of time in the 1990s and early 2000s, however, it was unclear to the research community what roles voxel phantoms would play. If voxel phantoms were to replace stylized phantoms, how much improvement in dose estimates should be expected? There were strong

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

33

indications already that the approach of developing and applying the voxel phantoms was not perfect, as reviewed by Caon110 and more recently by Zaidi and Xu.111 For example, the segmentation of original images into organs and tissues required a very laborious and tedious manual process (there is only a limited number of organs such as the skeleton which can be processed automatically and semiautomatically), often taking months or even years to complete. The earlier phantoms, such as GOLEM, BABY, and CHILD, developed at GSF had relatively poor image slice thickness (from 4 to 8 mm), thus inevitably compromising the anatomical accuracy.27,28,31 Even today, there is no consensus as to what constitutes a true segmentation because the process often involves some level of user-specific assumption about the anatomy during the image analysis. Certain organs such as the GI tract have poor image contrast, and the segmentation is nearly impossible in CT without enhancement. A typical image data set at more than 2 mm × 2 mm pixel resolution is not fine enough to delineate many small radiosensitive organs. As a result, the skin of most existing voxel phantoms is defined artificially as the outermost layer of voxels. The segmentation of the red bone marrow is also challenging. Consequently, its dose is nearly always calculated empirically because it is not easy to model it directly in the phantom. When the developers of the VIP-Man phantoms reported that the red bone marrow was segmented from color pixels of 0.33 mm × 0.33 mm resolution, the work was scrutinized by others partially due to the lack of consensus about the segmentation process. The lack of standardized procedures contributed to the current situation that although many phantoms and dosimetry data are reported, the accuracy may be impossible to evaluate. Original voxel phantoms were realistic in depicting the anatomy, but they are personspecific. The anatomical differences between two equally realistic voxel phantoms surprised many developers who were used to the idea that a radiation protection phantom must represent the average population. Realizing that there would be likely only one set of such “reference” phantoms, many developers later rushed to revise the original voxel phantoms by adjusting the organ sizes in the original image data to match with the ICRPrecommended anatomical data. Others mixed anatomical sources from different subjects. In doing so, these phantoms lost the anatomical realism—which was the original motivation to abandon the stylized phantoms. In the history of voxel phantom development, Zubal was one of the first who shared the original image data with other users based on a mutual agreement. Heated debate continues regarding the intellectual property associated with the developed phantoms. It is often a technical necessity for a researcher to name a phantom that is associated with his or her research contribution. However, it is not clear who should own such a right because the four steps of developing a phantom discussed above can be carried out by different groups. One scenario is when the original images were acquired and segmented by one individual and then a different individual performed additional image processing and modification before implementing the data into a specific Monte Carlo code. Such changes produce a practically unique phantom, and proper naming is often required for research purposes. But the ownership of the product is not always agreed upon by involved individuals. Some have chosen not to share or to use phantoms partially due to this concern. Others are afraid that sharing may cost an advantage in research in a time when too many voxel phantoms exist. Since 1998, the ICRP’s Task Group on Dose Calculations (DOCAL) and the SNM’s MIRD Committee have been evaluating new dosimetry data from these tomographic phantoms. In particular, the DOCAL, which is administered by the ICRP Committee 2, has been developing guidelines to facilitate the shift from stylized phantoms. However, it is likely that other voxel phantoms will continue to be used in various applications. Recently, the ICRP has decided to establish the ICRP Reference Computational Phantoms, to be described in Chapter 15. The history of computational phantom development has shown that it is the need for application, not the need for police-making, that will determine the course of technological

34

Handbook of Anatomical Models for Radiation Dosimetry

advancement. The need for simulating organ motions for cardiac imaging resulted in the developments of MCAT (a CSG type) by Tsui et al.4 using quadric and superquadric surface equations and the NCAT models (a BREP type) by Segars81 using the NURBS technology. Xu and Shi (2005) adopted the method described by Segars to develop a respiration-simulating 4D VIP-Man model for the need to understand the effects of respiration on radiation treatment.83 Using the same approach, Lee et al.90 developed the size-adjustable pediatric models.90 More recently, the BREP-based pregnant females by Xu et al.87 and those by Stabin et al.93 are also examples of application-driven research that will likely continue to dominate the research horizon in the future. There are several issues that can be addressed in the very near future regarding the BREP-based methods. The NURBS geometries are flexible and computationally efficient, but fine details may be lost on certain organs that have a complex topology. On the other hand, polygonal models can be used to create very smooth surfaces with impressive anatomical detail by paying a price of having too many vertices. The human body is a particular challenge in modeling for that it consists of organ surfaces of complex and unique shapes. For cardiac and respiratory motions at the frequency range of 10–100 cycles per second, the mesh models may still be acceptable. However, previous work has also shown that the NURBS primitives were very easy to adopt for both real-time and non real-time applications. Therefore, the specific strategy should be based on the specific applications and user preference. Regardless of the specific BREP data structure, there is currently an urgent need for application-based software that can streamline the process described in this paper. Finally, it would be a breakthrough if the Monte Carlo calculations can be performed directly and accurately in the BREP geometry. To date, the history of phantom development has been centered on the “Reference Man” paradigm which mandates a computational phantom to match approximately the 50th percentile values in terms of body height and weight for a specific gender and age group. Given the anatomical specificity in any voxel phantom, the Reference Man concept works against the original wish to improve the dose estimate in a population of workers who are obviously different from the anatomy depicted by the one voxel phantom. In contrast, the BREP phantoms may have demonstrated the feasibility to develop new-generation phantoms that represent a much broader range of individuals in terms of body height and weight, as well as organ topology. These features were impossible even 10 years ago, but the technology and collective experience of the research community seems to support that idea that we should and can move beyond the “Reference Man” paradigm. In one attempt to collaborate on research, a special conference session was organized in 2005.143 During the conference, attendees proposed to form the Consortium on Computational Human Phantoms (CCHP; www.virtualphantoms.org) to facilitate data sharing, dissemination, and intercomparison. It is expected that, in the future, the concerted efforts, such as the CCHP, will allow a consensus to be developed involving as many individuals as possible in the field of human radiation dosimetry modeling. The last chapter of this book, Chapter 30, contains some thoughts on what are needed for near and long-term research.

1.7 Summary It is clear from the more than 40 years history of anatomical modeling that in the early days, between the 1960s and the 1980s, the research community relied on the first-generation of

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

35

CSG-type MIRD phantoms which required the least computational power. The attempt at developing realistic phantoms was first explored in the development of the CAM phantom which was little known outside of the space radiation dosimetry community. From the 1980s, the advent of medical imaging and more powerful computers made it possible to create voxel phantoms which were geometrically simple but anatomically realistic. Since the 2000s, however, several groups have demonstrated the feasibility of creating phantoms which are deformable using the BREP modeling techniques, thus opening the door for more advanced applications. For the first time, leading phantom developers and application experts collaborated to document the details of their historical research. In the subsequent chapters, you will witness a history of computational phantoms that is clearly in parallel with, and thus offers a unique perspective about, advances in computer technologies and medical imaging. By reading this book, you can obtain a unique sense of the computational phantom development process: the conception of an idea, the identification of original anatomical data, the solution of various computing problems, the ownership and sharing of results, as well as the satisfaction and frustration associated with any scientific endeavor. It is still an open question whether or not the recently developed ICRP Reference Computational Voxel Phantoms will actually improve radiation protection dosimetry. The long-time paradigm in representing a population of individuals with a limited number of “reference” phantoms seems to work against the power of image-based phantoms that are anatomical realistic, but person-specific. Is it necessary or feasible to bring about a paradigm change in this concept? If so, how should the research community participate? There is no doubt that we are at crossroads now, perhaps as we were 20 years ago. The history to be unfolded in the following chapters may help us arrive at some answers and hints about the future.

Acknowledgments Since 2003, I have been in close collaboration on projects related to computational phantoms with Drs. Michael Stabin, Randy Brill, Wesley Bolch, Harald Paganetti, and Paul Segars. Some of their ideas are reflected in this chapter. During the preparation of this chapter, Dr. Binquan Zhang, a visiting scholar, and Mr. Juying Zhang, a doctoral student, both from Rensselaer, helped me compile the phantom tables. My research in computational phantoms would have not been possible without support at different times by the following governmental agencies: National Science Foundation/CAREER Program (BES-9875532), National Library of Medicine (R03LM007964), National Cancer Institute (R42CA115122 via Vanderbilt), National Cancer Institute (R01CA116743), and National Library of Medicine (R01LM009362).

References 1. Kramer, R. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods: Part I. The male (ADAM) and female (EVA) adult mathematical phantoms, GSF-Report S-885, Neuherberg-Muenchen: Institut fuer Strahlenschutz, GSF-Forschungszentrum fuer Umwelt und Gesundheit, 1982.

36

Handbook of Anatomical Models for Radiation Dosimetry

2. Park, S., Lee, J.K., and Lee, C. Development of a Korean adult male computational phantom for internal dosimetry calculation, Radiat Prot Dosim, 121, 257, 2006. 3. Pretorius, P.H. et al. Evaluation of right and left ventricular volume and ejection fraction using a mathematical cardiac torso phantom, J Nucl Med, 38, 1528, 1997. 4. Tsui, B.M.W., Terry, J.A., and Gullberg, G.T. Evaluation of cardiac cone-beam single-photon emission computed-tomography using observer performance experiments and receiver operating characteristic analysis, Invest Radiol, 28, 1101, 1993. 5. Tsui, B.M.W. et al. Quantitative cardiac SPECT reconstruction with reduced image degradation due to patient anatomy, Nucl Sci IEEE Trans, 41, 2838, 1994. 6. Billings, M.P. and Yucker, W.R. The computerized anatomical man CAM model, NASA CR-134043, Washington, DC: Government Printing Office, 1973. 7. Fisher, H.L.J. and Snyder, W.S. Variation of dose delivered by 137Cs as a function of body size from infancy to adulthood, ORNL-4007, Oak Ridge, TN: Oak Ridge National Laboratory, p. 221, 1966. 8. Fish, H.L.J. and Snyder, W.S. Distribution of dose in the body from a source of gamma rays distributed uniformly in an organ, ORNL-4168, Oak Ridge, TN: Oak Ridge National Laboratory, p. 245, 1967. 9. Snyder, W.S. et al. Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, J Nucl Med, 10 (Suppl. 3), 7, 1969. 10. Snyder, W.S., Ford, M.R., and Warner, G.G. MIRD Pamphlet No. 5, Revised Estimates of specific absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, New York: Society of Nuclear Medicine, 1978. 11. Cristy, M. Mathematical phantoms representing children of various ages for use in estimates of internal dose, U.S. Nuclear Regulatory Commission Rep., NUREG/CR-1159, Also Oak Ridge National Laboratory Rep., ORNL/NUREG/TM-367, 1980. 12. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources I: Methods, ORNL/TM-8381/V1, Oak Ridge, TN: Oak Ridge National Laboratory, 1987. 13. Stabin, M.G. et al. Mathematical models and specific absorbed fractions of photon energy in the nonpregnant adult female and at the end of each trimester of pregnancy, ORNL/TM-12907, Oak Ridge, TN: Oak Ridge National Laboratory, 1995. 14. Chen, J. Mathematical models of the embryo and fetus for use in radiological protection, Health Phys, 86, 285, 2004. 15. Mason, P.A. et al. Effects of frequency, permittivity, and voxel size on predicted specific absorption rate values in biological tissue during electromagnetic-field exposure, IEEE Trans Microw Theory Tech, 48, 2050, 2000. 16. Wang, J.Q. et al. Computation with a parallel FDTD system of human-body effect on electromagnetic absorption for portable telephones, IEEE Trans Microw Theory Tech, 52, 53, 2004. 17. Zhang, B.Q. et al. CNMAN: A Chinese adult male voxel phantom constructed from color photographs of a visible anatomical data set, Radiat Prot Dosim, 124, 130, 2007. 18. Gjonaj, E. et al. High-resolution human anatomy models for advanced electromagnetic field computations, IEEE Trans Magnet, 38, 357, 2002. 19. Mazzurana, M. et al. A semi-automatic method for developing an anthropomorphic numerical model of dielectric anatomy by MRI, Phys Med Biol, 48, 3157, 2003. 20. Kramer, R. et al. All about MAX: A male adult voxel phantom for Monte Carlo calculations in radiation protection dosimetry, Phys Med Biol, 48, 1239, 2003. 21. Kramer, R. et al. All about FAX: A female adult voXel phantom for Monte Carlo calculation in radiation protection dosimetry, Phys Med Biol, 49, 5203, 2004. 22. Kramer, R. et al. MAX06 and FAX06: Update of two adult human phantoms for radiation protection dosimetry, Phys Med Biol, 51, 3331, 2006. 23. Caon, M., Bibbo, G., and Pattison, J. An EGS4-ready tomographic computational model of a 14-year-old female torso for calculating organ doses from CT examinations, Phys Med Biol, 44, 2213, 1999.

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

37

24. Caon, M., Bibbo, G., and Pattison, J. Monte Carlo calculated effective dose to teenage girls from computed tomography examinations, Radiat Prot Dosim, 90, 445, 2000. 25. Cech, R., Leitgeb, N., and Pediaditis, M. Fetal exposure to low frequency electric and magnetic fields, Phys Med Biol, 52, 879, 2007. 26. Cech, R., Leitgeb, N., and Pediaditis, M. Current densities in a pregnant woman model induced by simultaneous ELF electric and magnetic field exposure, Phys Med Biol, 53, 177, 2008. 27. Williams, G. et al. The calculation of dose from external photon exposures using reference and realistic human phantoms and Monte Carlo methods, Phys Med Biol, 31, 449, 1986. 28. Zankl, M. et al. The construction of computer tomographic phantoms and their application in radiology and radiation protection, Radiat Environ Biophys, 27, 153, 1988. 29. Fill, U.A. et al. Adult female voxel models of different stature and photon conversion coefficients for radiation protection, Health Phys, 86, 253, 2004. 30. Petoussi-Henss, N. et al. The GSF family of voxel phantoms, Phys Med Biol, 47, 89, 2002. 31. Zankl, M. et al. Organ dose conversion coefficients for external photon irradiation of male and female voxel models, Phys Med Biol, 47, 2367, 2002. 32. Zankl, M. et al. GSF male and female adult voxel models representing ICRP Reference Man— The present status, in Proceedings of the Monte Carlo Method: Versatility Unbounded in a Dynamic Computing World, Chattanooga, TN: American Nuclear Society, A Grange Park, 2005. 33. Becker, J. et al. About Katja, a virtual human phantom of a 24-week pregnant woman, in Proceedings of the 7th International Scientific Conference SATERRA “Human and Environment,” Mittweida, Germany, 2007. 34. Schlattl, H., Zankl, M., and Petoussi-Henss, N. Organ dose conversion coefficients for voxel models of the reference male and female from idealized photon exposures, Phys Med Biol, 52, 2123, 2007. 35. International Commission of Radiological Protection (ICRP). Adult Reference Computational Phantoms. ICRP publication 110, Elsevier, Amsterdam, 2009. 36. Lee, C., Lee, J., and Lee, C. Korean adult male voxel model KORMAN segmented from magnetic resonance images, Med Phys, 31, 1017, 2004. 37. Lee, C. and Lee, J. Reference Korean human models: Past, present, and future, in The Monte Carlo Method: Versatility Unbounded in a Dynamic Computing World, Chattanooga, TN, April 17–21, 2005. 38. Lee, C. et al. Development of the two Korean adult tomographic computational phantoms for organ dosimetry, Med Phys, 33, 380, 2006. 39. Choi, S.H. et al. Construction of a high-definition ‘Reference Korean’ voxel phantom fro organ and tissue radiation dose calculation, in World Congress on Medical Physics and Biomedical Engineering, Seoul, Korea, 2006. 40. Kim, C.H. et al. HDRK-Man: A whole-body voxel model based on high-resolution color slice images of a Korean adult male cadaver, Phys Med Biol, 53, 4093, 2008. 41. Dimbylow, P.J. The development of realistic voxel phantoms for electromagnetic field dosimetry, in Proceedings of the Workshop on Voxel Phantom Development, Chilton, U.K., 1996. 42. Dimbylow, P.J. FDTD calculations of the whole-body averaged SAR in an anatomically realistic voxel model of the human body from 1 MHz to 1 GHz, Phys Med Biol, 42, 479, 1997. 43. Jones, D.G. A realistic anthropomorphic phantom for calculating organ doses arising from external photon irradiation, Radiat Prot Dosim, 72, 21, 1997. 44. Dimbylow, P. Development of the female voxel phantom, NAOMI, and its application to calculations of induced current densities and electric fields from applied low frequency magnetic and electric fields, Phys Med Biol, 50, 1047, 2005. 45. Dimbylow, P. Resonance behaviour of whole-body averaged specific energy absorption rate (SAR) in the female voxel model, NAOMI, Phys Med Biol, 50, 4053, 2005. 46. Ferrari, P. and Gualdrini, G. An improved MCNP version of the NORMAN voxel phantom for dosimetry studies, Phys Med Biol, 50, 4299, 2005. 47. Dimbylow, P.J. Development of pregnant female, hybrid voxel-mathematical models and their application to the dosimetry of applied magnetic and electric fields at 50 Hz, Phys Med Biol, 51, 2383–2394, 2006.

38

Handbook of Anatomical Models for Radiation Dosimetry

48. Zhang, G. et al. Organ dose calculations by Monte Carlo modeling of the updated VCH adult male phantom against idealized external proton exposure, Phys Med Biol, 53, 3697, 2008. 49. Zhang, G. et al. The development and application of the visible Chinese human model for Monte Carlo dose calculations, Health Phys, 94, 118, 2008. 50. Zhang, G., Liu, Q., and Luo, Q.M. Monte Carlo simulations for external neutron dosimetry based on the visible Chinese human phantom, Phys Med Biol, 52, 7367, 2008. 51. Available at: http://www.itis.ethz.ch/index/index humanmodels.html. 52. Saito, K. et al. Construction of a computed tomographic phantom for a Japanese male adult and dose calculation system, Radiat Environ Biophys, 40, 69, 2001. 53. Sato, K. et al. Japanese adult male voxel phantom constructed on the basis of CT images, Radiat Prot Dosim, 123, 337, 2007. 54. Sato, K. et al. Development of a voxel phantom of Japanese adult male in upright posture, Radiat Prot Dosim, 127, 205, 2007. 55. Saito, K. et al. Construction of a voxel phantom based on CT data for a Japanese female adult and its use for calculation of organ doses from external electrons, Jpn J Health Phys, 43, 122, 2008. 56. Sato, K. et al. Development of a Japanese adult female voxel phantom, Radiat Environ Biophys, (to be submitted). 57. Lee, C., Nagaoka, T., and Lee, J.K. Implementation of Japanese male and female tomographic phantoms to multi-particle Monte Carlo code for ionizing radiation dosimetry, J Nucl Sci Technol, 43, 937, 2006. 58. Nagaoka, T. et al. Development of realistic high-resolution whole-body voxel models of Japanese adult males and females of average height and weight, and application of models to radio-frequency electromagnetic-field dosimetry, Phys Med Biol, 49, 1, 2004. 59. Nagaoka, T. et al. An anatomically realistic voxel model of the pregnant woman and numerical dosimetry for a whole-body exposure to RF electromagnetic fields, Engineering in Medicine and Biology Society, 2006. EMBS ‘06. 28th Annual International Conference of the IEEE, New York, 2006. 60. Nagaoka, T. et al. An anatomically realistic whole-body pregnant-woman model and specific absorption rates for pregnant-woman exposure to electromagnetic plane waves from 10 MHz to 2 GHz, Phys Med Biol, 52, 6731, 2007. 61. Nagaoka, T., Kunieda, E., and Watanabe, S. Proportion-corrected scaled voxel models for Japanese children and their application to the numerical dosimetry of specific absorption rate for frequencies from 30 MHz to 3 GHz, Phys Med Biol, 53, 6695, 2008. 62. Akkurt, H., Beka, K., and Eckerman, K. VOXMAT: Phantom model with combination of voxel and mathematical geometry, in 53rd Annual Health Physics Society Meeting, Pittsburgh, PA, July 13–17, 2008. 63. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Phys, 78, 476, 2000. 64. Shi, C. and Xu, X.G. Development of a 30-week-pregnant female tomographic model from computed tomography (CT) images for Monte Carlo organ dose calculations, Med Phys, 31, 2491, 2004. 65. Wang, B., Xu, X.G., and Kim, C.H. A Monte Carlo CT Model of the Rando Phantom, Am Nucl Soc Trans, 90, 473, 2004. 66. Li, J.L. et al. Organ dose conversion coefficients for external photon irradiation using the Chinese Voxel Phantom (CVP), Radiat Prot Dosim, 2008 (submitted). 67. Zeng, Z. et al. Dose assessment for space radiation using a proton differential dose spectrum, J Tsinghua Univ (Sci Technol), 46, 374, 2007. 68. Smans, K. et al. Calculation of organ doses in x-ray examinations of premature babies, Med Phys, 35, 556, 2008. 69. Nipper, J.C., Williams, J.L., and Bolch, W.E. Creation of two tomographic voxel models of paediatric patients in the first year of life, Phys Med Biol, 47, 3143, 2002.

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

39

70. Lee, C. et al. The UF series of tomographic computational phantoms of pediatric patients, Med Phys, 32, 3537, 2005. 71. Lee, C. et al. Whole-body voxel phantoms of paediatric patients—UF Series B, Phys Med Biol, 51, 4649, 2006. 72. Sachse, F.B. et al. MEET Man-Models for Simulation of Electromagnetic, Elastomechanic and Thermic Behavior of Man. Erstellung und technische Parameter, Institut für Biomedizinische Technik: Universität Karlsruhe, 1997. 73. Doerfel, H. and Heide, B. Calibration of a phoswich type partial body counter by Monte Carlo simulation of low-energy photon transport, Radiat Prot Dosim, 123, 464, 2007. 74. Tinniswood, A.D., Furse, C.M., and Gandhi, O.P. Power deposition in the head and neck of an anatomically based human body model for plane wave exposures, Phys Med Bio, 43, 2361, 1998. 75. Gibbs, S. and Pujol, J. A Monte Carlo method for patient dosimetry from diagnostic x-ray, Dentomaxillofac Radiol, 11, 25, 1982. 76. Gibbs, S. et al. Radiation doses to sensitive organs from intraoral dental radiography, Dentomaxillofac Radiol, 16, 67–77, 1987. 77. Gibbs, S.J. et al. Patient risk from interproximal radiography, Oral Surg Oral Med Oral Pathol Oral Radiol Endod, 58, 347, 1984. 78. Zubal, I.G. et al. Computerized three-dimensional segmented human anatomy, Med Phys, 21, 299, 1994. 79. Dawson, T.W., Caputa, K., and Stuchly, M.A. A comparison of 60 Hz uniform magnetic and electric induction in the human body, Phys Med Biol, 42, 2319, 1997. 80. Sjogreen, K. et al. Registration of emission and transmission whole-body scintillation-camera images, J Nucl Med, 42, 1563, 2001. 81. Segars, W.P. PhD thesis, University of North Carolina at Chapel Hill, p. 243, 2001. 82. Johns Hopkins Technology Transfer. http://www.jhtt.jhu.edu/. 83. Segars, W. and Tsui, B. 4D MOBY and NCAT phantoms for medical imaging simulation of mice and men, J Nucl Med Meet Abst, 48, 203P, 2007. 84. Segars, W.P. et al. Development of a 4-D digital mouse phantom for molecular imaging research, Mol Imag Biol, 6, 149, 2004. 85. Xu, X.G. and Shi, C. Preliminary development of a 4D anatomical model for Monte Carlo simulations, in Monte Carlo 2005 Topical Meeting. The Monte Carlo Method: Versatility Unbounded In a Dynamic Computing World, Chattanooga, TN, April 17–21, 2005. 86. Zhang, J. et al. Development of a geometry-based respiratory motion-simulating patient model for radiation treatment dosimetry, J Appl Clin Med Phys, 9, 16, 2008. 87. Xu, X.G. et al. A boundary-representation method for designing whole-body radiation dosimetry models: Pregnant females at the ends of three gestational periods—RPI-P3, -P6 and -P9, Phys Med Biol, 52, 7023, 2007. 88. Xu, X.G., Zhang, J.Y., and Na, Y.H. Preliminary data for mesh-based deformable phantom development: Is it possible to design person-specific phantoms on-demand. The International Conference on Radiation Shielding-11, Georgia, April 14–17, 2008. 89. Hegenbart, L. et al. A Monte Carlo study of lung counting efficiency for female workers of different breast sizes using deformable phantoms, Phys Med Biol, 53, 5527, 2008. 90. Lee, C. et al. Hybrid computational phantoms of the male and female newborn patient: NURBSbased whole-body models, Phys Med Biol, 52, 3309, 2007. 91. Lee, C. et al. Hybrid computational phantoms of the 15-year male and female adolescent: Applications to CT organ dosimetry for patients of variable morphometry, Med Phys, 35, 2366, 2008. 92. Wu, D.G. et al. Evaluations of specific absorption rate and temperature increase within pregnant female models in magnetic resonance imaging birdcage coils, IEEE Trans Microw Theory Tech, 54, 4472, 2006. 93. Stabin, M. et al. ICRP-89 based adult and pediatric phantom series, J Nucl Med Meet Abst, 49, 14, 2008.

40

Handbook of Anatomical Models for Radiation Dosimetry

94. Anderson, S.W. et al. An instrumented phantom system for analog computation of treatment plans, Am J Roentgenol, 87, 185, 1962. 95. The Phantom Library, http://www.phantomlab.com/rando.html. 96. Computerized Imaging Reference Systems (CIRS), http://www.cirsinc.com. 97. Kyoto Kagaku Co., LTD, http://www.kyotokagaku.com/. 98. Griffith, R.V. et al. Tissue equivalent torso phantom for intercalibration of in vivo transuranic nuclide counting facilities, in Advances in Radiation Protection Monitoring, STI/PUB/494 Proceedings of the IAEA Conference, IAEA-SM-229/56, Vienna: IAEA, 1978. 99. Kim, J.I. et al. Physical phantom of typical Korean male for radiation protection purpose, Radiat Prot Dosim, 118, 131, 2006. 100. Kramer, G.H., Burns, L., and Noel, L. The BRMD BOMAB phantom family, Health Phys, 61, 895, 1991. 101. Tresser, M.A. and Hintenlang, D.E. Construction of a newborn dosimetry phantom for measurement of effective dose, Health Phys, 76, S190, 1999. 102. Jones, A.K. et al. Tomographic physical phantom of the newborn child with real-time dosimetry I. Methods and techniques for construction, Med Phys, 33, 3274, 2006. 103. Staton, R.J. et al. A tomographic physical phantom of the newborn child with real-time dosimetry. II. Scaling factors for calculation of mean organ dose in pediatric radiography, Med Phys, 33, 3283, 2006. 104. Leyton, M. A Generative Theory of Shape, Berlin: Springer-Verlag, 2001. 105. Agostinelli, S. et al. Geant4 a simulation toolkit, Nucl Instrum Methods Phys Res Sec A, 506, 250, 2003. 106. Stroud, I. Boundary Representation Modeling Techniques, London: Springer-Verlag, 2006, ISBN 978-1-84628-312-3. 107. Geant4 Team 2007 Geant4 User’s Guide for Application Developers, http://geant4.web.cern.ch/geant4/ G4UsersDocuments/UsersGuides/ForApplicationDeveloper/html, Last accessed August 2007. 108. Wikipedia 2007, http://en.wikipedia.org/wiki/Wikipedia, a website maintained by Wikipedia, the free encyclopedia. Last accessed August 2007. 109. Zaidi, H. and Sgouros, G. Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine, Bristol: IoP, 2003. 110. Caon, M. Voxel-based computational models of real human anatomy: A review, Radiat Environ Biophys, 42, 229, 2004. 111. Zaidi, H. and Xu, X.G. Computational anthropomorphic models of the human anatomy: The path to realistic Monte Carlo modeling in radiological sciences, Annu Rev Biomed Eng, 9, 471, 2007. 112. International Commission on Radiological Protection. Report of the Task Group on Reference Man, ICRP Publication 23, 1975. 113. Hwang, J.M.L., Shoup, R.L., and Poston, J.W. Mathematical description of a newborn human for use in dosimetry calculations, ORNL/TM-5453, Oak Ridge, TN: Oak Ridge National Laboratory, 1976. 114. Jones, R.M. et al. The development and use of a fifteen-year-old equivalent mathematical phantom for internal dose calculations ORNL/TM-5278, Oak Ridge, TN: Oak Ridge National Laboratory, 1976. 115. Deus, S.F. and Poston, J.W. The development of a mathematical phantom representing a 10-yearold for use in internal dose calculations, in Proceedings of the Symposium on Radiopharmaceutical Dosimetry, Oak Ridge, TN: HEW Publication (FDA) 76–8044, 1976. 116. Bouchet, L.G. et al. MIRD pamphlet no. 15: Radionuclide S values in a revised dosimetric model of the adult head and brain, J Nuc Med, 40, 62, 1999. 117. International Commission on Radiological Protection. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Oxford: Elsevier, 2002. 118. ICRU. Phantoms and computational models in therapy, diagnosis and protection. ICRU Report 48, Bethesda, MD: International Commission on Radiation Units and Measurements, 1992.

Computational Phantoms for Radiation Dosimetry: A 40-Year History of Evolution

41

119. Gu, J.W., Dorgu, A., and Xu, X.G. Comparison of main software packages for CT dose reporting, Health Phys, 95, s1, 50, 2008. 120. Stovall, M., Smith, S.A., and Rosenstein, M. Tissue doses from radiotherapy of cancer of the uterine cervix, Med Phys, 16, 726, 1989. 121. Attix, F.H. Introduction to Radiological Physics and Radiation Dosimetry, New York: Wiley, 1986. 122. Hubbell, J.H. Photon cross sections, attenuation coefficients and energy absorption coefficients from 10 keV to 100 GeV, NSRDS-NBS 29, 1969. 123. Storm, E. and Israel, H.I. Photon cross sections from 1 keV to 100 MeV for elements Z = 1 to Z = 100, Nuclear Data Tables A, 7, 565, 1970. 124. Hammersley, J.M. and Handscomb, D.C. Monte Carlo Methods, London, New York: Methuen; Wiley, 1964. 125. Zhang, J.Y., Na, Y.H., and Xu, X.G. Development of whole-body phantoms representing an average adult male and female using surface-geometry methods, Med Phys, 35, 2875, 2008. 126. Furler, M. Methods of converting geometry in CAD to MCNP code. MS thesis, Rensselaer Polytechnic Institute, May 2007. 127. Liu, X et al. Development & application of MCNP auto-modeling tool: MCAM 3.0, in Proceedings of the 23rd Symposium on Fusion Technology, Venice, Italy, 2004. 128. Raeside, D.E. Monte Carlo principles and applications, Phys Med Biol, 21, 181, 1976. 129. Turner, J.E., Wright, H.A., and Hamm, R.N. A Monte Carlo primer for health physicists, Health Phys, 48, 717, 1985. 130. Andreo, P. Monte Carlo techniques in medical radiation physics, Phys Med Biol, 36, 861, 1991. 131. X-5 Monte Carlo team, 2003 X-5 Monte Carlo team. MCNP—A general Monte Carlo N-particle transport code, Version 5, volume I: Overview and theory, LA-UR-03-1987, Los Alamos National Laboratory, 2003. 132. Pelowitz, D.B. MCNPX User’s Manual Version 2.5.0, Los Alamos National Laboratory, Report LA-CP-05-0369, 2005. 133. http://www.irs.inms.nrc.ca/EGSnrc/EGSnrc.html 134. Agostinelli, S. et al. Geant4—A simulation toolkit. Nucl Instrum Methods Phys Res A, 506, 250, 2003. 135. Allison, J. et al. Geant4 developments and applications, IEEE Trans Nucl Sci, 53, 270, 2006. 136. http://geant4.web.cern.ch/geant4/collaboration/working_groups.shtml#wg.Run 137. http://www.nea.fr/html/dbprog/peneloperef.html 138. Salvat, F., Fernandez-Varea, J.M., and Sempau J. PENELOPE, a Code System for Monte Carlo Simulation of Electron and Photon Transport, France: OECD Publications, 2003, ISBN 92-64-02145-0. 139. Salvat, F., Fernández-Varea, J.M., and Sempau, J. PENELOPE, a code system for Monte Carlo simulation of electron and photon transport. Workshop Proceedings, Barcelona, Spain, 2006. 140. http://www.fluka.org/fluka.php 141. Fasso, A. et al. The FLUKA code: Description and benchmarking, in CERN-2005-10, INFN/ TC_05/11, SLAC-R-773, 2005. 142. Battistoni, G. et al. FLUKA: A multi-particle transport code, in Proceedings of the Hadronic Shower Simulation Workshop. September. Fermilab 6–8: AIP Conference Proceeding, Batavia, IL, 2006. 143. Tomographic models for radiation protection dosimetry session. Monte Carlo 2005 Topical Meeting: The Monte Carlo Method: Versatility Unbounded In A Dynamic Computing World, Chattanooga, TN, April 17–21, 2005.

2 Stylized Computational Phantoms Developed at ORNL and Elsewhere Keith F. Eckerman, John W. Poston, Sr., Wesley E. Bolch, and X. George Xu

CONTENTS 2.1 Introduction .................................................................................................................43 2.2 Historical Developments of Stylized Anthropomorphic Phantoms ................... 45 2.2.1 Phantoms Developed in Early 1960s by Reddy, Callahan, and Brownell ................................................................................................. 45 2.2.2 The MIRD-5 Phantom Developed in 1960s by Fisher and Snyder ........ 45 2.2.3 The Early Pediatric Phantoms Developed by Poston and Coworkers .............................................................................................. 48 2.2.4 The Stylized “Family” Phantom Series Developed in 1980s by Cristy and Eckerman .............................................................................. 49 2.2.5 Works Associated the MIRD Committee ..................................................54 2.2.6 Stylized Models of the Lower Abdomen .................................................. 56 2.2.7 Stylized Phantoms Representing Pregnancy Women ............................ 57 2.2.8 Other Stylized Models of the Human Anatomy ...................................... 58 2.2.9 GSF Gender-Specific Phantoms, ADAM and EVA .................................. 58 2.2.10 The CAM Phantom Developed by NASA for Space Radiation Dosimetry .................................................................................... 59 2.3 Summary ...................................................................................................................... 60 References ............................................................................................................................... 61

2.1 Introduction In internal radiation dosimetry, a phantom is a mathematical representation of the human body which, when coupled with a Monte Carlo radiation transport computer code, can be used to estimate the absorbed dose to tissue and organs of the body from radionuclides distributed in the body—either uniformly or located in specific organs. The phantom description includes information on the elemental compositions and densities of the body or specific organs to allow the Monte Carlo code to track the radiation interactions and energy deposition in the body. The first-generation computational phantoms were developed to better assess organ doses from internally deposited radioactive materials in workers and patients. Some of

43

44

Handbook of Anatomical Models for Radiation Dosimetry

the earliest dose assessment techniques were developed in the first third of the twentieth century primarily for use with interstitial radiation sources such as radium. According to Loevinger,1 the dosimetry of radioactive materials distributed in the body had been under consideration as early as the 1920s. Quimby has provided an excellent historical review of the early development of radiation dosimetry in nuclear medicine.2 The early techniques were adaptations of techniques used for external dose assessment with assumptions and corrections applied to account for the different types of radiation used.3 However, rather than being able to measure the exposure or the absorbed dose, an internal dose assessment required a calculation. Internal dose calculations were performed during early days using the formulation presented by Marinelli and his colleagues in the 1940s.4,5 These equations considered only the absorbed dose from beta-emitting radionuclides (classified as nonpenetrating radiation) and from gamma rays (penetrating radiation) emitted in the decay of these radiation sources. For beta radiation, the simple equation was Dβ = 73.8E βCT where Dβ is the absorbed dose (rad) Eβ is the average beta particle energy (MeV) C is the radionuclide concentration (μCi/g) T is the effective half-life of the radionuclide (days) This simple formula only required that the average beta energy, the concentration, and the effective half-life of the radionuclide be known. This equation assumes that the emitted beta-radiation energy was totally absorbed in the tissue or volume of interest. For a radionuclide that also emits gamma radiation, the equation was D γ = 0.0341CT Γg where Dγ is the absorbed dose (rad) C is the radionuclide concentration (μCi/g) T is the effective half-life of the radionuclide (days) Γ is the specific gamma-ray constant for the radionuclide (R/μCi–h) g is a geometry factor For gamma radiation, the geometry factor was applied to account for the radiation energy escaping the volume of interest (the organ) without contributing to the absorbed dose. Later, Loevinger and his colleagues elaborated on these formulations and provided dose calculations for simple shapes such as right circular cylinders as well as specific organs of interest, e.g., the thyroid gland.6,7 In 1959, the International Commission on Radiological Protection (ICRP) used very simple models for the internal dosimetry calculations associated with the Report of ICRP Committee II.8 In these calculations, each organ of the body was represented as a sphere with an “effective radius.” The radionuclide of interest was assumed to be located at the center of the sphere and the “effective absorbed energy” was calculated for each organ. Corrections were made for the photon energy lost from the sphere (similar to the Marinelli approach). In this approach, even the total body was represented as a 30 cm radius sphere. It is also interesting to note that the 30 cm radius sphere

Stylized Computational Phantoms Developed at ORNL and Elsewhere

45

was used for an organ designated as “muscle” as well as for the small intestine and the entire gastrointestinal tract. At the time, these approaches provided reasonably accurate estimates of the dose from a distributed radionuclide. However, most dosimetrists and researchers hoped for improved techniques and more accurate dosimetry estimates as technology developed. There was also a need for dose calculations for a number of new radionuclides introduced into nuclear medicine and more was known regarding the distribution and retention of these radionuclides in specific organs. Of course, the next step was to attempt to model individual organs of the body and ultimately the entire human body in a realistic manner. With the increase in the size and speed of computers, some progress occurred during the late 1950s and through the 1960s and eventually the efforts led to stylized anthropomorphic phantoms—those resemble the human anatomy.

2.2 Historical Developments of Stylized Anthropomorphic Phantoms 2.2.1 Phantoms Developed in Early 1960s by Reddy, Callahan, and Brownell Early phantoms were simple shapes: spheres, disks, and cylinders. To simplify the calculations, many early phantoms were assumed to be composed of water and no attempt was made to simulate the elemental composition of tissue.9,10 From 1964 through 1967, Reddy, Callahan, and Brownell published the results of their Monte Carlo calculations for photon dosimetry. They considered point and volume-distributed photon sources in organ phantoms represented by spheres, thick ellipsoids, flat ellipsoids, and elliptical cylinders. These authors focused on the concept of the absorbed fraction (AF) of energy, which is simply the fraction of the emitted photon energy absorbed in the region of interest. The concepts of the absorbed fraction of energy and the specific absorbed fraction, i.e., the absorbed fraction per gram of absorbing material were introduced by Loevinger and Berman in Medical Internal Radiation Dose (MIRD) Pamphlet No. 1.14 In 1968, Brownell et al. published an extensive compilation of these results for photon radiations as MIRD Pamphlet No. 3.15 In 1971, Ellett and Humes continued the calculation of absorbed fractions for small volumes containing photon-emitting radioactivity.16 The next step in development of phantoms was to construct models representing the entire trunk of an adult human. Initially, the trunk was simulated by a right circular cylinder. A typical phantom was a right cylinder 30 cm in diameter and 60 cm in height. This phantom was used for both external radiation sources as well as internal radiation sources. In some cases the phantom was subdivided into smaller regions, which could provide dose estimates for organs or tissues located in the vicinity of the target volume.17 At about the same time, models of specific, small organs (e.g., thyroid gland, gonads) were being placed inside the cylindrical phantom and again both external and internal radiation sources were being investigated. 2.2.2 The MIRD-5 Phantom Developed in 1960s by Fisher and Snyder A number of laboratories developed computational phantoms for use in their research during the late 1960s and early 1970s. To track all these developments would require an extensive discussion so the activities at the Oak Ridge National Laboratory (ORNL),

46

Handbook of Anatomical Models for Radiation Dosimetry

under the direction of Walter S. Snyder, will be discussed to illustrate the rapid development of realistic phantoms. In 1966, Fisher and Snyder described the development of an adult phantom of use in dose calculations.18 The adult phantom was assumed to be standing erect with the arms at the sides of the body. Three specific regions were defi ned; the head and neck, the trunk including the arms, and the legs. The head and neck were represented by a 14 cm × 20 cm elliptical cylinder with a height of 24 cm. The trunk and arms were modeled as a larger elliptical cylinder, 20 cm × 40 cm with a height of 70 cm. The legs below the buttocks were modeled as a truncated elliptical cone with a height of 80 cm. Regions of little dosimetric importance were not included, e.g., the hands, feet, ears, nose, etc. The composition of the phantom was assumed to be tissue distributed homogeneously throughout. No attempt was made to model the lungs or skeleton or to defi ne the locations of specific organs in the phantom. Approximately 120 subregions were defi ned in the phantom, which were used to assign approximate values of the absorbed doses to organs located within specific regions. In some cases, absorbed dose estimates for large organs required the evaluation of the doses deposited in several of these regions. In 1967, Fisher and Snyder reported on the development of an adult phantom with 22 internal organs and more than 100 subregions.19 This phantom represented the next step in the development of anthropomorphic phantoms for use in dose calculations. Although the skeleton and lung regions were represented in the phantom the phantom was homogeneous as these regions were not assigned specific densities and elemental compositions. This point was very misleading in that many of the early drawings indicated such regions. However, a careful reading of the research reports from the ORNL group will confirm the homogeneous composition. Estimates of the absorbed dose to the skeleton were obtained by summing all the energy deposited in the entire skeleton and dividing by the mass associated with the skeletal volume of unit density. No special treatments were applied to obtain absorbed dose estimates for the lungs. Calculations using the adult phantom agreed well with those of Ellett et al.11–13 but were about 50% lower than those obtained using the ICRP spherical organ methods.8 Even though the original phantom was designed for use with internally deposited radionuclides, Snyder saw many other applications. In addition, in 1967, he used the phantom to study the distribution of dose in the body from external, point sources of gamma rays.20 He studied four photon energies (0.07, 0.15, 0.5, and 1.0 MeV) and four different source locations at distances of 1 and 2 m from the center of the phantom. A heterogeneous phantom was the next logical step in the development of anthropomorphic phantoms. Snyder and his colleagues21 published an extensive compilation of data on the absorbed fraction of energy for monoenergetic photons sources, uniformly distributed in selected organs in this phantom. This phantom was composed of three regions: skeleton, lungs, and the remainder (soft tissue). The densities of these regions were about 1.5, 0.3, and 1.0 g/cm3, respectively. The organ masses were selected to follow as closely as possible the data of Reference Man.22 This phantom ultimately became known in the nuclear medicine community as the “MIRD phantom.” The MIRD phantom was developed by Snyder’s research group at ORNL and, even though Snyder chose to call the phantom a “standard man approximation,” it was based on information being compiled for ICRP Publication 2322 on Reference Man. In reality, there was a parallel effort at the time to provide improved estimates of absorbed dose and dose equivalent for the ICRP. These estimates were later published as the ICRP Publication 30 series.23

Stylized Computational Phantoms Developed at ORNL and Elsewhere

47

The heterogeneous phantom and its application contained three principal idealizations: 1. Simple geometrical shapes were used to approximate the shape of the body and its organs. Twenty-two internal organs were included in the design but other unimportant structures (the nose, hands, feet, etc.) were ignored. It was assumed that each organ or tissue was homogeneous in terms of composition and density. However, different compositions and densities were used for the skeletal region, the lungs, and the remainder of the body (soft tissue). 2. The monoenergetic radiation source was assumed to be uniformly distributed in one or more of the organs (i.e., the source organs). 3. Monte Carlo methods were used to obtain a set of photon histories upon which estimates of the energy deposited in the organs and tissues (i.e., target organs) of the phantom could be derived. Only photon transport was considered and energy transferred to electrons by photon interactions was assumed to be deposited locally. Fifteen source organs and 25 target organs or regions were included in the set of calculations. Twelve photon energies were considered ranging from 0.01 to 4.0 MeV. A limited number of histories were followed in each calculation (25,000–50,000 histories) and, for low-energy photons; the absorbed fractions of energy for many organs were unreliable and were not included in the tabulations. In 1978, Snyder et al. published the results of an extensive set of calculations of specific absorbed fractions of energy using an improved heterogeneous phantom.8 As with previous calculations, these results were for monoenergetic photon sources uniformly distributed in organs of the heterogeneous phantom. However, significant improvements had been made in the phantom. These improvements included 1. The head section was represented by a right elliptical cylinder topped by half and ellipsoid. 2. The leg regions consisted of the frustums of two circular cones. 3. The male genitalia were moved to a position outside the trunk and on the front of the revised model of the legs. 4. Detailed descriptions of the scapulae and the clavicles were included. 5. The stomach and upper and lower large intestine were modeled as organs with walls and contents. 6. A source and target region representing skin was added to the phantom. Twenty source organs and 20 target organs or regions were included in the set of calculations. Twelve photon energies were considered ranging from 0.01 to 4.0 MeV. The number of histories followed was increased to 60,000 in hopes of increasing the reliability of the results. However, as before, calculations for low-energy photons were very unreliable and other methods were used to provide an estimate of the specific absorbed fractions in these cases. Over the years, a number of changes (improvements) to the heterogeneous phantom have been made. However, the fundamental use of the phantom for internal dose calculations has remained essentially the same since its inception. Initially, Monte Carlo transport codes used in internal dose assessment were capable only of transporting photons. Perhaps one of the most widely used computer codes was

48

Handbook of Anatomical Models for Radiation Dosimetry

the ALGAM code developed at ORNL.25 This code was used with the original Snyder– Fisher phantom and continued to be the basic code used for calculations by the ORNL group as the complexity of the phantom increased. In these calculations, it was assumed that electrons, created by photon interactions, deposited their energy at the point of creation. That is, the ALGAM code was not capable of transporting electrons. Computer codes, developed more recently, now have the capability of transporting both photons and electrons, including the production of bremsstrahlung, and considering other types of photon and electron interactions.26,27 Subsequently investigations using a coupled electron/photon transport code indicated that for most situations, this apparent deficiency was not significant.28,29 2.2.3 The Early Pediatric Phantoms Developed by Poston and Coworkers Development of the adult human phantom by Snyder and his colleagues was paralleled by the development of phantoms representing humans of other ages.18 These phantoms represented children with ages of 0 (newborn), 1, 5, 10, and 15. These early designs were assumed to have outer dimensions that represented the average height, surface area, and body mass of a child of the particular age. All “pediatric phantoms” were obtained by applying a set of simple transformations to the axes of the Cartesian coordinate system in which the adult phantom was defi ned. These phantoms became known as the “similitude phantoms” because of their resemblance to children. This approach had its limitations because children are generally not just “little adults.” However, these phantoms were the first developed to answer a real need in the nuclear medicine community.30 Improvements in the pediatric models were closely linked with the development of the heterogeneous adult phantom. Even though these new phantoms were heterogeneous, the pediatric phantoms were obtained through the same transformation method. The outside dimensions were obtained by applying a series of transformations to the coordinate system and no consideration was given to the actual organ sizes or shapes of the “transformed” organs. Although the masses of these “transformed organs” had to be known for the calculation of absorbed dose, these masses were never published.31 The limitations associated with transforming the major axes of the adult phantom should be clear. Children are not simply small adults and their organs are not necessarily “smaller adult organs.” Some organs are larger in a child than in the adult and get smaller as the individual approaches adulthood, e.g., the thymus. In addition, it was important that the actual size, shape, location, and mass of each organ be known for accurate dose calculations. For these reasons, and others, a significant effort was undertaken at ORNL during the mid-1970s to develop individual pediatric phantoms based upon a careful review of the existing literature for each particular age. This effort produced the next generation of mathematical phantoms that, although they appeared to be modeled after the adult, were designed independently. Three “individual phantoms” were designed by Hwang et al.32,33 These were the newborn, the 1- and 5-year-old models. A separate effort was undertaken by Jones et al.34 for the 15 year old, and Deus and Poston35 undertook the design of a 10 year old after the other four designs were complete. The development of the 10 year old was significantly different from those for the other four ages. In fact, this design was intended to point the way to the next generation of more realistic phantoms. Even though the design was completed and used for a limited number of dose calculations, it was not popular because of the very complex geometry and other approaches to the development of phantoms were followed.36,37

Stylized Computational Phantoms Developed at ORNL and Elsewhere

49

2.2.4 The Stylized “Family” Phantom Series Developed in 1980s by Cristy and Eckerman Cristy of ORNL developed a new series of stylized models of various ages in 1980.36 The series included an adult, a newborn, and individuals of ages 1, 5, 10, and 15 developed from anthropological data (legs, trunk, and head) and from age-specific organ masses published in ICRP Publication 23.22 Although some of the organ shapes and centroids were still obtained using the similitude rule from the Snyder–Fisher adult model, these phantoms represented a great improvement for pediatric dosimetry over the similitude pediatric phantoms. These hermaphrodite phantoms presented new regions and improvements such as a new stylized breast tissue region for all ages, the inclusion of the new model of the heart developed by Coffey,38,39 and a new model of the thyroid. While the ORNL pediatric model series was initially published in 1980 by Cristy,36 these models were not readily utilized until 1987 with the publication of ORNL/TM-8381 authored by Cristy and Eckerman.37 The mathematical phantoms were designed by Cristy36 after the adult phantom of Snyder et al.43 but has different densities and chemical compositions for lung, skeletal, and soft tissues. (The term “soft tissues” will be used herein for all near-unit-density tissues, i.e., density, 1 g/cm3.) These phantoms have been described by Cristy,36 but several changes were introduced since the 1980 report and are summarized in ORNL/TM-8381 authored by Cristy and Eckerman.37 One major change was that the age-15 phantom was modified to represent both a 15-year-old male and an adult female, following the observation that the body weight and dimensions of a reference adult female are approximately the same as those in the age-15 phantom.22 The breasts, the ovaries, and the uterus in the age-15 phantom were modified to be appropriate for an adult female. Also, the size of the liver was changed slightly, and the position of the gallbladder was changed so as not to overlap the new liver. These changes are noted in the description of these organs. This phantom is labeled “15-AF” in subsequent publications. These changes coincided with efforts at GSF, Germany to develop gender-specific phantoms, such as the so-called ADAM and EVA—a revised version of the MIRD-5 phantoms reported by Kramer et al.40 The latest phantoms were used with the ETRAN Monte Carlo photon transport code41,42 to calculate specific absorbed fractions of energy in all five pediatric phantoms, as well as in the adult male, for 12 photon energies (0.01–4.0 MeV). Electron transport was not considered in these simulations and the electron energy was assumed to be locally deposited. The phantom labeled “Adult male,” although a hermaphrodite, in the descriptions below is the Snyder adult phantom,43 with certain organs modified as described by Cristy.36 In brief, these modifications were the following: female breast tissue was added to the trunk (this phantom, like all the others, is hermaphroditic and could represent a larger than average adult female), and the improved heart model of Coffey39 was fitted into the trunk. The lungs had to be redesigned to accommodate the new heart; the difference in size between right and left lungs—not represented in the Snyder phantom— was incorporated into the new design. The head was redesigned to incorporate the ideas of Hwang, Shoup, and Poston, 32 including a change in position of the thyroid. The gallbladder of Hwang et al.33 was added. A modification of the descending colon was made to eliminate a small overlap with the pelvic skeleton and to make the wall thickness uniform. Other minor changes were made so that the “Adult male” phantom would be consistent with the manner in which certain organs were fitted into the pediatric phantoms: the position of the adrenals, the position of the gallbladder, the size

Handbook of Anatomical Models for Radiation Dosimetry

50

of the pancreas, and the shape and position of the thymus were all changed for this reason. Two additional modifications to the “Adult male” phantom were made by Cristy and Eckerman.37 The volumes of the breasts and the uterus have been changed slightly to be consistent with the “15-AF” phantom. Eckerman and Ryman44 revised the head region to include a neck, added the esophagus, and consideration of the extra thoracic airways. Another noteworthy change was the chemical composition and density of each type of tissue in the phantoms (skeletal, lung, and soft tissues). Also, compositions of the skeletal and soft tissues of the newborn became different from those at other ages. These changes also affected organ masses and whole body masses which were tabulated by Cristy and Eckerman.37 The same phantom descriptions and diagrams were purposely followed in a series of ORNL technical reports by Snyder et al.,21 Cristy, 36 and Cristy and Eckerman.37 As shown in Figure 2.1, each phantom consists of three major sections: (1) an elliptical cylinder representing the trunk and arms; (2) two truncated circular cones representing the legs and feet; and (3) a circular cylinder on which sets an elliptical cylinder capped by half an ellipsoid representing the neck and head. Attached to the legs section is a small region with a planar front surface to contain the testes. Attached to the trunk are portions of two ellipsoids representing the female breasts (not shown in Figure 2.1). The

2A1 2B1

ORNL-DWG. 74-9373

H2 2A2

H1

Phantom dimensions and dose regions Age (year)

Weight (kg)

H1 (cm)

H2 (cm)

H3 (cm)

A1 (cm)

B1 (cm)

A2 (cm)

0 1 5 10 15 Adult

3.148 9.112 18.12 30.57 53.95 69.88

23 33 45 54 65 70

13 16 20 22 23 24

16 28.8 46 64 78 80

5.5 8 11 14 18 20

5 7 7.5 8 9 10

4.5 6.5 6.5 6.5 7 7

H3

The adult human phantom FIGURE 2.1 The adult male phantom and associated dimensions. The same descriptions and diagrams like this were purposely followed in a series of ORNL technical reports by Snyder et al., 21 Cristy, 36 and Cristy and Eckerman. 37

Stylized Computational Phantoms Developed at ORNL and Elsewhere

51

ORNL–DWG 79–19955

0 year

1 year

5 year

10 year

15 year

Adult

FIGURE 2.2 External views of the age-specific phantom, phantoms representing an adult and children at 15 (adult female), 10, 5, 1, and 0 (newborn) years of age. When used for an adult female, the 15 year phantom has breasts appropriate for a reference adult female, which are not shown. (From Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381, Oak Ridge, TN: Oak Ridge National Laboratory, 1987.)

arms are embedded in the trunk, and minor appendages such as fingers, feet, chin, and nose are omitted. Drawings depicting the external features of all the family phantoms are shown in Figure 2.2. The pediatric phantoms were designed to form a developmentally consistent family with the existing Snyder adult phantom. The exterior of each phantom has approximately the form of the human body; but, as in their adult phantom, there has been no attempt to introduce small variations which would be presumed to have only a small effect on the scattering of photons. Similarly, the description of the interior organs, while approximately correct as to size, shape, position, composition, and density, are simplified to provide formulas which are readily calculated on a digital computer. The exact specifications of the phantom and the internal organs are given below. Figure 2.3 shows a schematic view of the principal organs. In these phantoms, the body is represented as erect with the position z-axis directed upward toward the head. The x-axis is directed to the phantom’s left, and the y-axis is directed toward the posterior side of the phantom. The origin is taken at the center of the base of the trunk section of the phantom. In general, the dimensions (cm) are given to two decimal places. The use of two decimal places does not imply that the average dimensions in some human population are known to such precision. This use is for convenience in designing the organs with correct volumes and spatial relationships. The trunk section includes the arms and the pelvic region to the crotch. The female breasts are appended to the outside of the trunk section. The volumes and masses for the trunk given above do not include the breasts. The head section includes a neck, represented by a right circular cylinder, and the head, consisting of a right elliptical cylinder topped by half an ellipsoid. The trunk, exclusive of the female breasts, is represented by a solid elliptical cylinder. The legs region of each phantom consists of the frustrums of two circular cones.

Handbook of Anatomical Models for Radiation Dosimetry

52

Brain Skull Spine

Arm bone

Ribs

ORNL-DWG 56-B212AR2 Organs not shown Adrenals Stomach Marrow Pancreas Skin Spleen Ovaries Testes Thymus Thyroid Lungs Leg bones Heart

Liver Upper large intestine Uterus Bladder

Gall bladder Kidneys Small intestine Lower large intestine

Pelvis

0

5

10

Centimeters FIGURE 2.3 Anterior view of the principal organs in the head and trunk of the adult phantom developed by Snyder et al.43 Although the heart and head have been modified, this schematic illustrates the simplicity of the geometries of the organs. (From Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381, Oak Ridge, TN: Oak Ridge National Laboratory, 1987.)

The phantom developers clearly understood the simplification of the stylized modeling techniques in the phantom development. Figure 2.4 compares the idealized model of the skeleton (on the left) and a more realistic representation (on the right).37 The regional distributions of the active (hematopoietic) bone marrow and the inactive (fatty) marrow vary greatly with age. The approximate weights of the total (active plus inactive) marrow, the active marrow, and the inactive marrow as a function of age are also provided in Figure 2.4.37 Data from Hudson,45 Custer,46 ICRP, 22 and Woodard and Holodny47 were used to estimate the weight of the total marrow. The weights of active and inactive marrow were calculated from the total marrow values by the method of Cristy.48 The marrow, active or inactive, was assumed to be distributed uniformly in the bone regions defined. In calculating an absorbed fraction for active and for inactive marrow in these regions by the Monte Carlo computer program, it was assumed that the marrow absorbs energy per gram as efficiently as did bone. This assumption was not grossly in

Stylized Computational Phantoms Developed at ORNL and Elsewhere

Skeletal region defined in phantom Skull Spine Ribs Scapulae Arm bones—upper portion Clavicles Leg bones—upper portion Pelvis Total amount of active bone marrow:

53

ORNL–DWG 70-4810R2A Distribution of active marrow in adult phantom – of this report – of Snyder et al. (1974) 13.1% 28.4 10.2 4.8 1.9

8.3% 29.9 19.2 2.9 2.3

1.6 3.8

0.8 3.4

36.2

33.3

1500 g

1120 g

Active bone marrow FIGURE 2.4 The idealized model of the skeleton is shown on the left and a more realistic representation on the right. The shaded areas indicate where the active marrow is located in the adult. The amount of active marrow in given bones, expressed as the percentage of the active marrow in the body, is also given for the adult. (From Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/ TM-8381, Oak Ridge, TN: Oak Ridge National Laboratory, 1987.)

error at energies of 200 keV or more; but it is increasingly inaccurate at energies below 100 keV, where the photoelectric effect dominates the photon interaction process. The effect is to overestimate the dose to marrow and to underestimate the dose to the bone mineral component of the mixture. The researchers noted that it was impossible to define in terms of conventional geometry the intricate microscopic intermixture of bone and marrow spaces in a more realistic fashion in the macroscopic characterization used in photon transport. As a consequence, another method of calculating this absorbed fraction was developed.37 The breasts in the age-15 male/adult female phantom have been changed from those given by Cristy36 for the age-15 phantom. The latter were designed to represent adolescent breasts. Note also that the breasts in the “Adult male” phantom as described in Cristy36 are modified slightly here to be consistent with the age-15 male/adult female phantom. The authors stated that there had been some disagreement between Kramer and coworkers49,51 and Cristy36,50 on the appropriate size of the breast for a reference adult female. Cristy36 recommends a volume of 190–200 mL for the size of a single breast, in accord with the 180 g mass recommended by the ICRP.22 Kramer and coworkers first recommended a volume of about 365 mL49 and later changed their recommendation to 260–270 mL.50 The difference in recommended representative breast sizes (~195 mL vs. ~265 mL) by Cristy and Eckerman37 is similar to the difference between the median (193 mL) and the mean (238 mL) in one study50,52 (52, 50), and the standard deviation of the mean is large (50%).

54

Handbook of Anatomical Models for Radiation Dosimetry

Cristy53 argues that this difference in breast size does not yield important differences in estimates of dose to the breast from either internal or external sources of photons, except at energies well below 0.025 MeV. At such low energies the phantoms may be too simple in design to give meaningful estimates of dose to the breasts for either internal or external sources—e.g., the distribution of the radiosensitive glandular tissue within the breast could become important here. Thus, this disagreement may be academic.53 The lobes of the thyroid lie between two concentric cylinders and are formed by a cutting surface. It was stated in Cristy36 that, when compared with the adult phantom of Snyder et al.,43 the “thyroid has been moved closer to the front surface of the body, after Hwang, Shoup, and Poston.33 The thyroid had been located too deeply within the neck-and-head region for external dose calculations.54 The new position is better for external sources anterior to the body, but it will remain unsuitable for external sources from the back or sides until a separate neck region is added to the phantom design. This difficulty is unimportant for internal emitters.” The uterus is an ellipsoid cut by a plane. The uterus in the age-15 male/adult female phantom has been changed from that in the age-15 phantom of Cristy,36 to represent an adult female rather than an adolescent. The volume was calculated from the data given in ICRP Publication 23.22 The uterus in the “Adult male” phantom was also modified to be consistent with the change in the “15-AF” phantom. The shapes are slightly different because of differences in trunk shape in the two phantoms. Generally organ shapes were allowed to change according to change in trunk shape, unless there was information to the contrary.37 2.2.5 Works Associated the MIRD Committee Since the publication of the stylized dosimetric model of Snyder et al. in MIRD Pamphlet 5 Revised,24 the MIRD committee has refined several internal organs to support the development of radioimaging tracers and therapeutic nuclear medicine. Modifications to the MIRD stylized model have been published as MIRD pamphlets, which include equations of the new geometries, tabulated absorbed fractions of energy for monoenergetic photons and electrons, and tabulated radionuclide S-values. In 1999, the MIRD committee adopted six new age-specific models of the head and brain55 representing average reference brain and head for a newborn, 1, 5, 10, and 15 years old (also representing the average adult female), and adult male. These phantoms were intended to provide better dosimetric tools in response to the increased number of neuroimaging radiopharmaceuticals.56 Due to the regional uptake of these new agents within the brain, accurate absorbed dose calculations required the use of a detailed model of the subregions of the brain not available with previous models. Similar to previous stylized models, simplistic geometrical shapes (intersection of ellipsoids, elliptical cylinders, tori, and planes) were used to represent the different regions of the head and brain, with volumes derived from published reference masses57 and shapes from analysis of MRI images. Twenty-one subregions were modeled within the head and neck, including five regions representing bony structures (simulated as an homogenized mixture of bone and red marrow with a density of 1.4 g/cm3) and 16 tissue regions (density of 1.04 g/cm3). Within the brain, eight subregions were delineated: the caudate nuclei, cerebellum, cerebral cortex, lateral ventricles, lentiform nuclei (a composite region of both the putamen and the globus pallidus), thalami, the third ventricle, and the white matter. Other regions considered within the head included the cranial cerebrospinal fluid (CSF), cranium, eyes, mandible, spinal cord, spinal cerebrospinal fluid, spinal skeleton, teeth, thyroid, upper face region, and the skin.

Stylized Computational Phantoms Developed at ORNL and Elsewhere

55

In both the ORNL phantoms of Cristy and Eckerman and in the MIRD adult model developed by Snyder and his colleagues, the kidneys were represented as two symmetric ellipsoids cut by a plane with no differentiation of their internal structure. However, because of their unique physiology, the kidneys are seen to concentrate radioactivity nonuniformly.58 Snyder and Ford59 designed a dosimetric model of the human kidney to study the administration of the mecurial diuretic chlormerodrin (neohydrin) labeled with 197Hg and 203Hg. Besides the kidneys, other internal organs of interest included the ovaries and the bladder. The kidneys were assumed to be of equal size (with a total volume of 305.8 cm3) and to be composed of three separate regions, the cortex, the medulla, and the papillary region. Dose estimates were provided for the cortex, medulla, ovaries, and bladder per millicurie-hour of residence of 197Hg and 203Hg in the cortex, in the medulla, in the bladder, and in the total body (blood). MacAfee published in 1970 a multiregion kidney model of an adult60 representing the renal cortex and medulla as two concentric elliptical shells, and the renal pelvis as a wedgeshaped hollow structure at the center of each kidney. In 1975, the MIRD committee used this model in MIRD Dose Estimate Report No. 6 to calculate the dose to the renal cortex and renal medulla from 197Hg- and 203Hg-labeled-Clormerodrin.61 Patel described a multicompartment kidney model in 1988.62 This model, which was similar to the model used by Snyder and Ford,59 consisted of three regions (the cortex, medulla, and papillae) that served as both the source and target regions. The geometry of the kidney was the same as in the original Snyder–Fisher phantom, i.e., the kidney was assumed to be an ellipsoid cut by a plane parallel to the z-axis of the phantom. This model was incorporated into the Snyder–Fisher heterogeneous phantom in a modified version of the ALGAM transport code25 and absorbed fractions and specific absorbed fractions of energy were calculated for 12 monoenergetic photon sources in the range 0.01–4.0 MeV. These results were used to obtain S-values for selected radionuclides for the three regions of the kidney as the sources. The radionuclides considered were 32P, 51Cr, 57Co, 67Ga, 99mTc, 111In, 123I, 131I, 127Xe, 133Xe, and 201Tl. A new kidney model has been adopted by the MIRD committee and was published as MIRD Pamphlet No. 19.63 Following the increased use of radiopharmaceuticals in therapeutic nuclear medicine and recent associated kidney toxicity,64 the MIRD committee developed six advanced stylised models of the kidney (newborn, 1, 5, 10, 15, and adult male). The outer dimensions of these models conformed to those used in the ORNL single-region kidney models while 12 interior structures were defi ned for each kidney: five medullary pyramids, five papillae, the renal pelvis, and the renal cortex. Although the number of medullary pyramids in these models was less than that seen in the real anatomy (6–18 pyramids), it represented a compromise between the mathematical simplicity needed for Monte Carlo transport calculations, and the need for an improved anatomic representation over the concentric ellipsoid-shell model of McAfee.60 Each region was derived from dimensions and volumes given in ICRP Publication 2322 for both the newborn and the adult, and assumed constant volume ratios between the different kidney subregions for the other ages. In these models, each medullary pyramid was modeled by half-ellipsoids (two vertical and three horizontal) with the papillae at its tip, the pelvis by a portion of an ellipsoid within the whole kidney, and the cortex was the remainder of the kidney. In both MIRD Pamphlet Nos. 15 and 19, the EGS4 Monte Carlo transport code26,65 was used for photon and electron transport. In these two pamphlets, absorbed fractions of energy were tabulated for selected source and target combinations (12 energies were simulated between 10 keV and 4 MeV). Following the MIRD method of internal dose calculation,66

56

Handbook of Anatomical Models for Radiation Dosimetry

mean absorbed doses to the target regions per unit cumulated activity in the source region (S-values) were tabulated for specific radionuclides. 2.2.6 Stylized Models of the Lower Abdomen The development of nuclear medicine imaging and therapy over the past decade has resulted in the need for more accurate dosimetry in regions that either were not represented or poorly represented in the MIRD and ORNL stylized phantoms. The lower abdomen is a particular region of the anatomy that is difficult to model properly due to the intricate geometry of its organs. Many assumptions were made in modeling these organs leading to either the organ not being represented, or being oversimplified. Development of new radioagents with specific uptake in the prostate or in the wall of the gastrointestinal tract has led to a need to modify the dosimetric model of the lower abdomen. In 1994, Stabin developed a mathematical model of the prostate gland and included it in the ORNL stylized model.67 This new organ was modeled as a single sphere located just below the urinary bladder, with a volume consistent with Reference Man of ICRP Publication 23.22 This model was used to calculate absorbed fractions of energy and S-values for selected radionuclides. In 1999, Mardirossian et al. recognized that the relative spatial position of the urinary bladder, rectum, and prostate were poorly represented in the ORNL phantom series. They developed a new model of the lower part of the colon,68 and separated the rectum from the lower large intestine. This new model included an explicitly defined rectum, anatomically correct sigmoid and descending colons, and a prostatic urethra and seminal duct. These modifications were implemented in the ORNL phantom series, after changing the position of the bladder and prostate gland to properly model the relative positions of these organs. These models were developed not only for the adult male model, but also for the other phantoms in the ORNL series using physiological and anatomical descriptions published in ICRP Publication 23.22 Because the intestinal tract and curvatures cannot be modeled with simple geometrical shapes that can be easily coupled to a Monte Carlo transport code, all models of this region have relied on thickening the wall region to preserve anatomical wall and content mass. The critical cells for these organs have been identified as the cells within the mucosa layer. In 1994, Poston et al. developed a revision of the GI tract of the MIRD adult phantom to better represent these sensitive targets.28,29 The actual wall of the GI tract was divided in its thickness into four regions of varying radiosensitivities; these layers were very difficult to model because the thickness of each layer varied from one section to another along the different regions of the GI tract. Poston et al. developed two methods to model this wall. The first method divided the tissue wall into 10 small, concentric layers (100 μm thick for the adult) and the dose to each layer was recorded separately. Then, the determination of the mucosa layer thickness for each section would give directly the dose to the mucosa. However, since it is not possible to determine directly the mucosa layer thickness for a specific patient, the subdivision into 10 regions has not been used for medical dose calculation. In a second method, Poston et al. measured the average thickness of the mucosa layer along the GI tract from anatomic slides of cross sections of a human GI tract. Different mucosa thicknesses for the stomach, the small intestine and the large intestine were obtained. This layer was included in the GI wall of the adult mathematical phantom and coupled to the EGS4 Monte Carlo transport code.26 Stubbs et al.69 presented calculations of the radiation-absorbed dose to the walls of hollow organs. These authors studied all four sections of the gastrointestinal tract but only

Stylized Computational Phantoms Developed at ORNL and Elsewhere

57

for four radionuclides important to nuclear medicine. They presented S-values (in units of Gy/Bq s) for 90Y, 99mTc, 123I, and 131I and concluded, as did Poston et al., that the ICRP “onehalf” assumption was overly conservative for nonpenetrating radiation. More recently, an improved (but very simple) geometric model for the small intestine has been developed and used in a study of electron energy deposition in the wall of the tract.70 Results were obtained using this model for monoenergetic electrons in the range 10–500 keV. Depth dose profiles were developed so that dose to the more sensitive cells could be assessed. Electron transport calculations using the computer code MCNP4A27 and the new model showed that only a small fraction of the available energy reaches the critical cells in the crypts of the wall. For electron energies below 330 keV, the only contribution to absorbed dose to the stem cells came from bremsstrahlung. For higher electron energies (i.e., >330 keV), the dose distribution curve was dominated by monoenergetic electrons. 2.2.7 Stylized Phantoms Representing Pregnancy Women In 1995, Stabin et al.71 added the fetus and placenta to the stylized adult female phantom and created phantoms of the pregnant female at each trimester (3-, 6-, and 9-months). The greatest changes to the adult female phantom involved the growth of the uterus and the existence of a compartment representing uterine contents. Stabin et al. represented the uterus in a manner similar to that described by Cloutier et al.72 Figure 2.5 shows a drawing of the cross-sectional view of the uterine region at 9-months in the Stabin et al. pregnant female phantom series.

Body surface

z΄

Placenta Uterine wall Other uterine tissue Fetal skeleton Fetal soft tissue y΄

50° Body surface z

x =0

y

FIGURE 2.5 Images of the 9-month uterine model in the Stabin et al. pregnant female phantom series. (From Stabin, M.G. et al., Mathematical models and specific absorbed fractions of photon energy in the nonpregnant adult female and at the end of each trimester of pregnancy, ORNL/TM-12907, Oak Ridge, TN: Oak Ridge National Laboratory, 1995.)

58

Handbook of Anatomical Models for Radiation Dosimetry

At the end of the first trimester (3 months) the uterus was represented by a right circular cone with a hemispherical cap. The axis of the cone was oriented in the Y-direction, formed at a 33° incline to the horizontal. After the third month of pregnancy, and until about the seventh month the main area of growth in the uterus occurred along its long axis with a small increase in breadth. In the 6-month phantom, the uterus is modeled as a cylinder capped at both end by hemispheres. The long axis of the cylinder ran in the Y-direction, tilted upward at an angle of 40° from the horizontal. In the 9-month phantom, the uterus was modeled as a cylinder capped at both ends by hemispheres, as in the 6-month phantom. The uterus was modeled to have extended in length as well as breadth, and the upper hemisphere was considerably larger than the lower hemisphere, and larger than in the 6-month phantom. The two hemispherical sections were connected by a section of a cone whose long axis of the cylinder ran in the Y-direction, and was tilted upward at an angle of 40° from the horizontal. The uterine contents of the 3-month phantom were modeled as a homogeneous mixture of soft tissue. No attempt was made at that time to model the presence of skeletal material. The uterine contents were thus used as a target to represent the fetus in this phantom. In the 6- and 9-month phantoms, the fetus, the placenta, and the amniotic fluid were explicitly modeled as separate regions. The fetal skeleton was also explicitly modeled in these phantoms and given the composition of the skeleton of the newborn phantom of Cristy and Eckerman.37 The placenta was also modeled as a hemispherical shell, in the 6- and 9-month phantoms. The whole body trunk was extended in the 6- and 9-month phantoms to accommodate the enlarged uterus, and a few of the abdominal organs were redesigned or moved. Urinary bladder and small intestine were significantly remodeled to model their significant displacement and, in the case of bladder, decreased in volume. In 2004, Chen extended the stylized pregnant female phantoms into four pregnancy periods, 8 weeks, 3, 6, and 9-months, for external ionizing radiation dosimetry.73 2.2.8 Other Stylized Models of the Human Anatomy Other modifications and additions to the ORNL and MIRD stylized models include a peritoneal cavity,74 a new model of the nasal cavity and major airway,75 and a new model of the long bones.76 The peritoneal cavity model was developed in 1989 by Watson et al.74 They modified the MIRD phantom to include a region representing the peritoneal cavity to support the dosimetry associated with several therapeutic and diagnostic techniques involving injection of radioactive material into the peritoneal cavity. Similarly, in 1997, Deloar et al. developed a model of the nasal cavity and major airway to support the dosimetry associated with 15O-labeled gases as positron emission tomography (PET) imaging agents.75 In 2000, Clairand et al. modified the model of the long bones of the ORNL phantom series to properly differentiate the cortical bone, trabecular bone, and medullary cavity.76 This effort was done to support bone dosimetry for photon sources and to allow more accurate marrow dose calculations for children. Because in children, the active marrow is not only found in the trabecular bone but also in the medullary cavities, the stylized models of the long bones of the legs and arms (truncated circular cones) of the ORNL phantom series did not allow for accurate bone dosimetry calculations. 2.2.9 GSF Gender-Specific Phantoms, ADAM and EVA A revised version of the MIRD-5 phantom24 was adopted by the GSF, Germany in 1975 and afterward used by Kramer and his colleagues in a series of studies related to external photon dosimetry. 77,78

Stylized Computational Phantoms Developed at ORNL and Elsewhere

59

The researchers noted that the MIRD-5 phantom represented a partially hermaphrodite adult only—the phantom has the dimensions of the male reference man including testes, ovaries, and uterus but no female breasts. Since the GSF researchers were interested in applying the phantoms to external dosimetry where the female breasts were important, it was believed that the MIRD-5 phantom could lead to inaccurate result for the female workers using the newly released tissue weighting factor.79 The GSF researchers decided to revise the MIRD-5 phantom by developing two separate sex-specific phantoms. In what became known as the ADAM and EVA phantoms, a number of changes in anatomical details were made.40 The EVA phantom was derived by shrinking all relevant volumes of the MIRD-5 phantom with the total whole body mass ratio of 0.83 that was revealed from the analysis of ICRP reference organ masses. Then, the female organ masses were modified to create space for neighboring organs. Finally, sex-specific organ such as testes, ovaries, uterus, and breasts were introduced into the appropriate phantom to yield ADAM and EVA, respectively. The chin was introduced by removing a section of the neck to created a more realistic external irradiation geometry for the thyroid. The female breasts were represented by two ellipsoid sections attached to the trunk of EVA. At the same time, at the birthplace of the MIRD-5 phantom, Cristy was developing agedependent hermaphrodite phantoms for internal dosimetry applications that were less sensitive to the shape and size of the female breasts.36 A debate about the differences of breast sizes of the reference female went on between the groups at GSF and ORNL.49–51 These two sets of stylized phantoms were later used by a large number of users worldwide. 2.2.10 The CAM Phantom Developed by NASA for Space Radiation Dosimetry The computational anatomical male (CAM) phantom was documented by Billings and Yucker in 1973 in a technical report to the National Aeronautics and Space Administration (NASA).80 The computational anatomical female (CAF) phantom was also developed.81 The approach used to develop this pair of phantoms appeared to be very different and aggressive and the CAM reportedly consisted of 1100 unique geometric surfaces and 2450 solid regions. Internal body geometry such as organs, voids, bones, and bone marrow were explicitly modeled using constructive solid geometry (CSG) techniques, according to the authors. A computer program called CAMERA was also developed for performing analyses with the CAM phantom. The authors state that “extremely detailed geometrical model of the human anatomy, the most detailed yet prepared, has been developed for use in investigations dealing with exposure of astronauts to the natural space radiation environment. The model is equally applicable to investigations dealing with exposure of humans to radiation associated with nuclear weapon and nuclear power system environments as well as medical applications such as radiotherapy and radiography.” Indeed the surface geometry was so detailed that one may wonder how this was possible in the 1970s with much less capable computers. Unfortunately, CAM and CAF phantoms were adopted exclusively for studies involving space radiation environments of interest to the aerospace industry.81 Very little information about the original phantom development was available to the research community outside the centers associated with the NASA until recently. Jordan, a long-time NASA-contracted phantom user, recently released some of the images (http://cmpwg.ans.org/phantoms/camera.pdf). It is interesting to note one unique exterior anatomical feature of the CAM phantom: the arms are separated from the trunk, unlike the MIRD-5 phantom and its successors. Two images of the CAM phantom are shown in Figure 2.6.

Handbook of Anatomical Models for Radiation Dosimetry

60

(a)

(b)

FIGURE 2.6 (See color insert following page 524.) Surface plots of the CAM phantom. (a) The whole body view showing arms separated from the trunk (http://cmpwg.ans.org/phantoms/camera.pdf). (b) The close-up view of the facial details.

2.3 Summary Mathematical formulations of the organs and tissues of the body used in the dosimetry of internally distributed radionuclides existed as early as the 1940s, although the first anthropomorphic phantom was not reported until in the 1960s. In the 1970s and 1980s, the sophistication of these stylized phantoms increased significantly. This evolution began with the specification of a single organ mass, followed by the use of simple shapes to simulate organs or the entire body of an adult human. More sophisticated models were later developed which used simple shapes to model the geometry of select organs or the entire human body of an adult human. The desire to model the entire body of a “Reference Man” and to specify the location, shape, volume, and mass of organs in the body as realistically as possible has remained the same to this day. The climax for stylized phantoms was reached in the 1980s when the gender- and age-specific family phantoms were systematically documented and widely adopted for various studies in internal and external radiation dosimetry, as well as in medical imaging and radiotherapy. By that time, Monte Carlo codes and personal computers were accessible to a large number of researchers.

Stylized Computational Phantoms Developed at ORNL and Elsewhere

61

The research on stylized human models which was carried out at ORNL through the 1980s played an essential role in the history of computational phantoms. The sex-specific adult phantoms at GSF, Germany in the early 1980s were revisions of the MIRD-5 phantom that was originally developed at ORNL. Major extensions in the 1990s, for examples, on the pregnant women and brain/head models, were also closely tied to the earlier work at ORNL. The direct involvement of ORNL’s scientists in the Society of Nuclear Medicine’s (SNM’s) MIRD committee facilitated the necessary standardization process. It is clear that close collaborations between leading developers were a key factor contributing to the success of these first-generation computational phantoms. Not all phantoms of this generation enjoyed the same recognition in the history. In fact, a few phantoms such as the CAM have been practically unknown by the mainstream radiation protection dosimetry community. As described in Chapter 1 of this book, the late 1980s would go into the history also as the beginning of an exciting new era of voxel phantoms. Collaboration in the information age required new approaches and no single developer would dominate the new research agenda as ORNL once did.

References 1. Loevinger, R. Distributed radionuclide sources. In Radiation Dosimetry Volume III, 2nd edn., Attix, F.H. and Tochilin, E., eds. (New York: Academic Press), p. 51, 1969. 2. Quimby, E.H. The development of radiation dosimetry in nuclear medicine. In Medical Radionuclides: Radiation Dose and Effects, Cloutier, R.J., Edwards, C.L., and Snyder, W.S., eds., AEC Symposium Series 20 (CONF-691212) ), (Washington, DC: U.S. Atomic Energy Commission) p. 7, 1970. 3. NCRP. The experimental basis for absorbed-dose calculations in medical uses of radionuclides. NCRP Report No 83 (Bethesda, MD: National Council on Radiation Protections and Measurements), 1985. 4. Marinelli, L.D. Dosage determination with radioactive isotopes, Am. J. Roentgenol. Rad. Ther., 47, 210, 1942. 5. Marinelli, L.D., Quimby, E.H., and Hine, G.J. Dosage determination with radioactive isotopes. II. Practical considerations in therapy and protection, Am. J. Roentgenol. Rad. Ther., 59, 260, 1948. 6. Loevinger, R., Japha, E.M., and Brownell, G.L. Discrete radioisotope sources. In Radiation Dosimetry, Hine, G.J. and Brownell, G.L., eds. (New York: Academic Press), p. 693, 1965. 7. Loevinger, R., Holt, J.G., and Hine, G.J. Internally administered radionuclides. In Radiation Dosimetry, Hine, G.J. and Brownell, G.L., eds. (New York: Academic Press), p. 801, 1956. 8. ICRP. Report of Committee II on Permissible Dose for Internal Radiation International Commission on Radiological Protection (Oxford: Pergamon Press), 1959. 9. Berger, M.J. Mird Pamphlet No. 2: Energy Deposition in Water by Photons from Point Isotropic Sources (New York: Society of Nuclear Medicine), 1968. 10. Berger, M.J. Mird Pamphlet No. 2: Energy Deposition in Water by Photons from Point Isotropic Sources (New York: Society of Nuclear Medicine), 1971. 11. Ellett, W.H., Callahan, A.B., and Brownell, G.L. Gamma-ray dosimetry of internal emitters, I. Monte Carlo calculations of absorbed doses from point sources, Brit. J. Radiol., 37, 45, 1964. 12. Ellett, W.H., Callahan, A.B., and Brownell, G.L. Gamma-ray dosimetry of internal emitters, II. Monte Carlo calculations of absorbed doses from uniform sources, Brit. J. Radiol., 38, 541, 1965. 13. Reddy, A.R., Ellett, W.H., and Brownell, G.L. Gamma-ray dosimetry of internal emitters, I. Monte Carlo calculations of absorbed doses for low-energy gamma-rays, Brit. J. Radiol., 42, 512, 1967.

62

Handbook of Anatomical Models for Radiation Dosimetry

14. Loevinger, R. and Berman, M. MIRD Pamphlet No. 1 A Schema for Absorbed-Dose Calculations for Biologically Distributed Radionuclides (New York: Society of Nuclear Medicine), 1968. 15. Brownell, G.L., Ellett, W.H., and Reddy, A.R. MIRD Pamphlet No. 3 Absorbed Fractions for Photon Dosimetry (New York: Society of Nuclear Medicine), 1968. 16. Ellett, W.H. and Humes, R.M. MIRD Pamphlet No. 8 Absorbed Fractions for Small Volumes Containing Photon-Emitting Radioactivity (New York: Society of Nuclear Medicine), 1971. 17. Auxier, J.A., Snyder, W.S., and Jones, T.D. Neutron interactions and penetration in tissue. In Radiation Dosimetry Volume I, 2nd edn., Attix, F.H. and Tochilin., eds. (New York: Academic Press), p. 275, 1968. 18. Fisher, H.L.J. and Snyder, W.S. Variation of dose delivered by 137Cs as a function of body size from infancy to adulthood, ORNL-4007 (Oak Ridge, TN: Oak Ridge National Laboratory), P. 221, 1966. 19. Fisher, H.L.J. and Snyder, W.S. Distribution of dose delivered in the body size from a source of gamma rays distributed uniformly in an organ, ORNL-4168 (Oak Ridge, TN: Oak Ridge National Laboratory), p. 245, 1967. 20. Snyder, W.S. The variation of dose in man from exposure to a point source of gamma rays, ORNL-4168 (Oak Ridge, TN: Oak Ridge National Laboratory), p. 257, 1967. 21. Snyder, W.S. et al. MIRD Pamphlet No. 5 Estimates of Absorbed Fractions for Monoenergetic Photon Sources Uniformly Distributed in Various Organs of a Heterogeneous Phantom (New York: Society of Nuclear Medicine), 1969. 22. ICRP. Report on the Task Group on Reference Man, ICRP Publication 23 (Oxford: International Commission on Radiological Protection), 1975. 23. ICRP. Limits for Intakes of Radionuclides by Workers, ICRP Publication 30 (Oxford: International Commission on Radiological Protection), 1979. 24. Snyder, W.S., Ford, M.R., and Warner, G.G. Mird Pamphlet No. 5, Revised Estimates of Specific Absorbed Fractions for Monoenergetic Photon Sources Uniformly Distributed in Various Organs of a Heterogeneous Phantom (New York: Society of Nuclear Medicine), 1978. 25. Warner, G.G. and Craig, A.N.J. ALGAM: A computer program for estimating internal dose in a man phantom, ORNL-TM-2250 (Oak Ridge, TN: Oak Ridge National Laboratory), 1968. 26. Nelson, W.R., Hirayama, R.H., and Rogers, D.W.O. The EGS4 code system SLAC Report 265 (Stanford, CA: Stanford Linear Accelerator Center), 1985. 27. Briesmeister, J.F. MCNP—A general Monte Carlo N-particle transport code, version 4a LA-12625 (Los Alamos, NM: Los Alamos National Laboratory), 1993. 28. Poston, J.W. Jr. et al. Calculation of absorbed energy in the gastrointestinal tract, Health Phys., 71, 300, 1996. 29. Poston, J.W. Jr. et al. A revised model for the calculation of absorbed energy in the gastrointestinal tract, Health Phys., 71, 307, 1996. 30. Kereiakes, J.G. et al. Doses to infants and children: A plea for a standard child, Health Phys., 11, 999, 1965. 31. Poston, J.W. The development of early pediatric models and their application to radiation absorbed dose calculations. In Dosimetry of Administered Radionuclides, Adelstein, S.J., Kassis, A.I., and Burt, R.W., eds. (Washington, DC: American College of Nuclear Physicians), p. 105, 1990. 32. Hwang, J.M.L. et al. Mathematical description of a one- and five-year-old child for use in dosimetry calculations, ORNL/TM-5293 (Oak Ridge, TN: Oak Ridge National Laboratory), 1976. 33. Hwang, J.M.L., Shoup, R.L., and Poston, J.W. Mathematical description of a newborn human for use in dosimetry calculations, ORNL/TM-5453 (Oak Ridge, TN: Oak Ridge National Laboratory), 1976. 34. Jones, R.M. et al. The development and use of a fifteen-year-old equivalent mathematical phantom for internal dose calculations, ORNL/TM-5278 (Oak Ridge, TN: Oak Ridge National Laboratory), 1976. 35. Deus, S.F. and Poston, J.W. The development of a mathematical phantom representing a 10-year-old for use in internal dose calculations. In Proceedings of the Symposium on Radiopharmaceutical Dosimetry, HEW Publication (FDA) 76-8044 (Oak Ridge, TN: Oak Ridge National Laboratory), 1976.

Stylized Computational Phantoms Developed at ORNL and Elsewhere

63

36. Cristy, M. Mathematical phantoms representing children of various ages for use in estimates of internal dose, ORNL/NUREG/TM-367 (Oak Ridge, TN: Oak Ridge National Laboratory), 1980. 37. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381 (Oak Ridge, TN: Oak Ridge National Laboratory), 1987. 38. Coffey, J.L. and Watson, E.E. S-values for selected radionuclides and organs with the heart wall and heart contents as source organs. In Third International Radiopharmaceutical Dosimetry Symposium (Rockville, MD: U.S. Department of Health and Human Services), 1981. 39. Coffey, J.L. A revised mathematical model of the heart for use in radiation absorbed dose calculation, MS Thesis University of Tennessee, (Knoxville, TN), 1978. 40. Kramer, R. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods: Part I. The male (ADAM) and female (EVA) adult mathematical phantoms, GSF-Report S-885 (Neuherberg-Muenchen: Institut fuer Strahlenschutz, GSF-Forschungszentrum fuer Umwelt und Gesundheit), 1982. 41. Seltzer, S.M. An overview of ETRAN Monte Carlo methods. In Monte Carlo Transport of Electrons and Photons (New York: Plenum Press), 1988. 42. Seltzer, S.M. Electron-photon Monte Carlo calculations: The ETRAN code, Appl. Radiat. Isot., 42, 917, 1991. 43. Snyder, W.S. et al. A tabulation of dose equivalent per microcurie-day for source and target organs of an adult for various radionuclides: Part 1. ORNL-5000 (Oak Ridge, TN: Oak Ridge National Laboratory), 1974. 44. Eckerman, K.F. and Ryman, J.C. External exposure to radionuclides in air, water, and soil. Federal Guidance Report 12, PB94-114451, (Washington, DC) 1993. 45. Hudson, G. Bone-marrow volume in the human foetus and newborn, Brit. J. Haemat., 11, 446, 1965. 46. Custer, R.P. An Atlas of the Blood and Bone Marrow, 2nd edn. (Philadelphia: W. B. Saunders), 1974. 47. Woodard, H.Q. and Holodny, E. A summary of the data of Mechanik on the distribution of human bone marrow, Phys. Med. Biol., 5, 57, 1960. 48. Cristy, M. Active bone marrow distribution as a function of age in humans, Phys. Med. Biol., 26, 389, 1981. 49. Kramer, R. and Drexler, G. Representative breast size of reference female, Health Phys., 40, 913, 1981. 50. Cristy, M. Representative breast size of reference female, Health Phys., 43, 930, 1982. 51. Kramer, R., Williams, G., and Drexler, G. Reply to M. Cristy, Health Phys., 43, 932, 1982. 52. Katch, V.L. et al. Contribution of breast volume and weight to body fat distribution in females, Am. J. Phys. Anthropol., 53, 93–100, 1980. 53. Cristy, M. Calculation of annual limits of intake of radionuclides by workers: Significance of breast as an explicitly represented tissue, Health Phys., 46, 283, 1984. 54. Kerr, G.D. Organ dose estimates for the Japanese atomic-bomb survivors, Health Phys., 37, 487, 1979. 55. Bouchet, L.G. et al. MIRD Pamphlet No. 15: Radionuclide S values in a revised dosimetric model of the adult head and brain, J. Nucl. Med., 40, 3, 62S, 1999. 56. London, E.D. Imaging Drug Action in the Brain (Boca Raton, FL: CRC), 1993. 57. Blinkov, S.M. and Glezer, I.I. The Human Brain in Figures and Tables (New York: Plenum), 1968. 58. Petegnief, Y. et al. Quantitative autoradiography using a radioimager based on a multiwire proportional chamber, Phys. Med. Biol., 43, 3629, 1998. 59. Snyder, W.S. and Ford, M.R. A dosimetric study for the administration of neohydrin labeled with 203Hg and 197Hg, ORNL-4168 (Oak Ridge, TN: Oak Ridge National Laboratory), p. 267, 1967. 60. McAfee, J.G. Problems in evaluating the radiation dose for radionuclides excreted by the kidneys. In Medical Radionuclides: Radiation Dose and Effects, Cloutier, R.J., Edwards, C.L., and Snyder, W.S., eds. (Oak Ridge, TN: U.S. Atomic Energy Commission), p. 271, 1969. 61. Blau, M. et al. MIRD dose estimate No. 6: Hg-197 and Hg-203-labeled chlormerodrin, J. Nucl. Med., 16, 1214, 1975.

64

Handbook of Anatomical Models for Radiation Dosimetry

62. Patel, J.S. A Revised Model of Kidney for Medical Internal Radiation Dose (College Station, TX: Texas A&M University), 1988. 63. Bouchet, L.G. et al. MIRD Pamphlet No. 19: Absorbed fractions and radionuclide s values for six age-dependent multiregion models of the kidney, J. Nucl. Med., 44(7), 1113, 2002. 64. Boerman, O.C., Oyen, W.J., and Corstens, F.H. Between the Scylla and Charybdis of peptide radionuclide therapy: Hitting the tumor and saving the kidney, Eur. J. Nucl. Med., 28, 1447, 2001. 65. Ford, R.L. and Nelson, W.R. The EGS code system: Computer Programs for Monte Carlo Simulation of Electromagnetic Cascade Shower Report 210 (Stanford: Stanford Linear Accelerator Center), 1978. 66. Loevinger, R., Budinger, T.F., and Watson, E.E. MIRD Primer for Absorbed Dose Calculations (New York: The Society of Nuclear Medicine), p. 128, 1991. 67. Stabin, M. A model of the prostate gland for use in internal dosimetry, J. Nucl. Med., 35(3), 516, 1994. 68. Mardirossian, G. et al. A new rectal model for dosimetry applications, J. Nucl. Med., 40(9), 1524, 1999. 69. Stubbs, J.B., Evans, J.F., and Stabin, M.G. Radiation absorbed doses to the walls of hollow organs, J. Nucl. Med., 39, 11, 1989–1995. 70. Bhuiyan, N.U. A Revised Dosimetric Model for Calculation of Electron Dose in the Small Intestine (College Station, TX: Texas A&M University), 2000. 71. Stabin, M.G. et al. Mathematical models and specific absorbed fractions of photon energy in the nonpregnant adult female and at the end of each trimester of pregnancy, ORNL/TM-12907 (Oak Ridge, TN: Oak Ridge National Laboratory), 1995. 72. Cloutier, R.J. et al. Dose to fetus from radionuclides in bladder, Health Phys., 25, 147, 1973. 73. Chen, J. Mathematical models of the embryo and fetus for use in radiological protection, Health Phys., 86, 285, 2004. 74. Watson, E.E. et al. A model of the peritoneal cavity for use in internal dosimetry, J. Nucl. Med., 30, 2002, 1989. 75. Deloar, H.M. et al. Internal dose estimation including the nasal cavity and major airway for continuous inhalation of C15O2, 15O2, and C15O using the thermoluminescent dosimeter method, J. Nucl. Med., 38, 1603, 1997. 76. Clairand, I., Bouchet, L.G., and Bolch, W.E. A new macroscopic model of the long bones for skeletal dosimetry, J. Nucl. Med., 41, 1062, 2000. 77. Kramer, R. and Drexler, G. The dose equivalent index (DEI) as a function of angular distribution of photons, Health Phys., 38, 426, 1980. 78. Kramer, R. and Drexler G. Dose equivalent conversion factors for external photon irradiation, in Radiation Protection: A Systematic Approach to Safety. Proceedings of the 5th Congress of the International Radiation Protection Society, Jerusalem (Oxford: Pergamon Press), p. 311, 1980. 79. Kramer, R. and Drexler, G. On the calculation of the effective dose equivalent, Rad. Prot. Dosim., 3, 13, 1982. 80. Billings, M.P. and Yucker, W.R. The Computerized Anatomical Man CAM Model, NASA CR-134043 (Washington, DC: Government Printing Office), 1973. 81. Atwell, W. Anatomical models for space radiation applications: An overview, Adv. Space Res., Oct; 14(10), 415, 1994.

3 The GSF* Voxel Computational Phantom Family Maria Zankl

CONTENTS 3.1 Introduction .................................................................................................................65 3.2 Construction of the GSF Voxel Phantoms ............................................................... 66 3.2.1 Region Growing ............................................................................................. 67 3.2.2 Threshold and Morphological Operations................................................. 67 3.2.3 Interactive Drawing of the Borderline ........................................................ 68 3.2.4 Manual Drawing of Organs ......................................................................... 68 3.2.5 Assignment of Organ Identification Numbers .......................................... 68 3.2.6 Estimation of Bone Marrow Distribution .................................................. 68 3.3 Description of the GSF Voxel Phantoms .................................................................. 69 3.4 Applications of the GSF Voxel Phantoms in Radiation Dosimetry ..................... 76 3.4.1 Monte Carlo Codes ........................................................................................ 76 3.4.2 Bone Dosimetry.............................................................................................. 76 3.4.3 Idealized Geometries (External) .................................................................. 76 3.4.4 Environmental Dosimetry............................................................................ 78 3.4.5 Dosimetry for Medical Imaging .................................................................. 78 3.4.6 Specific Absorbed Fraction Calculations for Organ Self-Absorption (Original Source Masses) .................................................80 3.4.7 SAF Calculations for Organ Cross-Fire ......................................................80 3.4.8 Applications Outside the Helmholtz Zentrum München .......................80 3.5 Conclusions .................................................................................................................. 82 References ............................................................................................................................... 82

3.1 Introduction To protect against ionizing radiation (occupational, environmental, and medical), researchers must determine the radiation dose to specific body organs and tissues for the scope of risk assessment. For this purpose, a series of computational phantoms (hereafter “phantoms”) of the human body were designed in the past that have been used together with

* On January 1, 2008, the “GSF—National Research Center for Environment and Health” has changed its name to “Helmholtz Zentrum München—German Research Center for Environmental Health.” However, since the voxel models have been known as “GSF voxel phantoms” for many years, this name is at present retained for our phantom family.

65

66

Handbook of Anatomical Models for Radiation Dosimetry

computer codes, to simulate the radiation transport and energy deposition in the body. The first generation of computational body phantoms were the so-called mathematical computational phantoms: mathematical expressions representing simple geometrical bodies that were used to describe idealized arrangements of body organs.1–5 This type of computational phantom has also been called the “MIRD-type,” due to the fact that its most famous representatives have been published in documents of the Medical Internal Radiation Dose (MIRD) Committee.1,2 During the last two decades, various groups have developed a new generation of body computational phantoms, “tomographic” or “voxel computational phantoms” (from voxel = volume element), as an extension and improvement to these earlier computational phantoms. These computational phantoms are segmented from computed or magnetic resonance tomographic data of real persons and, thus, offer a more realistic replication of human anatomy.6–13 This is only a small selection, and is not exhaustive, since the most important of these phantoms are described explicitly in other chapters of this book. The first voxel phantoms were developed at Vanderbilt University and at the GSF— National Research Center for Environment and Health, independently and more or less simultaneously. The phantom from Vanderbilt University was an adult female,14,15 and those from GSF were pediatric phantoms—an 8-week-old baby and a 7-year-old child.6,16–18 At the GSF, the pediatric phantoms, Baby and Child, were followed first by Golem, the voxel phantom of an adult male,11 and then by nine individual voxel phantoms: three male (Frank, Godwin, and the Visible Human), and six female (Donna, Helga, Irene, Katja, Laura, and Klara), where one of them—Katja—was pregnant.12,19–22

3.2 Construction of the GSF Voxel Phantoms All voxel phantoms constructed so far by our working group were based on computed tomographic (CT) image data of living patients, with the exception of the 8-week-old baby that was constructed from the image data of a dead body. All patients were scanned with a large number of contiguous axial slices. Each slice consists of a matrix of typically 256 × 256 or 512 × 512 pixels (picture elements, in a planar image). The volume elements are the pixels multiplied by the thickness of the slice. The single slice images are stacked, resulting in a three-dimensional (3D) array of voxels. In the primary image data, each pixel has a value that is characteristic of a certain physical property of the respective volume element; in the case of CT images this property is attenuation of x-rays of a specific radiation quality. The pixel values are the “Hounsfield numbers” or “CT numbers” that relate the attenuation property of a pixel to that of water (having the value 0) and range typically from −1000 (for air) to approximately between 500 and 1000 (for bone). Many image-processing software tools convert these Hounsfield numbers to nonnegative integer “gray values” by adding a constant value to the Hounsfield number of each pixel. Today, typically 12 bits are used to store these values, thus permitting a data range between 0 and 4095. Formerly, when storage needed to be used sparingly, the gray values were often limited to 255, thus reducing the storage requirement to 1 byte per pixel. Clearly, this renormalization was accompanied by a loss of image contrast. The image property coded by the gray values—namely the attenuation of the x-ray beam of the CT equipment—is related to the tissue electron density, not to the anatomical site. That means the anatomical boundaries between organs and tissues have to be determined

The GSF Voxel Computational Phantom Family

67

on the basis of the gray values and gray-value contrasts, and then have to be combined suitably with anatomical knowledge by the user. Each pixel is then assigned to an organ or tissue to which the corresponding voxel belongs. This process of assigning each voxel to a specific organ is called segmentation. Although image-processing methods are steadily improving, and researchers have been aiming for automatic segmentation for a long time, thorough segmentation of a large body section into many organs still involves a large amount of manual work. The reason is that there are only a small number of tissues that separate well from their neighborhood by gray-value contrast. For CT, these are mainly the bones and the lungs. The separation of muscle and adipose tissue can also be made by gray-value thresholds. The gray values of many individual organs, however, are inhomogeneous, and the gray-value ranges of various organs largely overlap. Therefore, organs that are in close contact to each other cannot be separated automatically on the basis of their gray values. Consequently, segmentation is still rather tedious and time-consuming. In principle, the organs and tissues to be segmented could be divided into three types: (1) organs and tissues that can be separated from the surrounding tissue or the neighboring organs by their gray value, i.e., they can be segmented using a threshold and/or region growing procedure; (2) organs and tissues that can be distinguished visually, i.e., an interactive method is applied and the user “draws” the borderline; and (3) organs and tissues that cannot be seen on the CT images, in which case the user has to draw the organ, consulting, if necessary, an anatomical atlas. Since the segmentation of the GSF voxel computational phantom family is a process extending over many years and is still ongoing, a variety of software tools and computer hardware have been involved in its construction. Therefore, the segmentation technique is described in general terms here. The following segmentation procedures were applied. 3.2.1 Region Growing For the segmentation of organs with good contrast and a small number of elements in each slice, for example the external body contour or the lungs, a “region growing” procedure was used. For this procedure, a seed point and a range of gray values are selected. A borderline is generated that encompasses all pixels around the seed point that have gray values in the selected range. At the next slice, the seed point and the range of gray values are taken again and the borderline is adjusted. Subsequently the borderline can be filled by a morphological method, i.e., all pixels inside the borderline are set to 1. The result is saved as a binary 3D data set in which those pixels inside the borderline have the value 1, and all the others have the value 0. 3.2.2 Threshold and Morphological Operations For organs with good contrast and several elements in each slice, a threshold method was applied to avoid using too many seed points. The threshold method sets all voxels between a minimum and maximum gray value to 1, and all other voxels to 0. This results in a 3D binary file. Subsequently, this was followed in most cases by a manual image editing, i.e., an erasing of the “wrong” pixels slice-by-slice, or a mathematical morphological operation. Some functions of morphological methods used in this work are the following: the “erode” operation peels off a layer from an organ using a specified structuring element; in our case mostly the smallest available element was used, which leads to a reduction of one pixel row within a single slice. The “dilate” operation expands the organ by adding a specified structuring element layer. The “close” operation is a dilation followed by an erosion,

68

Handbook of Anatomical Models for Radiation Dosimetry

both of which use the same structuring element. A closing will generally fill in thin gulfs and small holes in the binary volume. The “open” operation is an erosion followed by a dilation again using the same structuring element. This operation will generally suppress sharp capes and eliminate small objects. A “connect” operation identifies, counts and—if that option is selected—numbers the separate objects in an image. If a value greater than 1 is selected for the minimum allowable component size, all components smaller than this size are set to zero. This is another method to eliminate small objects. The final result, after all these possible manipulations, is again a 3D binary file in which all pixels belonging to the segmented tissue have the value 1 and all the others have the value 0. 3.2.3 Interactive Drawing of the Borderline As mentioned earlier, most of the organs to be segmented cannot be separated easily from the surrounding tissue by a threshold. Nevertheless, the user might be able to see or guess where each organ is located—if necessary, by consulting an anatomical atlas or a medical expert. In these cases, B-spline functions are useful, which allow drawing a borderline using a Bezier spline curve that is defined by a number of base points chosen by the user. The first advantage is that—in contrast to freehand drawn lines—the borderline curve is much smoother. The second advantage is that the line can usually be transferred into the next slice and adapted by slightly moving the base points to the actual contours of the organ. Typical organs segmented with this method are the kidneys, the liver, and the spleen. The resulting binary contour of the segmented organ can then be filled as already described. 3.2.4 Manual Drawing of Organs In some cases it is easier to “draw” the organ directly into the already segmented data set, because its location is known exactly, and its shape in the two-dimensional (2D) slice images can be represented by a simple geometric shape, such as a circle or an ellipse. Typical examples for these tissues are the eye lenses and the spinal cord. Another reason to draw organs freehand is that that their location is fixed only in general terms and the exact position is variable. Examples for this situation are the lymphatic nodes, a tissue type that was introduced in the voxel phantoms of the ICRP adult Reference Male and Reference Female (see Chapter 14). 3.2.5 Assignment of Organ Identification Numbers After an organ or tissue was segmented using one of the methods described above, it was stored as a binary file on hard disk. Subsequently, the binary files were assigned their specific organ identification number and compiled into one volume. Each organ or tissue is thus represented by those volume elements (voxels) identified as belonging to it from the CT slice images, having been assigned a common organ identification number instead of their original gray values. 3.2.6 Estimation of Bone Marrow Distribution The original gray values of the skeleton were stored in addition to the organ identification number, since they were used to estimate the bone marrow distribution. The proportion of bone marrow in each voxel was approximated using the original gray value of this voxel

The GSF Voxel Computational Phantom Family

69

and by linearly interpolating between a gray value corresponding to pure bone marrow and a gray value corresponding to pure hard bone. Here, different voxel densities are assumed to represent different mixtures of hard bone and bone marrow. This method, described in detail elsewhere,11 does not allow modeling of the complicated trabecular bone structure or the bone endosteum (also called “bone surface” which is the radiosensitive tissue adjacent to the mineral bone structures). The percentage of hard bone and bone marrow in each voxel in the skeleton, however, can be estimated from the CT pictures and, thus, the distribution of bone marrow in the whole skeleton can be reconstructed with a resolution corresponding to that of the CT scan. The method as it has so far been applied to the GSF voxel phantom family has three shortcomings. First, it has been assumed in the past that red and yellow bone marrow are equally distributed in all bones, except the long bones, where in the adult phantoms all bone marrow below mid of humeri and femurs is assumed inactive (i.e., yellow) following the literature data.23,24 In all other bones, constant (age-dependent) ratios of red and yellow bone marrow were assumed, disregarding the more detailed information on marrow cellularity in individual bones also available from the literature.23,24 This neglect was shown to have only a small dosimetric effect, 25 and it can be easily corrected in future studies. The second issue is due to the so-called partial volume effect. Voxels that are located at the boundaries of bones and surrounding soft tissue contain parts of both tissues and have, thus, attenuation properties that correspond to a mixture of these tissues. Applying the interpolation procedure assumes that these voxels contain bone marrow, whereas in reality the bones are enclosed by an outer shell of cortical bone. This leads to an overestimation of the red bone marrow (RBM) dose for radiation of low penetrability. The third shortcoming concerns the composition of the skeleton: so far, the skeleton of the GSF voxel phantoms was considered to consist of only mineral bone and bone marrow; the not-directly segmented proportion of cartilage and the so-called miscellaneous component have not been considered for the elemental composition of the skeleton. This results in an average density of the skeleton that is slightly higher than the “reference” density that arises when all skeletal constituents are accounted for. Golem’s average bone density, e.g., is 1.45 g∙cm−3, compared to 1.35 g∙cm−3 for the reference skeleton.26

3.3 Description of the GSF Voxel Phantoms Table 3.1 shows the GSF voxel phantoms available at the moment, given also the age, height, and weight of the individual from whose data the phantoms were constructed. The first voxel phantom constructed at the former GSF was the phantom “Baby,” segmented from the CT images of the dead body of an 8-week-old baby. The phantom “Child” was segmented next, modeled from the image data of a 7-year-old girl who underwent a whole body irradiation for leukaemia treatment, and who was also relatively small for her age. Figure 3.1 shows the two pediatric phantoms. The figure shows some of the main organs or tissues in a 3D representation and demonstrates the anatomical realism of this type of phantom. The fi rst adult voxel computational phantom constructed by our working group— “Golem”—was segmented from the image data of a patient who was 176 cm in height and had a weight of 68.9 kg. These data are in good agreement with the older ICRP data on the adult male Reference Man, 27 whose height and weight were 170 cm and 70 kg.

f 7 115 21.7 1.1

8

1.54

19.0 256 256 144 64 Whole body

f 8w 57 4.2 1.4

4

0.85

2.9 256 256 142 54 Whole body

Gender Age Height (cm) Weight (kg) Number of (nonzero) voxels (million) Slice thickness (voxel height, mm) Voxel in-plane resolution (mm) Voxel volume (mm3) Number of columns Number of rows Number of slices Number of organs Coverage

Child

Baby

Property

35.2 256 256 179 62 Whole body

1.875

10

f 40 176 79 2.2

Donna

2.7 512 512 193 62 Head and trunk

0.74

5

m 48 174 95 23.7

Frank

34.6 256 256 220 121 Whole body

2.08

8

m 38 176 69 1.9

Golem

Main Characteristics of the Members of the GSF Voxel Phantom Family

TABLE 3.1

9.6 512 512 114 62 Head to thigh, no arms

0.98

10

f 26 170 81 8.3

Helga

17.6 262 132 348 62 Whole body

1.875

5

f 32 163 51 3.0

Irene

17.6 240 132 346 88 Whole body

1.875

5

f 43 168 59 3.5

Laura

4.84

f 43 163 62.3 4.0

Katja

4.3 512 512 250 133 Head to thigh

15.2 299 146 346 136+19 Whole body plus fetus

0.91/0.94 1.775

5

m 38 180 103 20.1

Visible Human

34.8 256 256 220 88 Whole body

2.085

8.0

m 38 176 73.0 2.0

Godwin

15.1 256 256 346 88 Whole body

1.765

4.84

f 43 163 60.0 3.9

Klara

70 Handbook of Anatomical Models for Radiation Dosimetry

The GSF Voxel Computational Phantom Family

FIGURE 3.1 GSF pediatric voxel phantoms; baby (left) and child (right).

71

Then Donna, Frank, Helga, and Irene followed. Donna and Irene20 are whole-body phantoms of a tall person and a person of slight build, respectively. Both Donna and Irene have intestines that were segmented from a high-resolution CT data set of the intestinal region of another female patient and were then fitted into the pelvic regions of these computational phantoms.20 Helga is the head-to-thigh computational phantom of a large and heavy adult female, and Frank12 is the headand-trunk computational phantom of a heavily built male patient whose arms and part of the body contour are not contained in the slice images, since they were outside the field of view. The “Visible Human” was constructed at GSF from CT data from the Visible Human Project (VHP) of the American National Library of Medicine. Another voxel phantom of the same individual exists, called VIP-Man, segmented from the color photographic images.10,28 The differences between these two VHP image data sets are that the CT images have larger pixel sizes and the arms are missing in comparison with the color photographic images. Figure 3.2 shows the male voxel phantoms

FIGURE 3.2 (See color insert following page 524.) GSF male voxel phantoms; Golem (left), Frank (middle), and Visible Human (right).

72

Handbook of Anatomical Models for Radiation Dosimetry

FIGURE 3.3 (See color insert following page 524.) GSF female voxel phantoms: Donna, Helga, Irene, and Laura (from left to right).

Golem, Frank, and the Visible Human; the female GSF voxel phantoms Donna, Helga, Irene, and Laura are presented in Figure 3.3. When it became clear that our working group would undertake the task of constructing voxel computational phantoms that represented the ICRP adult Reference Male and Reference Female, we decided to modify the segmented voxel computational phantoms that should already conform as far as possible to the required external dimensions to limit the degree of anatomical distortion that had to be applied. Among the male phantoms, Golem was considered suitable, but none of the existing female voxel computational phantoms were close enough to the reference values of height and total body mass.29 Therefore, “Laura” was segmented specifically for this purpose. “Godwin” and “Klara”21 are the results of our first attempt to modify Golem and Laura such that the respective derivatives agree with the ICRP reference anatomical data.29 The final results of these endeavors—voxel computational phantoms that represent the ICRP adult Reference Male and Reference Female and that have been adopted by the ICRP—are presented separately in Chapter 14. For “Katja” a fetus was segmented from an magnetic resonance imaging (MRI) data set of the abdominal and pelvic regions of a female patient in the 24th week of gestation. The abdomen and pelvis of the reference female voxel phantom were then modified such that enough space was created to accommodate the segmented fetus and placenta. We did this by reducing the volume of the bladder contents, by packing the intestines more closely, by shifting them backwards, and by extending the body circumference appropriately. Katja is shown in Figure 3.4. All voxel computational phantoms except the Baby and the Visible Human were based on image data of live patients who had to undergo whole body or head-and-trunk

The GSF Voxel Computational Phantom Family

73

CT examinations for various reasons. The GSF voxel phantoms contain a large number of organs and tissues, including most of the ICRP critical organs, except for the bone endosteum (or bone surfaces), the fi ne soft tissue layer lining the surfaces of the trabeculae and of the long bones’ medullary cavities, and a detailed representation of the RBM located in the small marrow cavities in the spongiosa regions. Most of the adult phantoms have also some tissues relevant to inhalation, like extrathoracic airways (anterior nasal passages, larynx, pharynx), the trachea, and the main branches of bronchi. The latter tissues are thin layers and are represented in the voxel computational phantoms by a single pixel row. Because the pixels’ side length is larger than the thickness of these tissue layers, however, the masses of these tissues are larger in the voxel phantoms than in reality. Obviously, it is not possible to segment structures below voxel resolution and this problem is common to all existing voxel computational phantoms. For the mucous membranes of anterior and posterior nasal passage, which are only a few micrometers thick in reality, this FIGURE 3.4 mass difference is even larger. Despite these limitations, the GSF voxel phantom Katja of presence of these target tissues at their appropriate anatomia woman in the 24th week cal location is a clear advantage compared to the MIRD-type of pregnancy. phantoms, where the doses to these tissues have to be approximated by the doses to “surrogate” organs such as thyroid or lungs. The accuracy of the masses of the eye lenses and skin is also restricted by the voxel size. Concerning the latter tissue, one layer of voxels at the surface of the body represents skin. Consequently, the pixel side length defi nes the skin thickness. This is the reason that the skin masses of the voxel phantoms are somewhat higher than the ICRP reference value, and also differ between the individual voxel computational phantoms. The number and type of organs that were segmented in the individual voxel phantoms evolved in the course of time, and therefore the list of organ identification numbers is also different among the phantoms, depending on when they were segmented. The Baby and Child have 54 and 64 segmented objects, respectively. In Golem, 121 structures were segmented, 67 of them skeletal components, which is the largest number of individual bones among all our phantoms. Sixty-two organs and tissues were identified for Donna, Helga, Irene, and Frank, and 133 for the Visible Human, who has the same number of skeletal components as Golem. Laura, Godwin, and Klara have 88 segmented organs and tissues, respectively. Recently, Baby was complemented with several objects, and the resulting “Babynew” has now 66 segmented organs and tissues. Katja has 136 segmented objects, plus the placenta, the amniotic fluid, the umbilical cord, and 16 tissues that have been segmented in the fetus. Table 3.2 shows the masses of the main organs of the individual voxel phantoms. Multiplying the respective volumes with appropriate tissue densities derived the organ masses. The organ masses of the individual computational phantoms deviate from those of the ICRP reference values,29 since the voxel computational phantoms were constructed from data of real individuals, whereas the reference data represent values averaged over whole populations. However, if one considers those organs that are known to be radiosensitive,30,31 a majority of the organ masses agree within approximately 30% with the ICRP reference values.

Adipose tissue Adrenals Bladder wall Bladder contents Brain Breast Colon wall Colon contents Eye lens Gallbladder Heart Kidneys Liver Lungs Muscle Oesophagus Ovaries Pancreas

2,759. 4.21 2.55 11.5 376. 0.12c 25.8 16.5 0.23 1.03 43.7 30.3 182. 35.5 2,759.a 1.26 0.35 2.96

a

Baby-New

13,580 3.94 24.5 169. 1,316. – 84.6 49.5 0.56 – 206. 188. 733. 153. 13,580a – 2.67 30.1

a

Child 34,820 21.7 61.0 45.0 1,208. 43.9c 322. 309. 1.9 6.6 446.d 281. 1,585. 631. 25,420 27.7 12.1 41.2

Donna 30,590 13.6 56.6 218. 1,827. 1.8c 379. 666. 0.7 9.0 381.d 494. 2,072. 1,338. 13,820b 62.6 – 60.0

b

Frank 19,970 22.8 68.4 272. 1,218. – 297. 237. 0.9 8.3 716. 316. 1,592. 729. 26,970 30.1 – 71.9

Golem 39,800 6.6 60.8 22.1 1,279. 134.c 426. 609. 1.6 5.7 531.d 390. 1,757. 463. 21,340b 28.0 11.9 43.3

b

Helga

Masses (g) of Main Organs of the Members of the GSF Voxel Phantom Family

TABLE 3.2

11,630 12.4 39.0 25.5 1,255. 57.0c 271. 273. 1.6 19.3 472.d 212. 1,225. 685. 21,100 24.3 11.9 61.9

Irene 22,670 12.6 31.9 92.7 1,127. 84.4c 310. 310. 0.66 9.95 280.d 280. 1,334. 1,108. 18,680 21.8 11.3 93.5

Laura

26,040 7.2 51.9 41.2 1,429. – 790. 2,186. 0.5 3.1 637. 383. 2,037. 1,026. 40,970b 86.2 – 62.5

b

Visible Human

23,690 13.0 43.8 156.2 1,300 200c 360. 320 0.4 10.2 250.d 275. 1,400. 950. 17,500 35.0 11.0 120.

Katja

20,050 14.0 50.0 200. 1,450. 3.36c 370. 300. 1.3 13.8 330.d 310. 1,800. 1,200. 29,000 40.0 – 140.

Godwin

23,720 13.0 40.0 200. 1,300. 200.c 360. 320. 0.4 10.2 250.d 275. 1,400. 950. 17,500 35.0 11.0 120.

Klara

74 Handbook of Anatomical Models for Radiation Dosimetry

e

d

c

b

a

416. 211. 31.7 50.2

Skeleton Skin Small intestine Small intestine contents Spleen Stomach wall Stomach contents Testes Thymus Thyroid Uterus

151. 52.1 150. 1.91 30.3 4.96 14.6

2,048. 1,180. 490.e

– 1,228.

306. 195. 305. – 19.0 18.7 71.7

7,484. 4,351. 435. 363.

– 1,012.

174. 233. 140. 21.1 10.7 25.8 –

10,450. 4,703. 959.e

7,250.b 737.b 664. 382. 339. 127. 177. – 3.3 22.3 –

54.7 1,177.

23.3 1,363.b

Adipose tissue and muscle not separated (in these early phantoms). Corresponding mass of part of the arms and legs is missing. Glandular tissue only. Wall (muscle) only. Not separated.

14.7 6.55 12.7 1.28 13.5 1.35 1.35

– 50.8

Prostate RBM

298. 62.8 10.3 – 7.7 31.5 79.8

6,503.b 1,653.b 443. 637.

– 1,043.

203. 163. 205. – 25.3 20.0 25.0

8,201. 3,620. 396. 311.

– 916.

257. 125. 209. – 19.4 24.8 82.5

8,501. 3,012. 843. 424.

– 1,058.

266. 258. 166. 25.5 14.0 31.8 –

8,841.b 1,950.b 521. 767.

37.0 1,399.

130. 140. 230. – 20.0 17.0 484.

7,148. 2,740. 600. 280.

– 900.

150. 150. 250. 35.0 25.0 20.0 –

10,585. 4,404. 650. 350.

17.0 1,170.

130. 140. 230. – 20.0 17.0 80.0

7,148. 2,708. 600. 280.

– 900.

The GSF Voxel Computational Phantom Family 75

76

Handbook of Anatomical Models for Radiation Dosimetry

3.4 Applications of the GSF Voxel Phantoms in Radiation Dosimetry 3.4.1 Monte Carlo Codes For applications in radiation dosimetry, the voxel computational phantoms were combined with various Monte Carlo codes that simulated the radiation transport in material. In our working group, the main focus was on photon dosimetry. Two codes are used predominantly for our computations. The first is a program sometimes called the “GSF code” that was further developed at the GSF5,18,32 from the ALGAM photon transport code stemming from the Oak Ridge National Laboratory.33 This code assumes that the energy transferred at a point of inelastic photon interaction is deposited at that point; secondary electrons are not pursued further (“kerma approximation”). The main advantage of this technique is its high calculation speed, since the pursuit of secondary particles is rather time-consuming, especially in the high-energy domain, where the ranges of these particles are long. The kerma approximation is valid as long as there is approximate secondary particle equilibrium, which can be supposed for all points located well within the body. However, for superficial organs, such as skin and testes, the kerma approximation leads to overestimations of up to a factor of two at photon energy 10 MeV; for these organs, it is valid only below approximately 1 MeV.34,35 The second code used frequently for our computations is the EGSnrc code package.36 Furthermore, the phantoms Godwin and Klara were also combined with MCNPX37,38 and PENELOPE,39 and some preliminary calculations have been performed with these latter code packages.40,41 3.4.2 Bone Dosimetry We used approximations to evaluate the doses to the radiation sensitive tissues in bone: the bone endosteum and the RBM. The mean dose to the skeleton (including hard bone, and red and yellow bone marrow) was taken as a conservative estimate of the dose to the bone endosteum. For the RBM dosimetry, a method initially developed by Kramer32 and later adapted for the use with voxel phantoms19 is employed: the physical transport property of a bone voxel is determined by the mixture medium, from which the total energy deposited in that voxel can be deduced by the Monte Carlo transport code. This amount of energy is then partitioned to the individual bone components according to their mass proportions and mass energy–absorption coefficients. For active bone marrow an additional correction factor is applied which accounts for the extra photoelectrons produced in the bone trabeculae that enter the marrow cavities.42 3.4.3 Idealized Geometries (External) Organ dose conversion coefficients for seven adult voxel phantoms were calculated for those idealized beam geometries commonly assumed to represent occupational exposures, i.e., irradiation by broad parallel beams of monoenergetic photons. The directions of photon incidence were anterior–posterior (AP), posterior–anterior (PA), left lateral (LLAT), right lateral (RLAT), and a full 360° rotation of the photon beam around the longitudinal axis of the body (ROT). The photon energies ranged from 10 keV to 10 MeV. The voxel phantoms employed were Donna, Frank, Golem, Helga, Irene, and the Visible Human from the GSF voxel phantom family, and “Voxelman.”7,43 These calculations were used to get an idea of the range of dose values to be expected for individual phantoms, and to study systematic dosimetric differences between voxel and mathematical phantoms.19 Figure 3.5 shows the

The GSF Voxel Computational Phantom Family

77

Organ equivalent dose/air kerma (Sv Gy–1)

1.6 1.4 1.2 1.0 0.8 0.6

Donna Helga Irene Golem Frank Voxelman Visible Human

0.4 0.2 0.0 0.01

0.1

1

10

Photon energy (MeV) FIGURE 3.5 Equivalent dose conversion coefficients (organ equivalent dose normalized to air kerma free in air, in Sv Gy -1) for the stomach for seven individual voxel phantoms. The geometry is a whole body irradiation with a broad parallel beam of monoenergetic photons impinging in AP direction.

Organ equivalent dose/air kerma (Sv Gy–1)

1.6 1.4 1.2 1.0 0.8 Donna Helga Irene Golem Frank Voxelman Visible Human

0.6 0.4 0.2 0.0 0.01

0.1

1

10

Photon energy (MeV) FIGURE 3.6 Equivalent dose conversion coefficients (organ equivalent dose normalized to air kerma free in air, in Sv Gy -1) for the kidneys for seven individual voxel phantoms. The geometry is a whole body irradiation with a broad parallel beam of monoenergetic photons impinging in PA direction.

stomach equivalent dose conversion coefficients for the seven voxel phantoms for an AP irradiation with monoenergetic photons. The individual variability is only moderate. A further example is given in Figure 3.6 where the respective conversion coefficients for a PA irradiation and the kidneys are shown. In this case, the individual differences are much larger.

Handbook of Anatomical Models for Radiation Dosimetry

78

3.4.4 Environmental Dosimetry For environmental sources, we calculated organ dose conversion coefficients for the Baby and the Child,44,45 and more recently also for various adult voxel phantoms. The geometries considered were a semi-infinite volume source in air that mimics irradiation from a radioactive cloud, an infinite planar source at depth 0.5 g ∙ cm−2 in the ground simulating the deposition of radionuclides in the ground and allowing for surface roughness and initial migration with precipitation, and a semi-infinite volume source in the ground that reflects the homogeneous distribution of natural radioactivity in the ground. Figure 3.7 shows, as an example, liver equivalent doses normalized to the air kerma free in air at 1 m above the ground for the volume source in air and the phantoms Baby, Child, Golem, Visible Human, Voxelman, Donna, and Helga. The conversion coefficients increase substantially with decreasing body size due to a reduced self-shielding capability of smaller bodies. For the adult voxel phantoms, the individual differences—although following the same trend—are only moderate, as can be expected for environmental geometries, since these lead to a rather homogeneous dose distribution in the body with only limited influence of the exact location of an individual organ. A more detailed discussion of dosimetry for environmental exposures is presented in Chapter 15. 3.4.5 Dosimetry for Medical Imaging The evaluation of patient organ and effective dose is important, since the dose due to medical exposures is by far the larger contribution to the total dose received on average by an individual. For a study on the individual variability of organ doses from x-ray examinations, the voxel computational phantoms Donna, Irene, Golem, and Visible Human were employed.46 The following examinations (performed in supine position) were simulated:

Organ equivalent dose/air kerma (Sv Gy–1)

1.0

0.8

0.6

0.4

Baby Child Golem Visible Human Voxelman Donna Helga

0.2

0.0 0

0.1

1

10

Photon energy (MeV) FIGURE 3.7 Equivalent dose conversion coefficients (organ equivalent dose normalized to air kerma free in air 1 m above the ground, in Sv Gy -1) for the liver for seven individual voxel phantoms. The geometry is irradiation by monoenergetic photons from a volume source in air.

The GSF Voxel Computational Phantom Family

79

thorax AP (the most frequent examination in intensive care situations); lumbar spine AP and LAT; pelvis AP; ribs AP; abdomen AP; thoracic spine AP. The exposure conditions were chosen to be optimal for the examination (conventional radiography) and were taken from the European Guidelines on Quality Criteria for diagnostic radiographic images47 and respective German recommendations. The organ doses were found to depend highly on the exposure conditions; and the position of the projection field is critical, particularly for those organs that are partly in the field or near the edge of the field. For some examinations individual dose differences occur, attributed to the different amount of shielding of an organ, depending on the individual topology of these organs. In Table 3.3, organ doses

TABLE 3.3 Organ Equivalent Doses per Entrance Dose Free-in-Air (mSv ⋅ mGy −1) for an Abdomen AP Planar X-Ray Examination of the Voxel Phantoms Irene, Golem, Donna, and Visible Human and for Tube Voltages 75 and 90 kV 75 kV

Bladder (wall) Breast (glandular tissue) Colon (wall) Gonads Liver Lungs Oesophagus RBM Skeleton Skin Stomach (wall) Thyroid Adrenals Brain Kidneys Muscle Pancreas Uterus Small intestine (wall) Spleen Thymus Upper large intestine Lower large intestine Entrance skin surface Exit skin surface

90 kV

Irene

Golem

Donna

Visible Human

Irene

Golem

Donna

0.689 0.007

0.410 –

0.269 0.011

0.175 –

0.784 0.011

0.487 –

0.325 0.015

0.223 –

0.598 0.274 0.390 0.013 0.008 0.071 0.156 0.094 0.325 0.001 0.143 0.000 0.166 0.112 0.266 0.312 0.697

0.469 0.004 0.097 0.005 0.006 0.056 0.112 0.091 0.195 0.000 0.049 0.000 0.091 0.071 0.240 – 0.423

0.373 0.091 0.069 0.007 0.007 0.033 0.084 0.084 0.253 0.001 0.119 0.000 0.072 0.059 0.202 0.134 0.442

0.282 0.014 0.032 0.003 0.004 0.021 0.065 0.103 0.092 0.000 0.040 0.000 0.053 0.046 0.111 – 0.349

0.691 0.356 0.463 0.018 0.013 0.097 0.197 0.101 0.391 0.001 0.195 0.000 0.220 0.131 0.343 0.389 0.798

0.555 0.006 0.124 0.007 0.009 0.078 0.145 0.097 0.236 0.001 0.069 0.000 0.126 0.084 0.311 – 0.510

0.455 0.124 0.090 0.011 0.011 0.048 0.112 0.090 0.306 0.001 0.162 0.000 0.102 0.072 0.264 0.176 0.532

0.342 0.020 0.045 0.005 0.006 0.031 0.090 0.108 0.115 0.000 0.059 0.000 0.078 0.056 0.149 – 0.429

0.130 0.006 0.740

0.023 0.001 0.535

0.095 0.003 0.447

0.020 0.000 0.374

0.169 0.009 0.841

0.033 0.002 0.631

0.127 0.005 0.537

0.030 0.001 0.380

0.402

0.359

0.278

0.180

0.483

0.427

0.349

0.398

1.276

1.267

1.190

1.358

1.312

1.315

1.240

1.402

0.024

0.014

0.017

0.005

0.038

0.023

0.025

0.008

Note: The total filtration is 2.5 mm Al; the focus-to-image receptor distance 115 cm.

Visible Human

80

Handbook of Anatomical Models for Radiation Dosimetry

per entrance dose (free in air) are given for an abdomen AP examination of the four voxel computational phantoms at two different settings of the tube voltage. If the entrance dose is taken as the normalization quantity, the organ dose conversion coefficients increase with increasing tube voltage. However, this effect is more than compensated by the higher amount of radiation transmitted for higher tube voltages and thus reaching the image receptor. This is also indicated by the increasing dose conversion coefficients for the exit skin surface. In those cases where the image receptor demands a constant dose, therefore, the entrance dose required per image receptor dose decreases with increasing tube voltage, and the absolute organ equivalent doses decrease as well. Concerning patient size, the conversion coefficients decrease with increasing patient diameter for organs located in the beam, such as the colon (upper and lower large intestine), pancreas, and kidneys. Also, the seeming increase in conversion coefficients is more than compensated by the higher amount of radiation transmitted through a thinner body. For organs located at the edge of the beam, such as the stomach and the adrenals, there is no obvious dependence of the conversion coefficients on patient diameter, since the proportion of the organ that is inside the beam or the organ distance from the edge of the beam are dominated by the individual organ topology rather than by the patient thickness. 3.4.6 Specific Absorbed Fraction Calculations for Organ Self-Absorption (Original Source Masses) The calculation of absorbed dose to organs due to incorporated radionuclides is based on the quantity absorbed fraction (AF) that specifies the fraction of energy emitted by radioactivity in a “source” organ that is absorbed in a “target” organ. Dividing the AF by the target organ mass leads to the so-called specific absorbed fraction (SAF). Since photons have only a moderate ability to penetrate, especially at lower energies, the larger amount of energy is absorbed in the source organ itself. Consequently, the SAFs for organ self-absorption (target = source) depend strongly on source organ mass that can be very different among the individual voxel computational phantoms. An example is given in Figure 3.8 for self-irradiation of the thyroid.48 The large differences of the SAF values for the individual computational phantoms reflect the large range of thyroid masses that are between 6.2 g (for Voxelman) and 31.5 g (for Visible Human). The statement of the MIRD commission that the photon self-dose to an organ scales as (1/mass2/3)49 is exactly reflected by these data.50 3.4.7 SAF Calculations for Organ Cross-Fire The situation is different when irradiation of a target organ by photons released in a different source organ is considered. For this organ “cross-fire,” the source or target organ masses have no significant influence, and the variations are assumed to be attributable to individual differences in the organ topology, such as the distance of source and target organ. At the former GSF, SAF have been calculated for approximately 40 source organs and seven adult voxel phantoms.48 As an example, Figure 3.9 shows the SAF values for liver as source and stomach as target organ. 3.4.8 Applications Outside the Helmholtz Zentrum München Further to the studies performed by our working group, some members of the GSF voxel phantom family have been used by other researchers for a variety of applications,

The GSF Voxel Computational Phantom Family

81

Frank Golem Visible Human Voxelman Donna Helga Irene

SAF (kg–1)

100

10

1

0.01

0.1 1 Photon energy (MeV)

10

FIGURE 3.8 SAFs (kg−1) for self-irradiation of the thyroid with monoenergetic photons and for seven adult voxel phantoms.

SAF (kg–1)

0.1

0.01 Frank Golem Visible Human Voxelman Donna Helga Irene 0.001 0.01

0.1

1

10

Photon energy (MeV) FIGURE 3.9 SAFs (kg−1) for irradiation of the stomach with monoenergetic photons that originate in the liver, and for seven adult voxel phantoms.

such as dosimetry for multidetector CT,51 and dose calculations for exposure to space radiation.52,53 In principle, one could assign other types of properties to the segmented voxel array, such as resilience, stability, or other physical, chemical or biological properties. This might further broaden the applicability of this family of voxel phantoms.

Handbook of Anatomical Models for Radiation Dosimetry

82

3.5 Conclusions Voxel computational phantoms are the most developed, complete, and realistic computational phantoms of the human anatomy. Thus, they offer a clear improvement compared to the older mathematical phantoms whose organs are described by relatively simple geometrical bodies. Generally, voxel phantoms are a breakthrough in the realistic modeling of the human body and enable a better accuracy for many radiation protection dosimetry applications. There is, however, an obvious variability among the dose data for the individual voxel phantoms, due to differences in stature and individual anatomical features. Despite the obvious advantages of the existing voxel phantoms, compared to the stylized mathematical computational phantoms, it is also clear that a voxel phantom—being constructed from the image data of an individual—does not necessarily represent a whole population, or an average person. Especially in view of the significant influence of organ masses on the resulting organ doses in internal dosimetry, it is obvious that even a large variety of individual voxel phantoms cannot meet the demand for computational phantoms of standard persons. The emergence of such reference computational phantoms does not, however, mean that the other voxel phantoms are becoming obsolete. To a certain extent, they could be used as tools toward a more personalized dosimetry. Since the currently existing voxel phantoms range from slim persons to heavy persons, the dose values published so far give a dose range in which an individual dose may be expected to lie, together with an indication of the magnitude of dose differences to be expected between individual persons. Furthermore, it is believed that they can be used to roughly estimate the doses to an individual by selecting those for the voxel computational phantom fitting best to the person under consideration. It should be clear, however, that existing voxel computational phantoms—both individual and reference—cannot represent any real individual, and that organ dose conversion coefficients from literature cannot be directly applied to an individual person. Especially in those situations where a reliable dose assessment for an individual is required, this approach is not possible, e.g., for radiation treatment planning purposes.

References 1. Snyder, W.S. et al. Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, J. Nucl. Med., 10, Suppl 3, 5, 1969. 2. Snyder, W.S., Ford, M.R., and Warner, G.G. Estimates of specific absorbed fractions for photon sources uniformly distributed in various organs of a heterogeneous phantom, MIRD Pamphlet No. 5, revised, Society of Nuclear Medicine, New York, 1978. 3. Cristy, M. Mathematical phantoms representing children of various ages for use in estimates of internal dose, U.S. Nuclear Regulatory Commission Report NUREG/CR-1159 (also Oak Ridge National Laboratory Report ORNL/NUREG/TM-367), 1980. 4. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, Part I: Methods, TM-8381/V1, Oak Ridge National Laboratory, Oak Ridge, TN, 1987. 5. Kramer, R. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods: Part I. The male (ADAM) and female (EVA) adult mathematical phantoms, GSF-Report S-885, Institut fuer Strahlenschutz, GSF-Forschungszentrum fuer Umwelt und Gesundheit, Neuherberg-Muenchen, 1982.

The GSF Voxel Computational Phantom Family

83

6. Zankl, M. et al. The construction of computer tomographic phantoms and their application in radiology and radiation protection, Radiat. Environ. Biophys., 27, 153, 1988. 7. Zubal, I.G. et al. Computerized three-dimensional segmented human anatomy, Med. Phys., 21, 299, 1994. 8. Dimbylow, P.J. The development of realistic voxel phantoms for electromagnetic field dosimetry, Workshop on Voxel Phantom Development, Chilton, U.K., 1996. 9. Caon, M., Bibbo, G., and Pattison, J. An EGS4-ready tomographic computational model of a 14-year-old female torso for calculating organ doses from CT examinations, Phys. Med. Biol., 44, 2213, 1999. 10. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-MAN: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Phys., 78, 476, 2000. 11. Zankl, M. and Wittmann, A. The adult male voxel model “Golem” segmented from whole body CT patient data, Radiat. Environ. Biophys., 40, 153, 2001. 12. Petoussi-Henss, N. et al. The GSF family of voxel phantoms, Phys. Med. Biol., 47, 89, 2002. 13. Kramer, R. et al. MAX06 and FAX06: Update of two adult human phantoms for radiation protection dosimetry, Phys. Med. Biol., 51, 3331, 2006. 14. Gibbs, S.J. et al. Patient risk from interproximal radiography, Oral Surg. Oral Med. Oral Pathol. Oral Radiol. Endod., 58, 347, 1984. 15. Gibbs, S. et al. Radiation doses to sensitive organs from intraoral dental radiography Dentomaxillofac Radiol., 16, 67, 1987. 16. Williams, G. et al. The calculation of dose from external photon exposures using reference and realistic human phantoms and Monte Carlo methods, Phys. Med. Biol, 31, 449, 1986. 17. Williams, G. et al. The construction of 3D whole body images from CT data and the use of image processing methods to produce files for Monte Carlo dose calculations, CAR 87, Springer Verlag, Berlin, Germany, 1987, p. 148. 18. Veit, R. et al. Tomographic anthropomorphic models, Part I: Construction technique and description of models of an 8 week old baby and a 7 year old child, GSF-Report 3/89, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1989. 19. Zankl, M. et al. Organ dose conversion coefficients for external photon irradiation of male and female voxel models, Phys. Med. Biol., 47, 2367, 2002. 20. Fill, U.A. et al. Adult female voxel models of different stature and photon conversion coefficients for radiation protection, Health Phys., 86, 253, 2004. 21. Zankl, M. et al. GSF male and female adult voxel models representing ICRP Reference Man— the present status, The Monte Carlo Method: Versatility Unbounded in a Dynamic Computing World, Chattanooga, TN, 2005. 22. Becker, J. et al. About Katja, a virtual human phantom of a 24-week pregnant woman, Proceedings of the 7th International Scientific Conference SATERRA “Human and Environment,” Mittweida, Germany, 2007. 23. Cristy, M. Active bone marrow distribution as a function of age in humans, Phys. Med. Biol., 26, 389, 1981. 24. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: The Skeleton, ICRP Publication 70, Pergamon Press, Oxford, U.K., 1995. 25. Schlattl, H., Zankl, M., and Petoussi-Henss, N. Organ dose conversion coefficients for voxel models of the reference male and female from idealized photon exposures, Phys. Med. Biol., 52, 2123, 2007. 26. Zankl, M., Eckerman, K.F., and Bolch, W.E. Voxel-based models representing the male and female ICRP reference adult—the skeleton, Radiat. Prot. Dosim., 127, 174, 2007. 27. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Pergamon Press, Oxford, 1975. 28. Spitzer, V.M. and Whitlock, D.G. Atlas of the Visible Human Male, Jones and Bartlett Publishers, Sudbury, MA, 1998. 29. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, U.K., 2003.

84

Handbook of Anatomical Models for Radiation Dosimetry

30. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Pergamon Press, Oxford, U.K., 1991. 31. ICRP. The 2007 Recommendations of the International Commission on Radiological Protection, ICRP Publication 103, International Commission on Radiological Protection, Elsevier, Oxford, U.K., 2007. 32. Kramer, R. Ermittlung von Konversionsfaktoren zwischen Körperdosen und relevanten Strahlungskenngrößen bei externer Röntgen- und Gamma-Bestrahlung, S-556, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1979. 33. Warner, G.G. and Craig, A.M. ALGAM, a computer program for estimating internal dose from gamma-ray sources in a man phantom, TM-2250, Oak Ridge National Laboratory, Oak Ridge, TN, 1968. 34. Chao, T.C., Bozkurt, A., and Xu, X.G. Conversion coefficients based on the VIP-Man anatomical model and EGS4-VLSI code for external monoenergetic photons from 10 keV to 10 MeV, Health Phys., 81, 163, 2001. 35. Saito, K. et al. Construction of a computed tomographic phantom for a Japanese male adult and dose calculation system, Radiat. Environ. Biophys., 40, 69, 2001. 36. Kawrakow, I. and Rogers, D.W.O. The EGSnrc code system: Monte Carlo simulation of electron and photon transport, PIRS Report 701, National Research Council of Canada (NRCC), Ottawa, 2003. 37. Waters, L. MCNPX user’s manual, version 2.3.0, Los Alamos National Laboratory Report, LA-UR-02-2607, April 2002. 38. Hendricks, J.S. et al. MCNPX extensions, version 2.5.0, LA-UR-05-2675, LANL, Los Alamos, NM, 2005. 39. Salvat, F., Fernandez-Varea, J.M., and Sempau, J. PENELOPE, a code system for Monte Carlo simulation of electron and photon transport, OECD, 2003. 40. Taranenko, V. and Zankl, M. Photon and electron transport simulation in voxel geometry with PENELOPE, Biomed. Tech. (Berl), 50, 271, 2005. 41. Taranenko, V., Zankl, M., and Schlattl, H. Voxel phantom setup in MCNPX, in Proc. The Monte Carlo Method: Versatility Unbounded in a Dynamic Computing World, LaGrange Park, USA: American Nuclear Society, Chattanooga, TN, 2005. 42. King, S.D. and Spiers, F.W. Photoelectron enhancement of the absorbed dose from x-rays to human bone marrow: Experimental and theoretical studies, Br. J. Radiol., 58, 345, 1985. 43. Zubal, I.G. et al. Two dedicated software, voxel-based, anthropomorphic (torso and head) phantoms, in Proc. Workshop on Voxel Phantom Development, Dimbylow, P.J., Ed. National Radiological Protection Board, Chilton, U.K., 105, 1996. 44. Saito, K. et al. Organ doses as a function of body weight for environmental gamma rays, J. Nucl. Sci. Technol., 28, 627, 1991. 45. Saito, K. et al. The calculation of organ doses from environmental gamma rays using human phantoms and Monte Carlo methods, Part I: Monoenergetic sources and natural radionuclides in the ground, GSF—Report 2/90, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1990. 46. Petoussi-Henss, N., Zankl, M., and Panzer, W. Estimation of organ doses in radiology using voxel models describing different patients, Biomed. Tech. (Berl.), 50, 664, 2005. 47. European Commission. Quality criteria for diagnostic radiographic images, EUR 16260, Office for Official Publications of the European Communities, Luxembourg, 1996. 48. Zankl, M. et al. The application of voxel phantoms to the internal dosimetry of radionuclides, Radiat. Prot. Dosim., 105, 539, 2003. 49. Snyder, W.S. et al. “S” absorbed dose per unit cumulated activity for selected radionuclides and organs, MIRD Pamphlet 11, Revised, Society of Nuclear Medicine, New York, 1975. 50. Petoussi-Henss, N. et al. Patient-specific scaling of reference S-values for cross-organ radionuclide S-values: What is appropriate?, Radiat. Prot. Dosim., 127, 192, 2007.

The GSF Voxel Computational Phantom Family

85

51. DeMarco, J.J. et al. Estimating radiation doses from multidetector CT using Monte Carlo simulations: Effects of different size voxelized patient models on magnitudes of organ and effective dose, Phys. Med. Biol., 52, 2583, 2007. 52. Ballarini, F. et al. GCR and SPE organ doses in deep space with different shielding: Monte Carlo simulations based on the FLUKA code coupled to anthropomorphic phantoms, Advances in Space Research, 37, 1791, 2006. 53. Trovati, S. et al. Human exposure to space radiation: Role of primary and secondary particles, Radiat. Prot. Dosimetry, 122, 362, 2006.

4 The ADELAIDE Teenage Female Voxel Computational Phantom Martin Caon, Giovanni Bibbo, and John E. Pattison

CONTENTS 4.1 Introduction ................................................................................................................. 87 4.2 The ADELAIDE Phantom .......................................................................................... 88 4.2.1 The Constraints Imposed by the Available Images .................................. 88 4.2.2 Segmentation Method: Identifying the Organ Boundaries ..................... 89 4.2.3 ADELAIDE Body Size Compared to Mean Body Sizes for Age ............. 92 4.2.4 The Size of ADELAIDE’s Organs ................................................................ 93 4.2.5 The Shape and Locations of Organs in the Body ...................................... 97 4.2.6 The Location of Organs within the Body ................................................... 99 4.2.7 Conversion of Segmented Anatomy Images to an Input File for the EGS4 Monte Carlo Code ........................................................... 99 4.3 Organ Dose Calculations with ADELAIDE and EGS4........................................ 101 4.4 Limitations of the Computational Phantom ......................................................... 102 References ............................................................................................................................. 102

4.1 Introduction The motivation for producing a pediatric computational phantom was to determine the absorbed organ doses from computed tomography (CT) examinations using the public domain EGS4 Monte Carlo code.1 The project was conceived in about 1992 in response to the dearth of available absorbed dose data for children who underwent CT procedures. Inspiration was derived from the work of Cristy, who extrapolated the mathematical medical internal radiation dose (MIRD) computational phantoms to children2 and from the work of Zankl et al. who produced voxel computational phantoms of two children.3 The Monte Carlo dose calculations for CT procedures using “adult” MIRD computational phantoms by Shrimpton et al.4 and later for plain radiography in children with MIRD computational phantoms5 also inspired the ADELAIDE work. In the early 1990s, desktop PCs were not capable of performing Monte Carlo calculations; picture archiving and communication systems (PACS) were not yet available; the Digital Imaging and Communications in Medicine (DICOM) standard for distributing and viewing medical image from any origin was not operating; computers were not networked; the electronic

87

88

Handbook of Anatomical Models for Radiation Dosimetry

transfer of data was limited by the capacity of the “floppy disk”; and the Internet had not yet developed into the information distributing system that it currently is. Consequently, working with medical images on computers was not straightforward.6 The situation in 2007 is much better and the number of voxel computational phantoms has increased dramatically.7 Despite the improvements in computer hardware and software, segmenting the large number of medical images required for a voxel computational phantom is a timeconsuming task. Even though steps have been taken to semiautomate the task,8–10 segmentation is still a lengthy operation. This chapter describes the segmentation procedure used for ADELAIDE’s construction and the anatomy of the computational phantom.

4.2 The ADELAIDE Phantom 4.2.1 The Constraints Imposed by the Available Images Obtaining the medical images that are required to produce a voxel computational phantom was not an easy task in the 1990s. Medical images are confidential patient medical records and were not readily shared by medical centers. Furthermore, extensive CT examinations that spanned a large proportion of a patient’s anatomy were not common and in any case usually did not include the arms and legs. Consequently, prior to the construction of ADELAIDE11 only two voxel computational phantoms of children existed and they were modeled from the cadaver of an 8 week old baby and from a 7 year old patient who was undergoing treatment for leukemia.3 The purpose for constructing ADELAIDE was to calculate torso organ doses from CT. Thus, even though significant anatomy (the head, arms, and legs) were excluded from the computational phantom (see Figure 4.1), that anatomy is not of interest in many CT examinations so is not irradiated. However, the CT scanner patient table was also modeled so that the attenuation it produces was also simulated. Hence, the computational phantom suited its purpose. Indeed, the presence of arms alongside the torso would be detrimental to the accurate calculation of CT dose, as arms are routinely held out of the beam. Their inclusion would shield the underlying tissue from the x-ray beam. At the time of ADELAIDE’s construction (commencing in 1996), the DICOM standard had not been widely deployed to medical imaging equipment, so getting access to the GE format image files stored in the scanner’s computer was not straightforward. A software program that was able to interrogate the CT scanner and download the GE proprietary format files and convert them to be saved as.tif files was obtained from GE and used for the purpose. Fifty-four CT images at an interval of 10 mm of a female torso captured by a FIGURE 4.1 The span of anatomy included in the General Electric HiSpeed Advantage CT scanner were ADELAIDE voxel model. (From Caon, available for modeling. The image files were 512 × 512 M., Bibbo, G., and Pattison, J., Phys. Med. pixels in size and the field of view for the images was a Biol., 44, 2213, 1999. With permission.)

The ADELAIDE Teenage Female Voxel Computational Phantom

89

FIGURE 4.2 Left: Original 512 × 512.tif image. Right: Image showing extrapolated tissue. Bottom: Image resized to 128 × 128 and segmented.

circle of diameter 29 cm. As the patient’s width at the shoulders and hips was greater than 29 cm, the original images showed truncated anatomy in these regions (see for example Figure 4.2a, a slice through the chest superior to the heart). This problem was overcome by estimating the external contour of the skin and using image processing software to manually draw in the skin boundary for the missing anatomy in each image (Figure 4.2b). Some additional extrapolation was done on the bitmap data after exporting it to “MS Excel.” Each added pixel was then identified as containing either muscle, bone, subcutaneous fat or skin and its gray scale value assigned accordingly. 4.2.2 Segmentation Method: Identifying the Organ Boundaries Segmentation was done completely manually using Paint Shop Pro v4.14 (JASC Inc.) and Image-Pro Plus v3.0 (Media Cybernetics). Images were resized to 128 × 128 pixels which resulted in voxels of size 2.53 × 2.53 ×10.0 mm. As the slices were 10 mm apart in the z-direction, the loss of resolution in the x- and y-directions that resulted from the resizing was not seen as significant. Resizing also had the advantage of making the computational phantom manageable on the computer that was available. A disadvantage of the low (by modern standards) resolution was that the skin surface and the boundary between organs were “stepped” (Figure 4.2c). The image processing software allowed the images to be magnified so that individual pixels could be seen, their gray scale value to be viewed and the value altered as desired. In this way, a boundary between organs or tissue could be drawn and the gray scale values of the contained pixels altered to the value chosen for that tissue. Where necessary this was done pixel by pixel.

Handbook of Anatomical Models for Radiation Dosimetry

90

An underlying assumption used in preparing the voxel computational phantom is that the individual tissues or organs are homogeneous and so are assigned the same gray scale value. That is, we assume that all of the voxels assigned to a particular organ or tissue have the same composition and density. This, of course, is not the case as for example the pixels in subcutaneous fat tissue had gray scale values that typically ranged from 65 to 90, for skin a typical range is from 85 to 125, muscle pixels ranged from 138 to 173, liver from 153 to 204, spleen from 169 to 200 (where 0 is black for air/lung and 255 is white for cortical bone). The assigned gray scale was chosen to be typical of the pixels of that tissue while being sufficiently different from the chosen gray scale of neighboring organs to be distinguishable by the naked eye when viewed on the computer screen. For the purposes of identifying the different media in EGS4, each organ/tissue was in addition to a gray scale value, assigned an EGS4 medium index (Table 4.1). TABLE 4.1 Assigned Gray Scale Values to Organ/Tissue and EGS4 Medium Indices EGS4 Medium Index 1 2 3 4 5 6 7 8 10 11 13 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Tissue

Assigned Gray Scale

Air surrounding model Cortical bone, internal bone, and marrow Heart (and great blood vessels) Spinal cord Skin Subcutaneous fat Breasts Lungs Esophagus Muscle and soft tissue Kidney Liver Spleen Stomach and contents Blood (aorta and IVC) Gas (in digestive system) Bowel and contents Pancreas Gall bladder (contents) Uterus Ovaries Bladder (empty) Thymus Trachea Thyroid Colon and contents Bone surface

0 255 3 4 110 70 40 1 10 66 80 15 16 17 30 60 90 50 250 22 33 44 26 27 39 77 254

Note: Indices 9 and 12 are assigned to the CT scanner’s patient table. Index 14, initially assigned to spongy (cancellous) bone, is not used as it could not be reliably distinguished from cortical bone.

The ADELAIDE Teenage Female Voxel Computational Phantom

91

The process of segmentation is tedious due to the large number of organs to delineate in each CT slice and the large number of slices. The majority of organs such as the lungs, liver, spleen, kidneys, bone, muscle, and subcutaneous fat were easily distinguishable by eye on a computer screen and their edges were straightforward to identify. However, on a magnified version of the image, selecting the individual voxels that lie on the boundary of a tissue was often a matter of judgment. For example while skin, subcutaneous fat and muscle were distinguishable from each other, the gray scale values of the pixels at their common borders were intermediate in value between the gray scale values of pixels located in the center of fat and muscle tissue. Voxels inside the skin were usually assigned to fat if they were sufficiently dark to distinguish them from the higher gray scale value of muscle tissue. The precise boundary between breast tissue and fat was difficult to discern. Where there was significant doubt, voxels were assigned to breast to avoid underestimating their size. Consecutive CT slices were examined to ensure that the size and shape of the breasts were sequentially consistent. The subject had very little discernable visceral fat so tissue not included in an organ and deep to subcutaneous fat was assigned to muscle and soft tissue. In the case of the esophagus, pancreas, thymus, and ovaries, it was difficult to distinguish the organ from the surrounding tissue of similar gray scale. In this case, a radiologist and a CT anatomy atlas were used to identify the organs. The adrenals and kidneys were not separated and were given the same gray scale and medium index. The digestive system consists of the stomach, the small intestine, the large intestine, and their contents—which may be gas, semiliquid, or a semisolid paste. Gas in the system was easily distinguishable, as gas voxels had gray scale 0. Gas was also discernable in the trachea and esophagus. Where gas occurred within an organ, voxels containing gas were distinguished from voxels containing solid media and energy imparted to the gas was not included in the calculation of absorbed dose to the organ. Sometimes, the contents of the gut were also apparent. For example in the stomach, the horizontal boundary between liquid chyme (stomach contents) and gas indicated that below the boundary, chyme was present rather than stomach wall (see Figure 4.3). The ascending and descending colon were usually imaged in the CT slices as transverse sections so a reasonable attempt at distinguishing the contents from the wall could be made. However, the boundary between large intestine and small intestine was not clear. Nor was the difference between small intestine wall and its contents apparent due to the complicated pattern of folding of the intestine. Hence, contents and wall were given the same gray scale number. The stomach and descending colon were distinguished from each other and from the small intestine and ascending and transverse colon. These latter three organs of the gut (and their contents) were collectively called bowel and given the one gray scale and EGS4 medium index. Bone tissue stood out clearly in the CT images as the white pixels (gray scale 255) contrasted with the surrounding tissue. The gray scale value gradually decreases at the edges of bones so bone tissue was taken to occupy all pixels with gray scale greater than 225. The outermost bone voxels were further distinguished as being “surface bone voxels.” This was done to facilitate the calculation of absorbed dose to ICRP 60’s “bone surface.” Note that the thickness of this tissue, like the skin, is determined by the voxel FIGURE 4.3 size. Because of this limitation, the calculated Note the gas in the stomach and bowel.

92

Handbook of Anatomical Models for Radiation Dosimetry

dose obtained from ADELAIDE is to the surface voxels rather than to the “bone surface” referred to by ICRP 60. Dose to the 10 μm layer of tissue at the bone surface is enhanced (over dose to soft tissue not adjacent to a soft tissue/bone interface) due to the absorption of the energy of photoelectrons that are produced in bone. The dose to bone surface is in practice replaced by the mean dose in bone.12 It is argued that since the dose to bone is higher than to tissue, using mean dose to bone is a conservative estimate for the enhanced dose to the soft tissue layer adjacent to bone. Replacing it instead with mean dose to surface bone voxels (which are composed of cortical bone), as we have done, may overestimate the dose to bone surface even more as the attenuation coefficient for cortical bone is greater than for less dense bone. The less dense cancellous bone (containing marrow) was, in some images, distinct from the more dense cortical bone. However, it was decided not to distinguish cancellous from cortical bone as we were not confident that all the cancellous bone was able to be identified. The position of some slices and the images of bones displayed at times made it difficult to distinguish between pixels with gray scale less than 255 due to containing cancellous bone, from those due to being the edge of a bone imaged at a glancing angle. Hence cancellous bone and cortical bone (but not the surface voxels) were given the same gray scale and called collectively “internal bone.” Internal bone was considered to be a homogeneous mixture consisting of cortical bone, red marrow, and yellow marrow in the proportions 40%, 39%, and 21% by mass respectively. These percentages were chosen so that when the masses of voxels containing bone surface and internal bone were added to give total skeleton mass, the proportions of this mass due to the three components of the skeleton are 69% cortical bone, 20% red marrow, and 11% yellow marrow. These proportions (69%, 20%, and 11%) are the same as those Cristy2 used in his 15 year old computational phantom when only the bones that the Cristy computational phantom has in common with our torso are considered. Some tissues that are not specified by ICRP 60 as needed to determine effective dose (either as a tissue assigned a weighting factor or as one of the remainder tissues) are included in ADELAIDE because they were visible on the images. These were subcutaneous fat, the heart and adjacent great blood vessels, the descending aorta and inferior vena cava (and the blood they contain), the spinal cord, the trachea, that proportion of internal bone that is yellow marrow or cortical bone, and the contents of the gall bladder. The heart includes the heart wall and the enclosed blood and a portion of the great blood vessels where these were not able to be distinguished from the heart. When the images clearly showed the aortic arch, the descending aorta, and the inferior vena cava (and other smaller blood vessels), these tissues were identified as blood. Voxels assigned to “muscle and soft tissue” included the readily identified skeletal muscle, lymph tissue, any soft tissue adjacent to organs that was clearly not part of the organ and visceral fat. There was very little discernable visceral fat surrounding the organs. Hence, it was common for one organ to share a boundary with another organ without any intervening soft tissue being evident. If any soft tissue was actually present between such organs (without being discernable), it has been included within the organs. Consequently, it is unlikely that the size of organs has been underestimated in the process of segmentation and it is more likely that their size has been overestimated to a small extent. 4.2.3 ADELAIDE Body Size Compared to Mean Body Sizes for Age Hitchcock et al.13 in a survey of the body size of young Australians living in Perth found the mean and standard deviation of the mass of a sample of 175 girls aged 13 years (average age 13.5 years) to be 48.7 ± 8.0 kg. The figures for 145 girls of 14 years (average age 14.5 years)

The ADELAIDE Teenage Female Voxel Computational Phantom

93

were 52.1 ± 9.2 kg. The standard deviations indicate that it is common for the mass of girls of these ages to vary by up to 17% from the mean. The ADELAIDE torso was modeled from a patient of age 13 years and 11 months and mass about 48 kg. Hence, the mass of ADELAIDE is slightly less (by 1.5%) than the average mass of Australian 13 year old girls, but well within the normal range. The heights of the 13 and 14 year old girls in Hitchcock et al.’s survey were 158.1 ± 6.9 cm and 161.3 ± 6.3 cm, respectively. For the case of height, it can be expected that a 4% deviation from the mean height-for-age is not uncommon in girls in their early teens. ADELAIDE was constructed from a patient whose height was about 157 cm, hence is slightly less (by 0.7%) than, but very close to the average height of Australian 13 year old girls. Hitchcock et al. compared the median weights and heights from their survey with data from the United States (published by the National Center for Health Statistics in 1977) and found small differences in weight between the girls and no significant height differences. Hence the Australian averages are very close to the averages for girls in the United States. A single voxel computational phantom cannot be representative of all 13–14 year old girls, however ADELAIDE is close to the average stature for her age. 4.2.4 The Size of ADELAIDE’s Organs The Cristy 15 year old MIRD computational phantom 2 was designed to represent an average-sized teenager and is intermediate in size between a boy and girl of that age. On the other hand, ADELAIDE is representative of the average height and weight of an Australian 13 year old girl. Differences between the two computational phantoms are to be expected due to the different “ages” of the computational phantoms and also because ADELAIDE is female while the Cristy computational phantom is hermaphrodite. The anatomies of ADELAIDE and a MIRD style phantom differ markedly in the disposition of internal space to organs. The organs of a MIRD computational phantom, being described by mathematical equations, are constrained in shape to that which can be represented easily by equations. While the volumes of organs are based on anatomical data, the resulting shapes resemble the actual shape of organs only very approximately. They are further constrained in position by the need to occupy a space that does not overlap that of another organ. This results in some “free space” between organs—filled with soft tissue—that is not present in real anatomy. On the other hand, the organs of ADELAIDE are realistic in shape, fit snugly against each other and are separated (if at all) by the actual amount of soft tissue present in a real patient. A major difference is in the allocation of soft tissue (see Table 4.2). In ADELAIDE, subcutaneous fat (medium 6) was visible beneath the skin so is identified and excluded from the category “muscle and soft tissue” (medium 11). Similarly, 1023 cm3 of gas was visible in the digestive system and esophagus and this also is excluded. In the Cristy computational phantom, all the torso volume outside of the identified organs is deemed soft tissue. Hence 18,364 g (67%) of the 27,284 g torso is soft tissue. When all the soft tissues of ADELAIDE are totaled (media with indices: 4, 6, 10, 11, 18, and 27), the result is that 12,245 g (52%) of the 23,636 g torso is soft tissue. Another major difference between the Cristy computational phantom and ADELAIDE is the absence of arms in the latter. The Cristy computational phantom includes the soft tissue and bone of each arm within the elliptical cylinder of the torso envelope. To arrive at a mass for a Cristy “torso,” we have taken the two arms to constitute 10% of the total Cristy body mass (i.e., 5564 g of the 55,644 g mass) and in Table 4.2 have subtracted this mass from the mass (32,800 g) of the Cristy torso. Organ mass in the ADELAIDE computational phantom is determined by the formula: No.of voxels assigned to organ × volume of a voxel × density of voxel.

Handbook of Anatomical Models for Radiation Dosimetry

94

TABLE 4.2 Organ Masses of the ADELAIDE 14 Year Old Torso Compared with Cristy’s 15 Year Old Trunk and Data for 14 Year Old Japanese Girls Medium Index

Tissue

ADELAIDE Mass (g)

Cristy 15 Year Old Mass (g)

Red marrow (39% of “internal bone”) Yellow marrow (21% of int bone) Cortical bone (40% of int bone) Heart (ADELAIDE Inc., blood and great blood vessels) Spinal cord Skin Subcutaneous fat Breasts Lungs

866

845a

466

404

888

1703b

705

238 (+330 blood)

10 11 13

Esophagus Muscle and soft tissue Kidney

15 16 17

2

3 4 5 6 7 8

69 1018 2577 300 1085

Ogiu et al. Mass (g)g

237 ± 15

1059c 44 650

729 ± 63

31 9337 363 (inc. adrenals)

(31.4d) 18,420e 235 (+10 for adrenals)

232 ± 21 (+8 for adrenals)

Liver

1543

1220

1285 ± 157

Spleen

204

117

127 ± 35

128 (+208 cont)

112 (+185 cont)

221 1023 cm3 620 (+646 cont)

549 (+573 cont)

21

Stomach wall (38% of ADELAIDE stomach and contents) Blood (in aorta and IVC) Gas (in digestive system) Bowel walls (small and large) (49% of wall and contents) Pancreas

58

62

79 ± 11

22 23 24 25 26

Gall bladder (contents) Uterus Ovaries Bladder Thymus

8 35 11 19 (empty) 15

47 25 5 34 (+152 cont) 27

35 ± 15

27 28

Trachea Thyroid

10 18

12

12 ± 2

29

Colon wall (54% of ADELAIDE wall and contents) “Bone surface” Total

243 (+207 cont)

120 (+104 cont)

1736 23,636

27,280f

18 19 20

30 a

b

c d e f

g

Includes red marrow of clavicles, scapulae, ribs, pelvis (includes 50% of upper half of femur), spine, one-third of upper portion of arm bones. Mass of Cristy bone (clavicles, scapulae, ribs, pelvis, spine, one-third of upper portion of arm bones, excludes marrow). Volume of skin on trunk × skin specific gravity (1.105). Yamaguchi’s value.21 Cristy named tissues = 8860 g (unnamed soft tissue = 27,280 – 8860 = 18,420 g). Mass of Cristy 15 year old trunk = 32,800 g (add breasts (+44 g), minus arms each assumed to be 5% of total body weight (−5564 g for both arms) = 27,280 g). Mass values ± standard deviation.

The ADELAIDE Teenage Female Voxel Computational Phantom

95

Table 4.2 presents the organ masses determined in this way as well as the organ masses in Cristy’s 15 year old computational phantom and average organ masses of between 3 and 6 Japanese 14 year old girls.14 One would expect that the organs in the Cristy 15 years old, because it represents a larger-sized teenager than ADELAIDE, and considering the effect that the larger organs of boys would have on its organ sizes, would be larger than the organs in ADELAIDE. This is not the case. The contents of the gall bladder, the urinary bladder walls, and the thymus do have a smaller mass in ADELAIDE. In addition, the masses of six organs or tissues (red marrow, yellow marrow, skin, stomach wall, small and upper large intestine walls, and the pancreas) are within 15% of each other in the two computational phantoms. However, the other organs and tissues (except for muscle and soft tissue) are significantly greater in mass in ADELAIDE (see Table 4.2). We will consider muscle and soft tissue first. The Cristy computational phantom has 67.5% of its mass allocated to muscle and (unnamed) soft tissue, while ADELAIDE has 51.8% of its mass as muscle and soft tissue (including subcutaneous fat, spinal cord, esophagus, blood in blood vessels, and the trachea). Given that the Cristy computational phantom represents an average 15 years old and ADELAIDE is close to the average size of an Australian 13 years old, it could be expected that the proportion of body mass that is muscle and soft tissue would be more similar than it is. It is possible that the Cristy value is too high or the ADELAIDE value is too low (when compared to the “real” proportion) or that both computational phantoms fail to estimate the true proportion accurately. In support of the first contention, the MIRD computational phantom has significant gaps between its organs that contain soft tissue. In ADELAIDE, almost all of the volume within the thoracic and abdominopelvic cavities is assigned to identified organs. That is, there is no identifiable visceral fat and very little soft tissue separating the organs. The external profile of the MIRD computational phantom, by not narrowing at the waist, allows more soft tissue to be present between the digestive organs and the skin than is the case in real teenagers. In support of the second contention, the organs in ADELAIDE are likely to incorporate some of the connective tissue (mesenteries) adjacent to their borders and between them and the neighboring organ. If anything, this overestimates the size of the organs in ADELAIDE. For example, the diaphragm was not separately distinguished so the diaphragm tissue is included in the liver, stomach, or heart. Heart tissue includes the pericardium. Part of the omentum (tissue layer overlying the intestine) will have been assigned to the intestinal tissue. Hence, in ADELAIDE, the assigned mass of an organ is likely to overestimate the actual mass. However, in the cases where this has occurred, the overestimation is likely to be small, since if the amount of tissue between organs was not small, it would have been discernable in the CT images and identified. Thus, the third contention (that the proportion of body mass that is muscle and soft tissue in the Cristy computational phantom is too high, and in the ADELAIDE computational phantom is too low) is the most likely to be true, but the actual typical proportion of body mass that is muscle and soft tissue is probably closer to the proportion in ADELAIDE. To highlight the difference in the size of organs between the two torsos, the organ mass as a percentage of their respective torso mass is presented in Table 4.3 (for the organs used to calculate effective dose). Except for the bladder, the percentages for ADELAIDE’s organs are all greater than for Cristy’s 15 year old torso. The proportional organ mass for ADELAIDE is 50% greater than the Cristy torso for six organs and 100% greater or more for four of these. By noting the effect of reducing the size of ADELAIDE’s organs to approximately the size of Cristy’s and assigning the extra tissue to soft tissue, it is possible

Handbook of Anatomical Models for Radiation Dosimetry

96

TABLE 4.3 Mass of Organs in the ADELAIDE Torso as a Percentage of the Torso’s Mass, Compared to the Values for the Cristy Torso

Ovaries Red marrow Colon wall Lung Stomach wall Bladder Breast Liver Esophagus Thyroid Skin Bone surface Remainder Totals

ADELAIDE

Cristy 15 Years Old

ADELAIDE/Cristy

0.046 3.7 1.03 4.59 0.54 0.08 1.27 6.53 0.13 0.076 4.30 7.33 (surface voxels) 45.0 74.6a

0.018 3.1 0.44 2.38 0.41 0.12 0.16 4.47 0.099 (thymus) 0.044 3.88 10.82 (skeleton) 71.3 97.2b

2.6 1.2 2.3 1.9 1.3 0.7 7.9 1.5 1.3 1.7 1.1 0.7 0.6

Note: Cristy torso mass corrected to exclude mass of arms. a Remaining 23.3% is heart, gall bladder contents, subcutaneous fat, trachea, spinal cord, contents of digestive system, remaining “internal bone,” and blood. b Remaining 2.7% is heart and gall bladder contents.

to quantify any possible overestimate. Hence if the marrow and internal bone is decreased by 1000 g, the size of the heart by 200 g, the lungs by 400 g, kidney by 100 g, liver by 300 g, spleen by 80 g, bowel by 100 g, colon by 200 g, a sum of 2380 g of extra soft tissue is “produced” (and the organs of ADELAIDE and the Cristy 15 years old have approximately equal mass). The revised total soft tissue in ADELAIDE is 2380 + 12,245 g = 14,625 g of the 23,663 g torso, or 62% of the torso mass. This is closer to the Cristy 15 year old trunk’s value of 67.5%. However, it is unlikely that the size of ADELAIDE’s organs have been overestimated by this amount during the segmentation process. The masses of the organs in the Cristy computational phantom were obtained using ICRP 23,15 which in turn uses data from the first half of the twentieth century. The Cristy organ masses are almost identical to those of the preceding 15 year old computational phantom of Jones et al.,16 which used organ mass data from the same sources. The major differences are that Cristy used masses that were about 8% larger for the components of the digestive system, 10% larger for the pancreas and masses that were 10% and 17% smaller for the liver and spleen, respectively. The body size of children in most Western countries has been increasing during the twentieth century. In Australia, NSW school children have increased in height by 8–9 cm between 1908 and 1971.13 A further height gain of about 1.5 cm has occurred in Perth school children (aged from 5 to 16 years) when compared to NSW children in the period 1970–1972 to 1984.13 When this increase in children’s stature is considered along with the period from which the MIRD computational phantoms draw their anatomical data, it seems likely that organ masses would have increased along with the increase in stature of children. Consequently, it is likely that the organ masses in the MIRD computational phantoms underestimate the typical organ masses of children living in the 1990s.

The ADELAIDE Teenage Female Voxel Computational Phantom

97

Some support for this likelihood is provided by the organ mass data of Ogiu et al. for Japanese people.14 Their autopsy data include the masses of eight different organs from six 14 year old females. While it is not known whether the mass of an organ taken from a cadaver is the same as the organ mass in the live individual, the average masses of six of the eight organs is greater than the masses used by Cristy (the other two are about the same mass). That this is the case even though the Cristy computational phantom is a year “older” and includes organs that are the average of male and female organ masses supports the argument that the Cristy organ masses underestimate the organ masses of contemporary 15 years old. A clear example of an organ important for radiological protection whose size, in proportion to body size, seems to be underestimated is the female breast (see Table 4.3). The Cristy computational phantom includes breasts each of size 22 cm3. The 14 year old patient from who ADELAIDE was constructed had breasts each of about 150 cm3. ADELAIDE’s breasts, as a proportion of body size, are much larger than those of the Cristy 15 years old. Another is the lungs. ADELAIDE has lungs that are 4.6% of the torso mass while the Cristy 15 year old lungs are 2.4% of the trunk’s mass (Table 4.3). The Cristy 15 year old computational phantom’s lungs are assigned a mass of 650 g, volume 2200 mL and density 0.295 g cm−3. The average lung mass of 16 individuals of 14 years of age in the Ogiu autopsies is 800 g, which is significantly larger. The ADELAIDE torso has a lung mass—determined from lung volume, 4174 mL, and the ICRU density of 0.26 g cm−3 for an adult inflated lung—of 1085 g. The lung volume represents a maximum value achieved after inspiration and while breath-holding during the CT examination. Hence, the lung mass value depends on the density used for lung tissue. We will consider an alternatively derived value for lung density. The Cristy adult lungs have a mass of 999 g, volume 3400 mL, and density 3.4 g cm−3. An alternative value for mass, taken from an Anatomy and Physiology text,17 is 1180 g for males. The same book gives total lung capacity (which includes the amount of air that can be forcefully inspired after inspiration of the normal tidal volume) as 5800 mL. Hence, a maximally inflated lung due to a breath-holding procedure during a CT examination may have a density of 0.2 g cm−3. Using this density with ADELAIDE’s lung volume results in a lung mass of 835 g. This mass is similar to the Ogiu et al. value (800 g). Organ mass alone may not have a great influence on average absorbed dose to organs as dose is measured in joules per kilogram. If absorbed dose is reasonably evenly distributed within adjacent tissues of similar densities, tissue surrounding an organ that is inadvertently assigned to that organ will not greatly alter the joules per kilogram imparted to the organ. A larger organ will occupy a larger volume and this may alter the pattern of shielding and exposure to x-rays experienced by nearby organs and tissues. Hence it is the disposition of organs within the trunk and the shielding afforded by surrounding tissues that is more likely to affect absorbed dose than mass alone. 4.2.5 The Shape and Locations of Organs in the Body To some extent, the internal organs change their shape and position relative to each other as a result of the movements associated with breathing, heartbeat, food digestion, and body posture. For example, using images from a patient lying supine on the patient table of a CT scanner to construct ADELAIDE, has some consequences for the resulting representation of anatomy when compared to that of a person standing. One is that the organs are acted upon differently by gravity and they recline toward the dorsal surface of the body. That this is the case is apparent from the images of the stomach and duodenum, where

98

Handbook of Anatomical Models for Radiation Dosimetry

the horizontal surface of the liquid chyme (stomach contents) is clearly visible below the gas it also contains. If ADELAIDE had been modeled from a person standing erect, the organs would be shifted anteriorly and inferiorly in their position. Another consequence of using a living subject is that gas is present in the digestive system, esophagus, and trachea. ADELAIDE has about 1 L of gas in these organs. Radiation will not suffer the attenuation in gas that it does in tissue. Hence, the tissues adjacent to pockets of gas will be exposed differently to radiation than if surrounded by tissue. NORMAN, the adult male voxel computational phantom of the NRPB,18 had regions on the dorsal surface of the back that had been flattened by lying supine during MRI acquisition and which had to be “corrected” during segmentation. In the case of ADELAIDE, the torso is required for the calculation of organ doses from CT examinations which are conducted on patients lying supine. Hence the flattening of the buttocks and back is a desirable realistic feature. Using mathematical equations to approximate complex volumes, as the MIRD computational phantoms do, results in compromises that cause the shape of mathematical organs to depart from their real shape. In real anatomy, the upper surface of the liver and stomach are domed and project superiorly so that for several centimeters they are surrounded by the lower lobes of the lungs. Mini et al., in a survey of the CT scans of five men and five women, found that the section of the trunk where lungs, stomach, and liver overlapped contained 30% of the volume of the lungs, and about 50% of the volumes of the stomach and liver.19 This means that a significant part of the liver and stomach may be in the x-ray beam during a lung examination and a significant part of the lungs will be in the x-ray beam during a liver examination. MIRD computational phantoms represent the lower surface of the lungs and upper surface of the liver as horizontal planes so that liver and lungs do not share the same transverse slice. Consequently, using a MIRD computational phantom for a calculation of absorbed dose during a chest examination to image the lungs means that the liver and stomach will be totally out of the direct beam. As a consequence, the dose to these organs will be underestimated by using a MIRD computational phantom. In real anatomy, the external boundary of the lungs and the intestines follow the curve of the body wall and fill all the space within it. In MIRD computational phantoms, the lungs and intestines do not extend outward to the body wall. This results in the MIRD small and large intestine being too deep with respect to the abdominal wall. These organs in a MIRD computational phantom will experience an attenuated x-ray beam when compared to a voxel computational phantom. In the human rib cage, the fi rst seven ribs are joined to the sternum by their costal cartilage, ribs 8–10 are joined to a common cartilage that is, in turn, attached to the sternum. Ribs 11 and 12 do not extend across the front of the abdomen. Consequently, the lungs and liver are not completely enclosed by the rib cage at the front of the body. MIRD computational phantoms do not include a sternum and represent all 12 ribs as horizontal bands encircling the trunk, which results in the lungs and diaphragm being partly obscured by bone. Staniszewska has noted these deficiencies and that an underestimation of absorbed dose to the lungs should result.20 She modified the Cristy computational phantom and compared absorbed doses to organs (in mGy) to the Cristy computational phantom and her modified computational phantom resulting from PA chest radiography. Her modified 10 years old received 35% less dose to the ribs, 160% more dose to the lungs, and 5% more dose to the liver. A large proportion of ADELAIDE’s skeleton is assigned to “surface bone voxels.” This is the result of some bones, such as the scapulae and vertebrae, being irregular in shape. In addition, the scapulae are very thin. This thinness and their spine and coracoid and

The ADELAIDE Teenage Female Voxel Computational Phantom

99

acromion processes (bony projections) result in a large part of the scapulae being classified as “surface bone voxels.” Similarly, vertebrae have transverse and spinous processes and consequently a large surface area. 4.2.6 The Location of Organs within the Body For the purposes of calculating average absorbed dose to organs, there are some considerations that are more important than the size and shape of organs. One is the position of the organs relative to the x-ray beam. That is, whether the distribution of organs within a computational phantom places the organs wholly within the beam or not for a particular examination. A second consideration is the effect that the shielding provided by surrounding tissue has on the photon flux and spectrum incident on an organ. An organ directly irradiated by an x-ray beam will be imparted more energy than an organ outside the beam and receiving only scattered radiation. Hence, as discussed above, a real patient undergoing a CT lung examination will have part of their liver directly irradiated and one undergoing a liver examination will have the lower parts of their lungs irradiated. This is not necessarily the case with a MIRD computational phantom. Consequently, average absorbed dose to liver from a lung CT examination may be actually greater than is estimated by a simulation using a MIRD computational phantom. Similarly, average absorbed dose to lungs from a liver CT examination may be actually greater than is estimated by a MIRD simulation. The intestines of a MIRD computational phantom are quite compact and do not overlap into the transverse slices that contain the liver and stomach. Nor do they extend downward far enough to surround the bladder and uterus. In ADELAIDE, the intestines extend further in all directions than do the MIRD representation of intestines. Consequently, in the voxel computational phantom, the x-ray beam is more likely to be intercepted by the intestines. The ADELAIDE computational phantom has intestines overlain by about 2 cm of muscle and soft tissue whereas the MIRD computational phantom may have 5 cm of overlying tissue. Hence, the intestines of the latter are shielded to a greater extent than they should be. This shielding effect has been noticed by Yamaguchi in his report on the age dependence of effective doses for external photons incident on MIRD computational phantoms. He reached the following conclusion: “Since the thickness of the shielding tissues increases generally with body size, organs in older bodies receive correspondingly less equivalent dose.”21 and this conclusion is reiterated on page 44 of ICRP 74.12 The presently observed age dependence of effective doses may alter as the greater availability of pediatric voxel computational phantoms with more realistic amounts of overlying tissue is used to calculate effective dose. The presence of arms in the MIRD computational phantom and the construction of its rib cage further shield the MIRD organs in comparison to the anatomy represented by ADELAIDE. It is possible to remove the arms from the MIRD computational phantom as has been done for some radiography examinations considered in NRPB R-186.22 The MIRD computational phantom modified in this way would have been a more suitable one for CT examinations where imaging of the arms is not required. 4.2.7 Conversion of Segmented Anatomy Images to an Input File for the EGS4 Monte Carlo Code In order for the computational phantom ADELAIDE to be “EGS4 ready,” it must be included in an input file which has a transparent standard form and is accessible to users of EGS4. Our EGS4 input file (Adele.inp) is of similar format to those for the user-codes XYZDOS.

100

Handbook of Anatomical Models for Radiation Dosimetry

MOR and INHOMP.MOR which are included in the EGS4 code system distribution. In particular, the anatomy data for each pixel of an image is input row by row in the form: IL, IU, JL, JU, KL, KU, MEDIUM, DENSITY where IL and IU are the lower and upper I indices for a group of adjacent voxels all containing the same medium JL and JU are the J indices for those voxels and KL and KU are the K indices for the same voxels MEDIUM is the EGS4 medium index for the medium contained in the voxels (see Table 4.2) and DENSITY is the density of the medium (a value of 0.0 for density means that the default density as recorded in the PEGS4 data file is used). Each time the medium changes in adjacent voxels, a new line in the input file is required to define the medium in the next group of voxels. Hence, the number of data lines in the input file that define the anatomy is very large. It is essential that the medium index that appears in each line be the true value for the organ or tissue of interest. Ensuring that this is so is a time-consuming task due to the range of gray scale values that can occur within the one organ and to the overlapping ranges of gray scale values in different organs. How this was done is described below. Using the image processing software Image-Pro Plus, each segmented CT slice may be viewed as an image or as a bitmap (a two-dimensional array of pixels with their assigned gray scale values). The bitmap array of each image was exported to Microsoft Excel as a bitmap using the “dynamic data exchange” facility in Image-Pro Plus and saved as an Excel file. The two-dimensional arrays of assigned gray scale numbers in the Excel worksheets were extended if the array came from an image that had truncated shoulders and hips. This was done by adding extra columns to the worksheet and filling the cells in the extra columns with the appropriate EGS4 medium index for skin, subcutaneous fat, soft tissue, or bone as required. The gray scale values of all organs were checked to ensure that all pixels assigned to that organ had the same value. This was necessary as it is possible for a liver pixel (say) to have a gray scale value that differs from its assigned (after segmentation) value of 15 by an amount small enough to be indistinguishable by the naked eye from a pixel with value 15 when the image is viewed on a computer screen. The gray scale values of pixels in muscle or soft tissue located outside of the organs were searched to ensure that their value was not the same as that assigned to one of the organs. The Excel files after checking were saved as comma-delimited (.CSV) files. These 54.CSV files (one per K index), in turn, became the input for a FORTRAN program, “READDATA. MOR”—written by the first author (MC) for the purpose—which converted the organspecific gray scale numbers into the media indices (consecutive numbers as required by EGS4 ) and formatted the numbers one J index at a time into the form described above, that is IL, IU, JL, JU, KL, KU, MEDIUM, DENSITY This form is suitable for inclusion as a line in the input file for our EGS4 user-code. For example, nine lines from the input file “adele.inp” appear below.

The ADELAIDE Teenage Female Voxel Computational Phantom

101

39, 41, 93, 93, 56, 56, 5,0.0 42, 44, 93, 93, 56, 56, 6,0.0 45, 76, 93, 93, 56, 56, 11,0.0 77, 83, 93, 93, 56, 56, 6,0.0 84, 117, 93, 93, 56, 56, 5,0.0 118, 135, 93, 93, 56, 56, 6,0.0 136, 156, 93, 93, 56, 56, 11,0.0 157, 159, 93, 93, 56, 56, 6,0.0 160, 161, 93, 93, 56, 56, 5,0.0 They describe the voxels from slice 56 (KL = KU = 56), J index 93 (JL = JU = 93) and voxels with I indices between 39 and 161. The first line assigns three voxels (JL = 39, JU = 41) to skin (medium index = 5), the next line defines three voxels (JL = 42, JU = 44) to contain subcutaneous fat (medium index = 6), the third line defines the next 32 voxels (JL = 45, JU = 76) to contain muscle/soft tissue (medium index = 11), and so on. Voxels with I indices from 1 to 38 and from 162 to 200 are outside the computational phantom and so contain air. Each line ends with 0.0 indicating that the density of the medium is the same as that in the PEGS data file, that is, the default value. The resulting 54 fi les of formatted numbers were concatenated to form the single input fi le (of size 1.8 MB), which characterizes the anatomy of our computational phantom ADELAIDE.

4.3 Organ Dose Calculations with ADELAIDE and EGS4 The effective dose for CT examinations of the ADELAIDE computational phantom has been reported.11,23 The actual dose was calculated for a GE HiSpeed Advantage helical scanner and is specific to that scanner. Furthermore, the actual dose depends on the amount of anatomy imaged, the exposure factors used and to a lesser extent on the computational phantom used to calculate the scanner’s x-ray spectrum.24 Using tissue weighting factors from ICRP 60,25 the effective dose from a CT examination of the chest spanning 26 cm was 2.32 ± 0.05 mSv/100 mAs and for an abdomen examination spanning 24 cm was 2.16 ± 0.05 mSv/100 mAs. ADELAIDE is close to the average stature of a 13 year old female. By increasing (or decreasing) the size of the voxels by 5% (in each dimension), it is possible to extrapolate ADELAIDE to the average stature for a 16 (or 11–12) years old. When this is done, the effective dose to the 95% scaled ADELAIDE for CT examinations that spanned the equivalent anatomy was 8% higher for the chest examination and 4% higher for the abdomen examination.23 For the 105% scaled ADELAIDE, the dose for the chest examination was the same while the dose due to the abdomen examination was 5% less.

102

Handbook of Anatomical Models for Radiation Dosimetry

4.4 Limitations of the Computational Phantom The major limitation of ADELAIDE is that the anatomy modeled is not the complete anatomy of an individual—the arms, legs, and head are not present. Nevertheless, the computational phantom is suited to CT dose calculations but less suited to other applications. Another limitation is introduced by the size of the voxels (2.53 × 2.53 × 10 mm). For example, thin tissues such as the scapula and the skin may be modeled as larger (thicker) than they actually are. The effect on absorbed dose is probably small as the unit of absorbed dose is J kg−1 rather than joules. This supposition is reasonable provided that the shielding effect of (and on) surrounding tissues is not greatly affected. Bone marrow was not identified as active or inactive and was not distinguished from bone. That is, “bone” was modeled as a uniform mixture of 40% (by mass) of cortical bone, 39% red marrow, and 21% yellow marrow with a consequent density of 1.34 g cm−3. The adrenal glands are included in the volume of each kidney. While gas within the gut was segmented, the contents of the gut were not distinguished from the walls. The stomach was estimated to be 38% of the total stomach and contents (see Table 4.2), the bowel wall was 49% of the mass identified as bowel and contents, while the colon was estimated at 54% of the total combined colon wall and contents.

References 1. Nelson, W.R., Hirayama, H., and Rogers, D.W.O. The EGS4 Code System, SLAC-265, Stanford, CA: Stanford Linear Accelerator Centre, Stanford University, 1985. 2. Cristy, M. Mathematical Phantoms Representing Children of Various Ages for Use in Estimates of Internal Dose, U.S. Nuclear Regulatory Commission Rep. NUREG/CR-1159, Oak Ridge, TN: Oak Ridge National Laboratory Rep. ORNL/NUREG/TM-367, 1980. 3. Zankl, M. et al. The construction of computer tomographic phantoms and their application in radiology and radiation protection, Radiation and Environmental Biophysics, 27, 153, 1988. 4. Shrimpton, P.C., Jones, D.G.F., Hillier, M.C., Wall, B.F., Le Heron J.C., and Faulkner K. Survey of CT Practice in the UK Part 2: Dosimetric Aspects, NRPB-R249, Chilton, Didcot: National Radiological Protection Board, 1991. 5. Hart, D., Jones, D.G., and Wall, B.F. Coefficients for Estimating Effective Doses from Paediatric X-Ray Examinations, NRPB-R279, Chilton, Didcot: National Radiological Protection Board, 1996. 6. Caon, M., Bibbo, G., and Pattison, J. Running the EGS4 Monte Carlo code with Fortran 90 on a Pentium computer, Australasian Physical and Engineering Sciences in Medicine, 19, 201, 1996. 7. Caon, M. Voxel-based computational models of real human anatomy: A review, Radiation and Environmental Biophysics, 42, 229, 2003. 8. Caon, M. and Mohyla, J. Automating the segmentation of medical images for the production of voxel tomographic computational models, Australasian Physical and Engineering Sciences in Medicine, 24, 166, 2001. 9. Lee, C. et al. Hybrid computational phantoms of the male and female newborn patient: NURBSbased whole-body models, Physics in Medicine and Biology, 52, 3309, 2007. 10. Zankl, M. and Wittmann, A. The adult male voxel model “Golem” segmented from wholebody CT patient data, Radiation and Environmental Biophysics, 40, 153, 2001. 11. Caon, M., Bibbo, G., and Pattison, J. An EGS4-ready tomographic computational model of a 14-year-old female torso for calculating organ doses from CT examinations, Physics in Medicine and Biology, 44, 2213, 1999.

The ADELAIDE Teenage Female Voxel Computational Phantom

103

12. ICRP. Conversion Coefficients for Use in Radiation Protection against External Radiation. Publication 74, Annals of the ICRP, 26, 28, 1996. 13. Hitchcock, N.E., Maller, R.A., and Gilmour, A.I. Body size of young Australians aged five to 16 years, The Medical Journal of Australia, 145, 368, 1986. 14. Ogiu, N., Nakamura, Y., Ijiri, I., Hiraiwa, K., and Ogiu, T. A statistical analysis of the internal organ weights of normal Japanese people, Health Physics, 72, 368, 1997. 15. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Oxford: Pergamon Press, 1975. 16. Jones, R.M., Poston, J.W., Hwang, J.M.L., Jones, T.D., and Warner, G.G. The Development and Use of Fifteen Year-Old Equivalent Mathematical Phantom for Internal Dose Calculations, ORNL/ TM-5278, Oak Ridge Tennessee: Oak Ridge National Laboratory, 1976. 17. Seely, R., Stephens, T., and Tate, P. Anatomy and Physiology, 4th edn., Boston: WCB/McGrawHill, 1998. 18. Dimbylow, P.J. FDTD calculations of the whole-body averaged SAR in an anatomically realistic voxel model of the human body from 1 MHz to 1 GHz, Physics in Medicine and Biology, 42, 479, 1997. 19. Mini, R., Vock, P., Mury, R., and Schneeberger, P. Radiation exposure of patients who undergo CT of the trunk, Radiology, 195, 557, 1995. 20. Staniszewska, M.A. A modification to Cristy’s mathematical human phantoms for Monte Carlo simulations, Journal of Radiological Protection, 12, 85, 1992. 21. Yamaguchi, Y. Age dependent effective doses for external photons, Radiation Protection Dosimetry, 55, 123, 1994. 22. Jones, D.G. and Wall, B.F. Organ Doses from Medical X-Ray Examinations Calculated Using Monte Carlo Techniques, Report NRPB-R186, Chilton, Didcot: National Radiological Protection Board, 1985. 23. Caon, M., Bibbo, G., and Pattison, J. Monte Carlo calculated effective dose to teenage girls from computed tomography examinations, Radiation Protection Dosimetry, 90, 445, 2000. 24. Caon, M., Bibbo, G., Pattison, J., and Bhat, M. The effect on dose to computed tomography phantoms of varying the theoretical x-ray spectrum: A comparison of four diagnostic x-ray spectrum calculating codes, Medical Physics, 25, 1021, 1998. 25. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Oxford: Pergamon Press, 1991.

5 The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms W. Paul Segars and Benjamin M.W. Tsui

CONTENTS 5.1 Introduction ............................................................................................................... 105 5.2 The 4D Mathematical Cardiac-Torso Phantom ..................................................... 107 5.3 The 4D NURBS-Based Cardiac-Torso Phantom ................................................... 114 5.4 The 4D Extended Cardiac-Torso Phantom ............................................................ 121 5.5 The 4D Mouse Whole-Body Phantom.................................................................... 126 5.6 Summary .................................................................................................................... 132 References ............................................................................................................................. 132

5.1 Introduction In addition to the field of radiation dosimetry, computerized phantoms are also used in simulation studies to evaluate and improve medical imaging devices and techniques. Medical imaging simulation consists of virtual experiments that can be carried out entirely on the computer. Computer-generated phantoms serve as the subjects, while researchers use computer-based models of the imaging process to simulate predictive imaging data from them, as seen in Figure 5.1. A major advantage to using computer-generated phantoms in simulation studies is that the exact anatomy and physiological functions of the phantom are known, thus providing a standard form from which to quantitatively evaluate and improve medical imaging devices, image processing, and reconstruction techniques. Another advantage of using phantoms is that they can be easily altered to model different anatomies and medical situations, providing a large population of subjects with which to perform research. A vital aspect of a simulation is to have a realistic computational phantom of the subject’s anatomy. Without this, the results of the simulation may not be indicative of what would occur in actual patients or animal subjects, and would, therefore, have limited practical value. Previous chapters have highlighted the two main categories of phantoms and the advantages and disadvantages of each. Briefly, existing computational phantoms involve a trade-off between realism and flexibility, and this affects their applicability to imaging simulations. Since they are based on patient data, voxel-based phantoms are realistic, but they remain fixed to a particular anatomy and resolution. Studies of the effects of anatomical variations or motion on medical imaging can be limited, and the generation of 105

Handbook of Anatomical Models for Radiation Dosimetry

106

Computer phantom Model of imaging process

Simulated medical image

FIGURE 5.1 Computer-based medical imaging simulation. In this example, a chest x-ray is being generated from the computerized phantom.

XCAT

Evolution of computerized phantoms...

MOBY

MIRD

MCAT

NCAT

FIGURE 5.2 (See color insert following page 524.) Original MIRD phantom1 and phantoms developed in our laboratory that approach more ideal computerized phantoms.

the phantom at other resolutions requires interpolation, which introduces error. Stylized or mathematical phantoms, conversely, are defined mathematically, in order to allow for anatomical variation and generation at multiple resolutions. The simplicity of the mathematical equations, however, limits an exact modeling of the organ shapes. In order to get a more realistic phantom, the sets of equations defining the phantom have to become more complex; but with this increase in complexity comes a decrease in flexibility. Current work in phantom development has focused on the development of “hybrid” phantoms that seek to combine the realism of a patient-based voxelized phantom with the flexibility of a mathematical phantom. We have been leading the development of realistic and flexible digital phantoms for use in medical imaging research. Figure 5.2 shows the evolution of computerized phantoms toward more ideal hybrid computational phantoms and summarizes our work in this evolution with the development of the mathematical cardiac-torso (MCAT), NURBS-based cardiac-torso (NCAT), extended cardiac-torso (XCAT), and mouse wholebody (MOBY) phantoms.

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

107

5.2 The 4D Mathematical Cardiac-Torso Phantom The four-dimensional (4D) MCAT phantom, as seen in Figure 5.3, is a stylized anthropomorphic phantom that was developed in our laboratory for use in nuclear medicine imaging research, specifically for single-photon emission computed tomography (SPECT) and positron emission tomography (PET). The anatomy of the 4D MCAT phantom was constructed using simple geometric primitives based on the MIRD-5 computational phantom.1 We used overlap, cut planes, and intersections of the geometric objects, however, to form a more realistic anatomy. Using these mathematical formula, we modeled the size, shape, and configurations of the major thoracic structures and organs for imaging purposes. Table 5.1 shows a list of the organs modeled in the 4D MCAT phantom and their corresponding volumes. For the purpose of imaging simulation, the MCAT phantom is capable of simulating two physical models: a three-dimensional (3D) distribution of attenuation coefficients for a given photon energy and a 3D distribution of emission radionuclide activity for the various organs. Each of these computational phantoms is stored as a voxelized phantom of

Anterior

Posterior

Right lateral

FIGURE 5.3 3D surface renderings of the 4D MCAT phantom. Anterior, posterior, and right lateral views are shown.

TABLE 5.1 List of Organs Modeled in the MCAT Model and Their Corresponding Volume Organ

Volume (mL)

Liver

1825

Stomach (wall and contents)

367

Kidneys (2)

285

Spleen

175

Right lung

2216

Left lung

2150

Heart (blood and tissue)

640

Handbook of Anatomical Models for Radiation Dosimetry

108

any user-defined resolution. Since the phantom is mathematically defined, there are no errors associated with generating the phantom at different resolutions. These voxelized representations can be used in combination with analytical or Monte Carlo-based models of the imaging process to simulate transmission (x-ray, x-ray CT) and emission (SPECT, PET) imaging data. The 3D distribution of attenuation coefficients classifies the thoracic tissues into one of four tissue types: muscle (soft tissue), lung, spinal bone (less dense), and rib bone (more dense). To generate the attenuation coefficient or the transmission phantom, the various organs are set with their individual attenuation coefficients, a user-defined photon energy. We can calculate the attenuation coefficients for the different tissue types from the physical densities and the mass attenuation coefficients of the tissues. The mass attenuation coefficients of the tissues, at any energy from 1 to 1000 keV, are calculated from the elemental compositions of the tissues2 and the energy-dependent mass attenuation coefficients for the elements.3 Figure 5.4 illustrates the use of the 4D MCAT as an attenuation coefficient phantom for the 72 keV radionuclide Thallium-201. Projection images, similar to those acquired from a patient during transmission imaging, can be simulated from the voxelized attenuation coefficient phantom using a computational phantom of the projection process. The 3D radionuclide distribution models the uptake of a radiopharmaceutical in the various organs. To generate a radionuclide uptake or emission phantom, the intensity values of the organs are set to their individual uptake ratios for the desired radiopharmaceutical. Figure 5.5 shows the use of the 4D MCAT as a radiopharmaceutical uptake phantom for Thallium-201. The projection images simulated from the uptake phantom emulate those that would be acquired during an emission-imaging scan. The projection images reflect the uptake for the particular radiopharmaceutical in the organs. Areas of higher uptake are indicated by the brighter intensities, whereas areas of lower uptake are indicated by darker intensities. In order to study the effects of heart motion on cardiac SPECT and PET imaging, the 4D MCAT phantom includes a beating heart computational phantom.4 The cardiac computational phantom in the 4D MCAT is based on ellipsoids and simulates the changes in chamber volume, left ventricular wall thickness, and heart rotation that occur throughout

Transaxial slices for 72 keV radionuclide (T1-201) A

B

C

D

Transmission projection (chest x-ray)

D A 4D MCAT

Human

FIGURE 5.4 4D MCAT used as an attenuation coefficient or transmission phantom. Left: transaxial slices are shown for a phantom simulated with each organ set to its individual attenuation coefficient for Thallium-201. Right: comparison of a transmission projection of the MCAT phantom with an actual patient chest x-ray.

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

109

Transaxial slices with typical T1-201 uptake ratios Emission projections A

B

D

C

D

A Left

Anterior

FIGURE 5.5 4D MCAT used as a radiopharmaceutical uptake or emission phantom. Left: transaxial slices of the MCAT with organs set to typical uptake ratios for Thallium-201. Right: emission projections of the uptake phantom with the effects of attenuation. The increased uptake of the radiopharmaceutical by the liver and heart myocardium can be seen in the projections.

LV

RV

Heart: union of four partial ellipsoids

Z

LVext.

Z LVint.

X

X

LA RVint. RA

RVext. End diastole

End systole

End systole with end diastole overlay

FIGURE 5.6 Left: 4D MCAT cardiac model based on ellipsoids. The ventricles and atria are each defi ned by two ellipsoids, one for the inner and one for the outer boundary. Right: single LA slices of the beating heart during end-diastole and end-systole. End-systolic frame is shown with an end-diastolic overlay to illustrate the change in the LV during the cardiac cycle.

the cardiac cycle as seen in Figure 5.6. The beating heart motion in a normal human consists of a wringing-like twisting motion of the left ventricle (LV) and radial and longitudinal contraction of the heart walls.5–7 The changes and motion of the beating heart are simulated in the 4D MCAT by altering the parameters that define the ellipsoid models. A simplified version of the twisting motion of the heart is modeled by changing the rotation of the ellipsoids about the long-axis (LA) of the LV. The base of the ventricles is set to rotate clockwise, while the apex is set to rotate counterclockwise. The rotation is done as a pixel-by-pixel operation and is controlled by a parameter in the MCAT program that defines the maximum rotation. Altering the diameters of the axes that define the different ellipsoids in the phantom simulates the radial contraction and thickening of the heart walls during systole. We modeled this longitudinal contraction of the ventricles by changing the

Handbook of Anatomical Models for Radiation Dosimetry

110

Reference MCAT

120

LV volume (mL)

100

80

60

Systole

40 1

Diastole 6

11 Frame number

16

FIGURE 5.7 Volume curve for the LV of the 4D MCAT phantom as compared to that of an average male. (Adapted from Guyton, A. and Hall, J. Textbook of Medical Physiology, 9th edn., WB Saunders Co., Philadelphia, PA, 1996.)

valve plane between the atria and ventricles. The ellipsoids were altered so that the heart mass and volume remain constant throughout the cardiac cycle and equal to that of an average male,8 as seen in Figure 5.7. The heart computational phantom also includes the ability to model perfusion defects, such as those that would be seen in myocardial SPECT (Figure 5.8). Defects arise due to blocks in the coronary vessels that limit or cutoff bloodflow to the downstream myocardial tissue. These defects can be seen in myocardial SPECT as areas of low perfusion (low uptake of the radionuclide) in the left ventricular myocardium. Perfusion is indicative of the health of the tissue. Defects in the 4D MCAT are modeled as pie-shaped wedges in the wall of the LV by three parameters: circumferential width, LA width, and defect center.9,10 In producing the radionuclide uptake or emission phantom, the portion of the LV wall defined as the defect is given a lower uptake ratio than the rest of the LV myocardium. Figure 5.8 shows myocardial SPECT simulations with lesions located in three different locations within the LV. To investigate the effects of respiratory motion on SPECT and PET imaging, the 4D MCAT also includes a model for the breathing motion based on known respiratory mechanics.11 Respiratory motion involves the movement of the diaphragm, heart, thoracic cage, and lungs. During inspiration, the diaphragm contracts forcing the abdominal organs down and to the front of the body, expanding the volume of the chest. At the same time, the ribs rotate about an axis through their costal necks, moving outward and upward to further expand the chest volume. The lungs inflate due to the change in internal pressure that results from the increased chest volume.12 In order to simulate respiratory motion in

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

Short axis Base

111

Long axis

Circumferential width

Long-axis width Defect center

Apex FIGURE 5.8 (Left) Myocardial lesion in the 4D MCAT heart.9 The lesion is defined as a pie-shaped wedge in the left ventricular wall by the parameters: circumferential width, LA width, and defect center. (Right) Simulated myocardial SPECT imaging data showing three different locations for a lesion in the left ventricular wall. The lesions can be seen in the SA slices taken from above the LV and the LA slices taken along side of the LV. Lesions appear as areas of low perfusion (low intensity).

Heart

Liver

FIGURE 5.9 Parameters for the height of the diaphragm are altered to translate the diaphragm upward and downward in the geometry-based MCAT model. (Adapted from Segars, W.P., Lalush, D.S., and Tsui, B.M.W., IEEE Trans. Nucl. Sci., 48(1), 89, 2001.)

the MCAT phantom, we altered the geometric solids for the diaphragm, heart, ribs, and lungs through the manipulation of parameters defining them. To do this, we simulated the movement of the diaphragm during respiration by altering the parameters that define the height of the left and right diaphragm sections (Figure 5.9). The heart, liver, stomach, spleen, and kidneys were rigidly translated with the motion of the diaphragm. Tiled cut planes through the cylinder define the positions of the ribs in the MCAT (Figure 5.10a). This definition does not allow for the ribs to be rotated about the axis through their costal necks. Altering the tilt angle θ of the ribs and modifying the length of the rib short axis parameter allows us to approximate the rib length (Figure 5.10b). The RLA parameter was not adjusted. Figure 5.11 shows coronal cut slices of the 4D MCAT at end-inspiration and end-expiration. Two parameters, the height of the diaphragm and the anterior–posterior (AP) expansion of the chest, control the respiratory motion in the MCAT phantom. The height of the diaphragm controls the longitudinal motions of the liver, stomach, spleen, and heart while the AP

Handbook of Anatomical Models for Radiation Dosimetry

112

θ΄ Rib of length L

ΔHeight

θ

(a)

RSA

(b)

ΔRSA

Volume change (mL)

FIGURE 5.10 (a) Lateral view of the MCAT phantom showing the ribs defi ned as tilted cut planes through a cylinder. (b) Lateral view of the ribcage cylinder showing anteroposterior diameter change by altering the tilt angle and rib short axis RSA parameters. (Adapted from Segars, W.P., Lalush, D.S., and Tsui, B.M.W., IEEE Trans. Nucl. Sci., 48(1), 89, 2001.)

500

Expiration

Inspiration

250

0 0

1

2

3

4

5

Time (s)

3D anterior view

2D coronal slice

FIGURE 5.11 Top: normal respiratory curve. (Adapted from West, J. Respiratory Physiology, 5th edn., Williams & Wilkins, Baltimore, 1995.) Middle and bottom: 3D and 2D views of the MCAT at end-inspiration (left) and end-expiration (right).

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

113

expansion of the chest controls the lateral and AP motions of the organs as well as the motion of the ribcage. These time varying parameters were chosen to fit a volume curve for normal respiration12 (Figure 5.11). Time curves were derived for both the diaphragm motion and the AP expansion of the chest. The respiratory and cardiac models of the MCAT are parameterized so that a user can alter the magnitude or rates of each motion to simulate many different variations (normal and abnormal). In addition to motion, the MCAT phantom has the ability to model male and female anatomical variations. Adding breast extensions onto the male chest anatomy simulates female subjects. Variations in the male and female anatomy are generated by altering the parameters that define the different geometric shapes that compose the phantom. Figure 5.12 shows examples that demonstrate the ability of the phantom to vary patient anatomy. In each case, the phantom is altered to match the anatomy of the patient as determined by the patient’s PET scan.9,10 With this ability to modify the anatomy, the MCAT program can be used to simulate a patient population involved in patient studies. The 4D MCAT phantom provides a step up from the typical stylized phantom by providing a better representation of the anatomy while maintaining the flexibility to model anatomical variations and patient motion. With its capabilities, the 4D MCAT has been applied to many studies in emission imaging that seek to improve the quality of medical images. It has been used to research new image acquisition strategies and reconstruction algorithms, and to investigate the effects of physical factors, anatomy, and motion on medical images and to develop compensation methods for these effects. The 4D MCAT represents an advance in computerized modeling, but due to its geometrical design, is still lacking in terms of its level of realism that limits its applicability to higher resolution imaging techniques.

MCAT male population #92

#109

#117

MCAT female population #16

#43

PET

MCAT

FIGURE 5.12 MCAT phantoms created with varying anatomies. In each case, the MCAT phantom was altered through the parameters that defi ne the different structures to match the anatomy of the patient as determined by the patient’s PET scan. (Adapted from LaCroix, K.J. Evaluation of an attenuation compensation method with respect to defect detection in Tc-99m-MIBI myocardial SPECT images, PhD dissertation, The University of North Carolina at Chapel Hill, 1997; LaCroix, K.J. et al., J. Nucl. Med., 41, 502, 2000.)

Handbook of Anatomical Models for Radiation Dosimetry

114

5.3 The 4D NURBS-Based Cardiac-Torso Phantom In order to create a more ideal computational phantom, a primitive is needed that will allow more realistic modeling of human anatomy than is possible with geometrical solids while maintaining the ability to model anatomical variations and patient motion. An appropriate primitive that will permit the kind of modeling needed is the nonuniform rational B-spline (NURBS).13 NURBS are widely used in computer graphics and geometrical modeling to describe complex 3D surfaces. NURBS surfaces offer many advantages in our 4D mathematical modeling of the human body. Because it is a continuous surface, a NURBS permits representation of a surface at any resolution. Thus, it is possible to generate the phantom at any spatial resolution without introducing any errors that might result from interpolation. The principal disadvantage of the NURBS, compared to 3D geometric shapes like those used in the 3D MCAT, is that a NURBS requires more parameters to describe a surface. A 3D NURBS surface is defined by an n × m matrix of control points, two knot vectors (one for latitude and one for longitude), and a set of basis functions whereas a simple mathematical relationship is all that is required for geometric surfaces. Another disadvantage of NURBS is the mathematical complexities they introduce. More complicated algorithms are required to calculate intersections, produce solids, and to handle shared surfaces. These drawbacks are hardly consequential anymore, though, given the vast increase in computer processing power and memory. NURBS can therefore provide a powerful tool for computational phantom development. In terms of phantom representations, NURBS surfaces can be based on specific patient image data. By fitting NURBS to actual patient data, the phantom is more realistic than those based on solid geometry or simple mathematical relationships. Also, NURBS can be altered easily via affine and other transformations to model variations in anatomy among patients and patient motions. The shape of a NURBS surface can be modified through the control points that define the surface (Figure 5.13). The control points form a convex hull around the NURBS surface and determine its shape. By applying transformations to the control points, the shape of the surface can be modified or sculpted as if it were made of clay.13 To perform each transformation, one merely needs to multiply the control points of the surface to be altered by the appropriate transformation matrix. With its flexibility and its ability to realistically model anatomy, NURBS is an excellent basis for a realistic and flexible computerized phantom.

V

z

y

z x

U

y x

FIGURE 5.13 Modification of the shape of a NURBS surface through its control points. Modification to the shape of the surface is done by manipulating its control points. The shaded control points are translated upward altering the shape of the surface.

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

115

FIGURE 5.14 Anterior and posterior views of the 4D NCAT phantom.

Lungs Body Stomach Liver

Skeleton

Kidneys

Spleen

FIGURE 5.15 Segmentation of the torso organs using the SURFdriver surface reconstruction program. CT slices are contoured (left) then reconstructed into smooth polygon models (right).

With NURBS as its basis, the 4D NCAT phantom was developed as the next generation MCAT phantom14 (Figure 5.14). The NURBS surfaces in the torso were constructed based on the Visible Human Male CT data set.15 The CT data consisted of axial CT scans of the entire body taken at 1 mm intervals at a resolution of 512 pixels × 512 pixels (pixel size of 0.898 mm per pixel). The structures in the torso were manually segmented using SURFdriver16 to display the CT images as shown in Figure 5.15. The contours defined in the segmentation process for each organ models were reconstructed into smooth polygon models using SURFdriver’s rendering engine and smoothing function. The software program Rhinoceros17 was used to fit smooth, cubic NURBS surfaces to contours taken from the polygon models created by SURFdriver (Figure 5.16). Figure 5.14 shows the resulting torso anatomy of the NCAT. Since it was based on imaging data, the anatomy of the NCAT is more realistic than that of the MCAT. Table 5.2 lists the organs modeled in the 4D NCAT and their volumes. Like the 4D MCAT, the NCAT phantom was extended into four dimensions to model common patient motions such as the cardiac and respiratory motions. The beating heart

Handbook of Anatomical Models for Radiation Dosimetry

116

Lungs Body Liver

Skeleton

Kidneys

Stomach

Spleen

FIGURE 5.16 Generation of 3D NURBS models for the torso organs using the Rhinoceros NURBS program. Contours of the polygon models are lofted into NURBS surfaces (right). Contours obtained from the right kidney model are shown as example. The contours and their spacing are chosen to provide the best fit for the NURBS surface to the polygon model.

TABLE 5.2 List of Organs Modeled in the NCAT Phantom and Their Corresponding Volumes Organ Liver Gall bladder Stomach (wall and contents) Kidneys (2) Spleen Right lung Left lung Heart (blood and tissue)

Volume (mL) 1870 31 350 334 227 1900 1800 650

computational phantom of the 4D NCAT was based on 4D tagged magnetic resonance imaging (MRI) data obtained from Dr. Cengiz Ozturk of Johns Hopkins University, and Dr. Elliot McVeigh of the NIH and Johns Hopkins University, as seen in Figure 5.17. Three sets of tagged MR images from a normal subject were obtained and used to analyze the heart motion. The sets of data spanned the LV and were acquired for 26 time frames over the cardiac cycle. The data included two sets of parallel, short-axis (SA) images and one set of LA images. From the motion of the tag lines in the data, the full 3D motion of the heart over the cardiac cycle was analyzed and used to create time-dependent 3D NURBS surfaces for each of the four chambers of the heart. A 4D NURBS surface was then fit to the 3D surfaces creating a time-continuous 4D NURBS cardiac computational phantom (Figure 5.18) that was incorporated into the 3D anatomy of the NCAT phantom using the outer

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

117

SA image set 1 (SA1)

SA image set 2 (SA2)

LA image set 2 (LA1) FIGURE 5.17 Tagged MR data used as the basis for the NCAT beating heart. The data consisted of two sets of orthogonal SA images and one set of LA images. For each set, one image is shown at three different time frames from enddiastole (left) to end-systole (right). Tag lines begin as parallel stripes then deform with the movement of the heart tissue.

surface of the Visible Human heart as a guide. Since it was based on tagged MRI data, the NCAT heart illustrates the realistic contracting and twisting motion of the normal heart. With the flexibility of the NURBS surfaces, the NCAT heart was parameterized so that it could model a wide variety of beating heart motions, normal and abnormal. Many different parameters can be altered, such as ejection fraction, longitudinal and radial contraction, cardiac twist, heart rate, etc. Perfusion defects can be simulated in a similar manner to those of the MCAT (Figure 5.18). Global or regional cardiac motion abnormalities that result from these defects can also be simulated. Given the definition of an abnormal region, the control points that define the region in the heart are analyzed to characterize their normal motion throughout the cardiac cycle. The normal motion is parameterized in terms of the average regional radial (wall thickening) and longitudinal contractions and for the cardiac twist. These

Handbook of Anatomical Models for Radiation Dosimetry

118

End-diastole

RA

LA

RV

LV

End-systole

160 RV

140

Volume (mL)

120 100

LV RA

80

LA

60 40 20 0 0

0.2

0.4

0.6

0.8

1

Time (s) FIGURE 5.18 Cardiac model of the 4D NCAT. Plots of the volume change in the chambers are shown at the bottom.

parameters can be scaled to alter the motion of the selected region. The abnormal motion is blended in smoothly through transitional regions with that of the normal myocardium. With its realism and flexibility, the 4D NCAT cardiac computational phantom provides a useful tool in the study of cardiac imaging and the effects of cardiac motion in medical images. Respiration was modeled in the 4D NCAT based on a set of respiratory-gated CT data from the University of Iowa. The data were taken of a normal volunteer at 5%, 40%, 75%, and 100% of the total lung capacity (TLC). By identifying landmark points on and within the respiratory structures, and tracking their positions between the time frames, a general motion computational phantom for each respiratory structure and for different regions inside the lungs was formulated. The motions were scaled down to correspond to normal tidal breathing and incorporated into the phantom. We simulated the movement of the diaphragm by translating control points that define the left and right diaphragm surfaces (Figure 5.19). The heart, the stomach, and the spleen were translated up, down, backward, and forward with the movement of the diaphragm. The ribs were rotated about the axis through their costal necks to simulate their motion (Figure 5.20). Control points defining the lungs and body surfaces were altered, expanding or contracting them, depending on the rib and diaphragm motion. We set up the NCAT respiratory computational phantom in a similar fashion to that of the MCAT. Time curves were fit to the two time varying parameters for the diaphragm motion and the AP expansion in the chest. These motions

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

119

Heart

Liver FIGURE 5.19 The control points (open squares) of the diaphragm in the NCAT are translated upward and downward to simulate its respiratory movement. The right portion of the diaphragm (top of the liver) is shown in the above example. (Adapted from Segars, W.P., Lalush, D.S., and Tsui, B.M.W., IEEE Trans. Nucl. Sci., 48(1), 89, 2001.)

(Cx, Cy, Cz) Y axis

Costal neck

θR

Z axis

To sternum FIGURE 5.20 Rotation of a rib in the NCAT about the axis through its costal neck. Lateral view is shown. (Adapted from Segars, W.P., Lalush, D.S., and Tsui, B.M.W., IEEE Trans. Nucl. Sci., 48(1), 89, 2001.)

were setup to work in concert to produce a normal respiratory volume curve.12 Figure 5.21 shows 2D and 3D views of the NCAT defined at end-inspiration and end-expiration. The respiratory motion is more realistic than that of the 4D MCAT (Figure 5.21). Similar to the beating heart, the respiratory computational phantom was parameterized in terms of chest and diaphragm breathing so as to model different types of respiratory motions. In addition to motion, the flexibility of the NURBS surfaces also allows for anatomical variations. One models anatomical variations in the 4D NCAT by applying transformations to the base anatomy of the phantom.18,19 These transformations can be based upon an analysis of patient imaging data. Figure 5.22 shows a population of patients of varying anatomy used to perform studies of compensations methods in myocardial SPECT. To create the anatomical variations in this work, we set up the 4D NCAT with parameters to scale the torso and organ sizes as well as to change the size, shape, orientation, and position of the heart. These same parameters were randomly sampled from distributions obtained from the Emory PET Torso Model Database20 and used to generate 24 phantoms, half male and half female. Figure 5.22 shows simulated data from six of these phantoms.

Handbook of Anatomical Models for Radiation Dosimetry

Volume change (mL)

120

Inspiration

500

Expiration

250

0 0

1

2 3 Time (s)

4

5

3D anterior view

2D coronal slice

FIGURE 5.21 Top: Normal respiratory curve. Middle and bottom: 3D and 2D views of the NCAT at end-inspiration (left) and end-expiration (right). (Adapted from West, J. Respiratory Physiology, 5th edn., Williams & Wilkins, Baltimore, 1995.)

FIGURE 5.22 First row shows one slice from six different models. Second row shows corresponding attenuation maps, and third row shows reconstructed SPECT images simulated from the phantoms.

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

121

The 4D NCAT phantom offers a vast improvement over the geometry-based MCAT, by providing more realistic models of the anatomy, the cardiac system, and the respiratory motions of the human body. As such, the 4D NCAT has gained a widespread use in nuclear medicine imaging research, especially for evaluating and improving imaging instrumentation, data acquisition techniques, and image processing and reconstruction methods. It is widely used in myocardial SPECT, providing an excellent tool with which to study the effects of anatomy and patient motions.11,18,19,21–25 Despite this success, the 4D NCAT still has its limitations. The anatomy was based solely on the Visible Male CT dataset from the National Library of Medicine and was restricted to just the region of the torso. Also, as was the case with the MCAT, female subjects are modeled with the addition of user-defi ned breast extensions. Another limitation is that the phantom, although capable of being far more realistic, was originally designed for low-resolution imaging research and lacks the anatomical details for application to high-resolution imaging such as x-ray CT and MRI.

5.4 The 4D Extended Cardiac-Torso Phantom The 4D XCAT phantom was developed as the next version of the 4D NCAT. It includes more detailed and realistic anatomy and physiology, suitable for use in higher-resolution imaging applications. The XCAT phantom includes a whole-body male and female anatomy based on the high-resolution Visible Male and Female anatomical datasets from the NLM. The anatomical images are more detailed than that of the CT used to create the original NCAT male anatomy. The male data consist of 1878 anatomical slices of the body with a resolution of 2048 × 1216 and a pixel size of 0.33 mm and a slice width of 1 mm. The anatomical dataset of the female has the same characteristics as those of the male with one exception; the slices were obtained at 0.33 mm intervals. This results in over 5000 anatomical images over the body. Similar techniques as those used to create the NCAT organ models were used to create the detailed male and female anatomies for the XCAT phantom, as seen in Figures 5.23 and 5.24. As can be seen in the Figures, the XCAT includes more detailed organ models. Work is underway to include all the blood vessels and muscle tissue. Table 5.3 lists the current organs modeled in the male XCAT phantom. In addition to the structures shown, blood vessels, lymph nodes, and muscle tissue are also in the process of being included in the XCAT. In addition to the basic anatomy, we also updated the cardiac and respiratory motions in the XCAT phantom. High-resolution gated cardiac CT data of a healthy subject obtained from a dual source multislice CT (MSCT) scanner was used to define a more detailed anatomy for the cardiac computational phantom, as seen in Figure 5.25. The motion of the chamber surfaces was set up by combining information from the CT data as well as the gated tagged MRI data upon which the original 4D NCAT heart computational phantom was based. The cardiac twisting motion of the heart could not be ascertained from CT imaging data; therefore, the twisting motion in the original heart computational phantom was scaled to fit the new heart segmented from the CT data. Once the twisting motion was established, the radial and longitudinal contractions could be obtained by noting the epi- and endocardial borders in the gated MSCT images. We determined the motion of the vessels and other cardiac structures by tracking the landmark points located on or within

Handbook of Anatomical Models for Radiation Dosimetry

122

Anatomical detail

FIGURE 5.23 (See color insert following page 524.) Male (left) and female (right) anatomies of the 4D XCAT phantom.

Head Illac vessels Lymph nodes Vas deferens

Chest

Hypogastric vessels Uterus Bladder

Seminal Bladder vesicles Prostate

Fallopian tube Ovary Cervix

Vagina Testes Ant.

Post Male abdomen

Right

Left Female abdomen

FIGURE 5.24 The 4D XCAT includes a high level of anatomical detail.

them for each subsequent time frame. Time curves were defined for the control points defining each surface creating the enhanced 4D computational phantom for the heart. The resulting heart computational phantom is more detailed than that in the previous NCAT version, giving it the ability to be applied to high-resolution cardiac imaging research.

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

123

TABLE 5.3 List of Organs Modeled in the Male XCAT Model and Their Corresponding Volumes Organ Brain Liver Gall bladder Stomach (wall and contents) Kidneys (2) Spleen Right lung Left lung Heart (blood and tissue) Pancreas Large intestine (ascending, transcending, and descending) (wall and contents) Sigmoid/rectum (wall and contents) Small intestine (wall and contents) Bladder Prostate Testes (2) Thymus Thyroid Esophagus (wall and contents) Laryngopharynx Salivary glands Eyes (2) Adrenal (2) Pituitary Seminal vesicles Vas deferens Total skeleton Total body

Volume (mL) 1,400 1,853 31 462 348 240 1,893 1,826 670 90 1,680 212 1115 88 19 38 33 28 74 18 100 18 14 0.6 6 5 7,500 95,000

The respiratory motion of the 4D XCAT was similarly improved using more state-ofthe-art imaging data. A limitation to the respiratory motion of the 4D NCAT is that it was based on only one set of patient data. The data upon which it was based also had a resolution lower than that offered by more advanced CT scanners and consisted of only four time frames that did not adequately cover normal tidal breathing. Respiratory motion and its variations were better characterized in the 4D XCAT phantom through an analysis of several sets of 4D respiratory-gated CT image data obtained from Dr. George Chen of the Massachusetts General Hospital. Using automatic and semiautomatic techniques, the different respiratory structures were segmented from each time frame of each CT dataset. We used the time series of segmented structures to characterize the respiratory motion in each case. From an analysis of all patient datasets, the range in motion of the different structures was determined. This information was used to further parameterize the general respiratory computational phantom of the 4D XCAT to more realistically computational phantom respiratory variations. The respiratory motion was also extended to a wholebody computational phantom, including motion in the abdomen, as seen in Figure 5.25.

Handbook of Anatomical Models for Radiation Dosimetry

124

End-diastole

End-expiration

End-systole

End-inspiration

FIGURE 5.25 Cardiac and respiratory motions of the 4D XCAT.

As mentioned above, the original XCAT male and female anatomies were based on the Visible Human data. New anatomically variable XCAT phantoms are under development based on several sets of pediatric and adult CT data, as seen in Figure 5.26. We are currently analyzing many sets of adult and pediatric imaging data to define statistical distributions that will characterize the change in size and shape of the organs at different ages. To aid in this analysis, we developed an interactive graphical application that could import patient data and provided several tools (manual and automatic) that could be used to alter the male or female anatomy of the XCAT to match the patient data. In matching the patient data, statistical models were derived for the deformable template of each organ or structure and categorized based on age range. Based on the statistical models derived from the patient data, any number of random patients of varying anatomy can be created. Combined with accurate computational phantoms of the CT imaging process, the phantom can provide a wealth of simulated patient data for CT research. The new 4D XCAT approaches that of an ideal computational phantom with its basis upon human data and the inherent flexibility of the NURBS primitives. The NURBS basis for the 4D XCAT is capable of providing a mathematical description that is as realistic as a voxelized computational phantom segmented from human data. The flexibility of the NURBS basis offers a tremendous advantage over voxelized computational phantoms. From a template anatomy, the 4D XCAT can realistically model a multitude of different

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

125

FIGURE 5.26 Anatomical variations from adult to pediatric produced for the XCAT based on CT data.

anatomical variations and patient motions from both male and female subjects (including adults and children). Combined with accurate models for the imaging process, the phantom can provide a wealth of simulated image data that are far more consistent with that of actual patients, as seen in Figure 5.27. There is essentially no limitation. Any number of different anatomies, cardiac or respiratory motions or patterns, and spatial resolutions can be simulated to perform research.

Imaging simulations using the phantom

Myocardial SPECT

22

PET (Brain)

44

Chest x-ray

66

In-111 ProstaScint SPECT FIGURE 5.27 Imaging simulations performed using the XCAT phantom.

x-ray CT

126

Handbook of Anatomical Models for Radiation Dosimetry

5.5 The 4D Mouse Whole-Body Phantom The rapid growth in genetics and molecular biology in recent years, combined with the development of techniques for genetically engineering small animals, has led to increased interest in in vivo small animal imaging. With the rise of small animal imaging, new instrumentation, data acquisition strategies, and image processing and reconstruction techniques are being developed and researched. A major challenge remaining to researchers is how to evaluate the results of these new developments. Simulation techniques can provide a vital tool to evaluate and improve molecular imaging devices and techniques. Much research has been done in creating digital human phantoms for medical imaging research. Currently, there is a lack of realistic computer-generated phantoms modeling the mouse anatomy and physiological functions for use in molecular imaging research. The same methods and techniques used to develop the 4D NCAT and XCAT phantoms were used in the creation of a new 4D MOBY phantom.26 A 256 × 256 × 1024 3D magnetic resonance microscopy (MRM) dataset of a normal 16 week old male C57BL/6 mouse was used as the basis for the anatomy of the phantom. The dataset was obtained from G. Allan Johnson of the Duke Center for In Vivo Microscopy, an NIH Resource (P41 05959/R24 CA 92656). With a resolution of 110 μm over the whole body, the dataset is extremely detailed, allowing the creation of realistic models for several different anatomical structures. Figure 5.28 shows sample transaxial slices obtained from the MRM dataset.

FIGURE 5.28 Sample transaxial slices of the MRM dataset used to create the 3D anatomy for the mouse phantom. (From Segars, W.P. et al., Mol. Imag. Biol., 6, 149, 2004. With permission.)

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

127

Models for the different structures were created using the same techniques developed in our laboratory to construct the 4D NCAT and XCAT phantoms. The anatomical structures were manually segmented using the software program SURFdriver and then used as input to the Rhinoceros NURBS modeling software, where NURBS surfaces were fit to the segmented structures. A gated black-blood MRI (bb-MRI) cardiac data set (Figure 5.29) of a normal 15 week old male C57BL/6 mouse was used to create a 4D beating heart computational phantom for the phantom. The study was obtained from Stuart S. Berr of the University of Virginia (UVa) and consisted of 12 time frames over a complete cardiac cycle. At each time frame, the 256 × 256 SA MR images had a pixel size of 0.1 × 0.1 mm and a slice thickness of 1 mm. Using the technique described above, 3D NURBS surfaces were created for the principal structures of the heart: the right and LVs, the atria, and the large vessels at each time frame. We set up the time correspondence between the control points defining a surface over the time frames based on the 4D NCAT human heart computational phantom. We also scaled back the cardiac twisting motion illustrated in the 4D NCAT cardiac computational phantom14 to fit the smaller heart size of the mouse. Once the twisting motion was established, the radial and longitudinal contractions of the control points could be obtained by noting the epi- and endocardial borders in the gated MRI images. Using this technique,

FIGURE 5.29 One SA slice through 12 frames of gated black-blood MRI cardiac data. (From Segars, W.P. et al., Mol. Imag. Biol., 6, 149, 2004. With permission.)

Handbook of Anatomical Models for Radiation Dosimetry

128

we determined the 3D position of each control point defining a cardiac surface for each time frame and time curves were then defined for them completing the 4D computational phantom shown in Figure 5.30. Respiratory motion in the MOBY phantom was based on similar respiratory mechanics observed when creating the human NCAT and XCAT phantoms.11 The NURBS surfaces that define the respiratory structures in the phantom were altered through manipulation of their control points to simulate respiration. The diaphragm is defined in the phantom as the top of the liver that encompasses both left and right sections of the body. We simulated the movement of the diaphragm by simply translating the surface defining the liver (Figure 5.31). In addition to moving linearly up and down, the diaphragm was also set to move forward and backward with the changes in the AP diameter of the chest due to the movement of the ribs. This movement was observed in human respiration as well. The heart, stomach, spleen, and kidneys were translated with the movement of the diaphragm. In each case, the translation was applied to the control points defining the different structures in order to move them. The NURBS surfaces defining the lungs and body outline in the mouse phantom were set up to expand or contract with changes in the ribcage. Figure 5.32 shows 3D renderings of the respiratory computational phantom of the MOBY phantom.

End-diastole (ED) RA

LA

60

RA

LV

RV

End-systole (ES) LA

LV

RV

RV

Volume (μL)

50 40 LV 30 20 RA LA

10 0 0

5

10

15 Time frame

20

25

30

FIGURE 5.30 Top: 3D surface renderings of the epi- and endocardial surfaces of the RV and LV for the new NURBS-based mouse beating heart model at end-diastole and end-systole. Bottom: Volume curves for the atria and ventricles of the mouse heart model. (From Segars, W.P. et al., Mol. Imag. Biol., 6, 149, 2004. With permission.)

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

Heart

129

Heart

Liver Liver Stomach Stomach Spleen

Kidney

Spleen Kidneys

Kidney

FIGURE 5.31 Inspiratory motions of the liver (diaphragm), stomach, spleen, heart, and kidneys simulated in the mouse phantom. Expiratory motion was simulated as the reverse of the inspiratory motion. (From Segars, W.P. et al., Mol. Imag. Biol., 6, 149, 2004. With permission.) End-expiration

End-inspiration

FIGURE 5.32 Top: Anterior view of end-expiration (left) and end-inspiration (right) in the MOBY model. Bottom: Left lateral view at end-expiration (right) and end-inspiration (left). The dotted line indicates the movement of the diaphragm. (From Segars, W.P. et al., Mol. Imag. Biol., 6, 149, 2004. With permission.)

Handbook of Anatomical Models for Radiation Dosimetry

130

(a)

(b) FIGURE 5.33 Top: Reconstructed SPECT coronal images generated from the mouse phantom simulating the uptake of Tc-99m MDP. Bottom: Coronal SPECT images obtained experimentally from an actual mouse. (From Segars, W.P. et al., Mol. Imag. Biol., 6, 149, 2004. With permission.)

Used in combination with accurate models of the imaging process, the 4D MOBY phantom can produce realistic imaging data to serve as a standard from which other molecular imaging devices and techniques can be evaluated and improved. The top of Figure 5.33 shows reconstructed SPECT images generated from the phantom simulating the uptake of Tc-99m MDP in a normal mouse without respiratory motion. The bottom of Figure 5.33 shows reconstructed SPECT images obtained from imaging a mouse

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

131

(a)

(b) FIGURE 5.34 Top: Reconstructed cone-beam x-ray CT images generated from the mouse phantom. Bottom: Reconstructed cone-beam x-ray CT images obtained from a live mouse using a micro-CT system developed in our laboratory. (From Segars, W.P. et al., Mol. Imag. Biol., 6, 149, 2004. With permission.)

with the same radiopharmaceutical in our laboratory. Coronal image slices are shown. The top of Figure 5.34 shows reconstructed x-ray CT transaxial images simulated using the mouse phantom while the bottom of Figure 5.34 shows similar CT images obtained from a live mouse using a micro-CT system built in our laboratory. In both cases, the simulated images are comparable to those obtained experimentally. Like its human phantom counterparts, the MOBY phantom also has the ability to simulate different anatomies. Current work is underway to create anatomically variable models of the MOBY phantom as well as to create a computational phantom for the laboratory rat. Like the NCAT and XCAT, the 4D MOBY phantom can be applied to small animal imaging research to look at such things as the effects due to anatomy and motion, as well as

132

Handbook of Anatomical Models for Radiation Dosimetry

to investigate image acquisition protocols and to research new reconstruction techniques. It can also be used to test and validate new small animal imaging scanners before entering production process and performing research into dosimetry in small animals. The MOBY phantom, as such, provides a unique and useful tool in molecular imaging research.

5.6 Summary The above discussion presents our developments toward more ideal hybrid computational phantoms for use in medical imaging research. We have found that the NURBS modeling technique is an efficient and flexible way to describe the anatomy and physiology for realistic phantoms. NURBS surfaces have the ability to model complex anatomy based on actual patient data. This gives them the level of realism offered by voxel-based phantoms. In addition, NURBS surfaces have the flexibility of stylized phantoms, in that they can be altered easily to model variations in the 3D anatomy and be extended to 4D to model patient motion. The NURBS-based phantoms, therefore, offer a major evolutionary advance in the development of computerized computational phantoms. Using NURBS as a basis, we developed the NCAT, XCAT, and MOBY phantoms for the purpose of medical imaging research. With the ability to simulate realistic predictive imaging data of a population of human or animal subjects, the phantoms have found a wide use to develop, evaluate, and improve imaging devices and techniques and to investigate the effects of anatomy and motion. They may also provide useful tools in the field of radiation dosimetry as can be seen in other chapters.

References 1. Snyder, W.S. et al. Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, J. Nucl. Med., Suppl. 3, 7, 1969. 2. White, D.R. Tissue substitutes in experimental radiation physics, Med. Phys., 5, 467, 1978. 3. Hubbell, J.H. and Seltzer, S.M. Tables of X-Ray Mass Absorption Coefficients and Mass EnergyAbsorption Coefficients (Version 1.03) [online]. Available: http://physics.nist.gov, Gauthersbury, MD: National Institute of Standards and Technology, 1995. 4. Pretorius, P.H. et al. Evaluation of right and left ventricular volume and ejection fraction using a mathematical cardiac torso phantom, J. Nucl. Med., 38, 1528, 1997. 5. Park, J., Metaxas, D.N., and Axel, L. Analysis of left ventricular wall motion based on volumetric deformable models and MRI-SPAMM, Med. Imag. Anal., 1, 53, 1996. 6. Park, J., Metaxas, D.N., and Axel, L. Quantification and visualization of the 3D nonrigid motion of the left ventricle, in Proceedings of the SPIE Medical Imaging Conference (Physiology and Function), San Diego, CA, 177, 1997. 7. Park, J. et al. Deformable models with parameter functions for cardiac motion analysis from tagged MRI data, IEEE Trans. Med. Imag., 15, 278, 1996. 8. Guyton, A. and Hall, J. Textbook of Medical Physiology, 9th edn., Philadelphia, PA: WB Saunders Co., 1996. 9. LaCroix, K.J. Evaluation of an attenuation compensation method with respect to defect detection in Tc-99m-MIBI myocardial SPECT images, PhD dissertation, The University of North Carolina at Chapel Hill, 1997.

The MCAT, NCAT, XCAT, and MOBY Computational Human and Mouse Phantoms

133

10. LaCroix, K.J. et al. Receiver operating characteristic evaluation of iterative reconstruction with attenuation correction in 99mTc-Sestamibi myocardial SPECT Images, J. Nucl. Med., 41, 502, 2000. 11. Segars, W.P., Lalush, D.S., and Tsui, B.M.W. Modeling respiratory mechanics in the MCAT and spline-based MCAT phantoms, IEEE Trans. Nucl. Sci., 48, 89, 2001. 12. West, J. Respiratory Physiology, 5th edn., Baltimore: Williams & Wilkins, 1995. 13. Piegl, L. On nurbs—A survey, IEEE Comput. Graph. Appl., 11, 55, 1991. 14. Segars, W.P. Development and application of the new dynamic NURBS-based cardiac-torso (NCAT) phantom, PhD thesis, University of North Carolina at Chapel Hill, 2001. 15. Visible human male and female datasets, National Library of Medicine. Available at: http://www. nlm.nih.gov/research/visible/visible human.html. 16. Moody, D. and Lozanoff, S. A practical computer program for generating three-dimensional models of anatomical structures, Paper Presented at the 14th Annual Meeting of the American Association of Clinical Anatomists, Honolulu, HA, July 8, 1997. 17. Rhinoceros NURBS modeling software. Available at: http://www.rhino3d.com. 18. He, X. et al. Effect of anatomical and physiological factors and compensation methods on observer of performance for defect detection in myocardial perfusion, J. Nucl. Med., 44, 112P, 2003. 19. He, X. et al. A mathematical observer study for the evaluation and optimization of compensation methods for myocardial SPECT using a phantom population that realistically models patient variability, IEEE. Trans. Nucl. Sci., 51, 218, 2004. 20. Barclay, A.B., Eisner, R.L., and DiBella, E.V. PET Thorax Model Database, http://www.emory.edu/ CRL/abb/thoraxmodel, Atlanta, GA: Crawford Long Hospital of Emory University, 1996. 21. Segars, W.P. and Tsui, B.M.W. Study of the efficacy of respiratory gating in myocardial SPECT using the new 4-D NCAT phantom, IEEE Trans. Nucl. Sci., 49, 675, 2002. 22. Tsui, B.M.W., Segars, W.P., and Lalush, D.S. Effects of upward creep and respiratory motion in myocardial SPECT, IEEE Trans. Nucl. Sci., 47, 1192, 2000. 23. Smyczynski, M. et al. Impact of respiratory motion on the detection of solitary pulmonary nodules with SPECT imaging of NeoTect, in IEEE Medical Imaging Conference and Nuclear Science Symposium, Norfolk, VA, Nov. 10–16, 2002. 24. Smyczynski, M. et al. Modeling the respiratory motion of solitary pulmonary nodules for investigating SPECT tumor imaging, in IEEE Medical Imaging Conference and Nuclear Science Symposium, San Diego, CA, Nov. 4–10, 2001. 25. Tsui, B.M.W. et al. Quantitative cardiac SPECT reconstruction with reduced image degradation due to patient anatomy, IEEE Trans. Nucl. Sci., 41, 2838, 1994. 26. Segars, W.P. et al. Development of a 4-D digital mouse phantom for molecular imaging research, Mol. Imag. Biol., 6, 149, 2004.

6 The 3D and 4D VIP-Man Computational Phantoms X. George Xu, Tsi-Chian Ephraim Chao, Ahmet Bozkurt, Chengyu Shi, and Juying Zhang

CONTENTS 6.1 Introduction ............................................................................................................... 136 6.2 Development of VIP-Man and 4D VIP-Man.......................................................... 137 6.2.1 Original Images ........................................................................................... 137 6.2.2 Segmentation ................................................................................................ 138 6.2.2.1 General Organs............................................................................. 139 6.2.2.2 Red Bone Marrow ........................................................................ 139 6.2.2.3 GI Tract........................................................................................... 141 6.2.2.4 Miscellaneous Issues ................................................................... 142 6.2.3 Labeling ......................................................................................................... 144 6.2.4 Visualization ................................................................................................. 144 6.2.5 Implementation of VIP-Man into Monte Carlo Codes............................ 148 6.2.6 Extending 3D VIP-Man into 4D VIP-Man ................................................ 148 6.2.6.1 Conversion from Voxel Data to Surface Definition ....................................................................................... 149 6.2.6.2 Deformation of the Organ Surfaces ........................................... 149 6.2.6.3 Revoxelization of the 4D Computational Phantom for Each Respiratory Phase ......................................................... 150 6.3 Comparison between VIP-Man and Other Computational Phantoms .................................................................................................................... 151 6.4 Applications of the VIP-Man Computational Phantom ...................................... 152 6.4.1 Health Physics .............................................................................................. 152 6.4.1.1 External Photon Dosimetry ........................................................ 152 6.4.1.2 External Electron Dosimetry ...................................................... 153 6.4.1.3 External Neutron Dosimetry ...................................................... 153 6.4.1.4 External Proton Dosimetry ......................................................... 154 6.4.1.5 Internal Electron Dosimetry ....................................................... 154 6.4.1.6 Internal Photon Dosimetry for the GI Tract ............................. 155 6.4.1.7 RBM Dosimetry for External Irradiations ................................ 155 6.4.2 Applications in Radiological Imaging ...................................................... 156 6.4.2.1 Organ Doses from SPECT and PET Brain Imaging ................ 156 6.4.2.2 Organ Doses from X-Ray Radiographs ..................................... 156 6.4.2.3 Image Quality Optimization in Radiograph............................ 157 6.4.2.4 Organ Doses from Interventional Cardiological Examinations ................................................................................ 157 135

Handbook of Anatomical Models for Radiation Dosimetry

136

6.4.3

Applications in Radiotherapy .................................................................... 157 6.4.3.1 External Beam Selection in Prostate Radiation Treatment ....................................................................................... 157 6.4.3.2 Organ Doses from Proton Radiation Treatments .................... 158 6.4.3.3 Dosimetry for Respiration Management in IGRT ................... 158 6.4.3.4 Organ Doses from Cone-Beam CT Imaging for IGRT ......................................................................................... 158 6.5 Summary .................................................................................................................... 158 Acknowledgments ............................................................................................................... 160 References ............................................................................................................................. 160

6.1 Introduction The visible photographic man (VIP-Man) computational phantom was developed during 1996–2000 at Rensselaer Polytechnic Institute (RPI, Troy, New York). The interest in anatomical and dosimetric modeling by the group at RPI originated from the doctoral research Dr. X. George Xu performed in early 1990s at Texas A&M University (TAMU, College Station, TX), under the guidance of Dr. W. Dan Reece, Dr. John Poston, and Dr. Wesley Bolch (now with the University of Florida) served on his dissertation committee. Dr. Xu’s work at TAMU was to implement the stylized adult male and female phantoms into the Monte Carlo N-Particle (MCNP) code in order to develop an algorithm that would link personnel dosimeter readings to effective dose equivalent that was adopted by the U.S. Nuclear Regulatory Commission (see Chapter 18 for more details). Dr. Xu joined the RPI faculty in 1995. Later that year, an international workshop on “Voxel Phantom Development” was held at the National Radiological Protection Board (Chilton, U.K.). It was clear that the so-called voxel phantoms would soon replace the stylized phantoms. Voxel phantoms reported prior to 1995 were developed from either computer tomography (CT) or magnetic resonance imaging (MRI) of volunteers. The majority of the subjects were adult males. The image slices were relatively thick due to the need to reduce the total whole-body scanning time. Furthermore, the computer memory at that time severely limited the number of voxels in the modeling. Xu and two of his students at RPI decided to develop a new voxel phantom and soon became impressed with the unique images from the Visible Human Project® (VHP) that became available around that time.1,2 During late 1990s, the computer technologies were rapidly improving and the VHP’s images were made freely available for research by the National Library of Medicine (NLM). After a couple of years of planning, Dr. Xu received a 4-year research grant from the National Science Foundation’s CAREER program in 1998. Dr. Tsi-chian Chao and Dr. Ahmet Bozkurt, PhD students at that time, immediately became totally absorbed into the project and soon made several breakthroughs that led to a series of papers related to a voxel-based phantom known as the VIP-Man. In 2004, Dr. Chengyu Shi, a former PhD student at RPI, Dr. Xu and Mr. Juying Zhang, a current PhD student, added respirationsimulating features into the VIP-Man phantom. This chapter describes the development of three-dimensional (3D) and four-dimensional (4D) versions of the image-based whole-body computational phantom, VIP-Man. These are followed by a summary of numerous applications to projects in health physics, diagnostic imaging, and radiation treatment.

The 3D and 4D VIP-Man Computational Phantoms

137

6.2 Development of VIP-Man and 4D VIP-Man Three-dimensional medical imaging techniques, such as CT and MRI, allow us to easily visualize the internal structures of the body and to store the images in versatile digital formats. In the past two decades, the radiotherapy community has increasingly depended on the use of Monte Carlo techniques with patient CT images for clinical treatment dose optimization.3 Compared to the medical community, however, health physicists faced the following unique and intractable technical challenges: (1) whole-body computational phantoms are needed for most health physics applications, but medical images are taken only for a portion of the body (CT procedures expose the patients to intense x-rays); (2) a large amount of internal organs/tissues have to be identified and segmented for organ dose calculations in health physics, while, in radiotherapy, only the tumor volume and the critical organs inside the field need to be specified; (3) the size of a whole-body computational phantom can be potentially too large for computers and Monte Carlo codes to handle; and (4) health physics dosimetry involves photons, electrons, neutrons, and protons, but majority of the clinical radiotherapy procedures involve only photon/electron beams or seeds (a few centers also involve neutron or proton beams). Because of these issues, only a few groups successfully constructed image-based wholebody computational phantoms (see Chapter 1 for a complete list of voxel phantoms).4–8 In 2000, Xu et al.9 was eager to share their contribution to the literature by reporting an adult male phantom, named VIP-Man, which had the best image resolution of that time: pixel size of 0.33 mm × 0.33 mm and slice thickness of 1 mm. In such a high resolution, the phantom was believed to offer attractive advantages in modeling tiny radiosensitive structures such as red bone marrow (RBM), skin, eye lens, thyroid, optical nerves, etc. The development of a voxel phantom includes several steps: 1. Selection of an original image set: The quality of original raw data set is crucial in constructing a tomographic computational phantom. The resolution of original images determines the ability to perform segmentation and modeling of human anatomy. 2. Segmentation: Each pixel in the original images was identified as belonging to a tissue/organ by manual or semiautomatic procedures involving computer programs, knowledge about the anatomy and visual inspection of the images. 3. Labeling: Each pixel is assigned an index number that specifies not only which tissue/organ it belongs to but also its density and chemical composition. 4. Visualization: Two-dimensional (2D) or 3D visualization is necessary to display and inspect the segmented, labeled, and registered images. 5. Implementing this phantom into a Monte Carlo code for radiation transport purposes.

6.2.1 Original Images In the mid-1990s, several unique sets of whole-body CT/MR/color photographic images from the NLM’s VHP became available (http://www.nlm.nih.gov/research/visible). The ambitious goal of the VHP, which was conceived in 1988 and initiated in 1991, was to build the most detailed digital image library about the anatomy of an adult male and an adult

138

Handbook of Anatomical Models for Radiation Dosimetry

female.1,2 Cadavers that were considered “normal” and representative of a large population were evaluated. The donated body of an executed 38-year-old male from Texas was the first specimen to be selected for VHP. The individual was 186 cm tall and weighed 90 kg. Later, a 58-year-old female body was also obtained. To ensure the applicability, it was decided by NLM that the image data needed to be documented in several common formats used by radiologists and other physicians. Eventually, four modalities were used: traditional x-rays and CT scans to optimally visualize bone, MRI for soft tissue, and color photographs for definitive resolution. The color photographs, which had the finest resolution,9 were used to provide a standard for comparison. Generally, an image format consists of many pixels each representing a tissue volume in a 2D map. The 3D volume of the tissue is called a voxel, which is determined by multiplying the pixel size by the thickness of an image slice.10 In the VHP data set, the transversal MRI images of the head and neck and the longitudinal sections of the rest of the body, which were obtained at 4 mm intervals, are 256 × 256 pixels in resolution (each with 12 bits of gray tone resolution). The resulting voxel size for the MRI data set (torso portion) is 1.88 mm × 1.88 mm × 4 mm. The CT data consists of transversal CT scans of the entire body taken at 1 mm intervals at a resolution of 512 × 512 pixels where each pixel is made up of 12 bits of gray tone. The voxel size for the CT data set (torso portion) is 0.94 mm × 0.94 mm × 1 mm. The transversal anatomical photographs for both male and female cadavers are 2048 × 1216 pixels (each pixel is with 24 bits of color taken at 1 mm thick slices for the male cadaver and 0.33 mm for the female). For the visible male, there are a total of 1871 slices of both CT and anatomical photographs. The transversal anatomical images were obtained by photographing the top surface of the body block after removal of (by shaving) each successive millimeter (0.33 mm for the female) by a cryomacrotome. These color photographic data sets for whole body has a voxel size of 0.33 mm × 0.33 mm × 1 mm for the male (0.33 mm × 0.33 mm × 0.33 mm for the female). Since the first public debut on November 28, 1994, VHP images have been available in public domain. Since then, computer engineers and anatomists, working together, have devoted unprecedented effort to classify and visualize the data sets. Until recently, the Visible Human Male was the most complete computerized database of the human body ever assembled.11 Called “the greatest contribution to anatomy since Vesalius’s 1543 publication of De Humani Corporis Fabrica,” the VHP data sets are the seeds for a growing medical revolution. For more than a decade, scientists worldwide have been utilizing this national resource of anatomical information for biomedical sciences and engineering applications.12 Based primarily on the color photographic images, VIP-Man was constructed using the following detailed methods. 6.2.2 Segmentation To calculate organ doses, Monte Carlo simulations do not require the information of CT number, percentage distribution of water or color, but require the relationship between voxels and tissues/organs. Unfortunately, today’s computers are unable to automatically recognize tissues/organs from CT/MRI/cryosection images. Segmentation is a manual or semiautomatic procedure to recognize and classify the pixels on an original image into anatomical structures. VHP-cryosection images were previously segmented in the VHP by Spitzer and Whitlock to yield up to 1400 structures.11 Among these are 23 dosimetrically important tissues/organs. However, the other 16 tissues/organs were not segmented in the original data, including RBM, eye lenses, and gray/white matter in cerebrum, wall/content/mucosa in gastrointestinal (GI) tract, teeth, cerebrospinal fluid (CSF), and bladder wall/content. Several

The 3D and 4D VIP-Man Computational Phantoms

139

properties in the VHP-cryosection color photographs were helpful in identifying certain tissues and organs using various “measurement” techniques.13 For example, 1. Colors, such as red, green, and blue 2. Locations, such as x, y, z, distance to center, and distance to boundary 3. Connectivity, such as continuous component, boundary touching, hole, and isolated island 4. Others, such as shape, hue, saturation, intensity, etc. Organs with different compositions are expected to show variations in one, several, or all of the properties listed above. For this reason, the following sections will describe how these properties were used. 6.2.2.1 General Organs First of all this section will demonstrate how the “measurement” process works in segmenting the cerebrum, a tissue that is composed of the gray matter, white matter, and CSF. The color of all cerebrum pixels are examined and analyzed to yield the color histograms of red, green, and blue channels and the intensity. After several trials, the cerebrum is segmented by the following procedures. 1. Check the color of each pixel to see if R is larger than 190, G is larger than 160, and B is larger than 100, where R, G, and B are the intensity in red, green, and blue channels with a value from 0 (dark) to 255 (bright). If any two out of three checks mentioned above are passed, this pixel is identified as white matter. 2. Check the color of each pixel to see if R is smaller than 144, G is smaller than 102, and B is smaller than 50. If any two out of three checks mentioned above are passed, this pixel is identified as CSF. 3. Otherwise, this pixel is identified as gray matter. The gray matter, white matter, and CSF can be separated clearly by these procedures. Alternately, some tissues/organs cannot be segmented according to the color differences alone, such as the eye lenses/eyes, bladder wall/content, etc. The eye lenses can be segmented from the eye balls using the threshold of green channel (G > 114). However, other parts of the eyes also have the same green color as the eye lenses. The threshold of the green channel alone might yield some false pixels of the eye lenses. Methods like erosion, dilation, and median filter can be utilized jointly. In this method, the selection of the biggest connective compound followed by the median filter turned to be the favorite approach. The bladder content and the bladder wall were segmented as well in the similar way. 6.2.2.2 Red Bone Marrow The calculation of absorbed dose in the tissue of skeleton is a complex problem because of the difficulties in modeling the microscopic distribution of soft tissue and bone.14 In addition, there are two kinds of marrow inside the bone, yellow bone marrow (YBM) and RBM, both of which are difficult (if not impossible) to be identified and segmented directly from CT or MRI images. The distribution of RBM in the whole body has not been directly modeled in both stylized and voxel phantoms. However clinical measurements in cadavers

140

Handbook of Anatomical Models for Radiation Dosimetry

suggest that the RBM tends to be located more in flat bones or skull, with concentrations varying by sex and age.15 The difficulty in defining RBM distribution has led researchers to derive different methods to estimate the RBM dose. Two general methods are currently used for skeletal dosimetry from external photons: (1) Snyder et al.16 assume a homogeneous representation and identical efficiencies to absorb energy of the skeleton. Therefore, the mass ratios are used for portioning the energy deposition to various target tissues (endosteum, RBM, etc.). They also state that potential overestimation of RBM dose will increase for photons at energies lower than 200 keV. (2) Cristy and Eckerman17 suggest another way to estimate skeleton dosimetry from the photon fluence through the trabecular bone. Previous studies of the mean chord lengths and electron absorbed fractions are utilized to estimate how the secondary particles are generated and deposited due to the fluence. The high-resolution color photographs of the VHP brought a unique opportunity to realistically quantify the 3D spatial and quantitative representation for whole-body RBM distribution. Since the RBM cavities are less than 10 μm, smaller than each pixel of the VHP data set (0.33 mm), the microscopic segmentation was not possible. However, the distribution of RBM at the “macroscopic” level was satisfactorily segmented by a threshold determined by the statistical properties of the original color images. After a careful calibration, the “redness” and “saturation” were tested to be the best properties for separating RBM distribution. The “redness” is defined as Redness =

R × 255 R +G +B

(6.1)

where R is the red channel of the corresponding pixel G is the green channel B is the blue channel “saturation” is the purity of color, such as ⎧ 0, ⎪ ⎪ ⎪ max − min × 255, Saturation = ⎨ ⎪ max + min ⎪ max − min × 255, ⎪ ⎩ 2 − (max + min)

if max = min if max + min ≤ 1 if max + min > 1

where max =

max(R , G, B ) 255

min =

min(R , G, B ) 255

For example, the saturation is 255 for pure color, and is 0 for black/white/gray.

(6.2)

The 3D and 4D VIP-Man Computational Phantoms

141

First, the colors of all bone pixels were measured. The “redness” of each bone pixel was then calculated according to Equation 6.1 and the “saturation” from Equation 6.2. A plot of saturation versus redness for all the bone voxels in whole body showed that there was a cluster of voxels that had a redness greater than 120. These pixels were then identified as having considerable amount of RBM and were then grouped to represent the macroscopic distribution of RBM in the whole body. A 3D visualization of RBM distribution was obtained to visualize the distribution which did not exist until the work associated with the VIP-Man phantom. As shown in Figure 6.1b, VIP-Man has more RBM in flat bones than long bones, as predicted by clinical data.5,16 A comparison of masses of the RBM and skeletal tissue between VIP-Man and the International Commission on Radiological Protection (ICRP) 70 Reference Man15 is given in Table 6.1. Although VIP-Man (104 kg) is heavier than the Reference Man (73 kg), the skeletal and RBM masses were selected to be similar. However, VIP-Man has more RBM in lower body and less RBM in upper body compared to the Reference Man. It should be pointed out that the YBM is not segmentable from the bone in VIP-Man because it has the similar color as the bone. In addition, the ICRP skeleton computational phantom18 suggests that the trabecular bones and the cortical bone be considered separately. However, in this study, the RBM in VIP-Man was segmented without distinguishing trabecular or cortical bone. In the case of the RBM, the term “segmentation” means a process in which a consistent, computer-aided method is used to select pixels that contain a large fraction of RBM. This way, there is at least a chance to try to model the RBM directly from these cadaver-based color images. In essence, this approach is not different from a method commonly used by many developers who selected the outermost layer in the voxel phantom as the skin—a tissue that has a size smaller than a voxel. 6.2.2.3 GI Tract

(a)

(b)

FIGURE 6.1 (a) Anterior 2D visualization for VIP-Man model and (b) 3D visualization for RBM of VIP-Man.

The GI tract in VIP-Man consists of 14 different tissues or organs, including esophagus content, esophagus mucosa, esophagus wall, lower large intestine (LLI) content (includes descending and sigmoid colons), LLI mucosa, LLI wall, rectum, small intestine, stomach content, stomach mucosa, stomach wall, upper large intestine (ULI) content (includes ascending and transverse colons), ULI mucosa, and ULI wall. The differences in color and texture are the key to separate the wall and content of the GI tract. The stomach, esophagus, and large intestine were visually segmented into content and wall because the former is much redder (darker) than the latter. In the small intestine, however, the color difference of wall and contents were not significant enough to be useful. In addition, the folds and the villi on the surface of small intestine wall are too small to be identified in the original image used to create VIP-Man. Mucosa, which is a superficial tissue on the inner wall of esophagus, stomach, ULI, and LLI, has been known to be more radiosensitive than the

Handbook of Anatomical Models for Radiation Dosimetry

142

TABLE 6.1 Comparison of Skeleton and RBM Mass Distributions from VIP-Man and ICRP 70 Reference Man

Bone Structure

VIP-Man

RBM (g) ICRP 70 Reference Man

Skeleton (w/Marrow) (g) ICRP 70 Reference VIP-Man Man

Cranium

48.21

88.92

876.43

1,239.00

Mandible

2.00

9.36

89.14

126.00

Scapulae

46.43

32.76

319.24

378.00

Clavicles

13.41

9.36

102.68

84.00

Sternum

43.13

36.27

117.00

126.00

160.80

188.37

728.34

735.00

22.87

45.63

155.85

1,995.00

Ribs Cervical vertebrae Thoracic vertebrae

145.06

188.37

563.28

Lumber vertebrae

135.09

143.91

470.22

Sacrum

110.07

115.83

303.16

Innominate

315.93

204.75

1,071.43

1,113.00

41.37

78.39 –

2,027.67

1,606.50

1,376.93

1,186.50

768.96

661.50

647.61

556.50

379.68

378.00

244.64

241.50

Femora Tibiae

0.78

Other foot

2.01

Humeri

37.09

– 26.91 –

Radii and ulane

2.08

Other hand bone

1.11

–

Other

1.14

1.17

21.01

73.50

1,128.57

1,170.00

10,263.27

10,500.00

Total

Source: Cristy, M. and Eckerman, K.F., Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381/Volumes I–VII, Oak Ridge National Laboratory, Oak Ridge, TN, 1987.

wall tissue.19 From literature review, the average mucosal thickness can be 188 ± 52 μm 20 or 2.5 mm 21 in stomach, and 427 ± 98 μm in large intestine.20 In VIP-Man, a thin layer (0.33 mm for stomach and 0.67 mm for LLI and ULI) between wall and content is defined as the “mucosa.” The organ and tissue masses of VIP-Man are listed in Table 6.2. There is a significant difference in the mass of GI tract between VIP-Man and Reference Man, especially in the colon region (ULI + LLI). This suggests that the weight of adult intestine tract can be very different (varied from about 400 to 3400 g).21 6.2.2.4 Miscellaneous Issues 1. The teeth of VIP-Man were manually segmented by using an imaging software (Paint Shop Pro, version 6, Jasc Software, Inc., Eden Prairie, MN). 2. The male breasts of VIP-Man were 33.6 g in total weight, which were artificially constructed using the fat tissue that covered an area of 4.2 cm from a nipple and 0.99 cm beneath the skin. 3. Two layers of skin are important for radiation safety: 20–100 μm for epidermal effects, and 300–500 μm for dermal effects. The dose in the first layer is selected

The 3D and 4D VIP-Man Computational Phantoms

143

TABLE 6.2 Comparison of Organ Masses for VIP-Man, Scaled VIP-Man, and ICRP-23 Reference Mana Organs/Tissues Adrenals

VIP-Man (g)

Scaled VIP-Manb(g)

ICRP 23 (g)

8.3

5.8

Bladder (wall)

41.4

29.0

45.0

Bladder (urine)

43.2

30.2

102.0

1,574.0

1,101.9

1,429.0

Brain + nerve Breast (male)

14.0

33.6

23.5

26.0

265.1

185.6

121.0

Esophagus (wall)

38.9

27.2

Esophagus (lumen)

26.8

18.8

40.0 –

CSF

Esophagus (mucosa) Fat

–

3.5

2.5

36,326.6

25,430.7

17,200.0 10.0

Gallbladder (wall)

12.0

8.4

Gallbladder (bile)

21.0

14.7

60.0

Heart muscle

398.7

279.1

330.0

Kidneys

335.4

234.8

310.0

Lenses of eyes

0.4

0.4

1,937.9

1,356.6

1,800.0

LLI (wall)

290.8

203.6

160.0

LLI (lumen)

324.2

227.0

LLI (mucosa)

35.8

25.1

135.0 –

Liver

Lungs

0.54

910.5

637.4

1,000.0

43,002.6

30,104.3

28,000.0

Pancreas

82.9

58.0

100.0

Prostate

18.9

13.2

16.0

1,130.4

791.2

1,500.0

Muscle

RBM

1,0114.2

7,080.7

8,500.0

Skin

Skeleton

2,253.4

1,577.5

2,600.0

Small intestine

1,291.8

904.3

1,040.0

Spleen

244.0

170.8

180.0

Stomach (wall)

159.5

111.7

150.0

Stomach (content)

324.5

227.2

Stomach (mucosa)

13.7

9.6

250.0 –

Testes Thymus Thyroid

21 (1) 11.2

14.7

35.0

7.8

20.0

27.6

19.3

20.0

ULI (wall)

461.1

322.8

160.0

ULI (lumen)

905.7

634.0

ULI (mucosa)

63.4

44.4

135.0 –

Other Total a

b

1,688.0

1,181.7

4,382.0

104,277.2

73,000.0

70,000.0

Reference Man values are from ICRP 2322 and the MIRD model values from Cristy and Eckerman.1 Scaled VIP-Man was scaled down to be 176 cm in height and 73 kg in weight in accordance with ICRP 66.25

144

Handbook of Anatomical Models for Radiation Dosimetry

to represent the skin dose by ICRP.22 However, Wells23 suggests that the skin dose should be evaluated at depths between 300 and 500 μm according to data on late dermal atrophy. The skin in VIP-Man is about 0.7 mm thick in the front portion of the body, and becomes thicker in the back, especially near the hips. 4. There is only one testicle (according to the medical records of the cadaver) in VIP-Man. 5. The eyes of VIP-Man are closed (an important factor in external electron calculations). 6. VIP-Man is 186 cm in height and 104 kg in weight. In comparison, the Reference Man is only 73 kg and 176 cm.24 Since VIP-Man (like other voxel phantoms) is different from the Reference Man, calculations for effective dose should be performed and reported carefully. It is so different that one journal reviewer once insisted that we report our result in “effective VIP-Man dose.”25 6.2.3 Labeling An index number representing different tissues/organs was assigned to each segmented pixel. There are a total of 72 different tissues/organs segmented in VIP-Man, which are summarized in Table 6.3. The densities of tissues/organs, which are mainly adopted from ICRP Publication 23, are also listed in Table 6.3.21 It should be noted that the suggested density for skeleton in ICRP Publication 23 is 1.4 g cm−3, which is an average value for all skeleton tissues. Alternatively, the density of bone in VIP-Man is 1.55 g cm−3 because RBM, cartilages, disks, and tendons have been separated from bone tissue. Chemical composition data of tissues/organs are also vital for Monte Carlo simulation and are taken from International Commission on Radiation Units and Measurements (ICRU) Report 44.26 When a tissue/organ is not available in ICRP 23 or ICRU 44, its density and chemical composition is assigned using values from similar tissues/organs. The approach of assigning uniform density and chemical composition for each organ is also used in other voxel phantoms.27 6.2.4 Visualization After being segmented and labeled, VIP-Man phantom was inspected by 2D and 3D visualization. To make sure each of these procedure was correctly done, one would need to construct and compare the newly constructed anatomy with that depicted by the original VHP-photograph. Furthermore, in order to check the continuity between each transverse slice, the images in axial view was reconstructed into images in sagittal and coronal views. For example, Figure 6.1a shows a segmented slice in coronal view that assures the continuities of skin, muscle, bone, lung, liver, heart, etc. This comparison should be repeated slice by slice for the whole data set, 3D surface rendering is more efficient. The 3D visualization of VIP-Man can be utilized to assure that the shape and relative location of each organ is correct. Two types of 3D rendering are used in this study: 1. Volume rendering, which takes a set of voxels as input to render a 2D projection image from a 3D object. A C++ code was written to render the distance to surface of the rendered tissues/organs. 2. Surface rendering takes a set of surfaces as input to render a 2D projection image from a 3D object. Visualization toolkit (VTK)28 was used to generate the surface rendition for VIP-Man.

Brain Brain

38.06 891.96

122.70 681.37 440.49 1.43

1.55

1.55

2.10

1.20

0.26

0.26

1.04

1.04

1.04

1.04

1.06

1.06

1.04

1.03

1.03

Other

Teeth

Tendon

Bronchus R

Bronchus L

Caudate nucleus

Cerebellum

Cerebrum—gray

Cerebrum—white

Coronary L

Coronary R

Corpus callosum

CSF, spinal

CSF, skull

7,193.05

97.75

167.31

16.92

1.41

8.95

1.73

1.59

1,097.18

–

–

–

–

Brain –

–

–

–

Bone –

Bone

Bone

Bone

Spine

87.15

1.55

1.55

Bone

Adrenals –

– –

ICRP 60 Organ

Cranium

806.79

854.70

0.18 328.97 8.28

Mass (g)

Mandible

Bone

1.02

1.06

Blood

1.21 × 10−3 1.21 × 10−3

Density (g cm−3)

Inside Outside Adrenals

Air

VIP-Man Organ

Male breast

Content

Wall

Mucosa

LLI

Lung

Liver

Lentiform nucleus

Lenses of eyes

Lateral ventricle

Kidneys

Heart wall

Bile

Wall

Gallbladder

Fronix

Fat

Eye

Lumen Mucosa Wall

Esophagus

VIP-Man Organ

Organs Used in VIP-Man and Their Corresponding Names Defined in ICRP 60

TABLE 6.3

0.92

1.04

1.04

1.04

0.26

1.05

1.07

1.10

1.03

1.05

1.03

1.03

1.03

1.04

0.92

1.03

1.04 1.04 1.04

Density (g cm−3)

33.56

324.24

290.84

35.75

910.51

1,937.93

13.41

0.54

7.08

335.37

398.71

20.98

11.96

2.22

36,326.63

14.91

26.75 3.47 38.87

Mass (g)

(continued)

Breast

–

Colon

Colon

Lung

Liver

–

Lenses

–

Kidneys

–

–

–

–

–

–

– Esophagus Esophagus

ICRP 60 Organ

The 3D and 4D VIP-Man Computational Phantoms 145

18.91

1.05

1.04

1.05

1.04

0.98

1.10

1.04

1.06

Pancreas

Pons and middle cerebellar peduncle Prostate

Rectum

RBM

Skin

Small intestine

Spleen

1.04

Mucosa

51.54

82.86

1.75

13.72

324.53

244.04

1,291.77

2,253.43

1,130.44

Source: Wells, J., Br. J. Radiol. Suppl., 19, 146, 1986.

1.04

Content

Stomach

24.57

1.04

Optic nerve

0.33

1.04

Optic chiasma

43,002.59

Mass (g)

1.04

Density (g cm−3)

Muscle

VIP-Man Organ

Stomach

–

Spleen

SI

Skin

RBM

–

–

–

Pancreas

–

–

Muscle

ICRP 60 Organ

Vestibulocochlear

Content

Wall

Urinary bladder

Content

Mucosa

Wall

ULI

Trachea

Thyroid

Thymus

Thalamus

Testes

Wall

VIP-Man Organ

Organs Used in VIP-Man and Their Corresponding Names Defined in ICRP 60

TABLE 6.3 (continued)

1.04

1.04

1.02

1.04

1.04

1.04

0.26

1.05

1.03

1.04

1.04

1.04

Density (g cm−3)

0.07

43.22

41.38

905.70

63.36

461.12

7.32

27.56

11.22

8.07

21.00

159.52

Mass (g)

–

–

Bladder

–

Colon

Colon

–

Thyroid

Thymus

–

Gonads

Stomach

ICRP 60 Organ

146 Handbook of Anatomical Models for Radiation Dosimetry

The 3D and 4D VIP-Man Computational Phantoms

147

As shown in Figure 6.1b, volume rendering has the advantage to represent the tiny and complicated structures such as RBM. However, this rendition does not promise good depth perception. Alternately, as shown in Figure 6.2, surface rendering offers stereo rendition but omits tiny structures because a lot of smooth filters are involved. Figure 6.3 shows the comparison between 3D images of the Medical Internal Radiation Dose (MIRD)-type stylized phantom and VIP-Man. Compared to VIP-Man, the stylized phantom appears to be crude in representing the realistic shapes of the GI tract.

(a)

(b)

(c)

FIGURE 6.2 (See color insert following page 524.) VIP-Man in 3D views showing (a) whole-body skin and skeletal structure; (b) details of internal organs with lungs in red, stomach in gold, ULI in purple, kidney in red, liver in maroon, LLI in brown, etc.; and (c) details of the head and brain containing skull in gold, white matter in white, gray matter in gray, nerve in blue, spinal cord in gold, thyroid in red, skin in white, etc. VTK was used in the surface rendering of the voxelized images.

Thyroid Esophagus Heart Lungs Liver Stomach Small intestine Colon Rectum

FIGURE 6.3 The GI tract in the mathematical model is too simple to represent the very twisted GI tract compared to VIP-Man.

148

Handbook of Anatomical Models for Radiation Dosimetry

6.2.5 Implementation of VIP-Man into Monte Carlo Codes Due to its high resolution, implementation of VIP-Man into Monte Carlo codes required a huge amount of random accessible memory (RAM). In 1998, the maximum “useable” RAM for a typical personal computer (PC) was less than 1 GB and it was impossible to handle the size of VIP-Man, which contains a total of about 3.7 billion voxels and additional coding. A significant amount of effort was required to reduce the memory burden by using an innovative look-up table (LUT) algorithm. The LUT algorithm was successfully implemented in the Monte Carlo code EGS4, allowing the computer to store only the key anatomical and physical data; the details are unfolded from specific tables when needed. The memory saving with the LUT algorithm in VIP-Man/EGS4 is about a factor of 20.29 In 2001, VIPMan/EGS4 was run using a PC of 512-MB RAM at the original 0.33 mm × 0.33 mm × 1 mm voxel size. This made VIP-Man/EGS4 the “finest” voxel-based phantom ever developed for Monte Carlo calculations at that time. MCNP and Monte Carlo N-Particle eXtended (MCNPX), on the other hand, proved to be much harder to use for this phantom. The default code options had to be changed to optimize memory. Even with these improvements, however, the voxel size of VIP-Man/MCNP/X had to be compromised to 4 mm × 4 mm × 4 mm (or about 6 million voxels for the whole body) in order to run it on the same PC at that time. Others were reportedly able to handle a head computational phantom of 65 million voxels in MCNP4A using the Advanced Simulation and Computing Program (ASCI) Blue Mountain supercomputer (over 6000 Parallel central processing units [CPUs] from Silicon Graphics, Inc. [SGI]) at Los Alamos National Laboratory.30 In a later study, various attempts were made to increase the total number of voxel in the MCNP with limited success.31 Therefore, VIP-Man and other voxel phantoms from the Chinese VHP (see Chapter 11) that have original voxel size of 0.33 mm or less will continue to encounter the problem with the MCNP/X codes. Although the resolution for VIP-Man/MCNP/X is limited by the current computer technologies, VIP-Man/MCNP/X was the first voxel-based computational phantom constructed for neutron and proton dose calculations. All of our calculations were performed on PCs operated under Linux environment, which is a complete operating system that is similar but not identical to Unmultics Information and Computing System (UNIX). The Parallel Virtual Machine (PVM) in Linux has enabled us to use multiple CPUs for very time-consuming tasks. Compilers, such as g77, had to be used in EGS4 to accommodate the large integral format.29 Since both EGS4 and MCNP4B codes could transport photons and electrons, we were able to “benchmark” the modeling and Monte Carlo coding by making sure both codes yielded the same results for VIP-Man (at an identical voxel size of 4 mm × 4 mm × 4 mm resolution). A simple comparison was performed between organ doses calculated using different Monte Carlo codes for 1 MeV parallel photon beams under anterior–posterior (AP) irradiation. At the time this work was performed, the calculations took about 50 h for 10 million photons in MCNP and about 25 h for 25 million photons in EGS4 (on a 450 MHz Pentium II PC of 512 MB RAM). Both codes tracked electrons by different transport algorithms with carefully optimized electron step settings. Results indicate remarkable agreement within the statistical uncertainty between EGS4 and MCNP versions of the VIP-Man phantom.29 6.2.6 Extending 3D VIP-Man into 4D VIP-Man The sections above described the processes to develop the 3D, static VIP-Man computational phantom. For certain dosimetry applications, however, it is useful to consider organ

The 3D and 4D VIP-Man Computational Phantoms

149

motions and changing human postures that are dependent on the fourth dimension. In external beam radiation treatment of lung cancer, the treatment plan must take into account the degree and pattern of patient breathing. For example, to reconstruct a nuclear accident, the dose assessment will depend on how a worker moves around the radiation source. A time-dependent, 4D VIP-Man computational phantom was quite a challenge but at the same time it would open the door for interesting applications. The following section describes a method to extend the 3D VIP-Man into 4D by adding respiration-caused organ motions for the purpose of external beam treatment planning for the lung lesion.32,33 A 4D respiration-simulating anatomical computational phantom is a series of computergenerated 3D phantoms whose shape, size, and location change according to specified respiratory motion patterns. 6.2.6.1 Conversion from Voxel Data to Surface Deﬁnition The VIP-Man whole-body phantom contains voxels at a size of 0.33 mm × 0.33 mm × 1 mm. Figure 6.3 shows the internal organs of the 3D VIP-Man computational phantom adopted for this study. Organs, each originally defined as a group of voxels, were converted into polygon surfaces of anatomical features. Polygon models were then translated into nonuniform rational B-spline (NURBS) surfaces.34 Organ surfaces defined using NURBS could be easily deformed by changing control points as first demonstrated by Segars et al.35–37 and later by Xu and Shi32 and Zhang et al.33 The NURBS approach was also used to manually adjust shape/size for a series of pediatric computational phantoms.38 For this study, a commercial program called Rhinoceros (McNeel North America, Seattle, WA) was used to create, edit, analyze, and translate NURBS curves, surfaces, and solids. The VIP-Man organ files were changed into a format readable by Rhinoceros using a free software program called vtkEditor (http://www.esat.kuleuven.ac.be/∼vtkedit). Once the organs were imported into Rhinoceros, organ contours were regenerated and lofted into 3D NURBS surfaces. At that time, we performed this process on the lungs, heart, skin, rib cage, spine, kidneys, stomach, spleen, and liver. After generating the NURBS files, control points for each organ were obtained by exporting the NURBS organ file to a text file. These control points contain basic anatomical features of the original 3D VIP-Man computational phantom. 6.2.6.2 Deformation of the Organ Surfaces To simulate time-dependent deformation caused by the respiratory motion, the control points for each of the interlinked organs were transformed using a rigid motion defined in Equation 6.3 C new = S × R × C old + T where Cold is a 3 × N matrix defining the original control points Cnew is the 3 × N matrix of the translated points N is the number of one organ’s control points S is a scalar matrix R is a 3 × 3 matrix which defines rotation T is a 3 × N matrix which defines translation

(6.3)

150

Handbook of Anatomical Models for Radiation Dosimetry

By using parameters that are functions of time, t, in matrices S, R, and T, the control points are extended from the original 3D space into 4D. In this study, the respiratory motion patterns were based on the clinical data described previously by Segars et al.36 The rotation angle of the rib cage around its intersection with the spine was defined as a piecewise linear function with respect to time, t, as shown in Equation 6.4. Δθ × t ⎧ ⎪ θ = ⎨2 ⎪⎩ 3 × Δθ × (5 − t )

0≤t < 2 2≤t < 5

(6.4)

where we selected the angular step, Δθ, to be 2.5° and the entire respiration cycle to last 5 s. The first 2 s were used for the inspiration and the last 3 s were for the expiration. The positions of the sternum and skin were defined according to the positions of the rib tips. The lungs were modeled by referring to the position of the fifth rib in the transverse direction and the top of the liver in the longitudinal direction. For other organs, such as liver, stomach, heart, kidneys, and a lesion in the left lung, we first created a normalized motion curve as shown in Equation 6.5. ⎧ 1⎛ ⎛ π ⎞⎞ 1 − cos ⎜ × t ⎟ ⎟ ⎪ ⎜ ⎝ 2 ⎠⎠ ⎪ 2⎝ x=⎨ ⎪ 1 ⎛ 1 − cos ⎛ π × (5 − t )⎞ ⎞ ⎜⎝ ⎟⎠ ⎟ ⎪ 2 ⎜⎝ ⎠ 3 ⎩

0≤t < 2 (6.5) 2≤t < 5

where x is the normalized motion distance t is the time The magnitude of such organ motions was then controlled by multiplying different amplifying factors in the x, y, and z directions, respectively. We used MATLAB® (The MathWorks, Inc., Natick, MA), version 6.5, to develop a software program to automate the calculations and a NURBS toolbox (http://www.aria.uklinux.net/nurbs.php3) for MATLAB was used for the final integration of the 4D VIP-Man computational phantom. The reconstructed organ and body surfaces using the NURBS are shown in Figure 6.4, which displays the frontal and side views of the skin, lungs, heart, rib cage, spine, and liver.33 6.2.6.3 Revoxelization of the 4D Computational Phantom for Each Respiratory Phase For the calculation of radiation dose distributions under each of the respiratory phases, the 4D phantom was treated as a combination of a series of 3D phantoms representing the anatomy at different points of the respiratory cycle. Once the 4D VIP-Man phantom was constructed, each of the 3D phantoms was re-created from the 4D phantom for Monte Carlo calculations. This process was accomplished by converting the NURBS surfaces back to voxels for each given respiratory phase. The control points for an organ were saved in a 3D matrix and the contours were calculated by specifying the cutting plane coordinate in MATLAB. A total of eight respiration phases, each having 70 2D-slices, were sampled and revoxelized to represent the entire respiratory cycle in eight phases: peak exhale,

The 3D and 4D VIP-Man Computational Phantoms

(a)

151

(b)

FIGURE 6.4 (a) Front view and (b) side view of the reconstructed NURBS surfaces in phase 1 (middle inhale phase).

early inhale, middle inhale, late inhale, peak inhale, early exhale, middle exhale, and late exhale.33 The resolution for each of the revoxelized computational phantoms is 2.1 mm × 1.2 mm × 6.0 mm. Finer resolutions can be obtained if the 2D-slice size and cutting plane number are increased.

6.3 Comparison between VIP-Man and Other Computational Phantoms VIP-Man is first compared with the ICRP Reference Man21 in terms of organ mass. As listed in Table 6.2, VIP-Man is heavier, taller, with much more fat and muscle, and has heavier GI tract tissues compared to the Reference Man. For other internal organs, the mass differences between VIP-Man and the Reference Man are less significant and, in particular, the volume of the lungs is slightly smaller as expected from the cadaver images. Comparison of detailed anatomical information in VIP-Man with stylized mathematical computational phantoms was one of the motivations of this study. It is very obvious that the stylized mathematical computational phantoms have very simplified shapes for the GI tract and stomach. The relative locations of the stomach, large intestine, and thyroid are also quite different from those in the VIP-Man. VIP-Man also contains many tissues such as RBM, GI tract mucosa, teeth, gray/white matter of cerebrum, optic chiasma, and vestibulocochlear nerve that were previously not well defined (or not available) in stylized mathematical computational phantoms. Compared with other tomographic computational phantoms developed prior to 2006, VIP-Man has a smaller voxel size which allows some of the small organs to be more accurately defined. Our experiences with the VIP-Man phantom have made us recognize that, because all tomographic phantoms are developed from medical images taken from

152

Handbook of Anatomical Models for Radiation Dosimetry

real individuals, they are necessarily different in anatomy. As such, a composite of these computational phantoms will have to be used to represent a population of workers.

6.4 Applications of the VIP-Man Computational Phantom Since 2000, the group at RPI has used the VIP-Man phantom for a variety of radiological studies in health physics, diagnostic imaging, and radiotherapy. These studies are briefly summarized below. 6.4.1 Health Physics Health physics dosimetry typically involves organ dose and effective dose quantities for external and internal sources under standard irradiation conditions. The VIP-Man phantom offered an opportunity to perform quantitative comparison of these quantities with those derived previously from other phantoms. These studies were not meant to establish VIP-Man as a standard reference phantom. Instead, we were interested in demonstrating how similar or different the dosimetry data from VIP-Man, a large-sized individual of 40 years old, would be with respect to the ICRP Reference Man. Details of these studies have been reported for different radiation types including photons,29,39 electrons,40,41 neutrons,25,42 and protons.43 Here we provide brief descriptions. 6.4.1.1 External Photon Dosimetry Using the VIP-Man phantom, Chao et al. reported a new set of conversion coefficients from kerma free-in-air to absorbed dose and kerma free-in-air to “effective VIP-Man dose” for external monoenergetic photon beams from 10 keV to 10 MeV.30 (A correction for the reported data was subsequently published.44) This study noted that kerma approximation, which assumes secondary electrons from photon interactions to deposit their energies at an interaction site, could lead to potential uncertainty for high-energy photons incident on shallow tissues (such as breast, skin, eye lenses, or gonads). Photon sources considered in Chao et al. were monoenergetic parallel beams with energies from 10 keV to 10 MeV.30 The irradiation geometries included AP, posteroanterior (PA), left lateral (LLAT), right lateral (RLAT), rotational (ROT), and isotropic (ISO). For all sources, the photons were generated on a fixed plane as broad parallel beams. Results from this study were tabulated and compared with those obtained from the stylized computational phantoms, ADAM and EVA involving a GSF Monte Carlo code. The “effective VIP-Man doses” differed from the previously reported data by 10%–50% for photons between 100 keV and 10 MeV. The discrepancies were greater for lower energies and for individual organ doses. The study concluded that the size of computational phantom, kerma approximation, and the anatomical difference were three main factors in causing dosimetric discrepancies. These comparisons also suggested possible ways to improve the stylized phantoms. For example, the stomach is situated too close to the left side of the body compared to VIP-Man. The liver is situated too close to the right side of the body compared to VIP-Man, and the esophagus is too close to the back. VIP-Man has the smallest voxel size among existing computational phantoms for many years until the Chinese VHP generated several cadaver image sets that are 0.2 mm in thickness using specialized milling machines.45,46

The 3D and 4D VIP-Man Computational Phantoms

153

Although VIP-Man has its own limitations, it is the representation of a real person and contains many tissues that were previously not well defi ned (or not available) in other computational phantoms. 6.4.1.2 External Electron Dosimetry Compared to photon dosimetry, electron dosimetry is complicated because electrons undergo a large number of interactions when passing through materials. For radiation protection purposes, tables of conversion coefficients are needed to estimate the dose from the measurable field quantities, such as the particle fluence. Only a limited number of researchers have published organ dose conversion coefficients for electrons calculated from mathematical computational phantoms. The approach of using standard and simplified computational phantoms based on the Reference Man works well for penetrating radiation, such as photons and neutrons. Indeed, comparison of photon doses derived from different phantoms have showed impressive agreement.42 However, the uncertainty caused by the anatomical computational phantoms for electrons can be potentially very high because slight differences between body computational phantoms can cause dramatic dosimetric deviations for less-penetrating radiation. Chao et al.40 presented the results of organ dose calculations on the VIP-Man which is a taller and heavier computational phantom than the Reference Man. Organ doses for 10 MeV AP electrons are compared between results from VIP-Man using EGS4-VLSI Monte Carlo code, ADAM using MCNP4,47 hermaphrodite phantom using FLUKA,48 and MIRD-5 phantom using EGS4.49 For shallower organs such as skin and thyroid, the deviation between these results is about 80%. For deeper organs such as lung, stomach, bladders, liver, and esophagus, significant differences more than an order of two can be observed. The differences between VIPMan and the MIRD-based stylized phantoms suggest possible range of errors caused by using the MIRD-based stylized computational phantoms on a specific individual as large in body size as the VIP-Man. The study noted that Schultz and Zoetelief47 used thymus dose to substitute esophagus dose, which seemed to be a poor assumption. The lung dose reported by Ferrari et al.48 are much lower than those obtained from other phantoms because of the extra shielding by female breasts, which were inserted into a male hermaphrodite computational phantom used by the authors. This study suggested that a careful consideration is necessary before using hermaphrodite mathematical phantom for electron dosimetry. These comparisons suggest again that, at least for electron dosimetry, a single standard body computational phantom does a very poor job in representing individuals of diverse anatomy. The study further concluded that it was clear that a large number of voxel phantoms would need to be investigated before the degree of dose variation was understood. 6.4.1.3 External Neutron Dosimetry Using the VIP-Man phantom, Bozkurt and his coauthors reported a new set of fluenceto-absorbed dose and fluence-to-effective dose conversion coefficients calculated for both low-energy (10−9 to 20 MeV) and high-energy (20–10,000 MeV) neutrons.25,42 Organ dose calculations were performed using the Monte Carlo code MCNPX for 20 monoenergetic neutron beams under six different irradiation geometries: AP, PA, RLAT, LLAT, ROT, and ISO. The absorbed dose for 24 major organs and effective dose results based on the realistic VIP-Man were presented and compared with those based on the simplified MIRD-based phantoms reported in the literature. For high-energy neutrons, although VIP-Man has

154

Handbook of Anatomical Models for Radiation Dosimetry

detailed and realistic anatomical representations, neutron effective dose results did not vary significantly from the values obtained from much simpler MIRD-5 stylized phantoms for this range of energies. However, individual organ doses from VIP-Man were considerably different, suggesting that the existing neutron dose data need to be reevaluated at the organ levels. For low-energy neutrons, results in this study yielded up to 40% larger values for the effective dose, and for many organ doses, thus suggesting that the results reported in the past may not be conservative. The authors concluded that several factors might have contributed to the discrepancies: the differences in anatomical computational phantoms which cause around 10% difference in effective dose because VIP-Man is heavier and taller, and how the Monte Carlo codes treat the transport of high-energy particles, including the use of evaluated data and theoretical computational phantoms. 6.4.1.4 External Proton Dosimetry Bozkurt and Xu43 applied the VIP-Man phantom to calculate fluence-to-absorbed dose and fluence-to-effective dose conversion coefficients under high-energy proton environment. These organ dose calculations were performed using the Monte Carlo code MCNPX under six different irradiation geometries: AP, PA, LLAT, RLAT, ROT, and ISO and for 10 monoenergetic proton beams between 20 and 10,000 MeV. The absorbed dose results were presented for 24 major organs of VIP-Man and the calculated data were compared with those based on mathematical phantoms reported in the literature. Some discrepancies in organ dose and effective dose were observed which were within 40% due to the use of different transport computational phantoms employed by different Monte Carlo codes. 6.4.1.5 Internal Electron Dosimetry Based on the VIP-Man phantom, Chao and Xu calculated, for the first time, complete sets of specific absorbed fractions (SAF) for internal electron emitters.41 Electron emitters with energies from 100 keV to 4 MeV were studied and the results provided a set of complete dosimetry data for protection against internal electron exposures. This was also the first time to report internal electron data for walled organs such as esophagus, LLI, stomach, and ULI. Although electrons are considered as weakly penetrating radiation and researchers have usually ignored the dose to organs other than the source organ, results from this study shows that doses to neighbor organs and nearby organs can be too great to be neglected. In examining the effect of the partial penetrating of electrons in internal dosimetry, the authors classified the target organs into four categories according to the levels of absorbed dose. (1) The “highest target” for the organ receiving the highest dose which was usually the source organ itself. From the radiotherapeutic point of view, dose about a few tens of Sv is usually prescribed to the highest target. (2) On the other hand, compared to dose limits recommended by ICRP 60,22 a few tens of mSv of organ dose is critical to protect normal organs against radiation. Therefore, the organ receiving a dose less than 0.1% of the dose in the highest target was classified as the “irrelevant organ.” (3) If the SAF in the target organ was larger than 1% of SAF in the highest target, it was classified as the “neighbor target.” (4) Finally, the “nearby targets” were those receiving doses between 0.1% and 1% in the highest target. Chao and Xu41 also tabulated the neighbor targets and the nearby targets for 26 source organs emitting electrons from 100 keV to 2 MeV. This study provided convincing evidence that internal electrons do affect organs beyond the source organ.

The 3D and 4D VIP-Man Computational Phantoms

155

6.4.1.6 Internal Photon Dosimetry for the GI Tract In this study, the VIP-Man phantom was used by Chao and Xu to calculate SAFs for the GI tract.39 SAFs for sources in GI tract have been previously studied based on stylized phantoms. However, there are three shortcomings in these previous calculations: (1) the definition of GI tract in the stylized phantom, including those by Cristy and Eckerman17 and Poston et al.27 were anatomically simplified. (2) These stylized phantoms, including Smith et al.50 did not model the mucosal layer, resulting in a potential uncertainty in the assessed risk to the stomach. (3) The secondary electrons were often not considered in these Monte Carlo calculations because the range of secondary electrons was assumed to be too short. In this study using the VIP-Man phantom, the authors compared SAFs for the stomach wall from VIP-Man to those previously published by Cristy and Eckerman17 for photon sources in the stomach content. SAFs of stomach wall in the Cristy and Eckerman phantom is practically the same as those from VIP-Man, peaking at about 20 keV and leveling off at 100 keV and above. These results suggest that, for photon sources in the stomach content, the dose to the stomach wall is not dependent on the phantom (mass or anatomy). This is perhaps easy to understand for penetrating photon sources in the centrally located stomach content. On the other hand, the data line for the stomach mucosa as the target in VIP-Man is clearly above the data line for the stomach wall and the trend decreases as the energy increases from 10 to 100 keV. This suggests that, even for penetrating photons, the risk to the stomach may have been significantly underestimated (by about 60% at 20 keV) if the stomach wall is the only target under consideration, as was often done in the past. Results also show that the dose to the mucosa is much higher than dose to the wall for all the organs of the GI tract causing the possibility of underestimating the risk. This study clearly demonstrated the advantage of the VIP-Man phantom whose small voxel size allowed the dosimetry to be performed on small tissues structures, such as the mucosal layer in the GI tract. 6.4.1.7 RBM Dosimetry for External Irradiations In his doctoral research at RPI, Dr. Caracappa used two sets of Visible Human images for the same cadaver to develop an algorithm for assessing the dose to the RBM from external photon and electron sources.51 Dosimetry phantoms commonly in use assume that the marrow space throughout the body consists of a uniform mixture of active and inactive bone marrow. In reality, however, bone sites in different parts of the body are known to consist of varying combinations of active and inactive bone marrow. A Monte Carlo phantom was constructed in this study from the CT images of the VHP, and compared to the VIP-Man phantom, which was derived from color photographs of the same individual. These two data sets for the same individual offered interesting information that was not available elsewhere. RBM doses were calculated for the CT phantom using the uniform mixture assumptions and the cellularity factors adopted by ICRP. The goal was to test the previous assumptions and evaluate the accuracy of the computed doses in Monte Carlo simulations. Based on the newly developed algorithms, three dosimetry applications were investigated and tested. Broad beam photon irradiation in occupational exposure results in similar doses for high energies, but differences as great as 40% for low energies. In nonuniform photon exposures from selected CT examinations, the differences in the computed marrow dose are significant, 25% and 33% for the two cases modeled. An electron total body irradiation procedure for treating skin cancer is also studied, with a 39% difference in RBM dose between the existing method and the proposed revised method. These

156

Handbook of Anatomical Models for Radiation Dosimetry

results demonstrate the advantage of the new algorithms by accounting for marrow cellularity and distribution various bone sites in the anatomical and dosimetry computational phantoms. 6.4.2 Applications in Radiological Imaging 6.4.2.1 Organ Doses from SPECT and PET Brain Imaging To estimate internal dosimetry for brain imaging, a head and brain portion of the VIPMan was used to implement into the Monte Carlo code, EGS4-VLSI.52 Fifteen subregions were modeled including caudate nucleus, cerebellum, cerebral cortex, cerebral white matter, corpus callosum, eyes, lateral ventricles, lenses, lentiform nucleus, optic chiasma, optic nerve, pons and middle cerebellar peduncle, skull CSF, thalamus, and thyroid. This is the first time that several small structures are modeled for nuclear medicine dosimetry purposes. S-values were calculated for the most important sources and targets encountered in single photon emission computed tomography (SPECT) and positron emission tomography (PET) brain imaging. These results were then compared to those from the stylized head/brain phantom recommended by the MIRD.53 Although heavier individuals (such as VIP-Man) will usually receive lower radiation doses, however, the stylized head/brain phantom underestimates the S-values by 15% on average for a patient similar to the VIPMan phantom. More voxel head/brain phantoms are needed in order to compare various brain sizes and anatomical variations. Before such an intercomparison is performed, the results presented in this chapter are useful for patients who are similar to VIP-Man in body size or weight. 6.4.2.2 Organ Doses from X-Ray Radiographs VIP-Man was used by Mark Winslow, a PhD student at RPI, in collaboration with Dr. Walter Huda from University of Syracuse, to calculate values of energy imparted (ε) and effective dose (E) for monoenergetic photons (30–150 keV) in radiographic examinations. Energy deposition in the organs and tissues of the human phantom were obtained using Monte Carlo simulations. Ratios of effective doses to dose imparted (E/ε) were obtained for three common projections: AP, PA, and LAT of head, cervical spine, chest, and abdomen, respectively. For head radiographs, all three projections yielded similar results. At 30 keV, E/ε was 1.6 mSv J−1, which increased to 7 mSv J−1 for 150 keV photons. The AP cervical spine was the only projection investigated where the E/ε decreased with increasing photon energy. Above 70 keV, cervical spine E/ε showed little energy dependence and ranged between 8.5 mSv J−1 for PA projections and 17 mSv J−1 for AP projections. The values of E/ε for AP chest examinations showed very little variation with photon energy, and had E/ε of 23 mSv J−1. Values of E/ε for PA and LAT chest projections were substantially lower than the AP projections and increased with increasing photon energy. For abdominal radiographs, differences between PA and LAT projections were very small. All abdomen projections showed an increase in the E/ε ratio with increasing photon energy, and reached a maximum value of 13.5 mSv J−1 for AP projections, and 9.5 mSv J−1 for PA/lateral projections. These monoenergetic E/ε values can generate values of E/ε for any x-ray spectrum, and can be used to convert values of energy imparted into effective dose for patients undergoing common head and body radiological examinations.54

The 3D and 4D VIP-Man Computational Phantoms

157

6.4.2.3 Image Quality Optimization in Radiograph In his doctoral research, Mark Winslow, in collaboration with Professor Birsen Yazici of RPI, studied image optimization by examining the organ doses and image quality by analyzing approximately 2000 simulated chest x-ray images for the VIP-Man using the receiver operator characteristic/area under the ROC curve (ROC/AUC) analysis55,56 (see Chapter 23 for more information on that particular application). 6.4.2.4 Organ Doses from Interventional Cardiological Examinations After graduating from RPI, Dr. Bozkurt and his colleagues57 used the VIP-Man phantom to simulate both the patient and the physician during an interventional cardiological examination procedure. The patient was lying on the operation table while the physician was standing 15 cm from the patient. Organ equivalent doses and effective doses for both individuals were calculated simultaneously involving seven different x-ray spectra (between 60 and 120 kVp). The calculations were performed using the Monte Carlo code MCNP. The patient’s heart was irradiated by x-rays generated by a point source of a conical distribution. Five major projections typical to a coronary angiography examination were investigated. The authors reported the mean effective doses from 0.092 to 0.163 mSv/(Gy cm 2) for the patient and from 0.027 to 1.153 μSv/(Gy cm 2) for the shielded physician. The effective doses for the patient determined in the study were found to be lower than those reported in the literature partially because the VIP-Man phantom ignores possible higher ovarian dose contribution to the gonadal equivalent dose. On the other hand, the effective doses determined for the physician agreed reasonably well with the literature data. 6.4.3 Applications in Radiotherapy 6.4.3.1 External Beam Selection in Prostate Radiation Treatment The abdominal portion of the VIP-Man phantom was used in the doctoral research by Brian Wang at RPI to develop and demonstrate an Adjoint Monte Carlo (AMC) method for optimizing the external beam directions in the so-called 3D conformal radiation treatment of the prostate cancer.31 The AMC method had been widely used in nuclear reactor physics research but was never demonstrated for treatment planning in realistic 3D patient anatomy. With the VIP-Man phantom which was already implemented in the MCNP code that has multigroup adjoint cross sections, it was possible to test the theory in clinically relevant scenarios. This study was in collaboration with Dr. Moshe Goldstein, a nuclear engineering from Israel who first proposed the method during a sabbatical at ORNL, and Dr. Narayan Sahoo who was a clinical therapeutic physicist at Albany Medical Center. In this application, the adjoint fluxes for the prostate (Planning Target Volume [PTV]) and the rectum and bladder (organs at risk [OARs]) in the VIP-Man phantom were calculated on a spherical surface of approximately 1 m radius, centered at the center of gravity of PTV.31 An “importance” ratio, defined as the PTV dose divided by the weighted OAR doses, was calculated for each of the available beamlets to select the best beam angles. Finally, the doses in PTV and OAR were calculated using the forward Monte Carlo method. The Pinnacle treatment planning system was used to generate dose volume histograms (DVHs) for the 3D plan with beam angles obtained from the AMC method and a standard six-field conformal radiation therapy plan. Results showed that the DVHs for the prostate from these

158

Handbook of Anatomical Models for Radiation Dosimetry

two plans are very similar, suggesting that both methods can deliver necessary dose to the PTV. However, DVHs for bladder and rectum were found to be superior for the AMC method. This study demonstrated the feasibility of the AMC method in optimizing external beam directions based on anatomical information in a 3D and realistic patient anatomy. The study also identified issues to be further addressed before this method could become clinically useful.31 6.4.3.2 Organ Doses from Proton Radiation Treatments In a separate study, doctoral student Brian Wang, worked with Drs. Harald Paganetti and Hongyu Jiang of MGH in Boston to adopt the VIP-Man computational phantom to assess organ doses (including the RBM dose) from secondary radiation in proton radiation treatment.31 More information about this application for organ dose assessment from secondary radiation in proton treatment can be found in Chapter 25. 6.4.3.3 Dosimetry for Respiration Management in IGRT After graduating from RPI, Dr. Chengyu Shi worked with RPI doctoral student Mr. Juying Zhang, to apply the respiration-simulating 4D VIP-Man phantom for image-guide radiotherapy (IGRT) of lung cancer.33 Detail information can be found in Chapter 26. 6.4.3.4 Organ Doses from Cone-Beam CT Imaging for IGRT Recently, RPI doctoral students Mr. Jianwei Gu and Mr. Bryan Bednarz, used the VIPMan phantom to calculate organ doses using a recent procedure associated onboard imaging to localize the patient in IGRT.58 As the IGRT becomes widely and frequently practiced in clinics, it is becoming clear that such an “imaging dose” to the patient, in addition to scattered radiation dose to the healthy organs, is no longer at a level that can be entirely excluded from the treatment planning. The VIP-Man phantom and the MCNPX code were used in this study to simulate two imaging procedures: kV and MV cone beam CTs (CBCTs). The results indicate that thyroid receives the highest dose in head and neck scans for both kV and MV CBCTs, and the bladder receives the highest dose in prostate scan for both kV and MV CBCTs. The effective doses for H&N scan and for prostate scan are at the same level in both kV and MV CBCTs. This study provided a method to compute organ doses and effective dose that are useful in treatment planning and risk assessment.

6.5 Summary VIP-Man is a voxel-based computational phantom constructed in late 1990s from segmented color photographic images of the adult male from the VHP. The motivation of this development was to make available an anatomical realistic phantom that had better voxel resolution and more anatomical information than those developed earlier. Since the segmentation was based on color cadaver images of extra fine resolution, for several years, VIP-Man represented the world’s finest and most complete human anatomical computational phantom, containing small tissues, such as skin, GI tract mucosa, eye lenses,

The 3D and 4D VIP-Man Computational Phantoms

159

and RBM. The phantom was adopted into the state-of-the-art Monte Carlo codes, EGS4, MCNP, and MCNPX (and also GEANT4) for radiation transport studies involving nearly all radiation types of interest in radiological science such as photons, electrons, neutron, and protons. This allowed a systematical investigation on the potential errors the radiation protection dosimetry caused by a person-specific phantom. The work by Xu et al.7 contained a lot of detail that allowed readers to follow the procedure to construct the imagebased computational phantoms, especially from cadaver images later available in Korea and China. The VIP-Man phantom was used in many projects, such summarized in this chapter, to evaluate and compare a large number of important dosimetric quantities for external and internal sources under standard radiation protection dosimetry irradiation conditions, as well as various applications to imaging and radiotherapy. For more information about projects based on the VIP-Man phantom, please visit http://www.rpi.edu/dept/ radsafe/public_html/. To date, the VIP-Man phantom has been shared with more than 40 users internationally. The work by Xu et al. reviewed most of the radiation protection dosimetry data derived from the VIP-Man phantom.39 In operational personnel dosimetry, uncertainties in dose measurements of up to 50% are common as a result of dosimeter energy calibration, positioning, environmental noise, etc. It seems that a 30% improvement in the estimate of effective dose from external photon exposures, as reported for the VIP-Man and other phantoms, suggest that the decade-long effort to develop a new class of voxel phantoms did not directly improve upon the radiation protection dosimetry parameters. For electrons and protons, the dose differences caused by the anatomy are too great to rely on a single voxel phantom, being either stylized or tomographic. The authors further noted that the use of the newly developed standard voxel male and female phantoms might not lead to any real improvement in dose estimate for a population.38 Instead, the authors suggested that a library of person-specific voxel phantoms would have to be considered in the future. The authors then went on to predict that, although the tomographic images indeed provided realistic anatomical information of human body, the voxelized geometry was not technically suitable for developing dosimetry standards due to the difficulty in adjusting the voxel data. In contrast, advanced surface modeling tools, such as nonuniform rational B-splines or computer-aided design, were flexible in designing population-adjustable and time-dependent phantoms. The authors stated that if the same amount of resources were invested in refining the existing stylized phantoms, one could have achieved the same degree of dosimetric precision. The significance of the voxel phantom development was that it opened the door for person-specific dosimetry—a philosophy that the ICRP has yet to fully embrace. Using the versatile VIP-Man phantom and associated tools, exiting projects were carried at RPI one after the other in the past 10 years. In retrospect, the VIP-Man phantom brought the opportunity for us at RPI to aggressively experiment. Although the VIP-Man was originally developed for radiation protection dosimetry studies, it was recognized early on that, by shifting from stylized phantoms to voxel-based data, it became possible to also perform patient-specific radiation dosimetry studies in the fields of diagnostic imaging and radiation treatment where more precise dose assessment was required. It is interesting to note that the VIP-Man phantom was also successfully adopted for a few applications in surgical simulations by coupling the data with novel biomechanical information.59 As predicted by Xu et al.,7 by combining the fine anatomical information in the VIP-Man with physical properties that are radiological, electrical, thermal, chemical, mechanical, or biological, the VIP-Man phantom would someday become a useful tool for multidisciplinary applications.

160

Handbook of Anatomical Models for Radiation Dosimetry

Acknowledgments The development of VIP-Man computational phantom was crucial in our earlier research efforts at RPI. The authors are grateful to the National Science Foundation/Biomedical Engineering Program for a Faculty Early Career Development Award (#BES-9875532), to the National Library of Medicine and Dr. Victor Spitzer for the Visible Human images, to Dr. Keith Eckerman at Oak Ridge National Laboratory for his insightful discussions during the planning and execution of the project. The VIP-Man phantom was subsequently used and refined by other students at RPI and we would like to acknowledge their works that went into this chapter: Dr. Mark Winslow who, in collaboration with Dr. Walter Huda and Dr. Birsen Yazici, studied image quality optimization for radiography; Dr. Brian Wang who, in collaboration with Dr. Moshe Goldstein and Dr. Narayan Sahoo, studied adjoint Monte Carlo -based external beam optimization, and then, in collaboration with Dr. Harald Paganetti and Dr. Hongyu Jiang, on proton treatment; Dr. Peter Caracappa who developed new algorithms for red bone marrow dosimetry, and finally, Mr. Jianwei Gu and Dr. Bryan Bednarz, in collaboration with Dr. Steve Jiang, studied organ doses from image-guided procedures involving cone beam CTs.

References 1. National Library of Medicine (NLM). U.S. Board of Regents. Electronic imaging: Report of the Board of Regents, NIH Publication 90-2197, NLM, Bethesda, MD, 1990. 2. Ackerman, M.J. Accessing the visible human project. D-lib Magazine: The Magazine of the Digital Library Forum. http://www.dlib.org/dlib/october95/10ackerman.html, 1995. 3. Chetty, I.J. et al. Report of the AAPM Task Group No. 105: Issues associated with clinical implementation of Monte Carlo-based photon and electron external beam treatment planning, Med Phys, 34, 4818, 2007. 4. Zubal, I.G. et al. Computerized three-dimensional segmented human anatomy, Med Phys, 21, 299, 1994. 5. Jones, D.G. A realistic anthropomorphic phantom for calculating organ doses arising from external photon irradiation, Radiat Prot Dosimetry, 72, 21, 1997. 6. Hickman, D.P. and Firpo, M. Magnetic resonance image phantom program, UCRL-MA-118455, Lawrence Livermore National Laboratory, Livermore, CA, 1997. 7. Zankl, M. et al. Organ dose conversion coefficients for external photon irradiation of male and female voxel models, Phys Med Biol, 47, 2367, 2002. 8. Caon, M., Bibbo, G., and Pattison, J. An EGS4-ready tomographic computational model of a 14-year-old female torso for calculating organ doses from CT examinations, Phys Med Biol, 44, 2213, 1999. 9. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Phys, 78, 476, 2000. 10. Bushong, S.C. Radiological Science for Technologists, 6th edn., Mosby-Year Book, New York, 1997. 11. Spitzer, V.M. and Whitlock, D.G. Atlas of the Visible Human Male, Jones and Bartlett Publishers, Sudbury, MA, 1998. 12. National Library of Medicine, Proceedings of the Second Visible Human Project Conference, Bethesda, MD, October 1–2, 1998. 13. Russ, J.C. Image Processing Hand Book, 3rd edn. CRC Press, Boca Raton, FL, 1998.

The 3D and 4D VIP-Man Computational Phantoms

161

14. Eckerman, K.F. and Stabin, M.G. Electron absorbed fractions and dose conversion factors for marrow and bone by skeletal regions, Health Phys, 78, 199, 2000. 15. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: The Skeleton, ICRP Publication 70, Pergamon Press, Oxford, 1995. 16. Snyder, W.S., Ford, M.R., and Warner, G.G. Estimates of specific absorbed fractions for photon sources uniformly distributed in various organs of a heterogeneous phantom, MIRD Pamphlet No. 5, revised, Society of Nuclear Medicine, New York, 1978. 17. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381/Volumes I–VII, Oak Ridge National Laboratory, Oak Ridge, TN, 1987. 18. ICRP. Limits for Intake of Radionuclides by Workers, ICRP Publication 30, Pergamon Press, Oxford, 1979. 19. Pizzarello, D.J. and Witcofski, R.L. Basic Radiation Biology, Lea and Febiger, Philadelphia, PA, 1967. 20. Poston, J.W. et al. A revised model for the calculation of absorbed energy in the gastrointestinal tract, Health Phys, 71, 307, 1996. 21. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Pergamon Press, Oxford, 1975. 22. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Pergamon Press, Oxford, 1991. 23. Wells, J. Problems associated with localised skin exposures, Br J Radiol Suppl, 19, 146, 1986. 24. ICRP. Human Respiratory Tract Model for Radiological Protection: A Report of a Task Group of the International Commission on Radiological Protection, ICRP Publication 66, Pergamon Press, Oxford, 1994. 25. Bozkurt, A., Chao, T.C., and Xu, X.G. Fluence-to-dose conversion coefficients from monoenergetic neutrons below 20 MeV based on the VIP-man anatomical model, Phys Med Biol, 45, 3059, 2000. 26. ICRU. Tissue substitutes in radiation dosimetry and measurement, ICRU Report 44, Bethesda, MD, 1989. 27. Dimbylow, P.J. The development of realistic voxel phantoms for electromagnetic field dosimetry, Proceedings of the Workshop on Voxel Phantom Development, Chilton, U.K., 1996. 28. Schroeder, W., Martin, K.W., and Lorensen, W. The Visualization Toolkit: An Object-Oriented Approach to 3-D Graphics, Prentice Hall PTR, Upper Saddle River, NJ, 1996. 29. Chao, T.C., Bozkurt, A., and Xu, X.G. Conversion coefficients based on the VIP-Man anatomical model and EGS4-VLSI code for external monoenergetic photons from 10 keV to 10 MeV, Health Phys, 81, 163, 2001. 30. McKinney, G.W. Voxelized Model for MCNP, personal communication, 1999. 31. Yang, L. et al. A simulation for effects of RF electromagnetic radiation from a mobile handset on eyes model using the finite-difference time-domain method, Conf Proc IEEE Eng Med Biol Soc, 2007, 5294, 2007. 32. Xu, X.G. and Shi, C.Y. Preliminary development of a 4D anatomical model for Monte Carlo simulations, Monte Carlo 2005 Topical Meeting: The Monte Carlo Method: Versatility Unbounded in a Dynamic Computing World, Chattanooga, TN, 2005. 33. Zhang, J. et al. Development of a geometry-based respiratory motion-simulating patient model for radiation treatment dosimetry, J Appl Clin Med Phys, 9, 2700, 2008. 34. Piegl, L. On NURBS—A survey, IEEE Comput Graphics Appl, 11, 55, 1991. 35. Garrity, J.M. et al. Development of a dynamic model for the lung lobes and airway tree in the NCAT phantom, IEEE Trans Nucl Sci, 50, 378, 2003. 36. Segars, W.P. Development and Application of the New Dynamic NURBS-Based Cardiac-Torso (NCAT) Phantom, PhD thesis, University of North Carolina at Chapel Hill, Chapel Hill, NC, 2001. 37. Segars, W.P. et al. Development of a 4-D digital mouse phantom for molecular imaging research, Mol Imaging Biol, 6, 149, 2004.

162

Handbook of Anatomical Models for Radiation Dosimetry

38. Lee, C. et al. A series of 4D pediatric hybrid phantoms developed from the UF series B tomographic phantoms, Med Phys, 33, 2006, 2006. 39. Xu, X.G., Chao, T.C., and Bozkurt, A. Comparison of effective doses from various monoenergetic particles based on the stylised and the VIP-Man tomographic models, Radiat Prot Dosimetry, 115, 530, 2005. 40. Chao, T.C., Bozkurt, A., and Xu, X.G. Organ dose conversion coefficients for 0.1–10 MeV electrons calculated for the VIP-Man tomographic model, Health Phys, 81, 203, 2001. 41. Chao, T.C. and Xu, X.G. Specific absorbed fractions from the image-based VIP-Man body model and EGS4-VLSI Monte Carlo code: Internal electron emitters, Phys Med Biol, 46, 901, 2001. 42. Bozkurt, A., Chao, T.C., and Xu, X.G. Fluence-to-dose conversion coefficients based on the VIPMan anatomical model and MCNPX code for monoenergetic neutrons above 20 MeV, Health Phys, 81, 184, 2001. 43. Bozkurt, A. and Xu, X.G. Fluence-to-dose conversion coefficients for monoenergetic proton beams based on the VIP-Man anatomical model, Radiat Prot Dosimetry, 112, 219, 2004. 44. Chao, T.C., Bozkurt, A., and Xu, X.G. Correction to conversion coefficients based on the VIPMan anatomical model and EGS4-VLSI code for external monoenergetic photons from 10 keV to 10 MeV, Health Phys, 84, 390, 2003. 45. Zhang, B.Q. et al. CNMAN: A Chinese adult male voxel phantom constructed from color photographs of a visible anatomical data set, Radiat Prot Dosimetry, 124, 130, 2007. 46. Zhang, G.Z., Liu, Q., and Luo, Q.M. Monte Carlo simulations for external neutron dosimetry based on the visible Chinese human phantom, Phys Med Biol, 52, 7367, 2007. 47. Schultz, F.W. and Zoetelief, J. Organ and effective doses in the male phantom ADAM exposed in AP direction to broad unidirectional beams of monoenergetic electrons, Health Phys, 70, 498, 1996. 48. Ferrari, A., Pelliccioni, M., and Pillon, M. Fluence to effective dose and effective dose equivalent conversion coefficients for electrons from 5 MeV to 10 GeV, Radiat Prot Dosimetry, 69, 97, 1997. 49. Katagiri, M. et al. Effective doses and organ doses per unit fluence calculated for monoenergetic 0.1 Mev to 100 MeV electrons by the MIRD-5 phantom, Radiat Prot Dosimetry, 90, 393, 2000. 50. Smith, T., Petoussi-Henss, N., and Zankl, M. Comparison of internal radiation doses estimated by MIRD and voxel techniques for a “family” of phantoms, Eur J Nucl Med, 27, 1387, 2000. 51. Caracappa, P.F., Chao, T.C., and Xu, X.G. A study of predicted bone marrow dose from external radiation exposures using two sets of image data for the same individual. Health Phys, 96, 661, 2009. 52. Chao, T.C. and Xu, X.G. S-values calculated from a tomographic head/brain model for brain imaging, Phys Med Biol, 49, 4971, 2004. 53. Bouchet, L.G. et al. A revised dosimetric model of the adult head and brain, J Nucl Med, 37, 1226, 1996. 54. Winslow, M. et al. Use of the VIP-Man model to calculate energy imparted and effective dose for x-ray examinations, Health Phys, 86, 174, 2004. 55. Son, I.Y. et al. X-ray imaging optimization using virtual phantoms and computerized observer modelling, Phys Med Biol, 51, 4289, 2006. 56. Winslow, M., Xu, X.G., and Yazici, B. Development of a simulator for radiographic image optimization, Comput Methods Programs Biomed, 78, 179, 2005. 57. Bozkurt, A. and Bor, D. Simultaneous determination of equivalent dose to organs and tissues of the patient and of the physician in interventional radiology using the Monte Carlo method, Phys Med Biol, 52, 317, 2007. 58. Gu, J. et al. Assessment of patient organ doses and effective doses using the VIP-Man adult male phantom for selected cone-beam CT imaging procedures during image guided radiation therapy, Radiat Prot Dosimetry, ncn200, 2008. 59. Jin, W. et al. Improving the visual realism of virtual surgery, Proceedings of Medicine Meets Virtual Reality 13, Long Beach, CA, 2005.

7 The FAX06 and the MAX06 Computational Voxel Phantoms Richard Kramer, Helen Jamil Khoury, José Wilson Vieira, Vanildo Júnior de Melo Lima, Eduardo César de Miranda Loureiro, Gabriela Hoff, and Iwan Kawrakow

CONTENTS 7.1 Introduction ............................................................................................................... 163 7.2 Materials and Methods ............................................................................................ 165 7.2.1 Voxel Phantoms ............................................................................................ 165 7.2.2 The MAX Phantom ...................................................................................... 166 7.2.2.1 Database......................................................................................... 166 7.2.2.2 Adjustment of Organ and Tissue Masses ................................. 168 7.2.3 The FAX Phantom ........................................................................................ 168 7.2.3.1 Database......................................................................................... 168 7.2.3.2 Segmentation ................................................................................ 168 7.2.3.3 Addition of Head and Arms ....................................................... 170 7.2.3.4 Anatomical Corrections .............................................................. 170 7.2.4 The FAX06 and the MAX06 Phantoms ..................................................... 170 7.2.4.1 ICRP103 and New Concepts for Skeletal Dosimetry .............. 170 7.2.4.2 Segmentation of New Organs and Tissues .............................. 171 7.3 Results ......................................................................................................................... 177 7.3.1 Anatomical Results ...................................................................................... 177 7.3.1.1 The MAX06 Phantom .................................................................. 177 7.3.1.2 The FAX06 Phantom .................................................................... 179 7.3.2 Dosimetric Results ....................................................................................... 179 7.3.2.1 Skeletal Dosimetry Based on CT Images of Spongiosa .......... 179 7.4 Conclusions ................................................................................................................ 194 Acknowledgments ............................................................................................................... 194 References ............................................................................................................................. 194

7.1 Introduction Matter, after having been exposed to ionizing radiation, cannot express itself in terms of absorbed or equivalent dose. Consequently, the equivalent dose in tissues of the human body cannot be measured directly. Indirect measurements of equivalent dose can be made by radiation detectors, but they are restricted to locations on the surface of the human 163

164

Handbook of Anatomical Models for Radiation Dosimetry

body, thereby excluding most of the radiosensitive organs and tissues from this method of equivalent dose assessment. In order to establish relationships between protection quantities to be determined and operational quantities that can be measured, one has to make use of an exposure model: a physical or computational structure for the simultaneous determination of equivalent dose in the human body and of measurable quantities for exposure conditions relevant in radiation protection. It must therefore incorporate sources and fields of the radiations involved, a representation of the human body, a method for the determination of equivalent dose inside the human body, and a possibility to determine measurable quantities of interest. The results are usually expressed as conversion coefficients (CCs) as a function of the exposure conditions, which are ratios between equivalent dose to organs and tissues at risk and measurable quantities. Routine measurements in radiation protection can be interpreted in terms of organ and tissue equivalent dose by multiplying the instrument’s reading with the appropriate CC, provided that the irradiation conditions simulated with the exposure computational phantom correspond to the real exposure situation. Previously, CCs for radiation protection were determined using physical exposure model.1 The human body was represented by a physical phantom, which consisted of a real human skeleton embedded in tissue-equivalent material having the form and shape of a human body. At the same time, speed and memory capacity of computers increased significantly, which made it possible to develop computational exposure model that used virtual representations of the human body, called computational human phantoms (hereafter called phantoms), and radiation transport simulation methods to determine CCs.2 Today, absorbed or equivalent dose assessments in the human body are usually made with computational exposure model. The most important equivalent dose quantity in radiological protection is the effective dose, defined in 1991 by the International Commission on Radiological Protection (ICRP) in Publication 60.3 At that time, the effective dose represented a sum over the weighted equivalent doses in 23 organs and tissues at risk. Consequently, any type of human phantom should ideally contain at least these 23 organs and tissues. In 2000, at the Department of Nuclear Energy of the Federal University of Pernambuco (DEN/UFPE) in Recife/Brazil, a computational exposure model for the determination of CCs for the area of radiological protection using Monte Carlo (MC) codes was developed. For this purpose it was necessary to design two human phantoms. The FAX06 (Female Adult voXel) and the MAX06 (Male Adult voXel) phantoms have been developed in a two-step process. In the first step, two adult human voxel phantoms have been developed, called FAX4 and MAX,5 which contain homogeneously segmented skeletons and all 23 organs and tissues necessary to calculate the effective dose as defined by ICRP60.3 Care was taken that the masses of the organs and tissues matched the reference masses given by ICRP896 as closely as possible. The new recommendations of the ICRP, Publication 103,7 revise the concept of the effective dose with respect to the number of organs and tissues considered at risk and therefore included in the determination of the effective dose. Consequently, in the second step, the two adult phantoms underwent additional segmentation, which added six new organs and tissues according to ICRP103,7 and were called from then on FAX06 and MAX06.8 In order to open the possibility for the application of μCT images of trabecular bone in the FAX06 and the MAX06 phantoms, the second segmentation was extended to the skeletons by segmenting cortical bone, spongiosa (=trabecular bone filled with soft tissue),

The FAX06 and the MAX06 Computational Voxel Phantoms

165

medullary yellow bone marrow (YBM) in the long bones, and cartilage under observation of ICRP-based skeletal tissue masses.9 At the time of their introduction, MAX and FAX were the first human adult phantoms with soft-tissue organ masses based on the anatomical reference data recommended by ICRP Publication 896 and the updated versions FAX06 and MAX06 were the first human adult phantoms with ICRP-based skeletons, segmented into cortical bone, spongiosa, medullary YBM, and cartilage. Since then, the further development of the FAX06 and the MAX06 phantoms occurred mainly inside their skeletons where the spongiosa voxels have been further segmented into trabecular bone and marrow for the purposes of skeletal dosimetry.10,11 Over the years, the phantoms have been applied to many areas of radiation protection, such as medical exposures,12 occupational and environmental exposures,13 and to accidental exposures.14 With respect to patient dosimetry in x-ray diagnosis, a software tool for absorbed dose calculation based on the FAX06 and the MAX06 phantoms has been made available to the public.15,16 Until now, most of the absorbed or equivalent dose calculations in the FAX06 and the MAX06 phantoms have been made with the EGS417 and EGSnrc18,19 MC codes for photons and electrons. Meanwhile, the phantoms have been successfully connected to the GEANT420 MC code, which allows for the transport of many other particles besides electrons and photons in the future.21 During the two-step development of the FAX06 and the MAX06 phantoms, certain organs and tissues had to be resegmented and certain dosimetric methods had to be revised, especially in the area of skeletal dosimetry. The description given in this chapter will therefore focus on the updated versions of FAX06 and MAX06 and will report on their predecessors FAX and MAX only as far as necessary. For concepts that have now only “historical value,” the reader is kindly asked to consult the original publications.4,5 This book, in which this chapter appears, represents the most comprehensive and upto-date description of “Anatomical Computational Phantoms for Radiation Dosimetry.” Therefore, it is considered unnecessary to review the literature of phantom development in this chapter.

7.2 Materials and Methods 7.2.1 Voxel Phantoms A computerized tomography (CT) image is a two-dimensional (2D) picture of a crosssection through the body of a human, which is composed of pixels (picture elements) representing different gray values distributed mostly on a scale between 0 and 255. When presented in the format normally used in x-ray diagnosis, the human eye differentiates apparent homogeneous regions in CT images, which represent organs, tissues and, sometimes, pathological abnormalities. If amplified, however, these regions reveal that the homogeny is, in reality, composed of an intrinsic mixture of pixels with different gray values. Therefore, constructing anthropomorphic phantoms composed of homogeneous organs and tissues requires segmentation, a process, which divides a CT image into various organ-specific regions by assigning one label, called the organ ID number, to all pixels within that region, through which they become part of that organ or tissue.

166

Handbook of Anatomical Models for Radiation Dosimetry

A consecutive set of segmented CT images from head to toe represents a threedimensional voxel (volume element) matrix, called a human voxel phantom, and with ID numbers assigned to each voxel depending on the organ or tissue to which it belongs to. Human voxel phantoms can be introduced into the MC radiation transport codes for the purpose of equivalent dose calculation in segmented organs and tissues. 7.2.2 The MAX Phantom 7.2.2.1 Database Three segmented voxel phantoms were available on the Web site of Yale University22,23 at the time when the phantom development at the DEN/UFPE began • VOXELMAN: A torso voxel phantom with head, but without arms and legs • MANTISSUE3-6: The VOXELMAN phantom with legs and arms, which are closed in front of the abdomen • VOXTISS8: The MANTISSUE3-6 voxel phantom with the arms straightened laterally along the body All three phantoms have been constructed from the same database: namely 78 CT images acquired from neck to midthigh with 1 cm slice thickness, 51 CT images of the head and neck region with 0.5 cm slice thickness, and 124 high-resolution transverse MRI images with 0.15 cm slice thickness from a patient, who was scheduled for head, thorax, abdomen, and pelvic scans for a diagnosis of diffuse melanoma. His height was 175 cm and the weight was 70.2 kg. VOXELMAN represents the combination of the segmented head and body CT images with a 4 mm cubic voxel size. Later, Stuchly24 added arms and legs, segmented from the Visible Man’s red color cross-sections,25 to the torso phantom, which was called MANTISSUE3-6. This phantom has been rescaled to achieve a 3.6 mm cubic voxel size. Finally, Sjogreen26 straightened along the sides of the body the arms of the MANTISSUE3-6 phantom maintaining the 3.6 mm cubic voxel size. This version is called VOXTISS8, and it consists of 487 segmented body cross-sections, each of which expands into a 192 × 96 pixel matrix. About 40 organs and tissues have been segmented in the trunk, arms, and legs, and about 56 organs and tissues in the head. The VOXTISS8 phantom contains the high-resolution MRI head, and it was this voxel computational phantom, which has been chosen as database for the construction of the MAX phantom. Table 7.1 shows organs and tissue masses for the VOXTISS8 phantom and for the ICRP Reference Man.6 The last column shows the percentage deviation of the VOXTISS8 organ masses relative to the ICRP data. As one can see, the agreement between the two sets of data is poor. The differences are smaller than 10% only for the eyes, the brain, and the adipose tissue. For all other organs and tissues, the deviations are at least 15%, sometimes even higher than 100%. The thymus and one adrenal were not segmented at all in the VOXTISS8 phantom, and the volume of the other adrenal was too small. The bladder wall was too thick compared to the dimensions shown in anatomical textbooks or other phantoms. Voxels inside the body had the organ ID number of the testes, which made this organ too heavy, apart from causing an unrealistic distribution of its volume, and several transversal images showed that the lungs had been segmented sometimes outside the ribcage. Also it was found that about 4.5 kg of blood had falsely been segmented as muscle and adipose tissue.

The FAX06 and the MAX06 Computational Voxel Phantoms

167

TABLE 7.1 Comparison of Organ and Tissue Masses between the ICRP Reference Man and the VOXTISS8 Phantom

Organ/Tissue Adipose (fat) Adrenals Bladder wall Skeleton Brain Colon Eyes Kidneys Liver Lungs Muscle, skeletal Esophagus Pancreas RBM Small intestine Skin Spleen Stomach Testes Thymus Thyroid Trachea Total body Height

ICRP89

YALE

Percentage Differences

Reference Man (g)

VOXTISS8 (g)

VOXT/Reference Man (%)

14,600 14 50 10,500a 1,450 670 15 310 1,800 1,200 29,000 40 140 1170 1,000 3,300 150 400 35 25 20 10 73,000 176 cm

14,970 3 154.2 13,869.4b 1491 895.7 15.8 373.2 1434 756.5 36,070 31.5 38.8 570.9 1,296 6,676 272.8 251.5 72.5 Not segmented 5.1 14.3 81,728 175.3 cm

+2.5 −78.6 +208.4 +32.1 +2.8 +33.7 +5.3 +20.4 −20.3 −37.0 +24.4 −21.3 −72.3 −51.2 +29.6 +102.3 +81.9 −37.1 +107.1 – −74.5 +43.0 +12.0 −0.4

Source: Kramer, R. et al., Phys. Med. Biol., 48, 1239, 2003. With permission. Stomach, colon, and small intestine include contents. a Bone, marrow, cartilage, misc. b Bone and marrow.

The VOXTISS8 skeleton has a volume of 7884.7 cm3, 1387.1 cm3 of which have been segmented as “bone marrow.” The linear voxel dimensions of the VOXTISS8 phantom are 3.6 mm, whereas the linear dimensions of bone marrow cavities are in the range of 50–2000 μm.27 As it is impossible to segment an object with a pixel size that is greater than the linear dimension of the object to be segmented, consequently the volume segmented within the VOXTISS8 skeleton cannot represent bone marrow. In addition, one finds that the volume of that “bone marrow” represents only 17.6% of the total skeletal volume, whereas according to ICRP709 the bone marrow occupies 47.4% of the volume of the skeleton. Consequently with 13,869.4 g the weight of the VOXTISS8 skeleton is quite heavy for a person with a body height of 175.3 cm as Table 7.1 indicates. Apart from the bone marrow problem it was also found that part of the clavicles had not been segmented in the VOXTISS8 phantom.

168

Handbook of Anatomical Models for Radiation Dosimetry

7.2.2.2 Adjustment of Organ and Tissue Masses Our objective with the new segmentation was to change the volume of already segmented organs, and to add new organs that have not been segmented, such that the new voxel phantom corresponds with its organ and tissue masses as closely as possible to the organ masses given for the ICRP Reference Man.6 For the soft-tissue organs, a method of voxel exchange with surrounding general tissues, such as adipose, muscle, blood, etc. was applied, taking into account relevant information from anatomical textbooks. In the case of the testes, for example, the ID numbers of the voxels falsely segmented as testes inside the body have been changed into the ID numbers for muscle and/or adipose tissue, thereby reducing the volume of the testes. In order to reduce the volume of the bladder wall, voxels at the inner surface of the bladder have been changed into urine voxels until the remaining wall volume had the desired value. According to the mass differences shown in Table 7.1, the volumes of the adrenals, the liver, the lungs, the esophagus, the pancreas, the stomach, and the thyroid have been increased, while the volumes of the bladder wall, the colon, the kidneys, the skin, the small intestine, the spleen, and the testes have been decreased. The thymus and one adrenal have been added based on data found in ICRP2328 and anatomical textbooks. The thyroid has not only been enlarged, but its position has also been changed according to ICRP23, where 2 cm of tissue overlying the thyroid are mentioned for the adult male. While the trunk and head of the VOXTISS8 phantom represent a patient with a body height of 175.3 cm and a body weight of 70 kg, the arms and legs have been added to the torso from a person with a body height of 186 cm and a body weight of 104 kg. Therefore, we can assume that the muscle and adipose tissue of the arms and the legs are mainly responsible for the soft-tissue part of the 12% excess weight. In order to achieve a reduction of the total body weight, the superficial layer of skin voxel was uniformly removed from the arms and the legs. Afterwards the first layer of the remaining voxels received the ID number of the skin. 7.2.3 The FAX Phantom 7.2.3.1 Database The main set of data used for the construction of the FAX phantom consisted of 151 consecutive CT images of a 37-year-old female patient. The patient’s height was 165 cm, and her weight was 63.4 kg. The images covered the trunk, the neck, and the lower part of the head including the mandible with the lower teeth. The pixel size was 0.073 cm × 0.073 cm, and the distance between two consecutive images was 0.5 cm. CT Screening International, Irvine, California, provided the images in October 2002. A second set of data consisted of 206 consecutive CT images of the legs and feet of a 62-year-old woman. The pixel size was 0.07 cm × 0.07 cm, and the distance between two consecutive images was 0.25 cm. The images had been provided by the university hospital of the city of Porto Alegre, Brazil in September 2003. 7.2.3.2 Segmentation The CT images of the patients have been obtained in the digital imaging and communications in medicine (DICOM) format, and have been visualized by means of the software OSIRIS,29 which is available on the Internet. OSIRIS has several types of filters, which improve the visualization of boundaries between organs. After editing, the images have

The FAX06 and the MAX06 Computational Voxel Phantoms

169

been saved as bitmap files with the software PAINT, which is included in the Microsoft WINDOWS accessories. Segmentation was done in each of the 151 torso images with PAINT by manually painting every organ and tissue of interest with a different color, while the 206 images of the legs have been segmented automatically.30 Thereby the gray values of the pixels, which belong to a specific organ, were replaced by a specific color, which corresponds to a specific ID number. The automatic segmentation of the 206 images of the legs was possible, because the legs contain only a few different tissues, like skin, muscle, adipose, bone, and marrow. In the first step, a vector was defined to store in increasing order 256 gray values shown on the screen of the computer, beginning with the color white, then gradually turning into gray and finally reaching black. In this way, the position of a gray value stored in the vector corresponds to a “pixel value” to be used in the process of automatic segmentation, indicated by Figure 7.1, where (A) represents the original image in bitmap format. First, the size of the image was reduced from 512 × 512 original pixels to 256 × 256 new pixels (B), where each new pixel is the arithmetic mean of the gray values of four neighboring original pixels. This reduction provides a homogenization of the gray values that compose a certain region. Without changing the size of the image any further, this averaging process was repeated two more times (C) and (D), with the eight nearest neighbors of each pixel. Then a fi lter was used to remove “isolated pixels,” which represent pixels with a gray value significantly different from those of its neighbors (E). Finally, certain boundaries were chosen to define intervals on the scale between 0 and 255. Then, to all pixels with values lying within a certain interval, one value and its corresponding color were assigned. The regions defined by the intervals correspond to organs or tissues and the “colored value” is their ID number (F). According to the definition of the effective dose given in ICRP Publication 60,3 the following organs and tissues have been segmented: the adrenals, the bladder wall, the

FIGURE 7.1 Graphical representation of steps (A) through (F) for automatic segmentation of the FAX06 legs.

170

Handbook of Anatomical Models for Radiation Dosimetry

skeleton, the brain, the breasts, the colon, the kidneys, the liver, the lungs, the muscle, the esophagus, the ovaries, the pancreas, the small intestine, the skin, the spleen, the stomach, the thymus, the thyroid, the trachea, and the uterus. Although not included in the effective dose of Publication 60, the heart and adipose tissue have also been segmented. 7.2.3.3 Addition of Head and Arms The head of the MAX phantom was attached to the neck of the FAX phantom, after scaling down the MAX phantom’s head according to the anatomical differences between the Male and the Female Reference Adult as defined by ICRP89.6 The same procedure was applied to the addition of the arms, which also have been taken from the MAX phantom. 7.2.3.4 Anatomical Corrections During screening, the patient was asked to raise her arms behind her head. Therefore, we have redesigned the shoulders and the upper part of the arms according to data from anatomical textbooks. The images of the trunk, the head, the arms, and the legs originated from three different patients. Therefore, with respect to the representation of the skeleton some adjustments had to be made again based on anatomical textbooks, but at the same time taking into account as much as possible the reference distribution of bone mass fractions given by ICRP89.6 The form of the breasts has also been modified in order to represent an upright standing adult female. As in the case for the MAX phantom, all segmented organ and tissue volumes of the FAX phantom were adjusted based on voxel exchange to match the reference masses given by ICRP89.6 Finally all segmented images have been rescaled to achieve the same voxel size of 0.36 cm × 0.36 cm × 0.36 cm = 0.046656 cm3 as in the case of the MAX phantom, and the total number of slices has been adjusted to 453, which corresponds to a body height of 163 cm. 7.2.4 The FAX06 and the MAX06 Phantoms 7.2.4.1 ICRP103 and New Concepts for Skeletal Dosimetry According to the recommendations of the ICRP released in Publication 60, 3 the determination of the effective dose, the most important dose quantity in radiological protection, requires the calculation of the equivalent dose to radiosensitive organs and tissues shown in columns 1 and 2 of Table 7.2. The trachea replaced the upper large intestine,31 which was part of the original list of tissues,3 but actually is already included in the colon. Meanwhile, the ICRP extended by six the number of radiosensitive organs and tissues to be taken into account for the determination of the effective dose in its new recommendations released in Publication 103,7 due to recently reported new evidence about stochastic radiation effects in these organs and tissue. These new organs and tissues are shown in the third column of Table 7.2. The segmented organs and tissues of the FAX and the MAX phantoms include those mentioned in the first two columns of Table 7.2. The new organs and tissues from column 3 had either not been segmented (the gallbladder, the extra thoracic airways, the lymphatic nodes, the salivary glands) or their masses had not been adjusted (the heart wall, the prostate). In order to prepare the phantoms for the effective dose calculations based on ICRP103, it was decided to segment and/or adjust the organs and tissues shown in column 3 of

The FAX06 and the MAX06 Computational Voxel Phantoms

171

TABLE 7.2 Radiosensitive Organs and Tissues to be Included in the Calculation of the Effective Dose Organs and Tissues (ICRP60)

New Organs (ICRP103)

Adrenals Bladder Bone surface Brain Breast Colon Gonads Kidneys Liver Lungs Muscle

Extra thoracic airways Gallbladder Heart wall Lymphatic nodes Prostate Salivary glands

Esophagus Pancreas RBM Skin Small intestine Spleen Stomach Thymus Thyroid Trachea Uterus

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.

Table 7.2 in the MAX and the FAX phantoms, and to use this opportunity to segment the walls and contents in the stomach, the colon, and the small intestine, as well as the cortical bone, the medullary yellow the bone marrow, the spongiosa, and the cartilage in the skeletons, segmentations which had not been done during the design of the MAX and the FAX phantoms. Skeletal tissues had not been segmented because the CT number method, used to calculate the equivalent dose to the red bone marrow (RBM), required the application of the skeletal gray values contained in the original CT images of the patient. As these skeletal gray values have to match the segmented bone pixels exactly, it was not possible to segment regions inside the bones. New methods for skeletal dosimetry are based on CT images of spongiosa. Their applications to the skeletons of the MAX and the FAX phantoms, however, require the segmentation of the cortical bone, the spongiosa, the medullary bone marrow, and the cartilage in the skeletons. 7.2.4.2 Segmentation of New Organs and Tissues The segmentation of the new organs and tissues, in the FAX06 and MAX06 phantoms, was based on the original CT images, on anatomical textbooks, 32,33 on ICRP70,9 and on ICRP89,6 and if not stated otherwise, the following description refers to both phantoms. The images of the two phantoms have been edited with the SCION software, 34 and adjustments were made to the volumes by changing the organ and tissue identification numbers literally pixel by pixel. Tissue compositions and densities from ICRU4435 shown in Table 7.3 have been used to control the match with the ICRP reference masses. Adipose tissue, already segmented in the FAX and the MAX phantoms, was adjusted to match the reference mass and connective tissue was segmented for the first time. In 2006 when the phantom update was initiated, adipose and connective tissue were still considered by the ICRP as tissues at risk for the determination of the effective dose, but then eventually excluded in the fi nal version of the new recommendations, which was approved in 2007.

Handbook of Anatomical Models for Radiation Dosimetry

172

TABLE 7.3 Elemental Tissue Composition Used in the MAX06 and the FAX06 Phantoms Based on ICRU44

Atom H C N O Na Mg P S Cl K Ca Fe ρ (g cm−3)

Soft Tissue (%) 10.4 12.4 2.6 73.5 0.2 0.2 0.2 0.2 0.2 0.02 0.02 1.05

Muscle Tissue (%)

Adipose Tissue (%)

Skin Tissue (%)

Lungs Tissue (%)

Skeleton Skeleton Cartilage Cortical (%) (%)

10.2 14.3 3.4 71.0 0.1

11.4 59.8 0.7 27.8 0.1

10.0 20.4 4.2 64.5 0.2

10.3 10.5 3.1 74.9 0.2

9.6 9.9 2.2 74.4 0.5

0.2 0.3 0.1 0.4

0.1 0.1

0.1 0.2 0.3 0.1

0.2 0.3 0.3 0.2

2.2 0.9 0.3

3.4 15.5 4.2 43.5 0.1 0.2 10.3 0.3

22.5 1.05

0.95

1.09

0.26

1.1

1.92

Skeleton Spongiosa Skeleton (%) YBM (%) 8.5 40.4 2.8 36.7 0.1 0.1 3.4 0.2 0.2 0.1 7.4 0.1 1.18

11.5 64.4 0.7 23.1 0.1

0.1 0.1

0.98

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.

7.2.4.2.1 Adipose Tissue Adipose tissue had already been segmented in the original versions of the two phantoms. Only minor adjustments of the masses had to be made. 7.2.4.2.2 Connective Tissue Separable connective tissue includes primarily periarticular tissue, tendons, and fascia. Periarticular tissue has been segmented at the shoulders, the knees, the hips, and the elbows, between the vertebrae, between vertebrae and ribs, and between the sacrum and the ilium, and also at the sinfisis pubis for the FAX06 phantom. The fascia has been introduced between the deepest layer of the subcutaneous tissue and the muscle tissue distributed over the whole body, while the tendons have been segmented mainly at the joints. 7.2.4.2.3 Extra Thoracic Airways ICRP896 and ICRP1037 clarify that the extra thoracic airways consist of the anterior and posterior nasal passages, the mouth cavity, the pharynx and the larynx. Reference masses for the larynx are given in ICRP89.6 The volumes for the pharynx have been determined with data from anatomical textbooks, while estimates for the volumes of the nasal passages and the mouth cavity have been made based on data provided by ICRP896 on page 92 and also based on anatomical textbooks.32,33 7.2.4.2.4 Gallbladder The gallbladder was segmented below the liver, separately by wall and contents. 7.2.4.2.5 Heart Wall The heart walls had already been segmented in both phantoms, but now the volumes have been adjusted to match the masses recommended by ICRP89.6

The FAX06 and the MAX06 Computational Voxel Phantoms

173

7.2.4.2.6 Lymphatic Nodes Lymphatic nodes have been segmented at the armpit, at the groins, behind the knees, in front of the elbows, at the neck, and around abdominal soft-tissue organs. ICRP89 states “Except for the lymphocytes, which are present in most tissues, the lymphatic tissue is contained in the RBM and in the lymphatic organs: lymph nodes, spleen, thymus, mucous membranes, tonsils, adenoids, Peyer’s patches, and the vermiform appendix.”6 Therefore the segmented volume of the lymphatic nodes was confined to 50% of the volume recommended by ICRP89. 7.2.4.2.7 Prostate The prostate had already been segmented in the MAX phantom, but now the volume has been adjusted in order to match the mass recommended by ICRP89.6 7.2.4.2.8 Salivary Glands The salivary glands have been segmented around the mandible as parotid, submaxillary, and sublingual parts, observing the mass ratios 10:5:2 between them according to ICRP89.6 7.2.4.2.9 Stomach, Colon, and Small Intestine Introducing a wall thickness of one voxel layer, i.e., of 3.6 mm, the walls and contents of the stomach and the colon have been segmented. Because of the many twists and turns of the small intestine it is almost impossible to segment a continuous and steady wall based on voxels. Therefore the total volume of the small intestine was divided between the wall and contents by distributing the wall voxels homogeneously throughout the volume of the organ. 7.2.4.2.10 Skeleton The skeletons of the MAX and the FAX phantoms had been segmented with respect to the surrounding muscle tissue, but no attempt was made until now to segment tissues within the skeletal structures, like cortical bone, spongiosa, medullary YBM in the shafts of the long bones and cartilage. The images of the VOX_TISS8 phantom,23 which had been the basis for the development of the MAX phantom, show areas of segmented bone marrow, however, it was demonstrated that this bone marrow segmentation was faulty. In order to prepare the phantoms for advanced skeletal dosimetry in the future, we decided to segment the cortical bone, the spongiosa, the medullary YBM in the shafts of the long bones, and the cartilage in the skeletons of the MAX and the FAX phantoms, based on the original CT images, as well as by using anatomical textbooks,32,33 and the color photographs of the Visible Human.25 However, the new skeletons should also be based on recommendations published by the ICRP as far as anatomically possible, similar to the procedure applied to the segmented soft-tissue organs. Therefore the segmentation of the skeletal tissues took into account the following data provided by ICRP896 and ICRP70:9 1. Reference masses for bone, RBM, YBM, cartilage, and miscellaneous tissues (teeth, periostenum, and blood vessels) as shown in Table 7.4 for the reference adult male and female.6 For the cartilage masses only half of the recommended mass was taken into account because a part of the cartilage included by ICRP89 into the mass of the skeleton actually is located “off-bone,” like in the ear, in the nose, etc.,

Handbook of Anatomical Models for Radiation Dosimetry

174

TABLE 7.4 Skeletal Tissue Masses, Volumes, and Densities Based on ICRP89 and ICRU44 Skeletal Tissues Bone RBM YBM Cartilage Misc. Total

Density, r(g cm−3)

Reference, m(g)

Male, v(cm3)

Reference, m(g)

Female, v(cm3)

1.92a 1.03a 0.98a 1.10a 1.20b 1.37b

5500 1170 2480 550 250 9950

2864.6 1135.9 2530.6 500.0 208.3 7239.4

4000 900 1800 450 200 7350

2083.3 873.8 1836.7 409.1 166.7 5369.5

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. BM, bone marrow; MISC, teeth, periostenum, and blood vessels. a ICRU densities. b Calculated densities.

and also because in order to achieve smooth surfaces between bone and skeletal muscle, the cartilage is sometimes segmented as bone or sometimes as muscle. The density for the miscellaneous tissues has been calculated based on data from ICRP2328 as 1.20 g cm−3. With the ICRU4435 densities and the calculated densities shown in column 2, the corresponding skeletal tissue volumes have been determined and are presented in columns 4 and 6 of Table 7.4. The average density of 1.37 g cm−3 calculated from the total skeletal mass and volume is, of course, greater than the 1.3 g cm−3 recommended by ICRP89 because of the smaller cartilage mass. 2. For nine important bones or bone groups, Table 7.5 presents sex-specific mass fractions, the mass ratios between cortical and trabecular bone, the RBM mass fractions, taken or derived from data recommended in ICRP89,6 and the cellularity factors, which indicate the fraction of the marrow volume occupied by RBM, taken from ICRP70.9

TABLE 7.5 Skeletal Tissues Data Based on ICRP70 and ICRP89 ICRP 70/89 Skeletal Region

Male Bone Mass Fractions

Female Bone Mass Fractions

Mass Ratio of Bone Cortical/Trabecular

RBM Mass Fractions

Cellularity Factor

Lower arm bone Upper arm bone Ribcage Spine Skull Mandible Pelvis Upper leg bone Lower leg bone

0.059 0.053 0.126 0.190 0.118 0.012 0.106 0.153 0.183

0.056 0.047 0.104 0.204 0.119 0.012 0.105 0.159 0.193

87/13 80/20 94/6 30/70 95/5 90/10 90/10 70/30 80/20

0 0.023 0.228 0.422 0.076 0.008 0.176 0.067 0

0 0.25 0.60 0.70 0.38 0.38 0.48 0.25 0

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. Ribcage, ribs, sternum, clavicles, scapulae.

The FAX06 and the MAX06 Computational Voxel Phantoms

175

Tables 7.6 and 7.7 show theoretical distributions of skeletal tissue volumes for the reference adult male and female, respectively, which have been calculated with the data from Tables 7.4 and 7.5. The volumes of the miscellaneous tissues have been added to the volumes for cartilage. As an example, the calculation, which created the distribution of skeletal tissue volumes in Tables 7.6 and 7.7, will be demonstrated for the spine of the adult male: Application of the RBM mass fraction From Tables 7.4 and 7.5 one can calculate the amount of RBM mass in the spine as 1170 g × 0.422 = 493.7 g, or the RBM volume in the spine as 493.7 g/1.03 g cm−3 = 479.3 cm3.

TABLE 7.6 Theoretical Volume Distribution of Skeletal Tissues for the Adult Male Adult Male Skeletal Region Lower arm bone Upper arm bone Ribcage Spine Skull Mandible Pelvis Upper leg bone Lower leg bone Total volume

Cortical Bone, cm3 164.9 85.8 351.0 163.2 422.5 40.2 258.8 201.9 470.5 2158.8

Spongiosa, cm3

YBM Med., cm3

Cart/Misc., cm3

Total, cm3

144.4 215.8 454.1 1065.2 249.3 28.5 445.5 669.3 489.2 3761.3

59.0 89.9

42.6 37.4 89.7 134.8 83.1 8.5 74.7 108.8 128.8 708.4

410.9 428.9 894.8 1363.2 754.9 77.2 779.0 1257.4 1273.3 7239.6

277.4 184.8 611.1

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. Cart/misc., cartilage + miscellaneous tissues; ribcage, ribs, sternum, clavicles, scapulae; YBM med., medullary YBM in the shafts of the long bones.

TABLE 7.7 Theoretical Volume Distribution of Skeletal Tissues for the Adult Female Adult Female Skeletal Region Lower arm bone Upper arm bone Ribcage Spine Skull Mandible Pelvis Upper leg bone Lower leg bone Total volume

Cortical Bone, cm3 118.8 50.9 177.0 130.0 307.9 29.3 181.7 162.1 377.1 1,534.8

Spongiosa, cm3

YBM Med., cm3

Cart/Misc., cm3

Total, cm3

94.6 150.7 343.3 829.9 190.9 21.6 340.6 502.2 358.8 2,832.6

38.4 57.6

32.4 27.2 59.8 117.6 68.9 6.9 61.4 91.2 110.4 575.8

284.2 286.4 580.1 1,077.5 567.7 57.8 583.7 953.8 978.5 5,369.7

198.3 132.2 426.5

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. Cart/misc., Cartilage + miscellaneous tissues; ribcage, ribs, sternum, clavicles, scapulae; YBM med., medullary YBM in the shafts of the long bones.

176

Handbook of Anatomical Models for Radiation Dosimetry

Application of the cellularity factor According to the cellularity factor from Table 7.5 the RBM volume of 479.3 cm3 represents 70% of the total marrow volume in the spine. Therefore the YBM volume in the spine is 479.3 cm3 × 3/7 = 205.4 cm3, or the YBM mass in the spine is 205.4 cm3 × 0.98 g cm−3 = 201.3 g. Application of the male bone mass fraction From Tables 7.4 and 7.5 one can calculate the total mass of the spine as 9950 g × 0.190 = 1890.5 g. From Table 7.4 one can calculate the mass for Cart/misc as (550 g + 250 g)/9950 g × 1890.5 g = 152 g, or the volume for Cart/misc as 152 g/1.128 g cm−3 = 134.8 cm3. Application of the mass ratio between cortical and trabecular bone With the ratios between cortical and trabecular bone from Table 7.5 one can calculate the masses and volumes for cortical and trabecular bone in the spine as cortical mass = (1890.5 − 493.7 − 201.3 − 152 g) × 0.3 = 313.1 g, and the cortical volume as 313.1 g/1.92 g cm−3 = 163.2 cm3, and the trabecular mass = (1890.5 g − 493.7 g − 201.3 g − 152 g) × 0.7 = 730.5 g, and the trabecular volume as 730.5 g/1.92 g cm−3 = 380.5 cm3. The spongiosa volume is 479.3 + 205.4 + 380.5 = 1065.2 cm3. The application of this method to all nine bones yielded 1036 g of YBM in all RBM containing bones. Taking into account the total YBM mass from Table 7.4 one finds 2480 − 1036 = 1444 g of YBM which is located in the shafts of the long bones, in the spongiosa of the lower part of the upper long bones and in the lower long bones. These regions of spongiosa do not contain RBM. Finally, it was possible to determine the volume of the YBM in the shafts of the long bones, also called medullary YBM using the bone mass fraction of the long bones, and anatomical data from the CT images and the textbooks shown in column 4 of Tables 7.6 and 7.7. The segmentation of the skeleton into cortical bone, spongiosa, medullary YBM, and cartilage should take the volumes shown in Tables 7.6 and 7.7 into account as far as skeletal anatomy would permit to do so. In the bones of the human skeleton spongiosa is usually surrounded by regions of cortical bone. Measurements on cortical bone thickness in various bones in the CT images used for the design of the MAX06 and the FAX06 phantom have shown that for both sexes ca. 1.2 mm can be considered as a minimum thickness for cortical bone covering the spongiosa. This value has been confirmed independently by similar measurements performed by Brindle and Bolch.36 However, a cortical bone thickness of 1.2 mm cannot be represented by a 3.6 mm cubic voxel. Therefore the voxel matrices of both phantoms were rescaled in order to achieve a 1.2 mm cubic voxel matrix. The rescaling procedure divided each dimension of a 3.6 mm cubic voxel of the MAX and the FAX phantoms by 3, which gave 3 × 3 × 3 = 27 1.2 mm cubic voxels for every 3.6 mm cubic voxel. With this finer voxel resolution it became possible to realize anatomically meaningful distributions between cortical bone voxels and spongiosa voxels, as well as between skin voxels and adipose voxels. For example, in the MAX and the FAX phantoms the absorbed dose to the skin was calculated with a special algorithm in the first 1.5 and 1.2 mm depths of the surface voxels, respectively, because the voxel depth of 3.6 mm is not representative for the depth of the skin. In the new MAX06 and FAX06 phantoms the skin absorbed dose is now calculated as the absorbed dose to the 1.2 mm cubic surface voxel layer, i.e., averaged over a depth of 1.2 mm for both phantoms.

The FAX06 and the MAX06 Computational Voxel Phantoms

177

7.3 Results 7.3.1 Anatomical Results 7.3.1.1 The MAX06 Phantom For the most part, the skeleton of the MAX phantom is the skeleton of the VOX_TISS8 phantom.23 Table 7.8 shows the total volumes of bones and bone groups for the MAX phantom together with the ICRP-based data from Table 7.6. The percentage deviations in the last column are significant, especially for RBM containing bones, like the ribcage, the skull, the mandible, and the pelvis. At the first release of the MAX phantom5 the skeleton was hardly changed, because of the requirements of the CT number method as explained above, and also because the total volume of the MAX skeleton deviated only 9% from the ICRP-based total skeletal volume. But we felt that a segmentation of skeletal tissues should be based on anatomically reasonable volumes for the main bones and bone groups. The attempt to repair the falsely segmented MAX skeleton turned out to be complicated and showed that this process would probably take many months to achieve the objective. Therefore we decided to “borrow” the FAX skeleton, and to increase the dimensions of this female skeleton to arrive at the dimensions of the male skeleton, while at the same time observing the sex-specific mass ratios from Table 7.5. Table 7.9 shows the skeletal tissue volumes, which have been segmented in the MAX06 skeleton. For some bones it was possible to realize the ICRP-based volumes from Table 7.6 without compromising the skeletal anatomy. The total cortical bone volume of the MAX06 skeleton is 8.6% greater, and the total spongiosa volume is 4.9% smaller than the ICRPbased value from Table 7.6; however, for the total skeleton, the volume equals the theoretical value. After completion of the segmentation of the skeletal tissues, the soft-tissue organs of the MAX phantom, plus the newly segmented soft-tissue organs, were assembled together TABLE 7.8 Comparison of Bone Volumes between the MAX Phantom and the ICRP Reference Adult Male Adult Male Skeletal Region Lower arm bone Upper arm bone Ribcage Spine Skull Mandible Pelvis Upper leg bone Lower leg bone Total

MAX, cm3 308.0 600.2 1537.2 1131.1 497.0 102.4 1020.0 1132.0 1597.7 7925.6

ICRP-Based, cm3 410.9 428.9 894.8 1362.4 754.9 77.2 778.9 1257.5 1274.0 7239.5

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. Ribcage, ribs, sternum, clavicles, scapulae.

Difference, % −25 +40 +72 −17 −34 +33 +31 −10 +25 +9

Handbook of Anatomical Models for Radiation Dosimetry

178

TABLE 7.9 Volume Distribution of the Segmented MAX06 Skeletal Tissues MAX06 Phantom Skeletal Region

Cortical Bone, cm3

Spongiosa, cm3

YBM Med., cm3

Lower arm bone Upper arm bone Ribcage Spine Skull Mandible Pelvis Upper leg bone Lower leg bone Total volume

172.1 100.5 376.3 293.8 422.5 40.2 258.8 210.2 470.5 2344.9

137.1 201.1 429.7 934.6 249.0 28.5 445.5 661.1 489.1 3575.7

59.9 89.9

277.0 183.7 610.5

Cart/Misc., cm3 41.6 37.4 88.8 133.8 73.0 18.6 74.7 109.4 130.9 708.2

Total, cm3 410.7 428.9 894.8 1362.2 744.5 87.3 779.0 1,257.7 1274.2 7239.3

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. Cart/misc., cartilage + miscellaneous tissues; ribcage, ribs, sternum, clavicles, scapulae; YBM med., medullary YBM in the shafts of the long bones.

with the MAX06 skeleton to form the new MAX06 phantom, which consists of 1461 transversal images, each of which has 474 pixel × 222 pixel, i.e., the phantom matrix has 153738108 voxel, 41461410 of which are filled with human tissues. Figures 7.2 through 7.5 present images of the MAX06 phantom, which show some of the additionally segmented organs and tissues. Table 7.10 presents a comparison of the organ and tissue masses of the MAX06 phantom with those recommended in Table 2.8 of the ICRP89 report. The main differences between the organ and tissue masses of the two data sets occur for the skeleton (cartilage) and the lymphatic nodes for reasons which have been explained above; and for some tissues, like extra thoracic airways, spinal cord, etc., which are not listed in the Table of ICRP89. As these differences partly compensate, the total weight of the MAX06 phantom turns out to be about half a kilogram less than the reference weight of 73 kg, which corresponds to a difference of 0.7%, while at the same time the total sum of the ICRP recommended organ and tissue masses exceeds the reference weight by about half a kilogram.

Parotid

Submaxillar

Sublingual

FIGURE 7.2 MAX06 phantom: salivary glands. (From Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.)

The FAX06 and the MAX06 Computational Voxel Phantoms

179

MAX06, SLICE, 349 Skin Adipose Lungs Cortical bone ribcage Spongiosa ribcage Heart wall Soft tissue Lymphatic nodes Muscle Cortical bone upper arm bone Yellow marrow upper arm bone Esophagus Cortical bone spine Spongiosa spine Ribcage cartilage Spinal chord Spinel cartilage

FIGURE 7.3 MAX06 phantom: transversal image in the heart region. (From Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.)

7.3.1.2 The FAX06 Phantom Table 7.11 shows the skeletal tissue volumes segmented for the FAX06 phantom, and again for some bones the spongiosa and the cortical bone volume match the ICRP-based value exactly. The total cortical bone volume is 17.1% greater and the total spongiosa volume 9.3% smaller than the corresponding value from Table 7.7. These numbers differ from those found for the MAX06 skeleton above, because both sexes have different bone mass fractions as Table 7.5 indicates. Again the total skeletal volume equals the theoretical value from Table 7.7. Figures 7.6 through 7.8 present images of the FAX06 phantom, which show especially some of the additionally segmented organs and tissues. The comparison of the organ and tissue masses between the ICRP reference adult female and the FAX06 phantom is presented in Table 7.12. Here the sum of the ICRP masses from Table 2.8 of Publication 896 exceeds the reference weight of 60 kg by 193 g. Apart from the differences between the masses of the skeleton (cartilage) and the lymphatic nodes, the FAX 06 phantom has 1 kg of fat less than the reference adult FIGURE 7.4 MAX06 phantom. female. The FAX06 phantom consists of 1359 transversal images each of (From Kramer, R. which has 474 pixels × 222 pixels, i.e., the phantom matrix has 143004852 et al., Phys. Med. voxel, 34208854 of which are fi lled with human tissues. Biol., 51, 3331, 2006. Figures 7.9 and 7.10 show three-dimensional representation of the With permission.) skeletons and the body surface of the MAX06 and the FAX06 phantom, respectively, while Figure 7.11 presents the skeletons and internal organs of the FAX06 and the MAX06 phantoms with adipose and muscle tissue removed. 7.3.2 Dosimetric Results 7.3.2.1 Skeletal Dosimetry Based on CT Images of Spongiosa 7.3.2.1.1 Skeletal Soft Tissues at Risk The two skeletal tissues at risk, when the human body is exposed to ionizing radiation, are the hematopoietic stem cells of the marrow, called RBM, and the osteogenic cells on

Handbook of Anatomical Models for Radiation Dosimetry

180

the endosteal surfaces, called bone surface cells (BSCs) or bone endosteum, both of which are located in the irregularly shaped marrow cavities of trabecular bone with diameters ranging from 50 to 2000 μm.27 Skeletal dosimetry is concerned with the determination of the equivalent dose to the RBM and the BSC, which is a quite challenging task because of the complicated microstructure of trabecular bone. “Trabecular bone plus its supported soft tissue is sometimes also referred to as spongiosa.”9 Discussion about the final location and the distribution of the two skeletal tissues at risk is still ongoing with respect to the revision of the thickness of the BSC layer from 10 to 50 m, the exclusion of the Haversian canals of cortical bone, the inclusion of cortical surfaces of the medullary cavities,37 the consideration of trabecular bone remodeling,38 and the inhomogeneous distribution of RBM cells in the marrow.39 For the methods discussed here, the BSC represent the part of the marrow volume that is located within a

Tendons Periarticular tissue Fascia

FIGURE 7.5 MAX06 phantom: lymphatic nodes and connective tissues. (From Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.)

TABLE 7.10 Comparison of Organ and Tissue Masses between the MAX06 Phantom and the ICRP Reference Adult Male Organ/Tissue Adult Male Adrenals Salivary glands Esophagus Stomach wall Small intestine wall Colon wall Liver Gallbladder wall Pancreas Brain Heart wall Adipose Skin Muscle Lungs Skeleton Spleen Thymus Thyroid Kidneys Bladder wall Testes Prostate

ICRP89 (g)

MAX06 (g)

14.0 85.0 40.0 150.0 650.0 370.0 1,800.0

14.0 85.0 40.0 150.0 650.0 370.0 1,800.0

10.0 140.0 1,450.0 330.0 14,500.0 3,300.0 29,000.0 1,200.0 10,500.0 150.0 25.0 20.0 310.0 50.0 35.0 17.0 64,146.0

10.0 140.0 1,450.0 330.0 14,544.1 3,383.9 29,000.0 1,200.0 9,950.4 150.0 25.0 20.0 310.0 50.0 35.0 17.0 63,724.4

Difference (g)

+44.1 = +0.3% +83.9 = +2.5%

−549.6 = −5.2%

−421.6 = −0.7%

The FAX06 and the MAX06 Computational Voxel Phantoms

181

TABLE 7.10 (continued) Comparison of Organ and Tissue Masses between the MAX06 Phantom and the ICRP Reference Adult Male Organ/Tissue Adult Male Breasts Tongue Larynx Extra thoracic airways GI content Gallbladder content Trachea Tonsils Ureter/urethra Epididymes Pituitary gland Eyes Optic nerve Blood Hard palate Feces Spinal cord

ICRP89 (g)

Difference (g)

25.0 73.0 28.0 900.0 58.0 10.0 3.0 26.0 4.0 0.6 15.0

133.4 900.0 58.0 10.3

15.1 1.6

+0.1 = +0.7%

4900a

70,188.6 Connective tissue Lymphatic nodes

MAX06 (g)

2,600.0 730.0 73,518.6

Soft tissue 73,518.6

33.6 39.2 183.8 65,099.4 2,600.0 365.0 68,064.4 4,426.3b 72,490.7

−5,089.2 = −7.3% −365 = −50% −5,454.2 = −7.4% −1,027.9 = −1.4%

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. Without lungs. b Includes blood. a

TABLE 7.11 Volume Distribution of the Segmented FAX06 Skeletal Tissues FAX06 Phantom Skeletal Region Lower arm bone Upper arm bone Ribcage Spine Skull Mandible Pelvis Upper leg bone Lower leg bone Total volume

Cortical Bone, cm3 130.1 93.2 280.5 273.0 278.1 29.1 174.2 162.2 377.2 1797.6

Spongiosa, cm3

YBM Med., cm3

Cart/Misc., cm3

Total, cm3

83.2 108.4 239.6 686.9 220.0 21.6 348.5 502.2 358.8 2569.2

38.4 57.6

32.4 27.2 59.8 117.6 59.7 16.8 60.9 91.0 110.4 575.8

284.1 286.4 579.9 1077.5 557.8 67.5 583.6 953.7 978.6 5369.1

198.3 132.2 426.5

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. Cart/misc., cartilage + miscellaneous tissues; ribcage, ribs, sternum, clavicles, scapulae; YBM med., medullary YBM in the shafts of the long bones.

Handbook of Anatomical Models for Radiation Dosimetry

182

FAX06. SLICE 497

Adipose Small intestine content Small intestine wall Colon content Colon wall Lymphatic nodes Kidneys Spongiosa spine Spinal chord Soft tissue Cortical bone lower arm bone Yellow marrow lower arm bone Connective tissue (fascia) Muscle Cortical bone spine Periarticular tissue (connect. tissue)

FIGURE 7.6 The FAX06 phantom: transversal image in the abdominal region. (From Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.)

FAX06

Anterior nasal passage Posterior nasal passage

Pharynx

Mouth Larynx

FIGURE 7.7 The FAX06 phantom. (From Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.)

distance of 10 m from the surfaces of trabecular bone,40 while the RBM occupies a part of the remaining marrow volume given by the cellularity factor.9 7.3.2.1.2 Previous Methods of Skeletal Dosimetry The history of skeletal dosimetry is connected with the name of F. W. Spiers. Spiers and his coworkers at the University of Leeds in England can be considered the most important contributors to the development of skeletal dosimetry from its beginning in the early 1950s until the late 1980s. Conversion factors for marrow in trabecular bone from exposure to

The FAX06 and the MAX06 Computational Voxel Phantoms

183

Skin Pharynx Soft tissue Muscle Skull cartilage Brain Spongiosa skull Cortical bone skull Adipose

FIGURE 7.8 The FAX06 phantom: transversal extra thoracic airways image between mouth and nasal passage. (From Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.)

TABLE 7.12 Comparison of Organ and Tissue Masses between the FAX06 Phantom and the ICRP Reference Adult Female Organ/Tissue Adult Female Adrenals Salivary glands Esophagus Stomach wall Small intestine wall Colon wall Liver Gallbladder wall Pancreas Brain Breasts Heart wall Adipose Skin Muscle Lungs Skeleton Spleen Thymus Thyroid Kidneys Bladder wall Ovaries

ICRP89 (g)

FAX06 (g)

13.0 70.0 35.0 140.0 600.0 360.0 1,400.0 8.0 120.0 1,300.0 500.0 250.0 19,000.0 2,300.0 17,500.0 950.0 7,800.0 130.0 20.0 17.0 275.0 40.0 11.0

13.0 70.0 35.0 140.0 600.0 360.0 1,400.0 8.0 120.0 1,300.0 500.0 250.0 18,000.0 2,300.0 17,497.9 950.0 7,355.5 130.0 20.0 17.0 275.0 40.0 11.0

Difference (g)

−1,000 = −5.3% −2.1 < −0.1% −444.5 = −5.7%

(continued)

Handbook of Anatomical Models for Radiation Dosimetry

184

TABLE 7.12 (continued) Comparison of Organ and Tissue Masses between the FAX06 Phantom and the ICRP Reference Adult Female Organ/Tissue Adult Female Uterus Tongue Larynx Extra thoracic airways GI content Gallbladder content Trachea Tonsils Ureter/urethra Fallopian Tub. Pituitary gland Eyes Optic nerve Blood Hard palate Feces Spinal cord Connective tissue Lymphatic nodes

ICRP89 (g)

FAX06 (g)

Difference (g)

80.0 52,919.0 60.0 19.0

80.0 51,472.4

−1,446.6 = −2.7%

112.6 830.0 48.0 8.1

+0.1 = + 1.3%

830.0 48.0 8.0 3.0 18.0 2.1 0.6 15.0

15.0 1.3

3570a

57,492.7 2,100.0 600.0 60,192.7

Soft tissue 60,192.7

30.0 33.6 72.2 52,623.3 2,100.0 300.0 55,023.2 3,979.8b 59,003.0

−4869.5 = −8.5% −300 = −50% −5,169.5 = −8.6% −1,189.7 = −2.0%

Source: Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission. a Without lungs. b Includes blood.

x-rays,27,41,42 the measurement of chord-length distributions in trabecular bone,43 and the calculation of skeletal dose factors for incorporated radionuclides44–50 can be considered to be fundamental contributions to skeletal dosimetry, and many of them became the data base for research made in scientific laboratories all over the world. Based on the work of the group at the University of Leeds, another important center for skeletal dosimetry was developed at the Oak Ridge National Laboratory (ORNL) by Eckerman in connection with the development of the MIRD5 phantom series.51,52 With respect to the investigation presented here, the photon fluence-to-dose response (FDR) functions for RBM and BSC have to be mentioned, 53 which are based on the chord-length distributions measured by Darley54 and Beddoe et al.43 Application of the conversion factors for marrow in trabecular bone from external exposure to x-rays42 has initiated the development of skeletal dosimetry at the National Research Center for Environment and Health in Munich in Germany. Based on a proposal by Rosenstein,55 Kramer56 developed an algorithm to be applied to the MIRD5-type phantoms ADAM and EVA57 for external photon radiation, which used three correction factors (3CF), among them rad/Roentgen conversion factors for marrow in five different

The FAX06 and the MAX06 Computational Voxel Phantoms

185

FIGURE 7.9 Lateral and frontal three dimensional. (From Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.)

bones of the human skeleton published by Spiers.42 These so-called Spiers factors, or later “King–Spiers factors,”41 have also been integrated into the CT number method, developed by Zankl and Wittmann,58 which for the first time applied a heterogeneous tissue distribution to the skeleton of a phantom. While the University of Leeds was the center of skeletal dosimetry for about four decades, today the most advanced methods in this area come from the University of Florida. Bolch and his coworkers from the Nuclear and Radiological Engineering Department succeeded in segmenting trabecular bone and marrow in μCT and NMR images of the spongiosa of human bone samples, and subsequently connected these segmented images to a MC radiation transport code for the purposes of skeletal dosimetry.59–62 Spongiosa in the human skeleton is usually surrounded by cortical bone. Therefore, the group at the University of Florida developed a special MC transport code, called paired-image radiation transport (PIRT), which transports the particles through a “macro” matrix with a voxel size of some hundred micron, representing spongiosa, cortical bone and surrounding soft tissues, and at the same time through a “micro” matrix with cubic voxel sizes down to 30 μm, representing the microstructure of spongiosa with segmented volumes of marrow and trabecular bone. So far this new method has mainly been applied to isolated bone samples for nuclear medicine applications, but not to a complete skeleton embedded in a human body so far. 7.3.2.1.3 The 8 SP (Systematic–Periodic) Cluster Method The 8 SP cluster method for skeletal dosimetry was introduced in two publications.10,11 The first paper developed the fundamentals of the method, while the second paper

186

Handbook of Anatomical Models for Radiation Dosimetry

added the algorithm, which allowed for the calculation of the BSC equivalent dose in subvolumes of segmented trabecular microvoxels. The description given here is a compressed one. Readers interested in all the details are kindly requested to consult the original publications.10,11 In recent years, samples of trabecular bone extracted from human skeletons have been scanned for various reasons with microcomputed tomography using resolutions typically between 15 and 80 μm. The results are three-dimensional μCT images of human spongiosa, which depict the trabecular bone structure and the marrow cavities, like the one shown in Figure 7.12, which represents one of many images available to the computational dosimetry group at the DEN/UFPE for skeletal dosimetry studies. These μCT images are stacks of rectangular 2D digital images, which form parallepipeds with dimensions varying between 6.7 mm × 6.7 mm × 6.0 mm and 22.8 mm × 17.1 mm × 4.2 mm. The segmented spongiosa macrovoxels of the FAX06 and the MAX06 phantoms have cubic voxel dimensions of 1.2 mm. Introducing the μCT images into the spongiosa macrovoxels means that one has to extract 1.2 mm cubes from the μCT images, here called micromatrices. The MAX06 and the FAX06 phantoms have about 2 million and 1.5 million 1.2 mm cubic spongiosa macrovoxels, respectively. Applying the μCT images of spongiosa to skeletal

FIGURE 7.10 Three-dimensional frontal view of the skeletons of the FAX06 phantom view of the surface of the FAX06 (on the left) and the MAX06 (on the right) phantoms. (From Kramer, R. et al., Phys. Med. Biol., 51, 3331, 2006. With permission.)

FIGURE 7.11 (See color insert following page 524.) Frontal and lateral views of the skeletons and internal organs of the FAX06 and the MAX06 phantoms (adipose and muscle tissues removed).

The FAX06 and the MAX06 Computational Voxel Phantoms

187

FIGURE 7.12 3D μCT image of vertebral trabecular bone.

dosimetry one might think at first about filling these millions of spongiosa macrovoxels of a human skeleton with different micromatrices made of microvoxels containing marrow or trabecular bone. Apart from the expenditure of coding millions of micromatrices, however, this would lead to an explosion of memory space and execution time of the MC code; i.e., this is simply not feasible. Instead, for a given trabecular bone volume fraction, the method developed at the DEN/ UFPE calculates the equivalent dose to the RBM and the BSC in the spongiosa of the MAX06 and the FAX06 phantoms based on a cluster (=parallelepiped made of a small number) of micromatrices observing the following conditions: 1. The cluster of micromatrices must be selected from the CT image in such a way that the trabecular bone volume fraction of the cluster is equal to that of the whole CT image 2. The number of micromatrices of the cluster can be reduced as long as the equivalent doses to the RBM and the BSC do not change within the margins of the statistical errors During MC radiation transport, every time a particle enters a spongiosa voxel coming from a cortical bone voxel the transport is transferred to a micromatrix made of trabecular bone and cavities filled with marrow and BSC, whose matrix index (= its position in the cluster) is selected depending on the particle’s position in the phantom’s macromatrix and in the cluster. However, when a particle travels from one spongiosa voxel to another, then the index of the micromatrix is determined as the index of the neighboring micromatrix in the cluster,

188

Handbook of Anatomical Models for Radiation Dosimetry

or the algorithm assumes that the micromatrices repeat periodically if the particle leaves the cluster. Compared to a random selection of micromatrices, this procedure is systematic and periodic, thereby preserving coherently the spongiosa structure represented by the cluster. If the particle leaves the spongiosa, transport is retransferred to the macromatrix of the phantom. During particle transport through the micromatrices, energy deposition is registered in the marrow cavities. If the interaction takes place within a distance of 10 μm from trabecular bone surfaces or from cortical bone surfaces neighboring spongiosa, a special algorithm assigns a fraction of the energy to be deposited to the BSC based on the step length inside the 10 m layer. Otherwise the entire energy of that step is scored to the RBM equivalent dose taking into account the cellularity factor for the corresponding bone. Kramer et al.10,11 have shown that the SP selection applied to a cluster of only 2 × 2 × 2 = 8 micromatrices is capable of accurately calculating the whole-body RBM and BSC equivalent doses for external exposure to photons. Figure 7.13 represents a graphical display of the 8 SP cluster method.

3D micro-CT image of spongiosa Resolution: 30 μm Trabecular bone volume: 15% Size: 7.9 mm × 7.9 mm × 7.7 mm

Cluster of 81.2 mm cubic micromatrices with 15% trabecular bone volume extracted from the 3D micro-CT image

1 micromatrix systematically periodically selected from the cluster at runtime to be used in a spongiosa voxel

Spongiosa

FIGURE 7.13 The use of a cluster of eight micromatrices with 15% trabecular bone volume extracted from a μCT image scanned at a resolution of 30 μm. (From Kramer, R. et al., Phys. Med. Biol., 51, 6265, 2006. With permission.)

The FAX06 and the MAX06 Computational Voxel Phantoms

189

7.3.2.1.4 RBM and BSC Equivalent Doses for External Exposure to Photons 7.3.2.1.4.1 Comparison with RBM Data Determined by Other Methods The results of the 8 SP cluster method based on a vertebral bone sample scanned at 30 μm resolution are compared in Figure 7.14 with the results of the 3CF and the FDR method for the equivalent dose to the RBM of the FAX06 phantoms. The 3CF and FDR CCs are greater than the 8 SP cluster CC for the whole range of incident photon energies between 10 keV and 10 MeV. The reasons for this difference are the additional use of cortical bone and medullary YBM apart from spongiosa for RBM equivalent dose calculation and the application of kerma approximation when using the 3CF/FDR methods. In contrast, the 8 SP cluster method calculates the equivalent dose to the RBM only in soft tissue filled cavities of spongiosa surrounded by cortical and trabecular bone based on secondary electron transport. For external exposure to a real human skeleton, the radiosensitive soft tissues are shielded by cortical, and to a lesser extent, by trabecular bone, especially for photon energies below 500 keV. What serves as a shielding in real human skeletons, becomes a region contributing to the RBM equivalent dose in the case of the homogeneous skeleton used by the 3CF/FDR methods. Figure 7.14 shows that energy depositions by photons in “homogenized cortical bone” and also in “homogenized medullary YBM” increase the equivalent dose to the RBM above the values given by the 8 SP cluster method. The latter considers as contributing regions only the trabecular cavities filled with marrow having densities between 0.99 and 1.02 g cm−3 depending on the cellularity, whereas the 3CF/FDR methods use all skeletal regions with a density of 1.4 g cm−3 as contributors to the RBM equivalent dose, which increases the equivalent dose even more. For incident photon energies above 500 keV the shielding effect of cortical bone becomes less important, because secondary electrons have increasingly enough kinetic energy to penetrate the cortical bone shielding surrounding the spongiosa voxels, and consequently the differences between the 8 SP cluster and the 3CF/FDR curves in Figure 7.14 become smaller. But the 3CF/FDR data remain greater than the 8 SP cluster results especially when incident photon energies approach the range of 3–10 MeV, because the use of the kerma

Equivalent dose/air kerma (Sv/Gy)

1.0 0.8 0.6 0.4 RED BONE MARROW Fax06 phantom Whole skeleton AP BSC thickness: 10 μm

0.2 0.0 0.01

0.1

Photon energy (MeV)

8 SP cluster

FDR

1

10 3CF

FIGURE 7.14 RBM equivalent dose per air kerma free-in-air in the FAX06 phantom as a function of the photon energy for AP-incidence for different methods of skeletal dosimetry. FDR, fluence-to-dose response functions; 3CF, three correction factor method; 8 SP cluster, 8 SP cluster method based on a vertebral bone sample scanned at 30 μm resolution. (From Kramer, R. et al., Phys. Med. Biol., 51, 6265, 2006. With permission.)

Handbook of Anatomical Models for Radiation Dosimetry

190

Equivalent dose/air kerma (Sv/GY)

approximation “keeps the energy deposited in the skeleton,” whereas the 8 SP cluster results suggest that the RBM equivalent dose from the secondary electrons entering, created in and leaving the marrow cavities remains constant or finally decreases, implicating a net electron escape from the marrow cavities with increasing energy. Differences between RBM equivalent doses calculated with the 8 SP cluster method and the older methods may also arise from different trabecular bone volume distributions, a fact which on the other hand may explain additionally the relatively good agreement between the 3CF and the FDR methods, because, apart from both methods using homogeneous skeletons and kerma approximation, these two methods are based on the same trabecular chord-length distribution published by Darley54 and Beddoe et al.43 Explanations given above for the differences between the 8 SP cluster and the 3CF/FDR results apply also to the interpretation of the differences to be observed in Figure 7.15, which compares BSC equivalent doses calculated with different methods. Like before, “FDR” represents the FDR functions, here calculated for a thickness of 10 m, while “SKEL” represents the average equivalent dose to the homogeneous skeletal mixture, because the 3CF method does not provide correction factors for the calculation of the BSC equivalent dose. In the past, the “SKEL” equivalent dose was considered to be a conservative estimate for the BSC equivalent dose, which, according to Figure 7.15, is not true, compared to the FDR and the 8 SP cluster results for incident photon energies above 80 and 200 keV, respectively. The 8 SP cluster results are smaller than the FDR results for the above-mentioned reasons, namely the shielding by cortical bone and the secondary electron escape from the spongiosa for high incident photon energies. All BSC CCs are significantly greater than the RBM CCs shown in Figure 7.14, however, because of the additional contribution to the BSC equivalent dose by the photoelectrons released in trabecular bone entering the marrow cavities. These electrons have short ranges; i.e., they deposit their energy in the cavities mainly close to the trabecular bone surface. Figures 7.14 and 7.15 indicate that methods based on energy deposition in all skeletal regions filled with the same average elemental composition using kerma approximation 2.5 Bone surface cells Max06 phantom Whole skeleton AP BSC thickness: 10 μm

2.0 1.5 1.0 0.5 0.0 0.01

0.1

1

10

Photon energy (MeV) 8 SP cluster

FDR

SKEL

FIGURE 7.15 BSC equivalent dose per air kerma free-in-air in the FAX06 phantom as a function of the photon energy for AP-incidence for different methods of skeletal dosimetry. FDR, fluence-to-dose response functions; SKEL, average equivalent dose in the homogeneous mixture; 8 SP cluster, 8 SP cluster method based on a vertebral bone sample scanned at 30 μm resolution.

The FAX06 and the MAX06 Computational Voxel Phantoms

191

will usually overestimate the equivalent dose to the RBM and the BSC, mainly because these methods neglect the shielding effect of cortical bone for lower photon energies; and, additionally the escape of secondary electrons from the trabecular cavities for higher photon energies. At least, usually these earlier methods give a conservative estimate of the equivalent dose to the RBM and the BSC.

Equivalent dose/kerma in air (SV/Gy)

7.3.2.1.4.2 Comparison with RBM Data Determined with Other Exposure Models Figure 7.16 shows male RBM equivalent doses per air kerma free-in-air as a function of the photon energy for AP-incidence for the MAX06 phantom calculated with the 8 SP cluster method, and for the following male phantoms: MAX,5 ADAM,63 GOLEM,64 VIP-MAN,65,66 NORMAN-05,67 and REX.68 ADAM is a MIRD5-type mathematical phantom, while all other computational phantoms are true to nature voxel-based phantoms. MAX06, MAX, NORMAN-05, and REX have organ and tissue masses adjusted to ICRP-recommended data, whereas VIP-MAN represents an adult male much taller and heavier than the ICRP reference adult male, and GOLEM has about 4 kg less whole-body weight compared to the reference weight. The methods applied to the calculation of RBM equivalent dose vary significantly. For the ADAM phantom the 3CF method based on early Spiers factors27 and on an outdated distribution of RBM mass28 was used in a homogeneous skeleton. The CTN method combined with the 3CF method using the KS factors41 was applied to a heterogeneous skeleton in the MAX phantom, while the same method but with the older Spiers factors was used for the GOLEM phantom. For the NORMAN-5 phantom the authors developed a dose factor based on the photon energy and the MEA coefficients of RBM and the skeletal mixture to be multiplied with the photon fluence in a homogeneous skeleton with bone-specific materials taken from ICRU 46,69 which actually represents a type of FDR method. For the REX phantom, considered the adult male ICRP reference phantom, the authors apply a type of modified 3CF method. They use this method in segmented spongiosa

1.2 1.0

Red bone marrow AP

0.8 0.6

CTN: CT number method

0.4

8SP: 8SP cluster method 3CF: 3 correction factor method

0.2 0.0 0.01

0.1

1

10

Photon energy (MeV) MAX_CTN/3CF

MAX06_8 SP

GOLEM_CTN/3CF

NORMAN-05

VIPMAN

REX

ADAM_3CF

FIGURE 7.16 RBM equivalent dose per air kerma free-in-air as a function of the photon energy for AP-incidence for different male phantoms and methods of skeletal dosimetry.

192

Handbook of Anatomical Models for Radiation Dosimetry

voxels applying, among others, the KS factors. As the KS factors have been determined by King and Spiers41 for photon radiation incident on a vertebral body, their application only in the spongiosa is problematic because the shielding by the cortical bone shell is neglected. As a consequence, the photon energy used for the KS factors is too high, because not degraded by the cortical bone shell, which probably leads to elevated RBM equivalent doses. Another reason for the REX RBM equivalent doses being higher than the MAX06 RBM equivalent doses could be the fact that in order to match ICRPrecommended skeletal volumes exactly, the authors70 had to make anatomical compromises with regard to the representation of the cortical bone shell surrounding the spongiosa. There are some regions in the skeleton of the REX phantom where the spongiosa is not embedded in cortical bone, which leads to higher RBM equivalent doses because of the missing shielding. Taking advantage of the color photographs of the Visible Human,25 the authors of the VIP-MAN phantom65,66 have segmented the RBM in order to provide a direct macroscopic calculation of RBM equivalent dose. The VIP-MAN RBM CC shown in Figure 7.16 is significantly greater and smaller than the CCs for all other phantoms below 70 keV and above 3 MeV, respectively. This suggests that below 70 keV the VIP-MAN RBM is less shielded by cortical and trabecular bone than the RBM in all other phantoms. Above 3 MeV the VIPMAN RBM decreases because of secondary electron escape. The MAX06 and the REX CCs include also secondary electron escape, however, for these phantoms this effect becomes visible only above 6 MeV showing a much smaller decrease. A comparison of the RBM equivalent doses of the MAX, GOLEM, ADAM, and VIP-MAN phantoms has been discussed earlier,5 identifying, for example, the use of the older Spiers factors, a significantly higher calcium content in bone and less body mass as main reasons for the GOLEM RBM curve showing greater values than the MAX RBM curve. Although the CTN method has a conceptual defect as discussed by Kramer et al.,8 the MAX curve and the MAX06 curve agree reasonably well, because both methods use heterogeneous skeletons, where spongiosa voxels are surrounded by cortical bone voxels. It seems that the shielding effect of cortical bone especially at low photon energies is mainly responsible for the MAX and the MAX06 RBM curves being well below the NORMAN-05 RBM curve, which uses a homogeneous skeleton with relatively great average skeletal densities. Reduced RBM and BSC equivalent doses due to shielding by cortical bone compared to the results from methods that use homogeneous skeletal mixtures have also been observed by Lee et al.71 in a study on external photon exposure to different bone specimens using the PIRT method.62 In addition these authors also found good agreement between their CT image-based method and the CTN method. Figure 7.17 shows female RBM equivalent doses per air kerma free-in-air as a function of the photon energy for AP-incidence for the FAX06 phantom calculated with the 8 SP cluster method, for the FAX phantom calculated with the CTN/3CF method, and for the following female phantoms: EVA,63 DONNA,64 and REGINA.68 The arguments of the discussion for the male CCs in Figure 7.16 also apply here for the RBM equivalent doses of the female phantoms. The whole-body mass of the DONNA phantom, however, is not less, but is significantly greater than, the reference mass. Consequently the RBM equivalent dose of DONNA is less than the RBM equivalent dose of GOLEM. 7.3.2.1.5 Effective Dose for External Exposure to Photons Kramer et al.4,5 have discussed effective dose comparisons between the FAX and the MAX phantoms and other computational phantoms. Similar data for the FAX06 and the MAX06 phantoms will be calculated soon, because the discussion within ICRP with respect to the

The FAX06 and the MAX06 Computational Voxel Phantoms

193

Equivalent dose/kerma in air (Sv/Gy)

1.2 1.0

Red bone marrow AP

0.8 0.6 CTN: CT number method

0.4

3CF: 3 correction factor method 0.2

8SP: 8SP cluster method

0.0 0.01

0.1

1

10

Photon energy (MeV) FAX_CTN/3CF

FAX06_8 SP

EVA_3CF

REGINA

DONNA_CTN/3CF

Equivalent dose/kerma in air (Sv/Gy)

FIGURE 7.17 RBM equivalent dose per air kerma free-in-air as a function of the photon energy for AP-incidence for different female phantoms and methods of skeletal dosimetry. 1.5 Effective dose AP 1.0

0.5

0.0 0.01

0.1

1

10

Photon energy (MeV) REX/REGINA

MAX06/FAX06

FIGURE 7.18 Effective dose per air kerma free-in-air as a function of the photon energy for AP-incidence for the MAX06/ FAX06 and the REX/REGINA phantoms.

tissues and organs to be included in the calculation of the effective dose terminated only shortly before the approval of the new recommendations in March 2007. As a preview, Figure 7.18 shows a comparison of the effective dose between the FAX06/ MAX06 and the REX/REGINA68 exposure computational phantoms. The percentage differences are 60%, 31.4%, 19.3%, 10.0%, 5.7%, 2.9%, and 1.8% for incident photon energies of 10, 15, 20, 30, 40, 50, and 60 keV, respectively. It is difficult to explain these differences because anatomical descriptions of the REX and the REGINA phantoms have not been published so far. But one can assume that anatomical and/or compositional differences between frontal superficial organs and tissues of the phantoms involved

194

Handbook of Anatomical Models for Radiation Dosimetry

are responsible for the observed percentage differences. The value of the REX/REGINA effective dose for 70 keV seems to be too small, an error also to be observed for the RBM equivalent doses in Figures 7.16 and 7.17. Between 100 keV and 1 MeV the effective doses for the two exposure computational phantoms agree very well, which also supports the assumption that the deviations seen for lower energies are related to differences between superficial organs. Noticeable differences can be observed also above 1 MeV; however they never exceed 5%.

7.4 Conclusions Human phantom development is an ongoing process. Meanwhile new types of phantoms, called hybrid phantoms,72,73 have been published, which supersede the traditional voxel phantoms with respect to the true to nature representation of anatomical structures. Nevertheless, FAX06 and MAX06 represent milestones in the phantom development, because they were the first adult human voxel phantoms with ICRP89-based organ and tissue masses, with ICRP89-based segmented cortical bone and spongiosa in their skeletons and with a trabecular microstructure introduced into their spongiosa voxels where skeletal dosimetry based on μCT images of trabecular bone has been successfully applied for the first time for a complete skeleton embedded in a human body. The ICRP is going to recommend the use of traditional voxel phantoms for radiation protection. Therefore it seems that this type of human phantom will be still around for some time to come. Readers interested in using the FAX06 and/or the MAX06 phantoms or software tools based on these phantoms should visit the Web site http://www.grupodoin.com and follow the link Caldose.

Acknowledgments We acknowledge with thanks the permission of IOP Publishing Ltd. in allowing reproduction of previously published material from the journal Physics in Medicine and Biology. The authors would also like to thank the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and the Fundação de Amparo à Ciência do Estado de Pernambuco (FACEPE) for their financial support.

References 1. Jones, A.R. Proposed calibration factors for various dosimeters at different energies, Health Phys, 12, 663, 1966. 2. Fisher, H.L. and Snyder, W.S. Distribution of dose in the body from a source of gamma rays distributed uniformly in an organ, Report No. ORNL-4168, Oak Ridge National Laboratory, Oak Ridge, TN, 1967. 3. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Pergamon Press, Oxford, 1991.

The FAX06 and the MAX06 Computational Voxel Phantoms

195

4. Kramer, R. et al. All about FAX: A Female Adult voXel phantom for Monte Carlo calculation in radiation protection dosimetry, Phys Med Biol, 49, 5203, 2004. 5. Kramer, R. et al. All about MAX: A male adult voxel phantom for Monte Carlo calculations in radiation protection dosimetry, Phys Med Biol, 48, 1239, 2003. 6. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, 2003. 7. ICRP. The 2007 recommendations of the International Commission on Radiological Protection, ICRP publication 103, Ann ICRP, 37, 1, 2007. 8. Kramer, R. et al. MAX06 and FAX06: Update of two adult human phantoms for radiation protection dosimetry, Phys Med Biol, 51, 3331, 2006. 9. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: The Skeleton, ICRP Publication 70, Pergamon Press, Oxford, 1995. 10. Kramer, R. et al. Skeletal dosimetry in the MAX06 and the FAX06 phantoms for external exposure to photons based on vertebral 3D-microCT images, Phys Med Biol, 51, 6265, 2006. 11. Kramer, R. et al. Skeletal dosimetry for external exposure to photons based on microCT images of spongiosa from different bone sites, Phys Med Biol, 52, 6697, 2007. 12. Kramer, R. et al. Equivalent dose to organs and tissues in hysterosalpingography calculated with the FAX (Female Adult voXel) phantom, Br J Radiol, 79, 893, 2006. 13. Kramer, R. et al. MAX meets ADAM: A dosimetric comparison between a voxel-based and a mathematical model for external exposure to photons, Phys Med Biol, 49, 887, 2004. 14. Kramer, R. et al. Application of the MAX/EGS4 exposure model to the dosimetry of the Yanango radiation accident, Phys Med Biol, 50, 3681, 2005. 15. Kramer, R., Khoury, H.J., and Vieira, J.W. CALDOS_X—An on-line interface for absorbed dose assessment in diagnostic radiology, 11th International Conference on Radiation Shielding, Callaway Gardens, Pine Mountain, GA, April 13–18, 2008. 16. Kramer, R., Khoury, H.J., and Vieira, J.W. CALDOS_X—A software tool for the assessment of organ and tissue absorbed doses in diagnostic radiology, 12th International Congress of the International Radiation Protection Association, Buenos Aires, Argentina, 19–24 October, 2008. 17. Nelson, W.R., Hirayama, H., and Rogers, D.W.O. The EGS4 Code System, SLAC-265, Stanford Linear Accelerator Centre, Stanford University, Stanford, CA, 1985. 18. Kawrakow, I. Accurate condensed history Monte Carlo simulation of electron transport: I. EGSnrc, the new EGS4 version., Med Phys, 27, 485, 2000. 19. Kawrakow, I. and Rogers, D.W.O. The EGSnrc code system: Monte Carlo simulation of electron and photon transport, NRC Report PIRS-701, National Research Council of Canada, Ottawa, 2003. 20. Allison, J. et al. Geant4 developments and applications, IEEE Trans Nucl Sci, 53, 270, 2006. 21. Cassola, V.F. et al. Organ equivalent doses in the MAX06 and the FAX06 phantoms for external exposure to photon radiation calculated with the EGSnrc and the GEANT4 Monte Carlo codes, International Nuclear Atlantic Conference, Santos, SP, Brazil, September 30–October 5, 2007. 22. Zubal, I.G. et al. Computerized three-dimensional segmented human anatomy, Med Phys, 21, 299, 1994. 23. Zubal, I.G. et al. High resolution, MRI-based, segmented, computerized head phantom, in The Zubal Phantom Data, Voxel-Based Anthropomorphic Phantoms, http://noodle.med.yale.edu/ phantom, 2001. 24. Stuchly, M. The Zubal Phantom Data, Voxel-based Anthropomorphic Phantoms, http://noodle.med. yale.edu/phantom, 1996. 25. Spitzer, V.M. and Whitlock, D.G. Atlas of the Visible Human Male, Jones and Bartlett Publishers, Sudbury, MA, 1998. 26. Sjogreen, K. The Zubal Phantom Data, Voxel-Based Anthropomorphic Phantoms, http://noodle.med. yale.edu/phantom, 1998. 27. Spiers, F.W. Sources, fields, measurements and applications, in Transition-Zone Dosimetry Radiation Dosimetry, vol. III, 2nd edn., Academic Press, New York, 1969. 28. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Pergamon Press, Oxford, 1975.

196

Handbook of Anatomical Models for Radiation Dosimetry

29. Ligier, Y. et al. Osiris—A medical image-manipulation system, M D Computing, 11, 212, 1994. 30. Loureiro, E.C., Lima, F.R., and Stabin, M.G. A voxel-based head-and-neck phantom built from tomographic colored images, Cell Mol Biol (Noisy-le-grand), 48, 461, 2002. 31. ICRP. Human Respiratory Tract Model for Radiological Protection: A Report of a Task Group of the International Commission on Radiological Protection, ICRP Publication 66, Pergamon Press, Oxford, 1994. 32. Netter, F.H. Atlas de Anatomia Humana, 3rd edn., Porto Alegre, Artmed, 1998. 33. Sobotta, J. Atlas de Anatomia Humana, 20th edn., Guanabara Koogan, Rio de Janeiro, 1995. 34. SCION Image for WINDOWS. http://www.scioncorp.com, 2001. 35. ICRU. Tissue substitutes in radiation dosimetry and measurement, ICRU Report 44, Bethesda, MD, 1989. 36. Brindle, J. and Bolch, W. private communication, November 14, 2005. 37. Bolch, W.E. et al. Skeletal absorbed fractions for electrons in the adult male: Considerations of a revised 50-mu m definition of the bone endosteum, Radiat Prot Dosimetry, 127, 169, 2007. 38. Richardson, R.B., Nie, H.L., and Chettle, D.R. Monte Carlo simulation of trabecular bone remodelling and absorbed dose coefficients for tritium and C-14, Radiat Prot Dosimetry, 127, 158, 2007. 39. Watchman, C.J. et al. Spatial distribution of blood vessels and CD34(+) hematopoietic stem and progenitor cells within the marrow cavities of human cancellous bone, J Nucl Med, 48, 645, 2007. 40. ICRP. Limits for Intake of Radionuclides by Workers, ICRP Publication 30, Pergamon Press, Oxford, 1979. 41. King, S.D. and Spiers, F.W. Photoelectron enhancement of the absorbed dose from x-rays to human-bone marrow—Experimental and theoretical studies, Br J Radiol, 58, 345, 1985. 42. Spiers, F.W. Interim report on the determination of dose to bone marrow from radiological procedures, Br J Radiol, 36, 238, 1963. 43. Beddoe, A.H., Darley, P.J., and Spiers, F.W. Measurements of trabecular bone-structure in man, Phys Med Biol, 21, 589, 1976. 44. Beddoe, A.H. and Spiers, F.W. Comparative-study of the dosimetry of bone-seeking radionuclides in man, Rhesus-monkey, beagle, and miniature Pig, Radiat Res, 80, 423, 1979. 45. Spiers, F.W. Radionuclides and bone from Ra-226 to Sr-90—Silvanus Thompson Memorial Lecture, Br J Radiol, 47, 833, 1974. 46. Spiers, F.W. Particle dosimetry in bone and the toxicity of bone-seeking radionuclides, Phys Med Biol, 33, 395, 1988. 47. Spiers, F.W., Beddoe, A.H., and Whitwell, J.R. Mean skeletal dose factors for beta-particle emitters in human bone. 1. Volume-seeking radionuclides, Br J Radiol, 51, 622, 1978. 48. Spiers, F.W., Beddoe, A.H., and Whitwell, J.R. Mean skeletal dose factors for beta-particle emitters in human-bone. 2. Surface-seeking radionuclides, Br J Radiol, 54, 500, 1981. 49. Spiers, F.W., Whitwell, J.R., and Beddoe, A.H. Calculated dose factors for radiosensitive tissues in bone irradiated by surface-deposited radionuclides, Phys Med Biol, 23, 481, 1978. 50. Whitwell, J.R. and Spiers, F.W. Calculated beta-ray dose factors for trabecular bone, Phys Med Biol, 21, 16, 1976. 51. Eckerman, K.F. and Stabin, M.G. Electron absorbed fractions and dose conversion factors for marrow and bone by skeletal regions, Health Phys, 78, 199, 2000. 52. Stabin, M.G. et al. Evolution and status of bone and marrow dose models, Canc Biother Rad, 17, 427, 2002. 53. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources I: Methods, ORNL/TM-8381/V1, Oak Ridge National Laboratory, Oak Ridge, TN, 1987. 54. Darley, P.J. Measurements of linear path length in bone and bone marrow using a scanning technique, Proceedings on the Symposium on Microdosimetry, E.A.E.C. Report EUR d-f-e, Ispra, Italy, 1968.

The FAX06 and the MAX06 Computational Voxel Phantoms

197

55. Rosenstein, M. Organ doses in diagnostic radiology, DHEW Publication (FDA) 76-8030, US Government Printing Office, Washington, DC, 1976. 56. Kramer, R. Ermittlung von Konversionsfaktoren zwischen Körperdosen und relevanten Strahlungskenngrößen bei externer Röntgen- und Gamma-Bestrahlung, GSF-Report-S-556, Institut für Strahlenschutz, GSF-Forschungszentrum für Umwelt und Gesundheit, NeuherbergMuenchen, 1979. 57. Kramer, R. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods: Part I. The male (ADAM) and female (EVA) adult mathematical phantoms, GSF-Report S-885, Institut fuer Strahlenschutz, GSF-Forschungszentrum fuer Umwelt und Gesundheit, Neuherberg-Muenchen, 1982. 58. Zankl, M. and Wittmann, A. The adult male voxel model “Golem” segmented from wholebody CT patient data, Radiat Environ Biophys, 40, 153, 2001. 59. Shah, A. et al. Accounting for beta-particle energy loss to cortical bone via Paired-Image Radiation Transport (PIRT), Med Phys, 32, 1354, 2005. 60. Shah, A.P. et al. A paired-image radiation transport model for skeletal dosimetry, J Nucl Med, 46, 344, 2005. 61. Shah, A.P. et al. Chord-based versus voxel-based methods of electron transport in the skeletal tissues, Med Phys, 32, 3151, 2005. 62. Shah, A.P. et al. A comparison of skeletal chord-length distributions in the adult male, Health Phys, 89, 199, 2005. 63. Zankl, M. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods. Part VII: Organ doses due to parallel and environmental exposure geometries, GSF-Report 8/97, Institut für Strahlenschutz, GSF-Forschungszentrum für Umwelt und Gesundheit, Neuherberg-Muenchen, 1997. 64. Zankl, M. et al. Tomographic anthropomorphic models. Part IV: Organ doses for adults due to idealized external photon exposures, GSF-Report 13/02, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 2002. 65. Chao, T.C., Bozkurt, A., and Xu, X.G. Conversion coefficients based on the VIP-Man anatomical model and EGS4-VLSI code for external monoenergetic photons from 10 keV to 10 MeV, Health Phys, 81, 163, 2001. 66. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Phys, 78, 476, 2000. 67. Ferrari, P. and Gualdrini, G. An improved MCNP version of the NORMAN voxel phantom for dosimetry studies, Phys Med Biol, 50, 4299, 2005. 68. Schlattl, H., Zankl, M., and Petoussi-Henss, N. Organ dose conversion coefficients for voxel models of the reference male and female from idealized photon exposures, Phys Med Biol, 52, 2123, 2007. 69. ICRU. Photon, electron, proton and neutron interaction data for body tissues, Report 46, International Commission on Radiation Units and Measurements, Bethesda, MD, 1992. 70. Zankl, M., Eckerman, K.F., and Bolch, W.E. Voxel-based models representing the male and female ICRP reference adult—The skeleton, Radiat Prot Dosimetry, 127, 174, 2007. 71. Lee, C. et al. An assessment of bone marrow and bone endosteum dosimetry methods for photon sources, Phys Med Biol, 51, 5391, 2006. 72. Lee, C. et al. Hybrid computational phantoms of the male and female newborn patient: NURBSbased whole-body models, Phys Med Biol, 52, 3309, 2007. 73. Xu, X.G. et al. A boundary-representation method for designing whole-body radiation dosimetry models: Pregnant females at the ends of three gestational periods–RPI-P3, -P6 and -P9, Phys Med Biol, 52, 7023, 2007.

8 The University of Florida Pediatric Phantom Series Choonsik Lee, Daniel L. Lodwick, Deanna Hasenauer Pafundi, Scott R. Whalen, Jonathan L. Williams, and Wesley E. Bolch

CONTENTS 8.1 Introduction ............................................................................................................... 199 8.2 Materials and Methods ............................................................................................ 200 8.2.1 UF Newborn Voxel Phantom ..................................................................... 200 8.2.2 Phantom Development Procedure ............................................................ 200 8.2.2.1 Polygonization of the Voxel Phantom ....................................... 200 8.2.2.2 NURBS Modeling ......................................................................... 201 8.2.2.3 Voxelization of the Polygon Computational Phantom............ 202 8.2.3 Standardization of Hybrid Phantom ........................................................ 203 8.2.3.1 Match Body Dimension to Anthropometric Data ................... 203 8.2.3.2 The Matching of the Organ Volume to the ICRP Publication 89 Values ................................................................... 204 8.2.3.3 Alimentary and Respiratory Systems ....................................... 204 8.2.3.4 Creation of the Male Newborn Phantom from the Female Phantom .................................................................... 205 8.2.4 Extended Development of Pediatric Phantoms ....................................... 206 8.3 Results and Discussion ............................................................................................ 206 8.3.1 Voxelization Algorithm ............................................................................... 206 8.3.1.1 UF Hybrid Phantoms for Newborn Male and Female ........... 208 8.3.2 Comparison with UF Voxel Newborn Phantom and ICRP89 ............... 214 8.3.3 Advantages of Hybrid Newborn Phantom .............................................. 214 8.3.4 The Extended Development of Older Phantoms..................................... 216 8.3.5 Pediatric MicroCT-Based Skeletal Computational Phantoms ............... 217 8.4 Conclusions ................................................................................................................ 217 Acknowledgment ................................................................................................................. 218 References ............................................................................................................................. 218

8.1 Introduction This chapter reviews the development of a series of pediatric whole-body phantoms based on non-uniform rational B-spline (NURBS) surfaces as constructed at the Advanced Laboratory for Radiation Dosimetry Studies (ALRADS), University of FL, Gainesville, FL. We introduce the development procedure mainly for the newborn phantom, as that was the 199

200

Handbook of Anatomical Models for Radiation Dosimetry

first hybrid phantom in which NURBS technology was incorporated. Other older phantoms ranging from 1 to 15 years old were extensions of this newborn phantom. We further subdivided the phantom series into two groups—Group A and Group B. Group A is composed of male and female phantoms of the newborn, 1 year old, 5 years old, 10 years old, and 15 years old, where phantom stature, total weight, and individual organ masses are targeted to within 1% of ICRP Publication 89 reference values. While the newborn phantoms were constructed from whole-body CT scans of a newborn cadaver, the remaining phantoms of Group A were constructed from segmented head and chest–abdomen–pelvis CT images from live pediatric patients all of normal anatomy. Group B phantoms are constructed as upwardly and downwardly scaled versions of the Group A phantoms, thus providing a phantom at each 1 year interval, from newborn to 15 years old. The intent of the University of Florida (UF) pediatric series is to provide a reference library of phantoms for matching to individual patients in organ dose assessment in radiography, fluoroscopy, CT imaging, and radiotherapy. In traditional medical dose reconstruction, 50th percentile reference phantoms are assigned to individual patients based upon their age. In the UF approach, phantoms are assigned to individual patients, not based on age, but on trunk height, thus reducing residual uncertainties in organ volumes from as high as 50% to as low at 15%. Next, leg lengths are adjusted to match patient statue, and finally, the outer body contour of the phantom is adjusted to match as closely as possible the patient’s body distribution of subcutaneous fat and/or muscle. The overall approach thus yields a highly patient-specific phantom for dose assessment than afforded by existing applications. Finally, we discuss in this chapter our efforts to provide, through tissue sampling and microCT image analysis, detailed and agespecific skeletal computational phantoms for each phantom of the UF series.

8.2 Materials and Methods 8.2.1 UF Newborn Voxel Phantom UF hybrid newborn phantoms represent an evolution in the UF newborn voxel phantom developed by our same research group.1 We segmented the UF newborn voxel phantom from 485 CT slices of 6 day old female cadaver, which weighed 3.83 kg. The subject died in an attempt to correct congenital abnormalities of the great vessels, and was scanned within 24 h of death. Unlike typical CT scans of live patients, the arms were positioned parallel to the body and included within the scan coverage. We examined the CT images, and found that the cadaver was free of any physical defects that might cause problems in phantom development. We also employed two additional datasets from a 1 month old patient and a 2 month old male to replace the collapsed lungs of the 6 day old cadaver, and to model adrenal glands which were invisible in the original CT data, respectively. A total of 66 different regions were segmented through several image processing steps and manual contouring. 8.2.2 Phantom Development Procedure 8.2.2.1 Polygonization of the Voxel Phantom The UF voxel newborn phantom in the three-dimensional (3D) voxel matrix form should be converted into a polygon mesh form, which is the base framework for the NURBS surface

The University of Florida Pediatric Phantom Series

201

modeling. We used 3D-DOCTOR (Able Software Corp., Lexington, MA), a 3D modeling and image processing software for tomography data, to convert the voxel geometry of the UF newborn voxel phantom into a polygon mesh geometry. We imported a binary voxel phantom array consisting of 480 × 211 × 485 voxels into 3D-DOCTOR, and the organ contours were extracted from the binary file by utilizing an interactive segmentation tool organ-by-organ. The interactive segmentation tool automatically rendered organ contours on two-dimensional (2D) pixel slices with a certain organ tag assigned to the thresholding windows of the interactive segmentation tool. Contours for contents and gas in the intestinal tract such as the stomach, the colon, and the small intestine were not extracted, since we assumed that the intestinal tract was filled with soft tissue in the hybrid phantoms. We only obtained the outer contours from those walled organs, including the urinary bladder and the gall bladder, of which the wall thickness was exactly redesigned using NURBS surfaces. Both outer and inner contours, however, were acquired for bone sites such as cranium and vertebrae. Other than the internal organs, the skeletal contours were completely resegmented from the original CT data to more accurately describe the skeleton volumes of the newborn subject. After we obtained the contours of the whole anatomy in the voxel phantom, we generated 3D polygon mesh computational phantoms by the built-in 3D rendering function, and exported them into a Wavefront Object file format, which retains the dimensional, the position, and the organ tag information of the original voxel phantom. 8.2.2.2 NURBS Modeling We incorporated polygon mesh computational phantoms extracted from the voxel phantom into a NURBS modeling tool to generate NURBS anatomical surface computational phantoms. We employed Rhinoceros (McNeel North America, Seattle, WA), a NURBS modeling, rendering, and analysis software, to model the NURBS surfaces that exactly fit the polygon mesh computational phantoms generated from 3D-DOCTOR. To manipulate polygon anatomy computational phantoms independently and effectively, we generated Wavefront Object files from 3D-DOCTOR within five groups (body contour, alimentary system, respiratory system, other soft tissue organs, and the skeleton) and imported as five different layers in Rhinoceros. Each layer could be turned on or off, and was not affected by objects in other layers. Smooth NURBS surfaces were made from polygon mesh computational phantoms on an organ-by-organ basis. First, we obtained several contours from polygon mesh organ computational phantoms as needed, and the NURBS surfaces were fit to the contours using the LOFT tool. After generating NURBS surfaces, we removed the original polygon mesh computational phantoms from the geometry. The NURBS surfaces were generated for all of the internal organs and tissues, excluding the brain and skeleton, of which the shapes could not be effectively described by the NURBS surfaces. Pafundi et al. (2008) have presented the details of skeleton modeling in hybrid phantoms. Even though the eyeballs, lenses, ovaries, urinary bladder, and breast can be described by spherical or ellipsoidal shapes, it is usually difficult to model smooth 3D shapes by stacking manually segmented 2D contours in the development of a voxel phantom. In the context of the hybrid phantom, we carefully designed the NURBS-based stylistic computational phantoms by closely referring to their original shapes and positions. Eyes and ovaries were modeled as spherical objects, and the lens as a flatten ellipsoid. The breasts are not well developed at newborns, but were modeled by small button-like ellipsoid and placed at the corresponding position to facilitate breast dose calculation. The shape of the urinary bladder varies depending on the amount of urine inside, so this organ was also stylistically modeled by a walled ellipsoid. We modeled the contents

202

Handbook of Anatomical Models for Radiation Dosimetry

of the walled organs, such as the heart, the gall bladder, and the urinary bladder, by contracting organ surfaces. Each control points surrounding the NURBS surfaces was inwardly moved to a normal direction so that another NURBS surfaces for the contents could be generated. In this way, the volumes of walled organs were exactly matched to the reference data. We designed in a new way the tongue and the tonsils, of which the reference masses are presented in the International Commission on Radiological Protection (ICRP) Publication 89, in the UF hybrid phantoms. We described the tongue by bending and flattening a truncated ellipsoid. We modeled two palatine tonsils with spheres, and positioned them on the left and right sides toward the back of the tongue. We described the three layers of kidneys, the cortex, the medulla, and the pelvis, by contracting the original kidney volumes and matching each volume to the reference data of ICRP89: the cortex represents about 70% of the volume of the kidneys, the medulla about 25%, and the pelvis about 5%. 8.2.2.3 Voxelization of the Polygon Computational Phantom The existing Monte Carlo transport codes cannot recognize NURBS or polygon geometries, and so the NURBS computational phantoms cannot be directly incorporated into Monte Carlo radiation transport codes. A hybrid phantom voxelization process is thus key technology for the phantom in dosimetry studies. A voxelization algorithm for the purpose of this study should meet the following criteria. First, it must be able to voxelize multiple polygon objects involved in the whole-body NURBS computational phantom in the 3D voxel medium. Second, it should have the capability to handle outermost surfaces as well as artificial holes which are a problem in case of the cranium or vertebrae. Third, the original NURBS volumes must be as accurately maintained in the resulting voxel volume as much as possible. To meet these criteria, we wrote an in-house MATLAB® (The MathWorks, Inc., Natick, MA) code, named Voxelizer, exclusively for the purpose of this study. Voxelization was performed in three steps: triangulation, intersection calculation between polygon and z-grid, and ray-casting process. All smooth NURBS surfaces were triangulated using the built-in function of Rhinoceros, where the user can specify a meshing tolerance (MT) defined as the maximum angle between adjacent faces in the resulting polygon mesh. Smaller values of MT result in slower triangulation, more accurate meshes, and a higher polygon count. After the NURBS surfaces were triangulated into polygons, we saved the vertices of each polygon in ASCII Raw Triangles format. We then imported the vertex data into the Voxelizer, and calculated polygon-by-polygon the intersections between single polygons and z-grid, dictated by a predefined voxel resolution (VR). If the z-grid intersects a given polygon, two vertices of the intersection are stored in an array. Finally, we assigned user-defi ned organ tags to voxels involved in certain organ computational phantoms by determining if a voxel is inside or outside a given organ surface. In order to test whether or not a voxel is inside an object surface, one has to count the number of times a ray traveling from the center of the voxel to minus infinity of y-axis intersect surface of the polygon. If the number is odd, the point is inside the surface. How close the resulting voxel phantom is to the original NURBS computational phantom depends on two parameters: the MT and VR. To quantitatively understand the effect of those parameters on volume difference between NURBS computational phantom and voxel phantom, we performed a sensitivity study. The largest and smallest NURBS organs, the liver and the pituitary gland, were chosen for this sensitivity study. First, those organs were triangulated using five different MTs of 50°, 40°, 30°, 20°, and 10° and the volumes of

The University of Florida Pediatric Phantom Series

203

the resulting polygon computational phantoms were compared with those of the original NURBS computational phantoms. Second, they were voxelized using five different VRs, and the volumes of the voxelized organs were compared with the original polygon organ volumes. 8.2.3 Standardization of Hybrid Phantom It is generally desirable to make the computational phantom representative of various body dimensions and anatomical characteristics of a reference individual or patient to perform consistent and reproducible dosimetry calculations for radiation protection and medical situations. For the purpose of standardization, we made the following efforts by referring to the literature resources of anthropometric and anatomical data. 8.2.3.1 Match Body Dimension to Anthropometric Data To match the body dimensions of the newborn phantom to anthropometric data, we made two approaches to the NURBS computational phantom. First, we developed a deformable outer body contour and the initiative of a posture change to the original cadaver-based voxel phantom. We limited the original body contours extracted from UF voxel newborn phantom to fixed body dimensions and postures, and only the uniform scaling of the whole anatomy could be made by changing the segmented VR. We developed new body contours, in which dimensions of the head, torso, and limbs can be independently formed, from framework of the artistic 3D CHILD computational phantoms given by the 3D modeling community (http://www.3dcafe.com). We imported the polygon mesh computational phantom of the 3D CHILD into Rhinoceros, separately generated the NURBS surfaces for the head, the torso, the arms, the legs, the hands, and the feet, and combined them all into the whole-body contour. We carefully deformed each body contour to match the original body contour of the UF voxel newborn phantom by manipulating the control points surrounding the NURBS surfaces. We also deformed the body contours of the fingers and toes to fully represent the individual bones of the fingers and toes. Each body part of the resulting NURBS body contour is deformable to match any targeted anthropometric data. This deformability is also a useful feature used to accommodate the difference in body dimensions of other age groups, as well as different body shapes, such as the 25th or 75th height and weight percentiles. Second, we carefully modeled the original posture of the newborn cadaver in a curled posture with bent arms and legs, and thus represented a more neutral posture. We rotated and transported the arm bones (humerus, radii, ulnae, and hand bones) and the leg bones (femur, tibiae, fibulae, patellae, ankle, and feet bones) accordingly. Currently, three anthropometric data sets are available for the newborn: the standing height, the sitting height, and the head circumference (HC). We obtained the standing and the sitting heights for newborn from ICRP Publication 89.2 We took the Crown-Rump Length (CRL) of the newborn as a suitable substitute for sitting height. CRL is defined as the distance between the vertex of the skull and the ischial tuberosities. The standing height and the CRL of the reference newborn were 51 and 34 cm, respectively. We calculated the HC as 33.1 cm, using a correlation equation between HC and CRL as reported by Yang et al.3 We matched perfectly the standing height of the original UF voxel newborn phantom to the ICRP reference height, and the CRL was 2.9% shorter than ICRP data. The HC of the UF voxel newborn phantom was 32.5 cm, which was 1.7% shorter than the reference value.

204

Handbook of Anatomical Models for Radiation Dosimetry

8.2.3.2 The Matching of the Organ Volume to the ICRP Publication 89 Values We matched all of the organs and tissues in the hybrid phantom to the reference data of the ICRP Publication 89, with the exception of the brain, eye balls, skin, and walled organ contents. In the case of voxel phantom development, this matching process is usually performed by manually adding or removing pixels in 2D slice images, so that organ volumes can be accurately adjusted to the target values. As for hybrid phantoms, however, this matching process should be performed by manipulating the NURBS control points in the 3D medium, which is anatomically more realistic than pixel-based modification, as performed in voxel phantoms. Even though the volume of the eyeballs in the voxel phantom could be accurately matched to the reference values by modifying the pixel map of 2D slices, it is difficult to guarantee if the resulting 3D object would be anatomically realistic. As for hybrid phantoms, an expansion of the 3D spheres representing the eyeballs was not allowed beyond a certain space within the cranium, so it was impossible to increase the eye volume to the reference values. This issue was the same problem for the brain data, as the cranium restrained it in the same manner. No attempts were made to expand the cranium volume to accommodate these restrictions, since we gave first priority to a realistic skeletal computational phantom that faithfully represented the skeleton of the ICRP89 reference newborn. We adjusted the total body mass of the voxelized phantom to the reference mass of 3.5 kg by modifying the outer body contour. We generated skin by assigning a skin index to the outermost voxel layer following NURBS phantom voxelization. To match the skin volume to reference values, we adjusted the VR to the skin thickness of the Oak Ridge National Laboratory (ORNL) newborn phantom: 0.07 cm.4 We did not match the volumes of the contents of the walled organs, such as the small intestine, the colon, the stomach, the urinary bladder, and the heart, to the reference data as described in Lee et al.5 The NURBS surfaces of organs were so close to each other that a change in one organ’s volume would affect the volume of the neighboring organs. In an attempt to match the organ volumes to reference data at the NURBS modeling stage, every surface intersection should be avoided, since Rhinoceros cannot identify intersections between pairs of NURBS surfaces. Although overlapping volumes should be removed from those both organs, Rhinoceros independently calculates volumes of each organ without considering organ overlap. As a result, there could be differences in volumes reported in the NURBS computational phantom and the corresponding voxel phantoms. To avoid these volume change problems, we carefully adjusted and separated the NURBS organ surfaces from each other without sacrificing the original anatomical realism. Since it was difficult to completely recognize any intersections in the NURBS modeling stage, this adjustment was performed through an iterative process of adjusting the NURBS surfaces and recognizing the overlaps in the voxelized computational phantoms where the organ overlaps are easily seen in 2D image slices. 8.2.3.3 Alimentary and Respiratory Systems Special consideration was given to modeling the alimentary and respiratory systems. Although we obtained the organs in these two systems from the UF voxel newborn phantom, anatomical problems were evident, due to image discontinuities and other defects. These problems are most likely caused by poor vision-based segmentation of low contrast CT images. In the current study, we applied a stylistic approach to alimentary and respiratory systems in the context of hybrid phantom. First, the central luminal traces of the esophagus, the small intestine, the colon, and the rectosigmoid were approximately as given in the UF voxel newborn phantom and

The University of Florida Pediatric Phantom Series

205

her original CT images. In the small intestine, we found it nearly impossible to follow the central track, so that wavy contours were constructed according to reference anatomy resources; whereas we obtained those of the colon from the base voxel phantom. We generated a NURBS pipe structure of a given radius along these central tracks. We also made organ contents along the central track with a pipe of smaller radius. We did not attempt the separation of the wall and content for the esophagus, since it is nearly collapsed except when a food bolus is passing through. The stomach is transformable, depending on the stage of food digestion. Zankl et al.6 investigated the stomach dose of seven adult male and female voxel phantoms and reported that the stomach position varies largely among different individuals and even in a single individual. Therefore, a stylistic NURBS stomach computational phantom was redesigned referring to the stomach position and shape within the original voxel phantom, along with the anatomy literatures. We matched the anatomical features of the resulting NURBS alimentary system to reference data provided by the ICRP Publication 89.2 The publication reported reference lengths of the esophagus, the small intestine, and the colon as 10, 80, and 45 cm, respectively. These are physiological length measured in a living person, rather than anatomical lengths that are measured at autopsies, and are usually longer than the physiological lengths. We matched the central tracks of those three organs to the reference lengths within 5%. As for the colon, the original central track is much shorter than the reference length. Through a discussion with experienced radiologist, we extended the central track of transverse colon to the lower small intestine region. We also matched the organ masses of the esophagus, the stomach wall, the small intestine, and the colon to the reference masses of the ICRP 89. Radiosensitive tissues in the respiratory system are composed of an anterior nasal passage, a posterior nasal passage, a mouth, a nasal and an oral pharynx, a larynx, a trachea, a main bronchi, and the lungs.7 The anterior nasal passage is defined as extra-thoracic (ET)1, and ET2 comprises the posterior nasal passage, the mouth, the pharynx, and the larynx. Additionally, researchers categorize the trachea and the main bronchi into the bronchial region (BB). We designed and attached the external nose, including the two nostrils, to the face at the same position of the nose in the UF voxel newborn phantom. We obtained the central tracks of trachea and main bronchi from the newborn voxel phantom and the NURBS pipes, with a certain thickness generated along the tracks. We also matched the resulting respiratory system to the reference data provided in the ICRP 89. Two reference values, the masses of the larynx and the trachea, were available for the newborn individual: 1.18 and 0.45 g, respectively. We adjusted the volumes of the larynx and the trachea by manipulating the control points. The reference lung mass of newborn is 60 g, and the volume would be calculated as 202.7 cm3 by using a nominal lung density of 0.296 g cm−3 adopted by ORNL newborn stylized phantom.4 However, the volume of the NURBS lungs was 103.88 cm3, which is significantly smaller than the targeted reference volume at this density. Since any modification of the lung volumes was not permitted due to the fixed geometries of the rib cage, the heart, and the thymus, an effective lung density was assigned so that the total lung mass would match its reference mass. A similar approach was adopted by other investigators developing reference stylized or voxel phantoms.8,9 Lung density assigned to hybrid newborn phantom was 0.572 g cm−3.

8.2.3.4 Creation of the Male Newborn Phantom from the Female Phantom We developed the hybrid newborn phantom of the opposite sex using the original female phantom. The implicit assumption we made here is that there are no significant differences in organ volumes, positions, and depths between the reference male and the reference

Handbook of Anatomical Models for Radiation Dosimetry

206

female newborn child, with the exception of the reproductive organs. We obtained computational phantoms of the male-specific organs (scrotum, testes, and prostate) from the UF 9 month male voxel phantom10 and we inserted them in the hybrid female phantom following the corresponding removal of the ovaries and the uterus. We reduced the volumes of the male sex organs to match the reference data in ICRP 89. The bladder of the UF 9 month male voxel phantom also replaced the urinary bladder in the female phantom, since the uterus usually presses it anteriorly in the female anatomy. 8.2.4 Extended Development of Pediatric Phantoms We applied the same technology outlined above for the newborn phantom to the older phantom of the UF series: 1, 5, 10, and 15 year old, both male and females. Contours of major organs and tissues were converted or segmented from the previous UF voxel phantoms, and the computed tomography (CT) data as needed. We reconstructed and imported the polygon mesh computational phantoms for the major organs and tissues into Rhinoceros for the NURBS surface modeling. We matched the resulting NURBS/polygon mesh computational phantoms representing the body contour and the internal anatomy to anthropometric data and the reference organ mass data provided by the CDC and the ICRP, respectively. Finally, we completed a total of eight hybrid male and female phantoms, where we matched a total of eight anthropometric data categories to standard values within 4%, and organ masses matched to ICRP data within 1%, with the exception of the skin. We voxelized the hybrid phantoms from the NURBS phantoms at resolutions of skin thickness for Monte Carlo calculation.

8.3 Results and Discussion 8.3.1 Voxelization Algorithm To convert a NURBS-based hybrid newborn phantom into a voxel phantom, which is actually utilized for Monte Carlo dosimetry calculation, researchers developed a voxelization process. Figure 8.1 shows the voxelization process of newborn left lung as an example.

(a) Original voxel

(b) Polygon mesh

(c) NURBS

(d) Voxel (2 × 2 × 2 mm3)

(e) Voxel (1 × 1 × 1 mm3)

FIGURE 8.1 Example of voxelization process starting from (a) original voxel model of UF voxel newborn phantom through, (b) polygon mesh model, and (c) NURBS model to (d) voxelized model with different VR (1 × 1 × 1 mm3 and 2 × 2 × 2 mm3). (From Lee, C. et al., Phys. Med. Biol., 52, 3309, 2007. With permission.)

The University of Florida Pediatric Phantom Series

207

The original voxel lung of the UF voxel newborn phantom was converted to a polygon mesh lung computational phantom using 3D-DOCTOR (Figure 8.1a and b). We imported a polygon mesh lung computational phantom into Rhinoceros and generated a smooth NURBS-based lung computational phantom (Figure 8.1c). We triangulated the NURBSbased lung using a MT of 10°, and then voxelized it using two VRs: 2 × 2 × 2 mm3 and 1 × 1 × 1 mm3 (Figure 8.1d and e). Even though the resolutions of the resulting two voxel lung computational phantoms were lower than original VR of UF newborn voxel phantom, 0.0562 × 0.0562 × 0.0959 cm3, the smooth surfaces of real human anatomy were better described in hybrid lung computational phantoms. To study the effect of meshing the tolerance and the VR on the resulting organ volume, we chose to voxelize the largest and smallest NURBS organs, the liver and the pituitary gland, using difference parameters. First, we triangulated the NURBS organs using different MTs, 50°, 40°, 30°, 20°, and 10°. A comparison of the liver and the pituitary gland volumes for different MTs is tabulated in Table 8.1. The percent differences between the original NURBS volume and the triangulated polygon volume are presented in this table as well. As the MT decreases from 50° to 10°, the polygon organ volumes become closer to the NURBS volumes, and their corresponding percent differences decrease correspondingly. As for the liver, about 21 times more polygons were needed to reduce the volume difference from 1.904 to 0.105. The liver converged on its NURBS volume faster than the smaller pituitary gland did. Differences between the NURBS and the polygon computational phantoms were thus reduced to less than 1% by using an MT of 10°. We voxelized the polygon liver and pituitary gland that had been triangulated using MT of 10° by using different VRs. A comparison of organ volumes for different VRs is summarized in Table 8.2 with percent difference between polygon and voxel organs. As the VR increased, voxel organ volume converged to polygon organ volumes. We selected VRs to give a comparable difference for two different-sized organs. For both organs, we observed fluctuations in the volume differences for differences less than 1%. This feature may be caused by relatively low VRs, which reflect small amounts of voxel change below 1%.

TABLE 8.1 Effect of MT on Polygon Count, Organ Volume, Percent Differences in Organ Volumes for Both the Liver and Pituitary Gland Organs Liver

Pituitary gland

MT (°)

Polygon Count

Polygon Organ Volume (cm3)

Difference (%)a

50

3048

121.6750

−1.997

40

5098

122.8179

−1.076

30

8796

123.2308

−0.744

20

18816

123.7330

−0.339

10

64892

124.0242

−0.105

50

48

0.0744

−23.216

40

224

0.0908

−6.267

30

224

0.0908

−6.267

20

960

0.0954

−1.596

10

3968

0.0965

−0.401

Source: Lee, C. et al., Phys. Med. Biol., 52, 3309, 2007. With permission. Note: Volumes of the NURBS lungs and pituitary gland are 124.6643 and 0.0969 cm3, respectively. a Difference (%) = ( (Polygon volume – NURBS volume)/NURBS volume) × 100.

Handbook of Anatomical Models for Radiation Dosimetry

208

TABLE 8.2 Effect of VR on Organ Volume, Voxel Count, and Volume Differences for the Liver and Pituitary Gland That Were Polygonized from NURBS Models Using an MT of 10° Organs Liver

Pituitary gland

VR (cm)a

Voxel Organ Volume (cm3)

Voxel Count

Difference (%)b

1.5

131.6250

39

1

125.0000

125

5.694 0.374

0.5

122.7500

982

−1.432

0.2

124.0960

15512

−0.352

0.1

124.0250

124025

−0.409

0.1

0.1010

101

4.663

0.0663

0.0968

332

0.265

0.05

0.0973

778

0.777

0.02

0.0965

12061

−0.012

0.01

0.0965

96460

−0.041

0.005

0.0965

772024

0.003

Source: Lee, C. et al., Phys. Med. Biol., 52, 3309, 2007. With permission. Note: Volumes of the polygon liver and pituitary gland are 124.5338 and 0.0965 cm3, respectively. a X, Y, and Z sides of the all voxels are of the magnitude indicated in the column. b Difference (%) = ( (Voxelized volume – Polygon volume)/Polygon volume) × 100.

8.3.1.1 UF Hybrid Phantoms for Newborn Male and Female Researchers developed UF Hybrid (UFH) newborn phantoms by applying polygonization, NURBS modeling techniques, and voxelization processes to previous UF voxel newborn phantom. Researchers have developed hybrid phantoms with newborn male anatomy from female phantoms by replacing female-specific organs (ovaries and uterus) with male-specific organs (testes and prostate), which were obtained from UF 9-month voxel phantoms. Figure 8.2 shows the front views of the 3D rendering of the UF voxel newborn phantom, and the NURBS- and voxel-versions of UFH female newborn phantoms. We named the NURBS-version of hybrid phantom UFH-NURBS, and the voxel-version UF-voxel. We made the skin and the muscle transparent to better view the internal organs and the skeleton. As shown in Figure 8.2a, it is relatively difficult to distinguish organs in the abdominal region of the UF voxel newborn phantom, due to discontinuities in the z-direction, with the exception of the skeleton. NURBS surfaces in the UFH-NURBS newborn phantom have better continuity and more smoothness of the abdominal anatomy, especially in the large and small intestine, as shown in Figure 8.2b. We converted this detail to the UFH-voxel newborn phantom, even though its VR (3.43 × 10−4 cm3) is lower than that of the original UF voxel newborn phantom (3.03 × 10−4 cm3). The detail would be enhanced as the VR increases. The difference of anatomy between the male and female phantoms is depicted in Figure 8.3, where the left lateral 3D views of the lower abdominal regions with skin and left-hand bones are removed. We replaced the female-specific organs, the ovaries, and the uterus by the male-specific organs, testes, and prostate. We can observe the differences in the shape and position of the urinary bladder between the male and female anatomy. The organs and tissues in the UFH-NURBS male and female newborn phantoms are listed in Table 8.3, along with the mass and reference density obtained from the ICRU Report 46.11 We added comments on surrogate density for some organs, for which newborn

The University of Florida Pediatric Phantom Series

(a) UF voxel newborn phantom

209

(b)UFH-NURBS newborn phantom

(c)UFH-voxel newborn phantom

FIGURE 8.2 Front views of 3D rendering of (a) UF voxel newborn phantom, (b) UFH-NURBS female newborn phantom, and (c) UFH-voxel female newborn phantom. VR of (a) UF voxel and (c) UFH-voxel newborn phantoms were 0.0562 × 0.0562 × 0.0959 cm3 and 0.07 × 0.07 × 0.07 cm3, respectively. Body contours were made transparent for better viewing internal organs and skeleton. (From Lee, C. et al., Phys. Med. Biol., 52, 3309, 2007. With permission.)

Colon Small intestine Left ovary Urinary bladder Uterus

(a) UFH-NURBS female phantom

Colon Small intestine Urinary bladder Prostate Scrotum Left teticle Penis

(b) UFH-NURBS male phantom

FIGURE 8.3 Left lateral views of 3D rendering of (a) female and (b) male UFH newborn phantoms showing difference of gender-specific anatomy. Skin and left hand bones including cartilage were made transparent to better view gender-specific organs.

Alimentary system Tongue Salivary glands Tonsils Esophagus—wall Stomach—wall Stomach—contents Small Intestine—wall Small Intestine—contents Colon Right—wall Right—contents Left—wall Left—contents

Respiratory system ET1 (anterior nasal layer) ET2 (posterior nasal layer) ET2 (oral cavity layer) ET2 (larynx) ET2 (pharynx) Trachee Bronchi—extrapulmonary Lungs (inclusive of blood) Left lung Right lung Total

Organ System

Muscle (newborn) Ave soft tissue (male) Ave soft tissue (male) Gastrointestine Gastrointestine Ave soft tissue (male) Gastrointestine Ave soft tissue (male) Gastrointestine Ave soft tissue (male) Gastrointestine Ave soft tissue (male)

1.03 1.03 1.03 1.03

Calculated Calculated Calculated

0.62 0.62 0.62

1.05 1.03 1.03 1.03 1.03 1.03 1.03 1.03

Ave soft tissue (male) Ave soft tissue (male) Ave soft tissue (male) 50:50 soft tissue/cartilage Ave soft tissue (male) 50:50 soft tissue/cartilage 50:50 soft tissue/cartilage

(ICRU 46)

Comment

1.03 1.03 1.03 1.07 1.03 1.07 1.07

(g/cm )

3

Density

7.46 5.21 7.35 7.20

Not defined 5.99 Not defined 2.59 7.00 6.14 33.97 22.73

32.52

1.30 0.25 0.60 Not defined

Not defined

Mass (g)

7 −78 5 −40

30 0 −85 13 −59

−0.2

−46

20

0

% Diff

UF Voxel Female

28.86 31.16

7.01 17.93 7.01 17.93

3.52 5.99 0.10 1.99 6.99 25.44 30.12 54.92

60.02

0.09 0.85 0.74 1.30 0.31 0.50 0.39

Mass (g)

0.2 −25 0.2 49.4

0.7 −0.1 0.8 −0.6 −0.1 −36 0.4 −2

3.4 −2.9 0.0

−0.5

−0.3

% Diff

UFH-NURBS

28.75 30.92

7.01 12.01 7.01 12.01

3.50 5.98 0.10 2.00 6.97 25.34 29.86 29.97

59.68

0.08 0.72 0.72 1.31 0.31 0.50 0.38

Mass (g)

0.2 −50 0.2 0.0

−0.1 −0.4 −0.3 0.1 −0.4 −37 −0.5 −46

3.0 −3.7 −0.8

−0.2

0.9

% Diff

UFH-Voxel

28 32

7 24 7 12

3.5 6 0.1 2 7 40 30 56

60

1.3 Not defined 0.5 Not defined

Not defined Not defined

Mass (g)

ICRP 89

Summary of Organ Masses within Three Computational Phantoms of the Newborn: (1) UF Voxel Female, (2) UF Hybrid–NURBS, and (3) UF Hybrid–Voxel. The Latter Two Phantoms Include Separate Models of the Male and Female Newborn. These Masses Are Then Compared to ICRP Publication 89 Reference Masses by Organ, Organ System, and for Total Body Tissues and Total Body Masses for Both Sexes

TABLE 8.3

210 Handbook of Anatomical Models for Radiation Dosimetry

Cortical bone (infant) Red marrow (adult) Dentine Ave soft tissue (male) Effective ave density

1.65 1.03 1.30 1.03 1.47

1.1

Integumentary system Sking Skin (newborn)

Cartilage (adult)

1.10

Kidney (40 week fetus) Bladder (adult-empty) Urine of ave density Muscle (newborn) Ave soft tissue (male) Testes (adult) Ave soft tissue Ovaries (adult) Ovaries (adult)

Heart (40 week fetus) Blood (newborn)

Gastrointestine Ave soft tissue (male) Liver (40 week fetus) Ave soft tissue (male) Ave soft tissue (male) Ave soft tissue (male)

Skeletal system Bone associated cartilaged Bone tissues Bone (CB, TB) Active marrowe Teeth Miscellaneousf Total

1.03 1.04 1.01 1.05 1.03 1.04 1.03 1.05 1.05

1.04 1.07

Circulatory system Heart—wall Heart—content Bloodb

Urogenital system Kidneys (all regions) Urinary Bladder—wall Urinary Bladder—contentsc Penis Scrotum Testes (2) Prostate gland Ovaries (2) Uterus

1.03 1.03 1.04 1.03 1.03 1.03

Rectosigmoid—wall Rectosigmoid—contents Liver Gall Bladder—wall Gall Bladder—contentsa Pancreas

102.10

280.71

−42

17

−3 −12

0.29 3.52

Not defined

−14 0

21.58 4.00 6.48

−54

−78

1.33

21.13

−8 −75 −16 −34

2.77 2.99 109.13 2.19

Not defined

238.6

127.8

24.97 4.01 10.08 0.82 1.65 0.85 0.80 0.30 4.00

20.01 6.05

2.99 3.40 129.12 0.50 2.81 6.00

−0.6

0.4

0.3 −0.1 0.0 −0.1

−0.1 0.3 −16.0

0.1 −77

0.2 −72 −0.7 0.4 0.4 0.0

163.65

237.8

126.4

24.92 3.97 10.05 4.90 4.90 0.85 0.80 0.30 3.98

19.97 6.04

3.40 3.40 128.86 0.50 2.81 5.99

−6.5

−0.9

−0.7

−0.4 0.2 0.0 −0.5

−0.3 −0.7 −16.3

−0.2 −77

0.2 −72 0.2 0.3 0.3 −0.2

170.0 50.0 0.7 19.3

(continued)

175

240.0

127.32

25 4 12 Not defined Not defined 0.85 0.8 0.3 4

20 26 290

3 12 130 0.5 2.8 6

The University of Florida Pediatric Phantom Series 211

Totals by organ system Respiratory system Alimentary system—tissues of organ walls Alimentary system—Gl tract and gall bladder content

Additional tissues Adrenal glands (2) Brain Breasts (2) Eyeballs (2) Lens (2) Pituitary gland Spinal cord Spleen Thymus Thyroid Rest of body (ROB) Separable fat Skeletal muscle Separable connective tissues Fixed lymphatic tissuesh i Blood (large vessels) External nose Cartilage—ears j Miscellaneous ROB Total

Organ System Ave soft tissue (male) Brain (newborn) Adipose (newborn #2) Ave soft tissue (male) Eye lens (adult) Ave soft tissue (male) Brain (newborn) Spleen (40 week fetus) ICRP 89—Para 606 Thyroid (adult) Adipose (newborn #2) Muscle (newborn) Ave soft tissue (male) Ave soft tissue (male) Blood (newborn) 66:33 soft tiss/cartilage Cartilage (adult) Ave soft tissue (male) Effective ave density

0.99 1.05 1.03 1.03 1.07 1.05 1.10 1.03 1.02

(ICRU 46)

Comment

1.03 1.03 0.99 1.03 1.07 1.03 1.03 1.04 1.07 1.05

(g/cm )

3

Density

122.44

44.27

Not defined

0.45 1.43

6.00 322.39 0.09 6.01 0.13 0.10 6.76 9.51 13.00 1.30

Mass (g)

64.20 201.36

23

−20 −23 −57

−51

−50 −23

% Diff

0.1 0.0 −0.1

0.0 −15 0.1 0.2 0.6 −0.2

% Diff

UFH-NURBS

34.67 179.78

2509.17

3.01 291.38 Not defined 2.93 Not defined Not defined 15.13 7.64 10.00 0.56

Mass (g)

UF Voxel Female % Diff

85.54

63.69 201.19

2173.14

6.6

5.98 −0.4 321.33 −15.4 0.09 0.7 5.99 −0.2 0.13 −0.2 0.10 −0.3 6.44 9.47 −0.3 12.93 −0.5 1.29 −0.8

Mass (g)

UFH-Voxel

146.8

61.8 202.1

890 800 120 35 75 0.45 1.43 116.7 2038.8

6 380 0.09 6 0.13 0.1 Not defined 9.5 13 1.3

Mass (g)

ICRP 89

Summary of Organ Masses within Three Computational Phantoms of the Newborn: (1) UF Voxel Female, (2) UF Hybrid–NURBS, and (3) UF Hybrid–Voxel. The Latter Two Phantoms Include Separate Models of the Male and Female Newborn. These Masses Are Then Compared to ICRP Publication 89 Reference Masses by Organ, Organ System, and for Total Body Tissues and Total Body Masses for Both Sexes

TABLE 8.3 (continued)

212 Handbook of Anatomical Models for Radiation Dosimetry

5.10

3.33 127.84 238.59 Not defined 365.29 Not defined

3.81

0.00 0.00 280.71 102.10

2509.17

3538

3488 1.1

10.08

6.48

4.4

28.98

25.58

330.7

26.06

21.13

3389 3400 3485 3496

2173.14

363.75

237.77 163.65

126.40

10.65

5.08

10.05

28.89

25.92

1.4 1.7 −0.4 −0.1

3341 3342 3500 3501

2038.8

416.1

240.0 175.0

127.3

0.9

5.10

12.0

29.0

46.0

Source: Lee, C. et al., Phys. Med. Biol., 52, 3309, 2007. With permission. Note: UF hybrid—NURBS and UF hybrid—voxel include separate models of the male and female newborn. These masses are then compared to ICRP Publication 89 reference masses by organ, organ system, and for total body tissues and total body masses for both sexes. a In the UF Voxel Phantom, the wall and content of the heart and gall bladder are not separated—% difference given for the combined mass. b In this table, total blood volume is partitioned into three regions: (1) heart content, (2) individual organs, and (3) rest of body (major vessels). c No reference value is given in ICRP 89 and thus an approximate value of 12.3 g is used as defined in the ORNL stylized newborn phantom. d Skeletal cartilage excludes the following nonbone associated regions of cartilage: external nose and ears, larynx, trachea, and extrapulmonary bronchi. e Assumed to include the 7% of total blood volume (20.3 of 290 g) in the newborn as per Section 7.7.2 of ICRP 89. f As per Section 9.2.15 of ICRP 89, miscellaneous skeletal tissues include periosteum and blood vessels, but exclude periarticular tissue and blood. g Skin masses given here are for the female phantom, and are 0.15% higher in the male phantom due to the addition of the penls and scrotum. h Estimated from the reference adult values given in Section 7.8.2 of ICRP Publication 89 and scaled by newborn to adult total body mass. i Taken as 25.92% of total blood pool as per Section 7.7.2 of ICRP 89 (other tissues, arota, large arteries, large veins). j Miscellaneous rest-of-body is added to force the total body mass to its ICRP 89 reference value of 3500 g. k Male phantom masses additionally include soft tissuea occupied by the uterus and ovaries in the corresponding female phantom.

Total body tissues (F) Total body tissues (M) Total body mass (F) Total body mass (M)

Additional tissues—excluding rest of body Additional tissues—rest of body

Circulatory system—heart wall and content Urogenital system—kidneys and urinary bladder wall Urogenital system—urinary bladder content Urogenital system—Internal sex organs (ovaries, uterus, prostate)k Urogenital system—external sex organs (penis, scrotum, testes) Skeletal system—bone associated cartilage Skeletal system—bone tissues Integumentary system

The University of Florida Pediatric Phantom Series 213

214

Handbook of Anatomical Models for Radiation Dosimetry

reference data are not available. For example, the heart wall density for the newborn was not available, and the density of the adult heart was used in the surrogate. Also, we estimated the effective homogeneous lung density as the ratio of the reference lung masses given in the ICRP 89 to the lung volumes of the UFH phantoms. We separated the left and right organs for the adrenals, kidneys, and lungs. We also separated the wall and the content of some walled organs: the colon, the gall bladder, the heart, the rectosigmoid, the small intestine, the stomach, and the urinary bladder. The external nose, the pituitary gland, the tongue, and the tonsil were newly modeled in the UF hybrid phantoms. 8.3.2 Comparison with UF Voxel Newborn Phantom and ICRP89 The organ masses of the UFH-NURBS and UFH-voxel phantoms are compared to those of the original UF voxel newborn phantom and ICRP 89 as listed in Table 8.3. Even though we matched the total body mass of the UF voxel newborn phantom to the ICRP reference mass within 1%, other organs showed a significant mass difference from the reference mass. The Gall bladder wall was up to 194% heavier than reference. The percent differences of organ masses between the UFH-NURBS phantom and the ICRP 89 values were also evaluated in Table 8.3. All are within 1% for internal organs, except for the brain and the eyeballs. As mentioned previously, the expansion of the brain and the eyes was not attempted, as the skeleton—including the cranium—was already well matched to the ICRP 89 reference values. Skin mass was not available at the stage of NURBS modeling, since skin was generated only after the phantom voxelization. The UFH-NURBS newborn phantom was triangulated and voxelized using an MT of 10° and a VR of 0.07 × 0.07 × 0.07 cm3. According to the results of the sensitivity study, as shown in Tables 8.1 and 8.2, we expected these parameters to reduce the percent differences between the NURBS and the voxel computational phantoms below 1% for all organs. Table 8.3 evaluates and lists the percent difference of organ masses between the UFH-voxel phantom and the ICRP 89. The mass of the small intestine in the UF voxel phantom was 2.4% smaller than its reference value, while the NURBS small intestine was matched to within 1%. This difference was attributed to the self-overlap of small intestine, especially at the curved regions. Assigning a skin tag to the outermost voxel layer of body contour generated the skin. Since the VR was set to the skin thickness of ORNL newborn phantom, the skin thickness of the UFH-voxel phantom was the same as with the ORNL phantom. The resulting skin mass of the UFH-voxel phantom was 166.396 g, which was 4.92% smaller than its reference value. The number of voxels in the UFH-voxel phantom was 3.301 × 107, which was 32% smaller than in the original UF voxel newborn phantom. Hybrid phantoms thus make it possible to more effectively represent human anatomy using optimized computational resources. 8.3.3 Advantages of Hybrid Newborn Phantom Hybrid newborn phantoms developed in this study have several advantages over existing stylized and voxel phantoms. First, hybrid phantoms have both the flexibility of stylized phantoms, in changing the computational phantoms of individual organs, the outer body contour, and the extremities, while still preserving the realistic anatomy of voxel phantoms. Organs for which anatomical detail cannot be fully extracted from the original CT images due to the limited image contrast can be described by stylistic computational phantoms incorporated within the hybrid NURBS computational phantom. This advantage was highlighted in the case of the alimentary system. To show this advantage of hybrid phantoms, Figure 8.4 depicts a comparison between the alimentary systems of the stylized,

The University of Florida Pediatric Phantom Series

Esophagus

215

Esophagus

Esophagus

Stomach Stomach

Stomach Colon

Colon

Small intestine Small intestine Colon Rectosigmoid

(a) ORNL newborn

Small intestine Rectosigmoid

(b) UF voxel newborn

Rectosigmoid

(c) UFH voxel newborn

FIGURE 8.4 Alimentary systems of (a) ORNL stylized, (b) UF voxel, and (c) UFH-voxel newborn phantoms. Esophagus, stomach, small intestine, colon, and recto-sigmoid colon are labeled.

the voxel, and the UFH-NURBS newborn phantoms. Those of the stylized phantom and voxel phantom were obtained from the latest revised ORNL newborn phantom8 and UF newborn voxel phantom, respectively. The alimentary tract in the revised ORNL stylized newborn phantom is significantly simplified and unrealistic compared with the more realistic anatomy of both the voxel and hybrid phantoms. Even though the overall shapes of the large and small intestines in the UF voxel newborn phantom are more realistic than those of stylized phantom, they have partial volume defects and image slice discontinuities due to the limited resolution of the CT images. We voxelized the alimentary tracts of the UFH-voxel newborn phantom from the NURBS surfaces based on the original anatomical shapes in the UF voxel newborn phantom and the ICRP Publication 89 reference data. Other advantages of hybrid phantoms is (1) their flexibility to modify the outer body contour to represent nonreference individuals, (2) their ability to modify organ shape, position, and depth, and (3) their ability to generate voxel phantoms—whole-body or individual organ systems—at resolutions that are user-defined and thus optimized to the dosimetry problem at hand. Manipulations of control points surrounding the NURBS surfaces of body contours and internal organs allow the shapes to be modified to generate different sized organ or body shapes. Due to this feature, we can accurately match the organ masses of the UF hybrid newborn phantom to the reference data. It is also possible to add or remove certain organs as needed, which makes it possible to create a male phantom from the original female newborn phantom while maintaining the framework of all nonsex-organ anatomy. In the case of existing voxel phantoms, even though the x–y in-plane VR can be as high as the resolution of the original CT images, the resolution in the z-direction (slice thickness) cannot be increased, thus restricting it to the original scanning interval.

Handbook of Anatomical Models for Radiation Dosimetry

216

8.3.4 The Extended Development of Older Phantoms We developed the older members of the UF hybrid phantom series using the same technology as we incorporated into the newborn hybrid phantom. These additional phantoms are those of the ICRP 89 reference 1, 5, 10, and 15 year old male and female. We used the same CT-based anatomy for the development of the male and female at ages 1, 5, and 10 year, with the exception of the reproductive organs (i.e., the gonads, the uterus, and the prostate). The 15 year old male and female phantoms were developed independently from different CT data as described in Lee et al.12 A total of eight body dimensions were matched to the standard anthropometric data listed in the Table 8.4 within 4% for all phantoms. Figure 8.5 presents the 3D frontal view of the series of UF hybrid pediatric male and female phantom series. TABLE 8.4 Anthropometric Data Obtained from NHANES III/IV, Anthrokids, and ICRP89 0 M

1 F

M

5

15

10

F

M

F

M

F

M

F

109

109

138

138

167

161

Height Standing

51

51

76

76

Sitting

34

34

49.5

48.1

60.6

60.2

73.1

73.6

88.8

85.5

32.6

32.6

47.2

47

62

59.9

75

70.7

47.9

46.7

Length Arm Circumference 51.5

50.7

52.9

52.7

55.4

54.3

Neck

25.3

24.5

28.1

27.6

32.8

30.8

Waist

55

54.9

67.7

65.7

80.1

78.8

Buttock

57.3

58.4

74.5

75.8

92.5

93.4

25

24.9

31.2

31.1

38.8

36.3

Head

33.1

33.1

Breadth Biacromial

1-year male

1-year female

5-year male

5-year female

10-year male

10-year female

15-year male

15-year female

FIGURE 8.5 (See color insert following page 524.) Series of UF hybrid pediatric male (left) and female (right) phantoms: newborn, 1, 5, 10, and 15 year old phantoms.

The University of Florida Pediatric Phantom Series

217

FIGURE 8.6 Representative microCT images of age-dependent skeletal microstructure in the pediatric spine. Upper panel includes images of the cervical, thoracic, and lumbar vertebral bodies (left to right) of the newborn skeleton, while the lower panel gives corresponding images of 18 year old spine.

8.3.5 Pediatric MicroCT-Based Skeletal Computational Phantoms In their current configurations, the bones of the skeleton in the UF pediatric series of Figure 8.5 are homogeneous in tissue composition and skeletal microstructure, yet are of a reference total volume to contain all ICRP 89 reference tissue masses. Current efforts are devoted to endowing each phantom with a unique and age-dependent microCT-base skeletal computational phantom for not only internal radiation sources (alpha and beta particles as needed for nuclear medicine dose assessment), but also for the construction of fluence-to-dose response functions for assessing active marrow and endosteal tissue doses from both internal and external photon sources. Details of the dose–response function are given in Eckerman et al.13 Representative images of both newborn and midteen skeletal microstructures are shown in Figure 8.6 as needed for paired-image radiation transport simulation.14,15 Implementation of these skeletal computational phantoms requires subsegmentation of each pediatric bone in the UF hybrid series into regions of spongiosa, cortical bone, and medullary marrow using methods similar to that described by Zankl et al.16 and by Kramer et al.17 for their adult phantoms.

8.4 Conclusions To evaluate radiation dose distribution within human anatomy, researchers have developed and utilized simplified mathematical equation-based stylized phantoms for several dosimetry calculation purposes since the 1960s. Voxel phantoms have been subsequently

218

Handbook of Anatomical Models for Radiation Dosimetry

developed to overcome the anatomical limitations of stylized phantoms since the mid1980s. Even though the anatomical realism of computational phantoms has been improved significantly through the use of voxel phantoms, they are more difficult to match of individual patient body morphometry. To merge the advantages of the stylized and voxel phantoms, we have developed hybrid male and female newborn phantoms. We adopted NURBS, an advanced mathematical modeling tool, to replace the limited mathematical surface equations of stylized phantoms, and we utilized the precedent UF voxel newborn phantom as its realistic anatomical framework. The development of hybrid phantom was performed in three steps: polygonization of the UF voxel newborn phantom, NURBS surface modeling, and voxelization of NURBS computational phantom. We used two 3D graphic tools, 3D-DOCTOR and Rhinoceros, to triangulate voxel phantom and generate NURBS surfaces, and an in-house MATLAB code was written to voxelize resulting NURBS computational phantom into fi nal voxel phantom that is ready for Monte Carlo radiation transport calculation. The hybrid male newborn phantom was constructed from the female phantom by replacing female-specific organs with male-specific organs. We adjusted the resulting NURBS computational phantoms represent the body contour and the internal anatomy to match the anthropometric and the reference newborn data reported by ICRP and other sources. Finally, the UFH-voxel newborn male and female phantoms were constructed by voxelizing the UFH-NURBS phantoms used a targeted VR of 0.07 × 0.07 × 0.07 cm3. Researchers can make the NURBS-based hybrid phantoms realistic based on medical images, while maintaining the flexibility to model smooth organ surfaces and even organ motion. Body contours can be modified by manipulating control points to accommodate different body shape among individuals. The hybrid technologies established in this study, including polygonization of existing voxel phantom, NURBS surface modeling technique, and conversion of NURBS computational phantom into voxel phantom, can be applied to the development of a wide range of human phantoms in the future. This new class of anthropomorphic computational phantoms can be widely applied to dose evaluation in radiation protection, medical imaging, and radiation therapy, where realistic and flexible dynamic computational human phantoms are required.

Acknowledgment This work was performed under grant RO1 CA116743 from the National Cancer Institute (subcontract from Rensselaer Polytechnic Institute) with the University of Florida.

References 1. Nipper, J.C., Williams, J.L., and Bolch, W.E. Creation of two tomographic voxel models of paediatric patients in the first year of life, Phys Med Biol, 47, 3143, 2002. 2. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, Publication 89, International Commission on Radiological Protection, Oxford: Pergamon Press, 2002.

The University of Florida Pediatric Phantom Series

219

3. Yang, L. et al. A simulation for effects of RF electromagnetic radiation from a mobile handset on eyes model using the finite-difference time-domain method, Conf Proc IEEE Eng Med Biol Soc, 2007, 5294, 2007. 4. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381 V1-V7, Oak Ridge, TN: Oak Ridge National Laboratory, 1987. 5. Lee, C. et al. Hybrid computational phantoms of the male and female newborn patient: NURBSbased whole-body models, Phys Med Biol, 52, 3309, 2007. 6. Zankl, M. et al. Organ dose conversion coefficients for external photon irradiation of male and female voxel models, Phys Med Biol, 47, 2367, 2002. 7. ICRP. Human Respiratory Tract Model for Radiological Protection: A Report of a Task Group of the International Commission on Radiological Protection, ICRP Publication 66, Oxford: Pergamon Press, 1994. 8. Han, E., Bolch, W., and Eckerman, K. Revisions to the ORNL series of adult and pediatric computational phantoms for use with the MIRD schema, Health Phys, 90, 337, 2006. 9. Lee, C., Lee, C., and Bolch, W.E. Age-dependent organ and effective dose coefficients for external photons: A comparison of stylized and voxel-based paediatric phantoms, Phys Med Biol, 51, 4663, 2006. 10. Lee, C. et al. Whole-body voxel phantoms of paediatric patients—UF Series B, Phys Med Biol, 51, 4649, 2006. 11. ICRU. Tissue substitutes in radiation dosimetry and measurement, ICRU Report 44, Bethesda, MD, 1989. 12. Lee, C. et al. Hybrid computational phantoms of the 15-year male and female adolescent: Applications to CT organ dosimetry for patients of variable morphometry, Med Phys, 35, 2366, 2008. 13. Eckerman, K.F. et al. Response functions for computing absorbed dose to skeletal tissues from photon radiation, Radiat Prot Dosim, 127, 187, 2008. 14. Shah, A. et al. Accounting for beta-particle energy loss to cortical bone via Paired-Image Radiation Transport (PIRT), Med Phys, 32, 1354, 2005. 15. Shah, A.P. et al. A paired-image radiation transport model for skeletal dosimetry, J Nucl Med, 46, 344, 2005. 16. Zankl, M., Eckerman, K.F., and Bolch, W.E. Adult male and female voxel-based models representing the ICRP reference adult—The skeleton, Radiat Prot Dosim, 127, 174, 2007. 17. Kramer, R. et al. Skeletal dosimetry for external exposure to photons based on microCT images of spongiosa from different bone sites, Phys Med Biol, 52, 6697, 2007. 18 Pafundi, D. et al. Image-based pediatric skeletal dosimetry for the UF hybrid computational phantom series, 2008. Annual Meeting of the European Association of Nuclear Medicine, Munich Germany, October 11–15, 2008 (Supplement to Eur J Nucl Med 35(2), S135, 2008).

9 Japanese Computational Phantoms: Otoko, Onago, JM, JM2, JF, TARO, HANAKO, Pregnant Woman, and Deformable Child Kimiaki Saito, Kaoru Sato, Sakae Kinase, and Tomoaki Nagaoka

CONTENTS 9.1 Introduction ............................................................................................................... 221 9.2 Construction of Computational Phantoms ...........................................................222 9.2.1 JAEA Computational Phantoms ................................................................222 9.2.2 NICT Computational Phantoms ................................................................ 226 9.2.2.1 Adult Male and Female Computational Phantoms................. 229 9.2.2.2 Pregnant Women Computational Phantoms ........................... 230 9.2.2.3 Deformed Child Computational Phantoms ............................. 231 9.2.2.4 Computational Phantoms of Arbitrary Posture ...................... 232 9.2.2.5 Arbitrary High Resolution Human Voxel Computational Phantoms ....................................................................................... 233 9.2.2.6 Applications in Electromagnetic Dosimetry ............................ 233 9.3 Dose Calculation ....................................................................................................... 233 9.3.1 Monte Carlo Code Systems......................................................................... 233 9.3.2 External Dose Calculation .......................................................................... 236 9.3.2.1 Photon ............................................................................................ 236 9.3.2.2 Electron .......................................................................................... 239 9.3.3 Internal Dose Calculations ......................................................................... 240 9.3.3.1 Absorption Fractions ................................................................... 240 9.3.3.2 S Value............................................................................................ 242 9.3.4 Whole-Body Counting ................................................................................ 245 9.4 Application to Radiotherapy ................................................................................... 247 9.5 Conclusion (Future Plan) ......................................................................................... 250 References ............................................................................................................................. 251

9.1 Introduction The majority of human computational phantoms for radiation protection purposes, including stylized computational phantoms, have been constructed for Caucasian body types. For example, both the conventional MIRD computational phantoms (the Medical Internal 221

222

Handbook of Anatomical Models for Radiation Dosimetry

Radiation Dose Committee of the Society of Nuclear Medicine Pamphlet No. 5 type computational phantoms)1 and the newly released voxel computational phantoms by ICRP2 are both based on Caucasian anatomical data. Further, basic data for radiation protection, like dose conversion factors, have been prepared using Caucasian computational phantoms. Therefore, Asian researchers have desired a clarification of the difference in doses due to the anatomical differences between Caucasian and Asians. Voxel computational phantoms enable us to reasonably investigate the dose differences between these two different races, since they can simulate anatomical structures realistically. From this viewpoint, the Japan Atomic Energy Agency (JAEA) constructed the first Asian voxel computational phantom called Otoko in 2001,3 and several Japanese computational phantoms have been constructed since then, mainly for radiation protection purposes. Additionally, the National Institute of Information and Communication Technology (NICT) has developed Japanese voxel computational phantoms for the evaluation of the exposure to electromagnetic fields, which has become a concern in recent years. At NICT, several advanced computational phantoms, like a variable posture computational phantom and a pregnant computational phantom, have been developed. In this chapter, we introduce these Japanese computational phantoms, and explain their applications to ionizing and nonionizing radiation dosimetry in diverse conditions.

9.2 Construction of Computational Phantoms 9.2.1 JAEA Computational Phantoms At the JAEA, five Japanese voxel computational phantoms have been completed from computed tomography (CT) data: three male adult computational phantoms and two female adult computational phantoms. All CT data were taken for healthy volunteers at the Fijita Health University Hospital after receiving the approval by the Ethics Committee. Pictures of four of the developed voxel computational phantoms are shown in Figure 9.1, and the physical characteristics are listed in Table 9.1. Researchers segmented the first Asian computational phantom, Otoko,3 in collaboration with GSF in Germany with the commercially available image-processing equipment MIPRON (Kontron Elektonik, Eching, Germany), and the other computational phantoms by applying the GSF techniques with the commercial software Visilog4 (Noesis, Orsay, France) at JAEA. In these computational phantoms, compact bone and bone marrow are separately modeled in each skeletal voxel according to the CT value, enabling users to calculate doses with considering the bone marrow distribution in the body. The first generation computational phantoms Otoko3 and Onago4 developed by Saito et al. have a voxel size of 0.98 × 0.98 × 10 mm,3 while the second generation computational phantoms JM5 and JF6 developed by Sato et al. have a finer voxel size of 0.98 × 0.98 × 1 mm3. Performance of CT scanners has greatly improved in a short period, and it has become possible to take CT pictures at a high resolution with less exposure resulting in the development of the high-resolution computational phantoms. Researchers expect the former long-shaped voxel not to cause significant effects on dose accuracy, since the voxel volume is small; however, in some small organs, a voxel length of 10 mm could lead to insufficient resolution of the organ computational phantoms. Figure 9.2 shows examples of organ computational phantoms consisting of voxels at 10 mm thickness and at 1 mm. Voxels at 10 mm

Japanese Computational Phantoms

Otoko

223

Onago

JM

JF

FIGURE 9.1 (See color insert following page 524.) Japanese voxel models developed at JAEA.

TABLE 9.1 Physical Characteristics of Developed Voxel Phantoms at JAEA

Gender Weight (kg) Height (cm) Slice thickness (mm) Pixel side length (mm)

Otoko

Onago

JM

JF

Male 65 170 10 0.98

Female 57 162 10 0.98

Male 66 171 1 0.98

Female 44 152 1 0.98

Standard Asian Man Male 64 170 – –

Female 46 155 – –

thickness can properly model a stomach having a large size, while voxels at 1 mm thickness would be necessary for realistic modeling of a thyroid. Organ masses of the developed computational phantoms are shown in Tables 9.2 and 9.3. The both male computational phantoms Otoko and JM have body sizes close to the Asian Reference Man (ARM) defined by Tanaka;7 however, in general the organ masses of JM are closer to the reference values. The portion of organs whose masses coincide the reference values within 30% deviation is 85% for JM and 55% for Otoko. In Otoko, especially, small organs tend to deviate from the reference values. This is considered partly because the slice thickness of 10 mm is not small enough to precisely model small organs with high resolution. Organ masses of female computational phantoms should not be directly compared to the reference values, because Onago has larger body size than the Asian Reference Man, Female (ARMF) while JF has smaller body size. In Figure 9.3, we compare the body thicknesses and widths of the developed voxel computational phantoms with the reference values for the thorax, the abdomen, and the

224

Handbook of Anatomical Models for Radiation Dosimetry

Stomach

0.98 × 0.98 × 10 mm3

0.98 × 0.98 × 1 mm3 Thyroids

0.98 × 0.98 × 10 mm3

0.98 × 0.98 × 1 mm3

FIGURE 9.2 Examples of organ models consisting of voxels at 10 mm height and at 1 mm height.

buttocks. They are all within 2σ deviations around the reference values, and it was confirmed that the developed voxel computational phantoms do not much deviate from the average body shapes of the Japanese. In radiation protection, dose calculations are usually performed using an assumed upright position of human bodies, while voxel computational phantoms have been constructed from CT data taken in lying position. The structures of a human body and its organs are considered to slightly change according to the posture; therefore, researchers have investigated the effect of posture on organ doses. For this purpose, Sato et al. constructed the voxel computational phantom JM2 from CT data taken in an upright position9 for the same person as JM. Using a cone beam CT scanner, four sets of CT pictures of spherical regions at a diameter of 25 cm were taken after changing the height position. The CT pictures could not cover some peripheral parts of the body because of the limited diameter; the missing parts were complemented by the data of JM. This complement is considered not to affect the calculated dose significantly, since dominant organs were mostly included in the corn beam CT data.

Japanese Computational Phantoms

225

TABLE 9.2 Comparison of Organ and Tissue Masses of the Japanese Male Voxel Models Developed at JAEA and the ARM Ratio

Weight (g) Organ or Tissue

Otoko

JM

ARF

Otoko/ARM

Otoko/ARFM

Adrenals

20.9

11.9

14

1.49

0.85

Eyes

20

13.5

15

1.33

0.90

Lenses Gall bladder Esophagus Stomach Intestine

0.2

0.4

0.4

0.50

0.93

11.7

6.7

8

1.46

0.84 0.93

16.1 122.3 1,009

40

0.40

124

37.1

140

0.87

0.89

684

920

1.10

0.74

Small intestine

691

431

590

1.17

0.73

Large intestine

318

253

330

0.96

0.77

Upper large intestine

175

135

180

0.97

0.75

Lower large intestine

143

118

150

0.95

0.79

Heart

476

528

360

1.32

1.47

Kidney Liver Lung Muscle (skeletal)

320

0.83

1191

265 1,304

0.83

1,600

0.74

0.82

1,546

1,549

1,100

1.41

1.41

266

30,560

24,547

24,600

1.24

1.00

Pancreas

109

130

0.84

0.91

Skeleton

11,368

119 11,051

8,300

1.37

1.33

7,582

7,318

4,500

1.68

1.63

Hard bone Bone marrow

3,786

3,734

3,800

1.00

0.98

Skin

2,195

2,217

2,400

0.91

0.92

138

140

0.54

0.99

37

0.74

0.99

Spleen

75.7

Testes

27.4

36.8

Thymus

4.58

31.6

33

0.14

0.96

Thyroid

9.93

21.8

19

0.52

1.15

Trachea Urinary bladder

8.93 38.8

10.2

9

0.99

1.13

37.7

40

0.97

0.94

In Table 9.4, we compare the organ masses between the constructed upright computational phantom JF2 and JF. The organ masses agree reasonably, and this suggests the segmentation was properly carried out. The lateral CT pictures in Figure 9.4 indicate that the spine above the waist was bent backward for a standing position in comparison to the lying position. This bending is considered necessary to keep body balance in standing position. Further, the force direction by gravity is different between standing and lying positions resulting in different shapes and positions of flexible organs. The abdominal shape apparently differs between the two positions. We tabulated the distances between several organs for JM and JM2 in Table 9.5. Here, the gravity center of an organ was considered as the representative position. The distances relative to the brain hardly change according to the posture for the esophagus, the lungs, the

Handbook of Anatomical Models for Radiation Dosimetry

226

TABLE 9.3 Comparison of Organ and Tissue Masses of the Japanese Female Voxel Model Developed at JAEA and the ARMF Ratio

Weight (g) Organ or Tissue

Onago

JF

ARMF

Onago/ARMF

JF/ARFM

Adrenals

19.7

6.2

13

1.52

0.48

Eyes

23.3

15.6

12

1.94

1.30

Lenses

0.7

0.3

0.3

2.33

1.09

Gall bladder

4.1

3.8

6

0.68

0.63

30

Esophagus Stomach Intestine

9.3 99.4 914

0.31

1.65

106

49.5

110

0.90

0.97

790

980

0.93

0.81

Small intestine

742

552

720

1.03

0.77

Large intestine

172

238

260

0.66

0.91

Upper large intestine

106

121

140

0.76

0.87

Lower large intestine

67

117

120

0.56

0.97

Heart

476

280

320

1.49

0.88

Kidney

257

213

280

0.92

0.76

Liver

1,448

1,179

1,400

1.03

0.84

Lung

996

1,245

910

1.09

1.37

Ovaries

10.4

6.8

11

0.95

0.61

Pancreas

52.6

95.9

110

0.48

0.87

Skeleton Hard bone

10,354

7,389

6,400

1.62

1.15

7,133

4,658

3,400

2.10

1.37

Bone marrow

3,222

2,731

3,000

1.07

0.91

Skin

1,975

1,753

1,800

1.10

0.97

Spleen Thymus

90.6

55.5

120

0.76

0.46

1.6

19.4

29

0.05

0.67 0.43

Thyroid

5.8

7.3

17

0.34

Trachea

6.2

17.5

7

0.91

2.57

23.5

20.2

30

0.78

0.67

47.5

70

2.18

0.68

Urinary bladder Uterus

152

lower large intestine, and the urinary bladder; while the positions of the liver, the stomach, the gall bladder, and the kidneys in standing posture shifted to a lower direction because of gravity. The change in distance is within several percent in most cases. Positional relations among neighboring organs in the middle of body are difficult to explain with simple principles; some cases show a large distance change like the case for stomach content and pancreas. This change in positional relation affects the specific absorption fraction (SAF) as discussed later. 9.2.2 NICT Computational Phantoms This section gives an overview of the work at the NICT on voxel computational phantoms developed for electromagnetic-field dosimetry. In the mid-1990s, anatomically realistic

Japanese Computational Phantoms

227

400

300 2σ

Width (mm)

Thickness (mm)

300 200

100

× Reference JM JF Otoko Onago

0

200

× Reference JM JF Otoko Onago

100

0 T

A B Male

T

A B Female

T

A B Male

T

A B Female

FIGURE 9.3 Body thicknesses and widths of the developed voxel models at the thorax (T), abdomen (A), and buttock (B) parts. The reference values are shown together.

TABLE 9.4 Organ Masses of JM and JM2 Mass (kg) Organ or Tissue

JM

JM2

Adrenals

0.012

0.011 (0.93)

Brain

1.704

1.704 (1.00)

Heart wall

0.528

0.501 (0.95)

Kidneys

0.265

0.263 (0.99)

Liver

1.304

1.400 (1.07)

Lower large intestine wall

0.118

0.117 (0.99)

Lungs

1.549

1.603 (1.03)

Pancreas

0.119

0.117 (1.00)

Skin

2.217

2.237 (1.01)

Small intestine wall

0.431

0.425 (0.99)

Spleen

0.138

0.139 (1.01)

Stomach wall

0.124

0.122 (0.98)

Testes

0.037

0.037 (1.01)

Thymus

0.032

0.032 (1.00)

Thyroid

0.022

0.022 (0.99)

Upper large intestine wall

0.135

0.137 (1.02)

Urinary bladder wall

0.038

0.037 (0.98)

Note: Values in parentheses are the ratios of organ masses of JM2 to JM.

Handbook of Anatomical Models for Radiation Dosimetry

228

Spine

Abdomen

(a) Lying position

(b) Standing position

FIGURE 9.4 Comparison of body structures between JM in lying position and JM2 in standing position.

TABLE 9.5 Distances between Several Dominant Organs in JM2 and JM in Terms of the Gravity Center Organ Distance (mm) Organs

JM

JM2

Ratios of JM2 to JM

Brain and esophagus Brain and lungs Brain and liver Brain and stomach wall Brain and gall bladder wall Brain and kidneys Brain and lower large intestine wall Brain and urinary bladder wall Kidneys and liver Kidneys and lower large intestine wall Kidneys and pancreas Stomach content and liver Stomach content and lower large intestine wall Stomach content and pancreas

322 355 480 511 510 544 710

318 351 490 526 529 563 713

0.99 0.99 1.02 1.03 1.04 1.03 1

748 94 174

748 98 155

1 1.05 0.89

64 109 206

62 116 198

0.97 1.06 0.96

31

47

1.5

Japanese Computational Phantoms

(a)

(b)

229

(c)

(d)

(e)

(f)

FIGURE 9.5 3D images of voxel models developed at NICT: (a) adult male and (b) adult female models; (c) 26 week pregnant woman model; (d) 7 year old, (e) 5 year old, and (f) 3 year old models.

voxel computational phantoms of a cranium were developed and have been used for dosimetry of human heads exposed to electromagnetic radiation from cellular phones.10–14 These head computational phantoms, however, cannot be used for safety evaluations of wearable communication terminals worn on other parts of the body. Since the late 1990s, several research groups have been developing high-resolution whole-body voxel computational phantoms for electromagnetic-field dosimetry.15–19 These computational phantoms, however, are based on anatomical data for non-Japanese subjects and their body proportions differ substantially to those of Japanese population. Therefore, researchers at NICT have developed realistic high-resolution whole-body voxel computational phantoms of Japanese body size, as shown in Figure 9.5.20–22 9.2.2.1 Adult Male and Female Computational Phantoms Nagaoka et al. have developed voxel computational phantoms by collecting magnetic resonance imaging (MRI) data from volunteer Japanese adult males and females with average body size (Figure 9.5a and b).22 The average height and weight of Japanese adults between the ages 18 and 30 years are 171.4 cm and 63.3 kg for male and 159.1 cm and 52.6 kg for female, respectively.23 On the bases of this data, a male and a female subject for the computational phantoms were chosen whose height and weight were close to the Japanese average values. The male volunteer, who was 22 years old, was 172.8 cm tall and weighed 65.0 kg; the female volunteer, who was also 22 years old, was 160.0 cm tall and weighed 53.0 kg. Magnetic resonance images of the male and female subjects were acquired using a 1.5 Tesla (T) MRI system. Analysts used whole-body transverse (axial) images for the male and female to identify tissues and organs for each voxel. It is impossible to automatically segment the voxels with sufficient accuracy using currently available image-processing technologies. All tissue- and organ-identification processing was therefore performed manually by medical personnel using PC software. Researchers checked the three-dimensional images and the three orthogonal planes and then smoothed the boundaries of the

Handbook of Anatomical Models for Radiation Dosimetry

230

TABLE 9.6 Identified Tissues and Organs in the Male and Female Models Developed at NIST No.

Tissue and Organ

No.

Tissue and Organ

No.

Tissue and Organ

1

Adrenals

20

Gray matter

38

Seminal vesiclea

2

Air (internal)

21

Heart

39

Skin

3

Bile

22

Hypothalamus

40

Small intestine

4

Bladder

23

Kidney

41

Small intestine contents

5

Blood

24

Large intestine

42

Spleen

6

Bone marrow and cancellous bone

25

Large intestine contents

43

Stomach

7

Breast fatb

26

Lens

44

Stomach contents

8

Cartilage

27

Ligament

45

Tendon

9

Cavernous bodya

28

Liver

46

Testisa

10

Cerebellum

29

Lung

47

Thalamus

11

Cornea

30

Muscle

48

Thyroid

12

Cortical bone

31

Nerve (spinal cord)

49

Tongue

13

CSF

32

Ovaryb

50

Tooth

14

Diaphragm

33

Pancreas

51

Trachea

15

Duodenum

34

Pineal glands

52

Urine

16

Esophagus

35

Pituitary

53

Uterusb

a

54

Vaginab

55

White matter

17

Eye tissue (sclera)

36

Prostate

18

Fat

37

Salivary gland

19

Gall bladder

a b

Only male model. Only female model.

following tissues and organs: axial, sagittal, and coronal. The final positioning and shaping of the tissues and organs were performed under the supervision of medical doctors using three-dimensional visualization software (INTAGE RV; Kubota Graphics Technologies, Inc., Tokyo, Japan). The developed male and female computational phantoms were assumed to stand in upright position with their hands at both sides of their bodies. The computational phantoms are composed of voxels of 2 × 2 × 2 mm3 and are divided into 51 different tissues and organs, as shown in Table 9.6. The masses of the main tissues and organs of the developed computational phantoms were compared with the average values of the Japanese Reference Man,24 while the masses of the tissues and organs not compiled in the data of the Japanese Reference Man were compared with those of the ARM.7 The masses of 60% of the identified tissues and organs in the male computational phantom were within 30% of the average values. For the female computational phantom, the masses of 80% of the identified tissues and organs were within 30% of the average values.22,24 9.2.2.2 Pregnant Women Computational Phantoms Developing a high-resolution computational phantom on the basis of only the medical images of pregnant women is difficult because of ethical issues regarding obtaining high-resolution whole-body tomographic images of healthy pregnant women. Therefore, Nagaoka et al. have developed a whole-body pregnant woman computational phantom by

Japanese Computational Phantoms

231

Control point

Skin

Muscle

Skin

Muscle

FIGURE 9.6 Change in shape of abdomen of female model obtained using FFD algorithm. Female model before deforming abdomen (left) and after deforming abdomen (right).

combining a computational phantom of a fetus and the adult female computational phantom (Figure 9.5c).21 The researchers segmented the tissues and organs of the fetus computational phantom using image analysis software (SliceOmatic ver. 4.3; Tomovision Inc., Montreal, Canada). They obtained the images of a healthy 26 week pregnant woman from an abdominal MRI, and according to the MR images a fetus computational phantom was constructed consisting of six tissues: the fetus, the fetal brain, the fetal eyes, the amniotic fluid, the placenta, and the uterine wall. In addition, the researchers dilated the abdomen of the female computational phantom by applying the free-form deformation (FFD) algorithm25 in order to match the abdominal shape at 26 weeks of pregnancy. The advantage of this technique is that the object shape is not limited and it can maintain the continuity of deformed objects. These tissues were deformed in relation to the original MR images (abdomen) of the pregnant woman. Figure 9.6 shows the deformation of the abdomen of the female computational phantom. Finally, using the original MR images of the subject as a reference, the fetus computational phantom was combined with the abdomen of the deformed pregnant woman computational phantom. The computational phantom is composed of voxels of 2 × 2 × 2 mm3, and it is divided into 56 different tissues and organs. The fetal occiput is directed toward the mother’s left anterior side as shown in Figure 9.7. 9.2.2.3 Deformed Child Computational Phantoms Nagaoka et al. transformed the adult male computational phantom into child figure computational phantoms (Figure 9.5d through f).20 The transformation was performed because developing new computational phantoms of children based on the MRI data of healthy child subjects is difficult due to ethical issues. It is not possible to rescale an adult

232

Handbook of Anatomical Models for Radiation Dosimetry

computational phantom to match the size of a child, because the resulting computational phantom does not accurately represent the specific figure of the child. Nagaoka et al. obtained their own reference data on 3, 5, and 7 year old children because statistical data are hardly available on the figures of children under the age of 7 years. The original adult male computational phantom was transformed into figure computational phantoms of children using the FFD algorithm.25 Researchers adjusted the child computational phantoms to have the average heights, weights, and proportions of Japanese children. Like the adult male computational phantom, the developed child computational phantoms consist of 2 × 2 × 2 mm3 voxels and are divided into 51 different tissues and organs. From Figure 9.5, it can be seen that the proportions of the child computational phantoms differ from those of the adult computational phantoms. The sectional images of the child computational phantoms in Figure 9.8 clearly show that the anatomical structures are maintained.

FIGURE 9.7 Position of fetus in pregnant woman model.

9.2.2.4 Computational Phantoms of Arbitrary Posture Because most of whole-body voxel computational phantoms are based on MRI or CT data, a straight posture is assumed. This limits the simulation of actual situations when wireless telecommunication devices are used. Therefore, we have been developing computational phantoms that can be transformed into any arbitrary posture.26,27 By applying FFD,25 we transform the posture of computational phantoms. An example of a voxel computational phantom in a different posture is shown in Figure 9.9.

FIGURE 9.8 Sagittal images of child models.

3 years old

5 years old

7 years old

Japanese Computational Phantoms

233

9.2.2.5 Arbitrary High Resolution Human Voxel Computational Phantoms

FIGURE 9.9 Example of variable posture model. Male model in sitting pose.

The 2 mm resolution20–22 computational phantoms currently used by NICT enable us to evaluate exposure to high-frequency electromagnetic radiation up to 3 GHz. We anticipate, however, that wireless communication devices will be used at frequencies above 3 GHz in the near future. This motivated us to develop higher resolution computational phantoms to be able to evaluate exposures to electromagnetic fields of higher frequencies. The spatial resolution of the computational phantoms is doubled when the voxel size is reduced by half. Therefore, it is possible to double the applicable frequency in this approximated manner. Nevertheless, problems could arise if the staircase shape of the curvature boundaries is maintained, because such rough modeling can cause significant effects on the simulation at higher frequencies. To solve the problem, Nagaoka and Watanabe created a technique for developing arbitrary high-resolution computational phantoms by smoothing the irregularities between tissue boundaries.28 Examples of computational phantoms on which this technique has been applied are shown in Figure 9.10. The figure shows that the improved smoothness of the outer and internal boundaries. 9.2.2.6 Applications in Electromagnetic Dosimetry

Researchers evaluated the safety of radio frequency electromagnetic field (RF-EMF) based on the specific absorption rate (SAR), which is the amount of RF energy absorbed per unit weight of a body. SAR is used as a measure of the thermal effects of RF-EMF exposure. SARs in the human body were estimated by applying the finite-difference time-domain (FDTD) method29 to realistic human voxel computational phantoms. The method is very effective for use with inhomogeneous human computational phantoms that have complex shapes because the region being simulated is divided into Yee cells (minute blocks) and the elements of the electromagnetic field of each Yee cell are calculated for every discrete time step. An example of an FDTD analysis to evaluate the SAR distribution in a human voxel computational phantom of a male in a sitting pose is shown in Figure 9.11.

9.3 Dose Calculation 9.3.1 Monte Carlo Code Systems Monte Carlo code systems combined with the voxel computational phantoms for calculating doses and related quantities have been constructed as user codes of the multipurpose photon and electron transport calculation code EGS4.30 The outlines of the constructed

234

Handbook of Anatomical Models for Radiation Dosimetry

Original model (2 mm resolution)

Improved model (1 mm resolution)

FIGURE 9.10 Example of arbitrary high-resolution human voxel model using female model.

code systems are listed in Table 9.7. In the whole code systems, we added the function to calculate radiation transport in voxel geometry. In addition to the main physical computational phantoms of the EGS4, Rayleigh scattering, Doppler broadening in Compton scattering, linearly polarized photon scattering, and electron impact ionization can also be considered. UCPIXEL3,31 calculates organ doses for external exposures in different irradiation conditions: (a) basic irradiation geometries of AP, PA, RLAT, LLAT, ROT, ISO, AB, BA with monoenergetic photons or electrons; (b) a cylindrical plane source around the voxel computational phantom emitting photons having arbitrary energy and angular spectra; (c) uniform volume source in air, or uniform plane source in ground emitting electrons having an arbitrary energy spectrum where the voxel computational phantom stands vertically on the ground; (d) a point source emitting monoenergetic photons or electrons uniformly over directions viewed by an arbitrary rectangle. When emitted photons have energy and angular spectra, the energy and angle of a radiation are sampled using the probability distribution functions previously prepared. UCSAF32 calculates absorption fractions (AFs) for internal exposures from source distributed uniformly in an assigned region that usually corresponds to an organ or a tissue, or in plural assigned regions. Arbitrary point sources having different intensities in the body can also be assumed. The energy can be selected from (a) monoenergy, (b) an arbitrary energy spectrum, (c) and a beta spectrum expressed by a theoretical formula.

Japanese Computational Phantoms

235

SAR(W/kg) 10–4

10–3

10–2

10–1

100

101

FIGURE 9.11 (See color insert following page 524.) Example of SAR distribution in a model exposed to plane wave (80 MHz). Incident power density is 1 mW cm−2.

TABLE 9.7 Developed Monte Carlo Code Systems at JAEA Code Name

Object

Obtained Quantity

UCPIXEL

External exposure

Organ dose

UCSAF

Internal exposure

Organ dose

UCWBC

In vivo measurement

Detection efficiency

UCRTP

Radiation therapy

Dose distribution

Specific absorption fraction Pulse height spectrum

UCWBC33 obtains calibration factors for in vivo measurements to quantify radionuclide uptake in the body. In this code system, combinatorial geometry is available in addition to voxel geometry, and this enables to easily model detectors set around the human body. Basically, the same source conditions as UCSAF are assumed; additional source distributed

236

Handbook of Anatomical Models for Radiation Dosimetry

uniformly in the whole body can be simulated considering radionuclides like potassium or cesium. Detector responses can be calculated for photons emitted in the body; in case of a scintillation detector the scintillation efficiency as a function of electron energy can be taken into account. UCRTP34 is a code system to calculate dose distribution in a patient body in radiation therapy. In this system, energy deposition is recorded per each voxel, while in other systems energy deposition is recorded per organ or tissue. Typical irradiation conditions for radiotherapy can be considered in this program. 9.3.2 External Dose Calculation 9.3.2.1 Photon Sato et al. investigated variations in organ doses for external photons8 using the developed voxel computational phantoms at JAEA. Organ doses were calculated for six typical external irradiation conditions of AP, PA, LLRAT, RLAT, ISO, ROT with 25 kinds of monoenergetic photons ranging from 0.01 to 10 MeV. In addition to the developed four Japanese voxel computational phantoms Otoko, Onago, JM, and JF, Caucasian male and female computational phantoms Rex and Regina2 having body sizes close to the reference man defined by ICRP35 Caucasian male computational phantom VIP-man36 having a large size, and stylized computational phantoms Adam and Eva37 based on the reference man data were used for analysis of dose variation by difference factors. The absorbed doses of 20 dominant organs were analyzed for the male and 22 dominant organs for the female. The organs considered are as follows: the bone marrow, the upper large intestine, the lower large intestine, the lung, the stomach, the bladder, the liver, the esophagus, the thyroids, the skin, the cortical bone, the adrenals, the brain, the small intestine, the kidneys, the pancreas, the spleen, the thymus, the testes (ovaries), the breasts, and the uterus. We adjusted the histories of the Monte Carlo calculations so that the statistical uncertainty of each considered organ would become less than 5%. Figure 9.12 gives the ratios of organ doses for the Japanese computational phantoms to those for the Caucasian computational phantoms. We selected five dominant organs and two energies of 0.05 and 1 MeV in this figure. We normalized organ doses for Japanese male computational phantoms to Rex, and doses for Japanese female computational phantoms to Regina. Several tendencies are confirmed from this figure. Obvious differences are larger for 0.05 MeV. In terms of irradiation geometries, the difference is large in the order of LAT > PA > AP, which is considered due to the path length difference to reach the organs. Thyroid doses indicate a large variation up to a several factor, while dose different for other organs seem within a factor of 2 in the energy range over 50 keV. Figure 9.13 shows energy dependencies of thyroid doses for Onago, JF, and Regina in LLAT geometry. The positions of thyroids tend to change according to individuals, and this could vary the dose. The positions of thyroid are clearly different even for Japanese computational phantoms as shown in Figure 9.14. The thyroids of Regina are expected to locate at deeper positions than those for Japanese, though the detailed anatomical data have not been released on Regina. It is not yet clear if the large difference is due to racial difference or due to individual difference, and further analysis is necessary for understanding systematically the difference. The bladder wall doses are shown in Figure 9.15 for AP and PA geometries. Doses are larger in AP geometry since the bladder exists near the front surface of a human body. In most cases, the VIP-man shows the smallest doses because it has a larger body size

Dose ratio (Japanese/Caucasian)

1.4

Dose ratio (Japanese/Caucasian)

Japanese Computational Phantoms

3

1.2

(0.05 MeV)

1.2

1

1 0.8

0.6

0.6

0.4

0.4

JM JF

0.2 Th

Otoko Onago

Lu

Li JM JF

St

Bl

1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2

Otoko Onago

RLAT (0.05 MeV)

5 1 5 0

Th Lu

Li

AP (1 MeV)

St

0

Bl

Otoko Onago

JM JF

0.2 0

5 2

2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0

1.4 AP

0.8

0

237

Th

Lu

Li

St

Bl Otoko Onago

JM JF

RLAT (1 MeV)

Th Lu Li

St

Bl

5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0

1.4 PA (0.05 MeV)

1.2

PA (1 MeV)

1 0.8 0.6 0.4 JM JF

0.2

Otoko Onago

Th Lu Li

St

Bl

0

JM JF

Otoko Onago

Th Lu Li

St

Bl

2.5 JM JF

Otoko Onago

LLAT (0.05 MeV)

JM JF

2

Otoko Onago

LLAT (1 MeV)

1.5 1 0.5

Th Lu

Li

St

Bl

0

Th Lu

Li

St

Bl

FIGURE 9.12 Ratios of organ doses for Japanese models to those for Caucasian models in the thyroid (Th), lungs (Lu), liver (Li), stomach (St), and bladder wall (Bl) for external photon exposure. Organ doses for Japanese male models were normalized to Rex, and doses for Japanese female models to Regina.

Organ dose per air-absorbed dose (Gy/Gy)

1.4 1.2 1.0 0.8 0.6 0.4

JF Onago

0.2 0.0 0.01

Regina

0.1

1

10

Photon energy (MeV) FIGURE 9.13 Comparison of thyroid doses among Onago, JF, and Regina for external photon exposure in LLAT geometry.

Handbook of Anatomical Models for Radiation Dosimetry

238

Organ dose per air-absorbed dose (Gy/Gy)

Organ dose per air-absorbed dose (Gy/Gy)

FIGURE 9.14 Thyroid positions of JF and Onago.

2 AP

1 Japanese JM Otoko JF Onago 0 0.01

Caucasian Rex Regina

0.1 1 Photon energy (MeV)

10

2 PA

1

0 0.01

0.1 1 Photon energy (MeV)

10

FIGURE 9.15 Comparison of bladder wall doses among different models for external photon exposure in AP and PA geometry.

than other computational phantoms. Doses for Japanese computational phantoms are smaller than those for Caucasian computational phantoms in AP geometry, and larger in PA geometry. This might indicate some anatomical difference between Caucasian and Japanese: possible reasons are that the bladder exists in deeper position in the Japanese, and that Caucasian has bigger hip size. It is difficult to draw a conclusion only from these data, and further investigation is desired. In total, concerning individual organs, the dose difference among different computational phantoms becomes up to a factor of 4–5 at 50 keV. Further, some differences need to be considered due to systematic difference that we observed. Further analyses are needed to clarify the reason of the difference: some systematic difference due to anatomical difference may exist like the case of bladder wall. The difference in effective dose was within 10% for almost all of cases and hardly exceeds 20% in the energy range of 50 keV to 10 MeV. The complete data set of the calculated organ doses is under preparation, and detailed discussion will be made elsewhere.

Japanese Computational Phantoms

239

It must be noted that, in case of partial irradiation by narrow beams, organ doses change drastically by the incident position and angle.38 In such cases, an appropriate individual computational phantom is considered to be necessary to obtain realistic organ doses and effective doses. 9.3.2.2 Electron Saito et al. calculated external electron doses for Otoko and Onago in AP, PA, and ISO geometries,4 and compared the data with a MIRD computational phantom by Ferrari.39 Figure 9.16 demonstrates examples of the dose coefficients for the three computational phantoms: the left data for liver doses in AP geometry is an example where a good agreement was observed among the computational phantoms; while the right illustrates the kidney doses in PA geometry. Which is an example where there was apparent difference between the MIRD computational phantom and the voxel computational phantoms. In external electron exposure, the organ depth is a dominant factor to determine the energy dependency of the dose coefficients. The Figure indicates the effective depth is quite similar for three computational phantoms in the left case, and obviously different in the right case. Figure 9.17 shows organ doses for Otoko and Onago normalized to those for the MIRD computational phantom. Large differences between different computational phantoms take place mainly around 10 MeV where the electron range is several centimeter. It makes a great difference if electrons reach organs or not; and dominant organs lie mostly at depths of several centimeter. The maximum dose difference observed in this comparison was a factor of 50 in kidney dose in PA geometry. A factor of 50 does not have any concrete meaning, but the electron dose for an individual organ sometimes shows a large discrepancy between different computational phantoms. The maximum difference observed in an effective dose was a factor of 2, which coincides with data of Kramer.40 Concerning external electron doses, we find that the available data are still limited and that we desire further investigations.

80 Equivalent dose per fluence (fSv/m2)

Equivalent dose per fluence (fSv/m2)

80 Liver (AP)

70 60 50 40 30 20

Otoko Onago MIRD

10 0

100

101

102

103

Electron energy (MeV)

104

Kidneys (PA)

70 60 50 40 30 20 10 0

100

101

102

103

104

Electron energy (MeV)

FIGURE 9.16 Examples of organ doses for external electron exposure calculated using Otoko, Onago and a MIRD model. (From Saito, K. et al., Jpn. J. Health Phys. 43, 122, 2008. With permission.)

Handbook of Anatomical Models for Radiation Dosimetry

10

Lungs (PA) Relative dose

Relative dose

10

1 Otoko/MIRD

10 Stomach (AP)

1

0.1 Otoko/MIRD

Onago/MIRD

0.1 101

102

103

Kidneys (PA) Relative dose

240

1

0.1 Otoko/MIRD Onago/MIRD

Onago/MIRD

104

0.01 101

10

102

103

104

0.01 101

102

103

104

10 Small int. (AP)

Bone marrow (AP)

Effective dose

Otoko/MIRD

1 Otoko/MIRD

Onago/MIRD

0.1 101

102 103 104 Electron energy (MeV)

Relative dose

1

Relative dose

Relative dose

2

1 AP PA ROT

Onago/MIRD

0.1 101

102 103 104 Electron energy (MeV)

0 101

102 103 104 Electron energy (MeV)

FIGURE 9.17 Organ doses and effective doses for Otoko and Onago normalized to those for a MIRD model in external electron exposure. (From Saito, K. et al., Jpn. J. Health Phys. 43, 122, 2008. With permission.)

9.3.3 Internal Dose Calculations 9.3.3.1 Absorption Fractions Internal dosimetry requires the fraction of energy emitted as a specified radiation in a source organ that is absorbed in a unit target organ, the so-called specific absorbed fractions (SAFs). The SAFs currently used by the International Commission on Radiological Protection (ICRP) have been obtained from calculations using MIRD computational phantoms. Positional relation among organs that is an important factor to determine SAFs could change significantly according to conditions; hence, SAF calculations for sophisticated computational phantoms are necessary to accurately evaluate internal doses. Kinase et al.32 calculated SAFs for photons in the Otoko, Onago, and MIRD computational phantom using the UCSAF code and then compared these results with published data to investigate the influence of several parameters on the SAFs. The source was assumed to emit monoenergetic photons in the energy range of 10 keV to 4 MeV and uniformly distributed in the source organ. The source organ was taken as the kidneys, but more than 100 target organs were considered. We show the calculated SAFs for kidneys as the target organ, as well as the source organ, in Figure 9.18, together with the data for Golem and Voxelman. A good agreement was found between the SAFs for all computational phantoms except for Voxelman, whose kidneys have a larger mass than those of the other computational phantoms, resulting in lower SAF values. Sato et al.5 calculated self-absorption fractions (Self-AFs) using Otoko, JM, and a MIRD computational phantom. Since Self-AFs are not explicitly affected by organ masses, unlike SAFs, they are suitable to discuss characteristics of energy deposition by radiations emitted from a source organ. Examples of Self-AFs are shown for six organs in Figure 9.19. It is clear that Self-AFs are similar for large organs having identical shapes among different

Japanese Computational Phantoms

241

101 Otoko Onago MIRD 5 type voxel MIRD 5 type Golem Voxelman

SAF (/kg)

100

10–1

10–2 10–2

100

10–1

101

Photon energy (MeV)

FIGURE 9.18 SAFs for the kidneys as the source and target organs calculated using several different models. (From Kinase, S. et al., Radiat. Prot. Dosim., 105, 557, 2003. With permission.)

Self-AF (–)

1

1

Brain

0.1

1

0.1

0.1

JM Otoko Voxel-MRD

0.01

0.01

(a)

0.1 1 Energy (MeV)

0.01

0.01

(b) 1

0.1 1 Energy (MeV)

0.01

0.01

(c) 1

Thyroid

Pancreas

Self-AF (–)

JM Otoko Voxel-MRD

JM Otoko Voxel-MRD

1

0.1 1 Energy (MeV) Urinary bladder wall

0.1 0.1

0.1

0.01 JM Otoko Voxel-MRD

0.01 (d)

Spleen

Kidneys

0.01

0.1 1 Energy (MeV)

0.01 (e)

JM Otoko Voxel-MRD

JM Otoko Voxel-MRD

0.01

0.1 1 Energy (MeV)

0.001 (f)

FIGURE 9.19 Self-AFs for six different organs calculated using Otoko, JM, and a MIRD model.

0.01

0.1 1 Energy (MeV)

242

Handbook of Anatomical Models for Radiation Dosimetry

computational phantoms, such as brain and kidneys. In case of small organs, however, the masses and shapes affect the Self-AFs significantly. When the shapes are identical, Self-AFs are determined by the masses: the portion of energy absorbed by the organ itself is largely affected by the size. The difference in spleen doses could be explained by the masses. While, in pancreas, thyroid and unitary bladder wall, the Self-AFs are not explained by the masses, and this indicated the shapes are important factors for these cases. Sato et al. investigated the effect of the posture of a human body on SAFs.41 In Figure 9.20, SAFs were compared between JM and JM2 for the esophagus, the lungs, the lower large intestine and the unitary bladder as target organs, when 16 organs were taken as source organs. JM is a computational phantom based on CT data taken in supine position, while JM2 is based on CT data in upright position. Data at photon energies of 0.03, 0.1, and 4 MeV are shown in this figure. Obviously, the difference is larger for lower energy photons whose large attenuations tend to affect the SAFs. We selected the four organs because they did not change the positions relative to the head between supine and upright position as indicated in Table 9.5. Other organs moved to lower directions in upright position because of the gravity. These movements are reflected in SAFs. Concerning the esophagus and lungs which are located in the upper parts of the body, the SAFs in upright position are generally smaller than those in supine position because the distances among organs increase according to the downward movements of other organs. Conversely, in case of the lower large intestine wall and unitary bladder wall located in the lower parts, the SAFs become larger in an upright position. Under the conditions considered here, the maximum observed difference was a factor of 3. Sato et al. examined the relation of SAFs to organ distance using Otoko, Onago, JF, and JM.42 In Figure 9.21, SAFs for several different combinations of source and target organs in the four computational phantoms are plotted together as a function of distance between the centers of gravity. The distance between the centers of gravity may not be the most suitable parameter to analyze SAFs in every case; nevertheless, it was confirmed that SAFs would be roughly expressed as a function of this distance. 9.3.3.2 S Value Kinase et al.43,44 evaluated S values—which are the mean absorbed doses per unit cumulated activity to the target organ from uniformly distributed radioactivity within the source organ—for several beta-ray emitters using the UCSAF code. S values have been used for dose estimates in radiological protection and medical diagnostic procedures. In particular, self-dose S values for positron emitters in the brain, the heart, and the urinary bladder play important roles for an accurate quantification of the doses to patients administered a radiopharmaceutical for clinical PET imaging. The evaluation of S values for the urinary bladder wall is indispensable for designing patient protocol strategies intended to minimize the dose for a specific radiopharmaceutical. In general, self-dose S values for positron emitters in the brain and the heart follow the inverse first power of the mass, especially if the absorbed fraction is approximated by unity (meaning no energy escape from the brain and the heart). S values for positron emitters within the urinary bladder are currently derived on the assumption that the dose at the surface of the content is approximately half the dose within their volume. Figure 9.22 shows the self-dose S values for the brain and the heart of the computational phantoms, for 11C, 13N, 15O, and 18F; the contributions from positrons and the two annihilation photons are distinguished. Readers can see that, as expected, the S values increase with increasing mean energy of beta ray. In emitters with high positron energies, positron

(a)

Number of source organs

0

5

10

0

5

Ratio of SAFs (upright/supine)

<2.5 <3

<2.5

<3

0.03 MeV 0.1 MeV 4 MeV

<0.3 <0.5 <0.7 <0.85 <0.95 <1.05 <1.15 <1.3 <1.5 <1.7 <2

LLIW

<1.7 <2

0.03 MeV 0.1 MeV 4 MeV

Ratio of SAFs (upright/supine)

<0.3 <0.5 <0.7 <0.85 <0.95 <1.05 <1.15 <1.3 <1.5

Esophagus

<4

<4

(b)

(d)

0

5

10

0

5

10

Ratio of SAFs (upright/supine)

<2.5

<3

0.03 MeV 0.1 MeV 4 MeV

<2.5

<3

0.03 MeV 0.1 MeV 4 MeV

<0.3 <0.5 <0.7 <0.85 <0.95 <1.05<1.15 <1.3 <1.5 <1.7 <2

UBW

Ratio of SAFs (upright/supine)

<0.3 <0.5 <0.7 <0.85 <0.95 <1.05<1.15 <1.3 <1.5 <1.7 <2

Lungs

<4

<4

FIGURE 9.20 Distribution of ratio of the SAF in upright posture to that in supine posture. Four target organs from (a) to (d) and 16 source organs were considered.

(c)

Number of source organs

Number of source organs Number of source organs

10

Japanese Computational Phantoms 243

Handbook of Anatomical Models for Radiation Dosimetry

244

10–1 Liver

Kidneys

Lower colon

SAF (kg–1)

Kidneys

10–3 Liver

Thyroid

Brain

Liver

Lower colon

Stomach Thyroid

Lower colon

10–5 0

20

40

60

Distance between the gravity centers FIGURE 9.21 SAFs as a function of distance between the gravity centers. Data calculated using Otoko, Onago, JM, and JF were shown together. 1.5 ×101

S value (μ Gy/MBq-s)

MIRD

Onago

Otoko

1.0 ×101

Two annihilation photons

5.0 ×102

Positrons

0.0 18

F

11

C

13

N

15

O 18F

11

C

13

N

15

O

18

F

11

C

13

N

15

O

Positron emitter

(a) Brain 3.5 × 10–1

MIRD

Otoko

Onago

S value (μ Gy/MBq-s)

3.0 × 10–1 2.5 × 10–1

Two annihilation photons

2.0 × 10–1 1.5 × 10–1 1.0 × 10–1

Positrons

0.5 × 10–2 0.0

18

F

11

C

13

N

15

O

18

F

11

C

13

N

15

O

18

F

11

C

13

N

15

O

Positron emitter (b) Heart FIGURE 9.22 Self-dose S values for positron emitters calculated using Otoko, Onago, and a MIRD model. (From Kinase, S. and Saito, K., Radiat. Prot. Dosim., 127, 197, 2007. With permission.)

Japanese Computational Phantoms

245

100

S value ratio (The simple assumption/EGS-UCSAF)

Otoko Onago

10

1

14

C

24

Na

32

P

60

Co

89

Sr 90Sr 90Y Beta-ray emitter

91

Y

137

Cs

147

Pm

204

Tl

FIGURE 9.23 Ratios of the S values for positron emitters derived from the currently used method to those by Monte Carlo calculations with voxel models. (From Kinase, S. et al., J. Nucl. Sci. Technol., Suppl. 4, 136, 2004. With permission.)

interactions with the organ greatly contribute to the self-dose S. The figure also shows the variation of self-dose S values among the computational phantoms. This is, as already known, due to the different masses of the organs. Figure 9.23 shows the ratios of the S values derived from the currently used simple assumption to those by Monte Carlo calculations using Otoko and Onago. The S values to the urinary bladder wall for several beta-ray emitters such as 14C, 24Na, 32P, 60Co, 89Sr, 90Sr, 90Y, 91Y, 137Cs, 147Pm, and 204Tl have been evaluated using the UCSAF code. The ratios increase as beta-ray energy decreases. The discrepancies between Otoko and Onago may be explained by the different ratios of the bladder wall mass to the bladder contents mass. The results substantiate that the S values from the simple assumption based on the bladder content masses of each computational phantom are conservative. However, the difference is quite large from the realistic value. The S value ratios between the Otoko/Onago computational phantom and the MIRD computational phantom are shown in Figure 9.24. The ratios do not vary much. The ratios for Otoko are almost twice those for Onago. This is considered to be due to the different masses of the bladder walls. Hence, the S value largely depends on the mass of the bladder wall. 9.3.4 Whole-Body Counting Figure 9.25 shows the peak efficiencies calculated by Kinase et al.33 using the voxel versions of the water-filled block-shape computational phantoms (water models) for an in vivo counting system at JAEA. The relationship of peak efficiency to photon energy shows a typical shape for the p-type high-purity Ge (HPGe) closed-ended coaxial detector: the

Handbook of Anatomical Models for Radiation Dosimetry

246

10

Otoko/MIRD

S value ratio

Onago/MIRD

1

14C

24Na

32P

60Co 89Sr

90Sr 90Y Beta-emitter

91Y

137Cs 147Pm 204Tl

FIGURE 9.24 S values for positron emitters in Otoko and Onago normalized to those in a MIRD model. (From Kinase, S. et al., J. Nucl. Sci. Technol., Suppl. 4, 136, 2004. With permission.)

10–3 Experiment Water-filled block-shape (voxel) Otoko Onago

Peak efficiency

Calculation

10–4

102

103 Energy (keV)

FIGURE 9.25 Peak efficiencies for in vivo measurements calculated using Otoko, Onago, and a water phantom model. Experimental data with the water phantom containing 137Cs and 40 K are shown together. (From Kinase, S. et al., Radiat. Prot. Dosim., 125, 189, 2007. With permission.)

Japanese Computational Phantoms

247

peak efficiency increases up to about 170 keV and then it decreases. The measured peak efficiencies for the physical water phantom containing 137Cs and 40K are also plotted in Figure 9.25. The error bars indicate the experimental uncertainties for the measurements of peak efficiencies. The calculated peak efficiencies for the water model agreed within 10% with the measured ones for photon energies 662 keV (137Cs) and 1461 keV (40K). The calculated peak efficiencies for Otoko and Onago are also shown in Figure 9.25. The peak efficiencies for Otoko and Onago are higher than those of the water models over the whole photon energy range calculated. At 662 keV, the peak efficiencies for Otoko and Onago are about 1.1 times greater than those of the water models. These discrepancies, which have a maximum (1.3 times) at 59.5 keV, are probably due to geometric effects, in particular the effective distances between the computational phantoms and the Ge semiconductor detector. This fact indicates that the peak efficiencies for the water models must be carefully used.

9.4 Application to Radiotherapy Radiation therapy is an important application of dose calculation technique, using voxel computational phantoms with which accurate individual doses can be obtained. The ICRP has recommended that, in radiation therapy, the evaluation of the dose accuracy be within 5%.45 Since this recommended dose accuracy should be attained considering overall possible errors, the net error in dose calculation should be within 1%–2%. This dose accuracy could be difficult to attain by conventional analytical methods, especially in some cases where the conditions are complicated: for example, when the elemental composition and density change drastically according to position in the patient body, the disequillibrium of electrons could lead to significant errors in the dose calculation. Monte Carlo calculations are potentially able to give sufficiently accurate doses, and commercial dose planning systems utilizing Monte Carlo calculations have come to be used in some clinics. However, Monte Carlo calculations need enormous computational time, and users also need some basic knowledge to properly perform the calculations. The requirements prevent the Monte Carlo systems from prevailing in Japan. Saito et al. developed a system for providing accurate dose distributions through networks to plural medical facilities based on voxel computational phantoms and Monte Carlo calculations.46,47 The basic concept of the system design is to provide reliable accurate doses for any conditions so that the calculated doses can be used as standard values for quality assurance and control (QA/QA); and to construct an environment where users without basic knowledge can easily employ the system. To attain the first requirement, the relation of patient body modeling to dose accuracy was carefully checked, and approximations in Monte Carlo calculations which could lead to degeneration of dose accuracy were avoided as much as possible. To attain the second requirement, currently used commercial therapy planning systems are supposed to be utilized as the user interface in the system. In this case, a user can prepare input data in exactly the same way as usual, and check the calculated dose distributions as usual. The schematic diagram of the system is shown in Figure 9.26. In the system, all dose calculations are performed at the dose calculation center using a high-performance computer. First, a user sends a data set of CT images and a treatment plan to the center through network; then, a sophisticated voxel computational phantom of the patient is automatically constructed in a

Handbook of Anatomical Models for Radiation Dosimetry

248

IMAGINE system

Dose calculation center

Hospital Transfer of CT data

Network

Monte Carlo calculation on the ITBL computer Construction of voxel phantom

Consideration of accurate spectrum

Transfer of dose distribution • Therapy planning • Quality assurance

FIGURE 9.26 Schematic diagram of the dose calculation system IMAGINE for supporting medical facilities through the network. (From Saito, K. et al., Radiat. Prot. Dosim., 116, 190, 2005. With permission.)

short time; and a Monte Carlo calculation is performed utilizing a super-parallel computing; then the calculated dose distribution is sent back to the clinic and used for therapy planning or QA/QC. Concerning system setup, all a user has to do is to buy an ordinary personal computer, install the specific server application, and connect the PC to the Internet. To obtain accurate doses, an important factor is to evaluate a realistic x-ray energy spectrum that cannot be directly measured. For this, a user is requested to prepare the design of the irradiation head, and dose distribution data measured in a water phantom irradiated by the irradiation equipment. Then, the proper x-ray spectrum is obtained from a Monte Carlo simulation using an irradiation head computational phantom precisely constructed based on the design data. In the simulation, parameters of electrons incident on the target are finely adjusted so that the calculated dose distribution in a water phantom would reconstruct the measured data well. Concerning patient computational phantoms, we investigated the relation of elemental compositions to dose accuracy as shown in Table 9.8. We constructed several voxel computational phantoms having different classification of the elemental compositions modifying Otoko and Onago. The computational phantom was basically segmented into six regions: air, skin, muscle, adipose, lung, and skeleton computational phantomed as a mixture of cortical bone and bone marrow. The most sophisticated computational phantom is Model A, in which appropriate elemental compositions are assigned to all tissues. The most simplified computational phantom is model F in which all voxels are approximated by water having adjusted densities. The computational phantoms were increasingly simplified from A to F, that is, different tissues having different elemental compositions approximated some regions. Several tumors were set at different positions in the computational phantom; rotational irradiation was simulated for each tumor; and the average dose in the tumor was compared among the computational phantoms. Three different x-ray energies covering wide energy range that could be used in radiotherapy were considered to investigate the effect of difference in interactions due to energy.

Cortical bone

Water

J

Cortical bone

G

Water

Cortical bone

H

Cortical bone

E

F

I

Cortical bone

Water

Bone marrow

Bone marrow

Bone marrow

Bone marrow

Bone marrow

Cortical bone

C

D

Bone marrow Bone marrow

Cortical bone

Cortical bone

A

Adipose

Adipose

Adipose

Water

Soft tissue

Ethanol

Adipose

Adipose

Adipose

Adipose

Elemental Compositions

Bone Marrow

B

Model

Cortical Bone

Regions

Muscle

Muscle

Water

Muscle

Muscle

Muscle

Muscle

Muscle

Skin

Skin

Skin

Skin

Skin

Effect of Model Simplification Concerning Elemental Compositions on Calculated Dose Accuracy

TABLE 9.8

Lung

Lung

Muscle

Lung

Lung

Lung

2.5

2.4

0.1

0.4

0.2

0.2

14

14

4.0

2.6

1.9

0.9

0.1

–

– 0.1

CTRTx (120 kV) LINAC (6 MV, 24 MV)

Maximum Dose Difference (% Deviation from Model A)

Japanese Computational Phantoms 249

250

Handbook of Anatomical Models for Radiation Dosimetry

Obviously, the inaccurate doses are obtained when other materials substitute for skeletons having a relatively high atomic number element of Ca. Further, it was found that substitution of adipose by other materials could cause not negligible error in some cases. This is because the adipose contains carbon having a relatively small atomic number at a large portion compared with other tissues like muscle, skin, or lung. We developed software to automatically construct a voxel computational phantom from CT data. This software first discriminates artificial materials such as the coach, the fixing tools, and the blanket from the patient body. Then, the patient body is segmented into the six tissues. Fundamentally the segmentation is performed setting threshold CT values specific to each tissue; then, segmented tissues are corrected using information on tissue positions. The first version of the system was completed in 2007, and the terminals of the system were set up at several research institutes supposing the Altix 3700 computer at JAEA Tokai establishment as the dose calculation center. The system performance has been checked by operation tests connecting these facilities. From many runs, it was confirmed that time required for data transfer and pre and post data processing is 5–10 min. Also, it takes about 1 h on Altix 3700 with 128 CPU to perform a Monte Carlo dose calculation of one billion histories.

9.5 Conclusion (Future Plan) More than 10 Japanese voxel computational phantoms have been developed by JAEA and NIST, and have been used for different purposes. From these studies, it has been found that voxel computational phantoms are quite effective to obtain realistic dosimetric quantities and to investigate variations of those quantities by anatomical reasons. The calculation of dosimetric quantities is still ongoing, and in near future data sets of dosimetric quantities using the developed voxel computational phantoms and Monte Carlo code systems will be compiled. On the basis of analyses of the compiled data together with data from other Asian computational phantoms, researchers should discuss if reference Asian voxel computational phantoms would be necessary, and what should be investigated using Japanese computational phantoms in future. The techniques-related voxel computational phantoms have a wide range of application. Obviously, one important application is in the medical fields. An example on the dose calculation system for x-ray therapy was shown in this chapter; meanwhile, a therapy planning system for boron neutron capture therapy (BNCT) has been developed at JAEA48 and played an important role in BNCT in Japan though details are not described here. As sophisticated radiotherapy like intensity modulated radiation therapy (IMRT) or BNCT is becoming popular, the necessity for advanced dose evaluation techniques would be increasing. Further, dose evaluation for radiation diagnoses would also be important, since advanced diagnoses leading to relatively high doses such as CT, PET, and SPECT have come to be used. We are planning to apply the voxel computational phantom techniques to dose evaluation for a variety of these medical fields. Another essential application after the new ICRP recommendation is dose evaluation for animals and plants. So far, simplified computational phantoms have been employed for radiation protection purposes; nevertheless, it would be anticipated that realistic voxel computational phantoms of reference animals and plants would be utilized in future. As an initial trial, a frog voxel computational phantom was developed.49 Further, to analyze

Japanese Computational Phantoms

251

animal experiments for medicine development or for radiation effect research, quite a few voxel computational phantoms for animals have been already constructed. This is also in our future scope; in fact, some dose calculations using voxel computational phantoms have been already started.

References 1. Snyder, W.S. et al. Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, Journal of Nuclear Medicine, 10(Suppl. 3), 7, 1969. 2. Schlattl, H., Zankl, M., and Petoussi-Henss, N. Organ dose conversion coefficients for voxel models of the reference male and female from idealized photon exposures, Physics in Medicine and Biology, 52, 2123, 2007. 3. Saito, K. et al. Construction of a computed tomographic phantom for a Japanese male adult and dose calculation system, Radiation and Environmental Biophysics, 40, 69, 2001. 4. Saito, K. et al. Construction of a voxel phantom based on CT data for a Japanese female adult and its use for calculation of organ doses from external electrons, Japanese Journal of Health and Physics, 43, 122, 2008. 5. Sato, K. et al. Japanese adult male voxel phantom constructed on the basis of CT images, Radiation Protection Dosimetry, 123, 337, 2007. 6. Sato, K. et al. Development of a Japanese adult female voxel phantom, Journal of Nuclear Science and Technology, (accepted). 7. Tanaka, G. and Kawamura, H. Anatomical and Physiological Characteristics for Asian Reference Man, NIRS-M-115, Hitachinaka, Japan: National Institute of Radiological Sciences, 1996. 8. Sato, K., Endo, A., and Saito, K. Dose conversion coefficients calculated using a series of adult Japanese voxel phantoms against external photon exposure. JAEA-Data/Code 2008–016, 2008. 9. Sato, K. et al. Development of a voxel phantom of Japanese adult male in upright posture, Radiation Protection Dosimetry, 127, 205, 2007. 10. Dimbylow, P.J. and Mann, S.M. SAR calculations in an anatomically realistic model of the head for mobile communication transceivers at 900 MHz and 1.8 GHz, Physics in Medicine and Biology, 39, 1537, 1994. 11. Gandhi, O.P., Lazzi, G., and Furse, C.M. Electromagnetic absorption in the human head and neck for mobile telephones at 835 and 1900 MHz, IEEE Transactions on Microwave Theory and Techniques, 44, 1884, 1996. 12. Okoniewski, M. and Stuchly, M.A. A study of the handset antenna and human body interaction, IEEE Transactions on Microwave Theory and Techniques, 44, 1855, 1996. 13. Wang, J.Q. and Fujiwara, O. FDTD analysis of dosimetry in human head model for a helical antenna portable telephone, IEICE Transactions on Communications, E83B, 549, 2000. 14. Watanabe, S. et al. Characteristics of the SAR distributions in a head exposed to electromagnetic fields radiated by a hand-held portable radio, IEEE Transactions on Microwave Theory and Techniques, 44, 1874, 1996. 15. Dawson, T.W., Caputa, K., and Stuchly, M.A. A comparison of 60 Hz uniform magnetic and electric induction in the human body, Physics in Medicine and Biology, 42, 2319, 1997. 16. Dimbylow, P. Resonance behaviour of whole-body averaged specific energy absorption rate (SAR) in the female voxel model, NAOMI, Physics in Medicine and Biology, 50, 4053, 2005. 17. Dimbylow, P.J. FDTD calculations of the whole-body averaged SAR in an anatomically realistic voxel model of the human body from 1 MHz to 1 GHz, Physics in Medicine and Biology, 42, 479, 1997.

252

Handbook of Anatomical Models for Radiation Dosimetry

18. Mason, P.A. et al. Recent advancements in dosimetry measurements and modeling, in Radio Frequency Radiation Dosimetry, Klauenberg, B.J. and Miklavcic, D., eds., Dordrecht: Kluwer Academic Publishers, 2000, pp. 141–155. 19. Tinniswood, A.D., Furse, C.M., and Gandhi, O.P. Power deposition in the head and neck of an anatomically based human body model for plane wave exposures, Physics in Medicine and Biology, 43, 2361, 1998. 20. Nagaoka, T., Kunieda, E., and Watanabe, S. Proportion–corrected scaled voxel models for Japanese children and their application to the numerical dosimetry of specific absorption rate for frequencies from 30 MHz to 3 GHz, Physics in Medicine and Biology, 53, 6695, 2008. 21. Nagaoka, T. et al. An anatomically realistic whole-body pregnant-woman model and specific absorption rates for pregnant-woman exposure to electromagnetic plane waves from 10 MHz to 2 GHz, Physics in Medicine and Biology, 52, 6731, 2007. 22. Nagaoka, T. et al. Development of realistic high-resolution whole-body voxel models of Japanese adult males and females of average height and weight, and application of models to radio-frequency electromagnetic-field dosimetry, Physics in Medicine and Biology, 49, 1, 2004. 23. NIBH. Human Body Dimensions Data for Ergonomic Design, Tokyo: Research Institute of Human Engineering for Quality Life, 1996. 24. Tanaka, G., Nakahara, Y., and Nakazima, Y. Japanese reference man 1988-IV. Studies on the weight and size of internal organs of Normal Japanese, Nippon Igaku Hoshasen Gakkai Zasshi, 49, 344, 1989. 25. Sederberg, T.W. and Parry, S.R. Free-form deformation of solid geometric models, ACM Computer Graphics (SIGGRAPH’86 Conference Proceedings), 20, 151, 1986. 26. Nagaoka, T. and Watanabe, S. Development and application of human voxel models in Japan, 17th International Zurich Symposium on Electromagnetic Compatibility, Suntec City, Singapore, 59, 2006. 27. Nagaoka, T. and Watanabe, S. Postured voxel-based human models for electromagnetic dosimetry, Physics and Medicine in Biology, 53, 7047, 2008. 28. Nagaoka, T. and Watanabe, S. Technique using implicit fairing and specific absorption rates to improve spatial resolution of whole-body human voxel models exposed to plane waves in GHz bands, URSI General Assembly, Chicago, IL, August 7–16, 2008. 29. Tavlove, A. Computational Electromagnetics: The Finite-Difference Time-Domain Method, Boston, MA: Artech House, 1995. 30. Nelson, W.R., Hirayama, H., and Rogers, D.W.O. The EGS4 Code System, SLAC-265, Stanford, CA: Stanford Linear Accelerator Centre, Stanford University, 1985. 31. Funabiki, J. et al. An user code with voxel geometry and a voxel phantom generation system, in KEK Proceedings, Ibaraki-ken, Japan, 2000. 32. Kinase, S. et al. Evaluation of specific absorbed fractions in voxel phantoms using Monte Carlo simulation, Radiation Protection Dosimetry, 105, 557, 2003. 33. Kinase, S. et al. Application of voxel phantoms and Monte Carlo method to whole-body counter calibration, Radiation Protection Dosimetry, 125, 189, 2007. 34. Funabiki, J. et al. An EGS4 Monte Carlo user code for radiation therapy planning, in KEK Proceedings 2001–22, Ibaraki-ken, Japan, 2001. 35. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Oxford: Pergamon Press, 2003. 36. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Physics, 78, 476, 2000. 37. Kramer, R. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods: Part I. The male (ADAM) and female (EVA) adult mathematical phantoms, GSF-Report S-885, Institut fuer Strahlenschutz, GSF-Forschungszentrum fuer Umwelt und Gesundheit, Neuherberg-Muenchen, 1982. 38. Ohnishi, S. et al. Analysis of localised dose distribution in human body by Monte Carlo code system for photon irradiation, Radiation Protection Dosimetry, 111, 65, 2004.

Japanese Computational Phantoms

253

39. Ferrari, A., Pelliccioni, M., and Pillon, M. Fluence to effective dose and effective dose equivalent conversion coefficients for electrons from 5 MeV to 10 GeV, Radiation Protection Dosimetry, 69, 97, 1997. 40. Kramer, R. Effective dose ratios for tomographic and stylized models from external exposure to electrons, in The Monte Carlo Method: Versatility Unbounded in A Dynamic Computing World, Chattanooga, TN, April 17–21, 2005. 41. Sato, K. and Endo, A. Analysis of effects of posture on organ doses by internal photon emitters using voxel phantoms, Physics in Medicine and Biology, 53, 4555, 2008. 42. Sato, K. et al. Development of a Japanese adult female voxel phantom, in 2005 Annual meeting of Atomic Energy Society of Japan, Hiratsuka, March 29–31, 2005. 43. Kinase, S. and Saito, K. Evaluation of self-dose S values for positron emitters in voxel phantoms, Radiation Protection Dosimetry, 127, 197, 2007. 44. Kinase, S. et al. Evaluation of S values for beta-ray emitters within the urinary bladder, Journal of Nuclear Science and Technology, Suppl. 4, 136, 2004. 45. ICRP. Determination of Absorbed Dose in a Patient Irradiated by Beams of X or Gamma Rays in Radiotherapy Procedures, ICRP Publication 24, Oxford: Pergamon, 1976. 46. Saito, K. et al. Development of the accurate dose calculation system IMAGINE for remotely aiding radiotherapy, in KEK-Proceedings 2003–15, Ibaraki-ken, Japan, 2004. 47. Saito, K. et al. Dose calculation system for remotely supporting radiotherapy, Radiation Protection Dosimetry, 116, 190, 2005. 48. Kumada, H. et al. Verification of the computational dosimetry system in JAERI (JCDS) for boron neutron capture therapy, Physics in Medicine and Biology, 49, 3353, 2004. 49. Sakae, K. Voxel-based frog phantom for internal dose evaluation, Journal of Nuclear Science and Technology, 45, 1049, 2008.

10 Korean Computational Phantoms: KMIRD, KORMAN, KORWOMAN, KTMAN-1, KTMAN-2, and HDRK-Man Choonsik Lee and Chan Hyeong Kim

CONTENTS 10.1 Introduction ............................................................................................................... 255 10.1.1 Computational Human Phantom .............................................................. 255 10.1.2 Need for Korean Computational Phantoms ............................................ 256 10.1.3 Korean Reference Man Project ................................................................... 256 10.2 Developments of Korean Computational Phantoms ........................................... 256 10.2.1 Korean Reference Organ Data ................................................................... 256 10.2.2 Korean Stylized Computational Phantoms ............................................. 260 10.2.3 Korean Voxel Computational Phantoms .................................................. 262 10.2.3.1 KORMAN and KORWOMAN.................................................... 262 10.2.3.2 KTMAN-1 and KTMAN-2 .......................................................... 265 10.2.3.3 High-Definition Reference Korean (HDRK)-Man ................... 268 10.3 Conclusions and Future Work ................................................................................ 274 References ............................................................................................................................. 276

10.1 Introduction 10.1.1 Computational Human Phantom In the vast majority of instances, direct measurement of radiation-absorbed dose to radiosensitive tissues within the human body is not feasible, and thus must be derived from measurable quantities with the assistance of dose conversion coefficients. Dose conversion coefficients, expressed as the organ-absorbed dose per unit entrance skin dose or air kerma, are in turn calculated via radiation transport simulations within computational anthropomorphic computational phantoms of the human body’s exterior and internal organ structure. The first heterogeneous computational phantom of the human body, known as Medical Internal Radiation Dose (MIRD) computational phantom, was designed at the Oak Ridge National Laboratory (ORNL) as representative of the 50th percentile adult human male and several modified or extended versions of the original MIRD computational phantom have been developed thereafter for use in organ dosimetry within diagnostic radiology, 255

256

Handbook of Anatomical Models for Radiation Dosimetry

nuclear medicine, and radiation protection.1–3 Although these MIRD-type or stylized computational phantoms have significantly contributed to computational radiation dosimetry, the potential for unrealistic dose estimates does exist due to their rather simplistic representation of human anatomy. One approach to overcoming this inherent limitation is to develop voxel (also called tomographic) computational phantoms, which can realistically describe both the shape of the external body and of the internal organs through the use of tomographic images. The voxel computational phantom was first introduced by Gibbs et al. to evaluate patient risk from interproximal radiography in 1984.4 Recently, Zaidi and Xu provide summary data on currently existing voxel computational phantoms (through 2007) in a review article.5 10.1.2 Need for Korean Computational Phantoms Most existing stylized and voxel computational phantoms are based on the reference data and tomographic images of individuals from Caucasian populations. However, the Caucasian population represents only ∼20% of the world’s total population, while Asian populations comprise more than 50% of that total population. Furthermore, since radiation dose distributions within the human body are greatly dependent upon physical and anatomical characteristics of the exposed individual, significant dosimetric discrepancies may result from the application of Caucasian-based anatomic computational phantoms to approximate individuals within other racial groups. In an effort to construct Asian voxel computational phantoms, Saito et al. introduced the first Asian adult voxel computational phantom OTOKO, segmented from whole-body computed tomography (CT) images of a Japanese patient whose body dimensions were in agreement with those of the Japanese Reference Man.6,7 Nagaoka et al. evaluated dosimetric effects of radio frequency electromagnetic field using other Japanese adult male and female voxel computational phantoms.8 10.1.3 Korean Reference Man Project Considerable effort has been made to establish a system of reference for Korean human computational phantoms at Hanyang University, Korea since 2000, through the support of the Korean Ministry of Science and Technology. An important goal of this project is the construction of Korean computational anatomic computational phantoms for use in both medical and radiation protection dosimetry. Healthy adult male volunteers whose body sizes and weights were in close agreement with the average Korean adult (50th percentile) were recruited only for the purpose of reference organ data collection and computational phantom development, whereas most existing voxel computational phantoms are based upon CT images of individual patients who may or may not display age-matched 50th percentile body morphometry.

10.2 Developments of Korean Computational Phantoms 10.2.1 Korean Reference Organ Data Since anatomy and body composition significantly affects the resulting radiation doses, it is important for non-Caucasian population to set up their own reference data

Korean Computational Phantoms

257

and evaluate dosimetric difference between that data and Caucasian-based data. In effort to establish an Asian Reference Man, research on “Compilation of Anatomical Physiological and Metabolic Characteristics for a Reference Asian Man” was conducted by International Atomic Energy Agency (IAEA) in 1998.9 The IAEA task group collected anthropomorphic parameters of the six age groups from nine Asian countries, and reported data on organ mass, physiological parameters, and daily dietary intake of the Reference Asian Man. The report concluded that the mean organ masses of Asian population tended to be less than the ICRP values, and therefore each Asian country should undertake a specific research program to establish specific national reference values. Tanaka et al. organized and published data collection and summarization of organ mass.10 In Korea, Kim et al. performed the measurement of organ dimension for the fi rst time and submitted the data to the IAEA report.11 They collected the mass of six internal organs, obtained from autopsies of 1921 adults by the National Institute of Scientific Investigation in Korea. To significantly expand this dataset, the extensive in vivo organ mass measurement was performed by the Korean Reference Man Project to provide a Korean average volume and mass of 15 organs and nine bones in adult male and female. Different from the previous anatomical data based on autopsy, in vivo measurement of organ size was performed using three-dimensional (3D) rendering from contiguous whole-body magnetic resonance (MR) images. The organ masses were calculated using nominal organ density, and compared with literature data of ICRP and Asian Reference Man. The healthy volunteers (66 males and 55 females) whose body size was close enough to the average body dimension for Korean adults were recruited for MR scanning. The average body dimensions for Korean adult male and female were obtained from a national anthropometric survey performed by Korea Research Institute of Standards and Science (KRISS) in 1997.12 The height of the recruited subjects ranged from 164 to 178 cm for males, and from 152 to 170 cm for females. The weight ranged from 45 to 64 kg for males, and from 40 to 56 kg for females. The ages of male and female volunteers recruited in this study ranged from 27 to 49 and from 21 to 50, respectively. Whole-body MR scans were performed at Hanil Hospital, Seoul, Korea using MagneTom vision 1.5T computational phantom (Siemens, Munich, Germany). Before MR scanning of the recruited volunteers, optimal protocol was acquired through preliminary scanning, since some organs not scanned in common clinical MR scan should be included to imaging criteria along with total organ boundary. The body scanning was divided into eight parts: head, neck, upper trunk, lower trunk, upper leg, lower leg, upper arm, and lower arm. Head, upper and lower trunk regions were scanned with a 6 mm slice interval in axial, coronal, and sagittal direction, respectively. Leg and arm regions were scanned with 5 and 2 mm slice interval in coronal and sagittal direction, respectively. Neck regions were scanned with 3 mm slice interval in axial and 5 mm slice interval in coronal and sagittal direction, respectively. Anatomists manually segmented the organ contour in MR images using ISG Allegro workstation (ISG Technologies, Inc., Ontario, Canada), which is widely used for image segmentation, reconstruction, and measurement in medical imaging field. Two anatomists segmented one image slice to minimize human error. The contour of 15 internal organs (brain, eye, gall bladder, heart, kidney, liver, lung, pancreas, stomach, spleen, testes, thymus, thyroid, urinary bladder, and uterus) and nine bones (femur, tibia + fibula, humerus, radius + ulna, pelvis, cervical spine, thoracic and lumber spine, skull, and clavicle) were manually segmented. The segmented 2D images were piled up consecutively to render 3D volume.

258

Handbook of Anatomical Models for Radiation Dosimetry

An in-between image of the two contiguous images was generated by linear interpolation in Allegro workstation to obtain smooth organ surface. The mass of segmented organ and bone was calculated by multiplying the measured volume to the Asian reference density reported by the IAEA.9 The Asian organ density is the best estimation at the moment until Korean reference organ density will be available. Since Asian nominal density of gall bladder, liver, urinary bladder was not available in the report, data from the International Commission on Radiation Units and Measurements (ICRU) were substituted for those three organs.13 Mean, median, percentiles, and standard deviation of volume data were evaluated using statistics software SPSS 12.0.1 (LEAD Technologies, Inc., Chicago). In the case of bone, median values were used in mass calculation instead of mean values as the sample size was relatively small and outliers existed. Bone density from ICRU46 was used for all bone sites because Asian reference bone density was not available. An observation was called an outlier if its value was greater than the upper limit or less than the lower limit. The upper limit and lower limit are defi ned as Upper limit = Q3 + 1.5(Q3 − Q1)

(10.1)

Lower limit = Q1 − 1.5(Q3 − Q1)

(10.2)

where Q1 and Q3 are the 25th percentile and the 75th percentile, respectively. The organs for which the median volume was used for mass calculation were the heart, the kidneys, the pancreas, the stomach and the thyroid for male, and the heart for female. The statistical results of the organ volume, mass, and density are tabulated in Table 10.1 for adult male and female. A relatively large standard deviation was shown in some organs. Respiratory (lung) and circulatory (heart) organs continuously moved during MR scanning and caused significant discrepancy, even for the same subject. Depending on the time passed after having a meal, the volume of organs in alimentary (stomach) and urogenital (urinary bladder) systems was also highly variable even in a single person. Although the brain does not move and its size is not variable, brain volume shows a significant difference among subjects. The gall bladder shows a significant difference of standard deviation between male and female as well. Between meals, bile secreted from the liver is stored in gall bladder. During a meal, the bile is injected from the gall bladder into the small intestine. Consequently, the gall bladder showed considerable variation in shape and size. Organs of Korean males were up to 25% (thyroid) heavier than those of Korean females, except the gall bladder and the pancreas, of which females were up to 30% and 70% heavier than those of males, respectively. The resulting organ mass was compared with those of Asian and ICRP reference adults, and is tabulated in Table 10.2. The results from the previous study performed by Kim et al. were also included for comparison. Figure 10.1 shows the relative Korean organ mass divided by the values of ICRP and Asian reference adults. The mass of all organs of Korean subjects agreed with those of ICRP and Asian reference data within 60%, except for that of the gall bladder and the pancreas. The mass of Korean adult gall bladder was up to 63% (male) and 183% (female), respectively, higher than that of Asian reference adults. The pancreas of Korean adults was 76% (male) and 52% (female) lighter than that of ICRP reference adults.

Korean Computational Phantoms

259

TABLE 10.1 Volume and Mass of Organs and Bones in Korean Adult Males (M) and Females (F) and Reference Density Used for Mass Calculation Volume (cm3) Mean

Median

SD

Mass (g)

Density (g/cm3)

M

F

M

F

M

F

M

F

1478.0 19.3 12.4 625.4 325.1 1357.0 4320.0 39.0 357.0 160.2 27.7 38.6 15.3 202.3

1362.0 18.4 16.3 557.3 276.0 1161.0 2857.0 55.0 306.6 144.5 – – 11.3 171.3

1467.0 19.2 12.4 618.9 321.6 1375.0 4474.0 32.0 314.4 156.6 26.9 39.0 14.6 173.3

1363.0 18.7 14.8 540.9 271.2 1138.0 2954.0 44.3 307.0 143.7 – – 10.8 110.4

125.0 0.7 0.2 148.5 42.5 196.6 885.4 17.4 171.7 50.1 9.4 22.3 5.2 112.5

105.5 2.7 9.4 142.1 33.2 182.6 742.3 29.0 154.6 48.6 – – 4.5 125.3

1522 20 13 637 338 1438 1123 34 330 170 29 40 15 210

1403 19 17 557 290 1231 743 58 322 153 – – 12 178

1.03 1.03 1.03a 1.03 1.05 1.06a 0.26 1.05 1.05 1.06 1.04 1.03 1.05 1.04a

–

104.7

–

95.1

–

39.4

–

109

1.04

Bone Femora Tibiae, fibula Humerus Radii, ulnae Pelvis

427.0 309.9 154.5 82.7 921.5

320.4 237.7 95.8 55.6 772.0

422.7 308.8 147.5 84.0 924.3

324.0 238.2 101.6 58.4 775.9

66.5 55.3 30.6 14.6 134.0

262.2 50.6 15.6 8.6 123.2

562 411 215 123 1192

431 317 148 85 1079

Cervical spine T- and L-spine Skull Clavicles

108.1 731.2 479.8 24.0

86.3 641.1 – 16.5

108.8 689.9 508.2 23.3

85.0 613.8 – 17.3

20.1 216.6 116.2 8.8

14.3 131.1 – 5.3

154 918 818 33

121 816 – 24

Internal organs Brain Eyes Gall bladder Heart Kidneys Liver Lung Pancreas Stomach Spleen Testes Thymus Thyroids Urinary bladder Uterus

1.33 1.33b 1.46 1.46c 1.29(M), 1.39(F)d 1.42e 1.33f 1.61g 1.41h

Source: Park, S.H. et al., Radiat. Prot. Dosim., 118(3), 275, 2006. With permission. a Organ density of ICRU46 was used for gall bladder, liver, and urinary bladder, of which Asian data were not available. b Density of femora was also used for tibiae and fibula. c Density of humerus was also used for radii and ulnae. d Density of sacrum was substituted for pelvis of Korean adults. Densities for male and female were separately given. e Density of vertebral column (C4) was substituted for whole cervical spine of Korean adults. f Density of vertebral column (D6, L3) was substituted for T- and L-spine of Korean adults. g Density of cranium in ICRU46 was substituted for skull of Korean adults. h Density of second and sixth ribs was assumed to be close to clavicles of Korean adults since density of clavicles was not available in ICRU46.

Handbook of Anatomical Models for Radiation Dosimetry

260

TABLE 10.2 Comparison of Organ Mass of Korean Adult Male (M) and Female (F) with the Reference Data of the Previous Korean Study by Kim et al., ICRP, and Asian Reference Model Korean (This Study) Organs Brain Eyes Gall bladder Heart-with blood Kidneys Liver Lung-with blood Pancreas Stomach Spleen Testes Thymus Thyroids Urinary bladder Uterus

Korean (Kim et al.)

ICRP 89

Asian

M

F

M

F

M

F

M

F

1522 20 13 637

1403 19 17 557

– – – 348.8a

– – – 301.6a

1450 15 10 840

1300 15 8 620

1610 15 8 780

1483 12 6 620

338 1438 1123

290 1231 743

251.6 1863.9 1204.4

227.7 1610.9 957.4

310 1800 1200

275 1400 950

320 1600 1200

280 1400 910

34 330 170 29 40 15 210c

58 322 153

56.4 – 67.3 – – – – –

54.0 – 58.2 – – – – –

140 400 150 35 25 20 – –

120 370 130 – 20 17 – 80

12 178c 109

130 380b 140 37 30 19 140c –

110 290b 120 – 29 17 115c 70

Source: Park, S.H. et al., Radiat. Prot. Dosim., 118(3), 275, 2006. With permission. a Mass of heart without blood. b Since stomach mass was not available in Tanaka’s paper, it was obtained from IAEA-TECDOC-1005. c Mass of urine-containing urinary bladder.

10.2.2 Korean Stylized Computational Phantoms Based on the Korean reference organ data collected from adult male and female volunteers, stylized computational phantoms were constructed to represent Korean reference adult male and female by modifying the equations of the outer body and the internal organs of the ORNL adult computational phantom.14 Along with the reference organ data, the reference anthropometric data, obtained from the Survey of National Physique Standard (SNPS) by KRISS, were employed.12 The SNPS has been conducted every 6 years since 1979. In the latest survey in 1997, the body size data of 13,062 Korean individuals (6578 males and 6484 females) across various age ranges were obtained. The data of adult males in age range from 25 to 50 were adopted. The average height and weight of Korean adult males are 170.2 cm, and 68.2 kg, respectively. Mathematical surface equations representing a total of 13 organs were modified to match Korean reference organ volume: brain, gall bladder, heart, kidneys, liver, lung, pancreas, spleen, stomach, testes, thymus, thyroid, and urinary bladder. The outer body dimension was modified to match the Korean anthropometric data, since the height of the ORNL computational phantom is 178.6 cm, whereas that of the average Korean male is 170.2 cm. Referring to the Korean reference body dimension, the height of the head, trunk, and legs was separately decreased, except that of head, which was increased due to the larger volume of average Korean brain than that of the ORNL adult computational phantom. The height of the head, including neck, was changed from 28.6 to 31 cm (+8.4%), that of the

Korean Computational Phantoms

261

Organ mass ratio (korean/reference data)

3.0 Korean vs. ICRP (M) Korean vs. ICRP (F) Korean vs. Asian (M) Korean vs. Asian (F)

2.5

2.0

1.5

1.0

0.5

Uterus

Urinary bladder

Thyroids

Thymus

Testes

Spleen

Stomach

Pancreas

Lung

Liver

Kidneys

Heart

Gall bladder

Eyes

Brain

0.0

FIGURE 10.1 Comparison of Korean adult male and female average organ mass with ICRP and Asian reference data. (From Park, S.H. et al., Radiat. Prot. Dosim., 118(3), 275, 2006. With permission.)

trunk was changed from 70 to 64 cm (−8.6%), and that of the legs was changed from 80 to 76 cm (−6.3%). The arms, including arm bones, of the ORNL adult computational phantom were extracted from the trunk to match the trunk width of the Korean computational phantom to the Korean male anthropometric data. The trunk width and thickness of the ORNL adult computational phantom, where arms and arm bones are included into trunk region, are 40 and 20 cm, respectively, whereas those of the average Korean male trunk are 32.6 and 22.6 cm, respectively. The extracted arms were designed as the cylinders with the diameter of the average Korean male arms, 8.6 cm. Arm bones were included at the center of the extracted arms. The resulting width including the trunk and the arms was 50.6 cm. Skeleton of the Korean computational phantom was also redesigned to match the modified anatomy. The position and thickness of ribs were especially changed from those of the ORNL adult computational phantom as the trunk width was decreased by 7.4 cm. The total volume of the resulting skeletal structure of the Korean computational phantom was 5866 cm3. The total mass, which was calculated from the bone volume by multiplying the bone mixture density 1.4 g/cm3 was 8.2 kg, was close to the Asian reference bone mass 8.4 kg. In case that Korean-specific organ data are not available, the organs of the ORNL adult computational phantom were identically implemented into the Korean computational phantom: adrenals, esophagus, small intestine, and large intestine. The thyroid of the Korean computational phantom was simplified as the two adjacent elliptical cylinders as the Monte Carlo code cannot express the fourth degree equation of the ORNL adult thyroid. The total volume of the thyroid was matched to that of the average Korean adult thyroid. For the gall bladder, the wall volume of the Korean computational phantom was matched to that of the average Korean data and the total volume of the wall plus the

262

Handbook of Anatomical Models for Radiation Dosimetry

contents was matched to that of the ORNL adult computational phantom. In the case of the other hollow organs such as the stomach, the heart, and the urinary bladder, the total volume of the wall and the contents of the Korean computational phantom were matched to that of the average Korean data. The volumes of the wall and the contents were separately scaled down with the same ratio of total volume. After modifying the dimension of the trunk, ribs, and organs surrounding lungs, the resulting maximum volume of the lungs was 2994 cm3. If the lung volume was expanded to the Korean reference volume, the lung was penetrated by the surrounding rib cage and organs. As an alternative approach, the original lung density of the ORNL adult computational phantom, 0.295 g/cm3 was adjusted to 0.375 g/cm3 to match the resulting lung mass to the average Korean lung mass 1123 g.15 This is the intrinsic limitation of the unrealistic shapes of lungs and ribs in the stylized approach, and matching the lung density might be the best one can do with the stylized manner that other investigators also employed.16 Figure 10.2 shows the 3D frontal and rear views of the internal organs and the skeletal structure of the Korean phantom where the skin and muscle were made transparent for better viewing the internal structure. The volumes of a total 13 organs are tabulated in Table 10.3 along with the data from the organ volumes of Korean average adult male and that of the ORNL adult phantom.

FIGURE 10.2 The 3D frontal (left) and rear (right) views of the internal organs and skeletal structure of the Korean stylized phantom. Skin and muscle were made transparent for better viewing the internal organs and skeleton. (From Park, S.H., Radiat. Prot. Dosim., 121(3), 257, 2006. With permission.)

10.2.3 Korean Voxel Computational Phantoms 10.2.3.1 KORMAN and KORWOMAN Although the Korean stylized computational phantom was developed and used for radiation protection calculation including internal and external dosimetry, more realistic anatomy computational phantoms were developed following the relatively simplified stylized computational phantom. The Korean adult male and female voxel computational phantoms were constructed by semiautomatically segmenting 28 individual critical organs and tissues from MR tomographic data. The MR tomographic images for use in construction of the voxel computational phantom were selected from a number of whole-body MR images of healthy adult males and females taken to determine reference organ sizes for Korean individuals. The selected male subject was a 30 year old healthy male of 170 cm in height and 68 kg in weight, which are well within the range of reference height and weight of male adult Korean: 170.2 cm and 68.2 kg, respectively.15 The female subject was a 35 year old healthy adult female whose height was 160 cm and weight 55 kg. The average height and weight of female adult Korean are 158 cm and 54.8 kg, respectively.15 The Siemens MagneTom vision 1.5T was used to obtain in the MR scans. Head, trunk, and lower abdomen at pelvis level were scanned with a 10 mm slice width, the neck was scanned at thyro id level, and legs and arms were scanned with a 2 mm slice width. All portions of the body were scanned in axial direction

Korean Computational Phantoms

263

TABLE 10.3 Organ Volumes (cm3) of the Korean Adult Male Stylized Model, Korean Reference Adult Male, and ORNL Adult Model Organs Brain Gall bladder wall Heart wall Contents Kidney Liver Lungb Pancreas Spleen Stomach wall Contents Testes Thymus Thyroid Urinary bladder wall Contents

KMIRD 1477.0 12.3 253.0 366.0 321.1 1357.0 2994.0 32.7 160.0 126.4 188.0 27.5 38.5 14.6 39.5 163.0

Korean Reference Data 1478.0 12.4 618.9a 321.6 1357.0 4320.0 32.0 160.2 314.4a 27.7 38.6 14.6 202.3a

ORNL Adult Model 1370.0 10.1 303.0 437.0 288.0 1830.0 3380.0 90.7 176.0 152.0 250.0 37.6 20.1 19.9 45.7 203.0

Source: Park, S.H., Radiat. Prot. Dosim., 121(3), 257, 2006. With permission. a The volume of wall and contents is only available. b Instead of lung volume, the lung mass was matched to the Korean reference value with the adjusted lung density since Korean reference lung volume was too large to be implemented into rib cage.

except legs and arms which were imaged coronally. Arms were positioned parallel to the body when trunk was imaged unlike most MR scans for diagnostic purposes. Because contiguous axial slices with same slice width should be prepared for wholebody voxel computational phantom construction, the 76 axial images from head to lower abdomen, except for neck region, were collected with a 10 mm slice width for male phantom. For the neck region, five slices with a 10 mm slice width were collected from 27 original images. The 89 axial images of the legs were reconstructed from coronal images by using the medical image viewer OSIRIS4 (Human Image Unit, Geneva, Switzerland). By combining these images, a total of 170 axial slices were prepared for male phantom development. As for female phantom, a total of 200 axial slices were obtained from original MR images and leg images reconstructed from coronal images. The original MR data in Digital Communications in Medicine (DICOM) format were converted into Joint Photographic coding Experts Group (JPEG) format image files consisting of a matrix of 512 × 512 pixels for segmentation and indexing process. The commercial image processing software, Photoshop (Adobe Systems, Inc., San Jose, CA) and graphic digitizer PL-400 (WACOM Co., Ltd, Japan) were utilized for segmentation. Photoshop was chosen due to the capability to manipulate the various image formats and independently define organ contours by using the built-in multi layer function. This function made it possible to separately segment several organs and tissues included in one slice of MR image and easily modify erroneously segmented contours, while overlapping other organs and tissues. As many organs and tissues were included in one slice,

264

Handbook of Anatomical Models for Radiation Dosimetry

many layers were needed for segmentation. Up to 14 layers were used for one slice at the abdominal level. Two kinds of segmentation methods, “manual drawing” on screen digitizer and “software-automated segmentation,” were utilized depending on the clearness of boundary between neighboring organs and tissues. Tag numbers were assigned to different organs and tissues by drawing the boundary of each contour and filling it with different gray scale color ranging from 0 to 255. Contours of the skin, lungs, brain, and leg bones were automatically segmented using a “magic-wand tool,” built in the software, when the boundaries between those organs and surroundings were clearly identified. Skin was described as one voxel layer outside the surface of the body, which resulted in a skin thickness of 2 mm, slightly thicker than the 1.3 mm reported for the thickness of epidermis and dermis in the Caucasian reference data.17 All skeletal components were assumed to be homogeneously distributed in the bone because the bone marrow was not identified in MR images. The bone surface could not be described with the voxel resolution of this study because the thickness of bone surface is estimated at 0.01 mm. In total, 28 individual organs and tissues were segmented and indexed. An experienced radiologist reviewed the results. The organs and tissues considered to be radiosensitive according to ICRP are sufficient to calculate the necessary radiation protection quantities with relatively faster calculation speed. The data array of 250 × 120 × 170 voxels requires the storage of 5.1 MB, where the single voxel size is 2 × 2 × 10 mm3 for male phantom. The female voxel phantom has 300 × 150 × 200 voxels where the voxel resolution is 1.7 × 1.7 × 8 mm3. Figure 10.3 shows the 3D reconstruction of the whole-body shapes of Korean Man (KORMAN) and Korean Woman (KORWOMAN), which are representing Korean adult male and female, respectively. Lower portion of the computational phantoms’ arms are absent, as shown in the figures because the original MR images did not contain the part of arms and hands. The 3D reconstruction of skeletal structure is somewhat poorer than that of CT image-based voxel computational phantoms currently available in the world because the skeleton is not as well identified in MR images as in CT images. The mass of the small bowel shows a considerable discrepancy in comparison with ICRP Reference data, which was difficult to distinguish from surrounding tissues, because it was segmented partly according to anatomical knowledge. Table 10.4 presents a comparison of KORMAN internal organ masses with those of ICRP Reference Man for 17 important organs and tissues.18 The data (a) KORMAN (b) KORWOMAN of KORMAN were calculated from the segmented volume data by multiplying by the above organ den- FIGURE 10.3 sities. Thirty-five percent of 17 organs and tissues in The 3D views of the Korean (a) male KORMAN and ICRP Reference Man showed an organ voxel phantom KORMAN and (b) female voxel phantom KORWOMAN mass difference of more than 20%. Among those (right) where skin and muscle were organs, the small bowel and the spleen of KORMAN made transparent for better viewing the are 42% and 31% lighter than those of ICRP Reference internal organs and skeleton.

Korean Computational Phantoms

265

TABLE 10.4 Organ Mass of KORMAN, ICRP Reference Man, and Standard Japanese Organs Adrenals Bladder (wall) Brain Colon (wall) Esophagus Heart Kidney Liver Lung Pancreas Skin Small bowel (wall) Spleen Stomach (wall) Testes Thymus Thyroid

KORMAN 13.7 64.8 1790.5 331.1 32.3 800.9 316.0 1579.9 986.2 122.2 2822.9 373.5 103.2 116.8 25.4 25 17.5

ICRP Reference Man Standard Japanese 14 50 1450 300 40 840 310 1800 1200 140 3300 650 150 150 35 25 20

14 40 – 330 40 360a 320 1600 1100 130 2400 590 140 140 37 33 19

Source: Lee, C. et al., Med. Phys., 31, 1017, 2004. With permission. a The volume of blood-free heart.

Man, respectively. The large discrepancies are mainly observed for organs in the digestive and urinary tracts (small bowel, bladder, and esophagus), which are highly variable even in a single person due to the time passed after having a meal.19 Organ masses of the Standard Japanese individual were also included for comparison.20 Even though the Korean and the Japanese belong to similar ethnic groups, there were some significant disagreements between the two data sets. 10.2.3.2 KTMAN-1 and KTMAN-2 To develop Korean reference voxel computational phantoms, two approaches for two different subject volunteers were employed. First, the MR image acquisition was performed for an adult volunteer representing Korean standard adult body dimension. The MR scanning was divided into three sessions to alleviate volunteer fatigue resulting from very long MRI scan times. The motion of the subject was also consequently minimized sufficiently to obtain high-quality MR images. Transversal MR images of the whole body including legs with a slice interval of 5 mm were obtained. Second, whole-body CT images were acquired from another adult male volunteer using a positron emission tomography (PET)/computed tomography (CT) machine with intravenous contrast agent, which resulted in high softtissue contrast. The CT modality also provided much better images of the skeletal structure than previous computational phantoms based on MR images. The two volunteers had heights and weights close to the average height and weight of Korean adult male which were estimated to be 170.2 ± 5.2 cm and 68.2 ± 8.6 kg, respectively. The height and weight of the recruited male subjects were within 1% and 5%, respectively, of the average values.

266

Handbook of Anatomical Models for Radiation Dosimetry

The first volunteer, a 25 year old adult male, was 172 cm in height and 65 kg in total body weight. Whole-body MR scans were performed using a Siemens MagneTom vision at 1.5T. A total of 537 slices of transverse and sagittal MR images were obtained across the entire anatomy of the subject. MR scanning was divided into three sessions to reduce volunteer fatigue and to avoid image blurring as much as possible. The total time required for the arrangement of the subject and image acquisition over the whole body was approximately 9 h with 2 h of actual image acquisition scan time. The arms were stretched upward since the bore of the MR scanner was too small for the subject to enter with arms at his side. Although sagittal slices of the arms were obtained, those images were not appropriate for this study. Therefore, the arm structure was not available in the resulting MR images. A total of 344 slices were chosen for voxel computational phantom construction with a 5 mm vertical slice thickness. The second volunteer, a 35 year old adult male, was 172 cm in height and weighed 68 kg. He was scanned using a PET/CT machine, a Siemens Somatom Emotion Duo system. PET/CT scanning service is a government-approved medical practice for those individuals wishing to have cancer screening with informed consent, and has been performed at the Radiation Health Research Institute of Korea since 2003. The volunteer was recruited for the voxel computational phantom development in conjunction with a cancer screening protocol. The volunteer was informed of the benefit and possible radiation risk involved during the procedure. Subsequent image review by staff radiologists indicated that the subject was cancer free. A whole-body CT scan as undertaken as part of the PET/CT examination, and the CT images were adopted for this study. The total time required for the patient arrangement and the image acquisition was ∼30 min. Since CT scanning was also intended to develop the Korean adult physical phantom, a slice interval of 1 mm was adopted during the CT scanning. A total of 1788 transverse slices across the whole body were obtained at the interval of 1 mm with 68 slices representing duplicate anatomic coverage. The final voxel computational phantom was then constructed using 344 slices selected at an interval of 5 mm. The construction of a corresponding physical computational phantom of this same individual is described elsewhere.21 The segmentation and indexing of the cross-sectional images require specialized anatomic knowledge. The authors made reference to an anatomical atlas to segment organs and tissues from the MR and CT source images.22 Finally, the segmented images were reviewed by an experienced radiologist on a slice-by-slice basis through comparison to the original MR or CT images. The authors adopted the commercial image processing software, Photoshop 7.0 (Adobe Systems, Inc., San Jose, CA) and screen digitizer PL-400 (WACOM Co., Ltd, Vancouver, WA) for image segmentation. Tomographic images, MR, and CT in DICOM format were converted into JPEG-formatted files using a preprocessing program, and then were imported to Photoshop 7.0 for segmentation. The software provides a multilayer tool that is useful to segment several organs and tissues from a single image slice. Contours for several organs and tissues in single slices were stored in different layers, all of which were still included in the original slice. Original CT or MR images were placed on the first layer, and organs and tissues were contoured and stored in other corresponding layers. In this manner, each single slice included both the original gray-scale image and the organ/tissue contoured image. The dimensions of the segmented contour were in agreement with the original images, as provided by the multilayer tool. The skeleton defined within the MR images was manually segmented via reference to human anatomical atlas. Other than leg bones, most of the bone sites were difficult to identify from the surrounding organs and tissues, and the full segmentation of skeleton in MR images took about half of the total development period. The skeleton defi ned

Korean Computational Phantoms

267

within the CT images was segmented using the thresholding method. All skeletal components were assumed to be composed of a homogenized mixture of bone mineral and red/yellow marrow as the trabecular microstructure was not identifiable in both the MR and CT images. The total bone was separately segmented into 19 skeletal sites according to its relative red bone marrow content as reported in ICRP Publication 70: the cranium, the mandible, the scapulae, the clavicle, the sternum, the cervical vertebra, the thoracic vertebra, the lumbar vertebra, the sacrum, the os coxae, the upper half femora, the lower half femora, the lower leg bones (the tibiae, the fibulae, and the patellae), the ankle and foot bones, the upper half humeri, the lower half humeri, the ulnae and radii, the wrist and hand bone, and the ribs.23 The skeleton of MR-based computational phantom was, as expected, not clearly visible. The ribs were especially difficult to recognize, and were manually segmented via reference to a human anatomical atlas. The lower half of the humeri, ulnae, radii, wrist, and hand bones were not available in MR-based computational phantom because the arms of the volunteer were stretched upward and out-of-field during the scan. Only a part of the upper half of the humeri was included within the MR-based computational phantom. The extrathoracic (ET) region surfaces were introduced as a radiosensitive tissue by ICRP in 1994 and are included within the remainder category of the effective dose as a replacement of the upper large intestine.24,25 The ET region is composed of the ET1 region (anterior nose) and ET2 region (posterior nose, mouth, pharynx, and larynx). The ET1 region was described as one voxel layer tissue adjacent to the air inside of the anterior nasal passages. The ET2 region was also segmented as a one-voxel layer of tissue adjacent to air inside of the posterior nasal passages and mouth, respectively. The pharynx and larynx were manually segmented with the aid of human anatomical atlas. Muscles in the CT images were segmented using the thresholding method. Muscles in the MR images were segmented using the region growing method by manually marking seed points inside of each muscle region. Adipose tissue in both CT and MR images was assumed to be the remaining regions exclusive of muscle, internal organs, skeleton, and skin. Skin was defined by automatically attaching one voxel layer to the outmost contour of the whole body. Although the voxel resolution of 2 × 2 × 5 cm3 would cause an overestimate of skin volume, it was assumed that the consequent dosimetric effect would be negligible. Wall and content were separately segmented for all hollow organs: the esophagus, the small intestine, the stomach, the colon, and the bladder. The lumen content of the alimentary organs was described as water and gas. Upper and lower large intestines were separated, and each intestine mass is listed for comparison with other data. Three components of kidney (pelvis, medulla, and cortex) were identified for the use in internal dosimetry. The three components were manually segmented and adjusted so that the volume fractions would match with those reported by ICRP Publication 89: pelvis (5%), medulla (25%), and cortex (70%).18 Three different segmentation methods were applied depending on the organ or tissue image contrast. The first method was “thresholding,” where relevant thresholds of gray scales were applied to organs and tissues having unique ranges of CT number, such as the skeletal tissues, muscle, lungs, and in-body air. This method automatically segmented those tissues within the original CT images by setting the minimum and maximum boundaries of the gray scale. In the case of MR images, it was not possible to apply this method due to the inherent limitation of MR to represent differences in relative tissues densities within the human body. The second segmentation method was “region-growing,” which was applied to the organs and tissues clearly separable from the surrounding tissues. Organ contours were automatically defined following manual selection of a starting seed point.

268

Handbook of Anatomical Models for Radiation Dosimetry

The third method was “manual-drawing,” which was applied to most other organs that were not visible or whose contents provided insufficient image contrast to apply the region growing method. Organs with very low image contrast were segmented by referring to human body atlases.22 Full segmentation of the MR and CT images took approximately 2 years. The segmentation of MR images took about 16 months, and that of CT images took about 8 months. A total of 29 organs and 19 bone sites were semiautomatically or manually segmented from 344 transversal images. Index numbers ranging from 0 to 255 were assigned to organs and tissues. The resulting data array of the MR-based phantom, named KTMAN-1(Korean Typical Man-1) was 300 × 150 × 344 voxels requiring storage of 14.7 MB, and the single voxel size was 2 × 2 × 5 mm3. The other phantom based on CT images, named KTMAN-2 (Korean Typical Man-2) had the same data array and single voxel size as given by KTMAN-1. Organ mass of KTMAN-1 and KTMAN-2 were calculated by multiplying the segmented volume by the reported reference tissue density as given in ICRU Report 46.26 The masses and densities of all organs and tissues are tabulated in Table 10.5. Frontal and back views of the skeleton and internal organs within KTMAN-1 and KTMAN-2 are shown in Figure 10.4a and b, respectively, along with representative organs and tissues. Skin, muscle, and adipose were made transparent for improved viewing of internal organs and the skeleton. 10.2.3.3 High-Deﬁnition Reference Korean (HDRK)-Man Following the MR- and CT-based voxel computational phantoms, a different approach using high-resolution color sectioned photos was taken to develop a more sophisticated voxel computational phantom. The high resolution voxel computational phantom, HDRK-Man, was constructed by using the color photographic images (Figure 10.5), which were acquired by the Visible Korean Human (VKH) Project performed by the researchers at Ajou University and Korea Institute of Science and Technology Information (KISTI) in Korea.27,28 The images were obtained from the cadaver of a 33 year old Korean male (164 cm in height and 55 kg in weight), who died of leukemia. The cadaver was fixed, embedded, and frozen in an immobilization box. The cadaver was then serially sectioned in the longitudinal direction at 0.2 mm interval with a cryomacrotome, and each sectioned surface was photographed with a digital camera Kodak DSC560, which provides the resolution of 3040 × 2008 pixels (Kodak, Rochester, NY). The images were saved as Tag Image File Format (TIFF) (24 bits color). A total of 8590 photographic images with the resolution of 0.1875 mm × 0.1875 mm were acquired from the cadaver. The color slice images with the high resolution made it possible to accurately segment the organs and tissues, especially small anatomical structures. The researchers also acquired CT and MR images, in addition to the color photographic images, at 1 mm interval from the same cadaver for possible future use. Major organs and tissues were segmented by professional anatomists, including the skeleton, the lung, the liver, the brain, the kidneys, the skin (external surface), the urinary bladder (external surface), the heart (internal surface), the gastrointestinal tract (internal surface), the main arteries (internal surface), and the respiratory tract (internal surface). The color anatomical images at every 2 mm interval were selected for computational phantom construction; that is, a total of 850 images out of 8590 images. The organs and tissues clearly distinguished by the color were segmented by an automatic process with Photoshop 7.0 (Abode Systems, Inc., San Jose, CA) and IDL 5.6. This automatic process was used for the eyes, lenses, skeleton, skin, muscle, colon, small intestine, ET region, red bone marrow, and gall bladder. Other organs that could not be automatically

Organ and Tissue

Circulatory system Urogenital system

Alimentary system

Respiratory system

ET1 ET2 Lungs Trachea Esophagus Salivary gland Small intestine Spleen Stomach Tongue Tonsil Upper large intestine Lower large intestine Heart (with blood) Bladder Kidney (cortex) Kidney (medulla) Kidney (pelvis) Prostate

1.03 1.03 0.296 1.1 1.03 1.03 1.03 1.06 1.03 1.05 1.05 1.03 1.03 1.06 1.04 1.05 1.05 1.05 1.03

Density (g/cm3)

Soft tissue

GI tract Muscle Muscle GI tract GI tract

Cartilage GI tract Soft tissue GI tract

Soft tissue Soft tissue

Surrogate

a

7.0 42.3 3,572.7 10.3 31.1 80.3 616.4 116.0 148.5 46.0 3.3 149.5 143.9 590.1 36.4 204.3 62.9 19.6 17.4

KTMAN-1 5.3 56.2 3,658.2 12.3 36.9 78.0 666.9 157.0 126.3 38.9 3.0 167.5 122.4 662.9 32.8 193.4 64.2 15.2 16.4

KTMAN-2

Volume (cm3)

7.3 43.6 1,057.5 11.3 32.0 82.7 634.9 123.0 153.0 48.3 3.5 154.0 148.2 625.5 37.9 214.6 66.0 20.6 17.9

KTMAN-1 5.5 57.9 1,082.8 13.5 38.0 80.3 686.9 166.4 130.0 40.8 3.1 172.5 126.1 702.7 34.1 203.1 67.4 15.9 16.9

16

1,200 9 40 82 590 140 140 67 4 180 150 780 40 320c

(continued)

17

840 50 310c

1,200 10 40 85 650 150 150 73 3 370b

KTMAN-2 Asian Ref. ICRP Ref.

Mass (g)

Mass of Organs and Tissues Segmented in KTMAN-1 and KTMAN-2, and Comparison with Those of Asian Reference Man and ICRP Reference Man

TABLE 10.5

Korean Computational Phantoms 269

Skeletal system

Organ and Tissue

522.6 411.5 173.0 726.1 462.2 699.3 770.5 501.0 126.3 – – – 653.4 6,528.0

1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4

Thoracic vertebrae Lumbar vertebrae Sacrum Os coxae Femora (upper half) Femora (lower half) Tibiae, fibulae, patellae Ankle and foot bones Humeri (upper half) Humeri (lower half) Ulnae and radii Wrist and hand bones Ribs Total skeleton

24.0 768.6 143.9 272.2 75.5 53.3 168.6

KTMAN-1

1.04 1.4 1.4 1.4 1.4 1.4 1.4

Surrogate

a

569.4 251.9 169.8 148.2 353.1 767.7 8,011.9

572.9 326.3 231.0 836.0 502.2 664.9 1,030.7

18.8 835.1 133.1 279.5 106.8 62.7 170.9

KTMAN-2

Volume (cm3)

Testes Cranium Mandible Scapulae Clavicles Sternum Cervical vertebrae

Density (g/cm3)

701.3 176.8 – – – 914.7 9,139.3

731.6 576.2 242.1 1,016.6 647.1 979.0 1,078.8

25.0 1,076.0 201.5 381.1 105.8 74.6 236.1

KTMAN-1

797.2 352.7 237.7 207.5 494.3 1,074.7 11,217.0

802.0 456.8 323.3 1,170.5 703.1 930.8 1,442.9

19.5 1,169.1 186.3 391.3 149.5 87.7 239.2

8,400

37

10,500

35

KTMAN-2 Asian Ref. ICRP Ref.

Mass (g)

Mass of Organs and Tissues Segmented in KTMAN-1 and KTMAN-2, and Comparison with Those of Asian Reference Man and ICRP Reference Man

TABLE 10.5 (continued)

270 Handbook of Anatomical Models for Radiation Dosimetry

Skin Adipose Adrenal Brain Eye Gall bladder Liver Muscle Pancreas Pituitary gland Thymus Thyroid

1.09 0.95 1.03 1.04 1 1.03 1.06 1.05 1.04 1.03 1.03 1.05 Soft tissue Soft tissue

Water Soft tissue

Soft tissue

3,295.2 11,999.0 10.4 1,195.8 16.5 44.2 1,458.2 22,046.6 55.4 0.7 21.6 13.4

2,963.6 16,459.1 10.3 1,350.1 16.0 29.0 1,194.4 27,552.4 66.0 0.8 26.2 12.5

3,591.8 11,399.0 10.7 1,243.7 16.5 45.6 1,545.7 23,149.0 57.6 0.7 22.2 14.1 54,381.7

3,230.3 15,636.2 10.6 1,404.1 16.0 29.9 1,266.0 28,930.0 68.7 0.8 27.0 13.1 66,275.2

2,400 11,000 14 1,470 15 58 1,600 25,000 130 0.54 30 19

3,300 12,020 14 1,450 15 68 1,800 29,000 140 0.6 25 20

Source: Lee, C. et al., Med. Phys., 33, 380, 2006. With permission. a Density of all organs and tissues was obtained from ICRU 46 except these organs, of which density was not available in ICRU 46. Density of the organs was assumed to be close to the surrogate organs tabulated. b ICRP 89 provides the mass of right colon, left colon, and rectosigmoid. The total mass was used for comparison with other values. c Only total mass of kidney, composed of cortex, medulla, and pelvis, was available in ICRP 89.

Total mass (g)

Integumentary system Additional organs

Korean Computational Phantoms 271

272

Handbook of Anatomical Models for Radiation Dosimetry

segmented were manually segmented on a screen digitizer CINTIQ 15X (WACOM Co., Ltd, Japan) and the “Magnetic Lasso” tool in Photoshop 7.0, which automatically marks the region of selection by identifying differences of color, significantly expedited the segmentation time. Thyroid, urinary bladder, prostate, salivary glands, adrenals, esophagus, spleen, stomach, lung, brain, liver, thymus, pancreas, gonads, kidneys, heart, oral mucosa, and blood were segmented manually. The skeleton was divided into nine bone sites, and the red bone marrow was segmented considering the distribution of red bone marrow in these bone sites.29 After completion of the segmentation, the resolution of the segmented images was reduced to 1.875 mm × 1.875 mm considering current computation speed and memory size, resulting in an intermediate voxel resolution (a) KTMAN-1 (b) KTMAN-2 of 1.875 × 1.875 × 2 mm3. The total body height of the voxel compu- FIGURE 10.4 tational phantom was adjusted by changing the (See color insert following page 524.) The 3D voxel resolution; that is, the height (164 cm) frontal views of the MR-based (a) KTMAN-1 was matched to the Reference Korean (171 cm) and CT-based (b) KTMAN-2 where skin and residual soft tissue was made semitransparent to by increasing the z-resolution from 2 to visualize internal organs and skeleton. 2.0854 mm. The total skeleton mass was not available in the Reference Korean data and, therefore, it was estimated by the method suggested by Clays et al.23: m = −10.7 + 0.119 × HT (kg)

(10.3)

where m is the skeleton mass in kg and HT is the height of a human body in cm. The height of the Reference Korean is 171 cm and the skeleton mass was calculated as 9.6 kg. The original skeleton mass 8.6 kg was then matched to 9.6 kg (including red bone marrow) by increasing the in-plane voxel size from 1.875 × 1.875 mm2 to 1.981 × 1.981 mm2. Consequently, the voxel resolution of the final computational phantom became 1.981 × 1.981 × 2.0854 mm3. The “Inner Grow” and “Outer Grow” functions in Photoshop 7.0 were used to adjust the dimensions of the individual organs to the Reference Korean data. The organs that were larger than the Reference Korean were decreased by erosion (Inner Grow), and the eroded regions were filled with adipose tissue. The smaller organs than those of the Reference Korean were increased by dilation (Outer Grow) by adding pixels onto the organ surfaces. Adipose tissue was assigned to the remaining volume other than organs FIGURE 10.5 and tissues, including all of the undefined organs Example of color photographic slice and tissues. On completion of the adjustment for the image utilized for model construction at body height, skeleton mass, and organ dimensions, the the middle level of liver.

Korean Computational Phantoms

273

weight of the computational phantom was 67.8 kg, which was less than the weight of the Reference Korean (68 kg) by 0.2 kg. The weight was then adjusted simply by adding 0.2 kg of adipose tissue on the lower parts of the legs. The average tissue compositions and densities in ICRU 46 were used to describe tissues except for the lung and skeleton.13 The lung density (0.296 g/cm3) and composition were taken from ICRP 23.30 The skeleton density (1.34 g/cm3) and composition were recalculated from the reference values in ICRP 70 and ICRP 23, respectively, by subtracting the elemental composition of red bone marrow from homogeneous skeleton composition as they were explicitly segmented in the phantom.13 Figure 10.6 shows the 3D views and major organs of HDRK-Man. The final voxel computational phantom is 171 cm in height and 68 kg in weight. The size of the voxels (voxel resolution) is 1.981 × 1.981 × 2.0854 mm3 and the voxel array size is 247 × 141 × 850 (29,602,950) in the x-, y-, and z-directions, which corresponds to a total array dimension of 489.307, 279.321, and 1,772.59 mm, respectively. The computational phantom includes a total of 30 organs and tissues that are required to calculate effective dose defined by ICRP. Table 10.6 shows the organ masses of HDRK-Man, with the reference values used for adjustment. The Reference Asian data were used for the prostate, the bladder, the adrenals, the colon, and the small intestines, for which Reference Korean data were not available.9 The organ masses are overall in good agreement with the Reference Korean data, within 7%, except for the following organs. The eye lenses were not adjusted due to the limitation of voxel resolution. The thickness of skin is known to be 1.47–2.45 mm but the skin was represented simply by one layer of voxels on the surface of the body.18 The adipose tissue of HDRK-Man is much heavier because not only the adipose tissue itself but also all the undefined organs were defined as adipose tissue. Eye Brain Oral mucosa Thyroid

Thymus

Salivary glands Esophagus Bone

Breast Heart

Lung

Liver Stomach

Spleen

Small intestine

Kidneys

Bladder Gonads

Colon

FIGURE 10.6 (See color insert following page 524.) The 3D whole-body frontal view with semitransparent skin (left) and major organs and tissues (right) of the HDRK-Man. (From Kim, C.-H., Phys. Med. Biol., 53, 4093, 2008. With permission.)

Handbook of Anatomical Models for Radiation Dosimetry

274

TABLE 10.6 Mass of 30 Organs and Tissues of the HDRK-Man and Reference Korean (the Difference between Two Data Was Included) Mass (g) Organ

HDRK-Man

Mass (g)

Reference Korean

Difference (%)

Bone

9,607

9,649

−0.4

Liver Lung Brain Kidneys Spleen Stomach

1,474 1,156 1,620 359 177 141

1,438 1,123 1,522 338 170 140

2.5 2.9 6.4 6.2 4.1 0.7

Pancreas Thymus Gonads

126 39 28

130 40 29

−3.1 −2.5 −3.4

Eyes Lens Muscle Bladder Colon

21 0.51 23,300 42 343

20 0.4 25,000 40a 330a

5.0 27.5 −6.8 5.0 3.9

Reference Korean

Difference (%)

Organ

HDRK-Man

Small intestine Esophagus Adrenals Skin ET region Thyroid Red bonemarrow Prostate Blood Salivary glands Gall bladder Oral mucosa Heart wall Breast Adipose tissue

602

590a

2.0

40 14 4,260 73 15 1,068

40 14a 2,400 – 15 1,000

0.0 0.0 77.5 – 0.0 6.8

12 254 87

12a – 82

0.0 – 6.1

13 21 391 23.3 23,400.2

13 – 380 22 11,000

0.0 – 2.9 5.9 112.7

Source: Kim, C.-H., Phys. Med. Biol., 53, 4093, 2008. With permission. a The Reference Asian data were used for the organs and tissues for which Reference Korean data are not available.

10.3 Conclusions and Future Work To some extent, mathematical formulations of organs and tissues of the human body have been widely used in the dosimetry calculation for nearly 60 years. Over this time, the accuracy of these computational phantoms has significantly increased. The advantages and disadvantages of the stylized and voxel computational anthropomorphic computational phantoms can be discussed from three points of view as follows. The fi rst point is the anatomical realism described by the computational phantoms. The voxel computational phantoms based on medical images of real individuals exceed the stylized computational phantoms. Several investigators have revealed the unrealistic anatomy of the stylized computational phantoms through a comparison study with the voxel computational phantoms for several irradiation conditions and radiation types.31–41 However, the anatomical limitations of the voxel computational phantoms also exist. The thickness of mucosa layers in gastrointestinal tract (less than 0.5 mm), which are actually radio sensitive, cannot be described by using the relatively low voxel resolution of the existing voxel computational phantoms, except VIP-man

Korean Computational Phantoms

275

whose voxel resolution is 0.33 × 0.33 × 1 mm3.42 However, the stylized computational phantoms can exactly express the thickness of the mucosa wall.16 Moreover, the cubeshaped voxel is reported to cause the voxel size effect, to which the surface overestimation is attributed.43,44 Alternately, the stylized computational phantoms have smooth organ surfaces free from the voxel effect. The second thing to be kept in mind is the need of reference computational phantoms. The international organizations concerning radiation protection have accepted the stylized computational phantoms as reference computational phantoms. Although the stylized computational phantoms are anatomically unrealistic, the volume, position, and shape of individual organs are standardized to reference data, since they are expressed by mathematical equations. The feature of the stylized computational phantom is finely tuned by adjusting individual parameters of the equations. Significant individual variations exist, however, among the voxel computational phantoms. Studies that analyze organ doses from several voxel computational phantoms have reported a large variability of individual organ doses, even among similar size adult computational phantoms.41,45 Even though the organ volume and external body dimensions are matched to reference data, in order to construct the reference voxel computational phantoms,29,46 the dosimetric response to internal and external radiation exposure is more dependent on the position and shape of organs rather than organ volume. In order to establish the real reference voxel computational phantoms, the position and shape of internal organs should be standardized as well as organ volume. Third, another obstacle to the voxel computational phantom development is that the high-resolution medical images, the crucial source for computational phantom development, are difficult to obtain whereas this is not the case with the stylized computational phantoms. In spite of the long scanning time to obtain high-resolution MR images, the motion artifacts and vague skeletal structure are inherently unavoidable in the resulting MR data. Conversely, CT images provide high resolution and clear contrast of organ contours by using a contrast agent. However, a significant radiation dose is transferred to the subject in whole-body CT scanning. To address these limitations of the stylized and voxel computational phantoms, a new approach to take only advantageous features from the two kinds of computational phantoms has been performed by some investigators. That is the hybrid computational phantom that combines the easy-to-standardize mathematical equations of the stylized computational phantoms and the anatomical realism of the voxel computational phantoms. Segars et al. have developed the hybrid computational phantoms by using the nonuniform rational B-splines (NURBS) technology and used for radiation imaging simulation.47,48 The internal organs and body contour are segmented from the medical images just like the manner used in voxel computational phantom development or converted from the existing voxel computational phantom data. Then, the computational phantoms composed of smooth NURBS surfaces or other surface-geometrybased methods are generated from the segmented contours. Several authors have been reporting the hybrid phantoms representing reference pediatric individuals as well as pregnant woman.49–51 Using this technology, one can realistically describe human anatomy and smooth organ surfaces without any voxel effect, and also easily transform the organ position and shape to match the reference data. The computational anthropomorphic computational phantoms that are close to real human anatomy and whose organ volume, shape, and position are matched to the reference data will be available in the future.

276

Handbook of Anatomical Models for Radiation Dosimetry

References 1. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381 V1-V7, Oak Ridge, TN: Oak Ridge National Laboratory, 1987. 2. Kramer, R. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods: Part I. The male (ADAM) and female (EVA) adult mathematical phantoms, GSF-Report S-885, Institut fuer Strahlenschutz, GSF-Forschungszentrum fuer Umwelt und Gesundheit, Neuherberg-Muenchen, 1982. 3. Stabin, M.G. et al. Mathematical models and specific absorbed fractions of photon energy in the nonpregnant adult female and at the end of each trimester of pregnancy, ORNL/TM-12907, Oak Ridge, TN: Oak Ridge National Laboratory, 1995. 4. Gibbs, S.J. et al. Patient risk from interproximal radiography, Oral Surgery Oral Medicine Oral Pathology Oral Radiology and Endodontics, 58, 347, 1984. 5. Zaidi, H. and Xu, X.G. Computational anthropomorphic models of the human anatomy: The path to realistic Monte Carlo modeling in radiological sciences, Annual Review of Biomedical Engineering, 9, 471, 2007. 6. Saito, K. et al. Construction of a computed tomographic phantom for a Japanese male adult and dose calculation system, Radiation and Environmental Biophysics, 40, 69, 2001. 7. Tanaka, G., Nakahara, Y., and Nakazima, Y. Japanese reference man 1988-IV. Studies on the weight and size of internal organs of Normal Japanese, Nippon Igaku Hoshasen Gakkai Zasshi, 49, 344, 1989. 8. Nagaoka, T. et al. Development of realistic high-resolution whole-body voxel models of Japanese adult males and females of average height and weight, and application of models to radio-frequency electromagnetic-field dosimetry, Physics in Medicine and Biology, 49, 1, 2004. 9. IAEA. Compilation of anatomical, physiological and metabolic characteristics for a Reference Asian Man, IAEA-TECDOC-1005, IAEA, Vienna, Austria, 1998. 10. Tanaka, G. et al. Reference man models for males and females of six age groups of Asian populations, Radiation Protection Dosimetry, 79, 383, 1998. 11. Kim, Y.J. et al. Studies on the Reference Korean and estimation of radiation exposure dose— Physical standard and estimation of inter-external radiation exposure dose, Journal of Korean Association of Radiation Protection, 7, 1, 1982. 12. Tek, H. and Kimia, B. Volumetric segmentation of medical images by three dimensional bubbles, Computer Vision and Image Understanding, 65, 246, 1997. 13. ICRU. Photon, electron, proton and neutron interaction data for body tissues, Report 46, International Commission on Radiation Units and Measurements, Bethesda, MD, 1992. 14. Eckerman, K.F., Cristy, M., and Ryman, J.C. The ORNL Mathematical Phantom Series, Oak Ridge, TN: Oak Ridge National Laboratory, 1996. 15. Park, S.H. et al. In vivo organ mass of Korean adults obtained from whole-body magnetic resonance data, Radiation Protection Dosimetry, 118, 275, 2006. 16. Han, E., Bolch, W., and Eckerman, K. Revisions to the ORNL series of adult and pediatric computational phantoms for use with the MIRD schema, Health Physics, 90, 337, 2006. 17. ICRP. Recommendations of the International Commission on Radiological Protection, Publication 26, Oxford: Pergamon, 1977. 18. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, Publication 89, Oxford: Pergamon Press, 2002. 19. Zankl, M. and Wittmann, A. The adult male voxel model “Golem” segmented from wholebody CT patient data, Radiation and Environmental Biophysics, 40, 153, 2001. 20. Tanaka, G.I., Kawamura, H., and Nakahara, Y. Reference Japanese Man.1. Mass of organs and other characteristics of normal Japanese, Health Physics, 36, 333, 1979. 21. Kim, J.I. et al. Physical phantom of typical Korean male for radiation protection purpose, Radiation Protection Dosimetry, 118, 131, 2006.

Korean Computational Phantoms

277

22. Pretterklieber, M.L. Pocket Atlas of Body CT Anatomy, 2nd edn., edited by W.W. Richard and M.B. Gotway, New York: Lippincott, Williams & Wilkins, 2002, 144 pp. ISBN 0-7817-3663-3; 29.55, European Journal of Radiology, 48, 327, 2003. 23. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: The Skeleton, ICRP Publication 70, Oxford: Pergamon Press, 1995. 24. ICRP. Dose Coefficients for Intake of Radionuclides by Workers: Replacement of ICRP Publication 61, ICRP Publication 68, Oxford: Pergamon Press, 1994. 25. ICRP. Human Respiratory Tract Model for Radiological Protection: A Report of a Task Group of the International Commission on Radiological Protection, ICRP Publication 66, Oxford: Pergamon Press, 1994. 26. International Commission on Radiation Units and Measurements. Photon, Electron, Proton and Neutron Interaction Data for Body Tissues, Bethesda, MD: International Commission on Radiation Unit and Measurement, 1992. 27. Park, J.S. et al. Visible Korean human: Its techniques and applications, Clinical Anatomy, 19, 216, 2006. 28. Stabin, M.G., Sparks, R.B., and Crowe, E. OLINDA/EXM: The second-generation personal computer software for internal dose assessment in nuclear medicine, Journal of Nuclear Medicine, 46, 1023, 2005. 29. Kramer, R. et al. All about MAX: A male adult voxel phantom for Monte Carlo calculations in radiation protection dosimetry, Physics and Medicine in Biology, 48, 1239, 2003. 30. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Oxford: Pergamon Press, 1975. 31. Bozkurt, A., Chao, T.C., and Xu, X.G. Fluence-to-dose conversion coefficients from monoenergetic neutrons below 20 MeV based on the VIP-man anatomical model, Physics and Medicine in Biology, 45, 3059, 2000. 32. Bozkurt, A. and Xu, X.G. Fluence-to-dose conversion coefficients for monoenergetic proton beams based on the VIP-Man anatomical model, Radiation Protection Dosimetry, 112, 219, 2004. 33. Chao, T.C., Bozkurt, A., and Xu, X.G. Conversion coefficients based on the VIP-Man anatomical model and EGS4-VLSI code for external monoenergetic photons from 10 keV to 10 MeV, Health Physics, 81, 163, 2001. 34. Jones, D.G. A realistic anthropomorphic phantom for calculating organ doses arising from external photon irradiation, Radiation Protection Dosimetry, 72, 21, 1997. 35. Jones, D.G. A realistic anthropomorphic phantom for calculating specific absorbed fractions of energy deposited from internal gamma emitters, Radiation Protection Dosimetry, 79, 411, 1998. 36. Kramer, R., Khoury, H.J., and Vieira, J.W. Comparison between effective doses for voxel-based and stylized exposure models from photon and electron irradiation, Physics and Medicine in Biology, 50, 5105, 2005. 37. Lee, C. and Lee, J. The dosimetric effect of unrealistic arm structure of stylized human model, Medical Physics, 32, 2100, 2005. 38. Lee, C., Lee, J., and Lee, C. Korean adult male voxel model KORMAN segmented from magnetic resonance images, Medical Physics, 31, 1017, 2004. 39. Staton, R.J. et al. A comparison of newborn stylized and tomographic models for dose assessment in paediatric radiology, Physics in Medicine and Biology, 48, 805, 2003. 40. Yoriyaz, H. et al. Absorbed fractions in a voxel-based phantom calculated with the MCNP-4B code, Medical Physics, 27, 1555, 2000. 41. Zankl, M. et al. Organ dose conversion coefficients for external photon irradiation of male and female voxel models, Physics and Medicine in Biology, 47, 2367, 2002. 42. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Physics, 78, 476, 2000. 43. Rajon, D.A. et al. Voxel effects within digital images of trabecular bone and their consequences on chord-length distribution measurements, Physics in Medicine and Biology, 47, 1741, 2002.

278

Handbook of Anatomical Models for Radiation Dosimetry

44. Rajon, D.A. et al. Surface area overestimation within three-dimensional digital images and its consequence for skeletal dosimetry, Medical Physics, 29, 682, 2002. 45. Lee, C. et al. Development of the two Korean adult tomographic computational phantoms for organ dosimetry, Medical Physics, 33, 380, 2006. 46. Kramer, R. et al. All about FAX: A female adult voXel phantom for Monte Carlo calculation in radiation protection dosimetry, Physics and Medicine in Biology, 49, 5203, 2004. 47. Segars, W.P., Lalush, D.S., and Tsui, B.M.W. A realistic spline-based dynamic heart phantom, IEEE Transactions on Nuclear Science, 46, 503, 1999. 48. Segars, W.P. et al. Development of a 4-D digital mouse phantom for molecular imaging research, Molecular Imaging and Biology, 6, 149, 2004. 49. Lee, C. et al. Hybrid computational phantoms of the male and female newborn patient: NURBSbased whole-body models, Physics and Medicine in Biology, 52, 3309, 2007. 50. Lee, C. et al. Hybrid computational phantoms of the 15-year male and female adolescent: Applications to CT organ dosimetry for patients of variable morphometry, Medical Physics, 35, 2366, 2008. 51. Xu, X.G. et al. A boundary-representation method for designing whole-body radiation dosimetry models: Pregnant females at the ends of three gestational periods—RPI-P3, -P6 and -P9, Physics and Medicine in Biology, 52, 7023, 2007. 52. Park, S.H., Lee, J.K., and Lee, C. Development of a Korean adult male computational phantom for internal dosimetry calculation, Radiation Protection Dosimetry, 121, 257, 2006. 53. Kim, C.H. et al. HDRK-man: A whole body voxel model based on high-resolution color slice images of a Korean adult made cadaver, Physics in Medicine and Biology, 53, 4093, 2008.

11 Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP Binquan Zhang, Jizeng Ma, Guozhi Zhang, Qian Liu, Rui Qiu, and Junli Li

CONTENTS 11.1 Introduction ............................................................................................................... 279 11.2 Construction of the Chinese Phantoms ................................................................. 280 11.2.1 The CNMAN Phantom ............................................................................... 280 11.2.2 The VCH Phantom ....................................................................................... 282 11.2.2.1 Specimen Selection and Evaluation.......................................... 282 11.2.2.2 Cadaver Sectioning and Image Acquisition ............................284 11.2.2.3 Image Registration and Segmentation .....................................284 11.2.2.4 3D Reconstruction ....................................................................... 286 11.2.2.5 Visualization ................................................................................ 287 11.2.2.6 Material Definition ...................................................................... 287 11.2.2.7 The VCH Computational Phantom .......................................... 288 11.2.3 The CVP Phantom ....................................................................................... 291 11.3 Applications of the Chinese Phantoms in Radiation Protection ........................ 293 11.3.1 The CNMAN Phantom ............................................................................... 293 11.3.1.1 Conversion Coefficients ............................................................. 293 11.3.1.2 Lung Counting ............................................................................ 294 11.3.1.3 Whole-Body Counting................................................................ 294 11.3.2 The VCH Phantom ....................................................................................... 295 11.3.3 The CVP Phantom ....................................................................................... 296 11.4 Conclusions and Future Works ...............................................................................300 References ............................................................................................................................. 301

11.1 Introduction Radiation protection dosimetric values calculated from the Caucasian computational phantoms may not be applicable to the Chinese population due to differences in anatomical, metabolic, and physiological parameters. The China Institute for Radiation Protection (CIRP) has been making great efforts to develop human phantoms using Chinese anatomical data for radiation protection. In the early of 1980s, a few physical whole-body phantoms that were similar to the bottle manikin absorption (BOMAB) phantoms1 in shapes were manufactured for calibrations of whole-body counting based on the average heights and 279

280

Handbook of Anatomical Models for Radiation Dosimetry

weights of Chinese adult males and females. In 2006, Mao et al. fabricated a torso physical phantom named CIRP-RTP-1.2 A male cadaver with 172 cm and 71 kg was selected as the basis of the phantom. The organs of the phantom including the lungs, liver, kidneys, heart, and rib cages were made from tissue equivalent materials and radionuclide was distributed in the lungs. This phantom is used for the calibration of lung counting systems involving radionuclide that emit low-energy photons. Other institutes also contributed to the developments of Chinese physical phantoms for radiation dosimetry. At the Sichuan University, a solid anthropomorphic phantom, which represented a Chinese adult male, was developed for experimental radiation dosimetry in radiotherapy, diagnostic imaging, and radiation protection.3 Using the Medical Internal Radiation Dose (MIRD) committee’s phantom4,5 and the anatomical data for Chinese reference man,6 a Chinese mathematical phantom was defined at the Tsinghua University for internal radiation dose calculations.7 In recent years, more than 30 voxel phantoms have been developed worldwide.8 Development of Chinese voxel phantoms (CVPs) benefited from the Chinese Digitized Human Project. In 2001, a scientific conference on the subject of “Key technology problems of Chinese digitized human body” was held in Fragrant Hill Hotel in Beijing.9 Experts in China who attended the conference suggested that the Chinese digitized virtual human body project should be supported as national technology development project. This suggestion and specific plans were adopted by the Chinese government immediately. During 2002–2003, several sets of the Chinese digitized virtual human data were acquired in the First Military Medical University10–16 and the Third Military Medical University.17–24 In 2006, the CIRP developed a Chinese adult male voxel phantom named Chinese Man (CNMAN) using one of the Chinese digitized virtual human datasets from the Third Military Medical University.25 In 2007, another CVP called the Visible Chinese Human (VCH) phantom was developed at the Huazhong University of Science and Technology using data from the First Military Medical University.26–28 In addition to these two cadaver-based phantoms, the Tsinghua University also developed the CVP from magnetic resonance imaging (MRI) of a young Chinese male.29,30 In this chapter, the construction and application of CNMAN, VCH, and CVP phantoms are introduced.

11.2 Construction of the Chinese Phantoms 11.2.1 The CNMAN Phantom The voxel phantom called CNMAN for radiation protection was created at the CIRP located in Taiyuan, Shanxi Province, China, using the color photographs of the first Chinese Visible Human (CVH) dataset.11,24 The CVH dataset was acquired by the CVH project research team in the Third Military Medical University, China. It was obtained from a 35-year-old Chinese male cadaver judged to represent normal human anatomy as much as possible.12,31 The man was 170 cm high and 65 kg in weight, which were close to those of the Chinese and Asian reference adult males.6,32 The cadaver was free of organic disease and lesions. After vascular perfusion and cadaver embedding, the cadaver was sectioned using a milling machine with a milling accuracy of 0.001 mm. Slices were acquired from head-to-toes in a low-temperature laboratory. The slices were spaced at 0.5 mm for head and neck region (0.1 mm at the skull base), 1.0 mm for other regions. The serial cross sections were photographed with a high-resolution digital camera. Each photograph was stored in TIFF format

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

(a) Head

(b) Chest

281

(c) Abdomen

FIGURE 11.1 Original transversal photograph images.

and had a resolution of 3072 × 2048 pixels, 24 bits per pixel. Figure 11.1 shows three original transversal photograph images of head, chest, and abdomen. A total of 1759 slices with 1 mm thick were used for the construction of CNMAN. In order to assign each voxel to the appropriate organ or tissue, organs or tissues were identified first. With the help of published atlases of human anatomy,33,34 tissues and organs in each slice were identified visually and segmented manually by enhancing their borders using the Adobe Photoshop™ (Adobe Systems Inc., San Jose, CA). These organs or tissues included the adrenals, aorta, bladder, brain, esophagus, eye lens, gallbladder, heart, kidneys, large intestine, liver, lungs, muscle, pancreas, prostate, small intestine, skeleton, skin, spleen, spinal cord, stomach, submandibular gland, testes, thymus, thyroid, trachea, and vitreous body. Color threshold segmentation was also performed in the RGB color space to obtain red bone marrow and yellow bone marrow automatically. Then the pixels that belonged to the same tissue or organ were filled with a specific color for visualization. The external pixel rows and columns outside the whole-body area were subtracted from the images to reduce pixel numbers, thus to reduce voxel numbers. Finally, each segmented image had 2790 × 1290 pixels (24 bits per pixel) and a file size about 10 MB. To reduce the image size, all the files were stored in the JPG format. Figure 11.2 shows the images before and after the image processing. For three-dimensional (3D) reconstruction of the voxel phantom, image registration was performed according to the four fiducial markers12 in each image. The pixel size

Heart Esophagus Stomach

Liver Skin

Yellow bone marrow Abdominal aorta Red bone marrow

Spleen

Lungs

Muscle Bone

Spinal cord

(a)

(b)

FIGURE 11.2 (a) Original transversal image; (b) the same slice after segmentation.

282

Handbook of Anatomical Models for Radiation Dosimetry

was derived from the length of the line, which was measured by the ruler in the images, divided by the corresponding total pixel number of this line. This resulted in an average pixel size of 0.16 mm with a standard deviation of 0.5%. The CNMAN phantom has a minimum voxel size of 0.16 mm × 0.16 mm × 1 mm and contains more than 6 billion voxels. Volumes of organs or tissues were calculated from the voxel size multiplied by their voxel numbers. The 3D view of CNMAN (Figure 11.3) was obtained by volume rendering using Visualization Toolkit (VTK).35 Each organ or tissue was then assigned a specific identification number to relate it to the appropriate physical property, such as composition and density, for Monte Carlo simulations. These physical data of organs or tissues were taken from ICRU Report 4436 and ICRP Publication 89.37 With these volumes and densities, the masses of each organ or tissue were calculated. Comparisons of the organ or tissue masses of CNMAN with those of the Chinese reference adult male6,32 and the ICRP reference adult male are given (b) in Table 11.1. Masses from VCH and CVP which will (a) be discussed in the following sections are also given FIGURE 11.3 in Table 11.1. (See color insert following page 524.) Almost half of the 23 organs and tissues in CNMAN 3D rendering of CNMAN: (a) whole body; and ICRP reference adult male show the mass differ- (b) skeleton. ences of more than 20%. Of the 13 tissues or organs of CNMAN, 9 had mass differences of 50% in comparison with the Chinese reference adult male. A large discrepancy is observed for the spleen. Since there is no internal lesion, the difference is due to individual anatomical variation. 11.2.2 The VCH Phantom The Britton Chance Center for Biomedical Photonics at the Huazhong University of Science and Technology (HUST) located in Wuhan, Hubei Province, China, constructed the VCH phantom for use in external photon and neutron dosimetry26,28 from one highquality VCH dataset which was obtained at the First Military Medical University (recently named the Southern China Medical University) located in Guangzhou, Guangdong Province, China. At the HUST, the original images were segmented, walled and smaller structures were revised, and skeletal marrow distributions were defi ned for external proton dosimetry.27 11.2.2.1 Specimen Selection and Evaluation A comprehensive evaluation system was developed to assess the anthropologic and genetic characteristics of the selected cadaver specimen, which was a 24-year-old man, 166 cm in height and 58 kg in weight. He displayed no specific physical disease in the medical history. The data of measurement on all required index for the VCH specimen showed an average physical stature of Chinese adult males.18

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

283

TABLE 11.1 Comparison of Organ Masses for CNMAN, VCH, CVP and Chinese, Asian, and ICRP Reference Data Masses (g) Organs or Tissues Adrenals Bladder Content Wall Bone Brain Bronchi Epididymes Eyeballs Fat Gallbladder Heart Content Wall Intestines Large Small Kidneys Larynx Liver Lungs Marrow, red Marrow, yellow Muscle Nasal vestibule Esophagus Oral mucosa Pancreas Pharynx Pituitary gland Prostate Salivary glands Skin Spinal cord Spleen Stomach Content Wall Teeth Testes Thymus Thyroid Tongue Trachea Urethra Total

CNMAN 14.6 227.9

8,395.3 1,443.6

17.4 19.1 503.9

2,179.9 1253 926.9 359.5 1,732.2 798.1 1,547.8 865.9 33,184 36.9 88.2

19.4 4,016.4 81.1 577.1 243.9

26.9 9.7 14.4

56,403.2

VCH 13.7 211.2 160.2 51 5,413.2 1,051.5 29.9 5.4 12.9 10,734 8.2 622.7 384.7 238 1,548.8

198 24.4 1,193.9 805.3 816.7 412.5 17,193.7 63.4 35.7 92.8 50.5 14.3 0.6 13.6 87.5 2,461.3 44.5 211.1 612.1 421.8 190.3 46 21.6 25.4 18.2 48.5 19.3 13.1 54,000

CVP 4.9 66.4

11,570.2 1,899.2

Chinese 14

8,000 1,460

Asian

ICRP

14

14

100 40 4,500 1,470 26

50 5,500 1,450

15 10,000 8

4 15 14,600 10

400 380 1,630

510 330 1,670

34,561.6

320 27 1,600 1,200 1,000 1,300 25,000

310 28 1,800 1,200 1,170 2,480 29,000

35.3

40

40

130

140

0.54 12 82 2,400 30 140 380 240 140 45 37 30 19 67 9 9 60,000

0.6 17 85 3,300

19.9 21.9 952.5

325

524.2

290

1,964.5 1,007.5

1,410 1,250

118.3

120

0.9 45.5 10,681.6

2,400

410.3 226.4

165

110.7

40 30 26

9

60,000

150 400 250 150 50 35 25 20 73 10 10 73,000

Handbook of Anatomical Models for Radiation Dosimetry

284

11.2.2.2 Cadaver Sectioning and Image Acquisition The cadaver specimen underwent whole-body computed tomography (CT) and MRI scans before sectioning. The cadaver was fixed and frozen afterwards in the standing posture for serial sectioning at 0.2 mm vertical intervals on a milling machine. This process generated a total of 8920 slices from the top of the head to the bottom of the feet (exclusive of those in void space above the head or under the feet). Digital color images were obtained by photographing the top surface of every successive slice at the 5440 × 4080 pixel resolution, corresponding to a voxel size of 0.1 mm × 0.1 mm × 0.2 mm. The CT, MRI, and photograph images of a sample slice at brain level are show in Figure 11.4. Improvement in the data collecting work on the VCH has brought considerable advantages for the subsequent procedures. Compared with the American Visible Human Project (VHP), the upstanding (instead of lying down) posture of the VCH cadaver specimen presents more realistic internal organ structures under the influence of gravity. Blood vessel perfusion procedures, where the perfusate was made of gelatin, vermilion, and starch, were promptly implemented for preservation and clear displaying of the whole-body vascular system. Extrinsic marker sites, which provided spatial reference points for image alignment and registration, were artificially added into the embedding materials. Moreover, the cadaver was continuously sectioned without any cutting to separate the body into smaller blocks as was done in the VHP. 11.2.2.3 Image Registration and Segmentation To preclude any geometrical distortion that may be caused by the instability of the milling machine or camera, all transversal photographs were first aligned and registered using a global method with respect to the markers in each slice. Segmentation work was then performed on the 24 bit color photographs. Unlike the CT and MRI image sets, the techniques here were less dependent on pixel gray values. The boundaries between different organs and tissues displayed in cross-sectional images were manually distinguished and identified using the Adobe® Photoshop® software under the direction of experienced anatomists. Local tissues and suborgans, such as the bone marrow, were semiautomatically segmented with the help of computerized RGB thresholds. The dividing value (threshold) for a designed training parameter, which characterizes the proportion of subquantity in the red channel with respect to the summation in all three component channels

(a)

(b)

(c)

FIGURE 11.4 (a) CT, (b) MRI, and (c) color images of a sample slice at brain level of the VCH phantom.

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

285

(red, green, and blue) under RGB color space, was carefully recalibrated in later work and set to 43.1% for machine recognition of the red marrow. It was more complicated to threshold the yellow marrow, of which the color is mainly derived from fat cells and was less distinguished under the device-dependent environment of color representation as RGB space. Conversely, the International Commission on Illumination (Commission Internationale d’Eclairage, CIE) 1976 (L* a* b*) color space—of which the three coordinates represent the lightness of color, its position between magenta and green, and its position between yellow and blue, respectively—better approximates the human vision and is much broader in color gamut. We employed it to segment the yellow bone marrow and the original images were converted to the CIE 1931 XYZ color space in advance, which is the master space for both RGB and L* a* b*, and predicts the relationship between spectral power distribution and color perception. The rationale for the color space transformation used in this study was based on CIE standards and the accuracy of image conversion was verified using the Adobe Photoshop software afterwards. Taking into account the fact that the wavelength of light is associated with a visual sensation of yellow when perceived by the naked eye, it is reasonable to adopt the hue (H) and saturation (S), two basic properties of color that could also be calculated from the RGB space, into the filtering criteria for pixels rich in yellow bone marrow. The entire process was completed by an Insight Segmentation and Registration Toolkit (ITK)38 based C++ program. Careful calibrations suggest that a judging condition ranging from 10 to 30 for H, from 0.28 to 0.41 for S, and from 18 to 27 for the b* dimension, is effective for the determination of the yellow marrow on the VCH image set. Following this approach, the yellow marrow was mostly recognized in the lower part of the skeleton while the red marrow had a comparatively uniform distribution. Distributions of the red and yellow marrow computed from the sample skeletal region are illustrated in Figure 11.5. Every segmented part was labeled with a certain identification number specific to an organ or tissue. Instead of directly replacing the pixel values by the ID numbers, the procedure in VCH was distinctive in several ways. Segmentation results were first recorded by monochrome images where the target organ or tissue was highlighted in white while the rest of the image slice was filled with black, as shown in Figure 11.6. Resulting image groups corresponding to individual structures were uniformly categorized and named

(a)

(b)

FIGURE 11.5 (See color insert following page 524.) Illustration of the computed distribution of the red and yellow bone marrows within a sample color image of the VCH phantom: (a) before segmentation; (b) after segmentation.

Handbook of Anatomical Models for Radiation Dosimetry

286

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

(k)

(l)

FIGURE 11.6 Original photograph image, segmented organs (muscle, artery, vein, bone, stomach, lung, liver, spleen, spine, and spinal cord) and recombined view of a sample slice at abdomen level of the VCH phantom.

under the same geometric system of the whole body so that the information could be accurately relocalized. The entire process of organ labeling and image segmentation took about 8 months to finish. 11.2.2.4 3D Reconstruction We grouped voxels corresponding to individual segmented organs and tissues to construct the whole computational phantom. According to the sectioning accuracy and slice photographing precision of the VCH phantom, the cubic resolution should be 0.1 × 0.1 × 0.2 mm3, and the whole-body block would yield more than 100 billion voxels, which was a

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

287

challenge for currently available computers and Monte Carlo codes to handle. Therefore, we adopted two data reducing approaches. Pixel picking, the first step, adjusted the voxel resolution by controlling the density of the sampling work. A C++ program was developed to extract voxels at a fixed interval. Every 20th voxel along both anterior–posterior (AP) and left–right directions in every 10th slice was chosen to approximately represent the surrounding region of 2 × 2 mm2. Consequently, the elementary volume was altered to 2 × 2 × 2 mm3 per voxel. For the second step, large areas of margin in each slice were ignored. We eliminated voxels beyond the central rectangle of 460 × 240 mm2 in each slice, since they were located far from the body and contributed nothing to anatomy. Ultimately, there were 230 voxels from the left to right, 120 voxels from back to front, and 892 voxels from the top to bottom. Considering the VCH segmentation results, we framed a proper sequence when integrating voxels of different parts into the whole phantom so that inside or local ones would not be overlaid and eliminated. The red and yellow marrow, for instance, was supposed to be included after the bones. Kidneys were resolved after their adipose capsule. The spatial location of each voxel was derived from the planar coordinates of the sampled point and the height of the corresponding slice. The whole body was immersed in air medium and faced the +y-direction with the long axis lying along the +z-direction under the righthanded coordinate system. 11.2.2.5 Visualization Visualization is an important approach to ensure the accuracy of both image segmentation and 3D reconstruction. The VCH phantom was first visualized at the original voxel resolution in formats of VTK35 and virtual reality modeling language (VRML). Considering the amount of image sets was extremely large, we carried out the visualization work on a high-performance computing cluster. The 3D views of the internal organs, whole-body skeleton, and vascular system of the VCH phantom are shown in Figure 11.7. Posttreatments, such as surface smoothening have been applied to volume delineating and 3D rendering. The detailed structures of the head, kidney, and liver are displayed in Figure 11.8. Comparatively, sectional viewing of the phantom are usually chosen to illustrate the effect and precision of the whole-body lattice-based computational phantom. The coronal and sagittal views of the VCH phantom at the fi nal voxel resolution of 2 × 2 × 2 mm3 are depicted in Figure 11.9. The visual plane in Figure 11.9a is selected around the center on the sagittal axis of the body. With respect to the posture and coronal thickness of the VCH specimen, Figure 11.9b is synthesized from three different planes for continuous presentation of the whole-body information in one picture: two coronal planes for the upper and lower parts with an inclined connective plane of 14° at the abdomen level. 11.2.2.6 Material Deﬁnition We based organ-specific elemental mass fractions and tissue densities used in the dosimetric phantom on ICRU Report 4436 and ICRP Publication 89.37 Organ masses were calculated by multiplying the accumulated voxel numbers by the corresponding reference density. For purposes of better reflecting the situation within the final computational phantom, we calculate the organ masses of the VCH phantom at the reduced resolution after voxelization other than pixel resolution in original images and might differ much for small organs.

Handbook of Anatomical Models for Radiation Dosimetry

288

(a)

(b)

(c)

FIGURE 11.7 (See color insert following page 524.) (a) Internal organs, (b) whole-body skeleton, and (c) vascular system of the VCH phantom.

(a)

(b)

(c)

FIGURE 11.8 The detailed 3D views of the (a) head, (b) kidney, and (c) liver of the VCH phantom.

11.2.2.7 The VCH Computational Phantom Table 11.1 shows information about the major organs incorporated in the VCH phantom with mass comparison against those of ICRP Publication 8937 reference adult male, and International Atomic Energy Agency (IAEA) reference Asian adult male.32 The VCH

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

289

computational phantom has a reconstructed whole-body mass of approximately 54 kg. Details of the voxel-based anatomical phantom are described as follows.

(a)

(b)

FIGURE 11.9 (a) Sagittal and (b) coronal views of the VCH computational phantom at 2 mm voxel resolution.

11.2.2.7.1 Alimentary System According to ICRP Publication 89, the paired parotid, submandibular, and sublingual glands were described together as salivary glands. The cavities within the ears, mouth, and nose were uniformly defined as air. In addition to estimated anatomical variations between individuals, the mass discrepancy for the stomach of the VCH phantom in comparison with the reference value should also be ascribed to a couple of factors: the inherent inaccuracy in representation of the 0.3–1.5 mm varying thickness of wall and mucous layers at a constant 2 mm voxel resolution, the volumetric fluctuation under different degrees of filling, and the influence from adjacent structures in specimen preparation and the standing posture other than supine as in normal autopsies, both of which largely decide the shape of the stomach. The total masses of the intestines show excellent agreement, although neither suborgans, including the small intestine, ascending colon, transverse colon, descending colon, sigmoid colon and rectum, nor intestinal subregions, including the wall, mucosa and content, were further discriminated within the computational phantom. The considerably smaller pancreas of the VCH phantom was explained as a common phenomenon among the Chinese population.18

11.2.2.7.2 Circulatory System The heart wall—inclusive of the pericardium, endocardium, and septa—was segmented based on the contour identified in the earlier phantom, while the blood content within the four chambers was separated from the heart and added to the whole-body vascular system. Owing to the vessel perfusion treatment on the VCH specimen, major arteries and veins were able to distinguish themselves in color on cryosectional color images and were therefore capable of segmentation. There are approximately 676 mL of arterial blood and 758 mL of veinal blood presented in the computational phantom at 2 mm voxel resolution. It should be noted that the large amount of blood contained in capillaries and individual organs with the exception of the lungs and heart is not accounted here. 11.2.2.7.3 Integumentary System We acquired the skin of the VCH phantom by an erosion algorithm on the boundaries of the body in each slice image, with an average inward shrinking distance equaling one-voxel thickness. We developed an in-house C++ program was developed to finish this work. 11.2.2.7.4 Respiratory System We segmented the mucous membrane found around the palate, gingivae, and dorsum of the tongue was segmented into oral mucosa. We also separated the trachea and a pair of bronchi of the VCH phantom, which had been treated as one structure in earlier studies

290

Handbook of Anatomical Models for Radiation Dosimetry

and evaluated for 19.3 and 29.9 g, respectively. The overestimate of the trachea mass resulted from the description of rings by one-voxel thickness as well as the discrepancy in mass proportion between cartilage and soft tissue in comparison with the reference value. We defined the content of lumen in all respiratory organs as air. We did not fully exclude in segmentation, however, the lumen within bronchi since it was difficult to identify. We segmented the epithelium and part of cartilage along the nasal passage as the nasal vestibule. We also calculated the mass density used for the lungs in the VCH phantom on the basis of the mass ratio from lungs tissue to pulmonary blood and their corresponding specific gravities. 11.2.2.7.5 Skeletal System The bone tissue in the VCH phantom was obtained by original manual segmentation, while the distribution of marrow was automatically computed from the occupied inner cavity based on chromatic thresholding in color images as described above. An overestimate of the red bone marrow could be observed in the earlier VCH phantom and no attempt was made to identify the yellow marrow by then. Following the approach described in Section 11.2.2.3, we were able to compute and update the whole-body distribution of the red and yellow bone marrow. Although the 412.5 g total mass of the yellow marrow within the VCH phantom at 2 mm voxel resolution was significantly lower than the reference value, the definition and differentiation of the two types of marrow should be considered satisfactory, since a large amount of the inactive marrow fails to appear in distinctive yellow color. 11.2.2.7.6 Urogenital System The pair of kidneys segmented in the VCH phantom was exclusive of the pelvis. Their outer covering capsules, which are composed of connective soft tissue, were defined together with the adipose tissue in dosimetry calculations. The volume of the bladder may differ substantially due to the large variation in the amount of stored urine and was estimated to be approximately 154 mL in the VCH phantom. The segmented urethra was comprised of 0.5 g membranous layer and 12.6 g spongy tissue. The bulbourethral gland with a mass of 0.4 g was specified independent of the urethra. The testes and epididymes, which were considered as one structure in the earlier phantom, were separated and reevaluated to be 21.6 and 5.4 g, respectively. There were 11.6 g of penis sponginess, 5.1 g of seminal vesicle, 3.8 g of spermaducts, and 2.6 g of ureters represented under the 2 mm voxel resolution in the urogenital system, although the estimates of masses for these small organs might contain a large uncertainty. 11.2.2.7.7 Additional Organs and Tissues The brain used in the VCH phantom incorporated the brain stem (31.4 g), cerebellum (90.6 g), cerebrum gray matter (646.4 g), and white matter (283.1 g). The remaining space within the cranial cavity was uniformly defined as the cerebrospinal fluid (CSF). The spinal cord was manually segmented other than computational recognition within the skeletal region. The adipose tissue situated in the whole-body subcutaneous area was segmented as the fat, which amounts to more than 10 kg in total. The masses of the adrenals and eyeballs were reevaluated. Part of the adrenal tissue that was falsely segmented into the renal adipose capsule previously has been corrected. There were 2.6 g lachrymal tissues segmented around the eyes as well. Unidentified area all across the body was substituted by soft tissues in dosimetry calculations and had a collective mass of ∼8.4 kg.

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

291

11.2.3 The CVP Phantom A Chinese voxel computational phantom named CVP was developed from MRI images for dosimetry calculations for radiation dosimetry by the Department of Engineering Physics in Tsinghua University and the Institute of Space Medicine and Medical Engineering, both located in Beijing, China.29 The whole-body MR images were scanned from a young Chinese healthy male volunteer. His height was 170 cm and his weight was 70 kg. The original MR images had a resolution of 256 × 256 pixels. Each pixel has 12 bits of gray tone resolution. Transversal MRI images were obtained at 2 mm intervals above the head and neck and at 4 mm intervals below the neck in order to derive a more precise description of the head and neck. Finally a total of 565 original MRI slices were obtained. The voxel size is 1 mm × 1 mm × 2 mm above the head and neck and 2 mm × 2 mm × 4 mm below the neck. After the image registration by the multiquadric method and the segmentation-counting algorithm,19 a set of images (880 slices with 256 × 256 pixel in one slice) was derived. The volume of each newly created voxel is 2 mm × 2 mm × 2 mm. Due to the diversity and complexity of organs and the fact that many organs share close gray-scale values in the images, it was difficult to obtain satisfactory results by using automatic segmentation. In this study, manual contouring was used as the segmentation method. According to the contours, the organs could be distinguished and reconstructed. There is a large statistical redundancy in the contours obtained this way and some of the dots on the contours are not needed in the surface reconstruction. Consequently, an algorithm of sampling contour point was adopted to effectively reduce statistical redundancy, greatly speeding up the display and preserving the real shape of the organs. Eventually a total of 24 organs or tissues were identified and defined. These organs include the adrenals, bladder, esophagus, gallbladder, stomach, heart, kidneys, intestine, liver, lungs, pancreas, prostate, skeletal components, skin, spleen, stomach, testes, thyroid, etc. The bone surface which is estimated to be 1–10 μm in size could not be described with the voxel resolution of this study.39 Some organs such as the thymus were not segmented. The small and large intestines were not separately delineated. Figure 11.10 shows the skeleton and some of the organs of the CVP phantom. In this study, the following procedures were carried out to reduce the size of the image and increase the calculation speed in Monte Carlo simulations: (1) some voxels at the edge of each slice that were full of air in the original image were deleted and the final image matrix becomes 256 × 156 × 880. (2) The voxel size was reduced to 4 mm × 4 mm × 4 mm with the pixel picking method, to yield the final dimension of the image matrix of 128 × 78 × 440. For Monte Carlo simulations, organs or tissues of interest have to be related to appropriate physical properties. The average tissue compositions and densities recommended in ICRP Publication 2340 were used to tag each voxel in the CVP phantom for comparison with the most existing dosimetry datasets. However, it should be noticed that ICRP Publication 8937 revised lots of information of these data compared to ICRP Publication 23. Calculations using ICRP Publication 89 could also be carried out following the same procedures. Table 11.1 lists the organ or tissue masses of CVP and also the comparisons of these masses to other phantoms. Each organ or tissue mass of CVP was calculated from its volume and the density recommended in ICRP Publication 23. It is noted that 14 of 23 organs or tissues in CVP and ICRP 89 reference adult male showed the organ mass difference of more than 20%. Significant discrepancies are observed in some organs such as the testes, skin, and spleen. Generally three potential reasons could contribute to this

Handbook of Anatomical Models for Radiation Dosimetry

292

(a)

(b)

(c)

FIGURE 11.10 3D anterior views of the CVP model: (a) outermost surface; (b) skeleton; (c) some organs.

difference: (1) the discrepancy between the individual and the reference adult male; (2) the manual segmentation of the organs based on MR images is partly dependent on the anatomical knowledge; (3) the mass change due to the amplification of the voxel size. Particularly the big difference of red bone marrow could be caused by the difficulty to identify the tissue in MR images. As for the skin, the thickness of epidermis and dermis in the Caucasian reference data was reported as 1.3 mm.40 The 4 mm voxel size has likely increased the mass of the skin significantly. The mass of the spleen of CVP is twice as much of both the ICRP 89 reference adult male and the Chinese reference adult male, but closer to the one of CNMAN. Therefore it should be due to individual anatomical variation. Since the calculated conversion coefficients using the CVP phantom will be compared with those calculated using VCH28 and Korean Man (KORMAN),41 the comparison between the masses of the organs in the three human voxel phantoms are necessary. Considering the height and the total weight of these phantoms, the CVP (170 cm and 70 kg) is closer to KORMAN (170 cm and 68 kg) compared to VCH (1.66 m and 54 kg). The difference of the masses of the organs also follows a similar trend. The organ masses are more similar between CVP and KORMAN than VCH for the brain, lung, liver, and pancreas. However, the mass of adrenal of CVP is almost the same as that of the VCH, but is about twice smaller than that of the KORMAN.

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

293

11.3 Applications of the Chinese Phantoms in Radiation Protection 11.3.1 The CNMAN Phantom The Monte Carlo transport code MCNP was used to calculate doses in the CNMAN for radiation protection purposes. As the CNMAN contains more than 6 billions voxels, which is over the limit of MCNP code, the resolution of CNMAN was reduced during calculation. The pixel sizes used for the head, torso, and leg were reduced to 1.6 mm × 1.6 mm × 1 mm, 1.6 mm × 1.6 mm × 2 mm, and 1.6 mm × 1.6 mm × 5 mm, respectively. Figure 11.11 shows the two-dimensional (2D) views of the voxel phantom created in MCNP. Based on the constructed voxel phantom, more study will be performed in the future, such as calculating conversion coefficients for external dosimetry, specific absorbed fractions (SAFs) for internal dosimetry, and calibrations of in vivo measurements. 11.3.1.1 Conversion Coefﬁcients

(a)

Conversion coefficients link the operational and protection quantities to the physical quantities such as fluence and air kerma free-in-air. The organ-absorbed doses per unit air kerma free-in-air have been calculated for the CNMAN. Comparisons of some organ doses with those from VIP-MAN and ICRP 74 are shown in Figure 11.12 (E = 100 keV).

(b)

FIGURE 11.11 Saggital and coronal views of the CNMAN phantom after its implementation in MCNP code. 2.5

ICRP CNMAN VIP-MAN

2

Gy/Gy

1.5

1

0.5

us

ns

ym Th

le

d Ey e

yr

oi

h Th

om

ac

in St

Sk

gu s

ng

op ha Es

Lu

er Li v

es st Te

Bl

ad

de

r

0

FIGURE 11.12 Comparison of organ-absorbed doses per unit air kerma free-in-air, DT/Ka of CNMAN, VIP-MAN, and ICRP74 for a broad parallel photon beam.

294

Handbook of Anatomical Models for Radiation Dosimetry

11.3.1.2 Lung Counting The lung counting system at CIRP was simulated with CNMAN and the counting efficiency of CNMAN was compared to that of Japan Atomic Energy Research Institute (JAERI) torso phantom. It was found that the counting efficiencies of these two phantom differed more than 20% at the photon energy below 50 keV. At higher energy, the differences of the counting efficiencies became smaller. Higher lung counting efficiencies could be achieved by placing the detectors properly so as to get a low uncertainty of measurement result and to save measurement time. Ma et al.42 optimized the detectors’ positions by calculating the photon flux maps (Figure 11.13) on the body surface using CNMAN and the Monte Carlo N-Particle eXtended (MCNPX) code. The detectors were placed at the positions with highest flux to get the higher counting efficiencies. 11.3.1.3 Whole-Body Counting The CNMAN phantom could also be used to calibrate the whole-body counting system. The dependency of whole-body counting efficiency on worker anatomy has been studied with CNMAN and other Caucasian phantoms.43 Results (Figure 11.14) showed that the whole-body counting efficiencies of CNMAN are about 10% smaller than those of the reference male BOMAB phantom and the normalized man (NORMAN) phantom, though CNMAN is similar to the NORMAN phantom and the reference male BOMAB phantom in weights and heights. This is most likely due to differences in the body proportion (i.e., height versus width) and the relative position of the detector to whole-body phantom.

2.0 × 10–4 6.0 × 10–4 1.0 × 10–4 1.4 × 10–4 1.8 × 10–4 2.2 × 10–4 2.6 × 10–4 3.0 × 10–4 3.4 × 10–4 3.8 × 10–4

FIGURE 11.13 Flux on the surface of CNMAN (photons with energy of 17.5 keV is emitted from the lungs).

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

295

0.002 NORMAN CNMAN

Efficiency (counts per photon emitted)

0.0018 0.0016

BOMAB

0.0014 0.0012 0.001 0.0008 0.0006 0.0004 0.0002 0

0

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 Photon energy (keV)

FIGURE 11.14 Whole-body counting efficiencies of CNMAN phantom, BOMBAB phantom, and the Norman phantom.

11.3.2 The VCH Phantom We adopted the MCNPX,44 a well-benchmarked multiparticle Monte Carlo code package from the Los Alamos National Laboratory (LANL) that represents a major extension of the Monte Carlo N-Particle (MCNP) code, since it inherits the routines of arbitrary 3D geometry and user configurable output options in MCNP, uses new variance reduction techniques, and absorbs the theoretical models from the Los Alamos high-energy transport (LAHET) code system to track all particles at nearly all energies. In the configuration of the standardized external radiation geometry, parallel source beams that directed perpendicularly to the longitudinal axis of the body were set for AP, posterior–anterior (PA), left-lateral (LLAT), and right-lateral (RLAT) geometries, respectively. A full 360° rotation of beams around the longitudinal axis of the fixed body was set for the rotational (ROT) geometry. Source beams projected with no preference in direction, where the particle flux per unit solid angle remained constant, were configured for the isotropic (ISO) geometry. Photons, neutrons, and protons have been considered for dose calculations with the VCH computational phantom. Data derived from the MIRD-type simplified phantom family have been summarized by a joint task group of ICRP and ICRU, and further have been documented in ICRP Publication 7439 and ICRU Report 57.45 Monoenergetic photons ranging from 0.015 to 10 MeV, neutrons ranging from 10−9 to 104 MeV, and protons ranging from 20 to 104 MeV were investigated, respectively. In lower energies, interactions are mostly determined on the basis of cross-sectional data. In case of high-energy irradiations, theoretical models are employed to determine the particle interactions. The new F6 heat detectors were used to tally the cumulative energy deposition delivered by all transported particles for all regions (organ or tissue) of interest. For neutron and proton dosimetry, in particular, there could be a rich set of secondary particles, including protons, photons, electrons, neutrons, deuterons, tritons, 3He ions, and alpha particles, especially at high

296

Handbook of Anatomical Models for Radiation Dosimetry

incident energies. The photon and neutron dosimetry simulations were operated on three IBM workstations, each equipped with a 3.0 GHz Intel processor and 4.0 GB random access memory (RAM), using the Microsoft Windows XP platform. The proton dosimetry simulations were operated on a high-performance Inspur™ sever that was equipped with four Intel Xeon™ Tulsa 7120 dual-core 3.0 GHz processors and 32.0 GB RAM in the Microsoft Windows™ Server 2003 system. A total of 30 million particle histories for AP, PA, LLAT, RLAT geometries and 300 million histories for ROT geometry were calculated in all three types of radiation. The number of particle history simulated for ISO geometry was 40, 60, and 35 million in photon, neutron, and proton dosimetry, respectively. The overall level of statistical uncertainty was quite satisfactory. Special dosimetric quantities proposed for radiological protection, including the absorbed dose, equivalent dose, and effective dose as developed by ICRP and ICRU, have been extensively studied in the context of rapid progress in nuclear engineering, aerospace industry, and clinical radiology. The dosimetry system, which was first conceived by ICRP in Publication 2646 and was revised later in ICRP Publication 60,47 has been further clarified by ICRP 2007 recommendations,48 in which the radiation and tissue weighting factors for the quantities equivalent and effective dose determinations, were updated according to the latest available scientific information on judgment of health effects attributable to radiation exposure. Special considerations or assumptions need to be made in the calculation of the effective dose on the VCH phantom. We specified a fat layer in the frontal chest level as the breast, and represented the gonads by the testes. We used the mean dose absorbed by the intestines for both the colon and the small intestine. We approximated the absorbed dose to bone (exclusive of marrow) for the bone surface. We treated the bone tissue all across the body as one target, although the different subsites have been distinguished. Since there are no lymphocytes segmented in the VCH phantom, the muscle dose was used as a surrogate for their similarity on spatial distribution. The arithmetic mean value from the nasal vestibule, larynx, and pharynx was calculated in place of the extrathoracic (ET) airways. There would be no sex averaging on the effective dose determination of the VCH phantom and the dose results should only be considered as person-specific values for this representative Chinese adult male. In order to relate the dosimetric results with external radiation fields for practical use in radiological protection, dose quantities need further normalization by source fluence to be expressed in terms of fluence to absorbed dose conversion coefficients. 11.3.3 The CVP Phantom The CVP phantom was implemented into MCNP in dual PIV–CPU machines with 512 MB RAM under Linux operating system. A set of conversion coefficients from kerma free-in-air to organ-absorbed dose are presented for external monoenergetic photon beams from 10 keV to 10 MeV based on a whole-body exposure. Six standard irradiation geometries defined in ICRP/ICRU reports were simulated: AP, PA, LLAT, RLAT, ROT, and ISO geometries (ICRP 199639; ICRU 199845). The photon energy ranged from 10 keV to 10 MeV. The calculated conversion coefficients using the CVP phantom are compared with those calculated using VCH28 and KORMAN41 for all the organs and conditions. This comparison of the conversion coefficients at photon energies of 0.05, 0.1, 2, and 10 MeV were carried out under AP, PA, and LAT irradiations, respectively, and shown in Figures 11.15 through 11.17. The ratios calculated for RLAT and LLAT geometries were averaged and depicted as LAT geometry.

FIGURE 11.15 Comparison of the organ dose conversion coefficients from CVP, VCH, and KORMAN in AP geometry at photon energies of 0.05, 0.1, 2, and 10 MeV.

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP 297

FIGURE 11.16 Comparison of the organ dose conversion coefficients from CVP, VCH, and KORMAN in PA geometry at photon energies of 0.05, 0.1, 2, and 10 MeV.

298 Handbook of Anatomical Models for Radiation Dosimetry

FIGURE 11.17 Comparison of the organ dose conversion coefficients from CVP, VCH, and KORMAN in LAT geometry at photon energies of 0.05, 0.1, 2, and 10 MeV.

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP 299

300

Handbook of Anatomical Models for Radiation Dosimetry

Generally, the organ dose conversion coefficients of these three phantoms show their relatively large differences at low-photon energies (0.05 and 0.1 MeV) but agree very well at high-photon energies (2 and 10 MeV) in the three geometries (AP, PA, and LAT). The discrepancies decrease with the increasing photon energy because those photon beams have higher penetrating power. For example, the difference of the dose conversion coefficient for liver from CVP and KORMAN reaches to 13.3% at photon energy of 0.05 MeV and decreases to 0.4% at photon energy of 10 MeV. However, with photon energy decreased, the depth of an organ within the human body becomes an increasingly crucial parameter strongly influencing the small organ dose conversion coefficients. Considering the difference of the data from the three phantoms, the ones from CVP and from KORMAN are closer. More than half of the organ doses from VCH are higher than those from CVP and KORMAN. Probably this is due to their different construction method and the small stature and weight of VCH. The differences between the data from VCH and those from CVP or KORMAN are larger in PA geometry than those in AP or LAT geometry. This may be explained if noticing that most of the organs are located toward the front of the body, the organ dose conversion coefficients for those organs are more sensitive to the trunk thickness. Comparing the organ dose conversion coefficients from CVP and VCH, the discrepancies are lower than 10% at 10 MeV. More than half of the organ dose conversion coefficients are lower than 10% in AP and LAT geometry even at low-photon energy. However, large discrepancies over 40% were observed for bone, esophagus, and bladder wall. As has mentioned, the discrepancy gets the largest in PA geometry. The data for all the listed organs except skin show a large discrepancy over 35% in PA geometry at the photon energy of 0.05 MeV, and two are over 50% such as 63% for esophagus and 59% for bladder wall. In the comparison of the data from CVP and KORMAN, it was found that they agree well for most of the organs. The discrepancies are lower than 20% at low-photon energy and lower than 10% at higher-photon energy. However, large discrepancies, 25%–34% in the three geometries at the low-photon energy of 0.05 MeV, were also observed for bladder wall. And large discrepancies of 33%, 46%, and 50% exists in the data for stomach in AP, PA, and LAT geometry, respectively, at low-photon energy. It is worth noting that particularly the positions of organs of the alimentary tract strongly depend on the individual person’s position, and the composition and time of the last meal. It could explain the large difference of the data for stomach. As for the bladder wall, its position strongly depends on the time after going to the toilet.

11.4 Conclusions and Future Works Constructions of the three Chinese adult male voxel phantoms, CNMAN, VCH, and CVP, have been introduced in this chapter. CNMAN and VCH phantoms were based on the anatomical data from the Chinese Digitized Human Project while MRI of a healthy volunteer was used in CVP. These three phantoms have been implemented into Monte Carlo code MCNP(X) for different dose calculations. The applications of these three phantoms such as external dose calculations and in vivo measurements have also been discussed in this chapter. Several limitations and disadvantages of the voxel-based modeling approach are discussed from the experience obtained from their developments and implementations. Each phantom

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

301

in essence remains an individualized phantom since it was developed from a single real person. Intended structural bridging will be necessary to satisfy the need for the application of radiation dosimetry on more individuals of different human groups and is particularly more important to cadaver-based phantoms than those constructed from CT, MRI, and other kinds of medical tomographic images. The curve surfaces of both internal organs and exterior layer of the body are not well approximate by cubic voxels. The voxel-based approach largely depends on the quality of image sets and has difficulties in coping with complex details in 3D space. For instance, the intestines of the VCH phantom have not been further segmented since the abdomen of the cadaveric specimen was distorted and even damaged by unexpected external stress or some other improper treatments in sample preparation. Appropriate manual edit was necessary to recover the original geometric state of entire gastrointestinal tract. All the three voxel phantoms above are Chinese adult male phantoms and their body weights and heights are close to the Chinese reference data. Large uncertainties might be introduced when the simulation data from these phantoms are adopted directly for other ethnic groups. Phantoms for males at different ages, body sizes, and also for Chinese females should be considered in the future.

References 1. American National Standards Institute (ANSI). Specifications for the Bottle Manikin Absorption Phantom, ANSI/HPS N13.35, ANSI, New York, 1999. 2. Mao, Y., Ma, R.W., and Fan, Y.G. Development of realistic torso phantom for lung counter, The Second Asian and Oceanic Congress for Radiological Protection (AOCRP-2), Beijing, October 2006. 3. ICRU. Phantoms and computational models in therapy, diagnosis and protection, ICRU Report 48, ICRU, Bethesda, MD, 1992. 4. Snyder, W.S. et al. Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, MIRD Pamphlet No. 5, S3, Journal of Nuclear Medicine, Society of Nuclear Medicine, New York, 1969. 5. Snyder, W.S., Ford, M.R., and Warner, G.G. Estimates of absorbed fractions for monoenergetic photon source uniformly distributed in various organs of a heterogeneous phantom, Medical Internal Radiation Dose (MIRD) Committee Pamphlet No.5 (revised), Society of Nuclear Medicine, New York, 1978. 6. Wang, J.X. et al. Data of Anatomical Physiological and Metabolic Characteristics for Chinese Reference Man, Atomic Energy Press, Beijing, 1998. 7. Wang, H.Y. et al. The internal radiation dose calculations based on Chinese mathematical phantom, Nuclear Electronics and Detection Technology, 26, 915, 2006. 8. Zaidi, H. and Xu, X.G. Computational anthropomorphic models of the human anatomy: The path to realistic Monte Carlo modeling in radiological sciences, Annual Review of Biomedical Engineering, 9, 471, 2007. 9. Li, Z.H. Science and technology issues of digital virtual human body in China—Summary of Xiangshan Science Conference No.174, China Basic Science, 35, 2002. 10. Chinese Visible Human. http://www.chinesevisiblehuman.com/. 11. Zhang, S.X. et al. Number one of Chinese digitized visible human completed, Acta Academiae Medicinae Militaris Tertiae, 24, 1231, 2002. 12. Zhang, S.X. et al. Creation of the Chinese visible human data set, Anatomical Record B New Anatomy, 275, 190, 2003.

302

Handbook of Anatomical Models for Radiation Dosimetry

13. Zhang, S.X. et al. The third Chinese visible human dataset produced, Acta Academiae Medicinae Militaris Tertiae, 25, 1332, 2003. 14. Zhang, S.X. et al. Dataset of first Chinese visible human female completed, Acta Academiae Medicinae Militaris Tertiae, 25, 371, 2003. 15. Zhang, S.X., Heng, P.A., and Liu, Z.J. Chinese visible human project: Dataset acquisition and its primary applications, Conference Proceedings of the IEEE Engineering in Medicine and Biology Society, 4, 4168, 2005. 16. Zhang, S.X., Heng, P.A., and Liu, Z.J. Chinese visible human project, Clinical Anatomy, 19, 204, 2006. 17. Tang, L. et al. Data collecting technology on Virtual Chinese Human, Chinese Journal of Clinical Anatomy, 20, 324–326, 2002. 18. Yuan, L. et al. Selecting specimen for digitized virtual Chinese human, Chinese Journal of Clinical Anatomy, 20, 334, 2002. 19. Zhang, Y., Jiang, G.P., and Li, S.X. Elastic registration of medical images through multiquadric method, Journal of First Military Medical University, 584, 2002. 20. Yuan, L. et al. The study of key technology of virtual Chinese human, Acta Anatomica Sinica, 34, 225, 2003. 21. Yuan, L. et al. Construction of dataset for Virtual Chinese Male No. 1, Journal of First Military Medical University, 23, 520–523, 2003. 22. Zhong, S.Z. et al. Research report of experiment database establishment of digitized virtual Chinese No. 1 female, Journal of First Military Medical University, 23, 196, 2003. 23. Tang, L. et al. Chinese Digital Human girl No. 1 dataset report, Chinese Journal of Clinical Anatomy, 22, 98, 2004. 24. Zhang, S.X. et al. The Chinese Visible Human (CVH) datasets incorporate technical and imaging advances on earlier digital humans, Journal of Anatomy, 204, 165, 2004. 25. Zhang, B.Q. et al. CNMAN: A Chinese adult male voxel phantom constructed from color photographs of a visible anatomical data set, Radiation Protection Dosimetry, 124, 130, 2007. 26. Zhang, G.Z., Liu, Q., and Luo, Q.M. Monte Carlo simulations for external neutron dosimetry based on the visible Chinese human phantom, Physics in Medicine and Biology, 52, 7367, 2007. 27. Zhang, G.Z. et al. Organ dose calculations by Monte Carlo modeling of the updated VCH adult male phantom against idealized external proton exposure, Physics in Medicine and Biology, 3697, 2008. 28. Zhang, G.Z. et al. The development and application of the Visible Chinese Human model for Monte Carlo dose calculations, Health Physics, 94, 118, 2008. 29. Zeng, Z. et al. Dose assessment for space radiation using a proton differential dose spectrum, Journal of Tsinghua University (Science and Technology), 46, 374, 2006. 30. Li, J.L. et al. Organ dose conversion coefficients for external photon irradiation using the Chinese voxel phantom (CVP), accepted by Radiation Protection Dosimetry, 2009. 31. Zhang, S.X., Heng, P.A., and Liu, Z.J. Chinese visible human project, Clinical Anatomy, 19, 204, 2006. 32. IAEA. Compilation of anatomical, physiological and metabolic characteristics for a reference Asian man, IAEA-TECDOC-1005, IAEA, Vienna, 1998. 33. Spitzer, V.M. and Whitlock, D.G. Atlas of the Visible Human Male, Jones and Bartlett Publishers, Sudbury, MA, 1998. 34. Zhang, S.X., Heng, P.A., and Liu, Z.J. Atlas of Chinese Visible Human (Male and Female), Science Press, Beijing, 2004. 35. VTK. http://www.vtk.org. 36. ICRU. Tissues substitutes in radiation dosimetry and measurement, ICRU Report 44, ICRU Report 44, ICRU, Bethesda, MD, 1992. 37. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, 2003. 38. itk. http://www.itk.org/.

Chinese Voxel Computational Phantoms: CNMAN, VCH, and CVP

303

39. ICRP. Conversion Coefficients for Use in Radiological Protection against External Radiation, ICRP Publication 74, Pergamon Press, Oxford, 1996. 40. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Pergamon Press, Oxford, 1975. 41. Lee, C., Lee, J., and Lee, C. Korean adult male voxel model KORMAN segmented from magnetic resonance images, Medical Physics, 31, 1017, 2004. 42. Ma, J.Z. et al. Determination of the detectors’ positions and efficiencies of the in vivo measurement system using human voxel phantom, Proceedings of National Workshop on Individual Dose Monitoring, Beijing, China, May 12–15, 2008. 43. Zhang, B.Q., Mille, M., and Xu, X.G. An analysis of dependency of counting efficiency on worker anatomy for in vivo measurements: Whole-body counting, Physics in Medicine and Biology, 53, 3463, 2008. 44. Pelowitz, D.B. MCNPX User’s Manual Version 2.5.0, Los Alamos National Laboratory Report LA-CP-05-0369, Los Alamos, NM, 2005. 45. ICRU. Conversion coefficients for use in radiological protection against external radiation, ICRU Report 57, ICRU, Bethesda, MD, 1998. 46. ICRP. Recommendations of the International Commission on Radiological Protection, ICRP Publication 26, Pergamon Press, Oxford, 1977. 47. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Pergamon Press, Oxford, 1991. 48. ICRP. The 2007 Recommendations of the International Commission on Radiological Protection. ICRP publication 103, Annual ICRP, 37, 1, 2007.

12 Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series Representing 3, 6, and 9 Month Gestational Periods X. George Xu, Chengyu Shi, Michael G. Stabin, and Valery Taranenko

CONTENTS 12.1 Introduction ...............................................................................................................305 12.2 Early Development at Oak Ridge National Laboratory (ORNL) Prior to 1995 ...............................................................................................................306 12.3 Developments by Others Since 1995.......................................................................308 12.4 Development at Rensselaer Polytechnic Institute on a CT-Based Pregnant Female Computational Phantom in 2004 ..............................................................309 12.5 Development at Rensselaer Polytechnic Institute on the RPI-P3, RPI-P6, and RPI-P9 Phantoms ................................................................................. 315 12.5.1 Materials and Methods ............................................................................... 316 12.5.1.1 Selection of Reference Anatomical Data ................................... 317 12.5.1.2 The Fetus .......................................................................................317 12.5.1.3 The Uterus and Placenta .............................................................317 12.5.1.4 The Breasts ....................................................................................318 12.5.1.5 The Rest of the Body ....................................................................318 12.5.1.6 Voxelization ...................................................................................322 12.5.1.7 Tissue Compositions ....................................................................323 12.5.1.8 Monte Carlo Modeling ................................................................ 323 12.6 Results and Discussions ........................................................................................... 326 12.6.1 Organs Volumes vs. Reference Data ......................................................... 326 12.6.2 Visual Inspection of the Anatomy............................................................. 330 12.6.3 Analysis of Organ Centroids...................................................................... 330 12.7 Conclusion.................................................................................................................. 332 Acknowledgments ............................................................................................................... 333 References ............................................................................................................................. 333

12.1 Introduction The protection of a pregnant female and her embryo/fetus against ionizing and nonionizing radiation is of particular interest in health and medical physics because of the radiosensitivity associated with fetal development.1 Both intentional and unintentional exposures 305

306

Handbook of Anatomical Models for Radiation Dosimetry

of childbearing mothers to radiation have been reported for medical procedures in nuclear medicine, diagnostic radiology, and radiation therapy.1 In one incidence, the pregnancy of a patient was not known until the CT images had been obtained.2,3 In other cases of medical exposures, the fetus is exposed with the assumption that there is a greater benefit to the mother in need of clinical care. A pregnant worker in a nuclear facility under the license of U.S. Nuclear Regulatory Commission is subjected to special protective measures.4 Occupational exposures to pregnant workers, however, have been reported.5 Survivors of the atomic bombing in Japan have included pregnant individuals whose dosimetry data were used in epidemiological studies of radiation risk.6,7 Potential radiation risk adds to the dangers a pregnant traveler can experience aboard an aircraft.8 For nonionizing radiation, radiofrequency-based devices that are usually placed in the vicinity of the body can cause pregnant women and their fetuses to be exposed to hazardous energy sources such as electromagnetic waves, induction heating cookers, handheld metal detectors, as well as personal cellular phones.9–14 The principal adverse biological effects of ionizing radiation on the mammalian embryo and fetus include death, malformations, growth impairment, mental retardation, induction of malignancies, and hereditary defects.1,15,16 The frequency and magnitude of these effects differ according to the absorbed dose, the type of radiation, and the gestational age at which the exposure occurs.16,17 Streffer18 discussed the effects of in utero irradiation, and considered the possibility for tissue weighting factors specific to fetus. The biological effects of nonionizing radiation on fetuses, on the other hand, are less known in comparison with those of ionizing radiation. Currently, interest in studying nonionizing radiation effects is steadily increasing due to heightened uses of various nonionizing radiation emitting devices. Radiation protection of pregnant patients and her embryo/fetus has been a particularly important issue to the radiation oncology community. The incidence rate of cancer in pregnant women is increasing because of the trend of delaying pregnancy into later reproductive years and the success in early detection of common cancers associated with pregnancy, including cervical cancer, breast cancer, Hodgkin’s disease, melanoma, lymphoma, and leukemia.19,20 According to the American Association of Physicists in Medicine’s Task Group Report No. 36, 3500–4000 new cancers are diagnosed in pregnant women each year in the United States.17 The decision to administer a radiation treatment must be justified by weighing the benefit to the mother to the potential risk to the embryo/fetus from exposed radiation during the treatment of the mother.1 A radiation treatment plan for the pregnant patient is required to optimize the radiation dose to the treatment target in the mother, and to substantially limit the radiation dose to other organs and the embryo/fetus. In this chapter, we first review several computational phantoms of pregnant women and compare various geometry modeling methods including the phantoms by Stabin et al.21 and the CT pregnant woman by Shi and Xu.2,3 We then discuss in more detail a set of new phantoms, called RPI-P series, developed by Xu et al. at Rensselaer Polytechnic Institute using advanced surface-geometry modeling tools.22

12.2 Early Development at Oak Ridge National Laboratory (ORNL) Prior to 1995 The history of anatomical computational phantoms for radiation dosimetry goes back to when the stylized anatomical computational phantoms were first conceived nearly 40 years ago at ORNL for the Medical Internal Radiation Dose (MIRD) Committee of

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

307

the Society of Nuclear Medicine.23–25 Quadric equations were used to defi ne the organs using descriptive and schematic anatomy references. The strategy of this approach made the geometry simple enough for the computers available at that time to handle. The average reference anatomical values defi ned the so-called Reference Man.26 Researchers recognized that, although the anatomical variation among individuals could be significant, the computers at that time would not be able to manage anything more complex. Later improvements at ORNL led to a family of stylized computational phantoms representing both sexes at various ages that became widely adopted by the radiation dosimetry community.27 Adapting the computational phantom that represents both a 15 year old male and an adult female, Stabin et al.21 added the fetus and placenta to create stylized computational phantoms of a pregnant female at the end of each trimester of pregnancy, the 3, 6, and 9 month computational phantoms, for various internal nuclear medicine applications. The greatest changes to the adult female phantom involved the growth of the uterus and the existence of a compartment representing uterine contents. Stabin et al. used a computational phantom for the uterus similar to that described by Cloutier et al.28 Figure 12.1 shows images of the uterine computational phantoms in the Stabin et al. pregnant female phantom series. The 3 month uterus was represented by a right circular cone with a hemispherical cap. The axis of the cone was oriented in the Y-direction, formed at a 33° incline to the horizontal. After the third month of pregnancy, and until about the seventh month, researchers assumed the main area of growth in the uterus occurred along its long axis with a small increase in breadth. In the 6 month computational phantom, the uterus is modeled as a cylinder capped at both end by hemispheres. The long axis of the cylinder ran in the Y-direction, tilted upward at an angle of 40° from the horizontal. In the 9 month computational phantom, the uterus was modeled as a cylinder capped at both ends by hemispheres, as in the 6 month computational phantom. The uterus was modeled to have extended in length as well as breadth, and the upper hemisphere was considerably larger than the lower hemisphere, and larger than in the 6 month computational phantom. The two hemispherical sections were connected by a section of a cone whose long axis of the cylinder ran in the Y-direction, and was tilted upward at an angle of 40° from the horizontal. The uterine contents of the 3 month computational phantom were modeled as a homogeneous mixture of soft tissue. No attempt was made at that time to model the presence of skeletal material. The uterine contents were thus used as a target to represent the fetus in this phantom. In the 6 and 9 month computational phantoms, the fetus, the placenta,

Body

z΄

Body

Surface z΄

Uterine wall

Surface

Body Surface

Placenta z΄

Uterine wall Uterine contents

Placenta Uterine wall

y΄

Other uterine tissue

y΄ Fetal skeleton

Fetal skeleton Fetal soft tissue Other uterine tissue

56.9°

Fetal soft tissue y΄

50° 50°

Body

(a)

Surface

Body

z

x=0

y

(b)

Surface

Body Surface

z

z y

x=0

(c)

x=0

y

FIGURE 12.1 Images of the uterine models in the Stabin et al. pregnant female phantom series [26]: (a) 3 month model; (b) 6 month model; (c) 9 month model.

308

Handbook of Anatomical Models for Radiation Dosimetry

and the amniotic fluid were explicitly modeled as separate regions. The fetal skeleton was also explicitly modeled in these phantoms and given the composition of the skeleton of the newborn phantom of Cristy and Eckerman.27 The placenta was also modeled as a hemispherical shell, in the 6 and 9 month computational phantoms. The whole-body trunk was extended in the 6 and 9 month computational phantoms to accommodate the enlarged uterus, and a few of the abdominal organs were redesigned or moved. Urinary bladder and small intestine (SI) were significantly remodeled to model their significant displacement and, in the case of bladder, decreased in volume. Masses and centroids of all of the organs were provided, and SAFs for all source/target combinations were provided and were implemented in the MIRDOSE (ref) and OLINDA/EXM computer codes for use in dose calculations.

12.3 Developments by Others Since 1995 Chen29 extended the stylized pregnant female computational phantoms into four pregnancy periods, 8 weeks, 3, 6, and 9 months, for external ionizing radiation dosimetry. Kainz et al.13 developed a semi heterogeneous pregnant female torso computational phantom based on laser-scanned body surface data and analytical equations. Dosimetry calculations involved induced current densities and specific absorption rates (SAR) for pregnant women exposed to handheld metal detectors.12 Using a similar approach, Kawai et al.14 reported a stylized abdominal computational phantom of a pregnant female and calculated the SAR for normalmode helical antennas (NHAs). Using medical images, voxel or tomographic computational phantoms gradually emerged in the past two decades.30 Derived from whole-body CT and MR images (and in some cases cadaver cross-sectional photographs), these computational phantoms are anatomically realistic, although the images had to be processed to assign organ information to each of the image pixels—a laborious process known as “segmentation and labeling.” By the late 1990s, the image-processing and Monte Carlo calculations of these voxel computational phantoms, which consisted of a repeated lattice structure, were no longer hindered by the large number of voxels, because most personal computers had enough processing power. Although nearly 30 voxelized computational phantoms of adult, teenager, and child have been developed around the world, medical images are seldom available for pregnant individuals. X-rays-based procedures are normally not applied to pregnant patients. Due to a lack of whole-body images for pregnant females, Dimbylow10,11 adopted a “hybrid” approach in which a previously developed adult female voxel computational phantom, NAOMI, was combined with stylized fetal computational phantoms developed separately by Chen.29 These fetal computational phantoms were first voxelized to a resolution of 2 mm × 2 mm × 2 mm to match with the original nonpregnant NAOMI computational phantom that had been adjusted to conform with the ICRP reference adult female.10 It was noted by the author that the process required direct voxel editing to adjust the overlap between the fetus and certain anatomical structures in the mother—an approach known to be time-consuming and potentially inaccurate because of the large amount of voxels involved. In addition, the fact that the fetal computational phantoms are overly simplified is a severe compromise in the anatomical realism. The combined mother/fetus computational phantom was to date used to study nonionizing radiation dosimetry.11 Similarly, Cech et al.9 developed a pregnant female computational phantom, SILVY, using MR images for the external body of a pregnant woman who was found to have a malformed

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

309

fetus at the beginning of the third trimester. The MR images of the fetus portion were replaced with those from the CT images of a 30 week fetus3 reported by us. As mentioned earlier, the CT images of the fetus have poor image resolution. Nevertheless, researchers constructed the SILVY computational phantom and used it to calculate current densities within the central nervous systems of the mother and fetus for homogeneous 50 Hz electric and magnetic fields.

12.4 Development at Rensselaer Polytechnic Institute on a CT-Based Pregnant Female Computational Phantom in 2004 In 2004 for his doctoral research, Dr. Chengyu Shi developed a 30 week (∼ 7 month) pregnant female computational phantom from a set of CT images that were taken when the pregnancy was unknown.2,3 However, the emergency-room diagnostic procedure only covered the body partially from the upper chest to the thigh, and the CT image resolution (0.94 × 0.94 × 7 mm) was a compromise for anatomical modeling. Figure 12.2 shows coronal, axial, and sagittal views of the image respectively.

(a)

(b)

(c) FIGURE 12.2 Three different views of the CT images of the 30 week pregnant woman.

Handbook of Anatomical Models for Radiation Dosimetry

310

The whole image set has been segmented using image-processing algorithms, such as “erode,” “dilate,”“and,” “or,” and “difference.” A user code, CT-segmentation, has been developed using Microsoft Visual C++ 6.0 to get clear images and help with segmenting some organs. To verify the segmentation, we used Visualization ToolKit (VTK) to visualize the location, shape, and size of organs/tissues in three-dimensional (3D). VTK is an open-source, freely available software system for 3D computer graphics, image processing, and visualization provided by Kitware, Inc. (http://public.kitware.com/ VTK/). We used Tcl/Tk as the interface layer for the 3D visualization. All the work was done using a Pentium IV computer running Microsoft Windows 2000 (2.0 GHz Intel Processor, 256 MB RAM). After segmentation, 34 organ IDs were reserved for identified organs and tissues. Table 12.1 shows organs/tissues to be segmented and the reserved IDs for each identified organ/ tissue. After segmentation, the organ IDs have been assigned to material properties for the Monte Carlo simulation as shown in Table 12.2. We used the density and chemical composition in the Monte Carlo simulation and took them from International Commission on Radiation Units and Measurements (ICRU) Report 44.31 For those that were not listed in the ICRU Report, the value was assigned according to similar organ/tissue. Here each organ had uniform density and chemical composition, and this kind of approach was also used by other tomographic computational phantoms, such as the VIP-Man.32

TABLE 12.1 Organs/Tissues to Be Segmented and the Reserved IDs for Each Identified Organ/Tissue ID

Organs

ID

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Air-outside Muscle Bone Soft tissues of fetus Uterus Fat and soft tissues Skin Lungs Heart Liver Kidneys Stomach LLI wall Ovaries ULI wall Urinary bladder Esophagus

17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

Notes:

Organs Spleen Pancreas Placenta SI wall Inferior vena cava Superior mesenteric artery Stomach contents LLI contents ULI contents Urinary bladder contents SI contents Gall bladder contents Gall bladder wall Adrenals Blood Rectum Skeleton of fetus

ULI, upper large intestine; LLI, lower large intestine; SI, small intestine.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

IDs

Air-outside Muscle Bone Soft tissue of fetus Uterus Fat and soft tissue Skin Lungs Heart Liver Kidneys Stomach wall LLI wall Ovaries ULI wall Urinary bladder Esophagus Spleen Pancreas Placenta SI Wall

Organs

0.001205 1.04 1.55 1.052 1.052 1.052 1.10 0.26 1.03 1.05 1.05 1.04 1.04 1.05 1.04 1.04 1.04 1.06 1.05 1.052 1.04

Density 10.2 3.4 10.2 10.5 10.6 10 10.3 10.3 10.2 10.3 10.1 10.6 10.5 10.6 10.1 10.1 10.3 10.6 10.6 10.6

H 0 14.3 15.5 14.3 9.3 31.5 20.4 10.5 12.1 13.9 13.2 11.1 11.5 9.3 11.5 11.1 11.1 11.3 16.9 31.5 11.5

C 75.5 3.4 4.2 3.4 2.4 2.4 4.2 3.1 3.2 3 3 2.6 2.2 2.4 2.2 2.6 2.6 3.2 2.2 2.4 2.2

N 23.2 71 43.5 70.8 76.8 54.7 64.5 74.9 73.4 71.6 72.4 76.2 75.1 76.8 75.1 76.2 76.2 74.1 69.4 54.7 75.1

O

0.3 0.2 0.2 0.1

0.1 0.2 0.2 0.1

0.2 10.3 0.3 0.2 0.2 0.1 0.2 0.1 0.3 0.2

P

0.1 0.2 0.1

0.2

Mg

0.1 0.2 0.1

0.1 0.1 0.3 0.2 0.2 0.2 0.2 0.1 0.2 0.2

Na

0.2 0.1 0.2 0.1

0.1 0.2 0.1

0.3 0.3 0.3 0.2 0.2 0.2 0.3 0.2 0.3 0.2

S

ID Number, Densities, and Chemical Compositions of Tissues/Organs in the Tomographic Pregnant Woman Model

TABLE 12.2

0.2 0.2 0.2 0.2

0.2 0.2 0.2

0.3 0.2 0.2 0.3 0.3 0.3 0.2 0.2

0.1

Cl

Ar 1.3

0.3 0.2 0.2 0.1

0.1 0.2 0.1

0.3 0.2 0.2 0.1 0.2 0.2 0.3 0.2

0.4

K

0.1

Fe

(continued)

0.1

22.5

Ca

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series 311

1.06 1.06 1.04 1.04 1.04 1.00 1.04 1.03 1.03 1.02 1.06 1.04 1.22

Density 10.2 10.2 10.1 10.6 10.6 66.7 10.6 10.1 10.1 10.1 10.2 10.6 0.1

H 3.3 3.3 2.6 2.2 2.2 2.2 2.6 2.6 2.6 3.3 2.2 0

11.5 11.1 11.1 11.1 11 11.5 0.1

N

11 11 11.1 11.5 11.5

C

ULI, upper large intestine; LLI, lower large intestine; SI, small intestine.

Inferior vena cava Superior mesenteric artery Stomach contents LLI contents ULI contents Urinary bladder contents SI contents Gall bladder contents Gall bladder wall Adrenals Blood Rectum Skeleton of fetus

21 22 23 24 25 26 27 28 29 30 31 32 33

Notes:

Organs

IDs 74.5 74.5 76.2 75.1 75.1 33.3 75.1 76.2 76.2 76.2 74.5 75.1 0.7

O

0.1 0.1 0

0.1 0.1 0

0

0.1

0.1

0.1 0.1

0.1 0.1

P 0.1 0.1

Mg

0.1 0.1

Na

0.2 0.1 0

0.1

0.1 0.1

0.2 0.2

S

ID Number, Densities, and Chemical Compositions of Tissues/Organs in the Tomographic Pregnant Woman Model

TABLE 12.2 (continued)

0.3 0.2 0

0.2

0.2 0.2

0.3 0.3

Cl

Ar

0.2 0.1 0

0.1

0.1 0.1

0.2 0.2

K

0.1

Ca

0.1

0.1 0.1

Fe

312 Handbook of Anatomical Models for Radiation Dosimetry

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

313

After carefully checking and verifying the segmented data, the segmented tomographic pregnant woman data were compressed by using a method similar to Run Length Encoding (RLE) algorithm, where four bytes were used to represent the organ ID and repeating times for the organ ID. The reason to use this RLE compression is that it is efficient when there are large amounts of repetitive data, such as black and white images with large black and white areas, and it does not waste too much CPU power, like other compression methods. The data were then implemented into EGS4-VLSI for Monte Carlo calculation. The procedure to implement the tomographic pregnant woman computational phantom into EGS4-VLSI included: 1. Creating the identical 3D array in the virtual world inside EGS4-VLSI. We need to create a total of 512 × 512 × 70 cubes. To generate such a large amount of cubes, one needs to define 513 Y–Z planes, 513 X–Z, and 71 X–Y planes. Because the size of each voxel equals to 0.94 mm × 0.94 mm × 7 mm, the separation between two adjacent Y–Z planes would be 0.94 mm, X–Z planes will be 0.94 mm, and X–Y planes will be 7 mm. 2. Assigning an address to each cube using Equation 12.1: Ad d ress (i , j , k ) = i + j × 512 + k × 512 × 512

(12.1)

where i (0 ≤ i ≤ 511), j (0 ≤ j ≤ 511), k (0 ≤ k ≤ 69) are the index of each cube in three dimensions. 3. Loading in all 70 slices. 4. Using the address of each cube to look up its ID number in slices as show in Figure 12.3. Related this cube to its media (tissue/organ), density, and chemical composition as shown in Table 12.2.

Z

512*512*70

X Origin

1

512*512

Y 512 FIGURE 12.3 Coordinates and voxel IDs for the tomographic pregnant woman.

Handbook of Anatomical Models for Radiation Dosimetry

314

SAF (fetus

liver) (kg–1)

10–3 10–4 10–5 Stabin et al.21 This study

10–6 10–7 10–8 10–1

100

Photon energy (MeV) FIGURE 12.4 Comparison of SAF (fetus ← liver) for the stylized 6 month model and the model in this study.

Based on the theory of radiation transportation, the Monte Carlo code will simulate the transportation processes of photons, electrons, neutrons, or other radiation particles. First, a particle with initial conditions (energy, direction, position, weighting factor) is produced, and this particle’s interactions with surrounding environment are then calculated. A random number is generated to determine which interaction will occur by comparing probabilities (i.e., cross sections) of interactions. The process is repeated, and a particle is tracked in the target until it deposits all its energy or escapes. The whole process is called a “history” of the particle. When a large number of histories (usually several millions) are studied in the same fashion, the results accurately predict the physical processes that may be experimentally determined. The developed computational phantom has been applied into the Monte Carlo simulation for the calculation of specific absorbed fraction (SAF) for both photon and electron emitters. Figure 12.4 shows the results of SAF (fetus ← liver) for photon emitters and comparison between the 6 month stylized computational phantom and this study. As for higher energy (>100 keV), the results show good agreement between 0.25% and 5.16%. But for lower energy (<50 keV), more significant differences are seen for the two sets of SAF (fetus ← liver), ranging from 0.37% to 895.43%. The result from the tomographic pregnant woman is slightly higher than the 6 month stylized computational phantom. The main reason is the overlap between organs (liver and fetus). For such lower photon energy, the photon particle is nonpenetrating and relative distance between organs has more effect on cross irradiation. Stabin and Yoriyaz33 found similar differences while comparing cross irradiation between the liver and kidneys using the phantom by Zubal et al.34 and the MIRD phantoms.21 Figure 12.5 showed the calculated energy self-absorbed fractions of 14 different organs with electron emitters.

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

1.1

Adrenals Fetus Stomach SI wall ULI wall LLI wall Kidneys Liver Ovaries Pancreas Placenta Spleen UB wall Uterus

1.0 Self-absorbed fraction

315

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Electron energy (MeV) FIGURE 12.5 The energy self-absorption fractions for electrons of 14 different organs. ULI, upper large intestine; LLI, lower large intestine; SI, small intestine; UB, urinary bladder.

12.5 Development at Rensselaer Polytechnic Institute on the RPI-P3, RPI-P6, and RPI-P9 Phantoms By 2006, it became clear that existing computational phantoms of pregnant females had many shortcomings—they were based on overly simplified shapes, partial-body images or hybrid geometries that contain compromised anatomical detail. The rapidly growing fetus inside the body of the mother during the pregnancy offered a challenging problem in anatomical modeling and radiation dosimetry. Although a variety of methods have been attempted, they all fail to produce pregnant female computational phantoms that are anatomically realistic and computationally versatile. An entirely new approach was needed. At that time, the difficulty in dealing with organ size adjustment in voxel-based phantom was well known from our work35 trying to change the size of the VIP-Man phantom to match the ICRP Reference Man. The idea of a purely surface-geometry based modeling approach was inspired by our experience in developing a Nonuniform Rational B-Splines (NURBS)-based respiration -simulating lung computational phantom.35 A few years earlier, such a NURBS method had been demonstrated by Segars36 to develop the well-known NCAT computational phantom for cardiac imaging. However, the pregnant female phantoms required more than just the NURBS type of geometries. We eventually came up with a mixture of methods that took advantage of CT, NURBS, and most important, polygonal meshes that had the necessary fine details for constructing one organ at a time and then assembling the organs into the human body with the desired size and volume. The rest of this chapter is devoted to a boundary representation (BREP) type of modeling method in which the anatomy is represented and adjusted in an organ-based surface geometric domain involving a mixture of data structures represented by voxels,

316

Handbook of Anatomical Models for Radiation Dosimetry

meshes, and NURBS. This method was applied in this study to develop a series of pregnant mother and fetus phantoms, named RPI-P3, RPI-P6, and RPI-P9, at the end of 3, 6, and 9 month gestational periods, respectively.22 The surface meshes (or NURBS) are derived for each organ using anatomical information available, for example, from medical images. These 3D BREP-type defi nitions, in general, allow the anatomical details of the organs to be preserved with smooth surfaces. In addition, the organ volumes (and masses) can be adjusted to match the reference values defi ned in the ICRP Publication 89.37 12.5.1 Materials and Methods A BREP modeling method has been largely unexplored by the radiation dosimetry community. The specific goal of our project was to design 3D representations of various organs of the pregnant female and fetus at various gestational stages. This means that the individual parts of the body would have to be designed and then assembled into a whole-body frame according to desired organ reference volumes without internal overlapping. We decided to adopt the BREP modeling approach instead of previously used constructive solid geometry (CSG) type of voxels because the latter is difficult to deform during the necessary process of adjusting the volumes and avoiding overlap. By dealing only with the surfaces in the BREP, we concentrated our efforts on the shape and location of organ boundaries and treated the materials within the organ as having a uniform Reference organ density—an assumption that has been adopted volumes in all previous dosimetry computational phanCT toms for health physics. This BREP modeling phantom approach has proven to be critical when the fetal computational phantoms were “inserted” into Organ the computational phantom of the mother. VIP-Man geometry 3D Figure 12.6 illustrates the general workflow phantom modeling of the BREP modeling procedure used to create the RPI pregnant-female series. It starts with the Mesh gathering of initial input data including reference organ Voxelization models organ volume information37 as well as anatomical information defi ned in CT images,2 the VIPMan computational phantom in both voxel and original anatomical cross-sectional photos, 32 and Organ volume selected mesh objects used in computer graphcheck ics. These data are used to construct 3D organ surface of a desired volume, one at a time, before integrating all organs into the body frame. The Tissue Voxel phantom assembly of organs requires the potential overcompositions lap of adjacent organs to be carefully avoided by adjusting the organ shape and location (while maintaining the volume). Because most of the Monte Carlo public-domain Monte Carlo codes do not directly calculations handle BREP type of geometry, we converted the surface defi nitions into voxels for radiation trans- FIGURE 12.6 port simulations. Such conversion is performed, Flowchart of the BREP modeling method used as described later, for any desired voxel size. to develop pregnant female models.

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

317

Next, the organ volumes in voxel format are once again checked by measuring the total voxels in an organ and by visualizing the anatomical shapes to make sure the voxelization process is faithful. Minor volume adjustment is automatically done by voxelization algorithm allowing no more than 1% discrepancy with the reference data. The last step is to link the voxelized organs and tissues with information on elemental compositions for Monte Carlo calculations. 12.5.1.1 Selection of Reference Anatomical Data Anatomical data for the pregnant female and the fetus are gathered from several origins in order to accurately specify the organ shapes, volumes, and locations. Three gestational periods at the end of the three trimesters were considered: 13, 26, and 38 weeks. We use the primary sources of information from published reference values for organ masses and densities in the ICRP Publication 89.37 When not given specifically by the ICRP, data from other sources were used, including Stabin et al.21 who developed the first series of stylized pregnant computational phantoms based on a previous publication of the ICRP Reference Man26 and Chen29 who amended Stabin et al.’s computational phantoms. ICRP37 has provided a vast amount of physiological and anatomical data as part of the defi nition of Reference Man for radiation protection purposes. However, some of the organs that have changing mass and density during the pregnancy are not included by the ICRP.37 Polynomial interpolation of data described in ICRP for other periods of pregnancy was necessary, and the order of the polynomials was chosen to ensure one solution for each polynomial. Below are the descriptions of reference data for several key organs that change during the course of pregnancy. 12.5.1.2 The Fetus Three anatomical components of the fetus were considered: the brain, the skeleton, and the soft tissue. They were also defined by Stabin et al.21 and Chen.29 Values for fetal brain and fetal body masses at the end of each trimester were interpolated based on ICRP37 data. Comparison with Chen28 data shows a good agreement, with our interpolated mass being 5% greater for 3 month gestation and 1% less for both 6 and 9 month gestations. Interpolated mass for fetal brain is higher than that reported by Chen by 7%, 5%, and 1% for 3, 6, and 9 month gestations, respectively. Estimates for fetal skeleton mass and density were taken from Chen.29 Density of fetal soft tissue was estimated based on its mass and volume. The mass was calculated as the difference between the mass of the entire fetus and sum of masses for brain and skeleton. The volume of soft tissue was calculated as a difference between total volume of fetus and sum of volumes for brain and skeleton. Density values reported by Chen29 and the interpolated ICRP masses were used to calculate the brain and skeleton volumes. The fetal crown-rump length is defined as the greatest distance between the vertex of the skull and the ischial tuberosities in the naturally curled position of the fetus.37 12.5.1.3 The Uterus and Placenta Similarly, we interpolated from data recommended in the ICRP Publication 89 the masses of placenta and uterine wall.37 A comparison with those reported by Chen29 shows systematically higher values used by Chen,29 except for the end of the third trimester. Namely, interpolated placenta mass is 28%, 17%, and 2% lower for 3, 6, and 9 months, respectively;

318

Handbook of Anatomical Models for Radiation Dosimetry

the uterine wall is 28%, 34%, and 4% lower, respectively. Uterine wall masses used by Chen and Stabin are the same. According to the ICRP,37 the volume of the uterus consists of four parts: the fetus, the placenta, the amniotic fluid, and unaccounted maternal stores. In our modeling, the latter two components were grouped into the uterine contents, leaving the placenta and the fetus as separate components. The mass of the uterine contents was calculated as a sum of masses of its components. A density of 1.03 g cm−3 was used in agreement with that of Chen.29 12.5.1.4 The Breasts ICRP values for the mass of the breasts were interpolated. Since the breasts were not separated in the mathematical computational phantom of Chen,29 there is no comparison for the breasts’ mass. 12.5.1.5 The Rest of the Body A pregnant female gains body weight during the pregnancy, and certain organs, such as the breasts, will have increased masses.37 These values were used to estimate the wholebody mass of the mother at the end of each trimester. The volume of the whole body was calculated using an average density of the human body excluding the skeleton and lung, because these two organs have considerably different densities. Hence, the density of the remainder was estimated from the mass of the remaining body (such as adipose and connective tissues) and volume. In addition, the following assumptions were made: the skin mass is constant at all stages of pregnancy; the mass of the bladder content varies 0–500 g, 37 and consequently the value of 130 g was adopted in the first version of the series; the esophagus has no contents, i.e., it consists entirely of esophagus wall. In the second version of the original RPI-P series, the following two minor modifications were adopted: 1. The volume of the urinary bladder contents was reduced in the RPI-P3, P6, and P9 models in accordance with that reported by Stabin et al.21 The masses became 128, 107, and 42.3 g, respectively, instead of a fixed 130 g used previously. This change reflects the smaller bladder that is generally found at later stages of pregnancy. 2. The transverse colon was shifted slightly upwards, covering the SI and the uterus from above. Such an arrangement does not affect the fetal shielding, but is more consistent with anatomical changes due to enlarged uterus.21 Once the reference values have been chosen, the organ shapes were specified to agree with given masses (and volumes). Such organ shapes were specified in this study with two-dimensional (2D) and 3D anatomical information available from the following existing data: 1. Segmented CT images of a 30 week pregnant female.2 This set of CT image covered the portion of the body between the lower breast and the upper thigh in 70 slices, each 7 mm thick. The image resolution was 512 × 512 pixels in a 48 cm × 48 cm view field, i.e., a voxel size of 0.94 mm × 0.94 mm × 7 mm. The images were processed to identify 34 organs and tissues including the fetus. We used this data to derive and to check the anatomical features of the fetus.

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

319

2. Segmented and original color images of the VIP-Man32 of various resolutions. This computational phantom was based on anatomical color images of the Visible Man from the Visible Human Project.38,39 The original image resolution of the Visible Man is 0.33 mm × 0.33 mm and the slice thickness is 1 mm, which allowed for small and radiosensitive structures to be identified and modeled. 3. 3D anatomical computational phantoms of selected organs in the form of polygonal meshes used for 3D computer graphics community.40,41 The anatomical accuracy of these mesh computational phantoms is confirmed by comparison against the anatomical data mentioned above. To adopt our previous work on segmented CT data of a 30 week pregnant female and VIP-Man computational phantoms for organ volume adjustment, it was necessary to convert the segmented 2D image slices into 3D surface descriptions by extracting the contour lines. Using the external uterine wall as an example, Figure 12.7 illustrates the process of contour extraction, followed by the 3D surface creation (in this example, the lofting), and fi nally integration of the uterine external wall into the whole-body assembly. The process starts with outlining the organ(s) of interest on each of the slices in the CT dataset. Each contour consists of a polygon with many nodes. Since the CT image slices have been previously segmented by semiautomatic methods,2 we adopted the original outlines with only minor adjustment to correct for some overlapping in the original outlines. The node editing was performed using Able Software Corp’s software, 3D-Doctor, which offers tools such as adding, deleting, moving, splitting, closing, etc. The number of nodes in each contour from the original CT images was optimized to facilitate the process of extracting surface features without compromising the anatomical details. The tolerance level for nodes reduction was chosen after trial-and-error experiments. The NURBS is used in this example because the surface of the uterine wall is relatively flat and smooth, and thus requires only a limited number of control points for deformation.

CT images

Stack of 2D contours

Contours extraction

Lofting

NURBS surface

Integrated organ

Integration

FIGURE 12.7 An example for extracting 2D contours from existing CT images to form 3D NURBS surface representation of external uterine wall that is later integrated into the BREP mother model.

320

Handbook of Anatomical Models for Radiation Dosimetry

After the organs have been individually prepared in processes described above, the 3D NURBS (or mesh surface computational phantoms as to be discussed below) were adjusted according to the ICRP reference volume data. Another commercial software, Rhinoceros, was used to perform scaling and fitting of organs directly in the 3D space. Manipulation in 3D space allows for a higher degree of interactivity and user-friendliness than those approaches operate directly in the voxel domain. The Rhinoceros software offers a number of useful 3D tools for geometry manipulations such as deformation, fitting, Boolean operations, volume checking, and a variety of 3D surface rendering options. In dealing with adjacent organs of simple topology, the use of automatic Boolean operation for excluding overlapping space can be helpful. In other cases, manual adjustments may be necessary. The Rhinoceros software also played an important role in refining the surface definition, for example, in the mesh format which is known to cause “holes” —the missing organ surface areas. The “unification of facets normals,” which has to do with orientation of each of the surface polygons, is important for accurately determining the volume of a 3D mesh object. Other mesh editing tools, such as “elimination of degenerate and duplicate facets” and “non-manifold edges,” were also part of this semiautomatic procedure in our work that took approximately 2 months for three individuals (two graduate students and one postdoctoral associate) to complete the final editing phase for all 3, 6, and 9 month pregnant female phantoms. Although this volume adjustment process has proven to be considerably more efficient than other approaches, such as these based directly on voxels, there is a need for specific-purpose software to further streamline the procedure. It is our experience that, for organs of complicated topology such as the skeleton, the lungs, and the liver, polygonal meshes can yield better anatomical accuracy than those defi ned in NURBS. Figure 12.8 is a snapshot of the Rhinoceros software during the procedure involving the liver computational phantom defi ned in the 3D mesh format

FIGURE 12.8 A 3D mesh model of the liver is assembled with the rest of the body of the mother to allow for the volume and mass to be adjusted according to the ICRP reference values. The potential overlaps with adjacent organs such as the lungs, heart, ribs, and uterine wall are carefully resolved using various modeling tools in the Rhinoceros software. (A movie showing the step is available at http://www.youtube.com/watch?v=H1_gIXi_V94.)

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

(a)

(b)

321

(c)

FIGURE 12.9 Models of the 9 month old fetus and the mother. (a) The adjusted skin surface model of the fetus model in mesh, (b) the adjusted skin model of the mother in mesh to accommodate for the fetus at 9 month gestation, and (c) a surface rendering showing the skin of the mother contains a 9 month old fetus.

and inserted into the mother’s body frame. Using the software editing tools, the volume (and thus the mass) of the liver is adjusted to match with the ICRP reference value. The potential overlaps of the liver with adjacent organs such as the lungs, the heart, the ribs, and the uterine wall were carefully resolved. Similar processes were employed for other organs, including notably the fetus, that have varying masses or positions during the pregnancy. The 3D mesh computational phantoms of the fetus representing the 9 month gestational period and the skin of the mother were obtained from Web3Dservice.41 Figure 12.9 shows the separate polygonal mesh computational phantoms of the fetus and the mother, as well as the RPI-P9 whole body assembled using the same steps. The orientation of the fetus is not clearly specified in the literature, and after consulting with colleagues and the previous work by Stabin et al.,21 we decided to adopt a 50° angle between vertical axis of female body and approximate central axis of the fetus both 6 and 9 months, and 60° for 3 month computational phantom, respectively. The fetal skeleton was adopted from the previously developed CT computational phantom 2 and the VIP-Man.32 Organs with complex shapes, such as the ribs, were described using polygonal meshes whereas simpler ones were described using NURBS surfaces. The former is known to be capable of creating faithful boundaries, but the large number of polygons can be time consuming for manual operations. On the other hand, our experience shows that the NURBS surfaces are easy to deform because of fewer control points involved. However, NURBS modeling can be difficult for branching structures like trachea, and there is a tendency for the NURBS surfaces to compromise the anatomical details. Branching requires the manual fitting of NURBS segments. The modification of such an object frequently resulted in holes appearing between the patches. Therefore, the complicated NURBS objects were converted to meshes before the volume adjustments were performed. The following organs were modeled in NURBS surfaces: the esophagus, the thyroid, the thymus, the trachea, and the adrenals, all based on the VIP-Man32; the bladder, the spleen, the uterus, the ovaries, and the placenta, all based on the original CT images of the

322

Handbook of Anatomical Models for Radiation Dosimetry

pregnant female.2 The rest of the organs were based on polygonal meshes, including the fetus, the skeleton of the mother, and the skin of the mother. In each of the three RPI-P pregnant female computational phantoms, a total of 35 organs and tissues were explicitly defi ned. There are two possible approaches in modeling the intestines that contain a rather complex topology. The streamlined process of shaping intestinal structures in the Rhinoceros software with subsequent voxelization process was modified in the following way. The thickness of the large intestine (LI) walls was reconstructed during voxelization step based on its external surface. Due to self-overlapping of the SI, the SI’s wall and contents were not separated in this work. Volume estimation in the Rhinoceros software was found to be unreliable, due to the self-overlapping of the intestinal tube. Therefore, the SI volume had to be further adjusted in a later step during the voxelization (to be discussed below). Repetitive tasks in Rhinoceros were automated using scripting. Rendering of various images and animation were also accomplished directly in Rhinoceros. We developed a method to improve the smoothness of the surface of the skin by increasing the number of polygons. Higher-resolution polymesh for skin surface was created using the “divide” tool available in ZBrush version 2.0 (Zbrush Pixologic, Inc, Los Angeles, CA). 12.5.1.6 Voxelization In order to finally define the computational phantom geometries in Monte Carlo codes for dose calculations, we developed a procedure to convert the finished surface computational phantoms into the voxels at a desired voxel size. First, each organ of the pregnant female was saved as a single 3D surface file (in the Wavefront file format) to create a library of organs. Second, a boundary of the entire computational phantom was defined to cover the body of the pregnant female. Then the resolution of the phantom was defined as desired, typically about 1 mm. Each polygonal organ computational phantom was voxelized using the binvox software.42,43 Given the resolution information, the binvox can change the mesh surface to voxels organ by organ. Binvox is a shareware that performs the conversion from a polygonal mesh surface to an object that consists of less than 1024 × 1024 × 1024 voxels. To cover the entire body, we divided the body into three sections: the head and neck, body trunk, and the legs and feet. We selected 1 mm as the typical voxel phantom resolution throughout the body, although we also used the 3 mm in most Monte Carlo simulations. Finally, these voxelized organs were assembled into a whole body after overlapping voxels had been cleaned. The mass and volume of each organ were against the ICRP reference values. Some organs, namely the LI, the esophagus and the trachea, have walls and contents. The separation of the wall/content in these walled organs was performed during the voxelization using ICRP reference data37 and anatomical data from computational phantoms developed previously.2,32 The skin. The skin was defi ned during voxelization process by adding a single layer of voxels around the body except for the eye lens and eyeballs. For this project, two voxel resolutions, 1 and 3 mm, were considered, resulting in two different volumes for the skin. Accordingly, the volumes of the skin for all three computational phantoms are different due to differences in the body surface area. The same reference mass of the skin 2.3 kg was assigned for all three phantoms and both voxel resolutions. Consequently, in order to yield the correct mass, the “skin density” was adjusted appropriately. Namely, the following skin densities were taken for 3, 6, and 9 months, respectively: 0.98, 0.93, and 0.88 g cm−3 for the 1 mm voxel resolution, and 0.32, 0.31, and 0.29 g cm−3 for the 3 mm voxel resolution. Compared to the reference ICRP37 density of 1.1 g cm−3, the values used for 3 mm skin

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

323

are three times lower. However, this configuration of the skin (large voxels with reduced density) yielded correct total mass and provided adequate radiation attenuation per unit surface area. Centroids coordinates. Information of organ centroids provides an idea on the relative organ positions that can directly affect the dose estimates in internal dosimetry. To calculate the centroids, we employed a coordinate system with the origin at the center of the RPI-P9 voxel grid. In order to compare centroids among the 3, 6, and 9 month computational phantoms, the RPI-P9 origin was taken as a reference point to align the other two computational phantoms so that the brain always had the same centroid coordinates. The X-axis is parallel to the shoulders with positive direction toward the phantom left. The Y-axis represented anterior–posterior direction (positive direction toward the back of the body). The positive Z-direction points upward to the head of the body. The voxelized 9, 6, and 3 month versions were used in the calculations of the mass center of each organ according to

Σ x=

n i =1

n

xi ,

(12.2)

where x- is the x-coordinate of organ centroid n is the number of organ voxels xi is the x-coordinate of each organ voxel Other centroids coordinates, i.e., y- and z-, are calculated in the same way. For organs consisting of two parts (e.g., the lungs, ovaries, etc.), we calculated their centroids separately. 12.5.1.7 Tissue Compositions After the mass and density of each organ were specified, the tissue elemental compositions for each organ were defined for the purposes of radiation transport simulations involving Monte Carlo methods. Table 12.3 presents organ-specific elemental compositions that were based on reference values of the ICRP37 and ICRU.44 12.5.1.8 Monte Carlo Modeling As part of our internal quality assurance procedure, the RPI-P series computational phantoms were implemented in three well-validated Monte Carlo codes—the EGS,45 MCNPX,46 and Penelope.47 Average absorbed doses to organs obtained from these two codes were compared against the same external photon beam irradiation geometries: anterior–posterior (AP), posterior–anterior (PA), left lateral (LLAT), right lateral (RLAT), rotational (ROT), and isotropic (ISO). The photon energies were 10, 15, 20, 30, 50, 100, 200, 500, 1000, 1500, 2000, and 4000 keV. The goal of this exercise was to minimize potential discrepancies arising from the difference in radiation physics, cross section, and dose computation algorithms. In addition, this step tests effectively against human errors that can incur during the translation of the computational phantom geometries into the EGS and MCPNX codes. More information about the treatment in each of the codes is provided below.

25.6 14.5 11.5 13.9 13.2 18.6 10.5 9.3 16.9 31.5 16 20.4 11.3 11.9 9.6 31.5 16 5.5 9 19.5 1.24 × 10−4

10.5 10.7 10.6 10.4 10.3 10.3 10.3 10.5 10.6 10.6 3.5 10 10.3 10.4 10.5 10.6 4.2 10.8 10.4 9.6

Adrenal, gall bladder wall and conts., esophagus wall and conts., trachea, thymus, fetal soft tissue Brain Stomach wall and conts., SI and LI wall and conts. Heart wall and conts. Kidneys Liver Lungs Ovaries Pancreas Uterine conts., remainder Skeleton Skin Spleen Thyroid Bladder wall and conts. Uterine wall Fetal skeleton Fetal brain Placenta Eyeballs and lenses Air, inside

C

H

Organ

2.2 2.2 2.9 3 2.8 3.1 2.4 2.2 2.4 4.2 4.2 3.2 2.4 2.6 2.4 4.5 1.1 2.6 5.7 0.7553

2.7

N

71.2 75.1 71.8 72.4 67.1 74.9 76.8 69.4 54.7 44.5 64.5 74.1 74.5 76.1 54.7 50.2 81.6 77.2 64.6 0.2318

60.2

O

0.2 0.2 0.1

0.2 0.1 0.1 0.2 0.2 0.2 0.2 0.2 0.1 0.3 0.2 0.1 0.2 0.2 0.1

0.1

Na

0.4 0.1 0.2 0.2 0.2 0.2 0.2 0.2 0.2 9.5 0.1 0.3 0.1 0.2 0.2 8 0.3 0.1 0.1

0.2

P

0.2 0.1 0.2 0.2 0.3 0.3 0.2 0.1 0.2 0.3 0.2 0.2 0.1 0.2 0.2 0.3 0.1 0.2 0.3

0.3

S

0.2 0.1

0.1 0.3 0.1 0.3 0.2

0.3 0.2 0.2 0.3 0.1 0.2 0.2 0.1

0.3 0.1 0.3 0.2 0.3 0.2 0.2 0.2 0.2

0.2

K

0.3 0.2 0.2 0.2 0.2 0.3 0.2 0.2 0.1

0.2

Cl

Other

Ar 0.0128

Mg 0.3, Ca 16.5

I 0.1

Mg 0.2, Ca 21.5

Ca 0.1

Organ-Specific Mass Fractions Used in Material Definition of RPI-P Series Based on ICRP 37 Except for Skin, Placenta, Air, Eye Balls, and Lenses Which Are Based on ICRU 44

TABLE 12.3

324 Handbook of Anatomical Models for Radiation Dosimetry

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

325

12.5.1.8.1 EGS External Photons Simulations A previously developed Monte Carlo user code, EGS4-VLSI,48 was adopted to implement the RPI-P computational phantom series at a voxel resolution of 1 mm (so the whole body consists of 400 million voxels). For the intercomparison, we are interested in the fluence-toabsorbed-dose conversion coefficients that were determined from: DT = Φ × C T

(12.3)

where Φ is the source photon fluence CT is the fluence-to-organ-dose conversion coefficient for the organ T and given source photon energy DT is the absorbed dose averaged over the entire organ Typically, 107 histories were simulated and the uncertainties were better than about 1% for most of the target organs. For organs that were relatively small and for energies that are very low, the statistical uncertainties were better than 10%. The cutoff energies for both electron and photon were set at 10 keV in the EGS4 code. Six personal computers, with CPU speed higher than 1 GHz with memory higher than 1 GB, were used for the simulations. 12.5.1.8.2 MCNPX External Photons Simulations The RPI-P computational phantom series were implemented at a voxel resolution of 3 mm in the MCNPX.46 A coarser voxel grid was used because of a limitation of the MCNPX code on the total number of geometry elements (about 25 million voxels). This limitation existed when we previously implemented the VIP-Man in the MCNPX code for external neutrons.32,49,50 In our research group, we typically use the MCNPX (and MCNP) codes for external photon, neutron, and proton sources. For this project, we implemented the RPI-P computational phantom series in a way to maximize the computational efficiency by taking advantage of recent improvements in dealing with standardized voxel geometries by the MCNP/X development teams.51,52 The procedure of the voxel data transformation into the code geometry is based on the repeated structures. First, we describe the surfaces that limit an elementary voxel. The description of a voxel is followed by the definition of a box that holds the whole 3D array of voxels. After this, we label the elementary voxels that belong to different materials. Finally, we fill the computational phantom box with voxels of different property (assigned to different organs) via the unique MCNPX feature of specifying repeated strictures. The cutoff energies for electrons and photons were 70 and 1 keV, respectively. The 70 keV cutoff energy corresponds to one-tenth of the shortest voxel linear size in the less dense material (lung) ensuring adequate electron equilibrium. The standard MCNPX v2.5.0 photon–electron cross-sections, MCPLIB04 and EL03 Tables, were employed. Standard physics computational phantoms for photon and electron transport were employed including fluorescence, coherent and incoherent scattering, pair production, and bremsstrahlung generation. Photonuclear generation was excluded in this comparison with EGS. For absorbed dose calculations, the MCNPX energy pulse-height tally (also known as energy balance tally *F8:P,E) was used with appropriate normalization for organ mass and source fluence. The simulations run successfully on a personal computer. The running time in case of external geometry of irradiation is ∼1–2 h per 10 million of primary photons, which is acceptable.

326

Handbook of Anatomical Models for Radiation Dosimetry

12.5.1.8.3 Penelope Simulations The maximum energy of the electron cross-sections in MCNPX is 1 GeV; therefore, the accuracy of dose estimates at higher energies had to be verified. To this end, the Penelope code was selected for benchmarking at all energies for the AP source using the RPI-P9 phantom voxelized at 3 mm.53 Penelope is a modern code system for Monte Carlo photon-electron transport simulation. Its principal advantage is the unique electron–positron transport model with parameterization that allows varying the details of electron tracks—from a heavily condensed representation to a very accurate step-by-step approach. Detailed photon transport is implemented in the code. The Penelope version 2005 was used as a basis for the tracking in the voxel geometry.54 In accordance with code paradigm, a special steering program (based on the pendoses example) was programmed that reads the voxel data, samples the source, and scores mean organabsorbed energy along with the variance. The key subroutine vxl_step was implemented in place of standard subroutine step from the subroutine package pengeom; it transports exclusively inside the voxel array. Standard Penelope transport parameters were used: C1 = C2 = 0.1 (average angular deflection and maximum average energy loss) Wcc = Wcr = 10 keV (cutoff energy loss of hard inelastic collision and bremsstrahlung emission) Maximum path length for electron and positron DSmax = 0.03 mm Energy cutoffs were used as in MCNPX: 1 and 70 keV for photons and electrons/positrons, respectively. Penelope cross-sections were also based on the EPDL 97 data. The results of benchmarking for the fetal doses showed a good agreement between the MCNPX and Penelope codes—within 1% on average for all energies with a maximum discrepancy of 5% at the highest energy of 10 GeV. One sigma relative error of the results did not exceed 2%. Penelope performed much faster than MCNPX: 19 times at 50 keV, and 2 times at 10 GeV. The correctness of the models setup in Penelope and MCNPX is concluded.

12.6 Results and Discussions In the earlier sections, we have described methods for various aspects of the computational phantom development and application. In this section, we summarize key anatomical parameters of the resulting RPI-P computational phantom series and present data on the dose intercomparison from two Monte Carlo codes. 12.6.1 Organs Volumes vs. Reference Data Although organ volumes were adjusted to the reference values during 3D modeling, a final check for the voxelized representation of a phantom was needed due to two reasons. Firstly, intricate polygonal mesh complications, i.e., self-overlapping and holes, can prevent Rhinoceros from reporting the correct volume of an object. Secondly, voxelization process may introduce approximations while converting from the 3D smooth surfaces to discrete voxels. In order to ensure the fidelity of the entire process of organ volume adjustment, a critical step to check on the final volume of each organ must be performed. The results of this effort are summarized in Table 12.4, which includes the various reference data from ICRP

Organs of constant volume Brain Eyeballs Eye lens Thyroid Trachea Thymus Lungs Heart wall Heart cont. Esophagus Stomach wall Stomach cont. Liver Gall bladder wall Gall bladder cont. Pancreas Spleen Kidneys Adrenals SI wall and cont.

Column A

Organ

1,299.49 14.63 0.40 17.00 7.99 20.00 950.34 250.15 369.96 35.00 139.56 230.67 1,399.81 8.00 48.19 119.68 129.58 275.04 13.00 N/A

Column B

RPI-3D Model

1,299.46 14.62 0.40 17.00 8.00 20.00 950.16 250.15 369.98 34.95 139.53 230.66 1,399.71 7.98 48.20 119.66 129.66 275.07 13.03 877.50

Column C

RPI-Voxel-1 mm

RPI-P Series

1,300.00 14.60 0.40 17.00 8.00 20.00 950.00 250.00 370.00 35.00 140.00 230.00 1,400.00 8.00 48.00 120.00 130.00 275.00 13.00 880.00

Column D

ICRP Reference Data

Mass (g)

0.0 0.1 0.1 0.0 0.0 0.0 0.0 0.1 0.0 −0.1 −0.3 0.3 0.0 −0.2 0.4 −0.3 −0.3 0.0 0.2 −0.3

Column E

RPI-Voxel-1 mm vs. Reference Data (%)

Organ Masses, Densities, and Volumes Used in the Phantom Construction (RPI; Version 2)

TABLE 12.4

1.04 1.03 1.10 1.05 1.03 1.03 0.25 1.03 1.06 1.03 1.04 1.04 1.05 1.03 1.03 1.05 1.06 1.05 1.02 1.04

Column F

Reference Density (g cm−3)

(continued)

1,249.48 14.25 0.36 16.19 7.77 19.51 3,800.63 242.86 349.04 33.93 134.16 221.79 1,333.05 7.75 46.79 113.96 122.32 261.97 12.78 843.74

Column G

Volume for RPIVoxel-1 mm (cm3)

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series 327

Column B

N/A N/A 11.00 40.04 7,862.62 N/A

128.00 107.00 42.33 270.11 549.94 1,046.85 859.61 3,657.42 4,108.67 47.50 319.26 650.00

LI wall LI cont. Ovaries Bladder wall Skeleton Skin

Organs of varying volume Bladder cont. 3 m Bladder cont. 6 m Bladder cont. 9 m Uterine wall 3 m 6m 9m Uterine cont. 3 m 6m 9m Placenta 3 m 6m 9m

RPI-3D Model

Column A

Organ

128.40 107.14 42.29 270.32 549.99 1,045.28 859.16 3,657.06 4,121.65 47.43 319.28 649.88

361.23 320.82 10.99 40.05 7,798.26 2,300.00

Column C

128.00a 107.00a 42.30a 270.10b 548.90b 1,050.00b 859.20 3,657.80 4,140.00 47.50c 319.20c 650.00c

360.00 320.00 11.00 40.00 7,800.00 2,300.00

Column D

ICRP Reference Data

Mass (g)

RPI-Voxel-1 mm

RPI-P Series

0.3 0.1 0.0 0.1 1.05 −0.4 0.0 0.0 −0.4 −0.1 0.0 0.0

0.3 0.3 −0.3 0.1 0.0 0.0

Column E

Reference Data (%)

RPI-Voxel-1 mm vs.

Organ Masses, Densities, and Volumes Used in the Phantom Construction (RPI; Version 2)

TABLE 12.4 (continued)

1.04 1.04 1.04 1.05 1.05 1.05 1.03 1.03 1.03 1.04 1.04 1.04

1.04 1.04 1.05 1.04 1.30 1.10

Column F

(g cm )

−3

Reference Density

123.46 103.02 40.67 257.45 523.80 995.51 834.14 3,550.54 4,001.60 46.18 308.48 624.88

347.34 308.48 10.47 38.51 5,998.66 Varies

Column G

Voxel-1 mm (cm3)

Volume for RPI-

328 Handbook of Anatomical Models for Radiation Dosimetry

70.92 902.21 2,754.45 N/A 78.38 371.00 13.60 133.93 369.95 569.52 796.60 905.52 N/A N/A N/A

71.30 902.59 2,747.98 N/A 77.92 370.41 13.60 133.92 370.04 569.94 796.76 904.56 42,486.78 42,997.29 45,047.37

71.30 902.50 2,760.00 N/A 78.00 370.00d 13.60 134.10 370.00 569.00 795.00 905.00 42,495.70 42,997.30 45,085.00

0.0 0.0 −0.4 N/A −0.1 0.1 0.0 −0.1 0.0 0.2 0.2 0.0 0.0 0.0 −0.1 0.97 1.00 N/A 1.22 1.22 1.03 1.03 1.03 1.05 1.05 1.05 0.97 0.97 0.97

74.17 931.57 2,756.81 N/A 63.87 303.61 13.36 130.28 359.26 542.80 758.82 861.49 43,800.80 44,327.10 46,440.59

Note: Reference data are based on ICRP37 unless specifically denoted. RPI-voxel-1 mm data correspond to 1 mm voxel phantom resolution. RPI-3DModel data are based on direct volume estimation for polygonal mesh. a Stabin et al.21 b These values for uterine wall are interpolated from the ICRP data. In contrast, Chen29 used 374 g for 3 months, 834 g for 6 months, and 1,095 g for the 9 months. c These values for placenta are interpolated from the ICRP data. In contrast, Chen29 used 66 g for 3 months, 383 g for 6 months, and 640 g for the 9 months. d Chen.29

Fetal soft tiss. 3 m 6m 9m Fetal skeleton 3 m 6m 9m Fetal brain 3 m 6m 9m Breasts 3 m 6m 9m Remainder 3 m 6m 9m

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series 329

330

Handbook of Anatomical Models for Radiation Dosimetry

and previous published papers. To organize the information, we separated organs into two groups in the first column: those with a fixed mass, and those whose masses will change during pregnancy. Organs, such as the LI, will change in location inside the abdomen during pregnancy. Organs of fixed masses are listed in the upper part of Table 12.4, whereas the others are given in the lower part of the table for each of the gestational stages. 12.6.2 Visual Inspection of the Anatomy From Table 12.4, it can be seen that the discrepancies between masses reported by Rhinoceros for the 3D surface computational phantoms (column B) and those calculated after 1 mm voxelization (column C) are small and practically negligible. Organ masses after 1 mm voxelization are also compared with the corresponding ICRP reference values (column D), and the differences are within 0.5% (column E) indicating that the entire process of computational phantom development has very high fidelity. When data were not available directly from ICRP, published papers were used. For convenience to the readers, tissue densities and organ volumes are also provided. Due to the limited space in the table, we are unable to include results for the 3 mm voxelization process that typically yielded 1% differences for large organs and 4% for very small organ such as the eye lenses. Visual inspections were frequently necessary during the project to ensure the anatomical fidelity. Figure 12.10a shows the RPI-P3, P6 and P9 computational phantoms as rendered from the 3D surface geometries in Rhinoceros with extremely smooth surfaces. After the voxelization, the smooth surface is replaced with discrete voxels and the rendering of the 1 mm RPI-P9 is shown in Figure 12.10b. After the voxel version has been implemented in the MCNPX, the plotting function of the MCNPX is used as a convenient way to verify the ultimate geometry, as shown in Figure 12.10c. 12.6.3 Analysis of Organ Centroids Organ centroids were calculated using the method described previously. Results for the 1 mm voxel resolution are presented in Table 12.5 for the three RPI-P computational phantoms. These data are useful in the understanding of internal dosimetry results that are

(a) Updated

(b)

(c)

FIGURE 12.10 (See color insert following page 524.) The finalized RPI-P9 models. (a) Rendering of 3D models of the RPI-P3, P6 and P9 (from left to right) plotted from Rhinoceros. (b) Rendering of the voxelized RPI-P9 model before translated into the MCNPX. (c) A direct MCNPX geometry plot showing a cross section view of the 3 mm voxel model of the RPI-P9 implemented for Monte Carlo radiation transport calculations. Visual inspections allow the anatomical geometries to be verified.

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

331

TABLE 12.5 Organ Centroids Calculated for 1 mm Voxels Centroid Coordinates RPI-P3 Organ

x

y

Fixed position organs (2 mm threshold) Brain 0.1 6.4 Eyeball, left 3.4 −2.0 Eyeball, right −3.1 −2.0 Eye lens, left 3.4 −2.9 Eye lens, right −3.1 −2.9 Thyroid −0.3 6.8 Trachea −0.2 7.7 Thymus −0.5 3.6 Lungs, left 7.6 9.3 Lungs, right −7.2 9.0 Heart wall 0.9 5.5 Heart cont. 1.1 5.2 Esophagus 0.1 9.4 Stomach wall 3.4 4.9 Stomach cont. 3.9 4.8 Liver −4.6 3.3 Gall bladder wall −4.3 1.8 Gall bladder cont. −4.5 1.8 Pancreas −0.5 3.8 Spleen 9.4 3.5 Kidney, left 5.2 8.3 Kidney, right −5.0 9.0 Adrenal, left 3.9 9.5 Adrenal, right −4.0 9.6 Ovary, left 3.0 10.5 Ovary, right −3.3 9.3 Variable position organs Uterine wall 0.2 Uterine cont. 0.3 Placenta 0.1 Fetal brain 0.0 Fetal skeleton N/A Fetal soft tissue 0.0 Fetus, total 0.0 Breast, left 10.7 Breast, right −10.9 SI wall and cont. −0.4 Bladder wall 0.4 Bladder cont. 0.1 Skeleton (mother) −0.1 Skin 0.0 LI wall −0.4 LI cont. −0.7

−0.9 −0.1 −1.3 5.1 N/A 2.3 2.7 −3.2 −3.0 3.8 6.7 6.8 7.7 7.3 4.3 3.2

RPI-P6 z

x

y

75.0 71.0 71.0 71.4 71.4 55.9 51.2 45.0 40.0 40.6 41.5 39.5 47.1 29.0 29.0 29.4 25.5 25.5 21.7 27.3 21.9 20.2 27.7 26.1 7.3 8.0

8.6 7.0 12.7 3.0 N/A 4.5 4.2 37.8 37.7 12.8 −0.1 −0.7 11.3 4.9 11.95 14

RPI-P9 z

x

y

z

The same as those for the P3

−0.4 0.5 0.4 −0.1 −0.2 −0.2 −0.2 10.9 −11.1 −1.8 0.2 0.2 0.1 0.0 0.0 −0.3

−6.5 −4.3 −6.8 3.0 −2.0 −4.9 −3.8 −3.5 −3.2 5.6 6.0 5.9 7.4 6.6 3.1 1.5

13.0 11.4 21.4 5.4 8.3 8.5 8.1 37.7 37.6 16.3 −1.1 −1.4 13.0 5.2 13 15.4

0.1 0.6 0.9 0.0 −0.1 −0.1 −0.1 10.9 −11.1 0.1 0.2 0.2 0.2 −0.1 0.4 0.05

−8.7 −6.1 −10.3 3.9 −2.2 −7.8 −6.1 −1.9 −2.5 4.3 6.7 6.7 7.5 6.1 2.5 0.2

15.6 12.0 23.5 4.2 8.2 9.0 8.4 38.3 37.9 16.8 −2.8 −3.0 12.9 4.5 15.4 19.4

Handbook of Anatomical Models for Radiation Dosimetry

332

TABLE 12.6 Distances in cm between Centroids of the Fetus and Other Organs, Calculated for a 3 mm Voxel Models and Compared with Those Reported by Stabin et al.21 and Shi and Xu2 Stabin et al.21

This Study Organ

3 Months

6 Months

9 Months

9 6 9 8 26

14 6 15 13 23

16 8 17 16 23

Placenta Uterine wall Ovary, left Ovary, right Liver

3 Months N/A 2 6 6 23

6 Months

9 Months

8 0 11 11 20

10 1 12 12 19

sensitive to organ-to-organ distance. The symmetry of paired organs can be observed in the table by comparing x-coordinates (the same or similar absolute value) due to the fact that x origin is located approximately in the middle between the shoulders. From data in Table 12.5, distances between centroids of the fetus and other organs can be derived and compared in Table 12.6 with those reported by Stabin et al.21 The results suggest that, while there is a general agreement in the trend, the distances in these two types of computational phantoms can be remarkably different due to the completely different topologies. Since the internal dose estimates are sensitive to organ-to-organ distances, we expected the dose results (to be reported in future papers) to exhibit profound differences especially for low energy photons between the RPI-P computational phantoms and those reported by Stabin et al.21

12.7 Conclusion We have developed a set of realistic computational phantoms of a pregnant female and her fetus at the end of three gestational periods of 3, 6, and 9 months—the so-called RPI-P3, RPI-P6, and RPI-P9 computational phantoms. We adjusted organ volumes and masses of the computational phantoms to agree with those recommended in the ICRP Publication 8937 for reference individuals. We also considered previously published reference data.21,29 The largely unexplored and powerful BREP modeling method, which involves a number of surface geometries such as polygon meshes and NURBS, has been systematically developed and carefully described in this paper. We also revoxelized and implemented the RPI-P computational phantoms into three Monte Carlo codes, EGS4, MCNPX, and Penelope, for quality assurance purposes in our group. The results suggest that the implementation of the computational phantoms in the Monte Carlo codes was reliable, and additional calculations can be performed for various internal and external exposures to ionizing radiation. The RPI-P series have been used for the following calculations: specific absorbed fractions for internal electron and photon emitters,55 dose conversion coefficients for external photons,53 neutrons and protons,56 and secondary exposures of external beam radiation treatment.57 These data have already demonstrated the versatility of this unique set of pregnant female computational phantoms. As we did with other computational phantoms, this set of pregnant female computational phantoms will be shared with other researchers.

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

333

The work presented in this paper was performed over approximately 9 months. We are convinced that it would have been impossible to develop such a complete, consistent, and anatomically realistic set of computational phantoms from medical images that are difficult to find. The task of segmenting and adjusting organs in the voxel domain to conform to reference values for a set of three computational phantoms would have been prohibitively time-consuming and labor-intensive. Although there was a steep, initial learning curve in using the BREP modeling method, the intrinsic advantage of the method became obvious even in early stages of the project.

Acknowledgments Work on the RPI-P series was funded by grants from National Cancer Institute: R01CA116743 (awarded to Rensselaer) and R42CA115122 (awarded to RADAR Inc). Inputs from Dr. A Randy Brill, Dr. Keith Eckerman, Dr. Wesley Bolch, Dr. W Paul Segars and Dr. Peter Caracappa during the project are appreciated. Mr. Juying Zhang and Mr. Di Zhang were part of the team that developed the RPI-P series.

References 1. ICRP. Pregnancy and Medical Radiation, ICRP Publication 84, Oxford: Pergamon Press, 2000. 2. Shi, C. and Xu, X.G. Development of a 30-week-pregnant female tomographic model from computed tomography (CT) images for Monte Carlo organ dose calculations, Medical Physics, 31, 2491, 2004. 3. Shi, C.Y., Xu, X.G., and Stabin, M.G. Specific absorbed fractions for internal photon emitters calculated for a tomographic model of a pregnant woman, Health Physics, 87, 507, 2004. 4. U.S. Nuclear Regulatory Commission. Standards for Protection against Radiation, Title 10 Part 20 of the Code of Federal Regulations, Washington, DC, 1991. 5. Osei, E.K. and Kotre, C.J. Equivalent dose to the fetus from occupational exposure of pregnant staff in diagnostic radiology, British Journal of Radiology, 74, 629, 2001. 6. Neel, J.V. and William, J.S. The Children of Atomic Bomb Survivors, a Genetic Study, Washington, DC: National Academy Press, 1991. 7. UNSCEAR. Sources and effects of atomic radiation, Report to the General Assembly, United Nations Scientific Committee on the Effects of Atomic Radiation, 2000. 8. Lewis, B.J. et al. Cosmic radiation exposure on Canadian-based commercial airline routes, Radiation Protection Dosimetry, 86, 7, 1999. 9. Cech, R., Leitgeb, N., and Pediaditis, M. Fetal exposure to low frequency electric and magnetic fields, Physics in Medicine and Biology, 52(4), 879, 2007. 10. Dimbylow, P. Development of pregnant female, hybrid voxel-mathematical models and their application to the dosimetry of applied magnetic and electric fields at 50 Hz, Physics in Medicine and Biology, 51, 2383, 2006. 11. Dimbylow, P. SAR in the mother and foetus for RF plane wave irradiation, Physics in Medicine and Biology, 52, 3791, 2007. 12. Kainz, W. et al. Development of pregnant woman models for nine gestational ages and calculation of fetus heating during magnetic resonance imaging (MRI), in Proceedings of the 27th BEMS and EBEA Meeting 12-3, Dublin, Ireland, 2005.

334

Handbook of Anatomical Models for Radiation Dosimetry

13. Kainz, W. et al. Calculation of induced current densities and specific absorption rates (SAR) for pregnant women exposed to hand-held metal detectors, Physics in Medicine and Biology, 48, 2551, 2003. 14. Kawai, H. et al. Simple modeling of an abdomen of pregnant women and its application to SAR estimation, IEICE Transactions on Communications, E89B, 3401, 2006. 15. ICRP. Limits for Intake of Radionuclides by Workers: An Addendum, Publication 30, Part 4. Ann. ICRP 19 (4), Oxford: Pergamon Press, 1988. 16. ICRP. Biological Effects after Prenatal Irradiation (Embryo and Fetus), ICRP Publication 90, Oxford: Pergamon Press, 2003. 17. Stovall, M. et al. Fetal dose from radiotherapy with photon beams—Report of Aapm-RadiationTherapy-Committee Task Group No-36, Medical Physics, 22, 63, 1995. 18. Streffer, C. Can tissue weighting factors be established for the embryo and fetus?, Radiation Protection Dosimetry, 112, 519, 2004. 19. Kal, H.B. and Struikmans, H. Radiotherapy during pregnancy: Fact and fiction, Lancet Oncology, 6, 328, 2005. 20. Pavlidis, N.A. Coexistence of pregnancy and malignancy, Oncologist, 7, 279, 2002. 21. Stabin, M.G. et al. Mathematical models and specific absorbed fractions of photon energy in the nonpregnant adult female and at the end of each trimester of pregnancy, ORNL/TM-12907, Oak Ridge, TN: Oak Ridge National Laboratory, 1995. 22. Xu, X.G. et al. A boundary-representation method for designing whole-body radiation dosimetry models: Pregnant females at the ends of three gestational periods—RPI-P3, -P6 and -P9, Physics in Medicine and Biology, 52, 7023, 2007. 23. Snyder, W.S. et al. Absorbed dose per unit cumulated activity for selected radionuclides and organs, MIRD Pamphlet No. 11, New York: Society of Nuclear Medicine, 1975. 24. Snyder, W.S. et al. Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, Journal of Nuclear Medicine, 10 (Suppl. 3), 7, 1969. 25. Snyder, W.S., Ford, M.R., and Warner, G.G. Estimates of specific absorbed fractions for photon sources uniformly distributed in various organs of a heterogeneous phantom, MIRD Pamphlet No. 5, Revised, New York: Society of Nuclear Medicine, 1978. 26. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Oxford: Pergamon Press, 1975. 27. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources. I: Methods, ORNL/TM-8381/V1, Oak Ridge: Oak Ridge National Laboratory, 1987. 28. Cloutier, R.J. et al. Dose to fetus from radionuclides in bladder, Health Physics, 25, 147, 1973. 29. Chen, J. Mathematical models of the embryo and fetus for use in radiological protection, Health Physics, 86, 285, 2004. 30. Zaidi, H. and Xu, X.G. Computational anthropomorphic models of the human anatomy: The path to realistic Monte Carlo modeling in radiological sciences, Annual Review of Biomedical Engineering, 9, 471, 2007. 31. ICRU. Tissue substitutes in radiation dosimetry and measurement, ICRU Report 44, Bethesda, MD, 1989. 32. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Physics, 78, 476, 2000. 33. Stabin, M.G. and Yoriyaz, H. Photon specific absorbed fractions calculated in the trunk of an adult male voxel-based phantom, Health Physics, 82, 21, 2002. 34. Zubal, I.G. et al. Two dedicated software, voxel-based, anthropomorphic (torso and head) phantoms, in Proceedings of an International Workshop Held at the National Radiological Protection Board, Chilton, U.K., 1995. 35. Xu, X.G. and Shi, C.Y. Preliminary development of a 4D anatomical model for Monte Carlo simulations, in Monte Carlo 2005 Topical Meeting: The Monte Carlo method: Versatility Unbounded in a Dynamic Computing World, Chattanooga, TN, 2005.

Pregnant Female/Fetus Computational Phantoms and the Latest RPI-P Series

335

36. Segars, W.P. Development and application of the new dynamic Nurbs-based cardiac-torso (NCAT) phantom, PhD thesis, University of North Carolina at Chapel Hill, 2001. 37. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Oxford: Pergamon Press, 2003. 38. Ackerman, M.J. Accessing the visible human project. D-lib Magazine: The Magazine of the Digital Library Forum. http://www.dlib.org/dlib/october95/10ackerman.html, 1995. 39. National Library of Medicine. U.6. Board of Regents, Electronic imaging: Report of the Board of Regents, NIH Publication 90, Bethesda, MD: National Library of Medicine, 1990. 40. INRIA. http://www-c.inria.fr/gamma/, a website maintained by The French National Institute for Research in Computer Science and Control (INRIA), last accessed August 2007. 41. Web3Dservice. http://www.web3dservice.com, a website maintained by an International Internet and Multimedia solutions provider, last accessed August 2007. 42. Binvox. http://www.cs.princeton.edu/~min/binvox/, last accessed August 2007. 43. Nooruddin, F.S. and Turk, G. Simplification and repair of polygonal models using volumetric techniques, IEEE Transactions on Visualization and Computer Graphics, 9, 191, 2003. 44. ICRU. Photon, electron, proton and neutron interaction data for body tissues, Report 46, Bethesda, MD: International Commission on Radiation Units and Measurements, 1992. 45. Nelson, W.R., Hirayama, H., and Rogers, D.W.O. The EGS4 Code System, SLAC-265, (Stanford, CA: Stanford Linear Accelerator Centre, Stanford University), 1985. 46. Pelowitz, D.B. MCNPX User’s Manual Version 2.5.0, Los Alamos National Laboratory Report LA-CP-05-0369, 2005. 47. Salvat, F., Fernandez-Varea, J.M., and Sempau, J. PENELOPE, a code system for Monte Carlo simulation of electron and photon transport, OECD-NEA, 2003. 48. Chao, T.C. and Xu, X.G. Specific absorbed fractions from the image-based VIP-Man body model and EGS4-VLSI Monte Carlo code: Internal electron emitters, Physics in Medicine and Biology, 46, 901, 2001. 49. Bozkurt, A., Chao, T.C., and Xu, X.G. Fluence-to-dose conversion coefficients from monoenergetic neutrons below 20 MeV based on the VIP-man anatomical model, Physics in Medicine and Biology, 45, 3059, 2000. 50. Bozkurt, A., Chao, T.C., and Xu, X.G. Fluence-to-dose conversion coefficients based on the VIPMAN anatomical model and mcnpx code for monoenergetic neutrons above 20 MeV, Health Physics, 81, 184, 2001. 51. Taranenko, V., Zankl, M., and Schlattl, H. Voxel phantom setup in MCNPX, in Proceedings of the Monte Carlo 2005 Topical Meeting. The Monte Carlo Method: Versatility Unbounded In A Dynamic Computing World, Chattanooga, TN, 2005. 52. Wang, B. et al. The use of MCNP code for an extremely large voxel model VIP-man, in The Monte Carlo Method: Versatility Unbounded In A Dynamic Computing World, LaGrange Park, IL: American Nuclear Society, 2005. 53. Taranenko, V. and Xu, X.G. Fluence to absorbed foetal dose conversion coefficients for photons in 50 Kev-10 Gev calculated using RPI-P models, Radiation Protection Dosimetry, 131, 159, 2008. 54. Taranenko, V. and Zankl, M. Photon and electron transport simulation in voxel geometry with Penelope, in Proceedings of the 14th International Conference of Medical Physics, Nuremberg, Germany, September 14–17, 2005. 55. Shi, C.Y., Xu, X.G., and Stabin, M.G. SAF values for internal photon emitters calculated for the RPI-P pregnant-female models using Monte Carlo methods, Medical Physics, 35, 3215, 2008. 56. Taranenko, V. and Xu, X.G. Fluence-to-absorbed-dose conversion coefficients for neutron beams from 0.001 eV to 100 GeV calculated for a set of pregnant female and fetus models, Physics in Medicine and Biology, 53, 1425, 2008. 57. Bednarz, B. and Xu, X.G. A feasibility study to calculate unshielded fetal doses to pregnant patients in 6-MV photon treatments using Monte Carlo methods and anatomically realistic phantoms, Medical Physics, 35, 3054, 2008.

13 The Vanderbilt University Reference Adult and Pediatric Phantom Series Michael G. Stabin, Mary Ann Emmons-Keenan, W. Paul Segars, and Michael J. Fernald

CONTENTS 13.1 Introduction ............................................................................................................... 337 13.2 Methods ...................................................................................................................... 339 13.3 Results .........................................................................................................................340 13.4 Discussion and Conclusions....................................................................................343 References .............................................................................................................................345

13.1 Introduction Anthropomorphic computational phantoms used for dose calculations in nuclear medicine for the past 30 years have used the stylized anatomical computational phantoms that were originally developed for the Medical Internal Radiation Dose (MIRD) committee of the Society of Nuclear Medicine in 1960s.1 Figure 13.1 illustrates the exterior and cutaway views of the original MIRD stylized adult male computational phantoms. These computational phantoms were defined in three sections: an elliptical cylinder representing the arm, torso, and hips; a truncated elliptical cone representing the legs and feet; and an elliptical cylinder representing the head and neck. A number of organs and tissues were mathematically defined as occupying finite spaces within the whole body space, and were comprised of three types of tissue: “soft tissue,” bone, and lung. The mathematical descriptions of the organs were formulated based on descriptive and schematic materials from general anatomy references. Later improvements led to a series of “family” stylized computational phantoms, which include both genders at several ages.2 This series of stylized computational phantoms, along with a set of stylized computational phantoms representing the pregnant female at four stages of gestation,3 were used in the MIRDOSE4 and OLINDA/EXM 1.05 personal computer codes to facilitate calculation of standardized internal dose calculations. See Chapter 11 for further discussion of the pregnant female computational phantoms, including the recent development of realistic computational phantoms to replace those of [3]. The stylized computational phantoms of the past 30 years are now being replaced with realistic body computational phantoms based ultimately on human image data. The creation of a human body computational phantom employing nonuniform rational b-spline 337

Handbook of Anatomical Models for Radiation Dosimetry

338

Stomach wall

e

stin

nte

ll i ma

S

e arg er l e Upp testin in rge a er l e Low testin n i

(a)

(b)

(c)

FIGURE 13.1 Stylized adult male model of Snyder et al.1 showing (a) exterior view, (b) skeleton and internal organs, and (c) geometric shapes representing stomach and intestines.

(NURBS) computational phantoms by Paul Segars of Duke University6 allowed a realistic rendering of the human body, both male and female, as well as rapid and easy scaling of organs and the body. Figure 13.2 shows the adult male and female NURBS computational phantoms originally developed by Dr. Segars. The realism of these computational

FIGURE 13.2 Anterior views of the NURBS models of the adult male (left) and adult female (right).

The Vanderbilt University Reference Adult and Pediatric Phantom Series

339

phantoms is clearly superior to that of the stylized computational phantoms of the past, and other investigators, who created several tomographic computational phantoms from CT images,7 reported over- and underestimates of several tens of percent for organ-absorbed fractions between styled computational phantoms and more realistic, image-based computational phantoms.

13.2 Methods Researchers can easily scale the NURBS computational phantoms to different sizes and shapes, and we therefore developed a series of computational phantoms representing adults and children of different ages scaled from these original adult computational phantoms to match the recommended age-dependent organ masses given in International Commission on Radiological Protection (ICRP) Publication 89.8 Instead of performing tedious, slice-byslice segmentation of individual organs from many medical scans, which can take months or even years to create and perfect, and which, in the end, creates computational phantoms based on different individuals with unique organ placements and shapes, we found this method to be more efficient and created computational phantoms that were internally consistent. Existing software tools developed by Dr. Segars of Duke University, allow for the rapid scaling of human or animal NURBS computational phantoms (Figure 13.3). In this figure, for example, we see that the right lung is selected as an example. One or more selected organs may be translated or rotated in the any direction, scaled linearly in any direction, uniformly in three dimensions, from the center by a fixed factor, and

FIGURE 13.3 Screen image from the NURBS model scaling program developed by Dr. Segars of Duke University.

340

Handbook of Anatomical Models for Radiation Dosimetry

otherwise modified by the user. The viewing pane on the right then allows slice-by-slice inspection of results. Our starting point was the ICRP 89 recommended organ mass values for adults and children of the reference ages evaluated (newborn, 1 year old, 5 year old, 10 year old, 15 year old, and adult). Organ and whole body computational phantoms were developed and scaled to these organ masses, with inspection of the representation of the computational phantoms in comparison to medical images of real individuals and review by physicians to assess realism and make necessary adjustments. When accepted, the modified computational phantoms were saved, voxelized, and submitted to Monte Carlo radiation transport simulation codes. Separate male and female computational phantoms were made for all ages. For 10-year-old computational phantoms and younger, however, the ICRP does not offer separate organ mass values, so the same computational phantom was used for both genders, with the sex organs added as appropriate. We used the GEANT4 (GEometry ANd Tracking) C++ particle transport toolkit9 to perform radiation transport calculations in the voxel-based computational phantoms that were derived from the NURBS computational phantoms. The NURBS computational phantoms are easily voxelized at most resolutions and matrix sizes. After the NURBS computational phantoms were finalized, they were voxelized and imported into the GEANT4 routines to simulate the transport of radiation in the human body computational phantoms. Final masses used for comparison with the ICRP 898 recommended values were made based on the reported volumes of the voxelized computational phantom, as this was what was used in the transport calculations. For most organs, the difference between the NURBS reported and voxelized volumes were about 3%–5%. For small organs, however, the difference was sometimes greater.

13.3 Results Figure 13.4a through d shows images of selected computational phantoms from the series. The complete ICRP 89 tables of recommended organ masses are large, and will not be reproduced in their entirety here. Table 13.1 gives a comparison of a number of selected organs. Figures 13.5 through 13.7 show sample plots of absorbed fractions from the computational phantoms, with comparison to values from the Cristy/Eckerman phantoms.2 These plots typify the major trends seen in the results. Figure 13.5 shows photon-specific absorbed fractions (SAFs) for liver irradiating liver in the adult male phantom—results are similar, as the organs are similar in mass, and minor differences in shape does not affect these numbers significantly. Figure 13.6 gives results for photons in the adult male phantom for liver irradiating lung—here the values for the NURBS computational phantoms are higher at all energies, owing to the close geometrical proximity of the two organs in the more realistic NURBS computational phantoms, compared to the necessary separation which occurs in stylized phantoms. Figure 13.7 shows photon SAFs in the adult female for spleen irradiating stomach, in which we can see a general trend of higher SAFs, but the differences are more pronounced at lower energies. Figure 13.8 shows electron AFs (not SAFs) for electron sources in the kidneys of the adult female irradiating either the kidneys or liver. Traditionally, all of the AFs for electron self-irradiation have been assumed to be 1.0 for an organ irradiating itself and 0.0 for other organs. Although this trend is generally true, when electron transport is performed, there is some departure from this assumption at higher energies. This trend is more important for the smaller organs, particularly in the smaller phantoms (Figure 13.9).

The Vanderbilt University Reference Adult and Pediatric Phantom Series

(a)

(b)

(c)

341

(d)

FIGURE 13.4 (See color insert following page 524.) Sample images of NURBS standardized models, scaled to the ICRP 89 recommended reference organ masses, developed for internal and external dose assessment: (a) newborn female model, (b) 5-year-old male model, (c) 10-year-old female model, and (d) 15-year-old male model.

TABLE 13.1 Sample Organ Mass Values—Comparison of the ICRP 89 Recommended Values to Those in the Realistic Adult and Pediatric Phantom Series Adult Female

Organ Brain Right eye Left eye Salivary glands Heart with blood Right lung Left lung Liver Gallbladder Spleen Right kidney Left kidney Right adrenal Left adrenal Pancreas Left thyroid Right thyroid

ICRP Reference Mass (g)

Adult Male

NURBS Model Ratio Mass (g) NURBS/ICRP

ICRP Reference Mass (g)

NURBS Model Mass (g)

Ratio NURBS/ICRP

1300 7.5 7.5 70 620

1282.8 7.5 7.5 70.2 626.1

0.987 1.003 0.996 1.003 1.010

1450 7.5 7.5 85 840

1444.2 7.4 7.5 84.9 834.4

0.9960 0.9933 0.9995 0.9987 0.9933

475 475 1400 56 130 137.5 138 6.5 6.5 120 8.5 8.5

483.1 475.8 1409.3 57.6 129.4 134.9 137.2 6.4 6.6 121.9 8.6 8.5

1.017 1.002 1.007 1.028 0.996 0.981 0.994 0.989 1.013 1.016 1.016 1.003

600 600 1800 68 150 155 155 7 7 140 10 10

600.0 608.8 1790.7 67.3 149.5 157.3 156.2 6.9 7.1 143.0 10.0 10.0

1.0000 1.0146 0.9949 0.9893 0.9967 1.0147 1.0080 0.9818 1.0146 1.0212 1.0017 1.0029 (continued)

Handbook of Anatomical Models for Radiation Dosimetry

342

TABLE 13.1 (continued) Sample Organ Mass Values—Comparison of the ICRP 89 Recommended Values to Those in the Realistic Adult and Pediatric Phantom Series 15-Year-Old Female

Organ Brain Right eye Left eye Salivary glands Heart with blood Right lung Left lung Liver Gallbladder Spleen Right kidney Left kidney Right adrenal Left adrenal Pancreas Left thyroid Right thyroid

ICRP Reference Mass (g)

15-Year-Old Male

NURBS Model Ratio Mass (g) NURBS/ICRP

ICRP Reference Mass (g)

Brain Right eye Left eye Salivary glands Heart with blood Right lung Left lung Liver Gallbladder Spleen Right kidney Left kidney Right adrenal Left adrenal Pancreas Left thyroid Right thyroid

Ratio NURBS/ICRP

1300 6.5 6.5 65 540

1279.0 6.4 6.5 64.6 537.5

0.9838 0.9821 1.0018 0.9946 0.9953

1420 7 7 68 660

1423.0 6.4 6.5 67.5 659.9

1.0021 0.9821 1.0035 0.9932 0.9999

375 375 1300 49.3 130 120 120 4.5 4.5 100 6 6

377.0 377.0 1309.0 50.7 128.3 120.2 122.7 4.6 4.7 99.9 6.0 6.1

1.0053 1.0054 1.0070 1.0283 0.9873 1.0016 1.0229 1.0148 1.0404 0.9987 1.0001 1.0197

450.0 450.0 1300 52.7 130 125 125 5.0 5.0 110 6.00 6.00

455.3 447.9 1309.9 53.7 130.8 126.7 123.4 5.1 5.0 110.6 6.0 6.0

1.0119 0.9954 1.0076 1.0192 1.0059 1.0135 0.9868 1.0165 1.0029 1.0059 0.9980 1.0080

10-Year-Old

Organ

NURBS Model Mass (g)

ICRP Reference Mass (g) 1400 5.5 5.5 44 370 250 250 830 30.4 80 90 90 3.5 3.5 60 3.95 3.95

5-Year Old

NURBS Model Ratio Mass (g) NURBS/ICRP

ICRP Reference Mass (g)

NURBS Model Mass (g)

Ratio NURBS/ICRP

1401.4 5.6 5.5 44.7 368.8

1.001 1.024 1.003 1.016 0.997

1310 5.5 5.5 34 220

1326.6 5.5 5.4 34.1 218.0

1.0127 0.9985 0.9901 1.0020 0.9909

256.3 250.1 828.5 30.6 80.7 90.8 90.7 3.5 3.5 60.1 4.0 4.0

1.025 1.000 0.998 1.008 1.009 1.009 1.007 1.006 0.987 1.002 1.008 1.002

167.3 132.7 570 17.6 50 55 55 2.5 2.5 35 1.7 1.7

167.8 134.5 565.8 17.4 50.6 54.5 55.8 2.5 2.5 35.0 1.7 1.7

1.0030 1.0135 0.9927 0.9885 1.0117 0.9916 1.0145 0.9914 1.0049 1.0002 1.0074 1.0074

The Vanderbilt University Reference Adult and Pediatric Phantom Series

343

SAF (1/g)

0.001

0.0001

0.00001 0.01

0.1

1

10

Energy (MeV) FIGURE 13.5 Sample photon SAF plot, showing comparison of NURBS model values (diamonds) with values from the Cristy/ Eckerman stylized series (squares): Adult male (liver ← liver). 0.0001

SAF (1/g)

0.00001

0.000001

0.0000001 0.01

0.1

1

10

Energy (MeV) FIGURE 13.6 Sample photon SAF plot, showing comparison of NURBS model values (diamonds) with values from the Cristy/ Eckerman stylized series (squares): Adult male (lungs ← liver).

13.4 Discussion and Conclusions The previous generation of anthropomorphic computational phantoms was well thought-out and designed, and has served the dosimetry community well for over 30 years. Original body computational phantoms used in dosimetry were spherical in nature10; this was mathematically convenient and provided conservative estimates of SAF and dose, but was not at all realistic. The development of the stylized phantom series by Cristy and Eckerman2

Handbook of Anatomical Models for Radiation Dosimetry

344

1.00E–03 1.00E–04

SAF (1/g)

1.00E–05 1.00E–06 1.00E–07 1.00E–08 1.00E–09 1.00E–10 1.00E–11 0.01

0.1

1

10

Energy (MeV) FIGURE 13.7 Sample photon SAF plot, showing comparison of NURBS model values (diamonds) with values from the Cristy/ Eckerman stylized series (squares): Adult female (stomach ← spleen).

1.00E+00

Absorbed fraction

1.00E–01 1.00E–02 1.00E–03 1.00E–04 1.00E–05 0.01

0.1

1

10

Energy (MeV) FIGURE 13.8 Electron-absorbed fractions in the kidneys of the adult female, kidneys as a target (diamonds) and liver as a target (squares).

represented a revolutionary change in methods for development of SAFs, and dramatically increased the realism of available computational phantoms. The complexity of the calculations necessarily increased, as Monte Carlo methods were needed to estimate SAF values for the organs, but the ability to calculate cross-irradiation SAFs was introduced, and the science of dose calculation, for either internal or external sources of radiation, was greatly improved. The advent of image-based technologies for use in dose calculations is most important for the ability to perform patient-individualized dose calculations in nuclear medicine therapy,11 but has also been exploited to create more realistic standardized computational phantoms for general dose assessment from internal or external sources. The computational phantoms presented here are intended as updates to the Cristy/ Eckerman series, with the organ masses based on those recommended by the ICRP for

The Vanderbilt University Reference Adult and Pediatric Phantom Series

345

Absorbed fraction

1

0.1 0.01

0.1

1

10

Energy (MeV) FIGURE 13.9 Electron-absorbed fractions in the testes of the newborn male.

standardized individuals of various ages.8 Changes in actual SAF values from the Cristy/ Eckerman computational phantom series are notable, but mostly small in magnitude. The impact on calculated dose estimates for radiopharmaceuticals, other internal sources, or external sources should be minor, but of interest for evaluation. These phantoms, in conjunction with those of the Rensselaer Polytechnic Institute (RPI) group for pregnant females (Xu et al., Chapter 11), which provide an update on the Stabin et al. stylized series of phantoms,3 will be used in new versions of the OLINDA/EXM computer code and made available to the user community through the Consortium of Computational Human Phantoms hosted at RPI.

References 1. Snyder, W.S., Ford, M.R., and Warner, G.G. Estimates of specific absorbed fractions for photon sources uniformly distributed in various organs of a heterogeneous phantom, MIRD Pamphlet No. 5, Revised, Society of Nuclear Medicine, New York, 1978. 2. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381 V1-V7, Oak Ridge National Laboratory, Oak Ridge, TN, 1987. 3. Stabin, M.G. et al. Mathematical models and specific absorbed fractions of photon energy in the nonpregnant adult female and at the end of each trimester of pregnancy, ORNL/TM-12907, Oak Ridge National Laboratory, Oak Ridge, TN, 1995. 4. Stabin, M.G. MIRDOSE: Personal computer software for internal dose assessment in nuclear medicine, Journal of Nuclear Medicine, 37, 538, 1996. 5. Stabin, M.G., Sparks, R.B., and Crowe, E. OLINDA/EXM: The second-generation personal computer software for internal dose assessment in nuclear medicine, Journal of Nuclear Medicine, 46, 1023, 2005. 6. Segars, W.P. Development and Application of the New Dynamic NURBS-Based Cardiac-Torso (NCAT) Phantom, PhD thesis, University of North Carolina at Chapel Hill, Chapel Hill, NC, 2001.

346

Handbook of Anatomical Models for Radiation Dosimetry

7. Petoussi-Henss, N. et al. The GSF family of voxel phantoms, Physics in Medicine and Biology, 47, 89, 2002. 8. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, International Commission on Radiological Protection (ICRP) Publication 89, Pergamon Press, Oxford, 2002. 9. Agostinelli, S. et al. GEANT4-a simulation toolkit, Nuclear Instruments and Methods in Physics Research Section A—Accelerators Spectrometers Detectors and Associated Equipment, 506, 250, 2003. 10. ICRP. Report of Committee II on Permissible Dose for Internal Radiation, International Commission on Radiological Protection (ICRP) Publication 2, Pergamon Press, New York, 1959. 11. Guy, M.J. et al. RMDP: A dedicated package for I-131 SPECT quantification, registration and patient-specific dosimetry, Cancer Biotherapy and Radiopharmaceuticals, 18, 61, 2003.

14 Mesh-Based and Anatomically Adjustable Adult Phantoms and a Case Study in Virtual Calibration of a Lung Counter for Female Workers Yong Hum Na, Juying Zhang, Aiping Ding, and X. George Xu

CONTENTS 14.1 Introduction ...............................................................................................................348 14.2 Materials and Methods ............................................................................................348 14.2.1 Candidate Anatomical Organ Models ...................................................... 350 14.2.2 Anatomical References................................................................................ 352 14.2.3 Development of Deformable Adult Male and Female Phantoms with ICRP Anatomical Parameters ........................................................... 352 14.2.3.1 Mesh Preprocessing ..................................................................... 352 14.2.3.2 Mesh Volume Calculation ........................................................... 353 14.2.3.3 Mesh Deformation Algorithm ....................................................354 14.2.3.4 Collision Detection to Avoid Organ Overlaps ......................... 355 14.2.3.5 Special Mesh Deformation Operations ..................................... 357 14.2.4 Voxelization of the Mesh Geometry.......................................................... 358 14.2.4.1 Voxelization Procedures .............................................................. 359 14.2.4.2 Preparation of MCNPX Input Files with Specific Organ and Physical Properties ............................................................... 360 14.2.5 Software Development for Automatic Adjustable Phantom ................. 360 14.3 Applications and Discussion ................................................................................... 370 14.3.1 Virtual Calibration of a Lung Counter Using Female Phantoms of Varying Breast Size ................................................................................. 370 14.3.2 Monte Carlo Simulations of Contaminated Worker and Phoswich Scintillation Detectors ................................................................................. 371 14.3.3 Results of Virtual Calibration of Lung Counting Efficiency for Female Workers ...................................................................................... 371 14.4 Conclusions ................................................................................................................ 373 Acknowledgments ............................................................................................................... 374 References ............................................................................................................................. 374

347

348

Handbook of Anatomical Models for Radiation Dosimetry

14.1 Introduction Size- and posture-adjustable phantoms are computational models of the human body that have the ability to automatically or semiautomatically deform in order to define various organ shapes and volumes, as well as body postures. This type of deformable phantoms was found to be easier to develop with Boundary REPresentation (BREP) geometry definition methods in the form of either nonuniform rational B-splines (NURBS), polygon mesh, or a combination of both.1–5 Compared to voxel models, BREP models are better suited for geometry deformation and adjustment because a richer set of computerized operations can be utilized. These operations include extrusion, chamfering, blending, drafting, shelling, and tweaking. One of the most attractive applications of adjustable phantoms is their ability to “morph” into an existing reference phantom or into a new anatomy of a worker or patient for whom their does not exist sufficient whole-body image data. From 1997 to 2005, our work in the area of virtual human body phantom development at RPI was focused on voxelized phantoms, such as the VIP-Man phantom which was developed from high-resolution color photographs. Our experience with deformable phantoms started in 2005 when we adopted the NURBS data structure to develop the four-dimensional (4D) VIP-Man Chest Phantom that simulated respiration for radiation treatment planning (see Chapter 6 for three-dimensional (3D) and 4D VIP-Man models). In 2007, we decided to adopt a hybrid approach using a mixture of NURBS/mesh/voxel data structures to develop pregnant female phantoms (see Chapter 12). The procedures used at the time successfully allowed the separate bodies of the mother and fetus to be carefully deformed and then assembled into a single body that was anatomically realistic.6 However, the algorithms used at the time were based on a mixture of NURBS/meshes/ voxel data structures that were difficult to implement automatically by a computer. The deformation and adjustment of organs required a large amount of manual operation time and the outcome of each step was dependent necessarily on the operator’s experience. These initial projects lead us to the conclusion that the NURBS data structure was very advantageous for real-time simulations. On the other hand, one disadvantage that was observed was the tendency of NURBS geometry to severely approximate the anatomical detail of thin objects. Consequently, our focus then turned to the polygonal mesh data structure as it became clear that it had the advantage of both anatomical realism and computational efficiency. This chapter describes the basic methodologies behind our preliminary work that aims to develop a pair of male and female deformable phantoms containing automatic geometry deformation algorithms. To demonstrate the usefulness of these deformable phantoms, a case study is presented in this chapter by applying the female phantom to the virtual calibrations of a lung counter for in vivo bioassay measurements involving female workers of various chest and breast sizes.

14.2 Materials and Methods Ideally, a “deformable phantom” should include two components: (1) a reference phantom with anatomical parameters that match reference values for “average” individuals defined by the International Commission on Radiological Protection (ICRP).7,8 (2) A software

Mesh-Based and Anatomically Adjustable Adult Phantoms

349

program that automatically morphs the reference phantom into a different phantom defined by a set of new anatomical parameters. Although in principle any solid-geometry structure is a likely candidate, we have decided to work with polygon meshes because they offer high anatomical fidelity and computational efficiency. Polygonal meshes are composed of a set of vertices, edges, and faces that identify the shape of a polyhedral object in three dimensions. A mesh surface is defined by a series of connected polygon-shaped faces, most commonly triangles. Each face is made up a certain number of linear edges, which are in turn defined by two vertices. This data structure allows very complex geometries such as human organs and the body surfaces to be described in sufficient anatomical detail without overwhelming a computer’s memory. In sections below, we first discuss the development of adult male and female meshbased phantoms that are compatible with the ICRP Reference.9 We then introduce the organ deformation methods, including a collision detection algorithm, as part of the automatic mesh deformation computational process. This is followed by a description of the voxelization of the deformed mesh-based phantoms for Monte Carlo calculations. Finally, the development of a supporting software package that integrates the phantoms with various tools is discussed. Figure 14.1 illustrates the general workflow for this project. The process starts with initial anatomical mesh-files that were chosen as candidate 3D organ surface models. A key step of this process is to match the surfaces of these candidate organ meshes with the ICRP references (or other anatomies in other applications) without creating surface overlaps between adjacent organs. Thus, an automatic collision detection algorithm was a very important part of the development of a deformable phantom. Moreover, using a local mesh deformation procedure, the whole-body frame can be deformed to represent various body

Polygonal mesh models

Mesh optimization and preparation

Software interface for automatic organ and body deformation

Organ shape and volume specification (ICRP reference and patient-specific data)

Organ overlap checking

New mesh phantom

Voxel phantom

Monte Carlo calculations

Tissue compositions FIGURE 14.1 Flowchart of the process used to develop a deformable phantom.

Handbook of Anatomical Models for Radiation Dosimetry

350

postures such as sitting and anatomical features such as breast sizes. Since most of the Monte Carlo codes do not currently handle the BREP-type of geometries directly, a meshbased phantom needs to be converted to voxels for Monte Carlo dose calculations. The final step of this workflow is to link the voxel phantom with correct tissue density and elemental composition information so that radiation transport through the human-body phantom is modeled correctly in a Monte Carlo code. 14.2.1 Candidate Anatomical Organ Models Over the years, we have created many mesh-based organ models from the VIP-Man and other anatomical images. In this project, however, we have decided to adopt the commercially available Anatomium 3D10 male and female models as the candidate or base wholebody models. This set of data consists of 140 internal organs and skeletal structures (out of a possible 500 in the dataset). The topological information of each organ was stored in a common triangular mesh-file format known as the Wavefront Object (OBJ). Individual organ mesh-files were collected and stored to establish two anatomical organ libraries: one for a male phantom and one for a female phantom with a main difference in genderspecific sex organs. A unique ID number was given to each organ in the male and female organ libraries. Table 14.1 summarizes the organs which are described in the male and female phantoms along with their corresponding ID numbers.

TABLE 14.1 List of Organs and Their Corresponding Organ IDs ID 1 2 3 5 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Organ/Tissue Adrenal, left Adrenal, right Extrathoracic (ET) Oral mucosa, tongue Trachea Bronchi Blood vessels, head Blood vessels, trunk Blood vessels, arms Blood vessels, legs Humeri, upper half, cortical Humeri, upper half, spongiosa Humeri, upper half, medullary cavity Humeri, lower half, cortical Humeri, lower half, spongiosa Humeri, lower half, medullary cavity Ulnae and radii:ulnae and radii, cortical Ulnae and radii, spongiosa Ulnae and radii, medullary cavity Wrists and hand bones, cortical Wrists and hand bones, spongiosa

ID 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

Organ/Tissue Clavicles, cortical Clavicles, spongiosa Cranium, cortical Cranium, spongiosa Femora, upper half, cortical Femora, upper half, spongiosa Femora, upper half, medullary cavity Femora, lower half, cortical Femora, lower half, spongiosa Femora, lower half, medullary cavity Tibiae, fibulae and patellae, cortical Tibiae, fibulae and patellae, spongiosa Tibiae, fibulae and patellae, medullary cavity Ankles and foot bones, cortical Ankles and foot bones, spongiosa Mandible, cortical Mandible, spongiosa Pelvis, cortical Pelvis, spongiosa Ribs, cortical Ribs, spongiosa

Mesh-Based and Anatomically Adjustable Adult Phantoms

351

TABLE 14.1 (continued) List of Organs and Their Corresponding Organ IDs ID

Organ/Tissue

ID

Organ/Tissue

45 46 47 48 49 50 51 52 53 54 55 56 58 61 62 63 64 65 66 67 68 69 70 71 72 73 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90

Scapulae, cortical Scapulae, spongiosa Cervical spine, cortical Cervical spine, spongiosa Thoracic spine, cortical Thoracic spine, spongiosa Lumbar spine, cortical Lumbar spine, spongiosa Sacrum, cortical Sacrum, spongiosa Sternum, cortical Sternum, spongiosa Cartilage Brain Breast, left, adipose tissue Breast, left, glandular tissue Breast, right, adipose tissue Breast, right, glandular tissue Eye lens, left Eye bulb, left Eye lens, right Eye bulb, right Gall bladder wall Gall bladder contents Stomach wall Stomach contents Small intestine Ascending colon wall Ascending colon contents Transverse colon wall, right Transverse colon contents, right Transverse colon wall, left Transverse colon contents, left Descending colon wall Descending colon contents Sigmoid colon wall Sigmoid colon contents Rectum wall Heart wall Heart contents (blood) Kidney, left, cortex Kidney, left, medulla

91 92 93 94 95 97 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 119 120 121 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

Kidney, left, pelvis Kidney, right, cortex Kidney, right, medulla Kidney, right, pelvis Liver Lung, left, tissue Lung, right, tissue Lymphatic nodes, ET airways Lymphatic nodes, thoracic airways Lymphatic nodes, head Lymphatic nodes, trunk Lymphatic nodes, arms Lymphatic nodes, legs Muscle, head Muscle, trunk Muscle, arms Muscle, legs Esophagus Ovary, left (female only) Ovary, right (female only) Pancreas Pituitary gland Prostate (male only) Residual tissue Salivary glands, left Salivary glands, right Skin Spinal cord Spleen Teeth Testis, left (male only) Testis, right (male only) Thymus Thyroid Tongue (inner part) Tonsils Ureter, left Ureter, right Urinary bladder wall Urinary bladder contents Uterus (female only) Air inside body

Handbook of Anatomical Models for Radiation Dosimetry

352

14.2.2 Anatomical References To create phantoms that represent the characteristics of the ICRP Reference Man, a list of anatomical reference data was carefully compiled and later used to adjust the mesh-based organ models. The key information considered was the height and weight, as well as the mass, density, and elemental contents of individual organs—much of this information was available from ICRP Publication 89.8 The ICRP reference volumes can be calculated from the mass and density data. Additional anatomical data from the VIP-Man phantom,11 the RPI-P series pregnant female phantoms,6 and the REX/REGINA12 were also considered in this project for the purposes of testing the ability for these mesh-based phantoms to flexibly deform into a desired anatomy. 14.2.3 Development of Deformable Adult Male and Female Phantoms with ICRP Anatomical Parameters 14.2.3.1 Mesh Preprocessing From 3D computer graphics literature, it is known that 3D objects can be defined with specific types of meshes, such as manifold or nonmanifold, connected or disconnected, and open or closed. Computational anatomical models, such as the Anatomium 3D dataset, have well-defined human anatomical shapes and structures, but the mesh geometries are not always conducive for morphing operations. In order to facilitate necessary procedures in this project, the organ geometries in the Anatomium 3D dataset were carefully preprocessed to ensure that they were all defined by connected and closed meshes. This was achieved through mesh triangulation and preprocessing methods described below. The first step in mesh triangulation is to index each of the three vertices of every triangular face in the organ mesh in a counterclockwise fashion (i.e., loop1: V3 Æ V1 Æ V2, loop2: V4 Æ V3 Æ V2, loop3: V4 Æ V2 Æ V5). The directions of triangular surface normals can then be determined by the right-hand rule as demonstrated in Figure 14.2. This ensures N4,3,2 V3 V4 Loop2

N4,2,5

Loop1

N3,1,2 V1

Loop3 V2 V5

FIGURE 14.2

Vertex numbering and triangular surface normals common edges: V3V2 , V2V4 ; surface normal: → → N3,1,2, N4,3,2, N4,2,5; faces (or triangles): ΔV3V1V2, ΔV4V3V2, ΔV4V2V5.

Mesh-Based and Anatomically Adjustable Adult Phantoms

353

Teeth

(a)

(b)

Mandible

FIGURE 14.3 Comparison of mesh quality with and without mesh preprocessing of the mandible and teeth. (a) Before the preprocessing, the meshes have a lot of surface cracks. (b) The corrected meshes are closed and result in smooth surfaces.

→

→

→

that the surface normals, N3,1,2, N4,3,2, and N 4,2,5 of triangles ΔV3V1V2, ΔV4V3V2, and ΔV4V2V5, are oriented in a consistent outward-pointing direction (when looking at the mesh geometry from the outside).13 Another requirement of a mesh is that common edges, such as V3V2 and V2V4 , should be associated with loops which traverse them in opposite directions. (That is, in loop1 and loop2, the edge V3V2 is traversed in opposite directions; V2 Æ V3 versus V3 Æ V2. Likewise, in loop2 and loop3, the edge V2V4 is traversed in opposite directions; V2 Æ V4 versus V4 Æ V2). If these requirements do not hold, the mesh is called “open” because this signifies the presence of a “hole” or a “bad connection.” The opposite of an “open” mesh is a “closed” mesh. The distinction between an open and closed mesh is important because the volume of organ mesh geometries is only well defined if they are described by closed surfaces.14,15 The aforementioned preprocessing methods ensure that all of the organ meshes in the two phantom libraries were ready for the later steps described below. Figure 14.3 compares the mesh quality of the mandible and teeth before and after the mesh preprocessing. In this example, the original mesh file for the “mandible and teeth” has a total of 32,622 vertices among which 8.85% are found to be duplicated and a total of 61,640 faces among which 2.82% are found to be duplicated, leading to many holes and cracks as illustrated in Figure 14.3a. Through the mesh preprocessing, the problematic open-meshes were carefully corrected. Figure 14.3b shows the corrected meshes that yield smooth surfaces. 14.2.3.2 Mesh Volume Calculation For a solid volume of 3D organ meshes, the surface normals of each face must all be oriented in the outward pointing direction as described in the section above. The key to calculating the volume defined by a triangular mesh is to decompose it into several elementary tetrahedrones.16 Figure 14.4 depicts an example of a mesh volume calculation involving a simplest polyhedron, a tetrahedron, with vertices. Note how the geometry is decomposed into four different tetrahedrons with the same origin V0 which is {0, 0, 0}: {V0, V1, V3, V2}, {V0, V1, V2, V4}, {V0, V3, V2, V4}, and {V0, V3, V1, V4}. The volume of a single tetrahedron, {V0, V1, V3, V2}, can then be calculated with the following equation:

Handbook of Anatomical Models for Radiation Dosimetry

354

V2 (X2, Y2, Z2) V4 (X4, Y4, Z4)

V1 (X1, Y1, Z1) N1,3,2

V3 (X3, Y3, Z3)

Z

V0 (X0, Y0, Z0)

Y

X FIGURE 14.4 Example of a simple polyhedron, where the V0 is {0, 0, 0} vertices: V0, V1, V3, V2; tetrahedron 1: {V0, V1, V3, V2}; tetrahedron 2: {V0, V1, V2, V4}; tetrahedron 3: {V0, V3, V2, V4}; tetrahedron 4: {V0, V3, V1, V4}.

Volume _ V0 ,1,3, 2 =

1 (− X3Y2 Z1 + X2Y3Z1 + X3Y1Z2 − X1Y3Z2 − X2Y1Z3 + X1Y2Z3 ) 6

(14.1)

In general, to calculate the volume of a closed triangular mesh of arbitrary shape, we can decompose it into many tetrahedral sub volumes. Then by using Equation 14.1, we can sum over the volumes of each of these tetrahedrons to obtain the volume of the original triangular mesh. The total volume of the triangular mesh is given by the following equations: Volume _ Vi ′ =

1 (−X i 3Yi 2Z i1 + X i 2Yi3 Z i1 + X i3Yi1Z i 2 − X i1Yi3 Z i 2 − X i 2Yi1 Z i3 + X i1Yi 2Z i3 ) 6

Volume _ Vt ′ =

∑

(14.2)

volume _ Vi ′

i

where i is the index of elementary tetrahedrons or triangles. Triangle i has the coordinates of vertices, {Vi1, Vi1, Vi1}, {Vi2, Vi2, Vi2}, and {Vi3, Vi3, Vi3} which are ordered in such a way that the triangle surface normals are consistent with each other.17 14.2.3.3 Mesh Deformation Algorithm After the mesh volume calculations have been performed, each mesh is ready for deformation according to a user-defi ned algorithm. In the case of the adult male and female phantoms, the fi rst goal is to match the reference organ-specific volume data recommended by the ICRP.7,8 There are two possible approaches for mesh a deformation operations. One approach involves the multiplication of a universal scale factor to “every” vertex.

Mesh-Based and Anatomically Adjustable Adult Phantoms

355

Cortical bone; light gray Spongiosa; dark gray

(a)

(b)

FIGURE 14.5 Comparisons of scale modification and vertex normal modification using an example of the seventh rib spongiosa (inner mesh) that is created by the seventh cortical rib (outer mesh). (a) When the seventh cortical bone (outer mesh) was deformed by the universal scale factor, the result of the spongiosa mesh is not located inside the cortical rib mesh. (b) Using the scalar multiplication to each of vertex normals, the deformed mesh (spongiosa) fits the inside of cortical bone.

This method yields a uniformly scaled mesh which keeps the overall mesh shape, but the mesh is either reduced or increased in volume. The second approach involves the multiplication of a unique scalar to “each” vertex along the direction of vertex normals. For our applications, the latter approach was primarily used for the elastic deformation of most organ meshes while maintaining both the position and reference volume information. Figure 14.5 demonstrates the visual difference between these two mesh deformation methods. Each deformation operation is an iterative process in which an “acceptance criterion” of relative error, typically 0.5%, in the adjusted organ volume is used. To achieve this small volume error, both the universal scale factor and the unique scalar were evaluated using the Newton’s method, a well-known numerical analysis iterative technique.18 A solution that meets the acceptance criterion could be found with minimal computational cost. Once the desired mesh volume and deformation factors had been chosen for a userdefined “acceptable error,” all the organ meshes were automatically deformed accordingly to account for this change in volume. The center of mass of each organ does not change during this deformation process. 14.2.3.4 Collision Detection to Avoid Organ Overlaps A mesh-based adult male or female phantom contains more than 100 organs or tissues. It is essential to avoid organ overlaps between adjacent organs during the deformation process. The ray-casting method19 was used to avoid surface collisions by examining the distance between adjacent vertices in two different organ meshes. In this approach, every vertex point was designated as an origin of a single ray with the normal pointing at the other organ surface, as shown in Figure 14.6. Then, the distance between the origin and the reflecting point of the ray on an adjacent organ surface was calculated. Two surfaces are considered to be in a collision if the distance between two vertices on each of the meshsurfaces is less than a predetermined value, such as 3 mm used for this project, as shown in Figure 14.6. After a surface collision has been detected, the colliding vertex (denoted as a ⊗ mark in Figure 14.6) stops in that position and does not undergo further deformation. This process continues until the entire surface of each organ has been examined and the volume of the organ has matched with the given volume information.

356

Handbook of Anatomical Models for Radiation Dosimetry

Surface A

Surface B

FIGURE 14.6 Illustration of the collision detection algorithm for Surfaces A and B representing two organ surfaces. A collision event between two surfaces occurs when the distance between a vertex (labeled as ⊗) and the other surface along the normal direction is less than a predetermined value. The detection continues until all the vertices on both surfaces have been examined this way.

It must be noted that this procedure assumes that one of the two adjacent organs has a higher priority in a collision. The “priority organ” can expand its volume while the other organ gives away the space without changing its volume. Ideally, the organ deformation method should take into account of point-wise, physics-based tissue elasticity. In this project, however, the recommended ICRP organ density information was used to decide the priority of deformation between two organ surfaces (i.e., a softer organ gives away the space in a collision). Figure 14.7 shows how the density is used to prioritize the deformation process involving the liver and the right lung. In this example, the bottom of the right lung which is less dense is pushed upward by the liver using the mesh deformation and collision detection methods. The algorithms were first developed and fine-tuned using the MATLAB 7.4®.20

FIGURE 14.7 The right lung and liver both increase in volume size during collision and deformation. (Left) Before the collision and deformation. (Right) The bottom surface of the right lung, which has a lower density, is deformed and pushed upward by the adjacent liver after collision detection and deformation.

Mesh-Based and Anatomically Adjustable Adult Phantoms

357

14.2.3.5 Special Mesh Deformation Operations Because several organs have very complex geometric shapes, a few specialized deformation maneuvers were required. Specialized procedures were applied for the skeleton, stomach, urinary bladder, intestines that consist of multiple outer–inner layers, or wall-contents structures. While a typical candidate organ mesh consists of only one closed surface, these special anatomical structures must be described by a double-layer mesh surface. In addition, special procedures were used to adjust the breast size of the adult female mesh-model for virtual lung counting applications. Details about these specialized deformation algorithms are provided below: The bones and organs with outer/inner layers. In order to describe these hollow organ geometries using both outer and inner layer structures in the mesh domain (see Figure 14.5b), the original meshes were first used to create the outer layer surface meshes according to the desired volume. Next, outer meshes are deformed to create the inner layer surface meshes through the deformation and collision detection algorithms described above. For bone components, each part of the outer layer surface mesh (e.g., cortical bones) of the bone should be decomposed to two different types of inner bone meshes (spongiosa and cavities) before the deformation takes place. Eventually, each of the two finalized mesh-based adult male/female phantoms was defined by a total of 52 internal organs, 6 sets of muscles, and 458 bone structures. Figure 14.8 shows an example of the detailed cortical bone, spongiosa, and cavities structures as well as anatomically realistic muscles. Deformation of the breast for female workers. Accurate in vivo lung counting measurements for female workers must account for the chest thickness. A “virtual calibration” of a lung counter for female workers can be performed using a female computational phantom with adjustable breast sizes. For this project, a mesh-based breast-size adjustable adult female phantom was demonstrated using data on female brassiere Cup Sizes (CSs). First, a female

Cortical bone Spongiosa

Brachioradialis

Cavity Extensors of hands

Opponens

(a)

(b)

FIGURE 14.8 Special mesh deformation operations for certain organs and tissues. (a) Arm bone structures. (b) Arm muscles.

Handbook of Anatomical Models for Radiation Dosimetry

358

TABLE 14.2 Measured Anthropometric Data of the Deformable Breast Phantoms Sorted by the CS, BG, and UBG CS Code Letter Model

BG (cm)

AA A B C D E E40 F G

UBG (cm)

CS (cm)

Glandularity (%)

Mass (g)

89.2 89.2 89.2 89.2 89.2 89.2 89.2 89.2 89.2

10.4 13.4 15.4 16.6 18.4 20.8 22.8 22.8 24.8

40 15 12 11 9 7 40 6 5

500 1317 1700 1897 2331 2791 2854 3304 3855

99.6 102.6 104.6 105.8 107.6 110.0 112.0 112.0 114.0

Source: Hegenbart, L. et al., Phys. Med. Biol., 53, 5527, 2008. With permission.

Code

AA

A

B

C

D

E

F

G

CS range (cm)

10–12

12–14

14 –16

16 –18

19 –20

20 –22

22 –24

24 –26

FIGURE 14.9 Different breast sizes of the female phantom according to the brassiere CS defi nition (AA to G) by European clothes sizes standard (EN 13402).

breast model was isolated from the female whole-body phantom. Then, the breast model was deformed automatically. Specific parameters of the Bust Girth (BG) and the Under Bust Girth (UBG) recommended by the European cloth size standard were considered.21–23 According to EN 13402-2 and -3, the brassiere CSs (i.e., AA, A, B, C, D, E, F, G) were characterized by the difference between BG and UBG, which were measured by the perimeter of the outermost breast skin contour. The brassiere CS is represented by the following Equation 14.3, and tabulated in Table 14.2. CS = BG − UBG

(14.3)

In addition, the female breast model was deformed by a 60° downward angle along each vertex normal direction to simulate the effect of gravity for a bra-wearing woman. Figure 14.9 summarizes the adjustable breast phantoms. The application of such a breastadjustable female phantom will be described later in this chapter. 14.2.4 Voxelization of the Mesh Geometry A voxelization tool was developed using Visuall C++ (VC++) to convert the mesh-based adjustable phantoms into voxel geometries that can be adopted into Monte Carlo radiation transport calculations. The voxel size can be specified by the user.

Mesh-Based and Anatomically Adjustable Adult Phantoms

359

14.2.4.1 Voxelization Procedures First, each organ of the mesh-based adult male/female models was saved as a single 3D surface file in the organ library. Second, a bounding box enclosing the entire mesh-based phantom is specified. Next, the user defines the resolution of the voxelized phantom, with the default size of 1 mm. The in-house voxelization software is based on the parity-counting and ray-stabbing methods24 for closed-mesh polygonal surfaces. The ray-stabbing method is suitable for perfect closed-mashes, but the parity-counting method was also used in this process for certain geometrically complicated tissues, such as the vessels and muscles, that may still contain open-meshes. Figure 14.10 provides a visual inspection of the skull before and after the mesh voxelization process. Such visual comparisons are useful to ensure the fidelity of the conversion process. The bones and organs with outer/inner layers. Each organ in this group was converted to the voxel format by first voxelizing the inner layer structures (i.e., the contents of a walled organ or the spongiosa of a bone). The outer layer structures (i.e., the surface of an organ) are then voxelized as the remaining space between the outer and inner layer surfaces by taking into account of their respective reference tissue information. The vessels and muscles. After the mesh preprocessing, the complicated vessel structures (i.e., for blood vessels and lymphatic vessels) and muscle structures (i.e., for head, body, arms, and legs) needed to be checked as they tend to produce overlaps with other organs. To this end, the volumes of these organs were reevaluated in the voxel domain at the last step after other internal organs have been finalized. The skin. The skin is difficult to model in either voxel or mesh format because it is a very thin layer on the surface of the body with a thickness less than 1 mm. A representation by a double-layer mesh for the skin is not acceptable in the voxel domain. For this reason, as it is done in most voxel phantoms in existence, the skin was defined during the voxelization process by adding a single layer of voxels around the body excluding the eye lenses and eye balls. This caused the skin’s volume to differ from the reference value, depending on the chosen voxel resolution. To maintain the total reference skin masses—3.3 kg for the reference male and 2.3 kg for the reference female—we defined the skin’s density as the reference mass divided by the skin’s volume in the voxel domain.6

(a)

(b)

FIGURE 14.10 Visual inspection of the male skull during the conversion from mesh geometry to voxel geometry: (a) mesh format; (b) voxel format.

360

Handbook of Anatomical Models for Radiation Dosimetry

The female phantoms of varying breast size. The voxelization for the female breast phantoms was based on the density and elemental composition information obtained from the ICRP references.7,8 The density of the breast skin in the AA-size phantom, however, was modified from the nominal value to agree with the recommended mass by the ICRP. In addition, the E-size phantom which has 7% glandularity was revised to have 40% glandular tissue based on the recommended ICRP8 for the reference adult female, and was named as the E40 to be distinguished from the original E-size phantom. 14.2.4.2 Preparation of MCNPX Input Files with Speciﬁc Organ and Physical Properties A voxel-phantom is constructed using a lattice of organ IDs so that the Monte Carlo N-Particle eXtended (MCNPX)25 code is able to correctly calculate energy deposition in each organ during the radiation transport simulations. In the recent versions of MCNPX, the maximum allowed voxel number is about 25 millions, so that the voxel resolutions of the adult male and adult female were set to 2.7 and 2.5 mm, respectively. The resulting phantoms consist of 24.65 million and 24.47 million voxels, respectively. Table 14.3 summarizes the organ elemental compositions data which were used. These values are based on the reference values provided by ICRP 898 and the International Commission on Radiation Units and Measurements (ICRU).26 According to the ICRP Publication 23, 7 the average densities of the reference adult male and female were 1.07 g cm−3 and 1.04 g cm−3, respectively. The average volume of the whole body could be derived from dividing the whole body’s reference mass by the density. For the purpose of defining the connecting tissues that do not belong to any specific organ, a “remainders” tissue was defined to make up the mass difference between the total body and all explicitly accounted organs. A software package was developed in-house to create MCNPX input files. The geometry routines of the MCNPX uses the “repeated structures” feature to handle the voxel geometry. Information about the resulting phantoms is summarized in Tables 14.4 and 14.5. The relative errors of the final organ masses are less than 0.5% except for the small eye lens of reference volume of 0.18182 cm3 that has a relative error of about 3%. 14.2.5 Software Development for Automatic Adjustable Phantom All of the above mesh deformation and voxelization algorithms involved in creating deformable phantoms were implemented in a software program. The software was developed using the programming language Microsoft Visual C#.27 Adopted in both the Windows and Web applications, C# is a more powerful and simpler platform than C++. Furthermore, C# has the additional advantage of being compatible with the .NET framework. The mesh deformation and collision detection algorithms can be implemented as the Component Object Model (COM),28 which is an interface standard component in Microsoft. It allows building applications through C#, Visual Basic (VB), Visual C++ (VC++), or any other languages that are able to support the COM. This component-packaging approach makes this software design very robust and convenient in both mesh deformation and visualization. The mesh-based phantoms can be inspected in the main window of the software using the Visualization Toolkit (VTK).29 Even though the organ volume and overlap were automatically checked during the entire procedures, visual inspections were still useful. Figure 14.11 illustrates 3D front and back views of the reference adult male and female phantoms. Figure 14.12 shows the software Graphical User Interface (GUI) and the tools in a pull-down menu used for the female beast phantoms. The software GUI was developed in Microsoft Visual C# for breast deformation.

1, 2, 7, 8, 70, 71, 110, 114, 115, 120, 121, 126, 131, 134–136 61 72–86 87 89–94 95 97,99 111, 112 113 62, 64, 119 13, 16, 19, 22, 24, 26, 28, 31, 34, 37, 39, 41, 43, 45, 47, 49, 51, 53, 55 125 127 132 137, 138 139 66–69 140 5, 106–109, 133 3 9–12, 88 128 129, 130 100–105

ID

0.256

0.145 0.115 0.139 0.132 0.186 0.105 0.093 0.169 0.315 0.16

0.204 0.113 0.119 0.096 0.315 0.195 0 0.143 0.099 0.11 0.095 0.099 0.041

0.107 0.106 0.104 0.103 0.103 0.103 0.105 0.106 0.106 0.035

0.1 0.103 0.104 0.105 0.106 0.096 – 0.102 0.096 0.102 0.022 0.106 0.108

C

0.105

H

0.042 0.032 0.024 0.026 0.024 0.057 0.755 0.034 0.022 0.033 0.029 0.02 0.011

0.022 0.022 0.029 0.03 0.028 0.031 0.024 0.022 0.024 0.042

0.027

N

0.645 0.741 0.745 0.761 0.547 0.646 0.232 0.71 0.744 0.745 0.421 0.766 0.832

0.712 0.751 0.718 0.724 0.671 0.749 0.768 0.694 0.547 0.445

0.602

O

0.002 0.001 0.002 0.002 0.001 0.001 – 0.001 0.005 0.001 – 0.002 0.003

0.002 0.001 0.001 0.002 0.002 0.002 0.002 0.002 0.001 0.003

0.001

Na

Organ-Specific Mass Fractions for Various Tissue Composition Materials

TABLE 14.3

0.001 0.003 0.001 0.002 0.002 0.001 – 0.002 0.022 0.001 0.137 0.001 –

0.004 0.001 0.002 0.002 0.002 0.002 0.002 0.002 0.002 0.095

0.002

P

0.002 0.002 0.001 0.002 0.002 0.003 – 0.003 0.009 0.002 – 0.002 0.001

0.002 0.001 0.002 0.002 0.003 0.003 0.002 0.001 0.002 0.003

0.003

S

Tissue Material

0.003 0.002 0.002 0.003 0.001 0.001 – 0.001 0.003 0.003 – 0.002 0.004

0.003 0.002 0.002 0.002 0.002 0.003 0.002 0.002 0.001 –

0.002

Cl

0.001 0.003 0.001 0.003 0.002 – – 0.004 – 0.002 – 0.002 –

0.003 0.001 0.003 0.002 0.003 0.002 0.002 0.002 0.002 –

0.002

K

– – – – – – – – – – 0.289 – –

– – – 0.001 – – – – – 0.215

–

Ca

– – – – – – – – – – 0.007 – –

– – – – – – – – – 0.002

–

Mg

(continued)

– – I:0.001 – – – Ar:0.0128 – – Fe:0.001 – – –

– – – – – – – – – –

–

Others

Mesh-Based and Anatomically Adjustable Adult Phantoms 361

63, 65 14 17, 20, 23, 32, 35, 38 25 27 29 40 42 44 46 48 50 52 54 56 15, 18, 21, 30, 33, 36

ID

0.114 0.084 0.096 0.09 0.089 0.093 0.081 0.093 0.088 0.086 0.103 0.099 0.094 0.105 0.104 0.115

H

0.598 0.3 0.443 0.361 0.35 0.398 0.276 0.38 0.315 0.322 0.438 0.41 0.369 0.46 0.449 0.644

C 0.007 0.027 0.017 0.024 0.025 0.022 0.027 0.025 0.029 0.026 0.027 0.027 0.028 0.027 0.027 0.007

N 0.278 0.481 0.373 0.439 0.448 0.413 0.497 0.431 0.48 0.467 0.405 0.422 0.447 0.391 0.398 0.231

O 0.001 0.003 0.002 0.002 0.002 0.002 0.003 0.002 0.002 0.003 0.001 0.002 0.002 0.001 0.001 0.001

Na

Organ-Specific Mass Fractions for Various Tissue Composition Materials

TABLE 14.3 (continued)

— 0.035 0.023 0.028 0.028 0.024 0.038 0.023 0.028 0.032 0.009 0.013 0.02 0.005 0.007 0.001

P 0.001 0.004 0.003 0.003 0.003 0.003 0.004 0.003 0.004 0.004 0.002 0.003 0.003 0.002 0.002 0.001

S

Tissue Material

0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 – – – – – –

Cl – – – – – – – – – – – – – – – –

K – 0.065 0.042 0.051 0.052 0.043 0.071 0.041 0.051 0.059 0.013 0.022 0.035 0.006 0.01 –

Ca

– 0.001 – 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 –

Mg

– – – – – – – – – – Fe:0.001 Fe:0.001 – Fe:0.001 Fe:0.001 –

Others

362 Handbook of Anatomical Models for Radiation Dosimetry

Mesh-Based and Anatomically Adjustable Adult Phantoms

363

TABLE 14.4 Organ/Tissue Masses, Densities, and Volumes for the Male Phantom Masses (g) RPI Adult Male ID 1 2 3 5 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43

Mesh-Based Phantom 7 7 39.44 30.73 10 30 68.3 145.69 36.76 88 135.24 184.84 33.48 128.02 59.73 37.13 270.76 181.89 22.66 179.71 139.57 47.77 53.05 562.77 451 261.64 471.97 25.78 294.05 438.53 80.88 531.3 729.31 78.66 232.55 507.73 76.11 73.89 398.55 681.09 365.13

Voxel-Based Phantom 7.03 6.99 39.43 30.71 9.99 29.98 0.86 271.9 15.73 83.04 135.22 184.83 33.51 128.04 59.71 37.11 270.74 181.88 22.72 179.7 139.68 47.81 53.06 562.75 451 261.63 472.17 25.77 295.04 438.99 80.9 531.31 729.29 78.64 232.49 506.62 76.07 73.89 398.55 679.73 365.1

ICRP Reference Data 7 7 39.44 31 10 30 0.85 271.93 15.72 83.04 135.26 184.86 33.48 128.03 59.73 37.13 270.8 181.91 22.67 179.74 139.59 47.78 53.06 562.85 451.06 261.68 472.05 25.78 294.09 438.57 80.89 531.35 729.38 78.67 232.56 507.78 76.12 73.9 398.62 681.18 365.16

Voxel versus Reference Error (%) 0.38 −0.19 −0.01 −0.93 −0.05 −0.05 0.4 −0.01 0.05 0 −0.03 −0.02 0.09 0.01 −0.03 −0.04 −0.02 −0.02 0.23 −0.02 0.07 0.05 0 −0.02 −0.01 −0.02 0.02 −0.03 0.32 0.1 0.01 −0.01 −0.01 −0.03 −0.03 −0.23 −0.06 −0.01 −0.02 −0.21 −0.02

Reference Density (g cm−3) 1.02 1.02 1.03 1.05 1.03 1.03 1.06 1.06 1.06 1.06 1.92 1.2 0.98 1.92 1.11 0.98 1.92 1.11 0.98 1.92 1.11 1.92 1.15 1.92 1.16 1.92 1.12 0.98 1.92 1.11 0.98 1.92 1.11 0.98 1.92 1.11 1.92 1.23 1.92 1.12 1.92

Volume for RPI Voxel 2.7 mm (cm3) 6.89 6.85 38.28 29.25 9.7 29.11 0.81 256.51 14.84 78.34 70.43 153.43 34.19 66.69 53.87 37.87 141.01 164.1 23.19 93.59 126.03 24.9 46.12 293.1 389.65 136.27 420.13 26.3 153.67 396.08 82.55 276.72 658 80.25 121.09 457.1 39.62 60.17 207.58 605.55 190.16 (continued)

Handbook of Anatomical Models for Radiation Dosimetry

364

TABLE 14.4 (continued) Organ/Tissue Masses, Densities, and Volumes for the Male Phantom Masses (g) RPI Adult Male ID 44 45 46 47 48 49 50 51 52 53 54 55 56 58 61 62 63 64 65 66 67 68 69 70 71 72 73 75 76 77 78 79 80 81 82 83 84 85 86 87 88

Mesh-Based Phantom 519.99 221.11 192.19 102.91 73.55 286.54 335.29 186.17 302.03 109.21 173.48 9.89 56.3 88.65 1450 0.36 0.18 0.27 0.13 0.2 7.3 0.2 7.3 10 58 149.97 249.97 999.9 89.97 55.02 59.99 94.99 59.99 40.01 89.97 34.99 40.01 74.97 29.98 329.94 510.43

Voxel-Based Phantom 519.99 221.08 192.18 102.91 73.54 286.54 335.28 186.16 302.01 109.22 172.98 9.86 56.16 88.64 1449.97 7.48 4.98 7.48 4.98 0.19 7.3 0.19 7.3 10.01 57.72 150.56 249.31 999.89 89.99 55.02 60 95 60 40.02 89.99 35 40 74.98 29.99 330.38 509.93

ICRP Reference Data 520.06 221.13 192.21 102.92 73.55 286.58 335.34 186.19 302.07 109.23 173.51 9.89 56.31 88.67 1450 7.5 4.99 7.5 4.99 0.2 7.3 0.2 7.3 10 58 150 250 1000 89.99 55.02 60 95 60 40.01 89.99 35 40.01 74.97 29.98 330 510

Voxel versus Reference Error (%) −0.01 −0.02 −0.01 −0.01 −0.01 −0.02 −0.02 −0.02 0.02 −0.01 −0.3 −0.28 −0.26 −0.03 0 −0.23 −0.3 −0.23 −0.3 −2.57 −0.02 −2.57 −0.02 0.15 −0.48 0.37 −0.28 −0.01 0 0.01 0 0 0 0.02 0 0.01 −0.03 0.02 0.03 0.12 −0.01

Reference Density (g cm−3) 1.17 1.92 1.18 1.92 1.05 1.92 1.07 1.92 1.11 1.92 1.03 1.92 1.04 1.1 1.04 0.95 1.02 0.95 1.02 1.1 1.03 1.1 1.03 1.03 1.03 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.05 1.06

Volume for RPI Voxel 2.7 mm (cm3) 446.16 115.15 162.42 53.6 70.03 149.24 312.21 96.96 271.68 56.88 167.72 5.14 53.97 80.58 1394.2 7.87 4.88 7.87 4.88 0.18 7.09 0.18 7.09 9.72 56.04 144.77 239.72 961.44 86.53 52.91 57.69 91.35 57.69 38.48 86.53 33.66 38.46 72.1 28.84 314.65 481.07

Mesh-Based and Anatomically Adjustable Adult Phantoms

365

TABLE 14.4 (continued) Organ/Tissue Masses, Densities, and Volumes for the Male Phantom Masses (g) RPI Adult Male ID

Mesh-Based Phantom

89 90 91 92 93 94 95 97 99 100 101 102 103 104 105 106 107 108 109 110 113 114 115 119 120 121 125 126 127 128 129 130 131 132 133 134 135 136 137 138 140

107.11 38.25 7.63 109.9 39.24 7.86 1799.75 552.99 646.99 2.26 6.4 5.95 87.98 6.7 9.45 1217.68 15,004.57 2750.14 10,023.56 40 140 0.58 17 N/A 42.09 44.35 N/A 30.04 150 50.03 17.48 17.52 25.12 20 42.27 3.01 8.5 7.49 50 200 N/A

Voxel-Based Phantom 107.12 38.23 7.63 109.91 39.23 7.85 1797.81 555.18 644.95 2.25 6.39 5.98 104.39 7.83 11.09 1217.69 14,996.38 2750.16 10,023.63 40.02 140.31 0.59 17.07 21,007.48 42.7 42.7 3300 30.05 150.2 50.01 17.46 17.52 25.12 19.98 42.26 3 8.49 7.5 50.21 199.79 N/A

ICRP Reference Data

Voxel versus Reference Error (%)

107.12 38.25 7.63 109.92 39.25 7.87 1800 553 647 2.26 6.4 5.98 104.4 7.83 11.1 1217.81 15,006.82 2750.53 10,024.97 40 140 0.6 17

0 −0.03 −0.1 −0.01 −0.06 −0.2 −0.12 0.39 −0.32 −0.33 −0.17 −0.05 −0.01 −0.02 −0.1 −0.01 −0.07 −0.01 −0.01 0.05 0.22 −2.01 0.42

42.5 42.5 3300 30 150 50 17.5 17.5 25 20 42 3 8.5 7.5 50 200 N/A

0.47 0.47 0.17 0.13 0.03 −0.22 0.13 0.47 −0.08 0.63 0.02 −0.07 0.02 0.43 −0.1 N/A

Reference Density (g cm−3) 1.05 1.05 1.05 1.05 1.05 1.05 1.05 0.25 0.25 1.03 1.03 1.03 1.03 1.03 1.03 1.05 1.05 1.05 1.05 1.03 1.05 1.03 1.05 0.97 1.05 1.05 0.45 1.04 1.06 2.75 1.04 1.04 1.03 1.05 1.05 1.03 1.03 1.03 1.04 1.04 0

Volume for RPI Voxel 2.7 mm (cm3) 102.02 36.41 7.26 104.67 37.36 7.48 1712.2 2220.7 2579.8 2.18 6.2 5.81 101.35 7.6 10.77 1159.7 ####### 2619.2 9546.31 38.85 133.63 0.57 16.34 ####### 40.67 40.67 7326.15 28.9 141.7 18.19 16.79 16.85 24.39 19.03 40.25 2.91 8.25 7.28 48.28 192.11 N/A

Handbook of Anatomical Models for Radiation Dosimetry

366

TABLE 14.5 Organ/Tissue Masses, Densities, and Volumes for the Female Phantom Masses (g) RPI Adult Female ID

Mesh-Based Phantom

Voxel-Based Phantom

ICRP Reference Data

Voxel versus Reference Error (%)

Reference Density (g cm−3)

Volume for RPI Voxel 2.5 mm (cm3)

1 2 3 5 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43

6.5 6.5 18.61 9.02 8 25 43.02 144.06 32.42 69.45 112.61 111.9 19.89 102.2 52.78 20.55 155.15 91.24 33.59 104.08 72.84 32.51 40.45 403.6 417.09 247.76 225.05 39.51 232.47 174.67 55.43 618.84 586.53 87.65 171.74 270.35 44.95 34.67 259.85 445.07 162.88

6.49 6.49 18.6 9.01 8 24.99 6.06 242.38 43.24 92.65 112.62 111.9 19.81 102.21 52.77 20.55 155.13 91.23 33.6 104.67 72.71 32.52 40.46 403.59 417.08 246.93 225.05 39.43 232.5 174.67 55.42 616.32 586.51 87.66 172.56 270.37 44.97 34.65 259.83 443.83 162.87

6.5 6.5 18.61 9 8 25 6.06 242.38 43.24 92.64 112.6 111.9 19.89 102.21 52.78 20.55 155.14 91.24 33.59 104.08 72.84 32.5 40.45 403.6 417.09 247.75 225.05 39.51 232.47 174.67 55.43 618.85 586.52 87.65 171.75 270.35 44.94 34.67 259.84 445.07 162.87

−0.21 −0.21 −0.04 0.08 −0.02 −0.03 0.01 0 0.01 0.02 0.01 0 −0.38 0 −0.01 0.01 −0.01 0 0 0.56 −0.17 0.07 0.02 0 0 −0.33 0 −0.21 0.01 0 −0.02 −0.41 0 0.02 0.47 0.01 0.06 −0.05 0 −0.28 0

1.02 1.02 1.03 1.05 1.03 1.03 1.06 1.06 1.06 1.06 1.92 1.18 0.98 1.92 1.12 0.98 1.92 1.12 0.98 1.92 1.12 1.92 1.19 1.92 1.2453 1.92 1.05 0.98 1.92 1.12 0.98 1.92 1.12 0.98 1.92 1.12 1.92 1.19 1.92 1.11 1.92

6.36 6.36 18.06 8.58 7.77 24.27 5.72 228.66 40.8 87.41 58.66 94.44 20.22 53.23 47.23 20.97 80.8 81.66 34.28 54.52 65.08 16.94 33.97 210.2 334.92 128.61 215.09 40.23 121.09 156.33 56.55 321 524.94 89.45 89.88 241.98 23.42 29.14 135.33 400.17 84.83

Mesh-Based and Anatomically Adjustable Adult Phantoms

367

TABLE 14.5 (continued) Organ/Tissue Masses, Densities, and Volumes for the Female Phantom Masses (g) RPI Adult Female ID

Mesh-Based Phantom

Voxel-Based Phantom

44 45 46 47 48 49 50 51 52 54 55 56 58 61 62 63 64 65 66 67 68 69 70 71 72 73 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89

258.96 120.44 96.86 70.88 72.81 203.77 252.56 154.62 261.28 140.44 1.86 47.42 313.61 1299.99 150.05 47.16 150 46.1 0.2 7.3 −0.2 7.3 8 48 140.01 229.99 879.99 90 100.01 55 60 55 30 90 50 45.01 80 24.99 250.01 370 104.64

258.96 120.45 96.86 70.89 72.79 203.76 252.55 154.62 261.27 140.46 1.68 47.41 313.6 1293.76 150 99.99 150 99.99 0.21 7.29 0.21 7.29 8.01 47.9 139.91 228.85 879.99 89.99 100 55.01 60 55.01 30.01 89.99 50.02 45.01 79.98 25.01 250 369.99 104.64

ICRP Reference Data 258.96 120.45 96.87 70.88 72.81 203.78 252.56 154.62 261.28 140.44 1.67 47.41 313.61 1300 150 100 150 100 0.2 7.3 0.2 7.3 8 48 140 230 880 90 100 55 59.99 55 30.01 90 50 45.01 79.99 24.99 250 370 104.63

Voxel versus Reference Error (%)

Reference Density (g cm−3)

0 0 −0.01 0.01 −0.02 −0.01 0 0 0 0.01 0.67 −0.02 0 −0.48 0 −0.01 0 −0.01 3.13 −0.13 3.13 −0.13 0.18 −0.22 −0.06 −0.5 0 −0.01 0 0.01 0 0.01 0.03 −0.01 0.03 0.01 −0.01 0.06 0 0 0

1.09 1.92 1.13 1.92 1.14 1.92 1.08 1.92 1.17 1.05 1.92 1.08 1.1 1.04 0.95 1.02 0.95 1.02 1.1 1.03 1.1 1.03 1.03 1.03 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.04 1.05 1.06 1.05

Volume for RPI Voxel 2.5 mm (cm3) 237.31 62.73 85.88 36.92 64.13 106.13 233.05 80.53 223.17 133.5 0.88 44.06 285.09 1244 157.89 98.03 157.89 98.03 0.19 7.08 0.19 7.08 7.78 46.5 134.53 220.05 846.14 86.53 96.16 52.89 57.69 52.89 28.86 86.53 48.09 43.28 76.91 24.05 238.09 349.05 99.66 (continued)

Handbook of Anatomical Models for Radiation Dosimetry

368

TABLE 14.5 (continued) Organ/Tissue Masses, Densities, and Volumes for the Female Phantom Masses (g) RPI Adult Female ID

Mesh-Based Phantom

90 91 92 93 94 95 97 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 119 120 121 125 126 127 128 131 132 133 134 135 136 137 138 139 140

37.37 7.48 87.87 31.38 6.28 1403.29 421.25 528.72 0.53 1.54 2.5 7.05 13.08 11.59 401.97 8518.19 1524.92 7054.94 35 5.5 5.5 120 0.6 N/A 24.72 24.69 N/A 28 130 40 20 17 50.98 3 7.5 7.5 40 200 80 N/A

Voxel-Based Phantom 37.36 7.46 87.87 31.37 6.27 1403.01 420.98 527.33 1.34 3.86 2.58 57.65 3.91 9.79 401.95 8518.24 1524.91 7054.93 35 5.51 5.49 120.1 0.6 24,343.37 34.99 34.99 2300 28 130 39.96 19.99 17.01 50.97 2.99 7.52 7.52 40.01 199.85 79.65 N/A

ICRP Reference Data

Voxel versus Reference Error (%)

37.37 7.48 87.87 31.38 6.28 1400 422 528 1.34 3.86 2.58 57.66 3.9 9.8 401.97 8518.23 1524.87 7054.94 35 5.5 5.5 120 0.6

−0.04 −0.17 0 −0.04 −0.15 0.22 −0.24 −0.13 0.05 −0.03 −0.03 −0.02 0.4 −0.16 0 0 0 0 0.01 0.16 −0.14 0.08 −0.75

35 35 2300 28 130 40 20 17 50 3 7.5 7.5 40 200 80 N/A

−0.02 −0.02 0 0 0 −0.1 −0.06 0.08 1.95 −0.22 0.21 0.21 0.02 −0.08 −0.44 N/A

Reference Density (g cm−3) 1.05 1.05 1.05 1.05 1.05 1.05 0.25 0.25 1.03 1.03 1.03 1.03 1.03 1.03 1.05 1.05 1.05 1.05 1.03 1.04 1.04 1.05 1.03 #DIV/0! 1.05 1.05 #DIV/0! 1.04 1.06 2.75 1.03 1.05 1.05 1.03 1.03 1.03 1.04 1.04 1.05 0

Volume for RPI Voxel 2.5 mm (cm3) 35.58 7.11 83.69 29.88 5.97 1336.2 1683.9 2109.3 1.3 3.75 2.5 55.97 3.8 9.5 382.81 8112.61 1452.3 6718.98 33.98 5.3 5.28 114.38 0.58 25,307.8 33.33 33.33 5822.55 26.92 122.64 14.53 19.41 16.2 48.55 2.91 7.3 7.3 38.47 192.16 75.86 N/A

Mesh-Based and Anatomically Adjustable Adult Phantoms

(a)

369

(b)

FIGURE 14.11 (See color insert following page 524.) 3D front and back views of the phantoms: (a) the male; (b) the female.

FIGURE 14.12 Software GUI for deformable phantom operations showing various tools and the female phantom whose breast size is specified by a user.

370

Handbook of Anatomical Models for Radiation Dosimetry

The primary features of the software package are summarized below: 3D visualization of the mesh files. In the 3D display function, a mesh file can be easily handled, either in the wireframe data format or the surface rendering format. By using mouse clicking and wheel scrolling operations, a user can rotate and zooming in and out of 3D rendered mesh model. Interactive organ deformation. Based on a user’s initial definition of deformation factors such as organ volume, the mesh deformation can be carried out automatically in about 3 min on a PC operated in Intel Pentium(R) with a 2.66 GHz CPU and a 3 GB RAM. Given the breast CS in the software menu, the female reference phantom can be automatically deformed and adjusted. Other automatic operations. The software menu comes with several automatic operations that are critical to the deformable phantom software program. Using an in-house voxelization algorithm, the mesh-based phantom is converted to a voxel-based phantom for the purposes of Monte Carlo radiation transport simulations. During the voxelization processing, the data of physical properties for the organs are translated and updated using the Microsoft Access technology. The voxelized data are then automatically exported to a geometry format acceptable to the MCNPX code.25 Organ properties are defined by a material library that can be easily updated through an interactive editing function of the software.

14.3 Applications and Discussion The deformable phantoms can be used to study various radiological procedures involving radiation protection, Computer Tomography (CT) imaging, and radiation treatment. In this section, we describe the application of a female adjustable phantom to the virtual calibration of a lung counter for female workers.30 14.3.1 Virtual Calibration of a Lung Counter Using Female Phantoms of Varying Breast Size A bioassay procedure for a worker who has been exposed to airborne radioactive materials requires the estimation of radioactivity burden inside the lung. A lung counter is an in vivo measuring device which consists of radiation detectors sensitive to photons emitted from the radionuclides in the lungs. Placed on or near the surface of the body, a lung counter can be used to quantify internally accumulated radionuclides. However, a calibration must be performed before the detector response can be directly correlated with the retained radioactivity inside the lungs. In most cases, the calibration can be performed using a physical lung phantom with known radioactivity and type of radionuclides, such the Lawrence Livermore National Laboratory Torso-Phantom that has removable chest plates.31 In the case of female workers, however, a “virtual calibration” method using computer phantoms is more effective because a deformable mesh-based phantom can be used to account for a variety of breast sizes and shapes. To demonstrate the usefulness of the deformable phantoms described earlier, we have developed a set of eight female phantoms with a range of different breast sizes. Figure 14.13 shows the lung counter positions in the

Mesh-Based and Anatomically Adjustable Adult Phantoms

371

virtual calibration with the central axis indicating the center-of-mass of the two lungs as simulated in the MCNPX. 14.3.2 Monte Carlo Simulations of Contaminated Worker and Phoswich Scintillation Detectors As reported in a separate paper,30 three radionuclides, Am-241, Cs-137, and Co-60 listed in Table 14.6, were chosen to investigate the energy dependence of the counting efficiency. The radionuclides were uniformly distributed throughout the entire lung volumes. There are several types of lung counters, one of which is the so-called Phoswich scintillation detector.34 The thickness of the NaI(Tl)-crystal is 1.0 mm and the diameter is 203.2 mm for low-energy photons. For high-energy photons, the CsI(T1)-crystal, which had a 50.8 mm thickness and 203.2 mm diameter, is coupled to the NaI(Tl)-crystal. Another nondoped NaI-crystal was located beyond the detector crystals. All crystals were embedded in the Al2O3 power, which was sealed airtight in the FIGURE 14.13 steel housing with a Beryllium front window. To The position of the lung counter in the virtual simplify the detector model, other parts of deteccalibration with the central axis indicating tor (e.g., photomultiplier) were not considered. The the center-of-mass of the two lungs defined in Gaussian Energy Broadening (GEB) in the MCNPX the MCNPX. (From Hegenbart, L. et al., Phys. Med. Biol., 53, 5527, 2008. With permission.) code was used to improve the realism of the simulated gamma spectra empirical data for Full Width at Half Maximum (FWHM) for photopeak was adopted as the GEB parameters. The energy range for Am-241 was 20–80 keV. For Cs-137, the energy range was set to ±1.25 FWMH around its peak energy of 661.6 keV. For Co-60, the energy range was also set to ±1.25 FWMH around its peak energies of 1.17323 and 1.33249 MeV. For each Monte Carlo simulation, a total of 5 million particle histories were run to yield relative statistic uncertainties (one standard deviation) less than 0.4%. 14.3.3 Results of Virtual Calibration of Lung Counting Efficiency for Female Workers To compare the results from these female worker phantoms with different CSs directly, the counting efficiencies of both detectors were summed and normalized to the efficiency of the AA-model. For larger breast CSs, the attenuation of the photons is greater and the counting efficiency is smaller. Am-241, which emits low-energy photons, shows the greatest difference in counting efficiencies. For the G-sized phantom, the efficiency was about 50% of that for the A-sized phantom. This value was about 58% for the high-energy photon emitters, Co-60, and about 55% for Cs-137. Figure 14.14 summarizes the normalized counting efficiencies and the CSs. The values from the left to right are the breast mass in grams corresponding to the CS of AA, A, B,

Handbook of Anatomical Models for Radiation Dosimetry

372

TABLE 14.6 Nuclides and Their Photon Energies Used for the Virtual Calibration of the Lung Counter Nuclide

Photon Energy (MeV)

Emission Probability

Am-241

0.01376 0.0139 0.01754 0.02101 0.02634 0.05954 0.03182 0.03219 0.0364 0.6616 1.17323 1.33249

0.0108 0.1193 0.1861 0.0482 0.024 0.359 0.0195 0.0359 0.01055 0.85 0.9985 0.99983

Cs-137

Co-60

Source: Hegenbart, L. and Heide, B. Numerische Effizienzkalibrierung bei In-vivo-Messverfahren mittels an den Probanden angepassten Voxelmodellen, FZKA7330, Central Safety Department Annual Report, 2006. Note: Only photons with emission probabilities greater than 0.01 were considered.

1.00

Relative efficiency to AA-model

0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 0.45 250

500

750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250 3500 3750 4000

Total breast mass (g) FIGURE 14.14 Relative counting efficiencies (counts per particle history) of the three nuclides, Am-241 (squares, dotted line), Cs-137 (diamonds, solid line), Co-60 (triangles, dashed line), normalized by the AA-model as a function of the total breast mass. The effects of chest attenuation on the counting efficiency and potential improvement to the use of one-size physical phantom are demonstrated in this “virtual calibration.” (From Hegenbart, L. et al., Phys. Med. Biol., 53, 5527, 2008. With permission.)

Mesh-Based and Anatomically Adjustable Adult Phantoms

373

C, D, E, E40, F, and G. The relationship for each radionuclide can be approximated with the second-order polynomial fit in the Microsoft Excel 2008 with R 2-values of at least 99.2%. This application clearly shows that the virtual calibration with the unique female breast phantoms will allow the uncertainty caused by unknown chest attenuation in calibrations performed with a one-size physical phantom to be accurately quantified and corrected.30

14.4 Conclusions In this chapter, we have described the development and application of a pair of deformable adult male/female phantoms that are based entirely on polygonal meshes. Such phantoms contain two components: (1) a set of reference phantoms whose organ volumes and masses agree with those recommended by the ICRP Publication 89 for reference individuals and (2) a software program that automatically carries out a series of necessary operations such as deformation, collision detection, visual inspection, as well as voxelization for Monte Carlo calculations. To demonstrate the concept of a deformable phantom, this software was used to morph the mesh-based female phantom into a series of female phantoms having different breast sizes for the purposes of performing a “virtual calibration” of lung counting. This software is currently being refined by adding additional registration and deformation algorithms so the software can eventually model individuals in various weight/height percentiles. With these advanced software algorithms, we are also developing phantoms of adjustable postures such as a patient whose arms are raised during a medical imaging or treatment procedure (Figure 14.15a), a walking and sitting worker (Figures 14.15b and c) in a normal or accidental radiation environment. These phantoms

(a)

(b)

(c)

FIGURE 14.15 Selected postures of the deformable phantoms. (a) Raised arms; (b) walking; and (c) sitting.

374

Handbook of Anatomical Models for Radiation Dosimetry

must be voxelized for Monte Carlo calculations because existing codes do not directly handle the mesh format. Our experience has shown that the mesh-based data structure has clear advantages in designing deformable and adjustable phantoms. The deformation algorithm is relatively easy to develop and the computation is efficient and affordable. Furthermore, the anatomical details can be accurately preserved during the automated adjustment—a feature with the NURBS structures sometimes would compromise. However, the NURBS is generally faster to simulate and can be useful for real-time cardiac and respiratory motions. With advanced registration algorithms and software design, it is possible to eventually morph the entire mash-based reference adult male/female phantoms into a new patient or worker whose anatomical information is limited (e.g., body mass index or partial-body imaging data). Such organ and posture adjustable phantoms will make it possible to reduce dosimetric uncertainties caused by the use of a single reference phantom of fixed anatomy and posture.

Acknowledgments This project was funded by a grant from the National Cancer Institute (R01CA116743). Monte Carlo simulations of lung counting efficiency were performed by Mr. Lars Hegenbart, a visiting student at RPI from Karlsruhe Institute of Technology, Karlsruhe, Germany. The authors are grateful to Mr. Matt Mille for his helpful comments on the manuscript.

References 1. Wikipedia. http://en.wikipedia.org/wiki/Wikipedia, Last accessed August 2007. 2. Geant4 Team. Geant4 User’s Guide for Application Developers, http://geant4.web.cern.ch/ geant4/G4UsersDocuments/UsersGuides/ForApplicationDeveloper/html, Last accessed August 2007. 3. Agostinelli, S. et al. GEANT4—A simulation toolkit, Nuclear Instruments & Methods in Physics Research, Section A, Accelerators Spectrometers Detectors and Associated Equipment, 506, 250, 2003. 4. Leyton, M. A Generative Theory of Shape, London: Springer-Verlag, 2001. 5. Stroud, I. Boundary Representation Modelling Techniques, London: Springer-Verlag, 2006. 6. Xu, X.G. et al. A boundary-representation method for designing whole-body radiation dosimetry models: Pregnant females at the ends of three gestational periods—RPI-P3, -P6 and -P9, Physics in Medicine and Biology, 52, 7023, 2007. 7. ICRP. Report of the Task Group on Reference Man, Oxford: Pergamon Press, 1975. 8. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, Oxford: Pergamon Press, 2003. 9. Xu, X.G., Zhang, J.Y., and Na, Y.H. Preliminary data for mesh-based deformable phantom development: Is it possible to design person-specific phantoms on-demand?, The International Conference on Radiation Shielding-11, Pine Mountain, GA, 2008. 10. 3D Anatomy Model Sets, http://www.anatomium.com, Last accessed August 2007. 11. Xu, X.G. et al. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Physics, 78, 476, 2000.

Mesh-Based and Anatomically Adjustable Adult Phantoms

375

12. Schlattl, H. et al. Organ dose conversion coefficients for voxel models of the reference male and female from idealized photon exposures, Physics in Medicine and Biology, 52, 2123, 2007. 13. Beall, M.W. and Shephard, M.S. A general topology-based mesh data structure, International Journal for Numerical Methods in Engineering, 40, 1573, 1997. 14. Jain, V. and Zhang, H. A spectral approach to shape-based retrieval of articulated 3D models, Computer-Aided Design, 39, 398, 2007. 15. Ohbuchi, R. et al. Retrieving 3D shapes based on their appearance, in Proceedings of the 5th ACM SIGMM Workshop on Multimedia Information Retrieval (MIR), Berkeley, CA, 2003. 16. Ohanian, O.J. Mass properties in VTOL UAV conceptual design software, Part I: Overview and general algorithms, in Proceedings of the 64th Annual SAWE Conference, Annapolis, MD, May 16–18, 2005. 17. Zhang, C. and Chen, T. Efficient feature extraction for 2D/3D objects in mesh representation, in Image Processing, 2001. Proceedings of the 2001 International Conference on 2001, Thessaloniki, Greece, 2001. 18. Deuflhard, P. Newton Methods for Nonlinear Problems. Affine Invariance and Adaptive Algorithms. Springer Series in Computational Mathematics, Vol. 35, Berlin: Springer, 2004. 19. Amanatides, J. and Choi, K. Ray tracing triangular meshes, in Proceedings of the Eighth Western Computer Graphics Symposium, Whistler, BC, Canada, 1997. 20. The MathWorks®. MATLAB™ and Simulink for Technical Computing, http://www.mathworks. com/, Last accessed August 2007. 21. European Committee for Standardization (CEN). Size designation of clothes—Part 1: Terms, definitions and body measurement procedure, EN 13402-1, 2001. 22. European Committee for Standardization (CEN). Size designation of clothes—Part 2: Primary and secondary dimensions, EN 13402-2, 2002. 23. European Committee for Standardization (CEN). Size designation of clothes: Part 3. Measurements and intervals, EN 13402-3, 2004. 24. Nooruddin, F.S. and Turk, G. Simplification and repair of polygonal models using volumetric techniques, IEEE Transactions on Visualization and Computer Graphics, 9, 191, 2003. 25. Pelowitz, D.B. MCNPX User’s Manual Version 2.5.0, Los Alamos National Laboratory Report LA-CP-05-0369, 2005. 26. ICRU. Photon, electron, proton and neutron interaction data for body tissues, Report 46, Bethesda, MD: International Commission on Radiation Units and Measurements, 1992. 27. Microsoft® Visual C#, http://msdn.microsoft.com/en-us/vcsharp/default.aspx, Last accessed August 2007. 28. http://www.microsoft.com/com/default.mspx, Last accessed August 2007. 29. The Visualization ToolKit (VTK) http://www.vtk.org/, Last accessed August 2007. 30. Hegenbart, L. et al. A Monte Carlo study of lung counting efficiency for female workers of different breast sizes using deformable phantoms, Physics in Medicine and Biology, 53, 5527, 2008. 31. Griffith, R.V. et al. Tissue equivalent torso phantom for intercalibration of in vivo transuranic nuclide counting facilities, in Advances in Radiation Protection Monitoring, STI/PUB/494 Proceedings of the IAEA Conference, IAEA-SM-229/56, Vienna: IAEA, 1978. 32. Hegenbart, L. and Heide, B. Numerische Effizienzkalibrierung bei In-vivo-Messverfahren mittels an den Probanden angepassten Voxelmodellen, FZKA7330: Central Safety Department Annual Report, 2006. 33. Schötzig, U. and Schrader, H. Halbwertszeiten und Photonen-Emissionswahrscheinlichkeiten von häufig verwendeten Radionukliden, PTB-Bericht PTB-Ra-16/5: erweiterte und korrigierte Auflage, Braunschweig, Germany, 2000. 34. Doerfel, H. and Heide, B.S., M. Entwicklung eines Verfahrens zur numerischen Kalibrierung von Teilkörperzählern, FZKA 7238: Wissenschaftliche Berichte Forschungszentrum Karlsruhe, 2006.

15 The ICRP Reference Computational Phantoms Maria Zankl, Keith F. Eckerman, and Wesley E. Bolch

CONTENTS 15.1 Introduction ............................................................................................................... 377 15.2 Phantom Properties Requested by the ICRP ........................................................ 378 15.3 Construction of Reference Computational Phantoms from Existing Segmented Data ........................................................................................................ 379 15.4 Description of the ICRP Adult Reference Male and Female Computational Phantoms ........................................................................................ 381 15.4.1 General Features .......................................................................................... 382 15.4.2 Special Features of the Skeleton ................................................................384 15.4.3 Limitations .................................................................................................... 385 15.5 Applications in Radiation Dosimetry Planned by the ICRP .............................. 386 15.6 Conclusions ................................................................................................................ 387 Acknowledgments ............................................................................................................... 387 References ............................................................................................................................. 388

15.1 Introduction Computational phantoms of the human body—together with radiation transport codes— have been used for the evaluation of organ dose conversion coefficients in occupational, medical, and environmental radiation protection. During the last two decades, voxel computational phantoms were introduced that are derived mostly from (whole body) medical image data of real persons instead of the older mathematical Medical Internal Radiation Dose (MIRD)-type body computational phantoms. Among other laboratories, the Helmholtz Zentrum München—German Research Center for Environmental Health (i.e., the former GSF—National Research Center for Environment and Health) has developed 12 voxel phantoms of individuals of different stature and ages: 2 pediatric ones, 4 male, and 6 female adult computational phantoms (see also Chapter 3).1–4 It was shown that the schematic organ shapes of the MIRD-type phantoms presented an oversimplification, having an influence on the resulting dose coefficients, which in some cases deviate systematically from those calculated for voxel computational phantoms. The parameters influencing the organ doses for external radiation are mainly (1) the depth of the organ below the body surface, (2) the exterior shape of the trunk, and (3) the trunk diameter relative to the incoming radiation beam. For internal dosimetry, the influencing parameters are (1) the relative position of source and target organs (for organ cross fire) 377

378

Handbook of Anatomical Models for Radiation Dosimetry

and (2) the organ mass (for organ self-absorption). Hence, the International Commission on Radiological Protection (ICRP) decided to use voxel phantoms for the update of organ dose conversion coefficients, following the recent revision of the ICRP Recommendations.5 According to the ICRP, these voxel computational phantoms should be representative of the adult Reference Male and Reference Female6 with respect to their external dimensions, their organ topology, and their organ masses. To meet these requirements, the Helmholtz Zentrum München developed two voxel adult reference phantoms of a male and a female upon the request of the ICRP that are based on the voxel phantoms of two individuals whose body height and weight resembled the reference values. Approximately 140 organs and tissues were segmented, including objects that were not previously contained in the MIRD-type phantoms, such as the main blood vessels, cartilage, muscles, and lymphatic nodes. The organ masses of both computational phantoms were adjusted to the ICRP reference values, without spoiling their realistic anatomy. These phantoms are used by the ICRP in establishing radiation protection guidance, for example, effective dose coefficients and other secondary dosimetric quantities. They are referred to as ICRP-AM and ICRP-AF for the adult Reference Male and adult Reference Female, respectively.

15.2 Phantom Properties Requested by the ICRP In order to be able to use the voxel phantoms for calculations of organ dose conversion coefficients and specific absorbed fractions (SAFs) following the new ICRP Recommendations5 it is obvious that these phantoms should accommodate all organs that have been identified as subject to radiation injury (both stochastic and deterministic) and that contribute to the quantity “effective dose” or are protected by separate dose limits in occupational radiation protection or radiation protection of the public. Furthermore, additional target regions have been considered that have been identified in the Human Respiratory Tract7 and Human Alimentary Tract8 Models. When radioactive material is incorporated in the body, it is accumulated in certain organs, tissues, and body regions that are involved in the distribution, transport, and excretion of substances. These regions that accumulate radioactivity become thus “source organs” that irradiate the radiation-sensitive “target organs.” Most of the possible source regions are also target regions and as such are already included due to the above considerations. Additional source regions are located in the alimentary and respiratory tracts as well as in the skeleton. These additional source regions are, e.g., the oral cavity, the stomach contents, the intestine contents, the gallbladder content, the urinary bladder content, the surface of anterior and posterior nasal passages and the pharynx, bronchi, bronchioles, the blood vessels of the head, the trunk, the legs and the arms, cortical and trabecular bone mineral, as well as cortical and trabecular bone marrow. Due to the limited resolution of the medical image data used to construct the voxel phantoms (in the range of millimeters) and the small dimensions of some source and target tissues (tens of micrometers), not all of them could be segmented directly. Therefore, for some source and target tissues, “surrogate” regions had to be found or correction factors had to be applied to the calculated doses. These limitations of the phantoms are discussed in more detail below. It was requested that the height of the phantoms and the masses of all organs and tissues should possibly be in accordance with the reference data of ICRP Publication 89.6

The ICRP Reference Computational Phantoms

379

15.3 Construction of Reference Computational Phantoms from Existing Segmented Data In order to construct voxel phantoms of the Reference Male (respectively, Reference Female), appropriate voxel phantoms were used as starting points, i.e., computational phantoms with external dimensions close to the ICRP reference values6 since then the required modifications remained moderate, and the dangers of distorting the anatomical relations were small. As a basis for the male reference computational phantom, the male adult voxel phantom “Golem”9 was used that was constructed from whole-body CT images of a 38-year-old single individual patient who had a height of 176 cm and a whole body mass of slightly below 70 kg (Reference Male: 176 cm; 73 kg). The data set consisted of 220 slices of 256 × 256 pixels. The original voxel size was 8 mm in height with an in-plane resolution of 2.08 mm, resulting in a voxel volume of 34.6 mm3. Hundred and twenty-two individual objects were segmented (67 of these being bones or bone groups), including many—but not all—of the organs and tissues later identified in the ICRP reference anatomical data.6 Golem had been segmented at the end of the last decade, and contained nearly all organs that were then considered to be relevant for dose calculations in radiation protection, i.e., those organs that contributed to the evaluation of the effective dose, the quantity considered to be approximately characterizing the overall radiation risk from a certain radiation exposure, as defined in the previous ICRP Recommendations.10 Recently, segmentation of further structures took place to meet the new requirements. Due to the limited resolution of the image data (8 mm slice thickness) it was difficult to identify small structures, such as blood vessels much smaller than the large main vessels in the trunk. Therefore, only a small proportion of the blood pool could be segmented. Furthermore, since no cartilage had been considered in the original segmented computational phantom, and due to the limited dosimetric importance of this tissue, the effort toward its supplementary segmentation was further limited. The female reference voxel computational phantom was constructed on the basis of the voxel computational phantom “Laura.” The primary data were derived from a high-resolution whole body CT scan of a 43-year-old patient with 167 cm height and 59 kg of weight (corresponding ICRP reference values: 163 cm and 60 kg). The data set consisted of 174 slices of 5 mm width (head and trunk) and 43 slices of 2 cm width (legs), each with 256 × 256 pixels. The 2 cm slice images were resampled to result also in slices of 5 mm width, using data interpolation rather than simply repeating identical slices. The resulting data set consisted of 346 slices. The voxel size was then 5 mm height with an in-plane resolution of 1.875 mm, and this corresponds to a voxel volume of 17.6 mm3. A total of 88 objects were primarily segmented in Laura, and the number of different bone sites was 19.4 The following steps were then followed: (1) an adjustment of the body height and the skeleton mass of the segmented computational phantom to the reference data by voxel scaling; (2) an adjustment of the single organ masses to the reference values by adding or subtracting a respective number of organ voxels; and (3) an adjustment of the whole body mass to the reference values by adding or subtracting a respective number of adipose tissue voxels. We intended to keep the modifications to the skeleton shape to a minimum, so as to preserve the “frame” of the body. It was, however, not possible for both computational phantoms to accommodate the entire brain mass within the skull. Therefore, it was necessary to increase the volume of the skull. Golem had a noticeably narrow head, and other organs in the head were also small compared to the reference values. Therefore, we

Handbook of Anatomical Models for Radiation Dosimetry

380

decided to increase the voxel size of all the voxels of the entire head, and then resample this volume with the smaller voxel size of the rest of the body. This means that the reference phantom has a greater number of head voxels than Golem had. For the female reference phantom, only the skull was increased: the interior surface voxels of the skull were replaced by brain, and an additional layer of skull voxels was then added at the exterior surface. In order not to lose a layer of the surrounding tissues, this had to be preceded by an outward movement of the surrounding muscle, adipose tissue, and skin voxels. Furthermore, an outward movement of the female phantom’s ribs (as occurring also during breathing) was necessary as well to accommodate the liver. This was also done while the thickness of tissues covering the ribs—muscle, adipose tissue, and skin—was preserved again. Apart from these unavoidable modifications of the skeleton shape, the volume of the skeleton was adjusted to the reference value by voxel scaling. Since Golem’s body height corresponds to the reference value, the original voxel height was kept unmodified. Laura was taller than the ICRP adult Reference Female, so the voxel height for the female reference voxel computational phantom was reduced from 5.0 to 4.84 mm. After the mentioned moderate changes to the skulls of both phantoms, the numbers of segmented skeleton voxels (including the segmented cartilage) were 211,427 and 378,204 for the male and female reference voxel phantom, respectively. Table 15.1 shows the skeleton volumes of the ICRP adult Reference Male and Reference Female as derived from the reference mass data from ICRP Publication 89 and mass density data of ICRU Report 46.6,11 Voxel volumes of 36.54 and 15.25 mm3 for the male and female reference voxel phantoms were evaluated from these values and the segmented skeleton voxel numbers, by dividing the volume values aimed at by the respective numbers of skeleton voxels for both phantoms. The voxel height being fixed already, this resulted in voxel in-plane resolutions of 2.137 mm for the male and 1.775 mm for the female phantom, respectively. Compared to the original voxel sizes of Golem and Laura, this corresponds to an increase by 5.6% in voxel volume for the male reference voxel phantom and a decrease by 13.2% for the female reference voxel phantom.

TABLE 15.1 Reference Mass Values,6 Mass Density Values11 and Volumes Derived for Various Constituents of the Skeleton of the Adult ICRP Reference Individuals Mass (g)

Mineral bone Cortical bone Trabecular bone Cartilage Red bone marrow Yellow bone marrow Miscellaneous Total

Male

Female

5,500 4,400 1,100 1,100 1,170 2,480

4,000 3,200 800 900 900 1,800

200 10,450

160 7,760

Mass Density (g cm−3)

Volume (cm3) Male

Female

1.10 1.03 0.98

2,864.6 2,291.7 572.9 1,000.0 1,135.9 2,530.6

2,083.3 1,666.7 416.7 818.2 873.8 1,836.7

1.03 1.35

194.2 7,725.3

155.3 5,767.4

1.92

Source: Zankl, M. et al., Radiat. Prot. Dosim., 127, 174, 2007. With permission.

The ICRP Reference Computational Phantoms

381

Following the preparations described above, we performed an individual adjustment of the single organs. For this purpose, a software package called “VolumeChange” was designed.12 It uses the programming language IDL (Interactive Data Language) and represents each organ by its surface voxels, i.e., all voxels having at least one neighbor that does not belong to the same organ. The volumes are then modified by shifting surface voxels— inward for decreasing, outward for increasing the respective volume. The individual organs were adjusted to the respective reference values one by one, beginning with those that were larger than reference size in order to make room for those that had to be enlarged. Some fi ne structures could not be adjusted exactly to the reference values, due to limitations of voxel resolution and visibility. For most organs, however, a close approximation of the reference values could be achieved. The only limitation then was due to the fact that each organ has to consist of an integer number of voxels. That means that the resulting volumes may deviate from the value aimed at by at most half a voxel volume, i.e., ∼18.3 mm3 for the ICRP-AM and 7.6 mm3 for the ICRP-AF. At this stage, further anatomical details were segmented in the reference computational phantoms, going back to the original CT images from which Golem and Laura had been segmented. Some effort was made to identify a larger amount of blood vessels, which was especially demanding for the male phantom, due to the relatively large slice thickness, which resulted in a decreased detectability of fine structures. Furthermore, lymphatic nodes were incorporated into the phantoms. Since these objects could not be identified on the medical images, they were drawn manually, at locations specified in anatomical textbooks.13–16 Only a part of the lymphatic tissue reference mass was thus introduced; the distribution throughout the body and higher concentration at the specified locations was however correctly mirrored, such as in the groin, the axillae, etc. and to a certain extent also in the hollows of the knees and the crooks of the arms. When the individual organs had been adjusted to their reference mass values and additional structures had been incorporated, the internal anatomy was fixed. The final step was then to adjust the whole body masses to 73 and 60 kg for the male and female reference voxel computational phantoms, respectively. In both cases, the whole body masses were lower than the value aimed for, so the body had to be “wrapped” with additional layers of adipose tissue. Toward the end of this procedure, small iterations had to be made since each modification of the number of adipose tissue voxels resulted also in small changes to the skin mass, because the number of body surface voxels was modified. Finally, the whole body masses were adjusted to the reference values within 0.01 g.

15.4 Description of the ICRP Adult Reference Male and Female Computational Phantoms To clearly distinguish the “reference” voxel computational phantoms from the computational phantoms from which they originate, new names were given to them: for the male phantom, the name chosen initially by the developers was “Rex” (Reference adult male voxel computational phantom; Rex is also the Latin word for “king”), and to the female phantom, the corresponding female name—“Regina” (Latin for “queen”)—was given.17,18 The ICRP did not adopt these names, and chose ICRP Adult Male (ICRP-AM) and ICRP Adult Female (ICRP-AF) instead.

Handbook of Anatomical Models for Radiation Dosimetry

382

15.4.1 General Features A graphical representation of both reference adult voxel phantoms is shown in Figure 15.1; and their main characteristics are summarized in Table 15.2. A summary of all relevant source and target tissues, their segmented volumes and resulting masses of the reference adult male and female phantoms are given in Table 15.3.

FIGURE 15.1 (See color insert following page 524.) Frontal view of the ICRP-AM (left) and ICRP-AF (right), the voxel models representing the ICRP adult Reference Male and Reference Female.

TABLE 15.2 Main Characteristics of the Adult Reference Male and Reference Female Computational Phantoms Property Height (cm) Weight (kg) Number of (nonzero) voxels (millions) Slice thickness (voxel height, mm) Voxel in-plane resolution (mm) Voxel volume (mm3) Number of columns Number of rows Number of slices

ICRP-AM

ICRP-AF

176 73.0 1.95

163 60.0 3.89

8.0 2.13714 36.54 254 127 220

4.84 1.775 15.25 299 137 346

The ICRP Reference Computational Phantoms

383

TABLE 15.3 List of Source and Target Organs and Tissues, Their Segmented Volumes and Resulting Masses, Compared to the Reference Masses ICRP-AM Organ Adrenals Blood Bones (total skeleton) Brain Breast Eyes Gallbladder Gallbladder wall Gallbladder contents Gastrointestinal tract Stomach wall Stomach contents Small intestine wall Small intestine contents Right colon wall Right colon contents Left colon wall Left colon contents Recto-sigmoid colon wall Sigmoid colon contents Heart Heart wall Heart contents (blood) Kidneys Liver Lungs Lymphatic tissue Muscle tissue Esophagus Ovaries Pancreas Pituitary gland Prostate Residual (adipose) tissue Salivary glands Skin Spleen Teeth Testes Thymus Thyroid

ICRP-AF

Volume (cm3)

Mass (g)

Reference Mass (g)

Volume (cm3)

Mass (g)

Reference Mass (g)

13.6 973.7 7,725.3 1,381.0 25.6 14.3 66.0 13.5 52.5

14.0 1,032.1 10,450.0 1,450.0 25.0 15.0 68.0 13.9 54.1

14 5,600 10,450 1,450 25 15 68 10 58

12.6 807.4 5,767.4 1,238.1 511.9 14.3 54.3 9.9 44.4

13.0 855.8 7,760.1 1,300.0 500.0 15.0 56.0 10.2 45.8

13 4,100 7,760 1,300 500 15 56 8 48

144.2 240.4 625.0 336.6 144.2 144.3 144.2 72.1 67.3 72.1 795.4 314.3 481.1 295.3 1,714.3 2,891.3 134.0 27,619.0 38.8

150.0 250.0 650.0 350.0 150.0 150.0 150.0 75.0 70.0 75.0 840.0 330.0 510.0 310.0 1,800.0 1,200.0 138.0 29,000.0 40.0

150 250 650 350 150 150 150 75 70 75 840 330 510 310 1,800 1,200 730 29,000 40

133.3 0.6 16.5 21,535.2 82.5 3,420.2 144.2 18.2 33.7 24.3 19.2

140.0 0.6 17.0 20,458.4 85.0 3,728.0 150.0 50.0 35.0 25.0 20.0

140 0.6 17 18,200 85 3,300 150 50 35 25 20

134.6 221.2 576.9 269.2 139.4 153.8 139.4 76.9 67.3 76.9 587.2 238.1 349.1 261.9 1,333.3 2,300.8 76.8 16,666.7 34.0 10.6 114.3 0.6

140.0 230.0 600.0 280.0 145.0 160.0 145.0 80.0 70.0 80.0 620.0 250.0 370.0 275.0 1,400.0 950.0 79.1 17,500.0 35.0 11.0 120.0 0.6

140 230 600 280 145 160 145 80 70 80 620 250 370 275 1,400 950 600 17,500 35 11 120 0.6

24,838.3 68.0 2,496.8 125.0 14.6

23,596.4 70.0 2,721.5 130.0 40.0

22,500 70 2,300 130 40

19.4 16.4

20.0 17.0

20 17 (continued)

384

Handbook of Anatomical Models for Radiation Dosimetry

TABLE 15.3 (continued) List of Source and Target Organs and Tissues, Their Segmented Volumes and Resulting Masses, Compared to the Reference Masses ICRP-AM Organ Tongue Tonsils Ureters Urinary bladder wall Urinary bladder contents Uterus Total body

Volume (cm3)

Mass (g)

ICRP-AF Reference Mass (g)

69.5 2.9 15.5 48.1 192.3

73.0 3.0 16.0 50.0 200.0

73 3 16 50

71,109.9

73,000.0

73,000

Volume (cm3)

Mass (g)

57.1 2.9 14.6 38.5 192.3 77.7 59,258.0

60.0 3.0 15.0 40.0 200.0 80.0 60,000.0

Reference Mass (g) 60 3 15 40 80 60,000

Source: ICRP, Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, U.K., 2003.

15.4.2 Special Features of the Skeleton The skeleton is a highly complex structure of the body, composed of cortical bone, trabecular bone, red and yellow bone marrow, cartilage, and endosteal tissues (bone surfaces). The internal dimensions of most of these tissues are smaller than the resolution of the reference computational phantoms and, thus, these volumes could not be segmented in these voxel phantoms. Therefore, the skeletal dosimetry has to be based on the use of fluence-todose response functions that are multiplied with the particle fluence inside specific bone regions to give the dose quantities of interest to the target tissues. This methodology was also used in previous calculations employing the schematic mathematical phantoms.19 Nevertheless, an attempt was made to represent the gross spatial distribution of the source and target volumes in the voxel computational phantoms as realistically as possible at the given voxel resolution.18 Therefore, the skeleton was divided into those 19 bones and bone groups for which individual data on red bone marrow content and marrow cellularity are given in ICRP Publication 70.20 These bones are the upper halves of humeri, the lower halves of humeri, the lower arm bones (ulnae and radii), the wrists and hand bones, the clavicles, the cranium, the upper halves of femora, the lower halves of femora, the lower leg bones (tibiae, fibulae, and patellae), the ankles and the foot bones, the mandible, the pelvis, the ribs, the scapulae, the cervical spine, the thoracic spine, the lumbar spine, the sacrum, and the sternum. These individual bones were subsegmented into an outer shell of cortical bone and the enclosed spongious part of the bone. The long bones contain a medullary cavity as third component; this is again enclosed by cortical bone. This subdivision resulted in 44 different identification numbers in the skeleton: two—cortical bone and spongiosa—for each of the 19 bones mentioned above, and a medullary cavity for each of the six long bones (upper and lower halves of humeri, lower arm bones, upper and lower halves of femora, and lower leg bones). Furthermore, the amount of cartilage that could be identified on the CT images and could, thus, be segmented directly, was attributed to four body parts—head, trunk, arms, and legs. Hence, the skeleton covers a total of 48 individual identification numbers. The total volume of each bone results directly from the segmented number of voxels and the voxel volume. The cortical shell around the spongiosa was chosen to be one voxel layer; the cortical bone at the long bones’ shafts is thicker, and its thickness was adjusted

The ICRP Reference Computational Phantoms

385

such that the total cortical bone volume is in agreement with the reference value. For some bones, e.g., the sternum of the ICRP-AM and the sacrum and sternum of the ICRP-AF, the total bone marrow volume that is required to accommodate the ICRP reference volumes20 of red and yellow bone marrow (see below) was only marginally less than the segmented total volume of these bones. Consequently, there was not enough space left in these bones for an enclosing cortical shell. Cortical bone voxels around the sternum were, therefore, segmented for only a few slices of both computational phantoms. Although it would have been possible to spare a few voxels to accommodate a bit of cortical bone around the ICRPAM’s sacrum, this option was abandoned in order not to exceed the reference value of 1666.7 cm3 for the total cortical bone volume. Also, in the ribs of both computational phantoms, the space for cortical bone was limited; here only the larger portions are enclosed in a cortical shell; smaller parts—especially those with a large surface-to-volume ratio—are not given a cortical shell. For each of the 19 bones, various proportions of trabecular bone, red bone marrow, and yellow bone marrow compose the spongiosa. Furthermore, the additional volumes of “miscellaneous”6 and the not-directly segmented cartilage had to be accommodated in the skeleton; for practicability, these were merged within the spongiosa volume of all skeletal sites. In ICRP Publications 7020 and 89,6 reference data are given for the total masses of red (i.e., active) and yellow (i.e., inactive) bone marrow, the percentage distribution of the red bone marrow among individual bones, and the bone marrow cellularity in individual bones, based on earlier data by Cristy.21 Further data on the bone marrow distribution, especially of the yellow bone marrow, are not available. The volume of red bone marrow in each of the 19 bone groups can be calculated from the reference values of the total amount of red bone marrow and its percentage distribution. The bone marrow cellularity in an individual bone gives the proportion of the entire marrow in this bone that is still hematopoietic active: the red bone marrow fraction. From this value, the total bone marrow volume in that bone can be calculated. This permits the evaluation of the volume of yellow marrow. Of course, this is only possible for bones with a marrow cellularity that is nonzero. Those bones having zero cellularity do not contain red bone marrow, and the yellow marrow content cannot be estimated by this method. Therefore, the difference remaining between the total inactive marrow volume6,20 and that assigned to individual bones using the non ero cellularity values was distributed among those bones that do not contain red bone marrow, including the lower halves of humeri and femora, the lower arm and leg bones, and the hand and foot bones. A portion of the remaining yellow bone marrow is contained within the segmented medullary cavities of the humeri and femora and the lower arm and leg bones; the rest was attributed to the spongiosa regions of all zero-cellularity bones in relation to their respective volumes. Accordingly, each of the 19 bones or bone groups has its own unique bone-specific spongiosa composition. 15.4.3 Limitations During the process of adjustment to the reference values, some problems were encountered. First, the original intention to modify the skeleton exclusively by scaling could not be followed since in both cases it was not possible to accommodate the required brain mass inside the segmented skull. Therefore, the skulls of both phantoms had to be increased a bit, and for the female reference phantom, the ribs had to be moved slightly outwards—as during breathing—to create enough space for the liver. Second, since the image data used to create the phantoms have been acquired while the patients were in supine posture, the anatomy of the resulting voxel computational

386

Handbook of Anatomical Models for Radiation Dosimetry

phantoms had to be that of a lying person. That means that the abdomen is flatter than in a standing person, the abdominal organs shift upwards toward the chest, and the lungs are compressed. Furthermore, the curvature of the spine is slightly different from a person in upright posture. Although this effect is qualitatively obvious, there is little or no quantitative information on the amount of positional changes of individual organs, since usually no comparable examinations of the same person in different positions are available. In the only known study on these effects, an influence of posture on SAFs mainly for photon energies below 30 keV was found.22 Thus, believing that the dosimetric impact of the person’s position would be limited, the ICRP agreed to accept the voxel computational phantom geometry as-is, i.e., in the supine position. Consequently, the lung masses were adjusted to the reference values by increasing the density compared to the ICRU density value for a fully inflated lung. Further limitations of the resulting reference voxel computational phantoms are due to the fact that the dimensions of some tissues are smaller than the resolution of the available image data and could therefore • Not be identified on the images (e.g., medium and small blood vessels, fine bronchial structures, and lymphatic tissue) • Not be explicitly defined (e.g., the marrow cavities and the endosteum layer lining these cavities, the dimensions of which are only tens of micrometers) • Not be represented in their true size (like the mucous membranes of the extrathoracic airways) Since (1) not all small tissues could be exactly adjusted to their reference values; (2) the reference values of ICRP Publication 896 are rounded values; and (3) the adipose tissue was used to exactly adjust the whole body masses to the reference values, the resulting adipose tissue masses are 10% and 5% higher for the male and female phantoms, respectively, than the corresponding reference values. It should, furthermore, be clear that the reference computational phantoms still have their individual anatomical relations, although the organ masses have been adjusted to the reference values. For example, the arms of the ICRP-AM and ICRP-AF have slightly different orientations, they are both in supine posture, and there are more details that may deviate from any other person, even if such persons would exist with the same organ masses.

15.5 Applications in Radiation Dosimetry Planned by the ICRP The immediate plans of the ICRP involving usage of the male and female reference computational phantoms are dose calculations for the following topics: • • • •

Internal photon and neutron SAF values Dose conversion coefficients for external radiations Occupational intakes of radionuclides Interpretation of bioassay data

Subsequently, the phantoms will be released to the scientific community, and a more exhaustive description of the ICRP-AM and ICRP-AF will be issued by the ICRP in form of

The ICRP Reference Computational Phantoms

387

an ICRP Publication.23 When released to the scientific community, they can be applied not only to studies on ionizing radiation. There is already a wide field of dosimetric applications of voxel phantoms for nonionizing radiation, such as electromagnetic fields, radiation from mobile phones, etc. Voxel phantoms are the most developed, complete, and realistic computational phantoms of the human anatomy at the moment and can find a broad field of applications.

15.6 Conclusions While in the past, mathematical phantoms of the human body with simplified shapes of the body and the internal organs were used for all types of organ dose calculations, a variety of voxel computational phantoms became available in recent years that were based on medical image data of real persons. It was shown by a series of studies performed by different research groups that the voxel computational phantoms do not only have the advantage of a much more realistic anatomy—which is quite obvious—but that this difference has also a clear impact on the calculated organ doses. These findings have persuaded the ICRP to employ this new type of computational body computational phantoms for the next update of dose coefficients for external and internal exposures to ionizing radiation that is planned following the new ICRP Recommendations. The computational phantoms “ICRP-AM” and “ICRP-AF” described in this work present the effort undertaken at our working group upon request by ICRP’s DOCAL Task Group to construct voxel computational phantoms representing the adult Reference Male and Reference Female. The reference voxel phantoms presented in this work are the official computational phantoms representing the ICRP Reference Male and Reference Female. The ICRP will publish recommended values for dose coefficients for both internal and external exposures using the ICRP-AM and ICRP-AF phantoms. These will include SAFs for particles relevant to internal exposures, and dose coefficients for external radiation fields. It should, however, be clear that these phantoms still have their individual anatomical relations, although the organ masses have been adjusted to the reference values. It should, therefore, be clear that both phantoms cannot represent any real individual, and that they should not be used to assess doses for specific patients in medical applications. Especially for radiation treatment planning purposes, an image-based computational phantom, partial or whole body, for every individual patient is necessary, and in these cases the reference computational phantoms cannot be applied unless one is interested in secondary radiation to organs away from the target volume.

Acknowledgments The authors wish to express their gratitude to Dr. D. Gosch and Professor K. Friedrich, Centre for Radiology, University of Leipzig, Germany, and to Dr. D. Hebbinghaus and Professor B. Kimmig, Clinic for Radiation Therapy, University of Kiel, Germany, for providing the computed tomographic data. The development of the reference voxel phantoms was financially supported by the German Federal Ministry for the Environment, Nature Conservation and Nuclear Safety under contract number StSch 4256.

388

Handbook of Anatomical Models for Radiation Dosimetry

References 1. Petoussi-Henss, N. et al. The GSF family of voxel phantoms, Phys. Med. Biol., 47, 89, 2002. 2. Zankl, M. et al. Organ dose conversion coefficients for external photon irradiation of male and female voxel models, Phys. Med. Biol., 47, 2367, 2002. 3. Fill, U.A. et al. Adult female voxel models of different stature and photon conversion coefficients for radiation protection, Health Phys., 86, 253, 2004. 4. Zankl, M. et al. GSF male and female adult voxel models representing ICRP Reference Man— The present status, Proceedings of the Monte Carlo Method: Versatility Unbounded in a Dynamic Computing World, UCRL-CONF-208667, American Nuclear Society, Chattanooga, TN, 2005. 5. ICRP. The 2007 Recommendations of the International Commission on Radiological Protection, Annals of the ICRP Publication 103, International Commission on Radiological Protection, Elsevier, Amsterdam, 2007. 6. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, U.K., 2003. 7. ICRP. Human Respiratory Tract Model for Radiological Protection: A Report of a Task Group of the International Commission on Radiological Protection, ICRP Publication 66, Pergamon Press, Oxford, U.K., 1994. 8. ICRP. Human Alimentary Tract Model, ICRP Publication 100, Pergamon Press, Oxford, U.K., 2006. 9. Zankl, M. and Wittmann, A. The adult male voxel model “Golem” segmented from whole body CT patient data, Radiat. Environ. Biophys., 40, 153, 2001. 10. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Pergamon Press, Oxford, U.K., 1991. 11. ICRU. Tissue substitutes in radiation dosimetry and measurement, ICRU Report 46, Bethesda, MD, 1989. 12. Becker, J., Zankl, M., and Petoussi, N. A software tool for modification of human voxel models used for application in radiation protection, Phys. Med. Biol., 52, N195, 2007. 13. Brash, J.C. and Jamieson, E.B. Cunningham’s Text-Book of Anatomy, 8th edn., Oxford University Press, New York, 1943. 14. Möller, T.B. and Reif, E. Taschenatlas der Schnittbildanatomie—Computertomographie und Kernspintomographie. Band II: Thorax, Abdomen, Becken, Georg Thieme Verlag, Stuttgart, New York, 1993. 15. Möller, T.B. and Reif, E. Taschenatlas der Schnittbildanatomie—Computertomographie und Kernspintomographie. Band I: Kopf, Hals, Wirbelsäule, Gelenke, 2nd edn., Georg Thieme Verlag, Stuttgart, New York, 1997. 16. GEO kompakt. Das Wunder Mensch, In GEO kompakt, 2, Gruner + Jahr, Hamburg, 2005. 17. Schlattl, H., Zankl, M., and Petoussi-Henss, N. Organ dose conversion coefficients for voxel models of the reference male and female from idealized photon exposures, Phys. Med. Biol., 52, 2123, 2007. 18. Zankl, M., Eckerman, K.F., and Bolch, W.E. Voxel-based models representing the male and female ICRP reference adult—The skeleton, Radiat. Prot. Dosim., 127, 174, 2007. 19. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, Part I: Methods, TM-8381/V1, Oak Ridge National Laboratory, Oak Ridge, TN, 1987. 20. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: The Skeleton, ICRP Publication 70, Pergamon Press, Oxford, U.K., 1995. 21. Cristy, M. Active bone marrow distribution as a function of age in humans, Phys. Med. Biol., 26, 389, 1981. 22. Sato, K. et al. Development of a voxel phantom of Japanese adult male in upright posture, Radiat. Prot. Dosim., 127, 205, 2007. 23. ICRP. Adult Reference Computational Phantoms. ICRP Publication 110, International Commission of Radiological Protection, Elsevier, Amsterdam, 2009.

16 Physical Phantoms for Experimental Radiation Dosimetry David E. Hintenlang, William E. Moloney, and James Winslow

CONTENTS 16.1 Introduction ............................................................................................................... 389 16.2 Applications in Radiation Dosimetry .................................................................... 391 16.3 Attributes of Physical Phantoms............................................................................. 392 16.3.1 Tissue-Equivalent Materials ....................................................................... 392 16.3.2 Design and Development of Anatomy ..................................................... 393 16.3.3 Integration of Dosimetry ............................................................................ 394 16.4 Review of Existing Physical Anatomical Phantoms ............................................ 396 16.4.1 Research and University-Based Phantoms .............................................. 396 16.4.1.1 University of Florida Phantom Series ....................................... 396 16.4.1.2 Korean Male Phantom ................................................................. 399 16.4.2 Commercially Produced Phantoms .......................................................... 401 16.4.2.1 CIRS—Atom Phantom Series ..................................................... 401 16.4.2.2 CIRS—3D Sectional Torso Phantom ......................................... 403 16.4.2.3 RANDO Phantom ........................................................................ 403 16.4.2.4 Kyoto Kagaki Phantoms .............................................................. 406 16.5 Summary .................................................................................................................... 406 References ............................................................................................................................. 407

16.1 Introduction Physical phantoms are generally designed and fabricated for the purposes of performing either image quality or radiation dosimetry studies. A wide variety of physical phantoms have been developed for both imaging and radiation dosimetry studies, and for many of these phantoms the primary application is quality assurance (QA) of medical devices. The specific anatomical properties of the human subjects are not required, and therefore not integrated into the phantom. This applies to phantoms designed to quantify both image quality and/or radiation doses. In other situations, however, details of human anatomy become more important. For image quality studies, for example, it is appropriate to develop accurate images of human anatomy. These are most frequently found in training applications. Anatomical physical phantoms also play an important role in the accurate determination of human doses for 389

390

Handbook of Anatomical Models for Radiation Dosimetry

a wide variety of applications. An accurate reproduction of a specific human permits the accurate determination of radiation doses to that individual from a wide variety of radiation sources and exposure scenarios. The measured doses in these cases are limited to the specific geometry of the phantom used in the study. The extrapolation to doses delivered to other individuals of significantly different physical dimensions must be examined carefully in each individual case. In order to rigorously assess radiation doses for individuals of differing physical size and anatomical features, physical phantoms that realize these specific features must be fabricated. Examples of some situations where major features would affect radiation dosimetry include the differences between the adult male and female and the gross change of physical dimensions during human development, from the newborn to adulthood. For medical procedures, radiation may be directed to a limited portion of the body, and a phantom representing only this limited portion of anatomy may be sufficient to characterize radiation doses. Caution should be exercised when using some of these limited anatomical physical phantoms for either modeling or empirical studies since the scatter radiation to organs within the limited volume may not be accurately accounted for. This is of particular concern if the procedure delivers a significant portion of the primary radiation to volumes outside of the anatomical region represented by the phantom. The scope of this work addresses the general use of anatomical physical phantoms used for the determination of organ-specific radiation doses that may be used in a variety of external radiation fields. We therefore consider phantoms that represent a significant portion of the body, are appropriate for extracting multiple organ doses, and have accurate representations of the corresponding human anatomy. The anatomy associated with the physical phantoms discussed here provides two important contributions to the accurate evaluation of organ doses: 1. The anatomically correct positioning of organs of interest residing within the phantom 2. Accurate reproduction of the attenuation and scattering of incident radiations from a multitude of directions These requirements are fundamental to successfully achieving the primary goal of measuring the radiation dose deposited at specific organ locations within the body. Once organ doses are accurately measured, the data can be further processed to quantify the radiation weighted organ dose, effective dose, or other radiation dose quantities of interest. The evolution of computational and anatomical physical phantoms has historically developed along independent paths. Consider the well-known examples in each area that were considered state of the art some years ago: the RANDOâ phantom* and the Medical Internal Radiation Dose (MIRD) phantom.1 The RANDO phantom, developed by Alderson Research Laboratories, arose to fill the need to accurately measure doses from external radiation sources, and was fabricated using a human skeleton embedded in a more durable tissue-equivalent plastic representing a generalized human anatomy. The MIRD phantom developed as a computational tool to evaluate radiation doses from internal radiation sources. Due to the limitations of computational tools, including both the computing power of the existing hardware systems, and the level of sophistication of available computational software, the relevant human anatomy of the phantoms was fabricated

* The Phantom Laboratory, Salem, New York.

Physical Phantoms for Experimental Radiation Dosimetry

391

from an assemblage of simple geometric shapes designed to represent a variety of internal organs, and ultimately a complete human. Both the external and internal anatomy of these phantoms are significantly different. While the RANDO phantom is more anatomically realistic, physical phantoms representing the stylized MIRD phantoms have also been constructed and used to perform empirical studies, and to provide empirical validation of computational simulations based on the MIRD computational phantom. More recently, however the parallel developments of computational and physical anatomical phantoms have provided complimentary tools for the accurate evaluation of radiation doses for a wide variety of exposure scenarios. As previously mentioned, physical phantoms permit the measure of total dose at specific locations from a wide variety of radiation sources, regardless of the energy and source geometry. The measurements are limited, however, by the fact that they are made at small, localized volumes, not necessarily representative of average organ doses. When averaged organ doses are desired (i.e., for calculations of effective dose) it becomes extremely useful to have a complimentary computational phantom.

16.2 Applications in Radiation Dosimetry Physical anatomical phantoms are flexible tools that can be used for radiation dosimetry for a wide variety of exposure scenarios. In principle, the phantoms can be loaded with dosimeters at both external and internal body locations, and can be positioned in the radiation fields of interest, as the human subject of interest would be. They are then exposed for the period of time necessary for integrating dosimeters to accumulate a dose sufficient to permit accurate evaluations of the dose. Although this methodology is generally applicable to all radiation fields when using appropriately selected dosimeters, some subtle cautions are in order. Attention to the type of “tissue-equivalent” materials used in phantom fabrication is important. Historically, physical phantoms have been used for measurements in photon fields, comprised of machine-produced x-rays, or emissions from radionuclides. Correspondingly, the tissue-equivalent materials used in the phantoms have traditionally been optimized to simulate the attenuation of x-rays through human tissue. Most tissue-equivalent materials have been designed to match tissue attenuation characteristics at relatively high energies (around 1 MeV and higher) as required by most of the early applications. Phantoms constructed of these materials are therefore expected to respond accurately at these energies, but may not have comparable tissue equivalence for x-rays of lower energies, or other types of ionizing radiations. Alpha and beta radiation measurements can be made on the exterior of the phantom, as these particulate radiations do not penetrate to deep body tissues. The response of tissueequivalent materials to other penetrating forms of radiation, including high-energy electron beams, protons, or neutrons, however, should be approached more cautiously. While anatomical phantoms based on traditionally designed tissue-equivalent materials have been used in these applications, the results should be interpreted cautiously since there are limited studies of the performance of these materials in these radiation fields. For appropriate radiation fields, physical phantoms are useful tools for evaluating organ doses from a wide variety of exposure scenarios. The whole body anatomy provided is useful for simulating environmental radiation exposures, and a wide variety of occupational

392

Handbook of Anatomical Models for Radiation Dosimetry

exposures, since the radiation fields of interest are frequently whole body exposures. In these scenarios the direction of radiation incident on the body is usually the main variable of interest. It is therefore common, and reasonable, to use full body anatomical phantoms for these studies. Medical exposures are usually different in that only a limited portion of the human anatomy is exposed to radiation. Consequently, partial body phantoms, specific to particular medical procedures, are frequently used to evaluate radiation doses under these circumstances. However, full body phantoms are seeing increased validity as clinical procedures and exams irradiate increasingly large regions of the body using multidetector computed tomography (MDCT) and volume computed tomography (VCT), over extended regions of the body (i.e., abdomen, chest, etc.) and even whole body scans. The use of full-body phantoms in these cases not only permit the same phantom to be used for a wide variety of studies, but anatomy outside of the primary radiation fields may receive a substantial dose from scattered radiation, making these anatomical regions particularly relevant. Examples include radiation therapy, cone beam, and MDCT.

16.3 Attributes of Physical Phantoms The overriding goal of anatomical physical phantoms is to reproduce the radiation attenuation properties and geometry of the human body as accurately as possible. Although certain approximations are required in each of these areas, their effect on the resultant organ doses can be minimized by the appropriate selection of phantom components and design. Some of the important categories that must be considered for the development and application of anatomical physical phantoms for dosimetry include: the type of tissueequivalent materials utilized, the design and implementation of anatomical details and geometry, and the incorporation of dosimeters for dose measurements. 16.3.1 Tissue-Equivalent Materials Tissue-equivalent materials are designed as substitutes for real human tissues to allow greater flexibility and latitude in the use of phantoms. They are intended to provide a durable tissue substitute that accurately simulates the radiation attenuation and scattering properties of the tissues they replace. Tissue-equivalent materials must also maintain these properties over a long period of time without degrading from exposure to the environment or radiation damage. It is also useful for the materials to be mechanically resilient, since due to the large mass (particularly of whole body adult phantoms) they may be subject to rough treatment. The physical properties that are most desirable to model in a tissue simulant material for photon studies include the mass attenuation coefficient and the tissue density. Simulating these parameters, however, causes a problem: the mass attenuation coefficient is energy dependent. While it may be trivial to match the mass attenuation coefficient at a particular energy, it is much more challenging, and useful, to develop a phantom that simulates the attenuation of human tissue over a wide range of energies. This not only permits the phantom to be used for a broad range of source applications, but also ensures an appropriate response when used with x-ray sources having spectral components spanning a wide range of energies. A reasonable strategy in achieving this goal is to design a combination of elemental constituents that matches the effective atomic number (Zeff) of the desired tissue. To pursue

Physical Phantoms for Experimental Radiation Dosimetry

393

this goal, one needs to know the elemental constituents of the tissue being simulated. Useful sources for the elemental constituents of interest are the ICRU Reports 442 and 46,3 and in this respect the selection of tissue simulants is quite similar to that used for computational phantoms. The basis for tissue-equivalent materials for many years has been mixtures based on common epoxy resins. The epoxy resin may be combined with a variety of other molecular compounds to achieve the desired mass attenuation coefficient and density. Since the density of the epoxy resins is greater than that of soft tissues, the density must ultimately be lowered during the preparation process. The general strategy for developing tissue simulants based on the epoxy resins is to add a variety of constituents to the epoxy matrix in order to obtain a good approximation of the desired Zeff, then add a small amount of very low-density constituent, such as phenolic microspheres, to adjust the density to the desired value. A number of potential problems can complicate this process, as the parameters discussed above are not independent of each other. The elemental composition of some epoxy resins is maintained as proprietary information, or may not be precisely known by most manufacturers; and the epoxy-resin matrix can only accommodate a certain amount of additional constituents before it begins to phase separate. These issues make it important to empirically test and verify the attenuation and density characteristics of fabricated tissue-equivalent materials. Some anatomical phantoms have made use of human skeletal tissues for the skeletal structure of the phantom. These materials are extracted from cadavers and provide precise bone anatomy for accurate anatomical detail, but since they have lost bone marrow contents, and dried considerably, they may not be as accurate a tissue simulant for healthy bone as some synthetic bone materials are. Polymethylmethacrylate (PMMA), also labeled by trade names as acrylic, or Plexiglas, is also frequently suggested and used for phantom materials. It is, however, a more attenuating and higher density material than the preferred soft-tissue simulants. Tissue-equivalent materials may be fabricated to simulate a wide variety of tissues and organs. For radiation dosimetry studies, at the current state of the art, the differing levels of attenuation and density do not significantly affect dosimetry results. A single soft-tissue material is therefore commonly used for all soft-tissues with the exception of the lung. The lung volume has a similar elemental composition, but much lower density (∼0.32 g cm−3) than other soft tissues (∼1.04 g cm−3). It is frequently fabricated based on similar soft-tissue mixtures with the addition of surfactant and foaming agents to reduce the density. Skeletal structures are created that have a higher density, which is commonly achieved by using epoxy-resin mixtures with a reduced, or no, contribution of microspheres. While not common, anatomical physical phantoms may also incorporate adipose tissues, based on soft-tissue simulants having a reduced density. These become useful for simulating more extensive regions of adipose tissue, including female breast tissue and exterior adipose layers. 16.3.2 Design and Development of Anatomy The fabrication of anatomical features is a defining characteristic of anatomical physical phantoms. The degree of anatomical detail can vary significantly, perhaps most dramatically illustrated by comparison of MIRD-stylized physical phantoms with state-of-the-art phantoms designed for external radiation dosimetry. Anatomical physical phantoms have been designed to simulate human characteristics. Since individual human geometries vary, it is not surprising that the anatomical geometries of phantoms can, and should,

394

Handbook of Anatomical Models for Radiation Dosimetry

vary as well. Large-scale features of anatomical phantoms may be dictated by the ICRP Reference Man,4,5 such as overall height and weight, but most other anatomical details are more loosely standardized. Even with the availability of Reference Man specifications, the precise anatomical details for many anatomical physical phantoms remain largely undocumented. Those phantoms that utilize cadaver skeletons are subject to individual variation in the skeletal dimensions, although the overall external dimensions of the phantoms are consistent from phantom to phantom. Those that utilize synthetic materials throughout maintain constant dimensions from phantom to phantom. The source and rationale documenting the selection of the geometrical dimensions are rarely available for anatomical physical phantoms, although they remain generally representative of the size and geometry of the members of the population they are intended to reflect. The vagueness of specific phantom dimensions complicates the evaluation and comparison of specific organ doses. Since the phantom dimensions are not always rigorously developed from a specific anatomical phantom, the locations of individual organs cannot be well defined. Therefore, although it may be possible to measure a detailed dose map through the phantom, it is difficult, or perhaps impossible, to extract individual organ doses because of the uncertainties in specific organ locations. Notable exceptions to this trend are the phantoms developed at the University of Florida and the Korean Agency for Technology and Standards (KATS) (both are described in detail in Section 16.4). The geometry of these phantoms is based on CT scans of human subjects. The CT scan data is segmented into the various organs and tissues of the body. The complete data set is therefore available for the manufacture of the physical phantoms, including the precise location and extent of individual organs of interest. The same data set also provides the basis for complementary voxelized computational phantoms that provide an identical computational analog to this family of phantoms. 16.3.3 Integration of Dosimetry In order to perform dose measurements at interior locations of anatomical phantoms, one must incorporate an array of dosimeters throughout the phantom volume. The dosimeters selected for this task are typically small, with individual dimensions less than ∼10 mm3. The most common dosimeter used in these applications over the past years has been thermoluminescent dosimeters (TLDs). Most commercial phantoms incorporate an array of shallow voids in each slice of the phantom to accommodate the dosimeters. Recent advances in dosimeter development have resulted in additional types of dosimeters being integrated with anatomical physical phantoms. Optically stimulated luminescence (OSL) elements can also be incorporated into the dosimeter locations. They have a geometry and size generally similar to that of TLDs. Issues relevant to both TLD and OSL dosimeters are potential sensitivity to light, which suggests that they should be loaded into the phantom segments under subdued lighting conditions. Both also require separate read and erase processing to be performed. This necessitates that the dosimeters be removed and replaced in the phantom, which requires substantial disassembly and reassembly of the phantom between each set of measurements. These limitations have led some researchers to pursue other options for in-phantom dosimetry. Desirable features for these dosimeters include a mechanism for immediate and remote read-out of dosimeters, and a system capable of performing this operation for a large array of similar dosimeters. Such systems will permit the dosimeters from individual locations in the phantoms to be read shortly after the exposure of interest is concluded, and do

Physical Phantoms for Experimental Radiation Dosimetry

395

not require the dosimeter elements to be removed from the phantom between exposures. A dosimeter system that satisfies these requirements, and has been utilized for dosimetry in several anatomical physical phantoms, is the metal oxide semiconductor field effect transistor (MOSFET) dosimetry system originally developed by Thomson-Nielson, now a division of Best Medical Canada, Ltd.* The physical size of the MOSFET dosimeter element is slightly smaller than the TLD and OSL elements, and is connected via thin wires to a bias voltage supply and reader interface, which permits the rapid, sequential read-out of up to 20 individual dosimeters. The incorporation of these dosimeters within a phantom requires not only that a void be created in the phantom for the sensitive element of the dosimeter, but also that additional tissue-equivalent materials be removed to form a channel for the connecting wires to extend outside of the phantom. The dosimeters and associated wires are thin enough that they usually have a minimal effect on x-ray and photon attenuation, and the resulting dosimetry measurements. Their presence, however, can lead to imaging artifacts, which can be particularly spectacular in CT applications. Some degree of care should therefore be exercised when positioning the wire leads from large numbers of these dosimeters. Several research groups have successfully integrated MOSFET dosimeters for studies utilizing anatomical physical phantoms, including Duke University,6–8 the University of Florida,9–11 and Rensselaer Polytechnic Institute.12,13 Some manufacturers of phantoms are able to readily modify their phantoms to accommodate the requirements of any of these dosimeter types. An alternative dosimetry system that has been developed for in-phantom dosimetry at the University of Florida is based on the utilization of small scintillation elements coupled via optical fiber to a photomultiplier tube that remotely records the dose delivered at the sensitive element. These systems have the advantage of providing a small, sensitive detector volume, while the optical fiber cable is more tissue equivalent than the metallic leads used with electronic dosimeters. Dosimeter arrays have been developed and demonstrated for these fiber optic coupled (FOC) dosimeter systems, and have been applied to in-phantom measurement.14,15 The design and application of these arrays continue to be refi ned. A single element FOC dosimeter has recently become commercially available from Global Dosimetry.† This type of dosimeter was originally developed by the Naval Research Laboratory‡ and is based upon a Cu-doped quartz region for the sensitive element.16–20 Its applicability in the range of diagnostic x-ray energies has been further evaluated at the University of Florida.14,15,21 FOC systems continue to require a channel machined in the phantom to permit the optical fiber to extend from internal organ locations to the location of the reader (typically located a meter or two away). Dosimeters used for in-phantom dosimetry require a few special considerations regarding their calibration and interpretation. Dosimeters are usually calibrated under exposure to a primary beam of appropriate energy in air and compared against another wellcalibrated detector, such as an ion chamber. The calibration may be performed either with or without a backscatter media present. In the current discussion, it is expected that the dosimeter be utilized on the surface, or within the phantom, and it is therefore most appropriate to include backscatter contributions as part of the calibration. This method should give a good calibration for dosimeters located on, or near, the surface of * Best Medical Canada, Nepean, Ontario, Canada. † Global Dosimetry Solutions, Inc., Irvine, CA. ‡ U.S. Naval Research Laboratory, Washington, District of Columbia.

396

Handbook of Anatomical Models for Radiation Dosimetry

the phantom. Dosimeters located at greater depths in the phantom will receive increasing contributions from secondary radiations, including scattered photons and secondary electrons. The contributions typically increase with depth in the phantom, which may necessitate the use of various build-up layers of tissue-equivalent materials during calibration.12 The magnitude of these effects is dependent on the details of the dosimeters’ energy dependence and angular responses, but may result in the need for depth-dependent calibration factors. Another contributor requiring depth-dependent calibration factors is the potential for beam hardening of the primary beam as it passes through attenuating layers of tissue. This becomes most significant for diagnostic x-rays and interior dosimeter locations. While the primary beam hardens as it attenuates through tissue-equivalent material, eliminating low-energy components from the spectrum, the scatter/primary ratio increases with depth, adding low-energy components to the spectrum. The modified energy spectrum that strikes the dosimeter may require a different calibration factor to correctly interpret the dosimeter results, particularly if the dosimeter being used has a strong energy dependence. The resulting effects provide a good opportunity for validation of empirical results, and vice versa, using computational phantoms.

16.4 Review of Existing Physical Anatomical Phantoms 16.4.1 Research and University-Based Phantoms 16.4.1.1 University of Florida Phantom Series A series of anatomical physical phantoms have been fabricated for research applications at the University of Florida (UF). These phantoms have evolved throughout the past decade, beginning as direct analogs of the MIRD-stylized phantoms, and progressing to sophisticated anatomically accurate phantoms. A unique feature of this family of phantoms is that each one has been developed to correspond precisely with a complementary computational phantom. This permits a direct correlation of organ dosimetry between computational simulations and empirical measurements for identical phantoms. The phantom series has been constructed to represent a range of ages, illustrated in Figure 16.1, with particular emphasis on pediatric subjects. The computational analogs of these phantoms are discussed in detail in Chapter 8. This family of physical phantoms incorporates real-time dosimeters for a large number of organ locations in order to perform rapid measurements of both specific organ doses and effective dose. The phantom family incorporates tissue-equivalent materials using traditional epoxy-resin formulations, modified specifically to match the radiological parameters of tissues in the energy range of diagnostic x-rays. The properties of the tissue-equivalent materials are also modified to be representative of the tissues specific to pediatric subjects for bone and lung. More recent phantoms have been constructed using urethane-based soft tissue-equivalent materials. These materials provide a pliable soft tissue simulant that is homogenous and simplifies dosimeter placement. Since this research group utilizes arrays of real-time dosimeters, it is necessary to accommodate the dosimeter leads as they extend out of the phantom to the dosimeter reader. Epoxyresin-based materials form a hard plastic, requiring channels to be machined into each slice to accommodate the dosimeters, which can present concerns of radiation streaming

Physical Phantoms for Experimental Radiation Dosimetry

397

FIGURE 16.1 A series of physical phantoms developed at the University of Florida. Left to right: Adult Male, 1 year old, Adult Female, Newborn MIRD, Newborn Anthropomorphic, Adult Male. The lower torsos, including genitalia, and extremities of several phantoms are not shown, but can be attached when used for dosimetry measurements.

to the dosimeter location. The pliable urethane-based materials require only a thin slit to be cut into the material, which then conforms around the dosimeter and leads eliminating these concerns. The fi rst physical phantoms constructed at the UF were not anatomical physical phantoms by current standards, but were designed to reproduce precisely the stylized anatomy defi ned by the MIRD phantoms. The various tissues and organs within the phantom were defi ned by the mathematical equations of simple geometries as developed by Cristy and Eckerman.22 With a focus on pediatric exposure to x-rays, the phantoms were designed to serve as an empirical analog to the existing computational phantoms for newborn and 1 year old subjects. The phantoms are hermaphrodites; the same phantom being used to represent both male and female patients, although dosimeters located at sex-specific organs permits the evaluation of organ doses and dose equivalents for both male and females.23 The overall dimensions of these phantoms are 51 and 75 cm length and a 3.52 and 10.04 kg mass for the newborn and 1 year phantoms, respectively. Since the MIRD-style phantoms were originally developed for internal dosimetry applications, it is not surprising that the simplistic anatomy may lead to some significant discrepancies, compared to more realistic anatomy when attempting to measure organ doses from external radiation sources. A second generation of physical phantoms was subsequently fabricated by the UF that provided much more realistic anatomical details, again representing pediatric patients of newborn and 1 year ages.24 The anatomical detail of these phantoms was prepared from CT data sets of actual patients that had physical dimensions closely matching the previously constructed MIRD-style phantoms. The CT data sets were collected at 1 mm axial increments on recently deceased patients. The data sets were segmented to identify the specific locations and volumes of internal organs, and were used as

398

Handbook of Anatomical Models for Radiation Dosimetry

the basis for developing anatomical physical and computational phantoms having as near as practical identical dimensions. The physical phantoms were constructed with a 5 mm slice thickness using tissueequivalent materials similar to the epoxy-resin-based compositions developed by White,25 modified to more appropriately represent the age-dependent density of pediatric tissues. Each of the pediatric phantoms fabricated at the UF incorporates similar tissue-equivalent materials, air, a lung tissue equivalent, a soft tissue equivalent, and a bone tissue equivalent. Reference values for newborn, child, and adult tissues differ slightly, so these phantoms incorporate various filler materials that bring about the radiological equivalence to the age-specific reference tissues. More recently anatomical phantoms representing adult males (Reference Man) and adult females (Reference Woman) have been fabricated.26 The fabrication methodology is somewhat more advanced, with a series of CT images modified to match specific reference data, and then integrated with an automated machining system that permits the accurate fabrication of molds for a large number of 5 mm axial sections. The bone tissue used in this phantom is a rigid epoxy-resin-based tissue equivalent similar to that used in the other phantoms. The soft tissue and lung tissue-equivalent materials, however, utilize a proprietary urethane-based tissue-equivalent material that provides a pliable material enhancing phantom durability and integration of immediate read-out dosimeters. A unique feature of the UF phantom series is that they have all been configured to incorporate small, essentially point, dosimeters that provide an immediate read-out following exposure. The early studies utilized MOSFET dosimeters and were configured to monitor up to 20 individual organ locations.9 This dosimetry system has been incorporated into some commercially available phantoms by other researchers. More recently a high-sensitivity optically coupled dosimeter has been developed and integrated with these phantoms.12 The FOC dosimeter has demonstrated an improved sensitivity, angular dependence, and reproducibility when compared to the MOSFET dosimeters. The fiberoptic patch cables that connect the sensitive element to the reader are also more tissueequivalent and smaller, lending themselves to simple integration, and are less invasive when used with a phantom. In fact, when integrated with a phantom fabricated of pliable material, the optical dosimeters can be positioned with only a shallow slit cut into the tissue-equivalent material. The dosimeter cable is pushed into the slit and the pliable materials conform tightly around the cable and dosimeter, eliminating concerns regarding air-gaps and radiation streaming. The dosimetry measurements made at point locations in these phantoms can readily be extended to average organ doses since identical computational phantoms are also available. The ratio of the computationally determined organ dose averaged over the entire organ volume to the computationally simulated point doses defines the organ dose scaling factor. This scaling factor can then be multiplied by the measured point dose to obtain a reasonable, empirical measurement of specific organ doses.11 The early phantoms demonstrated the value of a physical phantom that could be utilized for dosimetry measurements at a wide variety of clinical facilities, and were the first to incorporate an array of immediate read-out dosimeters that facilitated the examination of doses for a series of radiographic examinations and varying techniques.27 Since the phantoms were also identically matched to a computational phantom, they also provided a unique opportunity to benchmark and compare the results of physical measurements with computational simulations.28

Physical Phantoms for Experimental Radiation Dosimetry

399

With the described methodologies, the UF phantoms have been used to examine organ doses that are delivered through a variety of diagnostic radiological procedures. These include traditional radiographic, fluoroscopy, and CT procedures. A series of characteristic doses for a pediatric patient undergoing a variety of radiographic procedures was compiled by Bower29 and subsequently extended to a large number of facilities by Hintenlang et al.27 This study examined organ doses and effective dose delivered from 17 commonly performed radiographic examinations performed at 10 different facilities. The effective doses generally scaled as expected for the various exams, but were surprisingly found to vary by nearly an order of magnitude between different facilities. The use of phantoms allowed an analysis of the distribution of examination parameters to identify those that could most effectively reduce pediatric patient dose. More recent investigations have focused on doses delivered by state-of-the-art CT systems. Helical and MDCT systems have introduced dose delivery that is not realistically evaluated using the traditional concept of the CT dose index (CTDI). Jones14 performed a comprehensive evaluation of the effect of CT acquisition parameters on organ doses and effective dose of a newborn. Using the anatomical phantom integrated with a real-time FOC dosimeter array, demonstrated the importance of having accurate anatomical geometries and orientations when performing dose measurements. The anatomical geometry directly affects dose contributions from the mA-modulation and significant scatter fields produced by the current generation of MDCT systems. 16.4.1.2 Korean Male Phantom The ICRP reference man is based on a Caucasian male of either Western European or North American decent. As a result, this phantom may not be entirely applicable to other populations with racial differences in custom, dietary habits, and climatic conditions. Therefore, a physical phantom representing an average Korean adult male was developed based on CT images by Kim et al.30 in Seoul, South Korea. The body dimensions of an average Korean male were obtained from the data of the Fourth Survey of National Physique Standard for the KATS. A healthy volunteer who fit the average Korean male was modeled (dimensions listed in Table 16.1) and source images were obtained using a PET/CT system (Siemens* Somatom Emotion Duo system). The physical phantom of the Korean male was constructed using whole body CT images and a Rapid Prototyping and Manufacturing (RP&M) technique. CT images in DICOM format were converted to STL format using V-works™ (CyberMed†). V-works performs volume rendering and surface rendering, and image segmentation was performed automatically for three tissue types. The phantom was then constructed using these three tissues: bone, lung, and soft tissue, made from SL5510 (Vantico, 3D systems‡), urethane foam, and polyurethane, respectively. CT data was separated into individual body parts (three tissue types) using SolidView™ (Solid Concepts§). Bone was constructed using a SLA3500™ RP&M Machine (3D systems‡). The same machine was used to build a cast for the lung phantom in which lungequivalent material was poured. Lung and bone phantoms were then positioned in an outermost skin cast into which tissue-equivalent (polyurethane) was poured and cured.

* Siemens Medical Solutions USA, Inc., Malvern, PA. † CyberMed, Ltd., Geumcheon-gu, Seoul, Korea. ‡ 3D Systems Corporation, Rock Hill, SC. § Solid Concepts Inc., Valencia, CA.

Handbook of Anatomical Models for Radiation Dosimetry

400

TABLE 16.1 Physical Feature of Anatomical Phantoms Phantom

Height (cm)

Reference Man Korean Male phantom RANDO Man RANDO Woman UF—Newborn UF—1 year old UF—Adult Male ATOM—Adult Male ATOM—Adult Female ATOM—Pediatric Newborn ATOM—Pediatric 1 year old ATOM—Pediatric 5 year old ATOM—Pediatric 10 year old

Weight (kg)

176 172 175 163 51 75 176 173 160 51 75 110 140

73 68 73.5 54 3.52 10.04 73 73 55 3.5 10 19 32

TABLE 16.2 Radiological Parameters of Anatomical Phantoms Soft Tissue Phantom

−3

Density (g · cm )

Bone −3

Zeff

Density (g · cm ) 1.23 1.6 Uses natural human skeleton 1.22 1.4 1.4 1.4862

Korean Adult Male ATOM RANDO

1.07 1.055 0.997

NA 7.15 7.6

UF—Newborn UF—1 year old UF—Adult Male ICRP 23 Reference Man

1.04 1.04 1.04 0.9869

6.77 6.69 NA 6.86

Lung Zeff NA 11.5

8.22 8.8 8.8 8.75

Density (g · cm−3)

Zeff

0.28 0.21 0.352

NA 7.38 7.11

0.3 0.3 0.33 0.2958

6.83 6.83 NA 7.14

The characteristics of the tissue-equivalent materials used in the Korean male phantom were close to ICRP recommendations and are listed in Table 16.2. The phantom was sectioned into 43 slices with an individual slice thickness of 2 cm, and hole grids were drilled 7 mm in diameter for TLD chip placement. Initial internal dosimetry measurements were performed by the manufacturers using LiF:Mg,Cu,P TLDs situated at the representative position of selected organs and tissues. Absorbed doses of these organs and tissues were calculated using MCNP 2.4 (using the same tomographic phantom constructed from the CT image data) in order to compare TLD relative response to absorbed dose. The relative response of the TLDs for a 662 keV 137Cs source and the simulated absorbed dose per photon were then calculated using a dose conversion coefficient of 2.96 × 10−32 Gy per relative response for soft tissue. The ratio of TLD to MCNP dose and measurements for a series of specific organs demonstrated agreement to within 7% between the measured and computed doses.

Physical Phantoms for Experimental Radiation Dosimetry

401

16.4.2 Commercially Produced Phantoms 16.4.2.1 CIRS—Atom Phantom Series The complete line of ATOMâ anthropomorphic phantoms are manufactured by Computerized Imaging Reference Systems, Inc. (CIRS).* Originally developed in Riga, Latvia, the ATOM dosimetry verification phantoms include pediatric phantoms for newborns, 1, 5, and 10 year olds, as well as adult male and female phantoms (physical dimensions provided in Table 16.1). Standard phantoms include head, torso, upper femur, and genitalia. Legs and arms are included with the newborn and 1 year old pediatric phantoms and may be manufactured upon request for the remaining phantoms and optional sinus, trachea, and bronchial cavities are available (Figures 16.2 and 16.3). Customized phantoms can be specifically manufactured, but the standard phantoms include tissue-equivalent materials representing bone, lung, and soft tissue, with the compositions formulated for accurate simulation of diagnostic and therapy energies. Photon attenuation values between 30 keV and 20 MeV based on ICRP recommendations are claimed to be within 1% for bone and soft tissue substitutes, and within 3% for lung substitute, by the manufacturer. Table 16.2 provides density and effective atomic number data for the tissue-equivalent materials used to represent human soft tissue, bone, and lung. Actual organ tissues included in the phantom are average soft tissue, average bone tissue, cartilage, spinal cord, spinal disks, lung, and brain. CIRS utilize a synthetic bone material to produce an accurate anatomical skeleton for the ATOM phantoms. Phantom skeletal structure can include separate trabecular and cortical bone densities and can vary significantly in both size and density. The skeletal anatomy of the ATOM phantom is based on a homogeneous bone tissue composition

FIGURE 16.2 Transverse section of the UF 1 year old anthropomorphic phantom illustrating soft, bone, and lung tissue structures. * Computerized Imaging Reference Systems, Norfolk, VA.

Handbook of Anatomical Models for Radiation Dosimetry

402

(a)

(b)

FIGURE 16.3 The commercially produced phantom. (a) The CIRS family of phantoms. (From CIRS Systems—Sales Info. With permission.) (b) RANDO adult phantom. (From The Phantom Laboratory—Sales Info. With permission.)

that averages known cortical to trabecular bone ratios and age-based mineral densities. For most organ-dosimetry applications, bone tissue can be reasonably simulated using homogenized bone properties, since details of the inner structure is not critical. This design feature provides dose verification for different applications, especially when comparative Monte Carlo calculations are necessary. While CIRS provides a map of internal organ locations, the traceability of the locations is not documented and there are no known corresponding computational reproductions to facilitate simulations of these phantoms. Standard phantoms consist of 25 mm thick contiguous sections. Each section contains a series of 5 mm diameter through holes with tissue-equivalent plugs. Hole locations are optimized for TLD placement in more than 20 internal organs. Alternative accommodations for dosimeters, including ion chamber cavities, grid patterns, and hole diameters are all available upon request. Purchase prices of the series range from approximately $9000–$20,000 depending on the specific phantom and options. The series of ATOM phantoms from CIRS have been recently implemented in a variety of dosimetric applications. The custom manufacturing of the phantom series allows the user flexibility in their dosimeter of choice. Furthermore, the chronology of the ATOM phantoms permits empirical assessment of dose over a range of patient ages and sizes. With regard to the doses delivered to pediatric patients in diagnostic radiology the ATOM phantom has seen much use. Frush et al.6 used 1 and 5 year old pediatric phantoms (phantoms 704 and 705) in combination with MOSFET dosimeters to assess the dose to pediatric patients undergoing conventional and CT angiography, as well as gated cardiac CT. Mukundan et al.8 used MOSFETs and the 5 year old phantom to assess the dose to the orbit during pediatric cranial CT with and without shielding. It was shown that dose to the eye was reduced by up to 42% with shielding, while the shield artifact fell outside the region of interest, thus not affecting image quality. Fricke et al.31 applied the newborn phantom 703 ATOM phantom with calibrated TLDs for individual organ dose measurements of the breast and lung. A pediatric radiologist performed image quality assessments of patients undergoing MDCT, and clinical protocols

Physical Phantoms for Experimental Radiation Dosimetry

403

were repeated on the anthropomorphic dosimetry phantom to assess the dose to individual organs. Results proved that the use of breast shields on female pediatric patients did not impair image quality and allowed a dose reduction of 6.7% in the lung and a 29% decrease to the breast. Blinov et al.32 utilized the phantom 701 adult male ATOM phantom in conjunction with TLDs placed for individual organ dose measurements of scattered radiation during digital fluoroscopy examinations. The adult male phantom embedded with TLDs was also used by Samei et al.33 to evaluate the scatter, dose, and effective detector quantum efficiency (DQE) performance of a slot-scan digital chest radiography system compared with that of a full-field digital radiography system. Hurwitz et al.7 modified the female anthropomorphic adult phantom (model 702) to reflect the expected location of a fetus during the first 3 months of gestation in order to determine the radiation dose to the fetus from clinical MDCT body scans. MOSFET and TLD placement in-phantom were used to measure primary radiation from abdomen– pelvis examinations, and to assess the scatter radiation from chest routines, and thus estimate the associated risk of individual examinations to the mother and her fetus. The model 702 female phantom has also been modified with circumferential layers of fat-equivalent tissues from CIRS, Inc. to simulate obese patients in a study on the effect of image quality and radiation dose in multislice CT for overweight individuals. Schindera et al.,34 with the modified female phantom and MOSFET dosimeters, showed that modifying CT protocols for obese patients can substantially improve image quality without a significant increase to abdominal organs. As evident in the clinical experiments discussed, the CIRS series of ATOM phantoms has shown a clear applicability to assessments in image quality and radiation dose comparisons. The customizability of these phantoms with regards to patient shape, size, and age, as well as the through holes for dosimeter placement, allows for their extensive use in nearly all applications in diagnostic and therapeutic radiology. 16.4.2.2 CIRS—3D Sectional Torso Phantom The CIRS Model 600 Anthropomorphic Torso phantom has recently been developed and is designed to accurately simulate an average torso (22 cm posterior–anterior (PA) thickness) for medical imaging and dosimetry applications. The phantom is comprised of 12 internal tissues and organs including the lungs, heart, liver, kidneys, spleen, and pancreas. The lower portion of the phantom contains a soft bolus simulating a mix of 30% adipose and 70% muscle tissue. Muscle-equivalent material covers the rib cage and vertebral column. The epoxy materials provide optimal tissue density as well as attenuation in the diagnostic and therapeutic energy ranges (40 keV to 20 MeV). The Torso phantom is proposed to be utilized for calibration, QA, and training purposes when specific internal organs are of interest. The phantom does not currently incorporate a specific set of features for integrating dosimeters, but can be configured and/or custom made to accommodate a multitude of dosimeters. 16.4.2.3 RANDO Phantom The RANDO phantom, currently manufactured by The Phantom Laboratory represents one of the oldest and most widely utilized anatomical dosimetry phantoms. The phantom was originally designed by Dr. Lawrence H. Lanzl in 1959. The Phantom Laboratory acquired the manufacturing rights from Alderson Research Laboratories in 1989. The phantom has

404

Handbook of Anatomical Models for Radiation Dosimetry

traditionally been used to provide detailed mapping of dose distributions necessary for evaluating radiotherapy treatment plans. The RANDO phantoms incorporate air, human cadaver skeleton or simulated bone material, a soft tissue-equivalent material, and a lung tissue-equivalent material. While these phantoms have traditionally incorporated real bone from a human cadaver skeleton, The Phantom Laboratory is currently transitioning to a simulated bone material for the skeletal structure. The soft and lung tissue-equivalent materials are manufactured using a proprietary urethane formulation designed for use in typical therapeutic energy ranges. These materials are cast to produce phantoms representing either an adult male or adult female. The RANDO soft tissue-equivalent material was designed to replicate the effective atomic number, electron density, and physical density of muscle tissue with randomly distributed fat. The RANDO lung tissue-equivalent material was designed similarly, replicating the effective atomic number, electron density, and physical density of lung tissue in a median respiratory state. The elemental composition for the RANDO tissue-equivalent materials have a large difference in weight percent of carbon and oxygen compared to human tissue. This is due to the low oxygen content in most polymers and resins compared to human tissues; an exact elemental compositional match is not easily attainable. Considering this limitation, carbon and oxygen are assumed interchangeable.35 The sum of carbon and oxygen should be similar to that of the tissue being represented. In the energy ranges dominated by Compton interactions, as is the case with radiotherapy, this assumption is reasonable. At typical diagnostic energy levels, this assumption is less forgiving due to the photoelectric interaction’s strong dependence on atomic number. Calculations of the mass attenuation coefficients and mass energy absorption coefficients of RANDO tissue-equivalent materials correspond well with those of the actual tissues represented in the energy range of 0.1–100 MeV; below 0.1 MeV the tissue equivalence begins to diminish.35,36 The anatomy of the RANDO phantoms is based on military anthropometry numbers taken from a 1950s U.S. Air Force Survey. There are two phantoms: The RANDO Man and the RANDO Woman. RANDO Man represents a 175 cm tall and 73.5 kg male figure. RANDO Woman represents a 163 cm tall and 54 kg female figure. Neither phantom has arms or legs. RANDO phantoms have traditionally constructed using a natural human skeleton. Minor adjustments are made to those skeletons lacking symmetry, with distorted joints, or not having the same size and shape as the molds. The lungs are hand-shaped and molded to fit the contours of the rib cage; the left lung is smaller than the right to allow for the heart. Air spaces found in the head, neck, and stem bronchi are duplicated. A variety of breast sizes are available for these phantoms; the breast shape is not natural. RANDO phantoms consist of 2.5 cm thick sections, and hole grids may be drilled into the soft and lung tissue-equivalent areas for each slice. Two standard hole grid sizes are available, 3 cm × 3 cm, and 1.5 cm × 1.5 cm, but custom holes can also be requested. Dose measurements and distributions can be obtained using film or individual dosimeters. Special capsules for lithium fluoride powders are also available. The current starting price for a RANDO phantom, clamping base, and storage case is $19,340. Various grid options for dosimeter placement can be custom manufactured for additional costs. Breast attachments ranging from size A through E for both male and female phantoms are available, and cost a few hundred dollars more, with an entire set costing as much as $788. The RANDO phantom has proven useful in many types of studies. They consistently serve as a tool with which to directly quantify patient dose. Recently, Wen et al.37 used TLDs and the RANDO phantom to measure the increased doses to patients receiving

Physical Phantoms for Experimental Radiation Dosimetry

405

IMRT for prostate cancer due to the use of daily cone beam CT pelvic scans for daily patient setup. Similarly, Luz et al.38 evaluated the dose resulting from four different 64-slice MDCT colonography protocols. Interestingly, measurements showed females receiving an elevated dose when compared with males. Fitousi et al.39 studied the effective dose to the patient during a vertebroplasty, which requires fluoroscopic exposure, using a RANDO phantom loaded with TLDs. They found that the effective dose to the patient was calculated to be 34.45 mGy, and the mean skin dose was 688 mGy. This mean skin dose was determined sufficiently high to recommend additional measures be taken to reduce the patient’s skin dose. Theocharopoulos et al.40 used the RANDO phantom in combination with TLDs to gather direct effective dose for anterior–posterior (AP) abdominal radiography, PA chest radiography, PA head radiography, and AP heart fluoroscopy. They compared these effective doses with two effective dose estimations. The first one relied on data published by the National Radiological Protection Board (NRPB) and entrance surface dose (ESD); the other relied on NRPB data and dose area product (DAP). They found that the ESD method differed from directly assessed values by 15% for the radiographic exposures and 60% for heart fluoroscopy, while the DAP method was generally in good agreement (<8%). As with the study by Theocharopoulos et al., the RANDO phantom is also used extensively to validate mathematical phantoms for dosimetry. For example, Compagnone et al.41 compared entrance skin dose values resulting from 11 standard radiographic examinations using three anthropomorphic, two simple cylindrical, and one mathematical phantom. The anthropomorphic phantoms, including one RANDO phantom, were treated as the “gold standard.” They found that their mathematical phantoms were accurate within 12%, and could therefore be used satisfactorily, providing a convenient estimation of dose. Deak et al.42 used a RANDO phantom to validate a Monte Carlo tool for patient-specific dose simulations in multislice CT using arbitrary protocols. The calculated dose distributions were found to be within 10% of phantom TLD measurements. Popescu et al.43 calculated absolute dose using a Monte Carlo method capable of incorporating monitor units that applies to any configuration of open and blocked fields, including IMRT. Their simulation results were within 2% difference of those obtained in the physical experiment. Osei et al.44 considered the occupational exposure to pregnant workers. They used a Monte Carlo simulation to determine coefficients for converting dosimeter readings to equivalent dose to the fetus. The female RANDO phantom was used to validate the results over a range of irradiation conditions. Here, the two sets of data indicated good agreement, with the Monte Carlo results being relatively greater than the experimental results, up to a maximum of 15%. The other use often found in the literature involving RANDO phantoms is as a patient substitute. This is especially useful in studies of exposures to occupational workers. For example, Kuon et al.45 mapped the fluoroscopic DAP/s during an invasive cardiology procedure as a function of angle using a RANDO phantom as the patient substitute and main scatter source. From this mapping, they could identify tube angulations that promise minimal radiation exposure to operators and patients. The study found that for their radiographic settings, the less-conventional PA and right anterior oblique views greater than 40° should be favored over steep left anterior oblique projections greater than 40°. In another study, Kuon et al.46 investigated the effectiveness of wearing a 0.5 mm leadequivalent cap (mask), with regard to the head exposure incurred at the operator position during an invasive cardiology fluoroscopic procedure. Here too, they used a RANDO phantom as a patient substitute and scatter source. The caps were shown to significantly reduce the scatter entrance skin air kerma to the operator.

406

Handbook of Anatomical Models for Radiation Dosimetry

16.4.2.4 Kyoto Kagaki Phantoms Another set of commercially available phantoms are the Kyoto Kagaki phantoms developed and manufactured in Japan by the Kyoto Kagaki, Co. Ltd.* These series of phantoms include whole body phantoms with good anatomical features. They include a Therapy Body phantom, model THRA-1 with representations for adult male and female, a Whole Body phantom (Model PBU-50), a CT Torso phantom (Model CTU-41), and a Chest phantom (Model N1). The Therapy Body phantoms follow the traditional style of the Rando phantom, providing both adult male and female representations and are constructed from a series of 30 mm slices. They extend from the top of the head to the top of the lower extremities, encompassing a 900 mm height, and utilize synthetic bones. As indicated by the phantom name, this phantom was developed to use as an aid in the treatment planning for radiotherapy applications. It incorporates three tissue-equivalent types of material representing, soft tissue, bone, and lung. The tissue-equivalent materials are epoxy-resin-based materials of proprietary compositions that were developed in a joint study with the National Institute of Radiological Sciences.† The dosimeter locations are normally laid out in a mesh, based on 30 mm × 30 mm centers although cavities for detectors can be customized on demand. The Whole Body, CT Torso, and Chest phantoms are designed as imaging and positioning/training phantoms and are not strictly intended as dosimetry phantoms. They have nicely detailed skeletal and anatomical features, with good internal organ detail available from diagnostic x-ray images. A series of individual organs are included, each with a Hounsfield number corresponding to that found in the human body. The soft tissue and organs are fabricated using tissue-equivalent urethane-based resins while the synthetic bone-equivalent tissues are fabricated using epoxy-resin-based materials. The urethanebased materials used for the soft tissues are also soft and somewhat flexible, which is an aid to their durability. A unique feature of the Whole Body phantom is that it includes full appendages with fully articulating joints of the major appendages. Since this group of phantoms is not specifically designed for anatomical dosimetry applications, they do not readily accommodate any particular dosimetry system, although with some effort, dosimeters could be integrated with the phantoms. The tissue-equivalence of the phantoms of this group of phantoms is not well documented, but an accurate reproduction of the CT Hounsfield number for various tissues would imply accurate tissue-equivalent attenuation. The reference for anatomical size and geometry is also not documented and is significantly smaller than Reference Man with an overall height of 165 cm and weight of 50 kg. The size of some of the bones and joints in the Whole Body phantom are also enlarged and modified to facilitate the structural strength of the phantom.

16.5 Summary The previous discussions describe the key components associated with the development and utilization of anatomical physical phantoms. The major features of several of the most commonly encountered phantoms are readily compared in Tables 16.1 and 16.2. These

* Kyoto Kagaki, Co. Ltd., Kyoto, Japan. † National Institute of Radiological Sciences, Chiba-shi, Chiba, Japan.

Physical Phantoms for Experimental Radiation Dosimetry

407

tables give the reader an overview of the different physical and radiological parameters that have been selected by the various phantom developers. The anatomical physical phantoms developed by research oriented groups frequently match an accompanying computational phantom, and the value of these phantoms for accurately evaluating organ doses and parametric studies has been illustrated. Various phantoms have been developed for specific purposes and while they are flexible tools, it is important that they be utilized in appropriate energy ranges. There are several examples of RANDO phantoms being utilized for studies in the diagnostic energy range, even though the tissue-equivalent materials they utilize are designed for optimal response at much higher energies associated with radiation therapy applications. While physical phantoms have made many valuable contributions to radiation dosimetry, their flexibility and future advances will likely permit them to improve the knowledge of radiation doses delivered as new procedures and exposure scenarios develop. The value of anatomical physical phantoms is being recognized for dose measurements associated with MDCT and VCT, where traditional dosimetry phantoms are too small to produce scatter fields similar to those produced by human subjects. Future improvements in utilization will likely come through advances in the associated dosimetry systems. Systems that accommodate large arrays of sensitive dosimeters that can be read immediately will permit the parametric studies over a wide number of internal organs and exposure scenarios necessary to develop and document human doses for realistic exposures.

References 1. Snyder, W.S. et al. Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, J Nucl Med, Suppl 3, 7, 1969. 2. ICRU. Tissue substitutes in radiation dosimetry and measurement, ICRU Report 44, Bethesda, MD, 1989. 3. ICRU. Photon, electron, proton and neutron interaction data for body tissues, Report 46, International Commission on Radiation Units and Measurements, Bethesda, MD, 1992. 4. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Pergamon Press, Oxford, 1975. 5. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, 2003. 6. Frush, D.P. and Yoshizumi, T. Conventional and CT angiography in children: Dosimetry and dose comparisons, Pediatr Radiol, 36, 154, 2006. 7. Hurwitz, L.M. et al. Radiation dose to the fetus from body MDCT during early gestation, Am J Roentgenol, 186, 871, 2006. 8. Mukundan, S. et al. MOSFET dosimetry for radiation dose assessment of bismuth shielding of the eye in children, Am J Roentgenol, 188, 1648, 2007. 9. Bower, M.W. and Hintenlang, D.E. The characterization of a commercial MOSFET dosimeter system for use in diagnostic x ray, Health Phys, 75, 197, 1998. 10. Jones, A.K. et al. MOSFET dosimeter depth-dose measurements in heterogeneous tissueequivalent phantoms at diagnostic x-ray energies, Med Phys, 32, 3209, 2005. 11. Staton, R.J. et al. A tomographic physical phantom of the newborn child with real-time dosimetry. II. Scaling factors for calculation of mean organ dose in pediatric radiography, Med Phys, 33, 3283, 2006. 12. Wang, B., Kim, C.H., and Xua, X.G. Monte Carlo modeling of a high-sensitivity MOSFET dosimeter for low- and medium-energy photon sources, Med Phys, 31, 1003, 2004.

408

Handbook of Anatomical Models for Radiation Dosimetry

13. Wang, B. and Xu, X.G. Measurements of non-target organ doses using MOSFET dosemeters for selected IMRT and 3D CRT radiation treatment procedures, Radiat Prot Dosimetry, 128, 336, 2008. 14. Jones, A.K. Dose versus image quality in pediatric radiology: Studies using a tomographic newborn physical model with an incorporated dosimetry system, PhD, University of Florida, FL, 2006. 15. Jones, A.K. and Hintenlang, D.E. Potential clinical utility of a fibre optic-coupled dosemeter (FOCD) for dose measurements in diagnostic radiology, Radiat Prot Dosimetry, 132, 80, 2008. 16. Huston, A.L. et al. Remote optical fiber dosimetry, Nucl Instr Methods Phys Res B Beam Interact Mater Atoms, 184, 55, 2001. 17. Huston, A.L. et al. Optically stimulated luminescent glass optical fibre dosemeter, Radiat Prot Dosimetry, 101, 23, 2002. 18. Justus, B.L. et al. Gated fiber-optic-coupled detector for in vivo real-time radiation dosimetry, Appl Optics, 43, 1663, 2004. 19. Justus, B.L. et al. Optically stimulated luminescence dosimetry using doped fused quartz, Radiat Prot Dosimetry, 84, 189, 1999. 20. Justus, B.L. et al. Optically and thermally stimulated luminescence characteristics of Cu+-doped fused quartz, Radiat Prot Dosimetry, 81, 5, 1999. 21. Benevides, L.A. et al. Characterization of a fiber-optic-coupled radioluminescent detector for application in the mammography energy range, Med Phys, 34, 2220, 2007. 22. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381 V1–V7, Oak Ridge National Laboratory, 1987. 23. Tresser, M.A. and Hintenlang, D.E. Construction of a newborn dosimetry phantom for measurement of effective dose, Health Phys, 76, S190, 1999. 24. Jones, A.K. et al. Tomographic physical phantom of the newborn child with real-time dosimetry I. Methods and techniques for construction, Med Phys, 33, 3274, 2006. 25. White, D.R. Formulation of tissue substitute materials using basic interaction data, Phys Med Biol, 22, 889, 1977. 26. Hintenlang, D., Moloney, W., and Winslow, J. An anthropomorphic adult physical phantom and fiber optic coupled point dosimetry system for the measurement of effective and average organ doses of CT patients, Med Phys, 35, 2652, 2008. 27. Hintenlang, K.M., Williams, J.L., and Hintenlang, D.E. A survey of radiation dose associated with pediatric plain-film chest x-ray examinations, Pediatr Radiol, 32, 771, 2002. 28. Sessions, J.B. et al. Comparisons of point and average organ dose within an anthropomorphic physical phantom and a computational model of the newborn patient, Med Phys, 29, 1080, 2002. 29. Bower, M.W. A physical anthropomorphic model of a one-year-old child with real-time dosimetry, PhD, University of Florida, 1997. 30. Kim, J.I. et al. Physical phantom of typical Korean male for radiation protection purpose, Radiat Prot Dosimetry, 118, 131, 2006. 31. Fricke, B.L. et al. In-plane bismuth breast shields for pediatric CT: Effects on radiation dose and image quality using experimental and clinical data, Am J Roentgenol, 180, 407, 2003. 32. Blinov, N.N. Jr. et al. A method for determination of the effective dose received by patient during digital scanning fluorographic examination from the results of measurement of radiation dose scattered by patient’s body, Med Tekh, 3, 2005. 33. Samei, E. et al. Comparative scatter and dose performance of slot-scan and full-field digital chest radiography systems, Radiology, 235, 940, 2005. 34. Schindera, S.T. et al. Abdominal multislice CT for obese patients: Effect on image quality and radiation dose in a phantom study, Acad Radiol, 14, 486, 2007. 35. White, D.R. Tissue substitutes in experimental radiation physics, Med Phys, 5, 467, 1978. 36. Shrimpton, P.C., Wall, B.F., and Fisher, E.S. The tissue-equivalence of the Alderson Rando anthropomorphic phantom for x-rays of diagnostic qualities, Phys Med Biol, 26, 133, 1981.

Physical Phantoms for Experimental Radiation Dosimetry

409

37. Wen, N. et al. Dose delivered from Varian’s CBCT to patients receiving IMRT for prostate cancer, Phys Med Biol, 52, 2267, 2007. 38. Luz, O. et al. Evaluation of dose exposure in 64-slice CT colonography, Eur Radiol, 17, 2616, 2007. 39. Fitousi, N.T. et al. Patient and staff dosimetry in vertebroplasty, Spine, 31, E884, 2006. 40. Theocharopoulos, N. et al. Comparison of four methods for assessing patient effective dose from radiological examinations, Med Phys, 29, 2070, 2002. 41. Compagnone, G., Pagan, L., and Bergamini, C. Comparison of six phantoms for entrance skin dose evaluation in 11 standard X-ray examinations, J Appl Clin Med Phys, 6, 101, 2005. 42. Deak, P. et al. Validation of a Monte Carlo tool for patient-specific dose simulations in multislice computed tomography, Eur Radiol, 18, 759, 2008. 43. Popescu, I.A. et al. Absolute dose calculations for Monte Carlo simulations of radiotherapy beams, Phys Med Biol, 50, 3375, 2005. 44. Osei, E.K. and Kotre, C.J. Equivalent dose to the fetus from occupational exposure of pregnant staff in diagnostic radiology, Br J Radiol, 74, 629, 2001. 45. Kuon, E. et al. Identification of less-irradiating tube angulations in invasive cardiology, J Am Coll Cardiol, 44, 1420, 2004. 46. Kuon, E. et al. Radiation exposure benefit of a lead cap in invasive cardiology, Heart, 89, 1205, 2003.

Part II

Applications

17 Applications to Environmental Exposures Nina Petoussi-Henss and Kimiaki Saito

CONTENTS 17.1 Introduction ............................................................................................................... 413 17.2 Calculation of Gamma Ray Fields in the Environment ...................................... 414 17.3 Simulation of a Secondary Source around the Computational Phantom ................................................................................... 415 17.4 Organ Dose Calculations for Monoenergetic Sources: The Voxel Computational Phantoms and the Monte Carlo Codes .................... 417 17.5 Calculation of Doses for Radionuclides................................................................. 421 References .............................................................................................................................423

17.1 Introduction Researchers can determine the dose coefficient for a specific radionuclide by the type, the intensity, the energy of the emitted radiations, the mode of exposure, and the anatomical characteristics. The dose coefficients incorporate the transport of emitted radiation in the environment, the simulation of the human anatomy, and the transport of radiation in the body. A number of publications have tabulated dose coefficients for external irradiation of the body from radionuclides distributed in the environment; for example, Poston and Snyder (1974); Dillman (1974); O’Brien and Sanna (1976); Koblinger and Nagy (1985); Jacob et al. (1986); DOE (1988); Saito et al. (1990); Petoussi et al. (1991); Eckerman and Ryan (1993).1–9 All are based on mathematical computational phantoms, mainly of adults. Data on organ doses for children and babies are very scarce. The first data based on voxel computational phantoms stemmed from Saito et al.7 and Petoussi et al.8 both of whom calculated the dose conversion factors normalized to source intensity and air kerma for environmental monoenergetic photon sources (15 keV to 10 MeV) and natural radionuclides for a 7-year-old child and a baby of 8 weeks using the voxels computational phantoms BABY and CHILD (see Chapter 3). Using these data, Jacob et al.10 published extensive tables of dose equivalent rates from artificial radionuclides for the BABY and CHILD in the air and deposited on the ground, in addition to the tables based on mathematical phantoms of adults. Saito et al.11 investigated organ dose as a function of body weight and the variation of effective dose (E) per air kerma under diverse exposure conditions.12 A subject’s body size was found to be the most important factor to affect the effective dose and differences could amount up to 90%, if one compares the values for an adult and for a baby at low energies. 413

414

Handbook of Anatomical Models for Radiation Dosimetry

The change in the posture of the human subject, as well as the biases of environmental sources, was found to affect the effective dose by about 10%. The variation of the effective dose with the posture of the human body (standing or lying) was found to be within 30%. The biases of environmental sources alternate the effective dose per air kerma mostly by 20% at the maximum, but in some case the alteration was found to be up to 40% due to the change in the energy spectrum. A similar trend is anticipated for the individual organ doses. Therefore, it was concluded that the conversion coefficients for the three typical environmental sources could be used as a reference set of values to derive both the organ doses and the effective doses of adults from air kerma or source activity obtained by measurement for a variety of environmental source configurations. The dose coefficients presented here are based partly on previously developed dosimetric methodologies (previous calculations of the energy and angular distributions of the radiations incident upon the body7) but also include recent results of the transport of these radiations within the body (i.e., voxel computational phantoms). The procedure described in the following was pursued at the Helmholtz-Zentrum München (former GSF—National Research Centre for Environment and Health) to estimate the organ doses from environmental photon sources. A three-step method was followed: 1. The calculation of the gamma ray transport in the environment (monoenergetic gamma rays and natural radionuclides) 2. The simulation of a secondary source around the computational phantom(s) 3. The calculation of organ doses using voxel computational phantoms Only photons and electrons contribute to the dose to tissues and organs of the body. By using the dose conversion coefficients for monoenergetic sources of photon and electron radiation, and by scaling them to the emissions of the radionuclides of interest, a researcher can derive dose coefficients from radionuclides in the environment related to a measurable quantity, i.e., activity concentration or air kerma.

17.2 Calculation of Gamma Ray Fields in the Environment It is a complicated computation task to estimate the doses to tissues of the body from radiations emitted by an arbitrary distribution of a radionuclide in the air or ground environment. Therefore, researchers use simplified and idealized exposure geometries instead. For the conversion factors presented here, the following three typical cases of environmental sources were considered: 1. A semi-infinite volume source in the air 2. An infinite plane source in the ground 3. A semi-infinite volume source in the ground The first source configuration models the gaseous radioactive release into the atmosphere at locations which are not too near the release point, by assuming a homogeneous contamination of the air up to a height of 1000 m above a smooth air–ground interface.

Applications to Environmental Exposures

415

The second source simulates the deposition of radionuclides in the ground, by assuming an infinite plane source in the soil. The source is shielded by a soil slab of 0.5 g cm−2, allowing for the surface roughness and initial migration with precipitation. The third source simulates the natural radioactivity in the ground and the dominant radionuclides of the 238U series, the 232Th series, and the 40K being homogeneously distributed to a depth of 1 m in the soil. In source 1, the dominant gamma rays come almost isotropically from upper 2π directions, while only a small amount of scattered gamma rays comes from the lower 2π directions. Source 3 shows the inverse tendency: the angular distribution is nearly uniform in the lower 2π directions with small components of scattered gamma rays stemming from upper 2π directions. In source 2, quite a large portion of the gamma rays comes from the horizontal directions. When the source distribution in the environment varies from the three typical source distributions, the angular and energy distributions also change. The environmental photon transport calculations, i.e., the calculations of the energy and angular distributions of the radiations, were performed with the Monte Carlo code YURI,13 originally developed to be used specially for environmental simulations. Experiments and other computation data have proved YURI. Compton scattering, photoelectric absorption, and pair production were considered as photon interaction processes. The air and the ground were assumed to contact each other with an infinite plane. Air was assumed to have a constant density of 1.2 × 10−3 g cm−3, corresponding to a temperature of 20°C and an air pressure of 0.1 MPa and to consist of N2, O2, and Ar having weight fractions of 75.5%, 23.2%, and 1.3%, respectively. Soil consists of SiO2, Al2O3, Fe2O3, and H2O with weight fractions of 58.3%, 16.7%, 8.3%, and 16.7%, respectively. A soil density of 1 g cm−3 was assumed in the calculations, since this value represents the upper 2 cm of soil. It should be noted that the environmental transport calculations were performed without the presence of the phantom; however, the perturbation caused by the human body was investigated and found to be insignificant. For the first two sources, calculations were performed for 20 initial gamma ray energies, from 15 keV to 10 MeV, to cover the wide energy spectra that many different artificial radionuclides have. The soil was modeled as a planar air–ground interface. Scatter and absorption of the radiation in both air and ground was considered in the calculation. From the transport calculations in the environment, double differential fluences, currents (i.e., fluences multiplied by the cosine of the angle of incidence), and air kerma values were obtained for points from 0 to 2 m above ground in steps of 20 cm. Table 17.1 shows calculated values of the air kerma rate free-in-air at 1 m height above the ground per unit activity concentration for a semi-infinite volume source in air and per unit activity per area for an infinite plane source in the ground; Table 17.2 shows the air kerma free-in-air at 1 m height above the ground per disintegration/kg for the semi-infinite volume source in the ground due to the natural radionuclides.

17.3 Simulation of a Secondary Source around the Computational Phantom From the results of the environmental calculations, a secondary source around the computational phantom was simulated in a form of a cylinder, slightly larger than the computational phantom mentioned in step two of the procedure, above.8 This was done by sampling the photons with the specific probability distribution functions as

Handbook of Anatomical Models for Radiation Dosimetry

416

TABLE 17.1 Calculated Air Kerma Rate at Height 1 m above the Ground per Unit of Activity Concentration for a Semi-Infinite Volume Source in Air and per Unit Activity per Unit Area for an Infinite Plane Source in the Ground

Energy (MeV)

Volume Source in Air

Plane Source in Ground

Air Kerma Rate per Unit of Activity Concentration [(Gy s−1)/(Bq m−3)]

Air Kerma Rate per Unit of Activity per Unit Area [(Gy s−1)/(Bq m−2)]

1.47 × 10−15 1.71 × 10−15 2.12 × 10−15 2.40 × 10−15 2.81 × 10−15 3.31 × 10−15 3.79 × 10−15 4.36 × 10−15 5.55 × 10−15 8.68 × 10−15 1.20 × 10−14 1.87 × 10−14 3.21 × 10−14 4.56 × 10−14 6.58 × 10−14 9.08 × 10−14 1.32 × 10−13 1.96 × 10−13 3.85 × 10−13 6.26 × 10−13

8.06 × 10−19 7.72 × 10−18 2.63 × 10−17 3.59 × 10−17 4.14 × 10−17 4.65 × 10−17 5.35 × 10−17 5.98 × 10−17 7.54 × 10−17 1.21 × 10−16 1.68 × 10−16 2.61 × 10−16 4.34 × 10−16 5.90 × 10−16 8.09 × 10−16 1.13 × 10−15 1.41 × 10−15 1.91 × 10−15 3.19 × 10−15 4.81 × 10−15

0.015 0.020 0.030 0.040 0.050 0.060 0.070 0.080 0.100 0.150 0.200 0.300 0.500 0.700 1.000 1.500 2.000 3.000 6.000 10.000

Source: Data from Saito, K. et al., Calculation of organ doses from environmental gamma rays using human phantoms and Monte Carlo methods. Part I: Monoenergetic sources and natural radionuclides in the ground, GSF-Report 2/90, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1990.

TABLE 17.2 Calculated Air Kerma at 1 m Height above the Ground per Disintegration per kg for a Semi-Infinite Volume Source in the Ground due to Natural Radionuclides Radionuclides

Air Kerma per Unit Source Intensity [Gy/(Disintegration kg−1)]

238

U series

1.29 × 10−13

232

Th series

1.68 × 10−13

40

K

1.16 × 10−14

Source: Data from Saito, K. et al., Calculation of organ doses from environmental gamma rays using human phantoms and Monte Carlo methods. Part I: Monoenergetic sources and natural radionuclides in the ground, GSF-Report 2/90, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1990.

Applications to Environmental Exposures

417

derived from the initial gamma ray fields, with respect to height, energy, and angle. These height-dependent double differential (with respect to angle of incidence and photon energy) gamma ray fields were then incorporated into the organ dose calculation with the voxel phantoms.

17.4 Organ Dose Calculations for Monoenergetic Sources: The Voxel Computational Phantoms and the Monte Carlo Codes To derive the organ doses, we performed Monte Carlo calculations for several voxel computational phantoms developed at the Helmholtz-Zentrum München (former GSF). Earlier calculations were performed for the voxel computational phantoms of a 7-year-old child and an 8-week-old baby.7,8,10 The Monte Carlo code used for the transport code in the body was the GSF code,14,15 which computes the dose deposited by photons from an external or internal source in various sections of a different media of the body. The code, which is a successor of the ALGAM code16 developed at Oak Ridge National Laboratory, is based on the fractional photon technique, and uses the kerma approximation. As an example of these calculations, Table 17.3 shows dose conversion factors normalized to source intensity due to natural radionuclides in the ground for the BABY computational phantom. The most recent calculations involve the new ICRP reference voxel adult male and female computational phantoms, ICRP-AM and ICRP-AF (Chapter 15 and Refs.17,18). These phantoms, described extensively in Chapter 15, are based on medical image data of real persons, and are consistent with the information given in ICRP Publication 8919 on the TABLE 17.3 Dose Conversion Factors Normalized to Source Intensity due to Natural Radionuclides in the Ground for the Model BABY of an 8-Week-Old Infant 238

Bladder Colon Liver Lungs Ovaries Red marrow Skeleton Skin Stomach Thyroid HE

U Series −1

1.13 × 10 1.14 × 10−1 1.08 × 10−1 1.10 × 10−1 1.38 × 10−1 1.11 × 10−1 1.31 × 10−1 1.22 × 10−1 1.12 × 10−1 1.07 × 10−1 1.17 × 10−1

232

Th Series −1

1.58 × 10 1.62 × 10−1 1.49 × 10−1 1.55 × 10−1 1.92 × 10−1 1.54 × 10−1 1.80 × 10−1 1.67 × 10−1 1.44 × 10−1 1.40 × 10−1 1.60 × 10−1

40

K

9.02 × 10−3 1.12 × 10−2 9.22 × 10−3 1.05 × 10−2 1.22 × 10−2 1.04 × 10−2 1.12 × 10−2 1.11 × 10−2 9.98 × 10−3 9.60 × 10−3 1.10 × 10−2

Source: Saito, K. et al., Calculation of organ doses from environmental gamma rays using human phantoms and Monte Carlo methods. Part I: Monoenergetic sources and natural radionuclides in the ground, GSF-Report 2/90, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1990. Note: Dose conversion factors in pSv per disintegration per kg.

418

Handbook of Anatomical Models for Radiation Dosimetry

reference anatomical and physiological parameters for both male and female subjects and represent, therefore, the reference Caucasian individual. Additionally, we employed other voxel computational phantoms developed at the Helmholtz-Zentrum München (former GSF).20 These computational phantoms, described in Chapter 3, cover both sexes and a wide range of ages and statures. Furthermore, we also used the adult voxel phantom of Zubal et al.21 The dose conversion coefficients have been computed with a user code22 that uses EGSnrc23 for the particle transport and scoring of the energy depositions. EGSnrc is a variant of EGS424 with improved physics and electron transport algorithm. In particular, we included bound Compton scattering and photoelectrons from K, L, and M shells. For both effects, we took into account the resulting fluorescence or Auger and Coster–Kronig electrons. In the computations of this work, the tracks of primary and secondary photons are followed down to kinetic energies of 2 keV, those of secondary electrons down to 20 keV. Electrons are assumed to deposit the energy continuously. The computational phantoms stand on the soil, simulated as a planar air–ground interface. The principles of calculation are common for all the computational phantoms used for the calculations. Both scatter and absorption of the radiation in both the air and the ground was considered in the calculation. Eighteen monochromatic photon energies were considered ranging from 15 keV to 10 MeV. For each irradiation, we simulated 7.5–20 million photon histories. This led to coefficients of variance for the calculated organ dose conversion coefficients that were generally around 0.2% for large organs and around to 1% for small organs. We estimated dose conversion coefficients for several organs, including the critical ones, i.e., active bone marrow, colon, lung, stomach, glandular breast tissue, gonads, urinary bladder, esophagus, liver, thyroid, endosteum (approximated by the mixture of the skeleton), brain, salivary glands, skin, adrenals, extrathoracic (ET) region, gallbladder, heart, kidneys, lymphatic nodes, muscle, oral mucosa, pancreas, prostate/uterus, small intestine, spleen, and thymus. From the dose conversion coefficients of the male ICRP-AM and the female ICRP-AF the effective dose, E, was calculated as defined at the ICRP recommendations of 2007.25 Figure 17.1 shows the dose conversion coefficients for the lungs submersion in a radioactive cloud (volume source in air), and Figure 17.2 shows the dose conversion coefficients for the gonads and for surface contamination in the voxel computational phantoms ICRP-AM, ICRP-AF, Visible Human, Voxelman, Donna, Helga and BABY.26 ICRP-AM and ICRP-AF have heights and weights almost identical with that of the ICRP Reference Man. The Visible Human, segmented at the GSF from the CT pictures obtained from the National Library of Medicine’s Visible Human Project, is tall and broadly built (180 cm height, 103 kg weight) and is more suitable to simulate bigger individuals. (Another voxel phantom of the same individual exists, segmented from whole body CT/MR/color photographic images27). Similarly, the female phantoms Helga (170 cm height, 81 kg weight) and Donna (176 cm height, 79 kg weight) represent overweight individuals. BABY is the voxel phantom of an 8-week-old baby and the data shown in Figures 17.1 and 17.2 stem from new calculations.26 Table 17.4 shows the effective dose per air kerma free-in-air for the ICRP adult reference computational phantoms and the BABY, in terms of air submersion and ground source. This has been calculated according to the ICRP 2007 recommendations.25 In order to calculate the effective dose of the BABY, the following surrogate organs need to be used: for endosteum, the mixture of the skeleton; for salivary glands, the thyroid; for ET, the thyroid; for lymphatic nodes, the muscle; for oral mucosa, the thyroid; for prostate, the uterus; furthermore, to the computational phantom, which stems from a CT of a female infant,

Applications to Environmental Exposures

419

Organ equivalent dose per air kerma (Sv/Gy)

1.2

Lungs

1.0

0.8

0.6

0.4 ICRP-AF ICRP-AM Donna Helga

0.2 0.0 0.01

0.10

Vishum Voxelman BABY

1.00

10.00

E (MeV) FIGURE 17.1 Dose conversion coefficients for lungs for submersion in a radioactive cloud (volume source in air) for the voxel models ICRP-AM, ICRP-AF Visible Human, Voxelman, Donna, Helga, and BABY.

Organ equivalent dose per air kerma (Sv/Gy)

1.2

Gonads

1.0

0.8

0.6

0.4 ICRP-AF ICRP-AM Donna Helga

0.2 0.0 0.01

0.10

1.00

Vishum Voxelman BABY

10.00

E (MeV) FIGURE 17.2 Dose conversion coefficients for gonads and for surface contamination, for the voxel models ICRP-AM, ICRP-AF Visible Human, Voxelman, Donna, Helga, and BABY.

testes were added. In Figure 17.3, the effective dose for the CHILD is shown, for air and ground geometry. The conversion coefficients for the reference female computational phantom ICRP-AF were found to be up to 10%–15% higher than those for the male computational phantom ICRP-AM, due to the smaller body size of the female computational phantom. As expected, the conversion coefficients for smaller or slender individuals—in this case for the BABY,

Handbook of Anatomical Models for Radiation Dosimetry

420

TABLE 17.4 Effective Dose in Sv/Gy, Calculated According to ICRP 10325 Definition, for the ICRP Adult Reference Models and the BABY, for Air Submersion and Ground Source Air Energy (MeV)

Effective dose/air kerma at 1 m above ground (Sv/Gy)

0.015 0.020 0.030 0.040 0.050 0.060 0.070 0.080 0.100 0.150 0.200 0.300 0.500 1.000 2.000 3.000 6.000 10.000

Ground

ICRP Adult

BABY

ICRP Adult

BABY

0.019 0.046 0.150 0.284 0.404 0.495 0.557 0.601 0.642 0.678 0.676 0.684 0.694 0.717 0.766 0.792 0.795 0.779

0.041 0.114 0.340 0.572 0.713 0.779 0.834 0.852 0.882 0.877 0.865 0.872 0.874 0.900 0.924 0.948 0.929 0.876

0.012 0.037 0.142 0.301 0.453 0.570 0.648 0.693 0.730 0.733 0.722 0.717 0.723 0.757 0.808 0.825 0.820 0.790

0.036 0.118 0.404 0.690 0.865 0.964 1.009 1.032 1.030 1.001 0.983 0.976 0.968 0.988 1.010 1.008 0.959 0.877

1.0

0.8

0.6 CHILD 0.4

0.2 Air Ground 0.0 0.01

0.1

1

10

Photon energy (MeV) FIGURE 17.3 Effective dose in Sv/Gy, calculated according to an ICRP 60 definition for the voxel model CHILD, for air submersion and ground source. (Modified from Saito, K. et al., Health Phys., 74, 698, 1998.)

Applications to Environmental Exposures

421

ICRP-AF, and Voxelman—were higher than those for bigger individuals like Donna and Visible Human, due to the reduced shielding of the single organs from the incident photon beams from the overlying tissue. This is, however, not always the case, as shown in other studies concerning external exposures, indicating that differences in organ topology from individual to individual have an influence that may overrule that of the external body dimensions. Considering the two different source types, we can see that the equivalent dose conversion coefficients for the volume source in air are generally lower than those for the plane source in the ground. This is due to the different angular distribution of the radiation impinging on the body: the gamma ray field from a source in the air is nearly isotropic with respect to directions from the upper hemisphere, while the incident directions of the gamma rays from a plane source have strong horizontal bias, and most photons come from horizontal directions. Since the human body standing vertically has a reduced shielding effect for photons coming from horizontal directions, this leads to the higher doses resulting from this geometry. However, in most cases, the differences in the conversion coefficients were found to be less than 20%.

17.5 Calculation of Doses for Radionuclides Only photons, including bremsstrahlung, and electrons emitted by the radionuclides are sufficiently penetrating to traverse the overlying tissues of the body and contribute to the dose to tissues and organs of the body. The energy spectra of emitted radiation are either discrete, as in the case of photons, or continuous, as in the case of beta particles and bremsstrahlung. The dose coefficient H TS for tissue T for any exposure mode S can be expressed in Equation 17.1 as ⎡ ⎢ H = ⎢ j =e , γ ⎣ S T

∑∑ i

y j (E i )H

∞

S T, j

⎤

(E i ) + ∫ y j (E i )H T,S j (E ) d (E )⎥ 0

⎥ ⎦

(17.1)

where yj(Ei) is the yield of radiations of type i and discrete energy Ei yj(E) is the yield of radiations per nuclear transformation with continuous energy between E and E + dE The other summation is overall electron and photon radiations. The contribution of the radiations to the dose in tissue or organ T is defined by the quantity H TS which is estimated by means of Monte Carlo calculations and is given as a function of energy for tissue and organ T for each exposure.9 For photons, we have tabulated these data for several target tissues of the body at each of the monoenergetic photon energies. The contribution of electrons to dose to organs other than skin need not be considered, due to the short range in tissue of electrons emitted by radionuclides.9 By using the dose conversion coefficients for monoenergetic photons and by scaling them to the emissions of the radionuclides of interest, we can derive dose coefficients from radionuclides in the environment. For this convolution, the nuclear decay data from the ICRP Publication 38 were used.28 Table 17.5 shows new data from the Helmholtz-Zentrum

Handbook of Anatomical Models for Radiation Dosimetry

422

TABLE 17.5 Dose Equivalent Rates Per Unit Air Volume (nSv h−1/Bq m−3) for Photon Exposure from a Semi-Infinite Volume Source in Air and Per Unit Area (nSv h−1/kBq m−2) for Photon Exposure from an Infinite Plane Source at a Depth of 0.5 g cm−2 in the Ground, for Some Radionuclides Air Submersion Effective Dose

Gonads Dose −1

−3

nSv h / Bq m Be-7 Na-22 Na-24 Mg-28 Sc-46 Cr-51 Mn-54 Co-57 Co-58 Co-60 Cu-64 Zn-65 Sr-85 Y-88 Sr-89 Zr-89 Nb-94 Nb-95 Mo-99 Tc-99 m Tc-99 Ru-103 Rh-105 Ru-105 Ag-108 m Cd-109 Sb-122 I-123 I-124 Te-129 m I-131 I-132 Ba-133 I-133 Cs-134 Cs-137 Ba-140 Ce-141 Ce-144 Pr-144

7.86E−03 3.69E−01 7.50E−01 2.32E−01 3.42E−01 4.85E−03 1.40E−01 1.62E−02 1.62E−01 4.35E−01 3.05E−02 9.95E−02 8.10E−02 4.73E−01 1.42E−05 1.93E−01 2.62E−01 1.27E−01 2.42E−02 1.74E−02 0.00E+00 7.52E−02 1.20E−02 1.27E−01 2.62E−01 8.30E−04 7.15E−02 2.19E−02 1.80E−01 5.10E−03 5.95E−02 3.74E−01 5.68E−02 9.81E−02 2.57E−01 0.00E+00 2.84E−02 1.01E−02 2.45E−03 7.11E−03

6.99E−03 3.27E−01 6.91E−01 2.05E−01 3.01E−01 4.26E−03 1.23E−01 1.45E−02 1.43E−01 3.84E−01 2.72E−02 8.75E−02 7.21E−02 4.21E−01 1.25E−05 1.70E−01 2.31E−01 1.12E−01 2.14E−02 1.55E−02 0.00E+00 6.69E−02 1.05E−02 1.12E−01 2.32E−01 6.83E−04 6.35E−02 1.95E−02 1.60E−01 4.49E−03 5.25E−02 3.30E−01 4.99E−02 8.71E−02 2.27E−01 0.00E+00 2.52E−02 9.00E−03 2.17E−03 6.37E−03

Plane Source in Ground Effective Dose

Gonads Dose −1

nSv h /kBq m−2 1.47E−01 5.54E+00 8.21E+00 3.02E+00 4.52E+00 7.35E−02 2.01E+00 2.48E−01 2.46E+00 5.39E+00 5.70E−01 1.28E+00 1.53E+00 5.75E+00 1.95E−04 2.84E+00 3.89E+00 1.92E+00 3.69E−01 2.63E−01 0.00E+00 1.42E+00 1.80E−01 2.01E+00 4.36E+00 8.44E−03 1.26E+00 3.34E−01 2.75E+00 7.90E−02 9.64E−01 5.63E+00 8.91E−01 1.71E+00 4.07E+00 0.00E+00 5.06E−01 1.53E−01 3.72E−02 9.38E−02

1.47E−01 5.53E+00 8.32E+00 3.02E+00 4.50E+00 7.45E−02 2.00E+00 2.60E−01 2.46E+00 5.38E+00 5.71E−01 1.28E+00 1.53E+00 5.74E+00 1.95E−04 2.83E+00 3.88E+00 1.92E+00 3.70E−01 2.74E−01 0.00E+00 1.43E+00 1.83E−01 2.02E+00 4.36E+00 1.09E−02 1.26E+00 3.48E−01 2.76E+00 8.02E−02 9.73E−01 5.63E+00 9.13E−01 1.71E+00 4.06E+00 0.00E+00 5.09E−01 1.60E−01 3.97E−02 9.45E−02

Applications to Environmental Exposures

423

TABLE 17.5 (continued) Dose Equivalent Rates Per Unit Air Volume (nSv h−1/Bq m−3) for Photon Exposure from a Semi-Infinite Volume Source in Air and Per Unit Area (nSv h−1/kBq m−2) for Photon Exposure from an Infinite Plane Source at a Depth of 0.5 g cm−2 in the Ground, for Some Radionuclides Air Submersion Effective Dose

Gonads Dose −1

−3

nSv h / Bq m Eu-154 Yb-169 Hf-181 Ta-182 Re-186 Ir-192 Hg-197 Tl-201 Hg-203 Pb-210 Ra-228 Np-239

2.06E−01 3.73E−02 8.55E−02 2.16E−01 2.57E−03 1.27E−01 7.22E−03 1.05E−02 3.57E−02 1.53E−04 3.13E−12 2.30E−02

1.82E−01 3.29E−02 7.60E−02 1.90E−01 2.28E−03 1.12E−01 6.36E−03 9.28E−03 3.13E−02 1.29E−04 1.12E−12 2.03E−02

Plane Source in Ground Effective Dose

Gonads Dose −1

nSv h / kBq m−2 2.76E+00 5.75E−01 1.55E+00 2.76E+00 3.96E−02 2.08E+00 1.17E−01 1.67E−01 5.26E−01 2.29E−03 1.60E−15 3.46E−01

2.76E+00 6.08E−01 1.56E+00 2.77E+00 4.19E−02 2.09E+00 1.27E−01 1.79E−01 5.37E−01 2.66E−03 8.32E−15 3.60E−01

München29 for the reference adult computational phantoms: dose equivalent rates (of the effective dose and gonad) per unit activity for the photon exposure from several radionuclides are shown, from radionuclides in the air and radionuclides deposited on the ground.

References 1. Dillman, L.T. Absorbed gamma dose rate for immersion in a semi-infinite radioactive cloud, Health Phys, 27, 571, 1974. 2. Poston, J.W. and Synder, W.S. A model for exposure to a semi-infinite cloud of a photon emitter, Health Phys, 26, 287, 1974. 3. O’Brien, K. and Sanna, R. The distribution of absorbed dose-rates in humans from exposure to environmental gamma rays, Health Phys, 30, 71, 1976. 4. Koblinger, L. and Nagy, G. Calculation of the relationship between gamma source distribution in the soil and external doses, Sci. Total Environ., 45, 357, 1985. 5. Jacob, P. et al. Effective dose equivalents for photon exposures from plane sources on the ground, Radiat. Prot. Dosimetry, 14, 299, 1986. 6. DOE (Department of Energy). External dose-rate conversion factors for calculation of dose to the public, DOE/EH-0070, DOE, Washington, DC, 1988. 7. Saito, K. et al. Calculation of organ doses from environmental gamma rays using human phantoms and Monte Carlo methods. Part I: Monoenergetic sources and natural radionuclides in the ground, GSF-Report 2/90, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1990.

424

Handbook of Anatomical Models for Radiation Dosimetry

8. Petoussi, N. et al. Organ doses for fetuses, babies, children and adults from environmental gamma-rays, Radiat. Prot. Dosimetry, 37, 31, 1991. 9. Eckerman, K.F. and Ryan, J.C. Federal Guidance Report No. 12, Oak Ridge National Laboratory, External exposure of radionuclides in air, water, and soil, Oak Ridge, TN, 1993. 10. Jacob, P. et al. Calculation of organ doses from environmental gamma rays using human phantoms and Monte Carlo methods, Part II: Radionuclides distributed in the air or deposited on the ground, GSF-Report 12/90, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1990. 11. Saito, K. et al. Organ doses as a function of body-weight for environmental gamma-rays, J. Nucl. Sci. Technol., 28, 627, 1991. 12. Saito, K., Petoussi-Henss, N., and Zankl, M. Calculation of the effective dose and its variation from environmental gamma ray sources, Health Phys., 74, 698, 1998. 13. Saito, K. and Moriuchi, S. Development of a Monte-Carlo Code for the calculation of gammaray transport in the natural-environment, Radiat. Prot. Dosimetry, 12, 21, 1985. 14. Kramer, R. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods: Part I. The male (ADAM) and female (EVA) adult mathematical phantoms, GSF-Report S-885, Institut fuer Strahlenschutz, GSF-Forschungszentrum fuer Umwelt und Gesundheit, Neuherberg-Muenchen, 1982. 15. Veit, R. et al. Tomographic anthropomorphic models, Part I: Construction technique and description of models of an 8 week old baby and a 7 year old child, GSF-Report 3/89, GSF—National Research Center for Environment and Health, Neuherberg, Germany, 1989. 16. Warner, G.G. and Craig, A.M. ALGAM, a computer program for estimating internal dose from gamma-ray sources in a man phantom, Report ORNL-TM-2250, Oak ridge National Laboratory, Oak Ridge, TN, 1968. 17. ICRP. Adult Reference Computational Phantoms. ICRP Publication 110, International Commission of Radiological Protection, Elsevier, Amsterdam, 2009. 18. Zankl, M., Eckerman, K.E., and Bolch, W.E. Voxel-based models representing the male and female ICRP reference adult—The skeleton, Radiat. Prot. Dosimetry, 127, 174, 2007. 19. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, 2003. 20. Petoussi-Henss, N. et al. The GSF family of voxel phantoms, Phys. Med. Biol., 47, 89, 2002. 21. Zubal, I.G. et al. Computerized three-dimensional segmented human anatomy, Med. Phys., 21, 299, 1994. 22. Schlattl, H., Zankl, M., and Petoussi-Henss, N. Organ dose conversion coefficients for voxel models of the reference male and female from idealized photon exposures, Phys. Med. Biol., 52, 2123, 2007. 23. Kawrakow, I. and D.W.O. Rogers. The EGSnrc code system: Monte Carlo simulation of electron and photon transport PIRS Report 701, National Research Council of Canada (NRCC) Ottawa, 2003. 24. Nelson, W.R. et al. The EGS4 Code System, SLAC Report 265, Stanford Linear Accelerator Center, Stanford, CA, 1985. 25. ICRP. The 2007 Recommendations of the International Commission on Radiological Protection. ICRP Publication 103, Ann. ICRP, 37, 1, 2007. 26. Schlattl, H. Personal communication, 2008. 27. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Phys., 78, 476, 2000. 28. ICRP. Radionuclide Transformations: Energy and Intensity of Emissions, ICRP Publication 38, Pergamon Press, Oxford, 1983. 29. Petoussi-Henss, N. Unpublished material, 2008.

18 Applications to External Radiation Exposures in Nuclear Power Plants Warren Dan Reece, Chan Hyeong Kim, and X. George Xu

CONTENTS 18.1 Introduction ...............................................................................................................425 18.2 Two-Dosimeter Algorithms for Effective Dose Estimation ................................ 427 18.2.1 Background................................................................................................... 427 18.2.2 Development of an Optimal Algorithm ................................................... 428 18.2.3 Application to Idealized Photon Beams ...................................................430 18.2.4 Application to Steam Generator Channel Head ......................................434 18.3 PRDC—A Software Package for Personnel Radiation Dose Calculation ......... 435 18.4 EPRI EDE Calculator—A Software Package for Assessing EDE and Effective Dose from Hot Particles on the Skin ..............................................442 18.5 Summary and Future Work ....................................................................................445 References .............................................................................................................................446

18.1 Introduction The health risk from radiation exposure can be assessed by the use of effective dose (E), or effective dose equivalent (EDE) (HE), recommended by the International Commission on Radiological Protection (ICRP) as the primary dose quantity in radiation protection. The concept of EDE was first introduced in 1977 and then revised in 1990, when the name was changed to effective dose; however, the U.S. Nuclear Regulatory Agency (U.S. NRC) has only adopted the former definition in the United States.1,2 The effective dose provides a technical basis to compare the radiation health risk with other kinds of risks. The same magnitude of effective dose can be assumed to produce the same degree of health risks irrespective of the organs or tissues involved in irradiation. Effective dose is defi ned as a weighted average of the doses to many organs and tissues in the body. It is not possible, however, to measure directly the absorbed doses in the organs and tissues. Therefore, a number of quantitative relationships have been developed, including a quantity called personal dose equivalent, Hp(d), which was developed as an operational quantity for personal dose monitoring.3 For strongly penetrating radiation, researchers have adopted a value of 10 mm, the depth below the body surface. This quantity is close to the deep dose equivalent (DDE) that is used to determine the whole-body dose in the United States. 425

426

Handbook of Anatomical Models for Radiation Dosimetry

For external exposures, Hp(10) has been thought to approximate the effective dose, or EDE. This is only true, however, when the dosimeter that is used to measure Hp(10) is facing the radiation source. In fact, Hp(10) will significantly underestimate the effective dose if the dosimeter is positioned on the opposite side of the photon entrance. For example, a personal dosimeter on the chest will significantly underestimate E (or HE) when a radiation worker is exposed to radiation from the back, because the radiation passes through the body and is attenuated before reaching the dosimeter. This underestimation can be significant (factors of 7–10) over the typical range of photon energies (0.08–1.0 MeV) and an estimated value of E (or HE) is often useless without knowing something about the irradiation geometry. Therefore, ICRP Publication 75 suggests that a personal dosimeter be worn at an appropriate position on the body in order to adequately indicate E (or HE), which is not possible practically, especially when the radiation field characteristics are unknown.4 This problem can be addressed by using two dosimeters, one on the chest and the other on the back, so that at least one dosimeter is always directly exposed to the radiation source. If a photon beam arrives from the back, then the dosimeter on the back would be directly exposed and would effectively respond to the photon beam, compensating for the underresponse of the dosimeter on the chest. This chapter discusses two-dosimeter algorithms that relate the dosimeter readings on the chest and back to the effective dose or EDE. A more detailed discussion follows on a two-dosimeter algorithm that was developed through a systematic optimization process using a computational human phantom and Monte Carlo dose calculations over many irradiation geometries. We also discuss the application of the two-dosimeter method to photon beams and to steam generator channel heads in nuclear power plants. The most popular method to determine effective dose is to use a computer simulation method, usually the Monte Carlo particle transport method. The Monte Carlo–based dose calculation, however, requires sophisticated software and computers as well as a good understanding of radiation physics. In particular, one first should construct, or acquire, if available, a computational phantom of the human body. Then, the computational phantom has to be imported to a Monte Carlo particle transport code and the code run to calculate organ and tissue doses. This kind of approach is complicated and is too difficult and timeconsuming to be used by a practicing health physicist. Section 18.3 reviews a software package, called personnel radiation dose calculation (PRDC),5 which was developed to calculate effective dose and radiation doses to the organs/tissues and to personnel dosimeters for external photon exposures. PRDC is easy to use and calculates the necessary dose quantities quickly. Hot particles (small activation products or fuel fragments) can be found on radiation workers’ skin or protective clothing, and can deliver a highly localized skin dose from the beta and low-energy gamma-ray emitters or a nonuniform whole-body dose from the high-energy gamma-ray emitters. The whole-body dose from an exposure to hot particles is relatively small compared to the dose to the skin, and software packages have been developed and used regularly only for skin dose calculations. In 2003, however, the U.S. NRC issued the Regulatory Information Summary (RIS) 2003–2004, which encouraged licensees to use the EDE in place of the DDE.6 Importantly, this RIS permits assessment of the EDE for localized skin contamination. This change in the regulatory environment stimulated the development of a software package that can calculate the effective dose from hot particles. Section 18.4 discusses a software package called EPRI EDE Calculator,7 which can be used to calculate the EDE (HE) and effective dose (E) for hot particles on the skin.

Applications to External Radiation Exposures in Nuclear Power Plants

427

18.2 Two-Dosimeter Algorithms for Effective Dose Estimation 18.2.1 Background The use of two dosimeters has been recommended by several authors to overcome problems associated with simple HP(10) concepts. Lakshmanan et al. used two CaSO4:Dy thermoluminescence dosimeter (TLD) badges to measure Hp(10) on a 30 cm3 water phantom.8 The TLD badges were placed on the front and back surfaces of the phantom to simulate personal dosimeters worn on the chest and back of a radiation worker. Three irradiation geometries were considered: antero posterior (AP), postero anterior (PA), and lateral (LAT). They concluded that the sum of the dosimeter readings from these two TLD badges divided by 1.5 provides a conservative estimate of HE. The underestimation was less than 5% for all irradiation geometries considered in their study. Reece, who was supported in the early 1990s by a grant from the Electric Power Research Institute, and his PhD student Xu at Texas A&M University, adopted the MIRD (Medical Internal Radiation Dose)–type stylized phantom (Figure 18.1) originally developed by Cristy and Eckerman to study the relationship between different exposure geometries and the two-dosimeter algorithms.9–12 Two air-filled 1 cm radius pseudo dosimeters, which were essentially isotropic-responding, were defined on the chest and back of the worker’s computational phantom to simulate readings of personal dosimeters. These studies considered broad parallel photon beams of six directions and point sources at various locations near the surface of the computational phantom. To report a personal dose, the two dosimeter readings were weighted according to the relative sizes of the dosimeter readings (higher dosimeter reading vs. lower dosimeter reading), not to the locations (front dosimeter reading vs. back dosimeter reading). They concluded that a weightings of 0.75 and 0.25 for the higher and lower dosimeter readings, respectively, were optimum, and does not underestimate HE by more than 11% for all irradiation geometries considered in their study. They also noted that this algorithm overestimated HE to a less degree for the LAT irradiation geometry than does the algorithm developed by Lakshmanan et al.8 Finally, the algorithm was accepted by the U.S. NRC, in the RIS 2003–2004, as the only acceptable algorithm for the estimation of effective dose.6 Claycamp sought to optimize dosimeter weighting factors through an optimization process using a commercial nonlinear optimization method (Quattro Pro) and various mixtures of AP and PA irradiation geometries.13 The primary objective of his study was to find an optimal combination of dosimeter readings that minimizes the uncertainties in the estimation of HE for an unknown combination of AP and PA irradiation geometries. He concluded that the best combination of dosimeter readings that not only minimizes the uncertainties in the estimation of HE for an unknown combination of AP and PA irradiation geometries, but also minimizes the likelihood of underestimation had a factor of 0.7 and 0.3 for the front and back dosimeter readings, FIGURE 18.1 respectively. The underestimation was less than 19% for all MIRD-type mathematical model combinations of AP and PA irradiation geometries and phoused to calculate effective dose ton energies considered in his study. (E) and EDE (HE).

428

Handbook of Anatomical Models for Radiation Dosimetry

In December 1995, the NCRP officially recommended using two dosimeters, one on the chest and the other on the back, for exposure scenarios “where the irradiation geometry or photon energy is unknown or difficult to characterize.”14 Through an optimization process similar to Claycamp’s, the NCRP also developed an algorithm, weighting 0.55 and 0.5 for the front and back dosimeter readings, respectively, with a comment that this algorithm generally does not underestimate HE by more than 10% (seldom by more than 5%) and generally does not overestimate HE by more than a factor of 2–3 (most often less than a factor of 2). These studies generally show that using two dosimeters and a suitable algorithm that combines readings from these two dosimeters can solve the underestimation problem of the single-dosimeter approach for posterior incident radiation. Unfortunately, these studies were not complete. First, Lakshmanan et al. used a simple cubic water phantom, which does not represent a human trunk for posterior incident radiation.8 Xu used a realistic anthropomorphic r, which can represent the human trunk even for posterior incident radiation, but did not completely the optimize dosimeter weighting factors.12 Claycamp and the NCRP, again, used a simple slab computational phantom that cannot appropriately represent a human trunk for lateral and posterior incident radiation, even though they were aware of Xu’s work.13,14 The optimization process was taken by Claycamp to find an optimal combination of dosimeter readings, but in this case only two irradiation geometries (AP and PA) were considered for optimization. Later, another PhD student, Kim et al.,15 investigated the use of two dosimeters for effective dose estimation more comprehensively. In this study, Kim determined an optimal combination of the chest and back dosimeter readings through the systematic optimization process by using the MIRD-type stylized phantom and the MCNP (Monte Carlo N-Particle) code, which is reviewed in the next section. 18.2.2 Development of an Optimal Algorithm There is no “ideal” algorithm, or an “ideal” combination of dosimeter readings, which would perfectly estimate E (or HE) for all irradiation geometries and photon energies. Therefore, an “optimal” combination of dosimeter readings was sought that would suitably estimate E (or HE) for most irradiation geometries and typical photon energies. The optimal algorithm was defi ned as an algorithm that neither underestimates E (or HE) by more than 10%, nor overestimates by more than 100% for the “most” incident directions of broad parallel photon beams for the typical range of photon energies (0.08–1.0 MeV). To calculate the distribution of radiation dose within the computational phantom and personal dosimeters on the chest and back, the exposure conditions and photon interactions in the worker’s body were simulated using the MIRD-type stylized phantoms and the MCNP code. Hundreds of incident beam directions (every 15° for polar angles and every 15° for azimuthal angles) and three photon energies (0.08, 0.3, and 1.0 MeV) were considered, which adequately bound the extremes of the exposure situations in nuclear utilities.16,17 A systematic optimization process was followed to derive the optimal algorithm. First, E (or HE) and two dosimeter readings, i.e., front/chest (Rf) and back (Rb), were determined for hundreds of incident beam directions. To find an optimal combination of dosimeter readings according to the locations of the dosimeter readings, E (or HE) was estimated as a combination of the “front (chest)” and “back” dosimeter readings as follows, with β varying between 0 and 1:

Applications to External Radiation Exposures in Nuclear Power Plants

429

E(est) or HE(est) = b × Rf + (1 – b) × Rb

(18.1)

Then, the E (or HE) assessment ratio (r) was calculated as the ratio of E(est)/E or HE(est)/HE for each beam direction. Note that an r value of unity means a “perfect” estimation of E (HE); r > 1, an overestimation; and r < 1, an underestimation. To be the optimal algorithm, an algorithm should show the least fluctuation in the distribution of r over all beam directions. The study used the ratio of the maximum and minimum values of r over all beam directions (=rmax/rmin) as the index of the fluctuation of r. Figure 18.2 shows rmax/rmin as a function of β. The optimal value of β, which minimizes the fluctuation in the distribution of r, is clear, about 0.58 (E: 0.575–0.595, HE: 0.565–0.585), for all photon energies considered in this study. The optimal combination of dosimeter readings is, then, E(est) or HE(est) = 0.58Rf + 0.42Rb

(18.2)

Although the optimal combination of dosimeter readings minimizes the fluctuation in the distribution of r, the formula may be systematically biased high or low. Therefore, the equation was further adjusted by normalizing against the dosimetry results of the AP direction, the most important and standard irradiation geometry in radiation protection. Also note that the minimum value of r typically occurs in the AP direction because the value of E (or HE) for the AP direction is greater than those for other directions while the dosimeter readings, and thus the calculated E(est) (HE(est)) value, for the AP direction is comparable to those for other directions. If the algorithm is normalized as follows, then, it seldom underestimates E (or HE) by more than 10% for all incident beam directions and photon energies: 20 0.08 MeV (HE) 0.3 MeV (HE) 1.0 MeV (HE)

rmax /rmin

15

0.08 MeV (E) 0.3 MeV (E) 1.0 MeV (E)

10

5 Minimum occures within 0.565–0.585 (HE) and 0.575–0.595 (E) for all energies (center: 0.58) 0 0.0

0.5 β

1.0

FIGURE 18.2 Ratios of the maximum and minimum values of HE and E assessment ratios over the beam directions (=rmax/rmin) as a function of β (the weighting factor for the front [chest] dosimeter reading). (From Kim, C.H. et al., Radiat. Prot. Dosim., 81, 101, 1999. With permission.)

Handbook of Anatomical Models for Radiation Dosimetry

430

E(est) = h(E)[0.58Rf + 0.42Rb]

(18.3)

HE(est) = h(HE)[0.58Rf + 0.42Rb]

(18.4)

or

If dosimeters are calibrated for personal dose equivalent on the ICRU slab phantom, Hp,slab(10),18 h(E) is calculated as 1.06, 1.06, 1.02, and 0.98 for 0.08, 0.3, 1.0, and 2.0 MeV photons, respectively, and an h(E) value of 1.02 can be used with less than 5% error unless the photon energy is very low (<80 keV). Likewise, h(HE) is calculated as 1.09, 1.09, 1.04, and 1.00 for these photon energies, respectively, and an h(HE) of 1.05 can be used with a 5% error. These h values also can be applied to dosimeters that are calibrated for DDE (Hd) with an insignificant error. The resulting algorithm is E(est) = 1.02[0.58Rf + 0.42Rb]

(18.5)

HE(est) = 1.05[0.58Rf + 0.42Rb]

(18.6)

18.2.3 Application to Idealized Photon Beams Figures 18.3 through 18.5 show the distributions of the E assessment ratio, r (=E(est)/E), over the incident beam directions for 0.08, 0.3, and 1.0 MeV photon beams, respectively, when one and two dosimeters are used to estimate E. Generally, a single dosimeter on the chest estimates E well for a broad range of frontal incident photon beams, neither underestimating E by more than few percent, nor overestimating by more than about 50%. As the azimuthal angle approaches 90° (i.e., LAT) or the polar angle approaches 0° (overhead, OH) or 180° (underfoot, UF), the single dosimeter on the chest significantly overestimates E.

Two dosimeter

7

7

6

6

5

5

4

4

r(=E(est)/E)

r(=E(est)/E)

Single dosimeter

3 2 1 0

3 2 1 0

PA 0

45

Pola r

PA 180

AP

135

90

ang l

e

90

135 180

hal mut A zi

45 0

a

nlge

0

45

180

AP

135

90

Pola r an gle 135

90 180

45 0

ge

l anl

utha Azim

FIGURE 18.3 Distributions of the E assessment ratio, r(=E(est)/E), over the incident beam directions of a 0.08 MeV photon beam when single- and two-dosimeter approaches are used.

Applications to External Radiation Exposures in Nuclear Power Plants

Two dosimeter

7

7

6

6

5

5

4

4

r(=E(est)/E)

r(=E(est)/E)

Single dosimeter

431

3 2 1 0

3 2 1 0

PA 0

45 AP Pola 90 135 r an gle 180 0

PA

180 135 90 lge l an 45 tha u m A zi

0

AP 45 Pola 90 135 r an gle 180 0

45

Az

180 135 lge al an

90 th imu

FIGURE 18.4 Distributions of the E assessment ratio, r(=E(est)/E), over the incident beam directions of a 0.30 MeV photon beam when single- and two-dosimeter approaches are used.

Two dosimeter 7

6

6

5

5

4

4

r(=E(est)/E)

r(=E(est)/E)

Single dosimeter 7

3 2 1

3 2 1 0

0

PA

PA 0

AP 45 Pola 90 r an 135 gle 180 0

180 135 e 90 anlg 45 thal mu i z A

0

AP 45 Pola 90 r an gle 135 180

0

180 135 90 nlge hal a mut

45 A zi

FIGURE 18.5 Distributions of the E assessment ratio, r(=E(est)/E), over the incident beam directions of a 1.00 MeV photon beam when single- and two-dosimeter approaches are used.

Finally, as the azimuthal angle exceeds 90° (i.e., the photon beam comes from the back), the single-dosimeter approach (on the chest) abruptly underestimates E. The maximum underestimations are by factors of 9.1, 6.7, and 6.7 for 0.08, 0.3, and 1.0 MeV photon beams, respectively. These underestimations occur because the body of the computational phantom (or a radiation worker) shields the dosimeter on the chest when the photon beam arrives from the back of the body. Obviously, these underestimations are so significant that an estimated value of E is almost meaningless unless the direction of a photon beam or irradiation geometry is known.

Handbook of Anatomical Models for Radiation Dosimetry

432

The use of two dosimeters and the associated optimal algorithm (two-dosimeter approach) completely solves the underestimation problem of the single-dosimeter approach for posterior incident photon beams. The two-dosimeter approach never underestimates E by more than 13% for all beam directions and photon energies between 0.08 and 1.0 MeV. When a photon beam comes from the back, the dosimeter on the back is directly exposed to the photon beam and compensates for the underresponse of the front (chest) worn dosimeter that is shielded by the body of a radiation worker. The two-dosimeter approach also estimates E well for a broad range of frontal incident photon beams, neither underestimating E by more than 11%, nor overestimating by more that about 50%, considering all photon energies. Alternately, the use of two dosimeters would worsen the overestimation problems for the LAT, OH, and UF directions. For the LAT direction, the two-dosimeter approach overestimates E by factors of 2.7, 2.6, and 2.0 for 0.08, 0.3, and 1.0 MeV, respectively, which are 3.0

3.0 θ = ±0°

r0(W )

2.5 2.0

2.0

1.5

1.5

1.0

1.0

0.5

0.5

0.0

0

30

θ = ±22.5°

2.5

60

90

120

150

180

0.0

0

30

60

90

120

150

180

3.0

3.0 0.08 MeV (single)

θ = ±52.5°

2.5

0.3 MeV (single) 1.0 MeV (single)

2.5

θ = ±82.5°

0.08 MeV (two)

2.0

0.3 MeV (two)

2.0

r0(W )

1.0 MeV (two)

1.5

1.5

1.0

1.0

0.5

0.5

0.0

0

120 150 180 30 60 90 Horizontal incident angle, w (degrees)

0.0

0

30 60 90 120 150 180 Horizontal incident angle, w (degrees)

FIGURE 18.6 Averaged E assessment ratios, r(w), as a function of the horizontal incident angles (to the chest dosimeter), w, for averaging angles (θ) of ±0° (static exposure situation), ±22.5°, ±52.5°, and ±82.5°. Note that r(w) is the averaged E assessment ratio for beams swinging from (w − θ)° to (w + θ)° of incident angles. Data points with lines: twodosimeter approach; data points only: single-dosimeter approach.

Applications to External Radiation Exposures in Nuclear Power Plants

433

somewhat larger than those of the single-dosimeter approach (1.9, 2.0, and 1.8 for these energies, respectively). For the OH direction, the overestimations are factors of 6.5, 6.1, and 4.4; for the UF direction, the overestimations are factors of 6.2, 5.8, and 5.2. All of the results presented so far, however, are true only for ideal or static exposure situations where the computational phantom or a radiation worker is exposed to a single-direction photon beam without any movement during exposure. In real exposure situations, radiation workers move during exposure and, therefore, experience some sort of averaged response even for a single-direction photon beam. This will result in less overestimation of E for the LAT and OH directions. Figures 18.6 and 18.7 show averaged E assessment ratios, rθ(w), as a function of horizontal and vertical incident angles (w), respectively, for averaging angles (θ) of ±0°, ±22.5°, ±52.5°, and ±82.5°. An rθ(w) value is calculated by averaging the r values over a range of incident angles from (w − θ)° to (w + θ)°. The averaging angle of ±0° represents an ideal or static exposure situation, where a radiation worker is exposed to a fixed photon beam without 7.0

7.0

r0(W )

6.0

6.0

θ = ±0°

5.0

5.0

4.0

4.0

3.0

3.0

2.0

2.0

1.0

1.0

0.0

0

30

60

90

120

150

180

7.0

0.0

θ = ±22.5°

0

30

60

90

120

150

180

7.0 0.08 MeV (single)

θ = ±52.5°

6.0

0.3 MeV (single)

6.0

θ = ±82.5°

1.0 MeV (single) 0.08 MeV (two)

5.0

5.0

r0(W )

0.3 MeV (two) 1.0 MeV (two)

4.0

4.0

3.0

3.0

2.0

2.0

1.0

1.0

0.0

0

120 150 180 30 60 90 Vertical incident angle, w (degrees)

0.0

0

30 60 90 120 150 Vertical incident angle, w (degrees)

180

FIGURE 18.7 Averaged E assessment ratios, r(w), as a function of the vertical incident angles (to the chest dosimeter), w, for averaging angles (θ) of ±0° (static exposure situation), ±22.5°, ±52.5°, and ±82.5°. Note that r(w) is the averaged E assessment ratio for beams swinging from (w − θ)° to (w + θ)° of incident angles. Data points with lines: twodosimeter approach; data points only: single-dosimeter approach.

Handbook of Anatomical Models for Radiation Dosimetry

434

any movement during exposure. For an ideal exposure situation (θ = ±0°), the overestimations of E for the LAT and OH directions are factors of 2.6 and 6.5, respectively, when the two-dosimeter approach is used to estimate E. As the r values are averaged over a small range of beam directions (θ = ±22.5°), the overestimations decrease to factors of 2 and 3 for the LAT and OH directions, respectively. As the averaging angle increases further, the overestimations for the LAT and OH directions disappear. 18.2.4 Application to Steam Generator Channel Head The steam generator channel head is one of the most important exposure situations in the nuclear power plants. The results given here are for the combustion engineering (CE)– type steam generator. In the steam generator channel head, the radiation field is usually dominated by only two activation products, 60Co and 58Co. Typically, 95% of the radiation field arises from these two products.19 There are some other activation products such as Cr-51, Mn-54, Fe-59, etc., but the dose contribution from these is, in general, relatively small. The dosimeter performance was not calculated for a specific combination of 60Co and 58Co. Instead, the emission spectra of 60Co and 58Co were simplified to 1.25 and 0.66 MeV, respectively, instead of 1.17 and 1.33 MeV for 60Co and 0.51 and 0.81 MeV for 58Co. The dosimeter performance was then calculated separately for each of these 1.25 and 0.66 MeV photon sources. The x-rays, scattered (and secondary) photons, and other low-energy photons from some minor nuclides were represented by the 0.08 MeV photon source. Fourteen most likely worker positions for a steam generator channel head environment were considered (see Figure 18.8). The S, L, R, U, D, L45, and R45 stand for standard, 45°

45°

50 cm

AU/BU

50 cm

AS/BS AR/BR AL/BL

30 cm 30 cm

AL45/BL45 AR45/BR45 AD/BD

FIGURE 18.8 Various worker positions in a steam generator channel head. The S, L, R, U, D, L45, and R45 stand for standard, left-shifted, right-shifted, up-shifted, down-shifted, left 45°-rotated, and right 45°-rotated, respectively. Also note that for each case the model faces either the divider plate (A) or the interior wall (B). (From Kim, C.H. and Reece, W.D., Health Phys., 83, 243, 2002. With permission.)

Applications to External Radiation Exposures in Nuclear Power Plants

435

left-shifted, right-shifted, up-shifted, down-shifted, left 45°-rotated, and right 45°-rotated, respectively. Note that for each case the computational phantom faces either the divider plate (A) or the interior wall (B). Table 18.1 shows HE assessment ratios, r (=HE(est)/HE), for various worker positions, source components, and photon energies when the two-dosimeter approach (chest and back) is used to estimate HE. These results show clearly that the two-dosimeter approach does not underestimate HE significantly considering all worker positions, source components, and photon energies in this study. The maximum underestimation was 4%, which occurs when the divider plate alone is contaminated by pure 60Co and the worker position is AR45. In this case, however, the underestimation of HE will be quickly canceled by the overestimation of the wall contamination (20% overestimation, column 3) and U-tube contamination (5% overestimation, column 4), resulting in an overall overestimation of about 7% (column 6). The two-dosimeter approach never underestimates HE for 0.66 MeV (58Co) and 0.08 MeV photon sources. The results also show that it is highly improbable that the two-dosimeter approach will overestimate HE by more than 60%, even for static exposure situations. If a radiation worker moves freely during the exposure, the maximum overestimation will be less (say, <30%). The overestimation will be further decreased by the angular response properties of personal dosimeters, but will depend on the detailed geometries of the dosimeters.20 Table 18.2 shows E assessment ratios, r (=E(est)/E), as a function of various worker positions, source components, and photon energies when the single, and two-dosimeter approaches are used to estimate E. Note that the r value of E is always greater than the r value of HE for all worker positions, source components, and photon energies. This is because E is always less than HE, as was discussed earlier. For the divider plate and interior wall, the r value of E is larger than that of HE by 4%–26% considering all worker positions and photon energies. For the U-tubes, the difference is somewhat larger, but still less than 54%. These differences between the r values of E and HE do not change the general conclusions. In summary, the results show that the use of two dosimeters (chest and back) is adequate in a steam generator channel head. The two-dosimeter approach does not underestimate the effective dose or EDE by more than 4% considering all worker positions and contamination situations in the steam generator channel head.

18.3 PRDC—A Software Package for Personnel Radiation Dose Calculation The computational method described earlier requires the use of a sophisticated computer, worker computational phantoms, and Monte Carlo code, as well as a thorough understanding of radiation transport and interactions in a human body. Such a capability may not be available to practicing health physicists. For this reason, a user-friendly software package called PRDC5 was developed, which calculates effective dose and radiation doses to various organs/tissues and personnel dosimeters based on a series of interpolations. PRDC was developed by (1) building a comprehensive and structured dosimetric database of organ/tissue doses and dosimeter responses by using the developed Monte Carlo simulation procedure for a fine grid set of point and beam sources and (2) writing a userinterface computer program that uses the developed dosimetric database to calculate organ/tissue doses, effective dose, and dosimeter responses by interpolations.

1.00 1.03 0.96 1.07 0.98 1.00 1.02 1.10 1.04 1.11 1.20 1.06 1.10 1.11

1.11 1.01 1.20 1.15 1.10 1.12 1.06 1.03 1.10 0.98 1.06 1.04 1.05 0.98

IW

1.72 1.05 1.05 1.72 1.72 1.72 1.72 1.72 1.05 1.05 1.72 1.72 1.72 1.72

UT

b

c

1.05 1.02 1.08 1.09 1.05 1.05 1.04 1.06 1.08 1.05 1.15 1.05 1.07 1.03

Case 1

1.25 MeV (Co-60)

a

1.27 1.03 1.07 1.30 1.27 1.27 1.26 1.28 1.07 1.05 1.34 1.27 1.29 1.26

Case 2

d

1.05 1.08 1.03 1.16 1.01 1.42 1.07 1.17 1.13 1.17 1.32 1.10 1.18 1.18

DP

a

1.19 1.08 1.28 1.23 1.19 1.21 1.14 1.10 1.17 1.04 1.13 1.12 1.12 1.04

IW 2.06 1.15 1.12 2.06 2.06 2.06 2.06 2.06 1.12 1.15 2.06 2.06 2.06 2.06

UT

b

c

1.11 1.08 1.15 1.18 1.12 1.34 1.10 1.13 1.15 1.11 1.25 1.11 1.15 1.10

Case 1

0.66 MeV (Co-58) a

1.42 1.10 1.14 1.47 1.43 1.58 1.42 1.44 1.14 1.12 1.52 1.43 1.45 1.42

Case 2

d

1.18 1.15 1.16 1.33 1.12 1.19 1.20 1.23 1.24 1.15 1.42 1.13 1.26 1.22

DP

a

1.26 1.17 1.33 1.28 1.27 1.30 1.20 1.21 1.27 1.15 1.23 1.26 1.24 1.16

IW

a

2.22 1.14 1.13 2.22 2.22 2.22 2.22 2.22 1.13 1.14 2.22 2.22 2.22 2.22

1.21 1.16 1.25 1.32 1.21 1.23 1.20 1.22 1.26 1.15 1.35 1.22 1.25 1.19

1.55 1.15 1.21 1.62 1.54 1.56 1.54 1.55 1.21 1.15 1.64 1.55 1.57 1.53

UTb Case 1c Case 2d

0.08 MeV

Source: Kim, C.H. and Reece, W.D., Health Phys., 83, 243, 2002. With permission. a The divider plate (DP) and interior wall (IW) were all assumed to be uniformly contaminated over the surface. b The photon field from the U-tubes (UT) was assumed to be fully collimated by the tubesheet (∼60 cm thick carbon steel with thousands of small holes). Also note that the beam source geometry was realized by averaging beams within polar angles of 20° with a fixed azimuthal angle (0°). c This is the case when divider plate (DP) and interior wall (IW) are equally contaminated, which is a reasonable assumption for the most of the cases. d This is the case when the H contribution from the U-tubes (UT) is comparable to those from the divider plate (DP) and interior wall (IW). E

AS AL45 AR45 AL AR U AD BS BL45 BR45 BL BR BU BD

DP

a

EDE (HE) Assessment Ratios, r(=HE(est)/HE), as a Function of Various Worker Positions, Source Components, and Photon Energies when the Two-Dosimeter Approach Is Used to Estimate HE

TABLE 18.1

436 Handbook of Anatomical Models for Radiation Dosimetry

1.10 1.27 1.02 1.29 1.04 1.08 1.14 1.19 1.09 1.33 1.39 1.12 1.18 1.22

1.18 1.05 1.35 1.21 1.20 1.18 1.15 1.11 1.27 1.03 1.13 1.14 1.12 1.07

IWa

2.52 1.10 1.08 2.52 2.52 2.52 2.52 2.52 1.08 1.10 2.52 2.52 2.52 2.52

UTb 1.13 1.17 1.18 1.27 1.14 1.12 1.14 1.15 1.19 1.18 1.29 1.13 1.15 1.13

Case 1c 1.60 1.15 1.15 1.68 1.60 1.59 1.60 1.60 1.15 1.15 1.70 1.59 1.60 1.59

Case 2d 1.18 1.36 1.10 1.43 1.08 1.16 1.22 1.29 1.20 1.43 1.56 1.18 1.28 1.32

DPa 1.29 1.14 1.48 1.32 1.32 1.30 1.26 1.19 1.38 1.11 1.21 1.25 1.21 1.15

IWa 3.01 1.24 1.18 3.01 3.01 3.01 3.01 3.01 1.18 1.24 3.01 3.01 3.01 3.01

UTb 1.23 1.26 1.28 1.40 1.22 1.22 1.24 1.24 1.31 1.26 1.42 1.23 1.24 1.22

Case 1c

0.66 MeV (Co-58)

1.82 1.25 1.25 1.93 1.82 1.81 1.83 1.83 1.26 1.25 1.95 1.82 1.83 1.81

Case 2d 1.32 1.41 1.24 1.64 1.20 1.31 1.36 1.37 1.35 1.39 1.68 1.23 1.39 1.38

DPa 1.39 1.25 1.53 1.40 1.43 1.42 1.34 1.32 1.49 1.23 1.31 1.40 1.33 1.28

IWa 3.30 1.25 1.22 3.30 3.30 3.30 3.30 3.30 1.22 1.25 3.30 3.30 3.30 3.30

UTb

0.08 MeV

1.35 1.34 1.38 1.57 1.33 1.35 1.35 1.34 1.44 1.31 1.53 1.34 1.36 1.32

Case 1c

2.00 1.31 1.32 2.15 1.98 2.00 2.00 1.99 1.36 1.29 2.12 1.99 2.00 1.98

Case 2d

Source: Kim, C.H. and Reece, W.D., Health Phys., 83, 243, 2002. With permission. a The divider plate (DP) and interior wall (IW) were all assumed to be uniformly contaminated over the surface. b The photon field from the U-tubes (UT) was assumed to be fully collimated by the tubesheet (∼60 cm thick carbon steel with thousands of small holes). Also note that the beam source geometry was realized by averaging beams within polar angles of 20° with a fixed azimuthal angle (0°). c This is the case when divider plate (DP) and interior wall (IW) are equally contaminated, which is a reasonable assumption for the most of the cases. d This is the case when the H contribution from the U-tubes (UT) is comparable to those from the divider plate (DP) and interior wall (IW). E

AS AL45 AR45 AL AR AU AD BS BL45 BR45 BL BR BU BD

DPa

1.25 MeV (Co-60)

Effective Dose (E) Assessment Ratios, r(=E(est)/E), as a Function of Various Worker Positions, Source Components, and Photon Energies when the Two-Dosimeter Approach Is Used to Estimate E

TABLE 18.2

Applications to External Radiation Exposures in Nuclear Power Plants 437

Handbook of Anatomical Models for Radiation Dosimetry

438

Using Monte Carlo simulation with the MIRD-type mathematical computational phantom and the MCP code, researchers calculated the dosimetric data of organ/tissue doses and dosimeter responses. Two air-filled, 3 cm radius, 1 cm thick disk-shaped pseudodosimeters were also defined at the centers of the chest and back of the computational phantom (at the height of 41 cm from the bottom of the trunk section) to model the personnel dosimeters on the chest and back. To calculate the organ doses and dosimeter responses, this study used MCNP F6 tally, which directly calculates the radiation dose or energy deposition per unit mass (MeV/g). This study also used MCNP F4 tally to calculate the detailed energy spectrum of photon fluence at the dosimeter locations. A dosimetric database was constructed from a large number of Monte Carlo simulations. For point sources, the PRDC database was constructed for a fi ne grid of source locations (4315 source locations between 20 and 900 cm from the computational phantom) and seven photon energies between 0.02 and 2 MeV (Table 18.3). For beam sources, the database was constructed for 15° increments of the polar angle and 15° azimuthal angle increments (referenced from the incident beam direction). From each Monte Carlo simulation, calculated values were organ/tissue-averaged equivalent doses to 23 organs TABLE 18.3 Dosimetric Database Structure Point Source Energy (MeV) 0.02 0.08 0.16 0.30 0.50 1.00 2.00

Radial Distance (cm) 20 30 40 50 70 100 120 140 170 200 250 300 400 600 900

Beam Source

Polar (°)

Azimuthal (°)

0 15 30 45 60 75 90 105 120 135 150 165 180

0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255 270 285 300 315 330 345 360

Energy (MeV) 0.02 0.08 0.16 0.30 0.50 1.00 2.00

Polar (°)

Azimuthal (°)

0 15 30 45 60 75 90 105 120 135 150 165 180

0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255 270 285 300 315 330 345 360

Applications to External Radiation Exposures in Nuclear Power Plants

439

and tissues in the computational phantom and detailed dosimeter responses (including energy spectrum) of the personnel dosimeters on the chest and back of the computational phantom. The calculated values were then indexed and compiled to form a database. Each record (i.e., a set of dosimetric data) in the database consists of 23 organ/tissue doses and personnel dosimeter responses. The database uses a spherical coordinate system to specify both point source locations and beam source directions. The radial distance of a point source is the distance of the point source from the center of the computational phantom (on the bottom surface of the trunk section). The polar angle varies from 0° (from above to below) to 180° (from below to above). The azimuthal angle varies from 0° (from front to back), to 180° (from back to front), and back to 360° (from front to back) in the clockwise direction. Note that the database does not include any point sources inside the computational phantom and therefore the database can be used only for external exposures cases. To calculate organ doses and dosimeter responses for a given point source or beam source based on the developed dosimetric database, PRDC performs a series of interpolations between energies, radial distances (for point sources), polar angles, and azimuthal angles. For most cases, PRDC uses a simple linear interpolation between two neighboring data points. For example, if a user enters 0.662 MeV for photon energy, the program will calculate the organ doses by interpolating dosimetric values between 0.5 and 1.0 MeV, assuming a linear variation between these two energies. The only exception to a simple linear interpolation is an interpolation between radial distances when the radial distance is greater than 200 cm. If the radial distance is greater than 200 cm, an interpolation is performed after the values in the database are weighted by the square of the radial distance (r2) to compensate for the 1/r2 behavior of the dosimetric values. The interpolated values are then divided by r2 after an interpolation is completed. This approach significantly reduces the error in interpolation for the radial distances greater than 200 cm, for which the database has rather sparse spatial locations. PRDC is designed to be user-friendly. On the input frame (Figure 18.9), a user first selects the source geometry (i.e., point source or beam source). The location of a point source can be specified in either the spherical coordinate system (r, p, a) or the Cartesian coordinate system (x, y, z). Interpolations in the space are performed in the spherical coordinate system that matches the structure of the dosimetric database. If the source location is given in the Cartesian coordinate system, PRDC first calculates the corresponding location in the spherical coordinate system and then performs the necessary interpolations in the spherical coordinate system. To execute PRDC, the user provides the photon energy in MeV and specifies the location of a point source or the direction of a photon beam in the input fields. The angular response factor (ARF) of the chest and back dosimeters can also be given to adjust the calculated dosimeter responses for the angular response properties of the dosimeters. The default ARF value is 1.0. On clicking the “Calculate” button, PRDC calculates the organ/tissue doses, effective dose, and personnel dosimeter responses within seconds. Figure 18.10 shows the output frame for a point source case. PRDC first displays the input data (i.e., source specification) in the upper section and then shows the radiation dose to 23 organs and tissue, calculated by interpolations. Finally, based on the organ/tissue doses, the program calculates the effective dose (E) and EDE (HE) in several different units. The dosimeter responses, which are also calculated by interpolations, are tissue kerma in the pseudo air dosimeters on the chest and back. These values can be used to predict the response of a tissue-equivalent personnel dosimeter on the chest and back after considering energy dependence and charged particle

440

Handbook of Anatomical Models for Radiation Dosimetry

FIGURE 18.9 PRDC input frame. (From Kim, C.H. et al., Radiat. Prot. Dosim., 118, 243, 2006. With permission.)

FIGURE 18.10 PRDC output frame. (From Kim, C.H. et al., Radiat. Prot. Dosim., 118, 243, 2006. With permission.)

equilibrium (CPE). PRDC also calculates the response of the two-dosimeter approach, which uses two dosimeters (chest and back) for an accurate estimation of the effective dose (E) and EDE (HE). In addition, for the beam sources and given photon energies (0.02, 0.08, 0.16, 0.3, 0.5, 1, and 2 MeV), PRDC provides a detailed energy spectrum of photon fluence and tissue kerma in the chest and back dosimeters (Figure 18.11).

Applications to External Radiation Exposures in Nuclear Power Plants

441

FIGURE 18.11 PRDC output–energy spectra at dosimeters. (From Kim, C.H. et al., Radiat. Prot. Dosim., 118, 243, 2006. With permission.)

It takes less than 10–20 min to learn how to use the program. This is a significant saving in time and efforts compared to learning a Monte Carlo simulation code and developing a computational anthropomorphic computational phantom. The program also calculates desired dosimetric quantities quickly. Computing time is usually a few seconds or less on a regular personal computer. PRDC does have some limitations, mainly due to the structure of the dosimetric database. For a point source, PRDC interpolates dosimetric values between eight spatial points, and all of these points should be located outside of the anthropomorphic computational phantom. This limitation confines the application of PRDC only to cases where the source is separated from the human body at least by a few centimeters. The database covers photon energies down to 0.02 MeV. Interpolations between 0.02 and 0.08 MeV, however, should be avoided. For the effective dose, the error from interpolation alone can be as large as ∼10% when the value is linearly interpolated between these two energy points. Dosimeter responses cannot be interpolated between 0.02 and 0.08 MeV because the dosimeter response does not change linearly between these two energy points, and a linear interpolation could result in more than 50% error. Interpolation above 0.08 MeV is dependable due the linear dependence of organ doses and dosimeter responses on photon energy. The photon energy spectra in the chest and back dosimeters are given only for beam sources. Furthermore, since the energy spectrum cannot be interpolated between different energies, the energy spectra can be retrieved only for the photon energies of 0.02, 0.08, 0.16, 0.3, 0,5, 1, 1.5, and 2 MeV. Although PRDC only calculates beam and point sources, one can use the point-kernel approach to calculate dosimetric values for volume or surface sources. For this, a volume or surface source can be first divided into a number of small source segments depending on the size and location of the source. Each source segment is then assumed as a point source, for which PRDC calculates the dosimetric values. Finally, the calculated dosimetric values are summed over the volume or surface of the source. To facilitate this procedure,

442

Handbook of Anatomical Models for Radiation Dosimetry

PRDC is equipped with a batch calculation mode, in which the program calculates the dosimetric values for a batch of point or beam sources. Note that the point-kernel approach always shows less error than the point sources since most errors cancel each other during the summation process.

18.4 EPRI EDE Calculator—A Software Package for Assessing EDE and Effective Dose from Hot Particles on the Skin Hot particles, or discrete radioactive particles, are small activation products or fuel fragments ranging between 10 and 250 μm in diameter. Hot particles contain 60Co or various combinations of fission products, including isotopes, that emit beta particles or a combination of beta particles and gamma rays. Due to their mobility, hot particles can be found on many parts of a worker’s skin or protective clothing, which results in a highly localized skin dose from beta- and low-energy gamma-ray emitters or in a nonuniform whole-body dose from high-energy gamma-ray emitters. For the past two decades, the U.S. nuclear power industry and regulatory agencies have had a special interest in the dosimetric issues related to hot particles.21–23 The whole-body dose from an exposure to hot particles is relatively small in comparison to the dose to the skin, and software packages have been developed and used regularly for skin dose calculations.24,25 In 2003, the U.S. NRC issued the RIS 2003–2004, which encouraged licensees to use the EDE (calculated using the NRC-approved tissue weighting factors) in place of the DDE in all situations that do not involve direct monitoring of external exposures using personnel dosimetry.6 Importantly, this RIS permits assessment of the EDE for localized skin contamination. This RIS is a major milestone in the more-than-a-decade effort to adopt the risk-based EDE methodology in radiation protection regulations in the United States. However, the latest Varskin 3 software package only calculates the absorbed dose to the exposed skin, not to the whole body.26 The software package PRDC, which is introduced in the previous section, computes the EDE and ED, but not for radiation sources close to or in contact with the body. To provide the necessary dosimetric data, EPRI sponsored a study at RPI to systematically calculate and tabulate the EDE and effective dose (ED) for hot particles that emit gamma rays.27 The new data covered point sources with photon energies between 0.1 and 2.0 MeV for 74 locations covering the entire body surface, using the MCNP Monte Carlo transport code and the MIRD-type mathematical computational phantom. The comprehensive data set from that study was later used to develop a software package called the EDE Calculator, which was specifically designed for hot particles.7 The EDE and ED values are indexed and compiled to form a database using Microsoft Access. Each record in the database is composed of location information (i.e., x, y, z coordinate system value for one specific position on a computational phantom), energy, EDE, and ED. A user can calculate the EDE and ED for any of the predetermined photon energies. Alternatively, the user can pick an isotope from a library of hot particles that are known to emit photons, often at several different energies: Cr-51, Mn-54, Co-57, Co-58, Fe-59, Co-60, Nb-95, Zr-95, Ru-103, Ru-106/Rh-106, Cs-134, Cs-137, Cs-137/Ba-137m, Ba-140, La-140, Ce-141,

Applications to External Radiation Exposures in Nuclear Power Plants

443

Ce-144/Pr-144, Pr-144, and Tm-170. The decay data for these isotopes were from online data provided by the National Nuclear Data Center (NNDC) as adopted by Stabin and da Luz.28,29 X-rays and gamma rays, and their corresponding yields, are listed for each isotope or a decay chain involving two isotopes. For the latter, the combined chain is listed by considering the “activity ratio after equilibrium” and the branching yield going from the parent to the daughter. For example, Cs-137 decays with a 94.4% yield to Ba-137m, which in turn emits a 661.66 keV gamma ray with a yield of 90.11%. The combined yield for a Cs-137/ Ba-137m source is, then, 85% for the 661.66 keV gamma-ray. The EDE Calculator was designed with user-friendly graphic user interfaces. A user needs to type their user name and password to enter the system. There are several system administrative functions at the login screen of the software for the authorized users to maintain the basic database. When the EDECal button on the Main Function Panel is selected, another graphic user interface will appear, as shown in Figure 18.12. In the top-left panel of the screen, the energy of the hot particle can be entered. This input is used to calculate the EDE and ED for a single energy. If, instead, a known isotope is used for the source, the name of the isotope can be selected by clicking the “Add” button on the right. A window will pop up allowing choice an isotope. Multiple isotopes can be added into the library. The user must also key in the length of exposure, the radioactivity of the source, as well as the source location. The position of the source is specified by either using the tree-view menu showing 74 location names or by clicking each of the zones directly on the front and back views of the computational phantom. The dose can be calculated for only one location on the body at a time. When a location zone is selected, a square will appear around the corresponding dot on the

FIGURE 18.12 A graphical user interface of the EDE Calculator. On the left panel, a user can key in the energy of the photon source or select from a library of isotopes. The user also can key in the radioactivity and exposure time. The position of the source is specified by either using the pull-down menu showing 74 location names or by clicking each of the zones directly on the model. (From Xu, X.G. et al., Health Phys., 91, 373, 2006. With permission.)

444

Handbook of Anatomical Models for Radiation Dosimetry

computational phantom. As the mouse moves over the blue dots, a message will appear in the bottom left of the window indicating which zone is highlighted and the Cartesian coordinates for that zone. To calculate the EDE and ED, the EDE Calculator performs a series of linear interpolations or extrapolations using the database. When the input data have been entered, the user can double-click the desired zone on the computational phantom or in the tree view to calculate the EDE and ED. The results will be displayed on the right side panel of the screen, including the name of the zone, the Cartesian coordinates, the energy or isotope used, the activity, time, EDE, and ED. A summary of the input data and the calculated doses can be automatically generated as a Microsoft Word document. This is done by clicking on the “Print Report” button on the screen to bring up a preview-reporting page as shown in Figure 18.13. The document can be printed as a hardcopy or can be saved on the computer. To exit the reporting screen, a user clicks the “Exit” button, and the computer will return to the EDE Calculator interface. A user-friendly graphic user interface is available to maintain and update the EDE Calculator database data. A user clicks the ISO/SEC button to launch the “Isotope/Section Management” screen, as shown in Figure 18.14. A user can also add or delete isotope data through this screen. The name and title of the user and information about their company can be edited from this screen. Users can also update computational phantom locations and names to better match the frequent hot particle exposures. To date, the EDE Calculator is the only software package that automatically computes the EDE and ED for various hot particle exposures. The software tool was developed to be a utility standard for calculating whole-body exposures from hot particles. It can certainly become an important tool in routine radiation control and risk evaluation in nuclear power plants.

FIGURE 18.13 A report of the calculated results is displayed in Microsoft Word automatically. (From Xu, X.G. et al., Health Phys., 91, 373, 2006. With permission.).

Applications to External Radiation Exposures in Nuclear Power Plants

445

FIGURE 18.14 Software interfaces allow for data to be updated easily. (From Xu, X.G. et al., Health Phys., 91, 373, 2006. With permission.)

18.5 Summary and Future Work Using the stylized worker computational phantoms, two kinds of methods to determine the effective dose, or EDE, were developed from original research performed at Texas A&M University: (1) a two-dosimeter method, which determines the effective dose, or EDE, based on the readings of the personnel dosimeters on the chest and back, and (2) user-friendly software packages that determine the effective dose and other dosimetry quantities based on a database of precalculated dose values, as direct applications of computational human phantoms in dosimetry for external radiation exposures in nuclear power plants. The two-dosimeter method helped to bring about a paradigm change in personnel dosimetry in the nuclear industry in the United States. In December 1995, the NCRP recommended using the two-dosimeter method for exposure scenarios “where the irradiation geometry or photon energy is unknown or difficult to characterize.”14 In 2003, the U.S. NRC issued the RIS 2003–2004 in order to encourage licensees to use the EDE in place of the DDE.6 This RIS is considered to be a major milestone in the more-than-a-decade effort to adopt the risk-based EDE methodology in radiation protection regulations in the United States. Internationally, the utilities in Korea have recently also decided to use the twodosimeter method to assess the effective dose of radiation for workers in highly gradient and nonuniform radiation fields. The regulatory agency in Korea is expected to approve the new method. Some issues remain to be addressed in the two-dosimeter method. The U.S. NRC approved the use of two dosimeters for estimating the EDE, but limited the algorithm to the algorithm developed by Xu,12 even though there were several advanced algorithms developed after the work. On the other hand, the nuclear industry in Korea has decided not to use the algorithm developed by Xu, but the 55/50 algorithm developed by the

446

Handbook of Anatomical Models for Radiation Dosimetry

NCRP. A study of the two-dosimeter method concluded that the optimal algorithm developed by Kim et al.15 gave the most reasonable estimation of the effective dose for the exposure situations in a steam generator channel head.30 In short, there are several algorithms available in the literature, and there should be some additional study on the evaluation of these two-dosimeter algorithms. Also note that the available algorithms were developed based solely on computer simulations, except for the primitive one developed by Lakshmanan et al.8; the two-dosimeter algorithms have never been tested under real exposure situations. Therefore, some experimental studies with real exposure situations in nuclear power plants would be necessary to confirm the two-dosimeter method in the real exposure situations. In this chapter, two software packages were reviewed—PRDC and EPRI EDE Calculator. PRDC calculates the effective dose and radiation doses to the organs/tissues and personnel dosimeters for external photon exposures from beam sources and point sources. The EPRI EDE Calculator calculates the effective dose and EDE from hot particles on the skin. These software packages are easy to use and practicing health physicists can quickly learn to use them. These programs, however, calculate the effective dose according to ICRP 60,2 which has been revised recently.31 The tissue weighting factors have changed significantly and new organs added. The new recommendation also assumes voxel computational phantoms, not MIRD-type mathematical computational phantoms, to represent the human body. Therefore, the database of the dose values in PRDC and EPRI EDE Calculator should be recalculated before the new recommendation can be implemented in the regulations. Due to the use of voxel computational phantoms, the calculation time of the database will be very significant; the calculation of ∼40,000 cases in PRDC will take about 4–5 years on a regular personal computer. Finally, it will be beneficial to add a direct Monte Carlo calculation module to the PRDC and EPRI EDE Calculator so that the user can have two options for dose calculation—the current fast interpolation-based calculation mode and a slower, but more accurate Monte Carlo simulation mode using a human computational phantom.

References 1. ICRP. Recommendations of the International Commission on Radiological Protection, Publication 26, Pergamon, Oxford, 1977. 2. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Pergamon Press, Oxford, 1991. 3. ICRU. Measurement of dose equivalents from external photon and electron radiations, International Commission on Radiation Units and Measurements, Bethesda, MD, Report No. 47, 1992. 4. ICRP. General Principles for the Radiation Protection of Workers, ICRP Publication 75, Pergamon Press, Oxford, 1997. 5. Kim, C.H., Cho, S.H., and Xu, X.G. PRDC—A software package for personnel radiation dose calculation, Radiation Protection Dosimetry, 118, 243, 2006. 6. Nuclear Regulatory Commission. Use of the effective dose equivalent in place of deep dose equivalent in dose assessments, NRC Regulatory Issue Summary 2003–04, Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory Commission, Washington, DC, 2003. 7. Xu, X.G., Su, H., and Bushart, S. The EPRI EDE calculator—A software package for assessing effective dose equivalent from hot particles on the skin, Health Physics, 91, 373, 2006.

Applications to External Radiation Exposures in Nuclear Power Plants

447

8. Lakshmanan, A.R., Kher, R.K., and Supe, S.J. Estimation of effective dose equivalent using individual dosimeters, Radiation Protection Dosimetry, 35, 247, 1991. 9. Cristy, M. and Eckerman, K.F. Specific Absorbed Fractions of Energy at Various Ages from Internal Photon Sources I: Methods, ORNL/TM-8381/V1, Oak Ridge National Laboratory, Oak Ridge, 1987. 10. Reece, W.D., Poston, J.W., and Xu, X.G. Assessment of Effective Dose Equivalent for External Photon Radiation: Calculational Results for Beam and Point Source Geometries, EPRI-3099-10, Electric Power Research Institute, Palo Alto, 1993. 11. Reece, W.D., Poston, J.W., and Xu, X.G. Determining the effective dose-equivalent for external photon radiation—Calculational results for beam and point-source geometries, Radiation Protection Dosimetry, 55, 5, 1994. 12. Xu, X.G. The assessment of effective dose equivalent using personnel dosimeters, Dissertation, Texas A&M University, College Station, 1994. 13. Claycamp, H.G. Optimisation of monitor weighting factors for the estimation of effective dose equivalent from external photon exposures, Radiation Protection Dosimetry, 63, 105, 1996. 14. National Council on Radiation Protection and Measurements. Use of personal monitors to estimate effective dose equivalent and effective dose to workers for external exposure to low-LET radiation, NCRP Report No. 122, NCRP, Bethesda, 1995. 15. Kim, C.H., Reece, W.D., and Poston, J.W. Development of a two-dosimeter algorithm for better estimation of effective dose equivalent and effective dose, Radiation Protection Dosimetry, 81, 101, 1999. 16. Kim, C.H. Use of two dosimeters for better estimation of effective dose, PhD dissertation, Texas A&M University, College Station, 1998. 17. Roberson, P.L. et al. Spectral and dosimetric measurements of photon fields at commercial nuclear sites, Report PNL-4915, NUREG/CR-3569, Pacific National Laboratory, Richland, 1984. 18. ICRP. Conversion Coefficients for Use in Radiation Protection against External Radiation, ICRP Publication 74, Pergamon Press, Oxford, 1996. 19. Polley, M.V., Garbett, K., and Pick, M.E. A Survey of the Effect of Primary Coolant pH on Westinghouse PWR Plant Radiation Fields, Report TR-104180, Electric Power Research Institute, Palo Alto, 1994. 20. Kim, C.H. and Reece, W.D. Effect of angular response properties of personal dosimeters on the estimation of effective dose using two dosimeters, Radiation Protection Dosimetry, 93, 215, 2001. 21. National Council on Radiation Protection and Measurements. Biological effects and exposure limits for “Hot Particles” NCRP Report No. 130, NCRP, Bethesda, 1999. 22. Nuclear Regulatory Commission. Highly radioactive particle control problems during spent fuel pool cleanout, Information Notice 2002–03, Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory Commission, Washington, DC, 2002. 23. United States Nuclear Regulatory Commission. Sources of unexpected occupational radiation exposures at spent fuel storage pools, Information Notice 90-33, Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory Commission, Washington, DC, 1990. 24. Durham, J.S. VARSKIN MOD2 and SADDE MOD2: Computer codes for assessing skin dose from skin contamination, NUREG/CR-5873, U.S. Nuclear Regulatory Commission, Washington, DC, 1992. 25. Traub, R.J. et al. Dose calculation for contamination of the skin using the computer code VARSKIN, NUREG/CR-4418, U.S. Nuclear Regulatory Commission, Washington, DC, 1987. 26. Durham, J.S. VARSKIN 3: A computer code for assessing skin dose from skin contamination, NUREG, Nuclear Regulatory Commission, Washington, DC, 2006. 27. Xu, X.G. The effective dose equivalent and effective dose for hot particles on the skin, Health Physics, 89, 53, 2005. 28. National Nuclear Data Center (NNDC). http://www.nndc.bnl.gov/, last accessed January, 2008. 29. Stabin, M.G. and da Luz, L.C.Q.P. Decay data for internal and external dose assessment, Health Physics, 83, 471, 2002. 30. Lee, G.S. Deviation of TL dosimeter responses in photon radiation fields from the effective doses, Masters thesis, 2000. 31. ICRP. The 2007 recommendations of the international commission on radiological protection. ICRP publication 103, Annals of the ICRP, 37, 1, 2007.

19 Applications to Bioassay for Internal Radiation Contamination Gary H. Kramer

CONTENTS 19.1 Introduction ............................................................................................................... 449 19.2 Indirect Methods .......................................................................................................454 19.2.1 Alpha Emitters.............................................................................................454 19.2.2 Beta Emitters ................................................................................................ 455 19.2.3 X-Ray and Gamma Emitters ...................................................................... 456 19.2.4 Dose Estimate from Indirect Methods..................................................... 457 19.3 Direct Methods .......................................................................................................... 457 19.3.1 Whole-Body Counting................................................................................ 458 19.3.2 Lung Counters ............................................................................................. 459 19.3.3 Organ Counting .......................................................................................... 460 19.3.4 Wound Counting......................................................................................... 461 19.3.5 Dose Estimates from Direct Methods ...................................................... 461 19.4 Calibration.................................................................................................................. 461 19.4.1 Alpha Counting ........................................................................................... 462 19.4.2 Beta Counting .............................................................................................. 462 19.4.3 Gamma Counting .......................................................................................463 19.4.4 Whole-Body Counting................................................................................ 463 19.4.5 Lung Counting ............................................................................................464 19.4.6 Organ Counting .......................................................................................... 466 19.4.7 Wound Counting......................................................................................... 467 19.4.8 Quality Assurance ...................................................................................... 467 19.5 Conclusion.................................................................................................................. 468 References ............................................................................................................................. 468

19.1 Introduction This chapter is an overview of the methods used to perform bioassay for the purpose of determining internal dose. It is not intended to be a compendium of methods, but sufficient information will be provided for the reader to follow-up if any particular methods are found to be interesting and worthy of further study.

449

450

Handbook of Anatomical Models for Radiation Dosimetry

Internal dose can only be estimated if the number of transformations in an organ, or collection of organs, is known. This requires knowledge of the metabolism of the nuclide, where it translocates within the body and at what speed, and how long it is expected to remain in any given location. Once the number of transformations is estimated, the amount of energy deposited in the organ must be determined. The dose resulting from all these interactions can be calculated using conversion factors that are currently based on mathematical models of the human body. For example, Snyder et al.1 described organ shapes by geometrical shapes (cylinders, cones, truncated cones, spheres, elliptical cylinders, etc.) of fixed size. However, there is a move to recalculate these factors using voxel models, which can more realistically present the human body. This work is expected to be completed in 2008 by one of the International Commission on Radiological Protection’s (ICRP) Working Groups. That being said, the voxel models that the group use to provide the new dose conversion factors will still represent a fixed body size corresponding to the reference male and female.2 The movement of nuclides through the body is described using compartmental models according to the experimental retention and excretion in the body; unfortunately, retention and excretion data are not always well known for human, and are sometimes based on animal experiments or on just a few human data points. These compartmental models are not physiological models, but simply mathematical tools that describe the observed results. It is worth pointing out that retention and excretion can vary widely from person to person. For example, cesium has, according to the ICRP, a long-term retention component of 110 days.3 A study of cesium retention4 showed that the ICRP two-component retention model was essentially confirmed, with 10% of the radiocesium body burden being excreted with a half time of about 2 days, and the remainder with a variable half-time (mean = 86 days; range = 49–186 days). The observed variability for the long-term compartment ranges from a factor of 2.4 times faster to a factor of 1.7 times slower compared with the default value of 110 days. Clearly, individuals with the faster clearance will have an internal dose lower than that calculated from the default value (110 days), and individuals with the longer retention will have an internal higher than the default value. Internal dose is often derived from the calculated intake, a quantity relatively easily obtained from the compartmental models mentioned above. Generally a computer must do these calculations as there are many equations to solve. An example of the complexity of a compartmental model is shown in Figure 19.1, which is the particle transport model in the lung.5 Each number on the left represents a rate of transport from one compartment to another. This compartment model must be integrated into all the other models that describe the retention and excretion of all the other organs in the body. The normal principles of radiation protection that can be used to reduce dose when faced with an exposure from a source that is external to the body, time distance, and shielding, do not apply well for internal exposures. Once a radionuclide has been incorporated into the body, either by inhalation, ingestion, skin absorption, or through a wound, then these principles break down. Clearly one cannot increase the distance from the source as it is carried around inside the body. Similarly, it cannot be shielded, as no shielding can be inserted into the body to surround the material. Time, however, can be affected. The exposure time can be reduced by increasing the elimination of the radioactive material from the body. Normally there are two ways in which a radioactive material will leave the body. The first, which cannot be changed, is the

Applications to Bioassay for Internal Radiation Contamination

451

FIGURE 19.1 Screen shot from integrated molecules for bioassay analysis (IMBA) professional showing the lung absorption model.

physical half-life, which leaves through the normal radioactive decay process. The second, which can be influenced in certain circumstances, is the biological half-life as the body is always replacing materials. For example, the biological half-life for hydrogen (which is the same for the radioactive isotope tritium) is about 10 days. This can be decreased by increasing the amount of fluids ingested. Workers who are trying to decrease the biological half-life of tritium frequently like to do this by ingesting beer to increase the water turnover in their bodies. Another method of increasing the excretion is chelation therapy. An example is diethylene triamine pentaacetic acid (DTPA)6 which can be applied to nuclides such as americium or plutonium7 or Prussian Blue for cesium.8 Unfortunately not every nuclide is susceptible to this treatment method. The two half-lives, physical and biological, are related and can be combined into one quantity known as the effective half-life. Their relationship is demonstrated in Figure 19.2 for a fictitious nuclide with a physical half-life of 8.00 days, a biological half-life of 15.00 days, and an effective half-life of 5.22 days. Once a radionuclide has entered the body, its chemical form and its physical form will determine its excretion. For the former, the chemical form will behave similarly to the nonradioactive type. In the example above tritium behaves as does hydrogen so that inorganic tritium will be like inorganic hydrogen, with a biological half-life of about 10 days; whereas organic tritium will be like organic hydrogen, with a biological half-life of about 40 days. For the latter, the solubility will be the main influence for its physical form. If a material is insoluble, and ingested, it will pass through the body quickly, and the dose will be less than if the material were soluble. A soluble material would have crossed the gut wall, would have entered the systemic system, and would then be subject to the normal

Handbook of Anatomical Models for Radiation Dosimetry

452

1.00

Physical

Biological

Effective

0.90

Fraction retained

0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 5

0

10

15

20 Time (days)

25

30

35

40

FIGURE 19.2 Relationship of physical, biological, and effective half-lives.

retention characteristics of that element. On the other hand, if an insoluble material is inhaled it will be slowly solublized. During that period, large doses can be delivered to the surrounding tissue. The ICRP has classified three solubility classes that are known as absorption parameters: Types F, M, and S. F is fast-rate clearance from the lung, M is medium-rate clearance from the lung, and S is slow-rate clearance from the lung.5 Clearly, a Type S material will be retained longer, and the preferred method of analysis is whole-body or organ counting providing there are emissions that can be measured outside the body. An example of how solubility affects the retention and excretion is shown in Figure 19.3. This figure shows

1.00E+00 Whole body

Urine (Bq/d)

Feces (Bq/d)

1.00E–01 1.00E–02 1.00E–03 1.00E–04 1.00E–05 1.00E–06 1.00E–07 0

50

100

150

FIGURE 19.3 Whole body retention, urinary and fecal excretion rates.

200

250

300

350

Applications to Bioassay for Internal Radiation Contamination

453

the whole-body retention, urinary, and fecal excretion rates for a Type S 241Am intake (this nuclide is usually a Type M material, but it has been changed to Type S for illustrative purposes). It is clear from Figure 19.3 that fecal measurements could be useful in the short-term, but obtaining a reliable intake estimate is difficult due to the steep slope of the excretion curve. The urinary excretion shows that any method relying on the this technique must have an extremely good detection limit as the amount in urine is between four and five orders of magnitude lower than the whole-body retention curve. That being said, a Type M inhalation of 241Am (the default case) gives a series of curves that only differ significantly for the urinary and fecal excretion rates. The urinary excretion rate is increased by about an order of magnitude, making urine analysis slightly more attractive; and the latter has an increased rate after 12 days, making it less attractive. Internal dose can be calculated if the time of intake is known (for acute), or the start of the intake is known (for chronic), or if the nature of the material (physical and chemical form), and the magnitude of the intake are known. In an accident situation, the time of intake is often well defined, but for an accidental intake the intake time may have to be inferred from work history. The chemical form of the material is often well known from the work history, but sometimes it may have to be inferred from its behavior in the body (i.e., decommissioning or working in old contaminated buildings). The magnitude of the intake almost always has to be inferred from bioassay measurement made after the event has occurred. The process flow for this is shown in Figure 19.4. There are two types of bioassay measurement that can be made and they are known as indirect and direct methods. The indirect methods measure radioactive content of excreta (e.g., urine, feces, breath) and use mathematical models to work back to the intake.3 Direct

Collect sample

Analyze for content

Correct for recovery, etc. Apply models

Adequate fit to data?

No

Refine parameters

Yes Compute intake

Estimate dose FIGURE 19.4 Flowchart describing the steps in a dose estimate from a bioassay test.

454

Handbook of Anatomical Models for Radiation Dosimetry

methods measure body content directly, which in some case can lead to a dose estimate directly, but mathematical models must still be used to derive the estimated intake. Once the intake has been obtained it is a simple matter to obtain a dose estimate, either by using computer codes such as IMBA9 or using one of the many ICRP publications.3 The two types of bioassay methods will be treated differently, but it should already be apparent that the solubility of the material will heavily influence the type of analysis that should be performed, as will the radioactive emission type (i.e., alpha, beta, x-ray, and gamma). In some cases it is advisable to perform both an indirect and direct analysis to be sure that if any amount has been incorporated into a person’s body, it is small enough to pose no adverse health effect. In many cases a negative (i.e., nondetectable) result is the best outcome, as it shows that any internal dose from material that remains undetectable is trivial (providing the minimum detectable activity is sufficiently low enough).

19.2 Indirect Methods 19.2.1 Alpha Emitters An alpha particle, a helium nucleus (two protons and two neutrons), is highly energetic. Yet, due to its heavy mass, alpha particle can only travel a short distance in tissue. The range of a 4 MeV alpha particle in air is about 2.5 cm and about 14 μm in tissue, whereas an 8 MeV alpha particle would have a range of 7 cm in air and 42 μm in tissue. As a result it can be ignored as an external hazard, as it cannot penetrate the dead layer of skin on one’s epidermis; but as an internal hazard, it is extremely dangerous. All of its energy is deposited in a very short track. If an alpha emitter gets incorporated in a cell nucleus, it will be responsible for double strand breaks of a DNA helix, something that is hard, if not impossible, to repair correctly and that leads to a variety of outcomes including: mutation, cell death, or cancer. This has been recognized by the ICRP and the radiation-weighting factor for alpha particles has been given the value of 20. By comparison, beta particle and gamma and x-rays have a radiation-weighting factor of 1. Therefore, in simplistic terms an alpha emitter is 20 times more dangerous than a nuclide that decays by another transformation route. Alpha particles are usually only emitted during the decay of nuclei that have an atomic mass greater than 209. Many of these materials that are encountered in the workplace have long physical half-lives (e.g., 235U: 7.038 × 108 years). It would be a simple matter to measure these nuclides using direct bioassay methods if sufficient gamma or x-ray radiation was emitted during the decay. Unfortunately, many of these nuclides emit no such radiation, or the amount that they emit is too small to be useful (for example 210Po, which is discussed in more detail in the following). By the time a direct method could be used, the amount of material in the body would be very large, and the associated internal dose would be unacceptably high. As a result, the body content is best estimated via an indirect method such as urinalysis. Because this sample is easy to collect, this is the method of choice. Fecal analysis can be performed sometimes, but it is generally only useful in the first few days following an intake and the samples are difficult to collect. A routine program where personnel are potentially exposed to a variety of nuclides will often use a screening program10 to check for gross contamination of an alpha emitter without necessarily knowing which emitter it is. Once contamination is found, a more specific

Applications to Bioassay for Internal Radiation Contamination

455

test may be applied to a resample, or the test sample may itself be subjected to a sequential analysis to identify the nuclide. Until recently, alpha emitters in urine were difficult and time-consuming to analyze. The urine sample had to be wet-ashed and all the organics destroyed. The remaining salts had to be removed by methods such as ion exchange, solvent extraction, or coprecipitation, and the final analyte prepared for analysis using either a coprecipitation methods or electroplating. The final analysis was carried out using an alpha spectrometer (or if the analyst was confident of the identity of the nuclide, an alpha counter). The alpha spectrometer offers the advantage of distinguishing multiple isotopes of the same nuclide, or different nuclides, in a single measurement. Many of the classic procedures can be found online; for example, the procedures used by the Environmental Measurements Laboratory known as HASL-300.11 Advances in science over the last decade have simplified the analysis process somewhat. Separations, if required, can be done using high-pressure liquid chromatography and analysis of many long-lived alpha emitters can be done using mass spectrometry,12–14 a technique that is often much more sensitive than the more conventional methods. The disadvantage is, of course, the cost of the machine, but this has dropped dramatically over the last decade. Where chemical separations are needed, the method of choice is often short chromatography columns manufactured by Eichrom, a company that was founded in 1990 to commercialize three novel extraction chromatographic resins developed at Argonne National Laboratory. These resins (Sr Resin, TRU Resin, and UTEVA Resin) advanced the analysis of actinides and radioactive strontium. Since that time, new extraction chromatographic materials, analytical grade ion exchange resins, and high-performance filters for alpha spectroscopy source preparation provide a comprehensive product line for the radiochemist.15,16 Many standards contain methods that have been built around these resins. One alpha emitter that deserves special mention is 210Po, the material used to kill the Russian dissident Alexander Litvinenko in November 2006.17 The nuclide has a short physical half-life (138 days) so it cannot be measured by mass spectrometry, and the nuclide also emits an energetic alpha particle (5.4 MeV) and a gamma ray; however, the probability of the gamma ray being emitted is very low (~1 in 100,000) so it cannot measured. The analysis of 210Po method is relatively simple. Once the urine sample has been acidified, polonium can simply be plated on a silver or nickel disk and counted directly in an alpha counter or an alpha spectrometer.11 Techniques commonly encountered when analyzing a sample for alpha content are wet ashing, separations (coprecipitation, solvent extraction, ion exchange, high-pressure liquid chromatography, electroplating), and measurement (mass spectrometry–thermal ionization, inductively coupled plasma, accelerator, alpha counting, alpha spectrometry, photon–electron rejecting alpha liquid scintillation, liquid scintillation, fluorometry, phosporimetry). The measurement of a signal is relative, as with all radiation analysis methods. That is to say, the amount of signal measured must be compared to a signal measured from a standard that has a known value of radioactivity. This will be further discussed in Section 19.4. 19.2.2 Beta Emitters A beta particle is an electron, so it is a much smaller and less energetic particle than the alpha particle. The range of beta particles is greater than alpha particles and a few millimeters of aluminum, or tissue are requested to stop beta particles with energy 1–2 MeV.

456

Handbook of Anatomical Models for Radiation Dosimetry

Because they are relatively light, beta particles do not travel in straight lines, but rather follow a random path through material. This makes more difficult to define the range of a beta particle. As with alpha emitters, not all beta emissions are accompanied with x-ray or gamma rays. Examples of these nuclides are 3H, 14C, 63Ni, 90Sr, and 147Pm. Similar to alpha emitters, the choice of analyte is urine. Often the sample can simply be counted in a liquid scintillation counter and the activity in the urine sample obtained quickly. In other cases separations are necessary,18 and these can be done using similar techniques to the ones already described above. The separated analyte can be counted again in a liquid scintillation counter or plated either as a coprecipitate or some other method (e.g., dry drop) on a planchet that is then measured in a planchet counter. Nuclides like 90Sr can cause difficulties if they undergo separation. This nuclide decays to 90Y, which has a physical half-life of ~60 h. Therefore, immediately after the 90Y is separated from 90Sr in whatever process is being used, it starts to grow back. This must be accounted for during the measurement, otherwise an incorrect activity will be obtained. This can be accomplished in one of two ways. The first, and simplest, is simply to wait for equilibrium to be reached and then measure the sample; however, this takes ~ 3–4 weeks and there is likely to be an unacceptable delay in estimating an internal does. The second, preferred method, is to note the time of the separation and calculate the percentage of 90Y in growth, and correct the measured signal accordingly. Recent advance in mass spectrometry techniques have lowered the half-life that can be measured with this technique. In some cases it is possible to measure nuclides with half-lives as low as 30 years. For example, 90Sr can now be measured using ICPMS.19,20 Techniques commonly encountered when analyzing a sample for beta content are wet ashing, separations (coprecipitation, solvent extraction, ion exchange, high-pressure liquid chromatography), and measurement (mass spectrometry, gas proportional counting, liquid scintillation counting). 19.2.3 X-Ray and Gamma Emitters X-ray and gamma radiation always accompany other types of decay, such as alpha, beta, positron (beta+), and electromagnetic radiation. As a result, x-ray and gamma ray are highly penetrating; therefore, heavily contaminated persons or their excreta (urine, feces, sweat, etc.) can pose a hazard to those nearby because they act as an external source. Gamma rays are typically found in the energy range of 0.1–3.0 MeV and they lose energy in three major ways: photoelectric, Compton scattering, and pair production. Further information on these processes can be found elsewhere.21 Gamma or x-rays are the simplest type of emission to measure, as the sample can simply be placed on a detector and the signal collected. The only thing that will affect the signal is the sample density and sample size, but that issue will generally be compensated for by the calibration. The two types of detectors commonly in use are the scintillation detector and the semiconductor detector. The former is usually sodium iodide, doped with thallium, and the latter is generally a high-purity germanium detector, which must be cooled by liquid nitrogen. There are now detectors than are electrically cooled and are slowly making their way into this application, and some are now in use in different laboratories around the world. Other types of detectors are available, such as cesium iodide, silicon (for x-rays), cadmium zinc telluride, bismuth germinate, etc. Relative newcomers are the lanthanum chloride and bromide detectors,22 these offer higher resolution (~3%) at room temperature

Applications to Bioassay for Internal Radiation Contamination

457

than other scintillation detectors (~7%). Their biggest disadvantage is the naturally occurring radioactive lanthanum that is in the detector crystal (138La half-life 1.02 × 1011 year, beta decay with two gammas: 788.7 keV, 33.6%, and 1435.8 keV, 66.4%) making them unsuitable for low-activity measurements. 19.2.4

Dose Estimate from Indirect Methods

Once the activity content of the sample has been determined, the analyzing laboratory will have a result in terms of activity per sample. This must now be converted to activity per day’s excretion. This can be achieved in several ways: 1. The ICRP has given guidance on how much the Reference Man (or Woman) is expected to excrete per day. The volume of the collected sample can be ratioed to this figure, and the activity content be adjusted accordingly. 2. The contaminated person can be asked to collect all voidings in a 24 h period and that that sample be analyzed for radioactive content. That result would be in the correct units. 3. The urine volume can be corrected to a 24 h sample by measuring the creatinine content and comparing it with the expected value. This is a similar technique to item 1 above. None of these methods are perfect and all have advantages and disadvantages. Assuming the intake time is known, or can be estimated, then the concentration of the nuclide in the urine can be used to estimate the intake either by using a computer code such as IMBA or by using the retention charts in the ICRP documents. It is a simple matter to determine the internal dose using the same tools once the intake has been determined. Further refi nements to the internal dose can be implemented using a stepby-step approach that has been developed elsewhere.23 Naturally, it is advisable to have multiple measurements to confirm that indeed a contamination has taken place and that it was not simply a contamination of the sample container. A series of measurements will show how the material’s excretion is behaving with time and will allow the dosimetrist to refine the dose estimate using person-specific retention parameters, although this process usually only takes place if the person is likely to exceed the national dose limit for a worker, or member of the public.

19.3 Direct Methods A direct method measures the radionuclide while it is in the body. Therefore, it avoids the correction to a 24 h volume (or weight in the case of fecal analysis) required for an indirect measurement. This correction can be the source of significant uncertainty in the final dose estimate. A direct measurement requires that the emission must be detectable outside of the body, thereby eliminating alphas emitters and most beta emitters (the exception to the latter will be discussed in the following). There are four type of direct measurement commonly in use: whole-body counting, lung counting, organ (e.g., thyroid) counting, and wound counting. Each will be discussed in turn.

458

Handbook of Anatomical Models for Radiation Dosimetry

19.3.1 Whole-Body Counting A radioactive material that is approximately homogeneously distributed throughout the body and emits a gamma ray of sufficient intensity and energy is best measured by this technique. Typical examples of radionuclides measured in this way include 57Co, 60 Co, 95Zr/95Nb, 137Cs. Whole-body counters typically measure gamma ray emitters that lie in the range 200 keV to 3 MeV, although 57Co (122 keV, 87%, and 136 keV, 12.3%) can often be measured in a whole-body counter. Radionuclides that only emit low-energy gamma rays or x-rays are not amenable to whole-body counting (e.g., 241Am 59.5 keV, 36%). In principle, the whole-body counting technique is simple: a contaminated subject is placed in front of a detector, or detector array(s), and measured for a period of time. The activity in the body is determined directly from the resulting spectrum. There are many types of whole-body counters that have different counting geometries and detector types. They fall into two basic categories: 1. Static systems and scanning systems 2. Scintillation detector or semiconductor detector In static systems, the subject stands, sits, or lies in front of a detector, or detector array(s). In scanning systems, the subject stands or lies under, in between, or over a detector or detector array(s). The subject can either be passed over the detectors on a movable bed or the detectors can scan across the subject who lies still on the bed. The detector types are usually sodium iodide or hyperpure germanium, which must be cooled by liquid nitrogen. There are now detectors than are electrically cooled and they are slowly making their way into this application. At the time of writing only a few whole-body counters worldwide are using electrically cooled detectors. The use of high-resolution detectors makes identification easier and also improves the detection limit due to the high resolution of the peak (i.e., less background channels are being measured). Shielding is an important component of these systems and the whole-body counter can either be in a heavily shielded counting room that has thick steel walls lined with lead to reduce the background in the lower energy region or it can be a shadow shield design. The shadow shield is a part of the counter around the detector that is heavily shielded, usually by steel and lead, and casts a low background “shadow” in that part of the counter. The subject being measured is usually transposed through this low background area on a moveable bed. A typical example of this type of counter is a shadow shield whole-body counter. These types of counters, or the static counters that have large arrays such as the Canberra FastScan whole-body counter, have the added advantage that the heterogeneity of the source is less important compared with the types of whole-body counter that are static and have a single detector. All types of whole-body counters are sensitive to the size (height and weight) of the person being measured to varying degrees. This will be discussed further in Section 19.4. Once the body burden has been obtained, the intake can be estimated by using a computer code or the ICRP documents for the intake retention fractions as a function of time. Naturally it is advisable to have more than one measurement so that the change in body burden can be tracked with time. The material’s behavior will confirm assumptions about the material’s physical characteristics or indicate that these assumptions need to be modified.

Applications to Bioassay for Internal Radiation Contamination

459

As mentioned above, beta emitters cannot be measured by direct methods unless they are accompanied by a gamma ray. However, there are exceptions to this. When a beta particle is emitted from a decaying nucleus, it may be slowed either by the medium it is entering or by the nucleus. As result of this slowing, x-rays are emitted, and this is known as bremsstrahlung (from the German for “braking radiation”) radiation. Depending on the type of slowing, it is further subdivided into inner and outer bremsstrahlung. “Inner bremsstrahlung” is the term applied to the less-frequent case of radiation emission during beta decay, resulting in the emission of a photon of energy less than or equal to the maximum energy available in the nuclear transition. Inner bremsstrahlung is caused by the abrupt change in the electric field in the region of the nucleus of the atom undergoing decay, in a manner similar to that which causes outer bremsstrahlung. Beta-particle-emitting substances sometimes exhibit a weak radiation with continuous spectrum that is due to both outer and inner bremsstrahlung, or to one of them alone. “Outer bremsstrahlung” is the term applied in cases where the energy loss by radiation greatly exceeds that by ionization as a stopping mechanism in matter. This is seen clearly for electrons with energies above 50 keV. It is, therefore, possible to measure some beta emitters if the detection system can measure the low-energy x-rays resulting from the bremsstrahlung process. This has been successfully carried out for a number of different “pure” beta emitters. For example, a phoswich detector system was used to aid in the evaluation of exposures to an insoluble 14C aerosol. The same detector system was also calibrated for other low-energy beta particle emitting radionuclides such as 63Ni and 147Pm.24 While the activity detectable in this manner was high compared with what direct measurements are usually able to achieve, this method was sensitive enough to be a useful tool. Outer bremsstrahlung can also be used in certain circumstances. For example, 32P can be measured in this way. In this example, whole-body counting relied on measuring the bremsstrahlung radiation produced as a result of the interaction of beta radiation with the body’s tissues. A 32P-spiked model was used as a calibration for this measurement.25 19.3.2 Lung Counters As the name suggests, this technique is applied to a person who has inhaled an insoluble radioactive material. Examples of these types of materials include uranium dioxide and high-fired plutonium compounds. The working range of a lung counter is from ~10 keV to about 400 keV. The types of detectors include sodium iodide, phoswich detectors, and hyperpure germanium detectors, which must be cooled by liquid nitrogen. There are now detectors than are electrically cooled and they are slowly making their way into this application, but at the time of this writing, no lung counter use electrically cooled detectors. All lung counting detectors have thin windows to collect as many of the low-energy photons as possible. In the past they were constructed of beryllium, but more recently the window material of choice has been carbon fiber. Typically the detectors are mounted in an array. The large diameter detectors (sodium iodide and phoswich) are generally grouped in pairs. The hyperpure germanium detectors were originally small in diameter and could be found in arrays of four per lung, for a total of eight detectors; however, the manufacturing process for these detectors is now capable of fabricating detectors up to 85 mm in diameter. Typically an array of two detectors per lung will be used for a total of four detectors.

460

Handbook of Anatomical Models for Radiation Dosimetry

Measuring low-energy photons has many difficulties and uncertainties associated with the process. In no particular order, the difficulties are 1. Transmission through muscle and adipose tissue is different, so accurate knowledge is required to make the best estimate. 2. Chest wall thickness is a critical parameter in determining lung burden. 3. Ribs will greatly attenuate the passage of photons, and while this is partially compensated by use of an anthropometric model, ribs in people are not all the same. 4. The distribution of activity is usually unknown. 5. Activity passing through the GI tract, or deposited in other organs, can greatly influence the results. 6. External contamination can be a problem, especially if near the detectors. 7. Size of the subjects’ lungs is likely to differ from the calibration models, and this will introduce a systematic bias into the activity estimate. 8. Chest wall profile is different from the calibration model and this will affect the result. 9. Finally, detector placement, especially when measuring low-energy photons, can have a great effect on the activity estimate. Lung counting is often a technique used in conjunction with an indirect method. The analysis of urine is very sensitive and has a much better detection limit than lung counting for most nuclides of interest (e.g., plutonium, uranium); however, it is always advisable to perform lung counting, as these materials are often excreted in small amounts. A negative (nondetectable) result from a lung count immediately puts an upper limit on any internal dose. Once a lung burden is obtained, the intake can be obtained by using a computer code or the retention figures and tables in the ICRP documents. Naturally it is advisable, as with any confirmed contamination, to use multiple measurements to determine the individual’s retention parameters. 19.3.3 Organ Counting Two examples are discussed here: thyroid counting and bone counting. The former is often carried out in hospitals that are using 125I or 131I. The former is also used in immunoassay while the latter is used for diagnosis and therapy of the thyroid gland. While it is true that both nuclides can be detected by urine analysis, the retention of iodine is such that great uncertainty could be introduced into the intake estimate because the excretion rate changes by several orders of magnitude over a short time. Back extrapolation from such a fast-changing function, even if the intake time is known very accurately, can cause great uncertainty in the estimate of the intake, and hence the dose. It is, therefore, much more accurate to measure iodine nuclides directly in the thyroid gland. This can be achieved with a short count time (60–300 s). Typically the detector will be a scintillation detector (sodium iodide) with only a few facilities using hyperpure germanium detectors. The hyperpure germanium detectors are generally used by facilities that are using the germanium detectors for other purposes (i.e., whole-body or lung counting). Once the activity in the thyroid has been obtained, the amount can be adjusted to the time of intake (if known) and the intake can be obtained by assuming that the uptake of the thyroid gland is 30%.3

Applications to Bioassay for Internal Radiation Contamination

461

Bone counting is performed for persons who have been historically exposed to a material that will be incorporated into bone as part of the metabolism process. Examples of these nuclides are 210Pb and 226Ra. Detectors are often the same as found in a lung counter. The areas of the body that are amenable to measurement are the skull and the knees. The measurements will be performed in a low background chamber, perhaps the same used by a facility for its lung counting, in order to achieve low detection limits. Once the activity in the area of bone that is being measured has been obtained, it must be converted into the total activity in the whole skeleton. This assumes that nuclide is deposited homogeneously through the bone type with which it was associated with surface seeker or volume seeker. The assumption is not true because the nuclide is actually deposited heterogeneously. 19.3.4 Wound Counting As the name suggests, this technique is the monitoring of a wound site using an external detector. Biopsy of the wound site is not discussed, as the analysis of that sample would be treated as an indirect method. Typically wound counting would be performed if an insoluble material was embedded in tissue of the wound site. Any of the previously mentioned detectors could be used for this application. Once an activity was determined, a judgment would have to be made to either treat the contaminated tissue to reduce the activity, or to remove the contaminated tissue by surgery. The decision would likely be based either on the potential dose that could be realized if all the material estimated in the wound site were solublized and incorporated into the body or the localized dose to the part of the body with the wound. 19.3.5 Dose Estimates from Direct Methods The intake can be obtained once the body burden (or the organ burden) has been established and if the intake time is known. In cases where the intake time is not known, the ICRP recommends that the midpoint of a monitoring period be used, but others26 have suggested that an unbiased estimate of the intake can be obtained if it assumed that there has been a chronic intake over the entire unknown period prior to the measurement. Naturally, the latter can only be done by use of computer codes. The former can be done using the tables and figures for the nuclide of interest found in the numerous ICRP publications. Naturally, it is advisable to have multiple measurements to confirm that a contamination has taken place and that it was not simply a contamination of the sample container. A series of measurements show how the material’s excretion behaves with time, and allows the dosimetrist to refine the dose estimate using person-specific retention parameters, although this process usually only takes place if the person is likely to exceed the national dose limit for a worker, or member of the public.

19.4 Calibration As stated above, any measure of radiation (or signal in a mass spectrometer) is a relative measurement. To obtain an amount of material present, the signal of the unknown sample

462

Handbook of Anatomical Models for Radiation Dosimetry

must be related to a sample that had a known amount of material in it. Some of the aspects of calibration of the equipment that have been mentioned above will be described in this section. 19.4.1 Alpha Counting The determination of the amount of an alpha emitter generally involves separations chemistry to remove interferents to the fi nal analysis method, and to reduce the matrix so the fi nal sample can be prepared for measurement. During this process some of the material of interest will be lost due to processing. In order to account for this loss, a tracer is often added to the sample. The amount of tracer that is recovered is then assumed to represent the recovery of the nuclide of interest. One problem with this methodology is that if the tracer is not in the same chemical and physical form as the nuclide in the sample, then one cannot be sure if the separations chemistry has treated the nuclide of interest and the tracer the same way. In other words, the recovery of the tracer and the nuclide of interest may be different. While not a great problem for materials in urine (as they were soluble to get there in the fi rst place) it can be a concern for fecal samples. In the latter case it is essential that any solubilization steps, such as a fusion, be 100% effective.27 The choice of tracer should be such that its alpha emission does not interfere with the nuclide of interest, and that both are easily separable. Naturally a tracer that emits alpha radiation can only be used if alpha spectrometry is being performed. If alpha counting (no spectroscopy) is being performed, then the tracer should be measurable by some other technique, such as gamma. The counting efficiency of the detector used in this process is determined by using a known quantity of material that has a similar, if not the same, alpha emission spectrum to the material being measured. During this process it is essential than no loss of activity occur between opening the standard solution and preparation of the standard. It is also essential that the calibration standard be in the same form as the sample being measured. For example, if the final product is a stainless steel disc with a filter paper containing a small amount of coprecipitate containing the nuclide of interest, then the standard used to calibrate the counter must be in that form, with a similar (if not same) mass. 19.4.2 Beta Counting Every consideration stated above for alpha counting can apply to beta counting. There are further considerations, as beta analysis can be done using liquid scintillation. For lower energy emitters, notably 3H, the sample quench can alter the results substantially so it is essential that the instrument has a quench curve to account for this. The two types of quench must be accounted for: color and chemical. Tea has been found to be a good mimic for urine and can be used to generate a quench curve. Planchet counting is often done using a coprecipitate whose mass can be somewhat variable. In this case it is essential that the calibration curve for counting the efficiency of the gas proportional counter be as a function of precipitate weight on the counting planchet. One of the problems encountered in this type of calibration curve can be due to clumping of the coprecipitate, which may be quite different in form from the calibration samples. The effect of this is something that can be investigated by Monte Carlo simulation, a technique that is more fully discussed in the following.

Applications to Bioassay for Internal Radiation Contamination

463

19.4.3 Gamma Counting A urine sample is generally placed in a Marinelli beaker of fixed volume, placed on an upright detector in a counting shield, and measured for an appropriate period of time. This instrument should be calibrated using a standard gamma source placed in the Marinelli beaker (of the same size) that is used in routine measurements. The volume of liquid should be fixed so that the standard solution and analysis solutions will be the same for all measurements. The only complicating factors that can occur are small density corrections between the standard solution, which was likely acidified water, and the urine sample (diluted to the same volume with water), and coincidence summing, which can be important if a calibration curve is being used instead of the actual nuclide being measured. In the latter case, if the same nuclide is used as a standard for what is being measured, and it is subject to coincidence summing, then the calibration process will largely compensate for this providing the activity levels are similar. Also, many of today’s gamma analysis software packages have coincidence summing corrections modules. Nevertheless, this is unlikely to be a large problem unless a person becomes heavily contaminated, in which case a direct method should be used. 19.4.4 Whole-Body Counting A whole-body counter must be calibrated with a radioactive source that represents the human body. The industry standard for this device is the Bottle Manikin Absorber (BOMAB) Phantom28 as shown in Figure 19.5. The BOMAB phantom, as originally designed,29 consisted of 10 elliptical and circular cylinders that could be filled with a water solution containing a radioactive material. Nominally, the BOMAB phantom represented the early definition of ICRP’s Reference Man30 that has been superceded by a new definition that is slightly taller and heavier.2 The BOMAB containing a radioactive solution is measured in the whole-body counter, and the counting efficiency for that energy is obtained. The process is repeated until a

FIGURE 19.5 BOMAB phantom family.

464

Handbook of Anatomical Models for Radiation Dosimetry

counting efficiency curve as a function of energy has been constructed. The process may be repeated using different sized BOMAB phantoms so that a facility has a counting efficiency at a given energy and for a given size; however, few facilities actually do this. As a result the counting efficiency for the Reference Man is generally used for all persons being measured, and this can introduce a large systematic bias into an activity estimate (of the order of a factor of two) if the person being measured is substantially different in size to the calibration phantom. Compared with the calibration phantom, the amount of adipose tissue a person has is a parameter that is largely ignored, but can be estimated with the use of overlay plates;31 however, this correction factor is difficult to apply. Monte Carlo simulations can be used to extend a facilities calibration data set, or can be used to investigate the effect of different parameters of the whole-body counter’s efficiency. Until recently, it was only possible to defi ne the whole-body counter virtually using a combination of geometric shapes (planes, cubes, cylinders, spheres, etc.) but recent advances in the generation of voxel models has made it possible to take realistic images from computer tomography scans or magnetic resonance imaging scans and convert them into a three-dimensional model of the human body. For example, a male has been made into a voxel model known as NORMAN32 and a female has been made into a voxel model known as NAOIMI.33 As a result, a whole-body counter can be “virtually” calibrated in a number of different ways to see the effect on the counting efficiency. For example, 1. 2. 3. 4.

The source location can be varied. The source distribution can be changed. The size of the individual can be changed. The relationship between a real person and the BOMAB phantom can be established for a given counting geometry. 5. The amount of overlying adipose tissue can be altered. 6. The effect of source movement as a result of metabolism can be studied.

On the physical side, phantoms for whole-body counting, and other applications, can be borrowed from the Phantom Library34 or from the International Atomic Energy Agency (for member states). 19.4.5 Lung Counting The low-energy photons that are generally measured by this technique require the calibration phantom be much more anthropometric than the BOMAB phantom. There are two such models available for the calibration of a lung counting system. They both have advantages and disadvantages. The first, which is the accepted industry standard in North America, is the Lawrence Livermore National Laboratory’s (LLNL) Torso Phantom35 shown in Figure 19.6. The LLNL phantom is available in three versions: 1. Original—three copies and made with real human bone. 2. Second generation—16 copies made by Radiology Support Devices (RSD, Radiology Support Devices Inc. 1904 E. Dominguez St., Long Beach, CA 90810), formerly known as Humanoid Systems Inc. 3. Commercial version—small mold changes, and also available from RSD.

Applications to Bioassay for Internal Radiation Contamination

465

FIGURE 19.6 The LLNL phantom showing internal organs, cover plate, and one overlay plate.

The phantom closely mimics the tissues and organ placement of a human torso to simulate the interaction of low energy photons (<200 keV) with bone, cartilage, muscle, and adipose tissues. It is constructed of muscle equivalent material and contains synthetic bone (after first generation) and synthetic cartilage. The overlay plates, which allow the user to simulate a range of chest wall thicknesses, can be constructed of different adipose–muscle equivalent mixtures. The torso cavity contains removable lungs, heart, liver, and other organs. All internal organs, except the lungs, are constructed of muscle equivalent material; the lungs are constructed of lung equivalent material. The LLNL phantom contains a rib cage, and a spinal tissue block substituted for the spine, 36 but no scapula. The LLNL phantom was originally supplied with three series of overlay plates, each set consisting of four overlay plates. The overlay plates simulate chest wall thicknesses varying from about 1.6 (no overlay) to 4.1 cm (thickest overlay plate added). Since then, two additional plates for the C-series have been constructed that extend the range of chest wall thickness values to 6.4 cm. The need for this extension was based on ultrasound measurements of the chest wall thickness of a large number of workers in the Canadian uranium industry.37 Each overlay plate series for the LLNL phantom simulates a different adipose/muscle composition. Series A simulates 87% adipose and 13% muscle, Series B simulates 50% adipose and 50% muscle, Series C simulates 0% adipose and 100% muscle. The adiposemuscle proportions of the overlay plates must be combined with the torso plate cover (100% muscle) so that the combined muscle–adipose compositions vary from 15%:85% (adipose:muscle) to 48%:52% (adipose:muscle) when the A- and B-series are used. The composition remains constant at 0%:100% (adipose:muscle) when the C-series overlay plates are used. The second phantom is the Japanese Atomic Energy Institute’s (JAERI) Torso Phantom38 shown in Figure 19.7. The JAERI phantom, as supplied by the manufacturer, consists of a torso core, a chest plate cover, sliced lungs, other internal organs, and six overlay plates. The adipose content

466

Handbook of Anatomical Models for Radiation Dosimetry

FIGURE 19.7 The JAERI phantom showing the internal organs, cover plate, and one overlay plate.

of the JAERI phantom is either 10%, 20%, or 30% depending on which overlay series is used, and the chest wall thickness can be varied from 1.9 to 3.2 cm (stacking multiple plates can extend this range). The phantom comes with sliced inert lungs so the user has the choice of manufacturing (or contracting) whole lungs sets with the activity homogeneously distributed throughout the lung tissue substitute, or of using laminated planar sources. The latter method has been shown to be equivalent39 to a lung set with the activity homogeneously distributed throughout the lung tissue substitute. Monte Carlo simulations can be used to make useful extensions to experimental calibrations. Humans are not all the same size and can differ significantly from the models’ shapes and sizes. As a result, the counting efficiency determined from the calibration model may not accurately represent the counting efficiency of the subject being measured. As the shape or size discrepancy between subject and phantom increases, so too will the error in the activity estimate increase. For example, lung size is a parameter that cannot be changed experimentally as neither of the existing phantoms have lung sets of different volumes, making this study physically impossible. This has been studied elsewhere using Monte Carlo simulations,40 and the fi ndings show that as a person’s lung volume is unknown, any lung count will have an inherent error in the activity estimate. The overestimate of the activity may be as high as a factor of 1.63 at 17 keV using an array of four 50 mm diameter detectors, and the underestimate of the activity may be as high as a factor of 1.27 at 17 keV using an array of four 85 mm diameter detectors. At 60 keV these values diminish to 1.41 and 1.20, respectively. 19.4.6 Organ Counting As for whole-body and lung counting, the accuracy of any measurement will depend on how complete the calibration is done. Thyroid counters can be calibrated using a thyroid phantom of which there are several available. They have been compared elsewhere.41 Three

Applications to Bioassay for Internal Radiation Contamination

467

of the four phantoms are equivalent when phantom to detector distances are greater than or equal to 12 cm, and all phantoms are essentially equivalent at 30 cm or more. The model for thyroid counter calibration need not be highly anthropomorphic, provided the calibration is not performed at short neck-detector distances. Bone counters, alternately, are generally placed on contact (or at least very close) to the subject’s skull or knee. As a result the calibration phantoms need to be anthropometric.42,43 Investigating counting geometry effects or extending the calibration curve, can easily be performed using Monte Carlo simulations, especially if a voxel model is used.44 19.4.7 Wound Counting An accident involving radioactive material can result in soluble or insoluble particles being embedded in a wound. In the former case, the soluble material will leave the wound site and is treated as an injection for dosimetry purposes, whereas an insoluble (or slowly solubilizing material) will remain in the wound site irradiating surrounding tissues and providing a reservoir for chronic systemic contamination. It is usually actinides that give the most cause for concern with wounds45 and the attenuation of the low-energy x-rays.46 Calibrations of such systems are not standardized and phantoms are generally made in-house for a specific case; however, Monte Carlo simulations have provided an alternate calibration route.47 19.4.8 Quality Assurance All methods described above must have a good, and preferably internationally recognized, quality assurance program in place to provide a high degree of confidence on the bioassay or in vivo result. Bioassay methods must, therefore, include the running of blank samples to test reagents and procedures. A reagent blank ensures that no contamination is present in the chemicals or equipment. A quality control sample, made up in the same matrix as the real sample, provides assurance that the procedure is working when the expected activity is obtained at the end of the analysis. If an internal tracer is used in the procedure, this type of sample can be replaced, but in cases where that is not possible, it provides assurance that the recovery is as expected. Standard sources and blank sources are used to check the instrumentation to ensure that the response is as expected and that there is no contamination present. In vivo methods must include check sources to validate that energy and efficiency response of the counting equipment. It is not necessary to use phantoms for this test. Backgrounds should be measured to ensure that no contamination is present, and the best way to do this is use an uncontaminated person. Where this is not possible, a phantom is a good alternative. Both bioassay and in vivo measurement programs should not rely on in-house sources to validate that procedures are working as expected. It is also advisable to participate in intercomparison exercises where the test sample is made by an outside, independent organization. Passing this test provides good assurances that not only are the procedures working as expected, but that the in-house quality control is working as expected. All in-house and external quality control tests should be displayed on a control chart so that long-term trends can be identified and remedial action taken before a failure occurs.

468

Handbook of Anatomical Models for Radiation Dosimetry

19.5 Conclusion Bioassay generally provides the starting point for evaluating dose. While there are many uncertainties in reaching a dose estimate, minimizing the uncertainty of the activity estimate is relatively easy compared with the other parameters, and should be done whenever possible. This is achievable if the analyzing laboratory develops or uses a well-tested method appropriate to the sample type, has a good quality assurance program in place, and bases the dose estimate on multiple measurements when investigating a contamination incident. The choice of what bioassay method, direct or indirect, will depend on a number of factors such as the solubility, the emission type (alpha, beta, gamma), the ease of sample collection, or the impact of measurement type on work practices; but it should be clear that the preferred method is always direct, as this avoids the assumption related to the amount of excreta defined as days-output. The calibration of the instrument used to assess the activity in the sample, or person, is critical to obtaining an accurate result. The calibration of equipment associated with indirect methods should be straightforward as the sample can easily be manipulated to match the standard sample that was used to calibrate the instrument (or be in a valid range if precipitate weight is a factor). It is not so simple to manipulate the sample to be the same as the standard used to calibrate an in vivo monitor. In this case, it should be clear that it is largely the other way round. The in vivo equipment should be calibrated over a wide range so the sample type (i.e., the person’s size) falls within the calibration range. For a whole-body counter, the counting efficiency should be characterized as a function of height and weight over the expected range of persons that would be measured by that facility. For a lung counter, the most important parameter is chest wall thickness, but there are others too as have been mentioned above. The bioassay measurement, whether direct or indirect, is generally used to estimate an intake that, in turn, can be used to estimate the internal dose. It is hoped that the reader now appreciates that this process can be more of an art than a science, and that the uncertainties on the final dose estimate can be many.

References 1. Snyder, W.S. et al., Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, Journal of Nuclear Medicine, Suppl 3, 7, 1969. 2. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, 2003. 3. ICRP. Individual Monitoring for Internal Exposure of Workers. Replacement of ICRP Publication 54, ICRP Publication 78, Pergamon Press, Oxford, 1997. 4. Tracy, B.L. et al., Human uptake of radiocesium from caribou meat, Radiation Protection Dosimetry, 48, 317, 1993. 5. ICRP, Human Respiratory Tract Model for Radiological Protection: A Report of a Task Group of the International Commission on Radiological Protection, ICRP Publication 66, Pergamon Press, Oxford, 1994.

Applications to Bioassay for Internal Radiation Contamination

469

6. Centres for Disease Control and Prevention, Facts about DTPA, http://www.bt.cdc.gov/radiation/dtpa.asp, last accessed November 2007. 7. Taylor, D.M., Hodgson, S.A., and Stradling, N., Treatment of human contamination with plutonium and americium: Would orally administered Ca- or Zn-DTPA be effective? Radiation Protection Dosimetry, 127, 469, 2007. 8. Centres for Disease Control and Prevention, Fact sheet Prussian Blue, http://www.bt.cdc.gov/ radiation/prussina blue.asp, last accessed November 2007. 9. Birchall, A. et al., IMBA expert (TM): Internal dosimetry made simple, Radiation Protection Dosimetry, 105, 421, 2003. 10. Eikenberg, J. et al., Fast radiochemical screening of transuranium radionuclides in urine using actinide extractive resin and low-level α/β LSC, Radioactivity and Radiochemistry, 10, 3, 1999. 11. Environmental Measurements Laboratory, The Procedures Manual of the Environmental Measurements Laboratory (HASL-300), http://www.eml.st.dhs.gov/publications/procman/, last accessed November 2007. 12. Karpas, Z. et al., Measurement of the U-234/U-238 ratio by MC-ICPMS in drinking water, hair, nails, and urine as an indicator of uranium exposure source, Health Physics, 89, 315, 2005. 13. Popplewell, D.S. et al., Isotopic composition of plutonium in human-tissue samples determined by mass-spectrometry, Radiation Protection Dosimetry, 26, 313, 1989. 14. Wyse, E.J. and Fisher, D.R., Radionuclide bioassay by inductively-coupled plasma-mass spectrometry (Icp/Ms), Radiation Protection Dosimetry, 55, 199, 1994. 15. Maxwell, S.L., Rapid analysis of emergency urine and water samples, Journal of Radioanalytical and Nuclear Chemistry, 275, 497, 2008. 16. Pappas, R.S., Ting, B.G., and Paschal, D.C., Rapid analysis for plutonium-239 in 1 mL of urine by magnetic sector inductively coupled plasma mass spectrometry with a desolvating introduction system, Journal of Analytical Atomic Spectrometry, 19, 762, 2004. 17. McDonald, J.C., Radiation protection and international intrigue, Radiation Protection Dosimetry, 123, 1, 2007. 18. Alvarez, A. and Navarro, N., Method for actinides and Sr-90 determination in urine samples, Applied Radiation and Isotopes, 47, 869, 1996. 19. Isnard, H. et al., Determination of Sr-90/U-238 ratio by double isotope dilution inductively coupled plasma mass spectrometer with multiple collection in spent nuclear fuel samples with in situ Sr-90/Zr-90 separation in a collision-reaction cell, Spectrochimica Acta Part B Atomic Spectroscopy, 61, 150, 2006. 20. Vonderheide, A.P. et al., Determination of Sr-90 at ultratrace levels in urine by ICP-MS, Journal of Analytical Atomic Spectrometry, 19, 675, 2004. 21. Evans, R.D., The Atomic Nucleus, McGraw-Hill, New York, 1958, p. 617. 22. Pani, R. et al., Lanthanum scintillation crystals for gamma ray imaging, Nuclear Instruments and Methods in Physics Research Section a—Accelerators Spectrometers Detectors and Associated Equipment, 567, 294, 2006. 23. Marsh, J.W., Bailey, M.R., and Birchall, A., A step-by-step procedure to aid the assessment of intake and doses from measurement data, Radiation Protection Dosimetry, 114, 491, 2005. 24. Johnson, J.R., Lamothe, E.S., and Kramer, G.H., The measurement of low-energy beta-particle emitting radionuclides in the lung using Phoswich detectors, Radiation Protection Dosimetry, 20, 267, 1987. 25. McCunney, R.J., Masse, F., and Galanek, M., Occupational ingestion of P-32: The value of monitoring techniques to determine dose—A case report, Journal of Occupational and Environmental Medicine, 41, 878, 1999. 26. Puncher, M., Marsh, J.W., and Birchall, A., Obtaining an unbiased estimate of intake in routine monitoring when the time of intake is unknown, Radiation Protection Dosimetry, 118, 280, 2006. 27. Sill, C.W., Simultaneous determination of U-238, U-234, Th-230, Ra-226, and Pb-210 in uranium ores, dusts, and mill tailings, Health Physics, 33, 393, 1977. 28. ANS, Specifications for the Bottle Mannequin Absorption Phantom, ANSI/HPS, N13.35, 1999.

470

Handbook of Anatomical Models for Radiation Dosimetry

29. Bush, F., The integral dose received from a uniformly distributed radioactive isotope, British Journal of Radiology, 22, 96, 1949. 30. ICRP, Report of the Task Group on Reference Man, ICRP Publication 23, Pergamon Press, Oxford, 1975. 31. Kramer, G.H., Burns, L., and Noel, L., The BRMD BOMAB phantom family, Health Physics, 61, 895, 1991. 32. Dimbylow, P.J., FDTD calculations of the whole-body averaged SAR in an anatomically realistic voxel model of the human body from 1 MHz to 1 GHz, Physics in Medicine and Biology, 42, 479, 1997. 33. Dimbylow, P.J., The calculation of SAR from limb current in the female voxel phantom, NAOMI, Radiat Prot Dosimetry, 121, 236, 2006. 34. United States Department of Energy, phantom library cataolg, http://www.pnl.gov/phantom, last accessed Jun 2008. 35. Griffith, R.V. et al., Tissue equivalent torso phantom for intercalibration of in vivo transuranic nuclide counting facilities, Advances in Radiation Protection Monitoring, STI/PUB/494 Proc. IAEA Conf. IAEA-SM-229/56, IAEA, Vienna, 1978. 36. Kramer, G.H. and Webber, C.E., Evaluation of the Lawrence Livermore National Laboratory (Llnl) Torso Phantom by Bone Densitometry and X-Ray, Applied Radiation and Isotopes, 43, 795, 1992. 37. Kramer, G.H., Hauck, B.M., and Allen, S.A., Chest wall thickness measurements and the dosimetric implications for male workers in the uranium industry, Health Physics, 80, 74, 2001. 38. Shirotani, T., Realistic torso phantom for calibration of in-vivo transuranic-nuclide counting facilities, Journal of Nuclear Science and Technology, 25, 875, 1988. 39. Kramer, G.H. and Guerriere, S., Comparison of sliced lungs with whole lung sets for a torso phantom measured with Ge detectors using Monte Carlo simulations (MCNP), Health Physics, 84, 222, 2003. 40. Kramer, G.H. and Capello, K., Effect of lung volume on counting efficiency: A Monte Carlo investigation, Health Physics, 88, 357, 2005. 41. Kramer, G.H. et al., Comparison of the ANSI, RSD, KKH, and BRMD thyroid-neck phantoms for I-125 thyroid monitoring, Health Physics, 70, 425, 1996. 42. Kellar, J., Fabrication of an Anthropomorphic Calibration Phantom for In Vivo Measurement of 152Eu in the Skull, University of Cincinnati, 1995. 43. Spitz, H. et al., A new anthropometric phantom for calibrating in vivo measurements of stable lead in the human leg using x-ray fluorescence, Health Physics, 78, 159, 2000. 44. Gomez-Ros, I.M. et al., A numerical method for the calibration of a whole body counter. application to in vivo measurements of Am-241 in skull, Health Physics, 84, S172, 2003. 45. Johnson, L.J. and Lawrence, J.N., Plutonium contaminated wound experience and assay techniques at Los-Alamos-Scientific-Laboratory, Health Physics, 27, 55, 1974. 46. Tyler, G.R., Self Absorption of x-rays by plutonium particles with special reference to plutonium in wound monitoring, Health Physics, 12, 509, 1966. 47. Borisov, N. et al., A new graphical user interface for fast construction of computation phantoms and MCNP calculations: Application to calibration of in vivo measurement systems, Health Physics, 83, 272, 2002.

20 Applications to Nuclear Medicine Michael G. Stabin and Manuel Bardiès

CONTENTS 20.1 Introduction ............................................................................................................... 471 20.2 Standardized Methods and Computational Phantoms for Internal Dose Calculations ................................................................................ 472 20.2.1 Absorbed Dose Calculational Schema...................................................... 472 20.3 Available Anthropomorphic Computational Phantoms ..................................... 474 20.3.1 Marrow Dose Computational Phantoms ................................................. 475 20.4 Currently Available DCFs ........................................................................................ 475 20.4.1 Standard Organ Masses .............................................................................. 476 20.5 Standard Dose Estimates for Radiopharmaceuticals .......................................... 476 20.5.1 Image-Based Approaches ........................................................................... 477 20.6 Input Data for Dosimetry Calculations ................................................................. 477 20.6.1 Quantification of Planar Data .................................................................... 479 20.7 Quantification of Tomographic Data ...................................................................... 481 20.8 Status and Evolution of Dose Calculational Approaches ................................... 482 References .............................................................................................................................484

20.1 Introduction Internal dose calculations for diagnostic and therapeutic applications in nuclear medicine are based currently on standardized computational phantoms for reference adults, children, and pregnant women. In contrast to external beam therapy, in which a patientindividualized computational phantom is developed for each patient, doses from these standardized subjects may be modified using patient-specific values of organ mass, for example. New technologies promise a new era for dose calculations, one in which imagebased computational phantoms may be used to develop more truly patient-specific dosimetry. For the present, however, standardized dose calculations predominate. This chapter examines the state of the art in this field, with descriptions of standard computational phantoms and methods for obtaining these estimates.

471

472

Handbook of Anatomical Models for Radiation Dosimetry

20.2 Standardized Methods and Computational Phantoms for Internal Dose Calculations Radiation dose estimates for diagnostic or therapeutic radiopharmaceuticals are routinely calculated to evaluate the risks and benefits of radiopharmaceutical use. The quantity of “absorbed dose,” i.e., the amount of energy from ionizing radiation that is absorbed per unit mass of any material, is the usual parameter of interest. One may also calculate the equivalent dose by use of appropriate radiation weighting factors (wR) as well as the “effective dose” as defined by the ICRP,1 which is the risk-weighted equivalent whole-body dose from a given administration. These quantities, however, are only used in the evaluation of diagnostic agents, and should not be applied for therapeutic agents. The risk has to be balanced with the goal of the radiopharmaceutical administration. For diagnostic procedures involving low radiopharmaceutical activity, the absorbed doses are usually well below the threshold for deterministic effects. This situation is very similar to that of radiology. In that sense, an estimate of absorbed doses is most often considered sufficient to establish the benefit/detriment ratio. For therapeutic procedures, the level of irradiation is higher, since the goal is to observe deterministic effects at the tumor level, while also sparing the surrounding healthy tissues. The level of accuracy required for the absorbed dose calculation is therefore higher than that needed for diagnostic procedures. The consequence of this higher dose is that the absorbed dose calculations in a therapeutic context are usually more demanding in terms of data acquisition/processing. Absorbed dose calculation methods, however, are versatile and can accommodate both diagnostic and therapeutic applications. 20.2.1 Absorbed Dose Calculational Schema A generic equation for the absorbed dose in an object T irradiated by a source S uniformly contaminated with radioactivity (for example, an organ or tissue with radiopharmaceutical uptake) is shown as ~

D(T ←S ) =

k AS

∑ y E φ (T ← S) i

i

i i

mT

(20.1)

where — D (T ← S) is the absorbed dose to a target region of interest (Gy or rad) yi is the number of radiations with energy Ei emitted per nuclear transition Ei is the energy per radiation for the ith radiation (MeV) φi(T ← S) is the fraction of energy emitted in a source region that is absorbed in a target region mT is the mass of the target region (kg or g) k is the proportionality constant (Gy-kg/MBq-s-MeV or rad-g/μCi-h-MeV) The proportionality constant k includes the various factors that are needed to obtain the absorbed dose in the desired units from the units employed for the other variables, and it is essential that this factor is properly calculated and applied. The quantity ÃS is often called the “cumulated activity” (Bq-s or μCi-h) and gives the total “number of disintegrations that have occurred over time in a source region.” ÃS is calculated as

Applications to Nuclear Medicine

473

∞

~

AS =

∞

∫ A (t )dt = A ∫ f (t )dt S

0

0

(20.2)

S

0

where A0 is the activity administered to the patient at time t = 0 fS(t) may be called the fractional distribution function for a source region (the fraction of administered activity present within the source region at time t) In many instances, the function fS(t) may be modeled as a sum of the exponential functions: f S (t) = f1e

− ( λ1 + λ p) t

+ f 2e

− ( λ 2 + λ p) t

+ + fNe

− ( λ N + λ p) t

(20.3)

where f1, …, f N represents the fractional uptake of the administered activity within the 1st to Nth compartments of the source region λ1, …, λN represents the biological elimination constants for these same compartments λp represents the physical decay constant for the radionuclide of interest Other functional expressions may be used to represent the fractional distribution function, but exponentials are most commonly encountered. A generalized expression for calculating the internal dose, which may describe the equations shown in publications by different authors, can be calculated by the following equation: D = N × DF

(20.4)

where N is the number of nuclear transitions that occur in source region S (identical to ÃS) DF is a “dose factor” It is important to note that by introducing a DF for the irradiation process (i.e., for all radiation emitted from N nuclear transitions) one implicitly assumes that the geometry of the irradiation remains constant over the period of integration of AS(t). The factor DF contains the various components shown in the formulas above for dose except ÃS; it combines decay data with absorbed fractions (AFs, values of φi(T ← S)), which are derived generally using the Monte Carlo simulation of radiation transport in computational phantoms of the body and its internal structures (organs, tumors, etc.):

DF =

k

∑ y E φw i

i

mT

i i Ri

(20.5)

As written, the above equations give only the dose from one source region to one target region, but they can be generalized easily to multiple source regions, and the radiation weighting factor (wR) may be included so that equivalent dose (H) may be calculated:

474

Handbook of Anatomical Models for Radiation Dosimetry

HT = D T × wR =

k

∑ A ∑ y E φ (T ← S)w ~

S

S

i

i

i i

Ri

mT

(20.6)

An interesting feature of this formalism is that it divides the dosimetric process into two different tasks: 1. N or A˜ S (cumulated activity) determination is performed in all regions where activity can be found during the irradiation process. The goal is to be able to determine the spatial and temporal behavior of the radiopharmaceutical. It is usually performed (but not exclusively) via quantitative scintigraphic imaging. 2. DF determination corresponds to the other term of the equation. According to the sought degree of refinement/accuracy, DF determination can be established for standardized computational phantoms of patient or for individual patients from patient image data (i.e., CT or MRI images). For diagnostic procedures, one is generally unable to establish N on a per-patient basis. Data are generally collected for a group of healthy volunteers, on a subset of patients, or extrapolated from animal data in order to give an estimate of standard pharmacokinetics. Computational phantoms are then used to derive absorbed doses in various compartments/organs of an “average” individual. This is seen in the dose estimates reports available from the International Commission on Radiological Protection (ICRP) for most radiopharmaceuticals used in routine diagnostic or therapeutic applications. Absorbed doses are usually given on a per-injected-activity basis, so that the absorbed dose estimates for a given patient merely require the knowledge of injected activity. For therapeutic procedures, dose estimates and the appropriate activity to administer should be assessed for each patient. Quantitative imaging is usually a complex task, and can be addressed by different approaches. DF calculations, as well, should be specific, i.e., based on patient data. A simpler approach is to adjust DF data obtained from available anthropomorphic computational phantoms in order to more closely model a given patient’s geometry. The computing power needed to calculate DFs in the use of patient-specific image-based geometry definitions is demanding, and has not been applied widely in clinical practice. It is possible to observe two converging trends in patient-specific absorbed dose calculation: one starts from patient computational phantoms and try to make these more “patient-like.” The other starts from patient data (CT images) and try to make it more manageable from a computing point of view. Both approaches are likely to merge in the near future.

20.3 Available Anthropomorphic Computational Phantoms The current generation of anthropomorphic computational phantoms began with the development of the Fisher–Snyder computational phantom,2 which employed a combination of geometric shapes—spheres, cylinders, cones, etc.—to create a simplified representation of the human body. Monte Carlo computer programs were used to simulate the creation and transport of photons through these various structures in the body, whose

Applications to Nuclear Medicine

475

atomic compositions and densities were based on data provided by the ICRP in its widely quoted report on “Reference Man,”3 now updated in a more recent report.4 These reports provide various anatomical data helpful in producing dose calculations for standardized individuals. The AFs and dose conversion factors (DCFs) (S values) as defined above, for over 100 radionuclides and over 20 source and target regions, were also published.5,6 Cristy and Eckerman7 modified the adult male computational phantom and developed computational phantoms for a series of individuals of different sizes and ages. They developed six computational phantoms, which were assumed to represent children of ages 0 (newborn), 1, 5, 10, and 15 year, and adults of both sexes. They published AFs for photons at discrete energies for these computational phantoms, which contained approximately 25 source and target regions. Tables of S values were never published, but ultimately were made available in the computer software called “MIRDOSE,”8 which was widely used by the nuclear medicine community. Stabin et al. developed a series of computational phantoms for the adult female, including a computational phantom of the nonpregnant adult female and women at three stages of pregnancy.9 These computational phantoms illustrated the changes to the uterus, the intestines, the bladder, and the other organs that occur during pregnancy, and included specific computational phantoms for the fetus, the fetal soft tissue, the fetal skeleton, and the placenta. S values for these phantoms were also made available to the dosimetry community through the MIRDOSE8 and OLINDA/ EXM10 software programs. As outlined in other chapters in this book, these computational phantoms are now being replaced with more realistic computational phantoms based on human image data.

20.3.1 Marrow Dose Computational Phantoms Spiers et al. at the University of Leeds11 first developed the electron AFs for bone and marrow for an adult male subject; these results were used to calculate DCFs, or S values, in MIRD Pamphlet No. 11.5 Eckerman and Stabin12 reevaluated this work and extended the results to derive DFs for 15 skeletal regions in six computational phantoms representing individuals of various ages. The results were also used in the MIRDOSE 3 software8 to provide average marrow dose, regional marrow dose, and dose–volume histograms for different aged individuals. Bouchet et al.13 used newer information on regional bone and marrow mass, and calculated new AFs using the EGS4 Monte Carlo code. Although the results of the Eckerman and Bouchet et al. computational phantoms were similar in most characteristics and reported results, the computational phantoms differed in a few important underlying assumptions. Researchers have derived a revised model14 that resolves these computational phantom differences in ways best supported by currently available data. New skeletal average AFs for all bone regions employed in the calculations in this study were implemented in the OLINDA/EXM10 computer code, designed as a successor to the MIRDOSE code.8

20.4 Currently Available DCFs For many years, the only source of DFs for use in practical calculations was to be found in the MIRD Pamphlet 11.5 In this publication, factors were given for about 25 organs, but

476

Handbook of Anatomical Models for Radiation Dosimetry

only in the adult male computational phantom for 117 radionuclides. The MIRDOSE code8 provided DFs for over 240 radionuclides and about 25 organs as well, and for the entire Cristy–Eckerman and Stabin et al. pediatric, adult, and pregnant female computational phantoms series (10 computational phantoms).7,9 Stabin and Siegel15 calculated DFs for over 800 radionuclides for 1. All source and target regions in the six computational phantoms in the Cristy– Eckerman computational phantom series7 2. All source and target regions in the four computational phantoms in the Stabin et al. pregnant female computational phantoms series9 3. All target regions in the Watson and Stabin peritoneal cavity computational phantom16 4. All target regions in the Stabin prostate gland computational phantom17 5. All source and target regions in the six computational phantoms of the MIRD head and brain computational phantom18 6. All source and target regions in the MIRD regional kidney computational phantom19 7. The unit density sphere computational phantoms of Stabin and Konijnenberg20 These DFs were based on decay data from the Brookhaven National Laboratory resource (http://www.nndc.bnl.gov/), and are useful tools for implementation in the dose equations described above. These DCFs use the child, adult, and pregnant woman computational phantoms and bone and marrow computational phantoms described above, and included standard modeling assumptions, as were described in that paper.15 20.4.1 Standard Organ Masses Cristy and Eckerman7 gave the masses of the target regions in the full-body computational phantoms, based on recommendations of the ICRP.3 The ICRP has recently recommended new standard organ masses,4 which currently uses them to design new computational phantoms for use in dose calculations. These masses, and all DFs, will be updated in future publications, as the data become available.

20.5 Standard Dose Estimates for Radiopharmaceuticals Many individuals and groups can calculate dose estimates for different radiopharmaceuticals by using the approaches and computational phantoms described above. It can sometimes be frustrating to some users to seek dosimetry information on a particular radiopharmaceutical and find several different sets of dose estimates, often with minor and sometimes significant discrepancies in the computational phantoms employed and the resulting doses calculated. Standardized dose calculations for a handful of radiopharmaceuticals (less than 20) were developed by the MIRD Committee over a 30 year period. Others were published at one time by the dosimetry information center in Oak Ridge.21

Applications to Nuclear Medicine

477

Most recently, a large compendium of dose estimates has been published by a working group of the ICRP for around 200 radiopharmaceuticals, based on the best known biodistribution data and using the standard computational phantoms described above.22 These values should be referenced in most cases in which standardized dose estimates for a particular radiopharmaceutical are needed. 20.5.1 Image-Based Approaches Several of the efforts to use image data to perform dose calculations move toward the idea of patient-specific dosimetry. Yoriyaz et al. provided DFs for a voxel-based anthropomorphic computational phantom.23 Clairand et al. provided variations of the standard adult anatomy in the DOSE3D code.24 Petoussi-Henss et al. created several realistic tomographic computational phantoms from CT images25 and found differences of several tens of percent for organ AFs from styled computational phantoms. Efforts to use subject-specific images with individual kinetic data include the 3D-ID code from the Memorial Sloan-Kettering Cancer Center,26 the SIMDOS code from the University of Lund,27 the RTDS code at the City of Hope Medical Center, 28 and the RMDP code from the Royal Marsden Hospital.29 The code with the most clinical experience to date is the 3D-ID code. These codes either rely on the standard geometrical phantoms (MABDose and DOSE3D) or patient-specific voxel phantom data (3DID and SIMDOS) and various in-house written routines to perform photon transport. Neither has a particularly robust or well-supported electron transport code, such as is available in EGS, 30 MCNP, 31 or GEANT.32 The PEREGRINE code33 has been proposed for three-dimensional (3D), computational dosimetry, and treatment planning in radioimmunotherapy. Oedipe, an IDL-based interface for patient-specific dose calculations carried out with MCNPX, has also been described.34 Many codes written by individual institutions assume that electron energy is absorbed wherever the electron is first produced. The development and support of electron transport methods are quite complex, as evidenced by ongoing intensive efforts by the EGS4, MCNP, and GEANT computer code working groups. It is not reasonable to expect in-house written codes to deal effectively with electron transport. In areas of highly nonuniform activity distribution, such as an organ with multiple tumors evidencing the enhanced uptake of an antibody, the explicit transport of both photons and electrons is needed to characterize dose distributions adequately.

20.6 Input Data for Dosimetry Calculations In the introduction to MIRD Pamphlet No. 16,35 the authors note that To determine the activity-time profile of the radioactivity in source regions, four questions need to be answered: 1) What regions are source regions? 2) How fast does the radioactivity accumulate in these source regions? 3) How long does the activity remain in the source regions? 4) How much activity is in the source regions?

478

Handbook of Anatomical Models for Radiation Dosimetry

The first question concerns identification of the source regions, while the second and third questions relate to the appropriate number of measurements to be made in the source regions as well as the timing of these measurements. The fourth question is addressed through quantitative external counting and/or sampling of tissues and excreta. Each source region must be identified and its uptake and retention of activity as a function of time must be determined. This provides the data required to calculate cumulated activity or residence time in all source regions. Each region exhibiting significant radionuclide uptake should be evaluated directly where possible. The remainder of the body (total body minus the source regions) must usually be considered as a potential source as well. Mathematical computational phantoms that describe the kinetic processes of a particular agent may be used to predict its behavior in regions where direct measurements are not possible, but where sufficient independent knowledge about the physiology of the region is available to specify its interrelationship with the regions or tissues whose uptake and retention can be measured directly. The statistical foundation of a data acquisition protocol designed for dosimetry requires that an adequate number of data points be obtained and that the timing of these points be carefully selected. As the number of measurements increases, the confidence in the fit to the data and in the estimates of unknown parameters in the computational phantom is improved. As a heuristic or general rule of thumb, at least as many data points should be obtained as the number of initially unknown variables in the mathematical curve-fitting function(s) or in the compartmental computational phantom applied to the data set. For example, each exponential term in a multiexponential curve-fitting function requires two data points to be adequately characterized. On the other hand, if it is known a priori that the activity retention in a region can be accurately represented by a monoexponential function, restrictions on sampling times are less stringent as long as enough data points are obtained to derive the fitted function. Because of problems inherent in the collection of patient data (e.g., patient motion, loss of specimen, etc.), the collection of data above the necessary minimum is advisable.” Gathering data for dosimetry requires characterization of the radiopharmaceutical kinetics, and the assignment of the image data to source regions defined in the human body computational phantoms for which DFs are defined. In principle, this must be done at the same resolution as that used for DF determination: if the dose calculation approach uses human body computational phantoms and organ-based dose calculations, then the cumulated activity should be assessed at the organ level. If the dose calculation approach allows for voxel-based dose calculation, however, activity should be assessed at the voxel level. Animal data may be extrapolated to humans to provide an initial estimate of dose, but ultimately human image data should be relied upon for the fi nal dose estimates. To determine the radiation dose from a human study, the counts detected in the field of view must be converted to the absolute values of activity (Bq or mCi), which require that a calibration factor for the camera must be known and that data that permit correction of the raw images are collected, i.e., for radiation attenuation and scatter. The variation of activity with time for voxel-based approaches is usually obtained via quantitative SPECT imaging. To date, positron emission tomography (PET) data have been little used for dosimetry, although PET quantification is an active area of research in its own right. PET is usually considered “more quantitative” than SPECT, due to a better spatial resolution and attenuation correction. The lack of suitable PET tracers currently restricts the application of PET-based biokinetics determination. Some studies have been presented within a context of pretherapy dose analyses, however, in which the kinetics of a beta emitter used for therapy were determined from PET studies carried out with a similar positron emitting radiopharmaceutical.36–38 This concept of beta “pairs” is likely to become more popular in the future, although current cumulated activity is usually determined via quantitative SPECT imaging.

Applications to Nuclear Medicine

479

Data acquisition parameters include the number and timing of scans. With allowances made for adequate statistics and spatial resolution, no special criteria are required for the number of projections or the matrix size, which will usually be performed according to standard local protocols. It is worth noting, however, that for many therapy scans, patients will have a relatively high level of activity, so camera sensitivity is not an issue. Data acquired soon after the administration may require short acquisition times, which will not impact greatly on the daily routine of the nuclear medicine department. Quantitative SPECT imaging can be performed in planar (2D static or whole-body) or in tomographic (3D) mode. However, gamma-cameras are not meant to provide direct quantitative information (i.e., activity or activity concentration). Many corrections have to be implemented in order to rely the detected events in the field of view to a quantitative index: 1. Deadtime corrections. Scintillation cameras are designed for use with low levels of Tc-99m and so are poorly adapted for use with high activities of, for example, I-131. A common problem with these systems is that of deadtime, whereby the counts registered do not increase linearly with the activity in the field of view. Both paralyzable and nonparalyzable systems may be characterized for their deadtime behavior so that correction factors may be obtained that can be directly applied to the image data.37 2. Attenuation correction. Attenuation correction is a crucial step in the quantification of emission data and a number of techniques have been used. In planar mode (both for static or whole-body), a transmission scan is usually acquired prior to the emission. This is at the basis of the geometric mean method described below. The most basic method employed in tomographic mode is that of Chang36 who assume that the imaged volume is uniformly filled with water. More complex solutions, that to date have seldom been used in clinical practice, involve adaptation of a patient CT scan to provide an attenuation map.39 3. Scatter correction. Scatter correction methods are generally performed by the application of scatter windows placed adjacent to the photopeak. Frequently, triple energy window (TEW) scatter correction is performed, as detailed above, although dual energy windows are also employed. A number of authors have explored the use of a large number of scatter windows40 or the acquisition of list mode data41 although limitations are often imposed by the system used. 4. Partial volume effect. Due to the overall poor spatial resolution of SPECT imaging, partial volume effect limits contrast recovery and absolute activity determination. Partial volume effect corrections have been proposed for tomographic scintigraphic imaging (both PET and SPECT), in general within a context of diagnostic for neurology. However, the relevance of partial volume effect correction fully applies to quantitative imaging for dosimetry and should be implemented in clinical studies.42,43

20.6.1 Quantification of Planar Data In planar imaging, the method most commonly used to obtain quantitative data for dosimetry is the external conjugate view counting pair (anterior/posterior). In this method, the source activity Aj is given by the expression:

Aj =

IAIP fj −μ t e e C

(20.7)

480

Handbook of Anatomical Models for Radiation Dosimetry

fj =

(μ jt j /2) sinh (μ jt j /2)

(20.8)

where IA and IP are the observed counts in the anterior and posterior projections (counts/ time) t is the overall patient thickness μe is the effective linear attenuation coefficient C is system calibration factor C (count rate per unit activity) the factor f represents a correction for the source region attenuation coefficient (μj) and source thickness (tj) (i.e., source self-attenuation correction) This expression assumes that the views are perfectly collimated (i.e., they are oriented toward each other without offset) under the computational phantom of narrow beam geometry without significant scattered radiation effects. Corrections for scatter are usually necessary, and a number of methods have been proposed. One relatively straightforward correction procedure for scatter compensation involves establishing adjacent windows on either side of the photopeak window, when the area of the two similar adjacent windows is equal to that of the photopeak. The corrected (true) photopeak counts CT are given by the expression: CT = Cpp − FS * (CLS + CUS )

(20.9)

where Cpp is the total count recorded within the photopeak window CLS and CUS are the counts within the lower and upper scatter windows, respectively If the areas of the scatter windows are not equal (in sum) to that of the photopeak window, then an appropriate scaling factor (FS) should be applied. Subtraction of the adjacent windows is assumed to compensate for the high-energy photon scatter tail upon which the true photopeak events are superimposed. Even if the areas of the scatter windows are equal to that of the photopeak window, the use of a scaling factor other than unity may provide the best correction for scatter in a given system with a particular radionuclide. This may be determined by the study of a known volume source in a water phantom whose dimensions are similar to that of a human subject. Other corrections are often required as well. Whenever an region of interest (ROI) is drawn over a source region on a projection image, some counts from the region will be from activity in the subject’s body that are outside of the identified source, including scattered radiation from other ROIs, background radiation, and other sources. The choice of locations and sizes of background ROIs is difficult to prescribe, and methods vary considerably between investigators, which can result in markedly different results for the final estimates of activity assigned to a source ROI. This process should be carried out with caution and attention to the above points for the best and most reproducible results. It is not uncommon for some organs or tumors to have overlapping regions on projection images. The right kidney and liver are frequently partially superimposed on such images, as are the left kidney and spleen, for example. When organ overlap occurs, an estimate of the total activity within a source can be obtained by a number of approximate methods. For paired organs, such as the kidneys and lungs, one approach is to quantify the activity in one of the organs for which there is no overlap with other organs

Applications to Nuclear Medicine

481

and multiply the number of counts in this organ by 2 to obtain the total counts in both organs. Another approach is to draw an ROI over the organ region in scans where there is overlap, count the number of pixels and note the average count rate per pixel, then use an ROI from another image in which there is no apparent overlap and the whole organ is clearly visible, count the number of pixels in a larger ROI drawn on this image, and then multiply the count rate per pixel from the first image by the number of pixels in the second image. Or, similarly, one would take the total number of counts in the first image and multiply by the ratio of the number of pixels in the second to the first image ROIs. If no image can be found in which a significant overlap with another organ does not obscure the organ boundaries, an approximate ROI may need to be drawn just from knowledge of the typical shapes of such organs. This kind of approximation is not ideal, but may be a necessary approximation. In addition, calibration coefficients for each radionuclide and gamma camera/collimator combination must be obtained by imaging a small source of known activity for a fixed amount of time. The attenuation characteristics of the camera may be studied by imaging this source with various known thicknesses of tissue-equivalent material interposed between the source and camera and fitting the results (counts versus thickness) to an exponential function.

20.7 Quantification of Tomographic Data Tomographic imaging offers the potential for improved dosimetric accuracy due to its increased contrast when compared with planar imaging. Tomographic data are particularly useful for dosimetry when heterogeneous uptake of activity in the source organ or under- or overlying background activity is suspected. Conversely, dosimetric approaches usually require that an accurate determination of the activity be carried out for the whole patient, a task that is difficult to perform in tomographic mode. Image reconstruction can be divided into filtered back projection and iterative techniques and can incorporate scatter and attenuation correction. Each of these techniques has a number of variables that may be employed, including smoothing parameters and the number of iterations. The effect on quantification of adjusting these parameters should be studied carefully on a case-specific basis. Conversion of counts to absolute values of activity may be performed in a number of ways. As specified above for the processing of planar data, it is possible to include a source of known activity within the field of view at the time of scanning. An approach used by some authors is to construct calibration computational phantoms that are aimed to emulate the patient. However, the ideal approach is to characterize the camera system so that, taking into account patient CT data, conversion factors may be obtained without recourse to computational phantoms or external sources. The quantification of image data has been considered for many years, although as yet there exist no standardized methods for quantifying SPECT data. If DFs are established for the organs of a reference anthropomorphic computational phantom, activity determination at the organ level is usually considered sufficient, and uncertainties in activity determination can be considered acceptable. However, with the appearance of new voxel-based computational dose calculation algorithms, activity determination has to be carried out at the voxel level as well. Quantitative scintigraphic imaging remains the largest single

482

Handbook of Anatomical Models for Radiation Dosimetry

obstacle to accurate dosimetry, and is currently a strong focus of research.27,44 It is probable that this task will be made easier with the advent of dual modality scanners and it is hoped that in time manufacturers will develop systems that are adapted to high energy high activity imaging, whereby camera sensitivity may be sacrificed to some extent in favor of spatial and energy resolution.

20.8 Status and Evolution of Dose Calculational Approaches Radiation dose calculations for nuclear medicine applications have been so far mostly considered as abstract and theoretical calculations, used to establish dosimetry for new agents and provide reasonable dose estimates to support radiopharmaceutical package inserts and for use in open literature publications. When patients are treated in therapy with radiopharmaceuticals, careful patient-specific optimization is often not performed. There are several reasons for this: 1. There are limitations on spatial resolution and accuracy of activity quantification with nuclear medicine cameras. Performing quantitative SPECT is not a trivial process, and the degree of confidence in activity estimates is seldom determined. 2. The realism and specificity to an individual patient of available body computational phantoms is limited. The computational phantoms described above were designed to represent the “reference” adult male and female, children, etc. Besides using geometric primitives to represent the body and its various organs, only one standardized computational phantom is available for any category of individual (e.g., an adult male). Therefore, the dose estimates calculated using this approach contain significant uncertainties when applied to any subject.45 3. Use of standardized dose calculations in therapy applications have produced reproducible results, but results which have generally correlated poorly with observed effects in patient populations. Physicians may understandably have low confidence in the relevance of such dosimetric estimates to plan individual subject therapy. Thus, unfortunately for the patients, a “one dose fits all” approach to therapy is usually employed, with significant caution resulting in administration of lower than optimum levels of activity to the majority of subjects. The use of image-based calculation frameworks, not only to develop new “reference” computational phantoms, but also to permit the use of patient-specific computational approaches for each therapy patient, is now developing quickly. The realism of the newer computational phantoms is shown in Figure 20.1, with comparison to the form of the existing computational phantoms developed and implemented in the historical MIRD system. Attempts to correlate the observed radiation effects with the radiation dose, particularly correlations of hematological toxicity with marrow dose, have not been particularly successful in many cases; due sometimes to uncertainties in the absorbed dose, but more often than not due to the difficulties in assessing each subject’s marrow functional status prior to therapy.46,47 Correlations of a number of marrow toxicity indices with marrow dose for 90Y labeled Zevalin, calculated using the reference adult computational phantom with over 150 subjects, were disappointing.48 This led to the approval of the compound

Applications to Nuclear Medicine

483

Brain

Pulmonary vessels

Abdominal vessels FIGURE 20.1 Comparison of the realism of the traditional MIRD body models with those being used to support current dose modeling efforts.

with no requirement for performing patient-individualized dose calculations. It is clear that the organ level dose calculations with standard, reference subjects will not produce dose calculations that will be of sufficiently high quality to be used in therapy planning. Characterization of patient-specific biokinetics, while necessary, is not sufficient. Use of the image-based, patient-specific dose modeling is needed to improve our understanding of observed radiation effects. Also, recent evidence indicates that the biologically effective dose (BED), not just the absorbed dose, is the parameter that should be characterized, in both internal and external dose calculations.49,50 Several investigators have shown recently that patient-specific modifications to standardized dose calculations can dramatically improve correlations between calculated dose and observed effects in tumors and normal tissues. Shen et al.51 using a 90Y-antibody in radioimmunotherapy, obtained an r value of 0.85 for correlation of marrow dose with observed marrow toxicity, using patient-specific marrow mass estimated from CT images and estimation of the total marrow mass from the mass of the marrow in three lumbar vertebrae. Siegel et al.52 obtained a correlation coefficient of 0.86 between platelet nadir and calculated marrow dose but with an ingenious modification based on the levels of a stimulatory cytokine (FLT3-L) measurable in peripheral blood that indicates the possible present status of a subject’s marrow, in the use of an 131I-131 anti cea antibody. While others have failed to find firm correlations between tumor dose and observed response, found a convincing relationship, in their study of 22 patients with 90Y Octreother. Kobe et al.53 evaluated the success of treatment of Graves’ disease in 571 subjects, with the goals of delivering 250 Gy to the thyroid and the end-point measure being the elimination of hyperthyroidism evaluated 12 months after the treatment. Relief from hyperthyroidism was achieved in 96% of patients who received more than 200 Gy, even for thyroid volumes >40 mL. Individually tailored patient thyroid dosimetry was made to the targeted total dose, with ultrasound measurement of subject thyroid mass and adjustment of the procedure to account for differences between observed effective retention half-times between studies involving the tracer activity and the therapy administration. These authors note

484

Handbook of Anatomical Models for Radiation Dosimetry

that success rates with more traditional treatments (not using individually tailored dosimetry) are typically at best 60%–80%. In conclusion, many sophisticated computational phantoms and methods are available for calculating the radiation dose to nuclear medicine patients. A convergence is occurring between sophisticated image-based modeling efforts and the use of standardized computational phantoms. Our understanding of radiation biology from internal emitters is essential to improvement of our understanding of dose/effect relationships in nuclear medicine therapy. It is important to convince physicians that such efforts are needed, as these relationships cannot be studied at all if all patients receive the same administration of activity, with no calculation of radiation dose. Providing better and more durable outcomes for cancer patients requires more aggressive and optimized therapy, which again is not possible without careful and accurate dosimetry. A paradigm shift is needed in the nuclear medicine clinic to accommodate these changes and improve patient therapy.

References 1. International Commission on Radiological Protection (ICRP). Limits for Intakes of Radionuclides by Workers, ICRP Publication 30, Pergamon Press, New York, 1979. 2. Snyder, W.S. et al. Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom, Journal of Nuclear Medicine, Suppl. 3, 7, 1969. 3. International Commission on Radiological Protection (ICRP). Report of the Task Group on Reference Man, ICRP Publication 23, Pergamon Press, Oxford, 1975. 4. International Commission on Radiological Protection (ICRP). Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, ICRP Publication 89, Pergamon Press, Oxford, 2003. 5. Snyder, W. et al. Absorbed dose per unit cumulated activity for selected radio nuclides and organs, MIRD Pamphlet No. 11, Society of Nuclear Medicine, New York, 1975. 6. Snyder, W.S., Ford, M.R., and Warner, G.G. Estimates of specific absorbed fractions for photon sources uniformly distributed in various organs of a heterogeneous phantom, MIRD Pamphlet No. 5, Revised, Society of Nuclear Medicine, New York, 1978. 7. Cristy, M. and Eckerman, K.F. Specific absorbed fractions of energy at various ages from internal photon sources, ORNL/TM-8381 V1-V7, Oak Ridge National Laboratory, Oak Ridge, TN, 1987. 8. Stabin, M.G. MIRDOSE: Personal computer software for internal dose assessment in nuclear medicine, Journal of Nuclear Medicine, 37, 538, 1996. 9. Stabin, M.G. et al. Mathematical models and specific absorbed fractions of photon energy in the nonpregnant adult female and at the end of each trimester of pregnancy, ORNL/TM-12907, Oak Ridge National Laboratory, Oak Ridge, TN, 1995. 10. Stabin, M.G., Sparks, R.B., and Crowe, E. OLINDA/EXM: The second-generation personal computer software for internal dose assessment in nuclear medicine, Journal of Nuclear Medicine, 46, 1023, 2005. 11. Spiers, F.W. Beta dosimetry in trabecular bone, in Delayed Effects of Bone-Seeking Radionuclides, Mays, C.W. (ed.), University of Utah Press, Salt Lake City, UT, 1969. 12. Eckerman, K.F. and Stabin, M.G. Electron absorbed fractions and dose conversion factors for marrow and bone by skeletal regions, Health Physics, 78, 199, 2000. 13. Bouchet, L.G. et al. S values for radionuclides localized within the skeleton, Journal of Nuclear Medicine, 41, 189, 2000.

Applications to Nuclear Medicine

485

14. Stabin, M.G. et al. Evolution and status of bone and marrow dose models, Cancer Biotherapy and Radiopharmaceuticals, 17, 427, 2002. 15. Stabin, M.G. and Siegel, J.A. Physical models and dose factors for use in internal dose assessment, Health Physics, 85, 294, 2003. 16. Watson, E.E. et al. A model of the peritoneal-cavity for use in internal dosimetry, Journal of Nuclear Medicine, 30, 2002, 1989. 17. Stabin, M.G. A model of the prostate-gland for use in internal dosimetry, Journal of Nuclear Medicine, 35, 516, 1994. 18. Bouchet, L.G. et al. MIRD pamphlet no. 15: Radionuclide S values in a revised dosimetric model of the adult head and brain, Journal of Nuclear Medicine, 40, 62S, 1999. 19. Bouchet, L.G. et al. MIRD pamphlet no. 19: Absorbed fractions and radionuclide S values for six age-dependent multiregion models of the kidney, Journal of Nuclear Medicine, 44, 1113, 2003. 20. Stabin, M.G. and Konijnenberg, M.W. Re-evaluation of absorbed fractions for photons and electrons in spheres of various sizes, Journal of Nuclear Medicine, 41, 149, 2000. 21. Stabin, M.G., Stubbs, J.B., and Toohey, R.E. Radiation dose estimates for radiopharmaceuticals, NUREG/CR-6345, Prepare for: US Nuclear Regulatory Commission, US Department of Energy, US Department of Health & Human Services, 1996. 22. ICRP. Radiation Dose Estimates for Radiopharmaceuticals, ICRP Publications 53 and 80, with addenda, Pergamon Press, Oxford, 1983–1991. 23. Yoriyaz, H., Stabin, M.G., and dos Santos, A. Monte Carlo MCNP-4B-based absorbed dose distribution estimates for patient-specific dosimetry, Journal of Nuclear Medicine, 42, 662, 2001. 24. Clairand, I. et al. DOSE3D: EGS4 Monte Carlo code-based software for internal radionuclide dosimetry, Journal of Nuclear Medicine, 40, 1517, 1999. 25. Petoussi-Henss, N. et al. The GSF family of voxel phantoms, Physics in Medicine and Biology, 47, 89, 2002. 26. Kolbert, K.S. et al. Implementation and evaluation of patient-specific three-dimensional internal dosimetry, Journal of Nuclear Medicine, 38, 301, 1997. 27. Dewaraja, Y.K. et al. Accurate dosimetry in I-131 radionuclide therapy using patient-specific, 3-dimensional methods for SPECT reconstruction and absorbed dose calculation, Journal of Nuclear Medicine, 46, 840, 2005. 28. Liu, A. et al. A radionuclide therapy treatment planning and dose estimation system, Journal of Nuclear Medicine, 40, 1151, 1999. 29. Guy, M.J. et al. RMDP: A dedicated package for I-131 SPECT quantification, registration and patient-specific dosimetry, Cancer Biotherapy and Radiopharmaceuticals, 18, 61, 2003. 30. Bielajew, A.F. and Rogers, D.W.O. Presta—The parameter reduced electron-step transport algorithm for electron Monte-Carlo transport, Nuclear Instruments & Methods in Physics Research Section B-Beam Interactions with Materials and Atoms, 18, 165, 1987. 31. Briesmeister, J.F. MCNP—A General Monte Carlo N-Particle Transport Code, Version 4C, LA-13709-M, Los Alamos National Laboratory, 2000. 32. Allison, J. et al. Geant4 developments and applications, IEEE Transactions on Nuclear Science, 53, 270, 2006. 33. Lehmann, J. et al. Monte Carlo treatment planning for molecular targeted radiotherapy within the MINERVA system, Physics in Medicine and Biology, 50, 947, 2005. 34. Chiavassa, S. et al. Validation of a personalized dosimetric evaluation tool (Oedipe) for targeted radiotherapy based on the Monte Carlo MCNPX code, Physics in Medicine and Biology, 51, 601, 2006. 35. Siegel, J.A. et al. MIRD pamphlet no. 16: Techniques for quantitative radiopharmaceutical biodistribution data acquisition and analysis for use in human radiation dose estimates, Journal of Nuclear Medicine, 40, 37S, 1999. 36. Chang, L.T. Method for attenuation correction in radionuclide computed tomography, IEEE Transactions on Nuclear Science, 25, 638, 1978.

486

Handbook of Anatomical Models for Radiation Dosimetry

37. Delpon, G. et al. Correction of count losses due to deadtime on a DST-XLi (SMVi-GE) camera during dosimetric studies in patients injected with iodine-131, Physics in Medicine and Biology, 47, N79, 2002. 38. Sgouros, G. et al. Patient-specific dosimetry for I-131 thyroid cancer therapy using I-124 PET and 3-dimensional-internal dosimetry (3D-ID) software, Journal of Nuclear Medicine, 45, 1366, 2004. 39. Fleming, J.S. A technique for using CT images in attenuation correction and quantification in SPECT, Nuclear Medicine Communications, 10, 83, 1989. 40. Buvat, I. et al. Comparative-assessment of 9 scatter correction methods based on spectralanalysis using Monte-Carlo simulations, Journal of Nuclear Medicine, 36, 1476, 1995. 41. Guy, M.J. et al. DETECT—Dual energy transmission estimation CT—for improved attenuation correction in SPECT and PET, IEEE Transactions on Nuclear Science, 45, 1261, 1998. 42. Du, Y., Tsui, B.M.W., and Frey, E.C. Partial volume effect compensation for quantitative brain SPECT imaging, IEEE Transactions on Medical Imaging, 24, 969, 2005. 43. Rousset, O.G., Ma, Y.L., and Evans, A.C. Correction for partial volume effects in PET: Principle and validation, Journal of Nuclear Medicine, 39, 904, 1998. 44. Autret, D. et al. Monte Carlo modeling of gamma cameras for I-131 imaging in targeted radiotherapy, Cancer Biotherapy and Radiopharmaceuticals, 20, 77, 2005. 45. Stabin, M.G. Uncertainties in internal dose calculations for radiopharmaceuticals, Journal of Nuclear Medicine, 49, 853, 2008. 46. DeNardo, D.A. et al. Imaging for improved prediction of myelotoxicity after radioimmunotherapy, Cancer, 80, 2558, 1997. 47. Juweid, M.E. et al. Prediction of hematologic toxicity after radioimmunotherapy with I-131labeled anticarcinoembryonic antigen monoclonal antibodies, Journal of Nuclear Medicine, 40, 1609, 1999. 48. Wiseman, G.A. et al. Radiation dosimetry results and safety correlations from Y-90-ibritumomab tiuxetan radioimmunotherapy for relapsed or refractory non-Hodgkin’s lymphoma: Combined data from 4 clinical trials, Journal of Nuclear Medicine, 44, 465, 2003. 49. Barone, R. et al. Patient-specific dosimetry in predicting renal toxicity with Y-90-DOTATOC: Relevance of kidney volume and dose rate in finding a dose-effect relationship, Journal of Nuclear Medicine, 46, 99S, 2005. 50. Dale, R. and Carabe-Fernandez, A. The radiobiology of conventional radiotherapy and its application to radionuclide therapy, Cancer Biotherapy and Radiopharmaceuticals, 20, 47, 2005. 51. Shen, S. et al. Improved prediction of myelotoxicity using a patient-specific imaging dose estimate for non-marrow-targeting Y-90-antibody therapy, Journal of Nuclear Medicine, 43, 1245, 2002. 52. Siegel, J.A. et al. Red marrow radiation dose adjustment using plasma FLT3-L cytokine levels: Improved correlations between hematologic toxicity and bone marrow dose for radioimmunotherapy patients, Journal of Nuclear Medicine, 44, 67, 2003. 53. Kobe, C. et al. Graves’ disease and radioiodine therapy—Is success of ablation dependent on the achieved dose above 200 Gy?, Nuklearmedizin-Nuclear Medicine, 47, 13, 2008.

21 Applications to Computed Tomography for Pediatric Patients Wesley E. Bolch, Choonsik Lee, Choonik Lee, Jorge Hurtado, and Jonathan L. Williams

CONTENTS 21.1 Introduction ............................................................................................................... 487 21.2 Past Studies on CT Dosimetry ................................................................................ 488 21.2.1 Physical Phantoms ....................................................................................... 488 21.2.2 Computational Human Phantoms ............................................................ 489 21.2.3 Voxel Phantoms ............................................................................................ 489 21.3 CT Dosimetry Study Using Series of UF Voxel Phantoms ................................. 490 21.3.1 Material and Methods ................................................................................. 490 21.3.2 Results and Discussion ............................................................................... 493 21.4 CT Dosimetry Study Using UF Hybrid Pediatric Phantoms..............................504 21.4.1 Material and Methods .................................................................................504 21.4.2 Results and Discussion ...............................................................................504 21.5 Conclusions and Future Research Needs ..............................................................508 Acknowledgment................................................................................................................. 509 References ............................................................................................................................. 509

21.1 Introduction Recent statistics compiled by the National Council on Radiation Protection and Measurements (NCRP) indicate that presently computed tomography (CT) accounts for only 12% of all radiological diagnostic examinations in the United States, yet it contributes upwards of 45% of the collective effective dose on an annual basis. While children constitute a small fraction of these examinations, they are undoubtedly at a higher risk following CT radiation exposure. Pediatric patients are more susceptible to the radiation-induced risks than are adult patients, owing to their more rapidly growing tissues, their wider and increased cellular distribution of red bone marrow, and their greater post exposure life expectancy.1–3 Consequently, even though pediatric patients only constitute a small fraction of medical radiation exposures, they are arguably at higher risk from potential radiation effects. With the introduction of helical CT in 1990 and multidetector CT scanners in 1998, a dramatic increase in the use of CT in the diagnosis of a variety of pathologic conditions has emerged. The annual number of CT examinations has increased from 3.6 million in 1980

487

488

Handbook of Anatomical Models for Radiation Dosimetry

to 33 million in 1998.4,5 According to the survey by Mettler et al.,6 approximately 10% of all CT examinations are performed in pediatric patients in the United States. Furthermore, this imaging modality accounts for about 67% of the patient collective dose.2,7 Medical practices involving a significant amount of radiation may be justified by the diagnostic benefits they provide to individual patients. As more radiological studies are utilized for patient health care, it becomes increasing important to accurately understand and to assess risks imposed by CT examination, especially in pediatric patients. In this chapter, we summarize past studies of CT organ dosimetry in children using both stylized (equation-based) and voxel (image-based) computational phantoms. Next, we discuss the use of the University of Florida (UF) pediatric series of Non-Uniform Rational B-Spline (NURBS) based hybrid phantoms, outlined in Chapter 9, in the assessment of modern-day multislice CT scanners. Using hybrid phantoms, one is able to match the patient body morphometry according to independent adjustments for the trunk height (shown to correlate strongly with internal organ volumes in the chest and abdomen), the total body height (through adjustments in leg length), and the total body mass (through adjustments in outer body contour defining the patients subcutaneous fat and muscle mass and body profile). The later techniques thus enable a high degree of patient specificity in phantom-to-patient matching, enabling a higher degree of accuracy in CT organ dose assessment than is capable with current age-matched 50th percentile reference phantoms. Applications of patient-sculpted phantoms are demonstrated as relevant to modeling mAmodulation techniques of modern CT scanners.

21.2 Past Studies on CT Dosimetry 21.2.1 Physical Phantoms The effective dose has been widely used to assess relative patient risk as it accounts for the absorbed dose and relative radiosensitivity of irradiated organs within the patient.8 Only a limited number of studies, however, have been performed in estimating organ doses and corresponding effective doses for pediatric patients in CT. In many of these studies, simple cylindrical physical phantoms of varying sizes have been used to estimate organ dose as a function of patient age in CT. Shrimpton and Wall9 used a 16 cm diameter polymethylmethacrylate (PMMA) phantom to derive reference dose indices, such as the weighted CT dose index (CTDIw) for single-slice scanning and the dose length product (DLP) for multiple-slice scanning. These dose indices are merely relative reference values for users to compare with their own measurements, and are not adequate for the determination of individual patient organ doses. These authors thus extended their efforts to calculate the effective dose using DLP measurement by utilizing normalization factors derived from simulation data taken from an adult anthropomorphic-stylized phantoms.10 Boone et al.11 evaluated size-dependent technique factors, including varying the tube currents and voltages, by using phantoms that ranged from 10 to 32 cm in diameter and reporting CT techniques that allowed constant image quality and reduced radiation dose in pediatric patients. Nickoloff et al.12 analyzed the effect of phantom size, tube voltage, tube current, and scanner type on the CTDI and found that CTDI is an exponential function of phantom diameter, and that it increases in phantoms with smaller diameters. Siegel et al.13 performed a similar study by using a wider array of physical phantoms (8, 16, 24, and 32 cm in diameters) for abdominal CT studies.

Applications to Computed Tomography for Pediatric Patients

489

Again, even though these studies suggest that the relative absorbed dose changes using the ion chamber inside of the cylinder physical phantoms, the measurement results cannot address real organ-absorbed doses or effective doses to pediatric patients. For organ dose evaluation, one must use either physical or computational anthropomorphic phantoms. Experimental measurements of organ doses using the anthropomorphic physical phantoms have been made by a few authors.14,15 Axelsson et al.14 used an anthropomorphic physical phantom of 1 year child16 and multiple thermoluminescent dosimeters (TLDs) to estimate organ doses for head exams under axial CT and in the lower trunk under spiral CT as performed using a General Electric HiSpeed advantage scanner operated at 120 kVp. The authors pointed out that the measurements can involve significant uncertainties, especially for the organs with partial exposure to the x-ray field. Chapple et al.15 used the whole series of pediatric physical phantoms of Varche,17 along with internally placed TLDs, to estimate organ and effective doses from the Siemens Somatom Plus 4 CT scanner used for head, chest, abdomen, and pelvis examinations. They also provided the conversion coefficients to translate the DLP measurement into estimates of effective dose for different phantom sizes. 21.2.2 Computational Human Phantoms A final technique for organ dose is to use Monte Carlo radiation transport methods with computational phantoms of pediatric patient anatomy. This method is one of the more accurate and versatile ones available. Several authors have conducted this type of approach to assess organ doses under CT examination.18–21 The Monte Carlo simulation of anthropomorphic phantoms can avoid the problems we see in physical phantom measurement, as one can assign detailed elemental compositions to each organ, and can accurately calculate the averaged absorbed doses to organs of interest as long as the phantom itself represents realistic human anatomy both internally and externally. Khurseed et al.21 utilized a series of pediatric-stylized phantoms to estimate effective dose in CT examinations. Phantoms of the newborn, 1, 5, 10 year child, and 15 year teenager from the Oak Ridge National Laboratory (ORNL) series10 were used to estimate the effective dose for three different CT machines at one tube voltage. The use of these stylized phantoms, however, can result in considerable discrepancies in organ dose, owing to the unrealistic torso shape and internal organ locations.22 In their study, the degree of the inaccuracies caused by the rather simple anatomy of the stylized phantoms was only partially addressed where the change in the esophagus geometry resulted in an absorbed dose estimate 12 times higher than previously reported.23 Furthermore, only axial scanning was modeled in this particular study and the phantom arms (which are inclusive of the trunk regions of the stylized phantoms) were not removed to simulate an arm-raised position as in normal clinical practice. 21.2.3 Voxel Phantoms Other studies using anthropomorphic computational phantoms were performed using more anatomically realistic voxel phantoms as described in Chapter 1.18–20 Voxel phantoms are constructed from the tomographic images of real patients and represent human anatomy in a more realistic fashion than permitted in the stylized phantoms. Excluding the UF pediatric phantoms of the present study, only three pediatric voxel phantoms have been utilized for the CT dosimetry: the BABY,18 CHILD,18 and ADELAIDE19,20 voxel phantoms.

490

Handbook of Anatomical Models for Radiation Dosimetry

Each of these phantoms is specific to the patient originally imaged, and in the case of the ADELAIDE phantom, only the torso region is available for dosimetry study. In each study, multiple axial rotations at varying positions were simulated, and thus the helical motion of the x-ray source was not explicitly considered. Jarry et al.24 published a study on the Monte Carlo method for estimating organ doses in adult patients undergoing either axial or helical CT examination. Organ doses were determined using both the adult-stylized ORNL phantom and a simple partial-body voxel phantom. In their study, the movement of the source was modeled explicitly, while the CT x-ray source profile was modeled using published information on the x-ray spectra with additional information provided by the manufacturer. Their work was the first Monte Carlo study where the helical movement of the x-ray source was explicitly modeled, along with detailed considerations of beam attenuation through the shaping filter. Staton et al.25 has recently used both a stylized and voxel phantoms of newborn patient to calculate the absorbed dose to organ and effective dose for multislice helical CT scanning used the modeling techniques of the Jarry et al.24 study.

21.3 CT Dosimetry Study Using Series of UF Voxel Phantoms 21.3.1 Material and Methods We use two classes of pediatric computational phantoms to quantitatively assess both individual organ doses and effective doses in pediatric patients undergoing helical multislice (16 detector) CT examinations of the head, chest, abdomen, pelvis, and torso. Trends in organ and effective dose for the pediatric patients are investigated as a function of peak tube potential, anatomical scan area, collimator setting, and patient anatomical phantom. The techniques used are similar to those used by Jarry et al.24 in their study of adult helical MSCT dosimetry, and are specifically applied to the Siemens SOMATOM Sensation 16 helical multislice CT scanner using manufacturer-provided information on x-ray spectra and beam shaping filtration. In a recent study by Lee et al.,26 the voxel phantoms of the UF Series B phantoms were used: a 9 month male, 4 year female, 8 year female, 11 year male, and 14 year male.27 The UF Series B phantoms were developed from their predecessor UF Series A phantoms28 which were in turn constructed through image segmentation of head and chest–abdomen–pelvis (CAP) CT scans of patients examined at Shands Children’s Hospital in Gainesville, FL. The UF Series A phantoms thus include patient-specific organ masses and body dimensions and, as the extremities are not typically within the scan coverage, these phantoms are of the head and torso only. The Series A phantoms were later extensively altered in the creation of the UF Series B phantoms29 through in inclusion of arms and legs taken as scaled segmented images from an adult CT-based phantom29 and through the reshaping of internal organs to more closely align them with age-interpolated ICRP Publication 89 reference masses.30 The arms of UF Series B phantoms were placed at the side of each phantom. In Lee et al.,26 the arms were once again removed to properly simulate their positioning during pediatric CT examination. In so doing, the clavicles, scapulae, humeri, and surrounding soft tissues of UF Series B were further remodeled to better simulate their position during upward arm extension—changes which were necessary to avoid dosimetric errors in organs such as the lungs, esophagus, thymus, and heart. Figure 21.1 shows the outer appearance of the modified and armless UF Series B pediatric phantom series.

Applications to Computed Tomography for Pediatric Patients

9 month male 4 year female

8 year female

11 year male

491

14 year male

FIGURE 21.1 UF Series B tomographic phantoms with removal of the arms for positioning under pediatric CT examination. (From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission.)

We also employed the pediatric phantoms from the ORNL series (newborn, 1, 5, 10, and 15 year) in this study. We generated the MCNPX inputs for the ORNL phantoms using the BodyBuilder 1.3 software (White Rock Science, White Rock, NM) as prepared with the hermaphrodite and non torso fat options. Figure 21.2 shows the outer appearance of the ORNL phantom series generated using the Bodybuilder software. As in the study of Khursheed et al.,21 we left the arms of the ORNL phantoms as is within the torso at each age so that dosimetric comparisons can be made that would be consistent with the practice of past studies. Furthermore, the BodyBuilder software does not permit an armless option, and thus users of the ORNL phantoms will not typically remove them. A SOMATOM Sensation 16 helical multislice CT scanner (Siemens Medical Solutions, Erlangen, Germany) was simulated within the MCNPX Monte Carlo radiation transport code.31 A source subroutine was written to allow the simulation of axial and variable detector pitch helical scans. The source subroutine generates the starting spatial coordinates and directional vectors for each simulated x-ray photon. The CT x-ray source was modeled as a fan beam originating from the focal spot with a fan beam angle of 52° and a focal spotto-axis distance of 57 cm. The helical path of the source was explicitly modeled based on the selection of the collimator setting, the detector pitch, and the scan length. The spatial coordinates of each photon were sampled randomly and uniformly over the helical path, and the direction of the photon was then randomly sampled within the fan beam. As such, mA modulation was not explicitly modeled in the Lee et al. study. As described in Staton et al.,25 radiographic film measurements of the beam profile and FWHM values as a function of focal spot size, body region filter, and kVp were acquired for this particular scanner. From this data, simulated beam widths of 14.7 and 26.8 mm were used in the Monte Carlo simulations for CT collimator settings of 12 and 24 mm, respectively, as used in pediatric imaging at Shands Children’s Hospital.

Handbook of Anatomical Models for Radiation Dosimetry

492

Newborn

1 year

5 years

10 years

15 years

FIGURE 21.2 ORNL phantom series with arm bones in place within the ellipsoidal torso region. (From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission.)

We obtained the CT x-ray energy spectra from the manufacturer for all peak tube potentials. We then implemented the energy spectra data in the Monte Carlo codes as lookup tables at intervals of 1 keV. The effects of the additional bowtie filtration were also included in the simulations. The manufacturer supplied information on the bowtie filter, including its dimensions, location, and composition. We did not simulate the radiation transport through the filter explicitly. Instead, we used a method described by Jarry et al.24 to calculate weighting factors that could be applied to source photon sampling. We calculated the pathlength through the bowtie filter for all angles within the fan beam based on the filter shape. We then generated a lookup table to correlate angles within the fan beam to pathlengths through the filter, which was then used along with the linear attenuation coefficient in the filter material for each photon energy, to calculate the exponential attenuation within the filter. We then used the attenuation value through the filter as a sourceweighting factor for each x-ray photon during the radiation transport simulations. For the present study, we selected MCNPX version 2.5.0 for radiation transport simulations. The MCNPX version 2.5.0 is capable of simulating both the stylized and voxel phantoms in a computationally efficient manner. The source subroutine was recompiled using the Compaq Visual Fortran compiler version 6.6B and Microsoft Visual C++. Each phantom was implemented into the MCNPX input deck with material cards that were specifically prepared for each age and tissue type according to ICRU Publication 4632 elemental compositions and densities. We simulated a variety of scan parameters to evaluate their influence on individual organ doses and effective doses within pediatric patients. Five different CT scan regions were simulated, including the head, chest, abdomen, pelvis, and CAP. For each CT examination, three tube potentials commonly used in pediatric CT scans—80, 100, and 120 kVp—were simulated at the two of the most commonly used collimator settings: 12 mm (16 detector

Applications to Computed Tomography for Pediatric Patients

493

array at 0.75 mm each) and 24 mm (16 detector array at 1.50 mm each). For each tube potential, two different x-ray spectra were utilized: one for the head CT examination, and one for the remaining examinations of the chest, abdomen, pelvis, and torso (CAP scan). A single detector pitch of 1.0 was selected for the simulations as the absorbed doses to internal organs are known to vary inversely with detector pitch. When using Monte Carlo radiation transport codes, tally outputs for an organ dose are given in the units of absorbed dose (mGy) per launched photon. In order to calculate the absolute organ-absorbed dose, researchers must create normalization factors to relate simulated dose values with experimental dose measurements; these factors (NF) are calculated using CTDI100 in-air simulations and ion chamber measurements.25 Shown in the following equation, NF is in units of photons per mAs and is calculated as a function of both peak tube potential kVp, and collimator setting C: NFkVp, C =

(K air-M )kVp, C (K air-S )kVp, C

(21.1)

In Equation 21.1, Kair-M is the air kerma measured free-in-air with the 10 cm ion chamber at the center of rotation for a given kVp and collimator setting C normalized to the mAs used in the measurement, which in this case was 100 mAs. Kair-M is therefore given in units of mGy per mAs. Similarly, Kair-S is air kerma given by the Monte Carlo simulation of ion chamber for the same measurement setup, and is given in units of mGy per photon history. Using the NF, the absolute tissue-absorbed dose, DAT, is thus calculated as

(D TA )kVp,C = (D TS )kVp,C × NFkVp,C × ⎛⎜⎝

mAs ⎞ ⎟ × (N rotation ) rotation ⎠

(21.2)

where DST is the Monte Carlo simulation estimate of the tissue-absorbed dose NFkVp,C is the normalization factor for a given kVp and collimator setting C, with the total mAs given as the product of the mAs per rotation and the total number of rotations in the scan, Nrotation For skeletal sites, energy-dependent volume-averaged photons fluences were calculated to assess the absorbed dose to the active marrow through the use of fluence-to-dose conversion factors published previously by Cristy and Eckerman.10 The total bone marrow dose was calculated by using a weighted average based on bone marrow doses to individual bones using published values of active bone marrow distribution within the pediatric skeleton.33 For calculations of the absorbed dose to the skeletal endosteum, a mass-weighted average dose to all the homogenous bone regions was used as a surrogate tissue.34 21.3.2 Results and Discussion The effective dose as derived within the two phantom types, the ORNL and UF Series B pediatric computational phantoms, was calculated for head, chest, abdomen, pelvis, and CAP CT examinations. Since the ORNL phantoms are hermaphrodites, having both male and female specific organs, separate male and female effective doses were calculated for the comparison against the gender-specific UF pediatric phantoms. As such, we calculated male effective doses using only equivalent doses of the male organs, while female effective

Handbook of Anatomical Models for Radiation Dosimetry

494

doses were evaluated using those of only the female organs. Figure 21.3a through e displays the effective dose at 80, 100, and 120 kVp tube potentials and for a 12 mm collimator setting in head, chest, abdomen, pelvis, and CAP examinations, respectively. In comparing pairs of figures for each CT examination, only small differences in the effective dose are seen between the two collimator settings. Two competing effects take

Head CT scan at 12 mm collimation

Normalized effective dose (mSv/100 mAs)

1.8

ORNL male phantoms ORNL female phantoms UF male phantoms UF female phantoms

1.6 1.4 1.2 1.0 0.8

120 kVp

0.6

100 kVp

0.4

80 kVp

0.2 0.0

0

2

(a)

Normalized effective dose (mSv/100 mAs)

14

16

18

Chest CT scan at 12 mm collimation

9.0

ORNL male phantoms ORNL female phantoms UF male phantoms UF female phantoms

8.0 7.0 6.0 5.0 4.0

120 kVp

3.0 100 kVp

2.0 1.0 0.0

(b)

12 10 4 6 8 Age of reference patient (years)

80 kVp 0

2

10 12 4 6 8 Age of reference patient (years)

14

16

18

FIGURE 21.3 Age-dependent effective doses for the ORNL phantoms (dashed lines) and the UF tomographic phantoms (closed circles—male, and open triangles—female) for (a) head, (b) chest, (c) abdominal, (d) pelvis, and (e) CAP CT examinations at 80, 100, and 120 kVp for collimator settings of (a) 12 mm and (b) 24 mm. (From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission.)

Applications to Computed Tomography for Pediatric Patients

495

place with a change in the collimator setting for scans of equal length. Firstly, since the penumbra is of constant magnitude for 12 and 24 mm collimator setting, the use of narrow collimation (12 mm) corresponds to a greater contribution from the penumbra to the total dose. Secondly, an additional half rotation is added on to each end of the helical scans. When the collimator setting increases, a larger additional irradiated region occurs at each edge of the scan due to the additional half rotation. Therefore, the absorbed doses for

Abdominal CT scan at 12 mm collimation

6.0

ORNL male phantoms ORNL female phantoms UF male phantoms UF female phantoms

Normalized effective dose (mSv/100 mAs)

5.5 5.0 4.5 4.0 3.5 3.0

120 kVp

2.5 2.0

100 kVp

1.5 1.0 80 kVp

0.5 0.0

0

2

(c)

10 12 4 6 8 Age of reference patient (years)

14

16

18

Pelvic CT scan at 12 mm collimation

Normalized effective dose (mSv/100 mAs)

6.0 ORNL male phantoms ORNL female phantoms UF male phantoms UF female phantoms

5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0

(d)

0

2

4

6

8

10

12

14

16

18

Age of reference patient (years)

FIGURE 21.3 (continued) (continued)

Handbook of Anatomical Models for Radiation Dosimetry

496

CAP CT scan at 12 mm collimation

Normalized effective dose (mSv/100 mAs)

16.0 ORNL male phantoms ORNL female phantoms UF male phantoms UF female phantoms

14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0 0

(e)

2

4 6 8 10 Age of reference patient (years)

12

14

16

FIGURE 21.3 (continued)

organs that fall on or near the edge of the scan boundary can be affected with the resulting change in collimator setting. The first effect will dominate for longer scans, while the second effect dominates for shorter scans. Furthermore, the impact on the effective dose is also dependent on which organs are nearest the edge of the scan. The change of the tube potential causes significant changes in the effective dose. In general, for both phantom types and all phantom ages, the patient effective dose at 100 kVp tube potential is ∼105% higher than seen at 80 kVp for the same tube current, while effective doses at 120 kVp deliver ∼210% higher effective doses than that seen at 80 kVp for the same tube current regardless of the collimator setting. These percent increases are consistent with CTDI100 values measured by other authors using PMMA phantoms.13 For all examinations exclusive of the head CT scans, noticeable differences in the effective dose are observed between the ORNL male and female phantoms. For chest and abdomen CT exams, where the female breasts are within the x-ray beam, the female ORNL patient phantoms receive overall higher effective doses than seen in the male ORNL phantoms. For pelvic exams, however, the external gonads of the male ORNL phantoms play an important role in increasing the effective dose over that seen in the female phantoms. As the UF phantoms do not, at present, have male and female phantoms of the same age, similar trends cannot be assessed in these voxel-based anatomic computational phantoms. The head CT exams do not show significant effective dose differences between the two phantom types. The UF phantoms, however, indicate lower effective doses for the chest exam and higher effective doses in the abdominal exam over those seen for the ORNL phantoms in both genders. For instance, in the chest exam, the effective dose of the UF 4 year phantom is 17% lower than the female effective dose of the ORNL 5 year phantom at a tube voltage of 120 kVp. These discrepancies are reversed for the abdominal CT exams, where the UF phantoms receive higher effective doses than seen in the ORNL phantoms. The effective dose for the UF 9 month phantom is 29% higher than the male effective dose for the ORNL 1 year phantom. These overestimates or underestimates in the effective dose in the chest and abdominal CT exams are due to less realistic torso thicknesses

Applications to Computed Tomography for Pediatric Patients

497

in the ORNL phantoms, which have torsos described by elliptical cylinders of constant thickness. In reality, the chest region is thicker than in the abdominal region of the patient, especially with their arms upwardly extended. Consequently, pediatric patients experience more internal organ shielding in chest exams and less so in the abdominal exams than seen in the ORNL patient phantoms. Obviously, these anatomical features are more realistically reflected in the UF voxel phantom series. For pelvic and CAP exams, there are no significant differences in the effective dose seen between two phantoms types, except for the female effective dose in the pelvic exams. The UF 4 year female receives 20% higher female effective dose than seen in the ORNL 5 year phantom. The cause of this difference is also found in the thinner torso of the UF phantom, where tissue shielding of the ovaries is more realistically simulated. Overall, the effective dose differences between the different phantom types increase at the higher tube potentials. For a given examination, the effective dose value was found to scale proportionally with the CTDI100 value for the kVp and collimator setting used in the scan. These trends are shown graphically in Figure 21.4 for head and CAP examinations for the youngest (9 month) and oldest (14 year) phantoms in the current UF series. Even though the effective dose provides useful information regarding the relative radiation risks to pediatric patients at different CT technique factors, it is also important to quantify individual organ-absorbed doses. Data on individual organ doses are essential for the construction of dose–response relationships in any radiation epidemiological studies of medically exposed patients. Furthermore, the values of the tissue weighting factors recommended by the ICRP are subject to change over time, and thus explicit and direct knowledge of individual organ-absorbed doses can be far more important than the rough estimates of the effective dose itself. Finally, it is important to realize that even though values of the effective dose are found to be in reasonable agreement between the UF and ORNL phantoms for several of the CT technique factors considered in this study, 16 Head scan 12 mm (9 months) Head scan 12 mm (14 years) CAP scan 12 mm (9 months) CAP scan 12 mm (14 years)

14

Effective dose (mSv)

12 10 8 6 4 2 0

0

5

10

15

20

CTDI100 (mGy) FIGURE 21.4 Variation of effective dose with CTDI100 for helical MSCT examinations of the head and CAP and patient age. (From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission.)

498

Handbook of Anatomical Models for Radiation Dosimetry

those agreements are in many cases the result of compensating errors in individual organabsorbed doses that contribute to the overall effective dose in the reference patient. Individual organ-absorbed doses and corresponding effective doses are tabulated and compared for selected phantoms in Tables 21.1 through 21.4 for the 12 mm collimator setting. Because the UF phantoms and the ORNL phantoms do not exactly match by reference age, phantoms of similar ages are paired for the side-by-side organ dose comparison: ORNL 1 year vs. UF 9 month, ORNL 5 year vs. UF 4 year, ORNL 10 year vs. UF 11 year, and the ORNL 15 year vs. UF 14 year phantoms. Since they are of different ages, percent differences may not be used for the absolute comparison between the series of phantoms. With proper caution, however, these comparisons do allow one to understand important differences between phantoms types. The gonads are assigned the highest tissue weighting factors in the effective dose as defined in ICRP Publication 60.8 The testes of the male phantoms do not show significant differences in the absorbed dose between phantom types when these tissues fall under the scanning area, except for the UF 11 year phantom and the ORNL 10 year phantom. The testes of the UF 11 year phantom are located deeper between the legs than in the other male phantoms and the shielding effects from the legs under lateral irradiation result in relatively less absorbed dose than seen in the ORNL 10 year phantom. This phenomenon can be understood as a characteristic of individual variability for this voxel phantom but at the same time it suggests the importance of organ location beyond simply the organ mass in predicting the magnitude of the organ dose. The female phantoms show distinct differences in ovary dose (see Table 21.2). The ovary dose of the UF 4 year phantom is 47.6% lower than the ORNL 5 year phantom under abdominal CT examination, while it is 25.6% higher in the pelvic exam. This observation implies that the location of the ovaries of the ORNL phantom is closer to the abdominal CT exam x-ray field than found in the corresponding UF 4 year phantom. For the pelvic exam, however, where the ovaries of both phantoms are in the field of the CT x-ray beam, the reasons for the 25.6% higher dose in the UF 4 year phantom are different. First, the thinner abdominal thickness of the UF 4 year phantom results in less shielding from the surrounding muscle, thus resulting in a higher absorbed dose to the ovaries. Second, the less realistic pelvic structure within the ORNL phantom, where the ovaries are surrounded by a greater extent by skeletal tissues, causes an increased shielding effect over that seen in the pelvic regions of the UF phantom. Figure 21.5 displays the pelvic skeletal structures of the two phantoms near the ovaries. The enhanced curvature of the pelvis of the ORNL phantom produces greater organ shielding than seen in the UF phantom for x-ray fields incident in the PA direction. The difference of the shielding effect of the pelvis was also found in the CAP exam, where the issue with ovary position was minimized as both ovaries were inside the field-of-view in the CAP exam. The UF phantom receives 17.6% higher ovary absorbed dose in this case compared to that seen in the corresponding ORNL phantom. It is difficult to find consistent differences in the red bone marrow dose between phantom types, except in the case of chest exams. For younger patients, the bone marrow of the UF phantoms receives a significantly lower absorbed dose than their counterparts in the ORNL phantom series (see Tables 21.1 through 21.3). For example, the bone marrow of the UF 9 month phantom receives 26.8% lower absorbed dose than that of the ORNL 1 year phantom for the chest exams. The overestimation of the bone marrow dose in chest exams of the younger ORNL phantoms is primarily due to the less realistic description of the rib cage and the inclusion of the arm structures within the phantom torso. The rib cage of the ORNL phantom is composed of multiple ellipsoidal closed rings each considered to enclose a homogeneous mixture of cortical bone and bone marrow. In contrast, red bone

0.00 5.75 0.01 0.26 0.04 0.01 0.06 0.28 1.28 2.95 9.27 0.07 16.74 1.28 0.02 0.04 0.15 0.06 0.07 0.39 1.36

ORNL 1 Year

0.01 5.54 0.06 0.58 0.16 0.02 0.17 0.81 3.00 3.94 20.14 0.10 17.70 6.99 0.06 0.07 3.50 0.11 0.16 1.07 1.35

UF 9 Months N/A −3.5 302.0 122.0 265.4 205.6 180.5 187.1 135.3 33.5 117.3 48.5 5.7 447.7 275.2 58.3 2275.2 72.5 127.0 172.2 −0.3

% Diff.

Abdomen

Pelvis

CAP

0.07 4.05 1.06 14.74 9.20 0.27 9.73 12.52 7.96 3.42 11.69 11.58 0.26 7.96 1.30 6.62 6.04 11.23 11.25 13.71 5.52

0.07 2.96 1.94 14.97 10.99 0.19 7.98 12.52 8.02 2.58 7.61 3.95 0.29 3.04 1.40 1.49 3.30 3.34 8.51 13.78 5.45

−2.8 −26.8 82.5 1.6 19.5 −29.9 −18.0 0.0 0.7 −24.7 −34.9 −65.9 10.9 −61.9 7.6 −77.5 −45.4 −70.3 −24.4 0.6 −1.3

0.35 2.23 7.08 2.74 12.65 1.15 12.15 4.23 0.20 2.20 5.78 10.71 0.03 0.20 9.24 12.52 4.63 11.48 12.15 0.55 4.31

0.50 2.03 11.10 4.41 13.77 1.97 13.68 5.52 0.68 2.92 5.07 13.45 0.06 0.42 13.06 14.51 3.88 14.06 14.35 1.07 5.56

42.2 −9.1 56.8 61.4 8.8 70.6 12.6 30.5 243.2 32.6 −12.4 25.5 134.1 113.1 41.3 15.9 −16.3 22.5 18.2 95.8 29.1

12.79 2.55 7.94 0.17 0.91 12.63 0.77 0.24 0.02 2.04 5.08 0.46 0.01 0.02 7.33 1.05 5.22 0.74 0.66 0.05 4.78

12.44 2.03 6.29 0.12 0.38 14.28 0.49 0.14 0.04 2.07 4.71 0.58 0.01 0.03 3.53 1.70 2.50 0.68 0.39 0.06 4.41

−2.7 −20.2 −20.7 −30.6 −58.6 13.1 −35.8 −43.2 83.1 1.3 −6.8 25.4 33.1 17.5 −51.8 62.4 −52.0 −8.6 −39.8 20.2 −7.7

12.80 7.35 13.80 15.28 14.78 13.56 14.47 13.43 8.42 6.40 18.71 13.60 0.27 8.42 14.16 14.13 13.47 14.04 14.21 13.90 11.98

12.66 6.02 15.63 15.95 15.62 15.62 15.64 13.81 7.63 6.43 14.80 14.79 0.32 3.19 15.76 16.16 8.20 15.45 15.94 14.21 12.41

−1.1 −18.1 13.3 4.4 5.6 15.2 8.1 2.8 −9.4 0.4 −20.9 8.8 17.8 −62.2 11.3 14.3 −39.1 10.0 12.1 2.2 3.6

ORNL 1 UF 9 ORNL 1 UF 9 ORNL 1 UF 9 ORNL 1 UF 9 Year Months % Diff. Year Months % Diff. Year Months % Diff. Year Months % Diff.

Chest

Source: From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission. Note: A tube potential of 120 kVp and the collimated beam thickness of 12 mm were used for the simulation (% diff. = (ORNL − UF)/ORNL × 100).

Tesles Bone marrow Colon Lungs Stomach Urinary bladder Liver Esophagus Thyroid Skin Bone surface Adrenals Brain ET region SI wall Kidney Muscle Pancreas Spleen Thymus Effective dose (male)

Organs

Head

Comparison of Organ-Absorbed Dose (mGy) and Effective Dose (mSv) for the ORNL 1 Year and the UF 9 Month Phantoms Normalized Per 100 mAs

TABLE 21.1

Applications to Computed Tomography for Pediatric Patients 499

ORNL 5 Years

N/A 22.1 129.6 69.5 174.3 N/A

210.7 179.5 194.8 26.7 18.5 44.6 69.3 8.8 214.5 147.6 144.7 1439.0 39.4 134.0 110.6 N/A 24.6

0.18 0.09 0.57 1.45 2.62 13.00 0.06 16.64 3.60 0.02 0.05 1.77 0.04 0.09 0.54 0.01 1.20

% Diff.

0.00 3.73 0.01 0.32 0.07 0.01

UF 4 Years

12.71 8.82 11.62 6.96 2.88 8.57 10.84 0.20 6.96 0.94 5.64 5.49 10.11 10.25 13.43 0.33 5.70

0.36 3.52 0.77 13.98 8.24 0.14

ORNL 5 Years

10.08 6.41 10.48 9.04 2.12 6.68 3.97 0.22 2.37 0.54 1.72 2.56 1.99 6.70 12.78 0.08 4.71

0.08 2.83 0.33 13.85 4.92 0.07

UF 4 Years

Chest

−20.7 −27.3 −9.8 29.9 −26.2 −22.1 −63.4 8.6 −65.9 −42.3 −69.5 −53.5 −80.3 −34.6 −4.9 −77.3 −17.4

−76.3 −19.4 −57.0 −0.9 −40.3 −49.3

% Diff.

0.66 11.27 3.57 0.11 1.80 3.94 9.90 0.01 0.11 8.11 11.93 4.04 10.42 11.17 0.44 2.21 4.45

2.36 2.14 6.29 2.02 11.82 0.84

ORNL 5 Years

0.83 12.80 3.39 0.42 2.19 3.87 11.50 0.04 0.26 12.74 14.14 2.95 13.06 13.28 0.73 1.24 4.87

1.24 1.88 8.30 3.35 12.62 0.92

UF 4 Years

Abdomen

25.3 13.5 −4.9 287.0 21.8 −1.9 16.1 174.5 140.8 57.1 18.6 −27.0 25.4 18.8 68.2 −43.6 9.5

−47.6 −12.0 32.0 65.5 6.7 9.5

% Diff.

0.03 0.55 0.13 0.00 1.72 3.59 0.28 0.00 0.00 6.59 0.76 4.77 0.48 0.43 0.03 10.90 4.20

10.49 2.74 7.25 0.09 0.67 12.03

ORNL 5 Years

0.03 0.35 0.08 0.02 2.11 6.00 0.39 0.00 0.02 4.54 0.77 3.05 0.67 0.30 0.03 13.01 4.97

13.17 2.70 8.40 0.08 0.45 14.41

UF 4 Years

Pelvis

7.7 −37.0 −38.7 N/A 22.9 67.2 40.8 N/A N/A −31.1 1.6 −36.1 37.7 −31.5 11.1 19.4 18.3

25.6 −1.5 15.9 −13.0 −32.7 19.8

% Diff.

13.19 13.37 12.34 7.53 5.47 13.69 12.48 0.21 7.53 12.96 13.23 12.40 12.79 13.10 13.65 12.81 11.89

12.24 7.18 12.64 14.34 13.64 12.72

ORNL 5 Years

10.67 14.59 11.40 7.13 5.72 14.85 13.02 0.25 2.38 15.64 15.29 7.68 14.22 14.75 13.07 13.89 12.80

14.39 6.64 15.47 14.75 14.13 15.05

−19.1 9.1 −7.6 −5.3 4.5 8.5 4.4 17.8 −68.3 20.6 15.5 −38.0 11.2 12.6 −4.3 8.4 7.7

17.6 −7.6 22.4 2.8 3.6 18.3

UF 4 Years % Diff.

CAP

Source: From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission. Note: The tube potential of 120 kVp and the collimated beam thickness of 12 mm were used for the simulation (% diff. = (ORNL − UF)/ORNL × 100).

Ovary 0.00 Bone marrow 3.06 Colon 0.01 Lungs 0.19 Stomach 0.02 Urinary 0.00 bladder Breast 0.06 Liver 0.03 Esophagus 0.19 Thyroid 1.15 Skin 2.21 Bone surface 8.99 Adrenals 0.04 Brain 15.30 ET region 1.15 SI wall 0.01 Kidney 0.02 Muscle 0.12 Pancreas 0.03 Spleen 0.04 Thymus 0.26 Uterus 0.00 Effective dose 0.96 (female)

Organs

Head

Comparison of Organ-Absorbed Dose (mGy) and Effective Dose (mSv) for the ORNL 5 Years and the UF 4 Year Phantoms Normalized Per 100 mAs

TABLE 21.2

500 Handbook of Anatomical Models for Radiation Dosimetry

0.00 1.88 0.01 0.14 0.03 0.00

0.04 0.21 0.50 1.72 8.11 0.03 13.94 1.96 0.01 0.02 0.81 0.02 0.04 0.23 0.74

0.00 2.11 0.00 0.10 0.01 0.00

0.01 0.10 0.76 1.60 6.45 0.01 14.65 0.76 0.00 0.01 0.07 0.01 0.02 0.12 0.76

184.2 102.1 −34.8 7.4 25.7 181.0 −4.9 157.8 N/A 186.3 1085.1 113.8 142.6 90.7 −2.7

N/A −11.0 N/A 36.1 289.1 N/A 7.97 10.02 2.78 2.37 7.09 9.83 0.13 2.78 0.64 4.74 4.93 8.86 8.85 12.68 4.42

0.02 3.38 0.54 12.63 7.14 0.08 7.21 8.99 5.22 1.58 5.01 4.94 0.10 1.33 0.47 2.55 1.78 2.63 7.53 10.79 3.82

0.00 2.63 0.45 12.04 5.92 0.02

ORNL UF 11 % Diff. 10 Years Years

ORNL UF 11 10 Years Years

−9.5 −10.3 88.0 −33.5 −29.3 −49.8 −21.4 −52.0 −26.7 −46.2 −63.9 −70.4 −14.9 −14.9 −13.4

N/A −22.3 −17.5 −4.7 −17.2 −78.6 9.90 2.89 0.07 1.43 2.98 8.81 0.01 0.07 5.91 10.88 3.35 9.19 9.94 0.31 3.36

0.07 1.82 4.87 1.55 10.59 0.48 10.85 3.69 0.37 1.71 3.59 9.90 0.02 0.16 10.26 12.30 2.02 11.32 11.78 0.59 4.05

0.06 2.33 6.86 2.96 11.12 0.30

ORNL UF 11 % Diff. 10 Years Years

Abdomen

9.6 27.9 440.1 19.6 20.3 12.4 153.5 166.6 73.5 13.0 −39.7 23.2 18.4 92.1 20.5

−11.7 27.8 40.9 91.6 5.0 −37.3 0.46 0.09 0.00 1.55 3.16 0.22 0.00 0.00 6.80 0.71 4.44 0.39 0.36 0.02 4.18

11.07 2.94 6.97 0.06 0.60 11.25 0.13 0.04 0.01 1.74 5.42 0.17 0.00 0.00 4.36 0.33 2.70 0.31 0.12 0.01 3.55

8.23 3.15 6.48 0.03 0.17 12.16

ORNL UF 11 % Diff. 10 Years Years

Pelvis

−70.8 −49.6 N/A 12.2 71.5 −21.7 N/A N/A −35.8 −54.3 −39.2 −20.2 −65.1 −42.3 −15.0

−25.6 7.3 −7.1 −39.2 −72.2 8.0 11.93 10.52 2.67 4.62 11.39 11.06 0.13 2.67 11.45 12.07 11.15 11.29 11.65 12.80 9.78

10.79 7.09 11.14 12.91 12.21 11.62

12.63 10.01 5.40 4.51 12.67 11.16 0.11 1.38 13.89 13.24 5.93 12.52 13.21 10.96 9.88

8.15 7.33 13.16 12.74 12.71 12.45

ORNL UF 11 % Diff. 10 Years Years

CAP

5.9 −4.8 102.5 −2.5 11.3 0.9 −15.2 −48.3 21.4 9.7 −46.8 10.9 13.4 −14.4 1.0

−24.5 3.4 18.1 −1.3 4.1 7.1

% Diff.

Source: From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission. Note: The tube potential of 120 kVp and the collimated beam thickness of 12 mm were used for the simulation (% diff. = (ORNL − UF)/ORNL × 100).

Testes Bone marrow Colon Lungs Stomach Urinary bladder Liver Esophagus Thyroid Skin Bone surface Adrenals Brain ET region SI wall Kidney Muscle Pancreas Spleen Thymus Effective dose (male)

Organs

Chest

Head

Comparison of Organ Dose (mGy) and Effective Dose (mSv) for the ORNL 10 Year and the UF 11 Year Phantoms Normalized Per 100 mAs

TABLE 21.3

Applications to Computed Tomography for Pediatric Patients 501

0.00 1.51 0.00 0.04 0.00 0.00 0.00 0.04 0.31 1.15 4.36 0.01 13.90 0.31 0.00 0.00 0.03 0.00 0.00 0.04 0.61

ORNL 15 Years

0.00 1.76 0.00 0.09 0.01 0.00 0.01 0.21 0.44 1.48 7.22 0.03 14.90 1.94 0.00 0.02 0.51 0.01 0.00 0.17 0.72

UF 14 Years

N/A 16.4 N/A 126.3 N/A N/A N/A 433.4 43.1 29.3 65.8 453.8 7.2 534.9 N/A N/A 1463.5 N/A N/A 372.4 19.1

0.01 3.19 0.35 10.33 5.90 0.03 6.77 8.05 4.16 2.12 5.99 6.16 0.08 4.18 0.41 3.88 4.24 7.16 7.24 11.19 3.79

ORNL % Diff. 15 Years 0.01 3.33 0.68 12.59 5.02 0.03 5.61 9.91 6.93 1.80 4.58 8.86 0.08 1.59 0.36 4.66 2.01 2.82 0.84 12.13 3.95

UF 14 Years

Chest

−27.3 4.3 92.6 21.9 −14.9 23.7 −17.2 23.1 65.7 −15.1 −23.6 8.7 −2.8 −61.8 −12.2 20.2 −52.6 −60.6 −88.5 8.3 4.4

0.06 1.70 3.80 1.46 8.82 0.28 8.14 2.41 0.01 1.21 2.48 7.56 0.01 0.01 4.53 9.37 2.79 7.43 8.23 0.25 2.81

ORNL % Diff. 15 Years 0.05 1.81 8.19 2.50 9.22 0.35 9.66 2.07 0.22 1.31 2.12 8.34 0.01 0.10 5.05 10.51 1.75 9.47 10.52 0.34 3.63

UF 14 Years

Abdomen

−5.7 6.3 115.6 71.1 4.6 22.6 18.6 −13.9 3924.3 8.8 −14.5 10.3 24.2 1766.1 11.5 12.2 −37.2 27.5 27.8 37.7 29.3

7.91 2.70 5.70 0.03 0.39 9.85 0.27 0.04 0.00 1.32 2.72 0.12 0.00 0.00 5.51 0.48 3.52 0.22 0.21 0.00 3.24

ORNL % Diff. 15 Years 7.78 3.28 3.95 0.05 0.48 11.66 0.33 0.05 0.01 1.62 5.01 0.16 0.00 0.01 8.00 0.51 2.38 0.76 1.93 0.01 3.23

UF 14 Years

Pelvis

−1.7 21.7 −30.6 78.5 22.0 18.4 24.8 6.0 N/A 23.3 83.7 33.3 N/A N/A 45.1 5.9 −32.4 247.2 834.3 N/A −0.4

7.85 6.63 9.04 10.54 10.33 9.96 9.52 8.22 4.19 4.02 9.74 8.60 0.09 4.19 9.46 10.12 9.24 8.82 9.28 10.92 8.15

ORNL % Diff. 15 Years

7.83 7.62 11.69 13.04 10.73 11.91 11.04 10.27 6.33 4.19 10.68 10.68 0.08 1.58 12.20 11.82 5.40 10.67 11.97 12.17 9.24

UF 14 Years

CAP

Source: From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission. Note: The tube potential of 120 kVp and the collimated beam thickness of 12 mm were used for the simulation (% diff. = (ORNL − UF)/ORNL × 100).

Testes Bone marrow Colon Lungs Stomach Urinary bladder Liver Esophagus Thyroid Skin Bone surface Adrenals Brain ET region SI wall Kidney Muscle Pancreas Spleen Thymus Effective dose (male)

Organs

Head

−0.2 14.9 29.3 23.8 3.9 19.6 16.1 25.0 51.2 4.2 9.7 24.2 −3.4 −62.3 29.0 16.8 −41.6 21.0 28.9 11.5 13.4

% Diff.

Comparison of Organ-Absorbed Dose (mGy) and Effective Dose (mSv) for the ORNL 15 Year and the UF 14 Year Phantoms Normalized Per 100 mAs

TABLE 21.4

502 Handbook of Anatomical Models for Radiation Dosimetry

Applications to Computed Tomography for Pediatric Patients

(a)

503

(b)

FIGURE 21.5 The skeletal structures near the ovaries of (a) the ORNL 5 year phantom and (b) the UF 4 year phantom. (From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission.)

marrow within the UF phantom is more properly confined to only the more lateral regions of the rib cage, and not the anterior locations of the cartilage-filled false ribs as shown in Figure 21.6. The presence of the additional skeletal tissues of the ribs and arm bones in the ORNL phantoms thus causes enhanced energy absorption to bone and subsequently to red bone marrow. The colon also demonstrates anatomical differences in the two phantom types. As mentioned earlier, because of torso thickness differences, the colon of the UF phantoms consistently receives higher absorbed doses than seen in the ORNL phantoms under CT examination of the abdomen. Finally, the thyroid under chest CT examination shows significantly different dosimetric characteristics, especially for the older phantoms. For example, the thyroid of the UF 11 year phantom receives 88% higher absorbed dose than

(a)

(b)

FIGURE 21.6 Anatomical regions irradiated during chest CT examination in (a) the ORNL 1 year phantom and (b) the UF 9 month phantom. (From Lee, C. et al., Med. Phys., 34, 1858, 2007. With permission.)

504

Handbook of Anatomical Models for Radiation Dosimetry

that seen in the ORNL 10 year phantom, even though the same organ of the UF 9 month phantom receives only 0.7% higher dose than in the ORNL 1 year phantom. As was also observed by other authors in phantoms of adult subjects,35 the thyroid within the ORNL phantom is not located within the chest, but is positioned solely within neck region of the phantoms. In contrast, the anatomically realistic thyroid of the voxel phantoms is positioned at a height similar to that of the clavicles, and thus is shielded to a greater extent under CT irradiation.

21.4 CT Dosimetry Study Using UF Hybrid Pediatric Phantoms 21.4.1 Material and Methods The UF hybrid 15 year old male and female phantoms representing the reference anatomy defined by ICRP89 were developed from the realistic anatomy sources: UF 14 year old male voxel phantom and a number of CT data of nearly 15 year old female subjects. The 50th weight percentile hybrid 15 year old male and female phantoms were named UFH15M50 and UFH15F50, respectively. A total of eight anthropometric parameters were matched to the standard values obtained from the four different resources within the error of 4%. Biacromial breadth in the UF hybrid 15 year old male phantom showed the largest discrepancy up to 4% from the standard data. All body organs and tissues are shown to be matched to within 1% with the exception of the skin (14.8% and 13.2% larger for male and female, respectively). Brain mass which was smaller than ICRP89 reference value in case of newborn phantom was exactly matched within 1% in the 15 year old phantoms. We generated the 10th and 90th weight percentile phantoms from the template 50th weight percentile phantoms by adjusting control points surrounding torso contour to match the standard total body weight reported by CDC. Figures 21.7 and 21.8 show the frontal and lateral 3D views of the three male and three female phantoms, respectively, with semitransparent skin and residual tissue. Total body masses of the 10th and 90th weight percentile male phantoms were 45.4 and 72.3 kg, respectively, and those of female phantoms were 42.5 and 68.5 kg, respectively. The total body masses were matched to the CDC standard values mentioned above within 0.3% accuracy. 21.4.2 Results and Discussion Absorbed doses to the organs in the field of view were calculated for chest and abdomen CT exams using the three male and three female 15 year old phantoms with the beam parameter of 120 kVp tube voltage and 1.2 collimator width. Normalized values of organabsorbed dose (mGy/100 mAs) are given in Table 21.5 for the male phantoms and in Table 21.6 for the female phantoms for both chest and abdominal CT examinations. In each table, organ doses are given for the traditional reference stylized phantom (e.g., ORNL 15 years), and three different versions of the UF hybrid phantom: one at the 50th weight percentile (reference phantom), and two additional at the 10th and 90th weight percentiles. For these latter phantoms, the differential in total body mass is assumed to be due only to variations in subcutaneous fat thicknesses as shown in Figures 21.7 and 21.8, and all hybrid phantoms are given at equivalent value of total body and trunk height.

Applications to Computed Tomography for Pediatric Patients

10th

50th

90th

(a)

505

10th

50th

90th

(b)

FIGURE 21.7 Frontal views of 3D rendering of the 10th (left), 50th (middle), and 90th (right) weight percentile UFH-NURBS 15 year old (a) male and (b) female phantoms. (From Lee, C., Med. Phys., 35, 2366, 2008. With permission.)

10th (a)

50th

90th

10th

50th

90th

(b)

FIGURE 21.8 (See color insert following page 524.) Lateral views of 3D rendering of the 10th (left), 50th (middle), and 90th (right) weight percentile UFH-NURBS 15 year old (a) male and (b) female phantoms. (From Lee, C., Med. Phys., 35, 2366, 2008. With permission.)

In Table 21.5, columns 2 and 3 show the absolute organ doses per unit integrated tube current for the ORNL and UFH15M50 phantoms, respectively, with their ratio given in column 4. These absorbed dose ratios range from 0.59 (kidneys) to 2.05 (thyroid) for the chest exam, and from 0.71 (esophagus) to 1.40 (lungs) for the abdominal exam. Only two

8.63E+00 1.11E+01 1.12E+01 2.28E+00 5.67E+00 1.06E+01 4.03E+00 9.81E+00 8.55E+00

4.34E+00 1.71E+00 1.49E+00 7.89E+00 8.23E+00 2.04E+00 7.99E+00 3.54E+00

8.05E+00 8.08E+00 1.00E+01 3.88E+00 6.77E+00 1.03E+01 5.90E+00 1.12E+01 4.18E+00

3.80E+00 2.41E+00 1.51E+00 9.37E+00 8.14E+00 1.46E+00 8.82E+00 4.53E+00

DORNL DUFH50 (mGy/100 mAs) (mGy/100 mAs)

1.14 0.71 0.98 0.84 1.01 1.40 0.91 0.78 0.71 1.40

1.07 1.38 1.12 0.59 0.84 1.03 0.68 0.88 2.05 0.59 2.05

DUFH50/DORNL

4.82E+00 1.82E+00 1.50E+00 9.08E+00 9.10E+00 2.09E+00 8.99E+00 4.18E+00

9.44E+00 1.15E+01 1.21E+01 2.36E+00 6.05E+00 1.14E+01 4.31E+00 1.04E+01 8.73E+00

DUFH10 (mGy/100 mAs)

1.11 1.06 1.01 1.15 1.11 1.03 1.13 1.18 1.01 1.18

1.09 1.04 1.08 1.03 1.07 1.08 1.07 1.06 1.02 1.02 1.09

DUFH10/DUFH50

3.09E+00 1.32E+00 1.27E+00 5.73E+00 5.56E+00 1.70E+00 5.50E+00 2.54E+00

6.65E+00 1.05E+01 8.50E+00 1.83E+00 4.07E+00 8.41E+00 3.07E+00 7.91E+00 7.43E+00

DUFH90 (mGy/100 mAs)

0.71 0.77 0.86 0.73 0.68 0.83 0.69 0.72 0.68 0.86

0.77 0.95 0.76 0.80 0.72 0.79 0.76 0.81 0.87 0.72 0.95

DUFH90/DUFH50

Source: From Lee, C., Med. Phys., 35, 2366, 2008. With permission. Note: The ratios of organ doses between the ORNL and 50th percentile phantoms, 10th and 50th phantoms, and 90th and 50th phantoms are given.

Chest CT Esophagus Breast Heart Kidney Liver Lungs Stomach wall Thymus Thyroid Min Max Abdomen CT Colon and Rectum Esophagus Heart Kidney Liver Lungs Stomach wall Small intestine Min Max

Male Phantoms

Normalized Values of Organ-Absorbed Dose (mGy/100 mAs) Calculated from the Reference ORNL 15 Year Phantom, and the 10th, 50th, and 90th Weight Percentile UF Hybrid 15 Year Old Male Phantoms for Chest and Abdominal CT Exams

TABLE 21.5

506 Handbook of Anatomical Models for Radiation Dosimetry

8.46E+00 9.43E+00 1.03E+01 1.18E+00 6.59E+00 1.05E+01 5.54E+00 9.31E+00 8.72E+00

5.46E+00 2.02E+00 1.68E+00 9.78E+00 8.94E+00 2.01E+00 9.20E+00 5.85E+00

3.80E+00 2.41E+00 1.51E+00 9.37E+00 8.14E+00 1.46E+00 8.82E+00 4.53E+00

DUFH50 (mGy/100 mAs)

8.05E+00 8.08E+00 1.00E+01 3.88E+00 6.77E+00 1.03E+01 5.90E+00 1.12E+01 4.18E+00

DORNL (mGy/100 mAs)

1.44 0.84 1.11 1.04 1.10 1.37 1.04 1.29 0.84 1.44

1.05 1.17 1.03 0.30 0.97 1.01 0.94 0.83 2.09 0.30 2.09

DUFH50/ DORNL

5.73E+00 2.18E+00 1.68E+00 1.05E+01 9.71E+00 2.04E+00 1.00E+01 6.35E+00

9.33E+00 1.09E+01 1.18E+01 1.16E+00 7.14E+00 1.17E+01 5.93E+00 9.92E+00 8.88E+00

DUFH10 (mGy/100 mAs)

1.05 1.08 1.00 1.08 1.09 1.02 1.09 1.09 1.00 1.09

1.10 1.15 1.14 0.98 1.08 1.11 1.07 1.07 1.02 0.98 1.15

DUFH10/ DUFH50

4.12E+00 1.91E+00 1.65E+00 7.00E+00 7.92E+00 1.95E+00 8.06E+00 4.14E+00

7.90E+00 9.37E+00 9.82E+00 1.18E+00 6.27E+00 9.77E+00 5.25E+00 8.82E+00 8.42E+00

DUFH80 (mGy/100 mAs)

0.75 0.95 0.99 0.72 0.89 0.97 0.88 0.71 0.71 0.99

0.93 0.99 0.95 1.00 0.95 0.93 0.95 0.95 0.96 0.93 1.00

DUFH99/ DUFH50

Source: From Lee, C., Med. Phys., 35, 2366, 2008. With permission. Note: The ratios of organ doses between the ORNL and 50th percentile phantoms, 10th and 50th phantoms, and 90th and 50th phantoms are given.

Chest CT Esophagus Breast Heart Kidney Liver Lungs Stomach wall Thymus Thyroid Min Max Abdomen CT Colon and rectum Esophagus Heart Kidney Liver Lungs Stomach wall Small intestine Min Max

Female Phantoms

Normalized Values of Organ-Absorbed Dose (mGy/100 mAs) Calculated from the Reference ORNL 15 Year Phantom, and the 10th, 50th, and 90th Weight Percentile UF Hybrid 15 Year Old Female Phantoms for Chest and Abdominal CT Exams

TABLE 21.6

Applications to Computed Tomography for Pediatric Patients 507

508

Handbook of Anatomical Models for Radiation Dosimetry

of nine organs are within 10% of each other for the chest exam, and only three of eight are within 10% for the abdominal exam. In each case, the organ masses are within 1% of their ICRP 89 reference values, and thus these differences are directly reflective of differences in anatomical shape, position, and depth of each organ with each phantom. Comparable differences are noted in comparing the two reference (50th percentile) female phantoms in column 4 of Table 21.6. Here, we note that organ dose ratios range from 0.30 (kidney) to 2.09 (thyroid) for the chest CT exams, and from 0.84 (esophagus) to 1.44 (colon and rectum) for the abdominal CT exams. The flexibility of the hybrid computational phantoms is further demonstrated in column 6 of Tables 21.5 and 21.6 for the male and female phantoms, respectively. For the 10th percentile phantoms, organ doses are shown to be within ±10% of those given by the corresponding 50th percentile reference hybrid phantom for nine of nine organs in the male chest exam, three of eight for the male abdominal exam, six of nine for the female chest exam, and eight of eight for the female abdominal exam. The composite range of dose ratios was a low of only 0.98 (kidneys—female chest CT) to a high of 1.18 (small intestine— male abdominal CT). Larger errors in dosimetry, however, can be seen when using 50th percentile reference phantoms to represent overweight children in CT body examinations. In column 8 of Tables 21.4 and 21.5, we have given the ratio of normalized organ-absorbed doses in chest and abdominal CT exams for the 90th percentile male (Table 21.5) and female (Table 21.6) to those given by the 50th percentile reference hybrid phantoms (male and female phantom, respectively). For the male chest CT, organ doses are underestimated by use of the reference phantom by as little of 5% (dose ratio of 0.95 for the male breast) to as much as 28% (dose ratio of 0.72 for the liver). Dose ratios range from 0.68 (liver) to 0.86 (heart) for the male abdominal CT exams. Closer agreement is seen between organ doses assessed in the 90th and 50th percentile female phantoms for the chest CT examination (see dose ratios in column 8 of Table 21.6). Here, only the lungs and esophagus show differences as much as 7%, and all organs are thus within a tolerance of ±10%. This is not surprising, however, as the additional 8.5 kg of total body weight is apportioned to the female phantom primarily in the abdominal region and upper thighs, and not within the upper regions of the torso. In column 8 of the lower half of Table 21.6, we thus show dose ratios of normalized organ doses between the 90th and 50th percentile female phantoms for the abdominal CT exam. Only three of eight organ doses are within a tolerance of ±10%, and organ doses are seen to be underestimated in the reference phantom by as high as 29%–25% (small intestine, colon, and kidneys).

21.5 Conclusions and Future Research Needs In this chapter, we summarized past studies of CT organ dosimetry in children using both stylized (equation-based) and voxel (image-based) computational phantoms, and then we discussed the use of the UF pediatric series of NURBS-based hybrid phantoms, outlined in Chapter 9, in the assessment of modern-day multislice CT scanners. Existing databases for reporting organ doses in pediatric CT imaging were developed primarily in the early to middle 1990s, and thus there is a critical need to update organ doses to include more modern-day CT scanner technologies. The use of hybrid phantoms to represent the scanned subject can also be used to model non reference (non-50th percentile) subjects,

Applications to Computed Tomography for Pediatric Patients

509

thus significantly improving the patient specificity of the dose estimates. Such an expanded database would be of immense help in not only the conduct of radiation epidemiological studies of CT imaging cancer risk, but additionally in providing patient-specific reporting of organ dose to the patient’s electronic medical charts at the time of examination.

Acknowledgment This work was performed under grant RO1 CA116743 from the National Cancer Institute (subcontract from Rensselaer Polytechnic Institute) with the University of Florida.

References 1. Brenner, D.J. Estimating cancer risks from pediatric CT: Going from the qualitative to the quantitative, Pediatr Radiol, 32, 228, 2002. 2. Pierce, D.A. et al. Studies of the mortality of atomic bomb survivors. Report 12, Part I. Cancer: 1950–1990, Radiat Res, 146, 1, 1996. 3. Preston, D.L. et al. Studies of mortality of atomic bomb survivors. Report 13: Solid cancer and noncancer disease mortality: 1950–1997, Radiat Res, 160, 381, 2003. 4. Mettler, F.A. Jr. et al. Use of radiology in U.S. general short-term hospitals: 1980–1990, Radiology, 189, 377, 1993. 5. Nickoloff, E.L. and Alderson, P.O. Radiation exposures to patients from CT: Reality, public perception, and policy, AJR Am J Roentgenol, 177, 285, 2001. 6. Mettler, F.A. Jr. et al. CT scanning: Patterns of use and dose, J Radiol Prot, 20, 353, 2000. 7. Brenner, D. et al. Estimated risks of radiation-induced fatal cancer from pediatric CT, AJR Am J Roentgenol, 176, 289, 2001. 8. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Oxford: Pergamon Press, 1991. 9. Shrimpton, P.C. and Wall, B.F. Reference doses for paediatric computed tomography, Radiat Prot Dosim, 90, 249, 2000. 10. Cristy, M. and Eckerman, K.F. Specific Absorbed Fractions of Energy at Various Ages from Internal Photon Sources, ORNL/TM-8381 V1–V7, Oak Ridge: Oak Ridge National Laboratory, 1987. 11. Boone, J.M. et al. Dose reduction in pediatric CT: A rational approach, Radiology, 228, 352, 2003. 12. Nickoloff, E.L., Dutta, A.K., and Lu, Z.F. Influence of phantom diameter, kVp and scan mode upon computed tomography dose index, Med Phys, 30, 395, 2003. 13. Siegel, M.J. et al. Radiation dose and image quality in pediatric CT: Effect of technical factors and phantom size and shape, Radiology, 233, 515, 2004. 14. Axelsson, B., Persliden, J., and Schuwert, P. Dosimetry for computed tomography examination of children, Radiat Prot Dosim, 64, 221, 1996. 15. Chapple, C.L., Willis, S., and Frame, J. Effective dose in paediatric computed tomography, Phys Med Biol, 47, 107, 2002. 16. Varchena, V. et al. Children’s heterogeneous phantoms and their application in roentgenology, Radiat Prot Dosim, 49, 77, 1993. 17. Varchena, V. Pediatric phantoms, Pediatr Radiol, 32, 280, 2002. 18. Zankl, M. et al. Organ doses for children from computed tomographic examinations, Radiat Prot Dosim, 57, 393, 1995.

510

Handbook of Anatomical Models for Radiation Dosimetry

19. Caon, M., Bibbo, G., and Pattison, J. An EGS4-ready tomographic computational model of a 14-year-old female torso for calculating organ doses from CT examinations, Phys Med Biol, 44, 2213, 1999. 20. Caon, M., Bibbo, G., and Pattison, J. Monte Carlo calculated effective dose to teenage girls from computed tomography examinations, Radiat Prot Dosim, 90, 445, 2000. 21. Khursheed, A. et al. Influence of patient age on normalized effective doses calculated for CT examinations, Br J Radiol, 75, 819, 2002. 22. Castellano, A., Dance, D.R., and Evans, P.M. CT dosimetry: Getting the best from the adult Cristy phantom, Radiat Prot Dosim, 114, 321, 2005. 23. Jones, D.G. and Shrimpton, P.C. Software: Normalized Organ Doses for X-Ray Computed Tomography Calculated Using Monte Carlo Techniques, NRPB Report SR250, Chilton, U.K.: National Radiological Protection Board, 1993. 24. Jarry, G. et al. A Monte Carlo-based method to estimate radiation dose from spiral CT: From phantom testing to patient-specific models, Phys Med Biol, 48, 2645, 2003. 25. Staton, R.J. et al. Organ and effective doses in newborn patients during helical multislice computed tomography examination, Phys Med Biol, 51, 5151, 2006. 26. Lee, C. et al. Organ and effective doses in pediatric patients undergoing helical multislice computed tomography examination, Med Phys, 34, 1858, 2007. 27. Lee, C. et al. Whole-body voxel phantoms of paediatric patients—UF Series B, Phys Med Biol, 51, 4649, 2006. 28. Lee, C. et al. The UF series of tomographic computational phantoms of pediatric patients, Med Phys, 32, 3537, 2005. 29. Lee, C. et al. Development of the two Korean adult tomographic computational phantoms for organ dosimetry, Med Phys, 33, 380, 2006. 30. ICRP. Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values, Publication 89, International Commission on Radiological Protection, Oxford; Pergamon Press, 2002. 31. Pelowitz, D.B. MCNPX User’s Manual Version 2.5.0, Los Alamos National Laboratory Report LA-CP-05-0369, 2005. 32. ICRU. Photon, Electron, Proton and Neutron Interaction Data for Body Tissues, Report 46, Bethesda, MD: International Commission on Radiation Units and Measurements, 1992. 33. Cristy, M. Active bone marrow distribution as a function of age in humans, Phys Med Biol, 26, 389, 1981. 34. Lee, C. and Bolch, W. Bone marrow and bone endosteum dosimetry methods comparison for external photons [Abstract], Med Phys, 32, 2101, 2005. 35. Lee, C., Lee, J., and Lee, C. The effect of unrealistic thyroid vertical position on thyroid dose in the MIRD phantom, Med Phys, 31, 2038, 2004. 36. Lee, C. et al. Hybrid computational phantoms of the 15-year male and female adolescent: Applications to computed tomography organ dosimetry for patients of variable morphometry, Med Phys, 35, 2366, 2008.

22 Applications to Computed Tomography for Adult Patients John J. DeMarco and Michael McNitt-Gray

CONTENTS 22.1 Introduction ............................................................................................................... 511 22.2 Materials and Methods ............................................................................................ 512 22.2.1 Overview of the MCNPX™ Monte Carlo Code ........................................ 512 22.2.2 Voxelized Patient Models............................................................................ 513 22.2.3 MCNP Lattice and Mesh Dose Scoring .................................................... 513 22.2.4 Computed Tomography Source Modeling in MCNP ............................. 514 22.2.5 Normalization and Simulation Benchmarks ........................................... 514 22.3 Simulation Benchmarks ........................................................................................... 517 22.4 Experiments Using Voxelized Phantom Data....................................................... 517 22.5 Experiments Using Voxelized Patient Data .......................................................... 519 22.6 Conclusion.................................................................................................................. 522 Acknowledgment................................................................................................................. 523 References ............................................................................................................................. 523

22.1 Introduction The increased technological capabilities, and complexity of multi detector row computed tomography (MDCT) as a diagnostic imaging tool has created a need for more accurate calculation tools to estimate the absorbed dose and radiological risk from CT examinations.1 An estimate of the effective dose to a patient undergoing a CT examination requires knowledge of the absorbed dose to each radiosensitive organ. Direct measurement of CT organ doses is not possible, and conventional dose estimates are difficult due to the complexity of modern, MDCT scanning protocols. Several approaches have been developed to approximate the behavior of current MDCT scanners such as those employed by imaging performance assessment of CT (ImPACT) in their dose calculation spreadsheet.2 The ImPACT CT dose spreadsheet is based on the simulations performed by the National Radiological Protection Board (NRPB) and the dose estimates for MDCT scan protocols are approximated from contiguous axial scan data and are not derived directly from helical acquisition simulations.3–5 In addition, the dose estimates are based upon the medical internal radiation dose, version V (MIRD-V) geometry model and do not account for patient-specific size variations. 511

512

Handbook of Anatomical Models for Radiation Dosimetry

Previous efforts have used Monte Carlo methods to estimate radiation dose from CT scans using realistic phantom models. Zankl et al. provided CT organ dose estimates to pediatric patients using voxelized models of their BABY and CHILD models.6 Caon et al. created a voxelized model of a 14 year old female to estimate effective dose from CT scan protocols.7,8 Khursheed et al. used the MCNPä code to calculate normalized effective dose values for three different scanners and mathematical anthropomorphic phantoms with ages ranging from newborn to adult.9 Schmidt and Kalender developed a fast Monte Carlo simulation method for calculating the dose in patient-specific geometries.10 More recently Tzedakis et al. simulated the effects of scan protocol, specifically overscanning, on voxelized cylindrical and anthropomorphic phantoms from a MDCT.11 Castellano et al. used Monte Carlo methods to compare doses to geometric patient models and different sized patient-based models,12 representing a teenager and a larger adult. Salvado et al. also used Monte Carlo-based methods to estimate doses to both anthropomorphic phantoms as well as some patient-based voxel models and compared these to CT dose index (CTDIw).13 Lee at al. have recently simulated a helical multislice CT to evaluate the organ and effective doses in pediatric patients.14 This chapter provides a review of previous efforts at the University of California (UCLA) to develop a comprehensive Monte Carlo-based CT scanner model that will simulate axial, helical, and variable tube current scan protocols. The CT simulation model has been applied to a family of patient-specific voxelized models in order to estimate absolute organ dose and whole-body effective dose as a function of patient size from arbitrary CT scan protocols. This type of simulation model is flexible, efficient, and can integrate other voxelized patient geometries where individual organs have been segmented.

22.2 Materials and Methods 22.2.1 Overview of the MCNPX‘ Monte Carlo Code MCNPX is a general-purpose Monte Carlo code and one of the family of MCNP codes developed and maintained at Los Alamos National Laboratory.15,16 All simulations were performed using the MCNPX (MCNP eXtended v2.6f) Monte Carlo code which supports the transport of 34 different particles including the applicable antiparticles. The MCNP family of simulation codes utilizes a combinatorial geometry system based upon planes, cylinders, cones, spheres, torii, and macrobodies to defi ne a three-dimensional (3D) simulation space. In addition to the conventional surfaces described above, the code also supports a 3D lattice geometry that is useful for simulating the voxelized geometries typically found in CT-based imaging systems. The code also supports a mesh tally feature that allows the user to overlay a Cartesian, cylindrical, or spherical tally mesh over an arbitrary simulation geometry. Simulations performed for CT diagnostic energies are typically run in photon-only mode and includes incoherent and coherent scattering, photoelectric absorption with the creation of K and L shell fluorescent photons, and doppler energy broadening of scattered photons. The photon transport model creates electrons but assumes that they travel in the direction of the primary photon, and that the electron energy is deposited at the photon interaction site, creating a condition of charged particle equilibrium (CPE). Under conditions of CPE, the assumption that collision kerma is equal to absorbed dose

Applications to Computed Tomography for Adult Patients

513

is valid and a collision-kerma tally can be used for the absorbed dose calculations. For all simulations, the collision kerma was calculated based upon a track-length estimate of the photon energy fluence. Energy fluence was converted to dose rate using the mass energy-absorption coefficients obtained by Hubbell and Seltzer.17 The current version of MCNPX incorporates the ENDF/B-VI photoatomic interaction data converted from the Lawrence Livermore evaluated photon, electron, and atomic data libraries. 22.2.2 Voxelized Patient Models Incorporating a voxelized patient model into MCNP requires a conversion from the original 3D patient model into the MCNP lattice geometry. The 3D patient model is typically acquired from CT data with each pixel represented as a Hounsfield number. DeMarco et al. have previously described a method for converting CT number into MCNP lattice material number. This requires a look-up table to convert a range of Hounsfield numbers into a single tissue-specific elemental composition and mass density. The result is a 3D matrix of MCNP material values each with its own material index and density. Although useful for creating 3D simulation geometries, the Hounsfield conversion method is based upon tissue composition and is not necessarily organ specific. Organspecific simulations require a preprocessing step that converts the original Hounsfield data into organ-specific index values. There are several examples of this type of preprocessing including the Visible Human Project,18,19 the GSF family of voxelized phantoms,20,21 and the University of Florida pediatric voxelized models.22,23 We have previously incorporated the GSF family of voxelized phantoms and the Visible Human model into the standard MCNP lattice structure.24 The native GSF and Visible Human data represents a 3D matrix of integer values with each integer representing a specific organ or tissue. The conversion to MCNP lattice geometry requires a similar look-up table with each organ index value related to a corresponding MCNP material index. In this case the MCNP material index may have identical elemental and density properties but is assigned an individual material index corresponding to a specific organ. In this book, GSF family phantoms were carefully described in Chapter 3. 22.2.3 MCNP Lattice and Mesh Dose Scoring The conversion methods described in Section 22.2.2 create 3D simulation geometries which require different methods for dose scoring. The voxel-by-voxel conversion method based upon Housfield number creates a 3D simulation lattice and the user has the option to tally the absorbed dose in individual voxel elements using the lattice or mesh tally geometry. The mesh tally is a virtual geometry and is only used to score particle fluence or dose. The Cartesian mesh tally is superimposed over the lattice geometry to efficiently tally the dose in a high-resolution Cartesian-coordinate lattice structure. The lattice geometry can also be used to score dose in individual voxel elements but becomes computationally inefficient for large 2D and 3D tally regions. Instead of scoring the dose in individual lattice elements, the mean organ dose is computed based upon the MCNP organ index value. The format of the lattice tally computes the mean organ dose by averaging the dose received by each voxel in a particular organ. A track-length estimate of energy fluence within the lattice voxel multiplied by the material-specific and energy-dependent mass energy-absorption coefficient produces collision kerma. The collision kerma is equivalent to absorbed dose under conditions of CPE. The assumption of CPE is reasonable almost everywhere in the model given the range of the resulting secondary electron fluence spectrum. The exception

514

Handbook of Anatomical Models for Radiation Dosimetry

is in the red bone marrow where the bone marrow is irradiated by photoelectrons released by the bony matrix. Because the Gesellschaft für Strahlenforschung (GSF) phantoms are contoured using a single bone index; we employed the method described by Zankl to estimate the absorbed dose to bone surface and bone marrow to account for these effects.21 A summation of the appropriate organ dose values weighted by the appropriate ICRP 60 tissue-weighting factors produced an estimate of the effective dose. 22.2.4 Computed Tomography Source Modeling in MCNP Jarry et al. previously described modifications to the MCNP source code to simulate a helical or axial CT source path.25 This modification is based upon changes to the MCNP source subroutine and requires the user to explicitly define the starting photon energy, position (x, y, z), and direction (cos θx, cos θy, cos θz). Simulation of commercial CT scanners requires additional information including photon emission spectra and bow-tie filter design. In order to simulate photons exiting the x-ray tube a cumulative distribution function (CDF) is precalculated for each photon emission energy. The CDF is formatted as a text file and an appropriate sequential sampling routine is implemented with the source.f subroutine. The CT bow-tie filter is not explicitly modeled in the simulation. Instead a table is precalculated that describes the half-fan emission angle versus the physical path length through the filter. This method assumes that the photons are emitted from a point source at the location of the x-ray target and the changes in path length due to the y- (i.e., longitudinal) dimension of the filter are negligible. A second sampling routine is implemented within source to uniformly sample through the CT fan-angle and choose the appropriate path length through the filter. This path-length value is used to calculate a first-collision attenuation factor based upon the appropriate filter material and photon energy. The physical shape of the filter is therefore accounted in a virtual sense by setting the photon weight equal to the attenuation factor. The movement of the CT source is based upon contiguous axial or helical source movement. For contiguous axial movement, the sampling algorithm requires the scan start and stop locations, and the nominal slice width thickness and spacing. The x, y, and z source positions are calculated by randomly sampling the CT gantry angle and the axial slice locations. Helical source movement requires scan start and stop, nominal beam slice width and table pitch. Figure 22.1 illustrates helical source movement as a function of table pitch by scoring the dose distribution using a rectangular mesh tally through the coronal plane of a homogenous test phantom. Variable tube current simulations are also possible by uniformly sampling for table position in the longitudinal direction. This produces a table index value that can be correlated with a corresponding gantry angle and mAs output setting. The mAs values can be normalized to the maximum tube current value and this normalized value can be applied to the photon weight exiting the CT source. Figures 22.2 and 22.3 illustrate variable mAs simulations through the sagittal plane of the CTDI body phantom. The variable mAs values are taken from a separate data set and are only used to test the simulation process. The practice of tube current modulation is gaining prevalence in modern MDCT scanners,26 and is used to dynamically reduce tube current when scanning thinner or less attenuating portions of patients. 22.2.5 Normalization and Simulation Benchmarks CT dosimetry measurements are typically performed using a pencil ionization chamber and the head and body CTDI phantoms. The pencil ionization chamber is an air-filled

0

0

13

13

25

25 Row

Row

Applications to Computed Tomography for Adult Patients

38

38

50

50

63

63

0

13

(a)

50

25 38 Column

515

63

0

13

25 38 Column

(b)

50

63

0 13

Row

25 38 50 63

(c)

0

13

25 38 Column

50

63

FIGURE 22.1 Helical irradiation patterns through the coronal midplane of a homogenous test phantom. The nominal beam energy is 120 kVp and scanning mode is helical. (a) Pitch = 1.25; (b) pitch = 1.0; (c) pitch = 0.8.

700 Tube current (mA)

600 500 400 300 200 100 0 0 (a)

50

100 150 200 250 Table position (mm)

300

350 (b)

FIGURE 22.2 Variable tube current simulation through the sagittal plane of the CTDI body phantom. (a) Tube current as a function of table position. (b) The sagittal dose distribution through the CTDI body phantom. The simulation is based upon a tube voltage of 120 kVp.

Handbook of Anatomical Models for Radiation Dosimetry

516

1.2 1.0 MCNPX Pencil chamber

0.8 0.6 0.4 0.2 0.0 0

5

10 15 Distance (cm)

20

25

FIGURE 22.3 In-air profile benchmark comparing pencil chamber exposure measurements versus the MCNPX simulation. The results have been normalized to the central axis of the fan beam.

cylindrical chamber with an active length of 100 mm surrounded by an air-equivalent plastic wall. The CTDI head and body phantoms are cylindrical phantoms made of polymethylmethacralate (PMMA), 15 cm in length with a diameter of 16 and 32 cm, respectively. Each phantom has five cylindrical sockets that will accept a PMMA insert or the pencil ion chamber. Standard CTDI measurements are typically presented in units of air kerma per mAs value for a given kVp and nominal slice thickness. All Monte Carlo codes provide tally results normalized on a per source particle basis and a conversion factor is necessary to provide an absolute dose normalization. We have previously reported on an in-air normalization method that is based upon pencil-chamber exposure readings taken at the center of the CT gantry.25 The CTDI100, in units of mGy per mAs, is calculated based upon Equation 22.1. CTDI100 =

(0.876) ⋅ X ⋅ L NBW ⋅ (mAs per rotation)

(22.1)

The value 0.876 represents the conversion from exposure to collision kerma in air, X is the exposure reading (mR), and NBW is the nominal beam width (mm) and L is the length of the pencil ionization chamber (typically 100 mm). The mAs per rotation is based upon the scanner mA setting and the time in seconds for a single axial rotation. The corresponding CTDI100 value can be efficiently calculated with Monte Carlo methods by using an energy fluence tally and multiplying by the corresponding mass energy-absorption coefficient for air. The simulated result will have units of mGy per source particle. The normalization factor (source particle per mAs) is defined using Equation 22.2 and is a function of scanner kVp and NBW. (NF)E ,NBW =

(CTDI100,measured )E,NBW (CTDI100,simulated )E,NBW

(22.2)

Applications to Computed Tomography for Adult Patients

517

TABLE 22.1 Calculated Normalization Factors Acquired for a Single Axial Scan from a Commercial MDCT Scanner kVp

Measured CTDI100 (mGy/mAs)

Calculated CTDI100 (mGy/History)

Monte Carlo 1s Standard Error

Normalization Factor

80

0.058

2.61E−12

0.0050

2.235E+10

100

0.117

2.52E−12

0.0049

4.654E+10

120

0.192

2.52E−12

0.0049

7.610E+10

140

0.286

2.58E−12

0.0049

1.108E+11

Note: Exposures were acquired in air at the scanner isocenter for beam energies of 80, 100, 120, and 140 kVp and a 24 mm nominal slice thickness. The tube current was kept constant at 200 mA and the rotation time was set to 0.5 s.

Table 22.1 provides an example data set for calculating the normalization factor from a commercial multidetector scanner. For any arbitrary CT dosimetry simulation, one should multiply the Monte Carlo results by the total number of mAs to obtain the absolute dose. In the case of a contiguous axial or helical protocol, the total number of mAs is equal to the mAs per rotation times the number of rotations (Equation 22.3).

(Dabsolute )E,NBW = (NF)E,NBW × (Dsimulated )E,NBW × (Total mAs)

(22.3)

22.3 Simulation Benchmarks With the CT tube in a fixed position at 90°, in-air profile measurements were taken to verify the shape of the bow-tie filter. The pencil ion chamber was positioned at the center of the CT gantry and exposure measurements were taken as a function of vertical table position. The identical simulation was performed to compare the measured versus simulated in-air profile shape. Figure 22.3 illustrates this comparison for a commercial scanner and body bow-tie filter with good agreement across the entire profile. The CTDI head and body phantoms are modeled including the five pencil-chamber inserts. One insert is positioned at the center of the phantom and the other four are positioned 1 cm below the surface at 12:00 (anterior), 3:00 (left lateral), 6:00 (posterior), and 9:00 (right lateral). The inserts can be filled with PMMA or by modeling the ion chamber air cavity depending on the position for the corresponding physical measurement. Table 22.2 provides a comparison data set with measurements obtained at the center and 12:00 positions. The agreement is good between measurement and simulation for all data points.

22.4 Experiments Using Voxelized Phantom Data We have previously compared measured versus simulated dose distributions using cylindrical and physical anthropomorphic phantoms scanned on a commercial MDCT

Handbook of Anatomical Models for Radiation Dosimetry

518

TABLE 22.2 Comparison of Measured versus Calculated CTDI100 Values Taken at the Center and 12:00 Position of the 32 cm CTDI Body Phantom kVp 80 100 120 140

Phantom Position

Measured CTDI100 (mGy/mAs)

Center 12:00 Center 12:00 Center 12:00 Center 12:00

0.99 2.03 2.32 4.38 4.13 7.70 6.61 11.8

Calculated CTDI100 (mGy/mAs) 0.99 2.07 2.34 4.61 4.21 7.96 6.70 12.3

Monte Carlo 1s Standard Error

% Difference

0.0094 0.0053 0.0085 0.0050 0.0080 0.0048 0.0076 0.0046

0.4 2.2 0.7 5.3 2.0 3.4 1.4 4.6

Note: Measurements were acquired at the scanner isocenter for beam energies of 80, 100, 120, and 140 kVp for a 24 mm nominal slice thickness. The tube current was kept constant at 200 mA and the rotation time was set to 1 s.

scanner.27 This was done by comparing physical surface measurements using MOSFET detectors. The MOSFET measurements (TN1002RD, Thomson-Nielsen, Ottawa, Ontario, Canada) were performed using 20 detectors and were made primarily to obtain profiles along the longitudinal (z) axis based upon a detector spacing of 1.6 mm intervals. The MOSFET detectors were normalized based upon an in-air exposure measurement (120 kVp, 4 × 5 mm beam collimation, body bow tie, 100 mAs) and the pencil ionization chamber. For each in-air measurement, a MOSFET detector was placed at isocenter and the corresponding pencil ionization chamber measurement was used to create a calibration factor based upon the ratio of the pencil-chamber reading versus the MOSFET signal with units of cGy per mV per mAs. A scanner technique of 120 kVp results in a MOSFET calibration factor of 0.069 ± 0.012 cGy/(mV mAs) averaged over a set of 20 detectors. The MOSFET detectors were not explicitly modeled when comparing the air-kerma variation along the surface of the CTDI body phantom. Using the cylindrical geometry of the CTDI phantom, the air-kerma MCNPX tally was performed in a cylindrical annulus at the surface of the phantom with an annulus thickness of 2 mm and a longitudinal spacing of 2 mm. The scanning parameters for this comparison were 120 kVp, 440 mA, 4 × 5 mm beam collimation, and 1 s rotation time. Figures 22.4 and 22.5 illustrate the comparison of the measured MOSFET values obtained at the surface of the phantom to those obtained by the MCNPX simulation for a contiguous axial and helical (pitch = 1.375) scanning protocol, respectively. Both measured and simulated results demonstrate approximately the same degree of variation from peak to trough with approximately the same period of variation. The results from the Monte Carlo simulations are much smoother as they represent ideal detectors, do not reflect any interdetector variations and can be obtained at a very large number of sample points rather than the limit of 20 MOSFET detectors we have for the physical measurements. The Monte Carlo results are based upon a percent relative error of less than 2% for all data points. The MOSFET measurements are presented with a y-axis mean relative error of approximately 0.05 based upon reproducibility measurements and an x-axis uncertainty based upon an approximate positioning error of ±1 mm relative to its adjacent neighbor(s).

Applications to Computed Tomography for Adult Patients

519

16.0 Simulation (Monte Carlo) MOSFET measurement (axial scan)

Surface dose (mGy per 100 mAs)

14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0 –2

–1

0

1

2

3

4

Distance from the CTDI phantom centerline (cm) FIGURE 22.4 Comparison of measured and simulated surface dose measurements along the longitudinal axis of the body CTDI phantom resulting from a contiguous axial scan. 16.0 Simulation (Monte Carlo) MOSFET pitch =1.375

Surface dose (mGy per 100 mAs)

14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0 –3

–2 –1 0 1 2 Distance from the CTDI phantom centerline (cm)

3

FIGURE 22.5 Comparison of measured and simulated surface dose measurements along the longitudinal axis of the body CTDI phantom resulting from a helical scan, pitch = 1.375.

22.5 Experiments Using Voxelized Patient Data We have previously studied the effects of patient size on organ dose from CT scan protocols from a MDCT scanner.28 The GSF family of voxelized patient models and the Visible Human model were implemented as input files using MCNPX. The relevant organs are

Handbook of Anatomical Models for Radiation Dosimetry

520

based upon the appropriate ICRP 60 tissue-weighting factors and the remainder organs (adrenals, brain, extrathoracic airways, small intestine, kidney, muscle, pancreas, spleen, thymus, and uterus) are assigned a common organ index and tissue material.1 Simulated CT scans of each voxelized patient model were performed using the MDCT source model, including the scanner-specific spectra, bow-tie filter, scanner geometry, and helical source path. The scan simulations in this work include a whole-body scan protocol and a thoracic CT scan protocol, each performed with fixed tube current. The whole-body CT scan simulation is based upon a tube voltage of 120 kVp, 4 × 5 mm collimation, and table pitch of 1. Figures 22.6 and 22.7 illustrate the effective dose and absolute dose of individual organs as a function of increasing patient weight for the wholebody scan protocol. A comparison with a conventional dose estimation method using the ImPACT spreadsheet yielded an effective dose of 0.14 mSv/mAs for the whole-body scan. This result is lower than the simulations on the voxelized model designated “Irene” (0.15 mSv/mAs) and higher than for the models “Donna” and “Golem” (0.12 mSv/mAs). Figure 22.7 illustrates the absolute dose for a few key organs (all having a tissue-weighting factor of 0.12 for the effective dose calculation) for all GSF models undergoing the wholebody scan. The lung doses range from 0.27 mGy/mAs (Baby) to 0.10 mGy/mAs (Visible Human) with the stomach and colon doses displaying similar ranges and trends with patient size. The thoracic scan simulation is based upon a tube voltage of 120 kVp, 4 × 5 mm collimation, and table pitch of 1.375 with the scan volume extending from 1 cm superior to the apex of the lungs to 1 cm inferior to the lung base. Figure 22.8 illustrates the effective dose as a function of increasing patient weight for the thoracic scan protocol. The values range from 0.047 mSv/mAs (Baby) to 0.020 mSv/mAs (Helga) suggesting a general trend in effective dose varying with patient size though there are exceptions as evidenced with Child and Golem. For the thoracic scan protocol, the ImPACT spreadsheet estimates an effective dose of 0.037 mSv/mAs, which falls between the calculated values for Irene (0.042 mSv/mAs)

Effective dose (mSv/mAs)

0.25

0.20

0.15

0.10

0.05

um sH Vi

an k Fr

ga el H

na on D

ol em G

en e Ir

ild Ch

Ba b

y

0.00

FIGURE 22.6 Plot of normalized effective dose (mSv/mAs) for each of the GSF models resulting from whole-body scan. In general, patient size increases from left to right.

Applications to Computed Tomography for Adult Patients

521

0.30 Lung Stomach

Mean organ dose (mGy/mAs)

0.25

Colon Marrow

0.20

0.15

0.10

0.05

0.00 Baby

Child

Irene

Golem Donna

Helga

Frank VisHum

FIGURE 22.7 Radiation doses (mGy) to organs with tissue-weighting factors of 0.12 from the whole-body CT scan. Patient weight increases from left to right. (Reproduced from DeMarco, J.J. et al., Phys. Med. Biol., 52, 2583, 2007. With permission.)

Eﬀective dose (mSv/mAs)

0.05

0.04

0.03

0.02

0.01

Vi sH um

nk Fr a

a H el g

na on D

em G

ol

e Ir en

ild Ch

Ba by

0.00

FIGURE 22.8 Plot of normalized effective dose (mSv/mAs) for each of the GSF models resulting from thoracic scans.

and Donna (0.031 mSv/mAs) and is higher relative to Golem (0.025 mSv/mAs). Figure 22.9 illustrates the radiation dose to several key organs (lung, thyroid, esophagus, and stomach) for each GSF model undergoing the thoracic scan simulation. The lung dose also demonstrates a general trend toward decreased dose with increased patient size with doses ranging from 0.15 mGy/mAs (Baby) to 0.05 mGy/mAs (Helga). Trends of radiation dose as a function of patient size are less obvious with the other selected organs, individual organ

Handbook of Anatomical Models for Radiation Dosimetry

522

0.16

Lung

Mean organ dose (mGy/mAs)

0.14

Thyroid Esophagus

0.12

Stomach 0.10 0.08 0.06 0.04 0.02

Vi sH um

k Fr an

a H el g

na on D

em G

ol

e Ir en

ild Ch

Ba by

0.00

FIGURE 22.9 Radiation doses to several key organs (all with weighting of 0.12 or 0.05 for effective dose calculation) for all GSF models from a thoracic CT scan. Patient weight increases from left to right. (Reproduced from DeMarco, J.J. et al., Phys. Med. Biol., 52, 2583, 2007. With permission.)

results being more indicative of that individual’s organ and its spatial relationship to the scan region. Certain organs may only be partially exposed in the thoracic scan relative to the location of the organ with respect to the scan’s start and stop locations. This will vary from individual to individual, even when using the same relative start and stop locations among all patient models and can also be influenced by the amount of overscan in the individual exam as well. The lung dose values for the thoracic scan protocol are lower (on a per mAs basis) than those of the whole-body scan for two major reasons: (1) the higher pitch value used in the thoracic protocol (1.375, compared to pitch 1 for whole-body scans) and (2) the fact that, while the lung is irradiated by primary radiation, the scan starts 1 cm superior to the lung apices and stops 1 cm inferior to the lung bases, thereby limiting scatter to the lungs from adjacent tissues. For the other organs, significant variations are found, which are due to variations in patient size as well as the organ’s location with respect to the scan’s start and stop locations.

22.6 Conclusion A comprehensive Monte Carlo-based CT scanner model has been developed that will simulate axial, helical, and variable tube current scan protocols. The CT simulation model has been applied to a family of patient-specific voxelized models in order to estimate absolute organ dose and whole-body effective dose as a function of patient size from arbitrary CT scan protocols. The model has been benchmarked using a variety of methods including CTDI phantoms, in-air profiles, and surface dosimetry using MOSFET detectors. This work has shown that the radiation dose patterns in CT can be complex, even for contiguous axial scans in homogeneous phantoms. To obtain accurate results from simulations,

Applications to Computed Tomography for Adult Patients

523

detailed models must be employed to accurately capture these variations. The integrated simulation model is used to provide estimates of both individual organ and effective doses for patient-based models of different sizes, using a detailed model of an MDCT scanner and simulating a whole body and thoracic scanning protocol. These methods were also recently applied to estimate fetal radiation dose from MDCT scans using models of pregnant patients of different gestational ages.29 Future work will extend this analysis and demonstrate the ability to estimate both individual organ and effective doses from different CT scan protocols on individual patient-based models and to provide estimates of the effect of patient size on these dose metrics.

Acknowledgment This work was funded in part by a grant from the National Institute of Biomedical Imaging and Bioengineering (NIBIB) 5R01EB004898.

References 1. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Pergamon Press, Oxford, 1991. 2. ImPACT. Imaging performance assessment of CT Scanners: A medical devices agency evaluation group. CT scanner matching data, tables of CTDI values in air, CTDIw, and phantom factor values. ImPACT Internet home page (http://www.impactscan.org), November 2006. 3. Jones, D.G. and Shrimpton, P.C. Survey of the practice in the UK. Part 3: Normalized organ doses calculated using Monte Carlo techniques, NRPB R-250, National Radiological Protection Board, Chilton, Didcot, 1991. 4. Jones, D.G. and Wall, B.F. Organ doses from medical x-ray examinations calculated using Monte Carlo techniques, Report NRPB-R186, National Radiological Protection Board, Chilton, Didcot, 1985. 5. Kramer, R. et al. The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods: Part I. The male (ADAM) and female (EVA) adult mathematical phantoms, GSF-Report S-885, Institut fuer Strahlenschutz, GSF-Forschungszentrum fuer Umwelt und Gesundheit, Neuherberg-Muenchen, 1982. 6. Zankl, M. et al. Organ doses for children from computed tomographic examinations, Radiat Prot Dosim, 57, 393, 1995. 7. Caon, M., Bibbo, G., and Pattison, J. An EGS4-ready tomographic computational model of a 14-year-old female torso for calculating organ doses from CT examinations, Phys Med Biol, 44, 2213, 1999. 8. Caon, M., Bibbo, G., and Pattison, J. Monte Carlo calculated effective dose to teenage girls from computed tomography examinations, Radiat Prot Dosim, 90, 445, 2000. 9. Khursheed, A. et al., Influence of patient age on normalized effective doses calculated for CT examinations, Br J Radiol, 75, 819, 2002. 10. Schmidt, B. and Kalender, W.A. A fast voxel-based Monte Carlo method for scanner- and patient-specific dose calculations in computed tomography, Phys Med, 18, 43, 2002. 11. Tzedakis, A. et al. The effect of z overscanning on patient effective dose from multidetector helical computed tomography examinations, Med Phys, 32, 1621, 2005.

524

Handbook of Anatomical Models for Radiation Dosimetry

12. Castellano, A., Dance, D.R., and Evans, P.M. CT dosimetry: Getting the best from the adult Cristy phanotm, Radiat Prot Dosim, 114, 321, 2005. 13. Salvado, M. et al. Monte Carlo calculation of radiation dose in CT examinations using phantom and patient tomographic models, Radiat Prot Dosim, 114, 364, 2005. 14. Lee, C. et al. Organ and effective doses in pediatric patients undergoing helical multislice computed tomography examination, Med Phys, 34, 1858, 2007. 15. Pelowitz, D., ed. MCNPXTM user’s manual version 2.6.0, LA-CP-07-XXXX, Los Alamos National Laboratory, Los Alamos, NM, 2007. 16. Waters, L., ed. MCNPX Version 2.5.C, LA-UR-03-2202, Los Alamos National Laboratory, Los Alamos, NM, 2003. 17. Hubbell, J.H. and Seltzer, S.M. Tables of x-ray mass absorption coefficients and mass energyabsorption coefficients (version 1.03) [online]. Available http://physics.nist.gov National Institute of Standards and Technology, Gauthersbury, MD, 1995. 18. Spitzer, V.M. and Whitlock, D.G. Atlas of the Visible Human Male, Jones & Bartlett Publishers, Sudbury, MA, 1998. 19. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Phys, 78, 476, 2000. 20. Petoussi-Henss, N. et al. The GSF family of voxel phantoms, Phys Med Biol, 47, 89, 2002. 21. Zankl, M. et al. Organ dose conversion coefficients for external photon irradiation of male and female voxel models, Phys Med Biol, 47, 2367, 2002. 22. Lee, C. et al. Whole-body voxel phantoms of paediatric patients—UF Series B, Phys Med Biol, 51, 4649, 2006. 23. Lee, C. et al. The UF series of tomographic computational phantoms of pediatric patients, Med Phys, 32, 3537, 2005. 24. DeMarco, J.J., Solberg, T.D., and Smathers, J.B. CT-based Monte Carlo simulation tool for dosimetry planning and analysis, Med Phys, 25, 1, 1998. 25. Jarry, G., Demarco, J., and McNitt-Gray, M. Monte Carlo dose verification of a commercial CT scanner with applications for patient specific dosimetry, Med Phys, 29, 1344, 2002. 26. McCollough, C.H., Bruesewitz, M.R., and Kofler, J.M. CT dose reduction and dose management tools: Overview of available options, Radiographics, 26, 503, 2006. 27. DeMarco, J.J. et al. A Monte Carlo based method to estimate radiation dose from multidetector CT (MDCT): Cylindrical and anthropomorphic phantoms, Phys Med Biol, 50, 3989, 2005. 28. DeMarco, J.J. et al. Estimating radiation doses from multidetector CT using Monte Carlo simulations: Effects of different size voxelized patient models on magnitudes of organ and effective dose, Phys Med Biol, 52, 2583, 2007. 29. Angel, E. et al. Radiation dose to the fetus for pregnant patients undergoing multidetector CT imaging: Monte Carlo simulations estimating fetal dose for a range of gestational age and patient size. Radiology, 249(1), 220–227, 2008.

23 Applications to Optimization of X-Ray Radiographic Imaging Birsen Yazıcı, Il-Young Son, An Jin, and X. George Xu

CONTENTS 23.1 Introduction ............................................................................................................... 525 23.2 Computational Phantom Generation and Effective Dose Calculations............ 526 23.2.1 Background of ViPRIS................................................................................. 526 23.2.2 The Process of X-Ray Image Generation .................................................. 527 23.2.3 Calculation of Effective Dose ..................................................................... 529 23.3 Image Quality Analysis ........................................................................................... 530 23.3.1 Observer Computational Phantoms ......................................................... 530 23.3.1.1 Non Prewhitening Match Filter with Eye Filter....................... 530 23.3.1.2 Hotelling Observer....................................................................... 531 23.3.1.3 Channelized and Laguerre–Gauss Hotellling Observers...... 531 23.3.2 Task Descriptions ......................................................................................... 532 23.3.2.1 SKE/BKE........................................................................................ 532 23.3.2.2 MAFC ............................................................................................. 533 23.3.2.3 SKEV and SKS............................................................................... 533 23.3.2.4 BKS .................................................................................................534 23.3.3 Performance Measures ................................................................................534 23.4 Data Generation and Optimization........................................................................ 535 23.5 Results and Discussion ............................................................................................ 538 23.6 Conclusions ................................................................................................................542 Acknowledgments ...............................................................................................................546 References .............................................................................................................................546

23.1 Introduction In the United States, approximately 250 million radiological examinations are performed each year, making diagnostic medical examinations the largest source of man-made radiation exposure.1 A major goal of radiography is to maximize the amount of diagnostic information while minimizing the radiation exposure to the patient. All radiographic x-ray examinations require the selection of beam parameters, which affect both the patient dose and the corresponding image quality. Optimization is difficult because

525

526

Handbook of Anatomical Models for Radiation Dosimetry

1. The x-ray imaging chain contains a large number of variables associated with the x-ray beam characteristics, the patient anatomy, the scatter removal grid, and the x-ray detector system. 2. Diagnostic x-ray examinations involve various body sites and tasks. 3. The effective dose is the sole indicator of patient risk but is not obtainable without a whole-body computational phantom of delineated organs. 4. An observer needs to evaluate the images with respect to quantitative image quality criteria, and should determine if the images meet the criteria. We previously developed the virtual photographic and radiographic imaging simulator (ViPRIS)2 to model the generation of radiography. At the core of this simulator are the image sets that have been made available to the public by the U.S. National Library of Medicine’s Visible Human Project (VHP).3 The user is permitted to specify different grid filtering parameters, detector efficiency, and nodules of varying composition and their location. A user-friendly software interface permits a user to develop thousands of simulated images and data sets to investigate the relationship between lesion detection threshold, lesion characteristics, patient dose, detector efficiency, and x-ray beam characteristics. The image quality optimizing technique optimizes the imaging chain by providing satisfactory image quality, while at the same time minimizing the patient’s dose. The image quality measure is based on the explicit calculation of optical density differences (using Monte Carlo simulations). Computerized observers evaluate the image quality in series of simulated experiments. Observer studies can be performed by either human volunteers or computerized observers. Optimizing the imaging process requires repeating observer studies many times with a large number of images until the ideal settings can be determined. Most such studies in the past utilized approximately 2,500,000 images. Such a large-scale survey is untenable using human volunteers. Alternately, computerized observer studies have, in recent years, been applied to x-ray image quality analysis. These techniques assess quality by how well a simulated observer detects the presence (or absence) of a lesion. Several different types of simulated observers have been developed, including Hotelling, channelized Hotelling, and non prewhitening matched filter with eye filter (NPWE).4–6 Optimization is performed with respect to two opposing objectives: 1. To maximize the detectability of a lesion 2. To minimize the dosage administered to the patient

23.2 Computational Phantom Generation and Effective Dose Calculations 23.2.1 Background of ViPRIS ViPRIS represents a first-generation imaging simulator that strives to produce realistic images using simulation software that authentically reproduces the process of creating a radiographic image.7 At the core of the VIPRIS are the image sets that have been made available to the public by the VHP. ViPRIS utilizes two phantoms constructed from the identical anatomy of the VHP. The simulated production of images is split into two steps. An image projection phantom constructed from the computed tomographic (CT) data set is used to

Applications to Optimization of X-Ray Radiographic Imaging

527

simulate the primary x-rays using a ray-tracing technique. We simulated scattered x-rays using the Visible Photographic Man (VIP-Man) phantom previously constructed from segmented color cryosectional photographs of the VHP.8 The VIP-Man phantom was originally developed to calculate radiation doses to an adult male. Image data from the VHP includes four sets of images. CT, magnetic resonance imaging (MRI), and x-ray images were obtained from the fresh body. The body was then frozen to allow 1 mm thick slices to be removed from the entire body for cross-sectional color photographs. The VHP resulted in a total of 1878 transverse color photographs to form a digital atlas of what was called the “Visible Man.”3 At Rensselaer Polytechnic Institute, segmentation and classification were performed to adopt 72 radiosensitive organs and tissues for radiation dosimetry studies in EGSnrc.8 Values of tissue densities and compositions for radiation dosimetry calculations were based on those recommended in ICRP 23.9 The original voxel size of the VIP-Man computational phantom was 0.33 mm × 0.33 mm × 1 mm. The whole body was represented by a total of 3,384,606,720 voxels. To reduce the computational burden, a simplified version of VIP-Man with a resolution of 1 mm × 1 mm × 1 mm was used for this study. We used the EGSnrc code in our study to calculate organ doses as well as to develop the scatter images. Because patients were instructed to lift their arms out of the field for this study, modifications were made to the VIP-Man phantom to remove the arms. 23.2.2 The Process of X-Ray Image Generation The generation of simulated x-ray images is done in two phases. For the given x-ray characteristics (such as the number of photons and energy), ViPRIS calculates the primary x-ray intensity (i.e., transmitted x-rays) by tracing each x-ray through the image projection phantom. ViPRIS then scales and calculates the scattered x-rays based on the stored data files from pregenerated Monte Carlo simulations. We then smooth the scattered data set to remove excess noise caused by statistical variations inherent in Monte Carlo simulations. Following this, we then modify each data set based on the specified detector quantum efficiency. We add quantum noise to the primary and scattered images using Box–Muller approximation of a Gaussian distribution as described in [7,10]. Finally, the two images are combined to form the simulated radiograph. The flow of the process is illustrated in Figure 23.1. The details of each step of the process are given in the following. The primary portion of an x-ray image is formed by photons that transmit through patient anatomy without interactions. The visible Human CT data can be converted into tissuespecific attenuation coefficients. Using the tissue-specific attenuation coefficient data and Beer–Lamberts law, we calculate the number of transmitted photons for each pixel using the fluence (i.e., the number of photons per unit area) specified by the user. To simulate lesions, ViPRIS replaces the attenuation coefficients of certain voxels within the selected chest region of the CT with those of lesions. The software allows a user to select a variety of lesion geometries and compositions in the form of a lesion Houndsfield units (HU).7 When an x-ray photon enters the human body, there is a probability that the photon may undergo scattering and move toward the film or detector grid, which results in added random background noise to the primary x-ray image. Monte Carlo techniques are ideally suited for modeling this random scattering process. Because the probability of Compton scattering is relatively small for x-rays and a large number of particle histories are necessary to reduce the statistical uncertainty inherent in the Monte Carlo method, this process was extremely time consuming and required ~ 2 days for simulation over 40 keV, and 10 days for 30 and 35 keV. Therefore, we chose to simulate the energy once and store the resulting image. A sample image can be seen in Figure 23.2. To further reduce the statistical

Handbook of Anatomical Models for Radiation Dosimetry

528

User selects parameters: keV, fluence, lesion, grid, and detector Calculate and store dose information and scatter profiles for all energies

Generate scatter portion of image

Generate primary portion of image

Use ESG4nrc to simulate radiograph of VIP-Man phantom

Rescale number of particles based on user supplied fluence values

Set incident photon intensity (Ie) equal to fluence value selected

Record energy deposited within VIP-Man and determine the mumber of particles that reach the detector Calculate effective dose per particle

Blur/smooth data Account for the efficiency of the detector Account for the attenuation of the grid Add Gaussian noise (Box-Muller) Multiply number of particles by the average scattered energy (move to energy domain)

Calcuate the transmission through the CT phantom and record Ip at the image plane Account for the attenuation of the grid Account for the efficiency of the detector Add Gaussian noise (Box–Muller) Multiply number of particles by the energy (move to energy domain)

Combine energies and scale to log (move to log energy domain) Calculate gray scale of image ranging from 0 to 255 display and save image, produce vectorized image of lesion area Use vectorized image and simulated ideal observers and calculate the quality of images Determine the minimum effective dose required to meet image quality criteria FIGURE 23.1 Flowchart of x-ray simulation and optimization process using ViPRIS and image quality analysis. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

uncertainty (mottle or noise), we smoothed, or denoized, each data set using a 24 × 24 boxcar filter.10 Once we calculated the primary and scatter images, ViPRIS accounts for various imaging components detailed below, and forms a final x-ray image. Prior to any operations the scatter data is scaled to match the fluence specified by the user. ViPRIS uses an approximation of a Gaussian distribution to add simulated quantum noise to the primary and scatter image data. We used the square root of the number of photons incident on a pixel to simulate the standard deviation of noise, as governed by Gaussian statistics. Because the quantum absorption efficiency (detection efficiency) of the detector is not unified,

Applications to Optimization of X-Ray Radiographic Imaging

529

FIGURE 23.2 A sample output of ViPRIS showing a simulated chest x-ray image. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

we use the user-specified efficiency for the primary and average scattered energies to modify the number of particles detected in all scattered and primary pixels. This allows for the simulation of any detector as well as x-ray system. To convert the photon fluence of the primary image into energy fluence, we multiply the number of surviving particles in each pixel by the user-specified energy. For the scattered portion the same procedure is followed; however, in this case the average scatter particle energy is used. To combine the scatter and primary portions of the radiograph, we utilized the specifications of the user-defi ned grid. ViPRIS simulates the presence of a grid by allowing the user to specify the transmission percentages of the primary and scattered image. We then scanned the data set for the lowest and highest numbers. We subtracted the lowest log energy fluence from each data point. A new maximum is calculated and divided by 255. We converted the data into a gray scale value between 0 and 255. 23.2.3 Calculation of Effective Dose Organs receive x-ray doses from the secondary electrons resulting from photoelectron absorption and Compton scattering. The energy deposition in each of the major organs of the VIP-Man phantom is saved during the formation of the scatter profile. The mean absorbed dose in an organ or tissue (DT) is calculated as the total energy deposited in organ T divided by the organ mass. The equivalent dose (HT) in organ T is calculated by multiplying the mean absorbed dose by the radiation-weighting factor, wR, which is unity for photons and electrons. Since the same equivalent dose value can cause different risks in different organs or tissues, a tissue-weighting factor (wT), has to be applied to yield the total risk, in terms of the effective dose (E) using Equation 23.1. E=

∑w H T

T

(23.1)

530

Handbook of Anatomical Models for Radiation Dosimetry

A thin layer of fat tissue around chest level is used as the male breast. Ten more organs and tissues are included in the remainder group of organs which together shares a total tissue-weighting factor of 0.05.11 Monoenergetic photons were considered in ViPRIS for dose calculations. The user specifies the energy to be used for each simulation. For each pregenerated scattered image there is a corresponding dosimetry data set. When the scatter data file is loaded, the researcher also specifies the associated dosimetry data set. The data set contains conversion factors for effective dose per particle at the specified energy. Since the user specifies a fluence to be simulated, he/she can calculate an effective dose by multiplying the dose conversion factor by the fluence.12

23.3 Image Quality Analysis In recent years, there has been extensive work on simulating the response of observers using computational phantoms. Image quality analysis is one tool for selecting optimal fluence and energy levels for the purposes of detecting lesions while keeping the patient dose at a minimum. 23.3.1 Observer Computational Phantoms The observer’s task is to take an image and decide whether the image contains a lesion or not. In general, such tasks are known as signal detection tasks. Each observer makes its decision based on a test statistic, λ, calculated using the input image and statistics derived from a given set of training data. In the case of linear observer computational phantoms, the equation for λ is simply λ = wTg

(23.2)

where w is the template used by the observer g is the vectorized test image The decision is made by comparing λ to a threshold, i.e., if λ is below the threshold, the image is classified as background; if not, it is classified as containing a lesion. 23.3.1.1 Non Prewhitening Match Filter with Eye Filter A natural conclusion to draw from the reduced efficiency of the human observer is that the human observer is unable to perform the prewhitening operation. The non prewhitening match filter (NPWMF) computational phantom predicts human performance in both correlated and uncorrelated noise. In signal-known-exactly/background-known-exactly (SKE/BKE) tasks, the NPWMF computational phantom correlates well with psychophysical data.13 The template of NPWMF is wNPWMF = ( g1 − gb )

(23.3)

Applications to Optimization of X-Ray Radiographic Imaging

531

and the test statistics are λ = ( g l − g b )T g

(23.4)

By combining an eye filter, an even larger body of human psychophysical data can be explained. This observer is called NPWE: wNPWE = ET ( gl − g b )

(23.5)

where E is a matrix containing the eye filter E( f ) = f ρ exp(−cf γ )

(23.6)

where f is the radial spatial frequency, c = 0.978, γ = 0.674, and ρ = 1.5034. The test statistics is λ = ( gl − g b )T EET g

(23.7)

23.3.1.2 Hotelling Observer Under the Gaussian noise assumption, a Hotelling observer is an optimal linear observer, maximizing the signal-to-noise ratio (SNR) of the test statistic. Researchers can derive it from the Bayesian decision rule, given that the noise is additive and Gaussian. The Hotelling observer uses fi rst- and second-order statistics (for both lesion and no-lesion images) to calculate its template. The template for Hotelling observer is given by5 wh = Σ −1( gl − g b )

(23.8)

where (−g1, Σ1) and (−gb, Σb) denote the sample mean and covariance matrices of the lesion and background images, respectively Σ = (Σ l + Σ b)/2 is the average of the two covariance matrices Σ is symmetric, so is Σ−1 Therefore, from Equation 23.2 λ = ( gl − g b )Σ −1 g

(23.9)

23.3.1.3 Channelized and Laguerre–Gauss Hotelling Observers Yao and Barrett14 showed that a modification to the Hotelling observer yields a good fit to a human observer. This modified observer is generally known as a channelized Hotelling observer. In this study, the Laguerre–Gauss channelized Hotelling observer is used as

Handbook of Anatomical Models for Radiation Dosimetry

532

described in15. In a generic channelized Hotelling observer, the images are first filtered through a set of frequency selective “channels.”13 The template is calculated based on the filtered training images, and then it is applied to the filtered test images. The psychophysical results for signal detection in a lumpy background correlates well with the channelized Hotelling computational phantom.14 Let U represent a matrix with each column equal to a channel, then the covariance matrix of the channelized images are Σ ch = U T ΣU

(23.10)

−1 T wch = Σ ch U ( gs − g b )

(23.11)

and the template is equal to

Using Equation 23.7, one can derive the test statistic λ of the channelized Hotelling observer as follows: −1 T λ = ( gs − g b )T U Σ ch U g

(23.12)

In this study, each filter is a Laguerre–Gauss function of different order. A Laguerre–Gauss function is nothing but the product of Gaussian function with the nth-order Laguerre polynomial U n (r, a ) =

⎛ −πr 2 ⎞ ⎛ 2πr 2 ⎞ 2 π exp ⎜ 2 ⎟ L n ⎜ 2 ⎟ a ⎝ a ⎠ ⎝ a ⎠

(23.13)

where Ln(x) is the nth-order Laguerre polynomial given by ⎛n ⎞ xm ( −1)m ⎜ ⎟ ⎝m⎠ m ! m =0 n

L n (x ) =

∑

(23.14)

23.3.2 Task Descriptions 23.3.2.1 SKE/BKE The simplest image quality analysis is the SKE/BKE task.16 In such a case, the researcher fixes each image location, i.e., the background remains fixed and the form of the “signal,” i.e., the lesion being detected, is known exactly prior to the analysis with no variations from trial to trial. The only variation comes from the random noise in the image due to quantum mottle. In this case, a single template is created for the computerized observers using statistics derived from the previously described lesion and background images. At each trial, the researcher presents the simulated observer with an image taken at the same location. The observer’s task is to determine whether or not a lesion is present in the image. To do this, the computerized observer sets a threshold value, say λ0, and applying a simple decision function. The decision function, D is a function of λ0 and is given by D (λ ; λ 0 ) = u (λ − λ 0 )

(23.15)

Applications to Optimization of X-Ray Radiographic Imaging

533

where u(x) is a unit step function which equals to 1 when x ≥ 0 and 0, otherwise. D(λ; λ0) = 1 indicates detection of a lesion. Due to the simplicity of the SKE/BKE task, it is not a good representative of real-world clinical situations. The task lacks the variability in the background and lesion that is present in a real-world situation. 23.3.2.2 MAFC Two-alternative forced-choice (2AFC) task and its generalization, M-alternative forcedchoice (MAFC) task extends SKE/BKE by introducing variations in the lesion location.16,17 In these tasks, an observer is presented with a set of M images each at different locations. The locations are fixed. Only one of these images contains the lesion. The form of the lesion (shape and size) being detected is known exactly and is fixed prior to analysis. The observer’s goal is to determine the location that contains the lesion. As with SKE/BKE, a single template is generated for the task based on training samples with various backgrounds. The observer can derive the optimal solution to MAFC task given that each locations are uncorrelated. The observer’s task procedure is to apply this template to each location and to select the one with the highest λ value to be the location with the lesion. 23.3.2.3 SKEV and SKS The MAFC task, although more realistic than SKE/BKE, still assumes a static lesion morphology. To implement a more realistic task, the MAFC can be further generalized to incorporate variations in the form of the lesion itself. There are two kinds of such tasks, signal-known-exactly but varied (SKEV) and signal-known-statistically (SKS). Both approaches have been studied18,19 and begin with a predetermined set of signals and backgrounds. Suppose J types of lesions, which may vary in any number of ways from shape or size to physiological characteristic of the lesion, are given and that M set of locations that may contain the lesion are identified. In the SKEV task, the observer is presented with the M locations and knows a priori for which lesion type to look. The observer’s goal is to identify which of the M locations contains this form of the lesion. As with MAFC, only one location contains the lesion at each trial. With computerized observers, this procedure translates to identifying the location that gives the highest λ value when applying a template specific to the type of lesion. Thus, a separate template must be generated for each lesion type using the appropriate set of training data. Although the SKEV task takes lesion variations into account, it does so with a priori knowledge about the type of lesion present for each trial. Thus, under the assumption that the locations are uncorrelated, SKEV is no different than an MAFC task at each individual trial. The introduction of the lesion variation from trial to trial, however, does affect the performance of the observer. The SKS task does not use this a priori information. The task remains fundamentally the same; that is, the observer must choose the location that contains the lesion. Now, however, the observer’s uncertainty about the lesion variation must also be taken into account. As with the SKEV task, a set of J templates are computed, one for each lesion type. However, unlike the SKEV task, the SKS is performed in two steps. First, the J × M set of test statistics are generated, one for each lesion type and for each location. Then these templates are combined to produce a “sum-of-likelihood” value for a lesion being present in a given location, where the sum is over all lesion types. Assuming J number of lesion variations and M pairwise uncorrelated locations, this likelihood for the ith location is given as J

li =

∑ j =1

M

⎡ 1 ⎤ ⎛ (λ i , j − λ 1, j )2 ⎞ ⎢ ⎥ exp ⎜ − ⎟ 2σ 2j ⎝ ⎠ ⎣⎢ σ j 2π ⎥⎦

⎡ ⎛ (λ m , j − λ 0, j )2 ⎞ ⎤ ⎢exp ⎜ − ⎟⎥ 2σ 2j ⎝ ⎠ ⎦⎥ ⎢ m =1 ⎣ M

∏

(1−δ m , j )

(23.16)

534

Handbook of Anatomical Models for Radiation Dosimetry

where δm, j denotes the Kronecker’s delta λ 1, j (λ 0, j ) is the statistical expectation of the test statistic if the jth type lesion is present (not present) λi, j is the test statistic at the ith location for the jth type lesion The location with the largest li is determined to be the location where the lesion is present. 23.3.2.4 BKS The background-known-statistically (BKS) task introduces the difference. One needs to introduce randomness in both the signal and the background (i.e., SKS/BKS). The stochastic background variation is also termed “clutter,” “object variability,” or “background variability.” One can achieve background variation through a statistically defined background, such as independent and identically-distributed (i.i.d) random noise or complex lumpy background. These statistically defined backgrounds can be treated analytically, as the computational phantom for lumpy background is K

B =b

∑ G (r , σ ) + B k

2

0

(23.17)

i =1

G(rk, σ2) is Gaussian blob with variation σ2 at random location rk, number of blobs K is a Poisson random variable, and b is a weight constant of this problem. 23.3.3 Performance Measures As a measure of quality for SKE/BKE and 2AFC images, researchers calculate and plot the receiver operator characteristic (ROC) curve and the area under the ROC curve (AUC) for each simulated observers, and then use these curves as measure of image quality. These curves have been studied extensively in other fields such as target detection in radar. The idea is to examine the AUC for varying fluence and energy levels to estimate the optimal dose to achieve desired performance in lesion detection. The ROC is a plot of sensitivity versus one minus the specificity as a function of threshold level. The sensitivity is given as the ratio between the number of true positives and the total number of actually positive cases Sensitivity =

NTP NTP + NFN

(23.18)

where NTP is the number of true positives NFN is the number of false negatives Specificity, on the other hand, is given by the ratio between the number of true negatives and the total number of negative results Specificity =

NTN NTN + NFP

(23.19)

Applications to Optimization of X-Ray Radiographic Imaging

535

where NTN is the number of true negatives NFP is the number of false positives A good measure of image quality is the AUC. A linear ROC curve with AUC of 0.5 indicates an observer whose decision is essentially random and suggests that image quality is poor; whereas a steep rise in ROC curve, corresponding to AUC close to 1, indicates high image quality, providing high sensitivity and specificity. The characteristics of the AUC suggest a probabilistic interpretation. In fact, if one interprets the test statistic score as a ranking criterion, it can be shown that AUC is equivalent to the probability of test statistic score for lesion ranking higher than the test statistic score for the background.20 The ROC and thus the AUC curves apply only to simple binary classification tasks, and cannot be constructed for MAFC, SKEV, and SKS, where the task detection of the location of a lesion varies, leaving more than two alternatives for an answer. Taking a cue from the probabilistic interpretation of the AUC, an alternative metric of image quality can be derived for MAFC, SKEV, and SKS tasks based on the probability of a lesion score ranking above the background. For MAFC, under the assumption that the M locations are uncorrelated, P is estimated as P=

1 N

N

∑ u (λ

1, n

− max(λ 0, n ))

(23.20)

n =1

where N is the total number of trials λ1,n denotes the test statistic for the location that has the lesion for the nth trial λ0,n, n denotes the test statistic for the locations without the lesion u(·) is the unit step function For the SKEV task, P is a weighted sum of Equation 23.15 calculated for each type of signal weighted by the estimated prior probability of each signal. To calculate P for the SKS task, one must first calculate the sum-of-likelihood for a signal being present at each of the M locations as given in Equation 23.12. Using the sum-of-likelihood, P is estimated by P=

1 N

N

∑ u (l

1, n

− max(l0, n ))

(23.21)

n=1

where N denotes the total number of trials.

23.4 Data Generation and Optimization For this study, we examined the effects of varying the incident energy and the effect on the amount of fluence required to meet a preset image quality threshold (IQT). The IQT

Handbook of Anatomical Models for Radiation Dosimetry

536

denotes the value of the image quality metric that the observer must perform in an equal to or better than fashion, in a given task. The detector quantum absorption efficiency (DQAE) was modeled after a cesium iodine screen as detailed.21 To take into account how absorption efficiency changes with energy, we changed the DQAE depending on the energy of the incident photons and the average energy of the scattered photons. Table 23.1 shows the DQAE value for each energy level. For the SKE/BKE task, the simulated lesion was kept at a constant location in the left lung clear of ribs with a diamond shape. Attenuation of the lesion was kept constant at 400 HU. We chose the HU value to represent a calcified lesion.22,23 For the MAFC task, we generated additional sets of data at three additional lesion locations for maximum of M = 4 locations. We assume that the locations were uncorrelated. The lesion size and HU value was kept constant, as in the SKE/BKE task. We also generated data sets with a smaller lesion for the SKEV task, for total of two possible lesion sizes. The set of locations used for these tasks was the same as those used in the MAFC task. The lesions were roughly diamond shaped. The larger lesion used in all the tasks was 11 pixels along the diagonal. The smaller lesion generated for the SKEV task was 7 pixels along the diagonal. Table 23.2 summarizes the locations, lesion types, and energy levels used for each task.

TABLE 23.1 DQAE in Percentage for Each Energy Level (keV) Energy DQAE Energy DQAE

30 56 75 34

35 91 80 31

40 88 90 23

45 77 100 18

50 68 110 14

55 60 120 10

60 51 130 7

65 46 140 5

70 40 150 4

Source: Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.

TABLE 23.2 Locations, Tumor Types, HU Value, and Energy-Level Range Used for Each Task Type

SKE/BKE 2AFC 3AFC 4AFC SKEV

Locations

Tumor Type

HU of Tumor

(1) (1), (2) (1)–(3) (1)–(4) (1)–(4)

(a) (a) (a) (a) (a), (b)

150, 400 400 400 400 400

Energy Range (keV) 30–150 30–150 30–150 30–150 30–150

Source: Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission. Each location is denoted by a number: (1) left lung, (2) right lung with obstruction by a rib, (3) heart, and (4) liver. Each tumor type (varied by the size) is denoted by a letter: (a) large (11 pixels) and (b) small (7 pixels).

Applications to Optimization of X-Ray Radiographic Imaging

537

We performed the optimization of the effective dose administered to the patient in the following three steps: 1. Constructed image quality curves for a fixed energy level as a function of fluence. As described previously, we used AUC as the image quality metric for SKE/BKE tasks requiring the construction of a series of ROC curves, each at a fixed fluence and energy level. A sample ROC curve is shown in Figure 23.3. The quality curve for SKE/BKE task was then the value of the AUC plotted against the varying fluence for each energy. For the MAFC, the SKEV, and the SKS tasks, the respective image quality metrics, as described in Section 23.3.3, were used in lieu of AUC. Figures 23.4 and 23.5 show the sample plots of these image quality curves for various tasks using Hotelling and Laguerre–Gauss Hotelling observers, respectively. Figure 23.4a shows the result for SKE/BKE task using a Hotelling observer. The figure shows that the performance is well done. Incorporating a more realistic observer computational phantom degrades the performance, as expected (see Figure 23.5a). The addition of more location (MAFC) and lesion variations (SKEV) also affects the performance, as seen in Figures 23.4b and c and 23.5b and c. 2. Examined for the lowest fluence level at predetermined IQT Once we generated the image quality curves, the next step is to examine these curves and identify the optimal fluence at each energy level. The optimality is defined as the lowest fluence at which the image generated meets a predetermined minimum IQT. A single optimal fluence level is identified per energy level. In general, raising the IQT increases the optimal fluence.

1 0.9 0.8

Sensitivity

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0

0.1

0.2

0.3

0.6 0.4 0.5 1-speciﬁcity

0.7

0.8

0.9

1

FIGURE 23.3 Plot of ROC curve for Laguerre–Gauss Hotelling observer at 70 keV and 1,000,000 m−2 fluence. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

Handbook of Anatomical Models for Radiation Dosimetry

538

1

1 SKE/BKE

0.85 0.8

0.85 0.8 0.75 0.7 0.65

0.75 0.7

2AFC 3AFC 4AFC

0.9

0.9

Quality (AUC)

Quality (AUC)

0.95

0.95

0.6 0

2

4

6

8

0.55

10 ×104

Fluence

0

2

4

6 Fluence

(a) SKE/BKE

8

10 ×104

Quality (AUC)

(b) MAFC 1 0.95 0.9 0.85 0.8 0.75 0.7 0.65 0.6 0.55 0.5

SKEV (M = 2) SKEV (M = 3) SKEV (M = 4)

0

(c) SKEV

2

4

6 Fluence

8

10 ×104

FIGURE 23.4 Plot of the quality measure versus fluence for various tasks performed. The plot shown are fixed at 50 keV. The plot shows the performance of Hotelling observer. The performance of the observer, i.e., the image quality metric, is, on average, decreased with addition of more location variations. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

3. Converted fluence level to effective dose levels The third step in the process is to convert the optimal fluence levels to their effective dose equivalent. The effective dose is a function of energy and the number of incident particles. The effective dose per incident particle for various chest x-ray configurations were previously calculated12 for various monoenergetic fields. These calculations were then used to estimate the effective dosage per unit area using the optimal fluences. Figure 23.6 shows the effective dosage per particle calculated at each energy level as found12. Using the effective dose calculations, the energy level and fluence at which the dose is minimized is then identified as optimal.12

23.5 Results and Discussion Figures 23.7 and 23.8 show the plots of optimal fluence versus energy level for various tasks using the Hotelling and Laguerre–Gauss Hotelling observers, respectively. In order

Applications to Optimization of X-Ray Radiographic Imaging

1

1

0.95

0.95 0.9

Quality (AUC)

Quality (AUC)

0.9 0.85 0.8 0.75 0.7 0.65 SKE/BKE

0.6

0.85 0.8 0.75 0.7 0.65 0.6

2AFC 3AFC 4AFC

0.55

0.55 0.5

539

0

(a) SKE/BKE

2

4

8

6

0.5

10

Fluence

×104

0

2

4

6

Fluence

(b) MAFC

8

10

×104

1 0.95

Quality (AUC)

0.9 0.85 0.8 0.75 0.7 0.65 0.6

SKEV (M = 2) SKEV (M = 3) SKEV (M = 4)

0.55 0.5

0

(c) SKEV

2

4

6

Fluence

8

10

×104

FIGURE 23.5 Plot of the quality measure versus fluence for various tasks performed. The plot shown are fixed at 50 keV. The plot shows the performance of Laguerre–Gauss Hotelling observer. The performance of the observer, i.e., the image quality metric, is, on average, decreased with addition of more location variations. Furthermore, compared with Figure 23.4, Laguerre–Gauss Hotelling performs worse than the Hotelling observer. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

to save space and minimize redundant information, we only show the results for SKE/ BKE, 3AFC, 4AFC, and SKEV (with M = 3). The following discussion applies to all other remaining tasks not shown. The performance and optimal dosage are influenced by three difference variables: the number of locations, the introduction of additional lesion type, and the IQT variations. The effect of each on optimal dosage will be discussed in turn. The graphs in Figure 23.7b are shifted up in fluence axis relative to graphs in Figure 23.7a. Similarly, the graphs in Figure 23.7c are shifted up relative to Figure 23.7b and a. The graphs in Figure 23.8 exhibit a similar pattern. It is clear from this pattern that the number of locations, M, affects the performance of the observer. This is not surprising, since increasing the number of locations increases the complexity of the task. More specifically, as the number of location variation increases, lesion detectability decreases. This pattern is also evident in the image quality plots of the Laguerre–Gauss Hotelling observer shown in Figure 23.5. To a lesser extent, this is also evident from the image quality plots of Hotelling observer shown in Figure 23.4. It is more difficult to differentiate between the performances of Hotelling observer for different values of M, due to the fact that the observer performs well in all tasks. This is to be expected since the Hotelling observer is optimal among all linear observers under Gaussian assumption. It is also established that the Hotelling observer outperforms human observers in all tasks.18 A more comparable

Handbook of Anatomical Models for Radiation Dosimetry

540

×10–13 3.5 AP PA RLAT

Eﬀective dose per particle

3 2.5 2 1.5 1 0.5 0 20

40

60

80 100 Energy (keV)

120

140

160

FIGURE 23.6 Effective dose (mSv) per incident particle for various chest x-ray examination. AP, anterior to posterior; PA, posterior to anterior; RLAT, right lateral. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

performance to human is achieved using Laguerre–Gauss Hotelling observer. The effect of the fluence requirement for larger M generally increases the dosage required. This can be assessed from comparing Figure 23.9a through c for the Hotelling obsesrver and Figure 23.10a through c for the Laguerre–Gauss Hotelling observer, which exhibits an analogous shift upwards in dosage. These results conform to the results from previous experiments with MAFC task.17,24–27 In all of these studies, it was found that the number of lesion locations degraded the performance of the computational phantom (and/or human) observers as observed in the present study. The degradation of the observer performance in turn increases the dosage requirement for target IQT. Increasing the IQT also affects the value of the optimal fluence. In all of the results for the Hotelling (Figure 23.7) and the Laguerre–Gauss Hotelling (Figure 23.8) observers, increasing the target IQT shifts the graph upwards in fluence axis. The reason is, it becomes more difficult for the computational phantom observer to achieve the desired level of performance as indicated by the IQT, i.e., the observer must detect more lesions correctly. The increased IQT hence increases the optimal fluence. The corresponding energy level at which the lowest optimal fluence is achieved, however, does not vary consistently with increased IQT (or with increased M). The results indicate that the energy level at which the lowest fluence is achieved ranges from 45 to 65 keV. The fluctuation in energy level corresponding to the optimal fluence may be due to the small number of energy samples (18 in all). Figures 23.9 and 23.10 show the conversion of the optimal fluences to their respective effective doses. Table 23.3 tabulates the lowest dosage in these graphs with their corresponding energy levels for several IQT values. The optimal dose

7

7

6.5

6.5

6

6

Log-fluence (m–2)

Log-fluence (m–2)

Applications to Optimization of X-Ray Radiographic Imaging

5.5 5 4.5 AUC - 0.8 AUC - 0.9 AUC - 0.95 AUC - 0.99

4 3.5

40

60

120

5.5 5 4.5

3.5

140

7

6.5

6.5

6

6

5.5 5 4.5 AUC - 0.8 AUC - 0.9 AUC - 0.95 AUC - 0.99

4

(c) AAFC

40

60

80 100 Energy (keV)

40

60

(b) 3AFC

7

3.5

AUC - 0.8 AUC - 0.9 AUC - 0.95 AUC - 0.99

4

Log-fluence (m–2)

Log-fluence (m–2)

(a) SKE/BKE

80 100 Energy (keV)

541

120

140

80 100 Energy (keV)

120

140

5.5 5 4.5 AUC - 0.8 AUC - 0.9 AUC - 0.95 AUC - 0.99

4 3.5

40

60

80 100 Energy (keV)

120

140

(d) SKEV (M = 3)

FIGURE 23.7 The logarithm of the optimal fluence (m−2) versus energy using Hotelling observer for (a) SKE/BKE, (b) 3AFC, and (c) SKEV (with M = 3) tasks. As the complexity of the task (i.e, number of location variations, number of lesion size variation) increases optimal fluence level also generally increases. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

is usually achieved at the energy level corresponding to the lowest fluence and spans the range of 45–65 keV. It is important to keep in mind, however, that there is a relatively positive relationship between the energy and the effective dose calculations as shown in Figure 23.6. The effect of this positive slope can be seen in some instances. For example, the graphs in Figure 23.7b show that the lowest fluence is achieved at around 55–60 keV for IQT value of 0.8. However, the lowest effective dose is actually achieved at 45 keV as shown in Table 23.3. The addition of lesion size variation further degrades the performance of the observers as task complexity is increased. This can be seen from the results in Figures 23.7 through 23.10, and in Table 23.3. These results show that the optimal fluence and effective dose for SKEV task rises relative to MAFC task with equal value of M. For example, comparing Figure 23.7b with Figure 23.7d, the fluence values are generally larger per energy level for SKEV task than for MAFC of equal M value. The performance degrading effect of introducing lesion size variation agrees with previous studies.19,28

Handbook of Anatomical Models for Radiation Dosimetry

7

7

6.5

6.5 Log-fluence (m–2)

Log-fluence (m–2)

542

6 5.5 5 4.5 AUC - 0.6 AUC - 0.7 AUC - 0.8 AUC - 0.9

4 3.5

40

60

80 100 Energy (keV)

120

5

7

6.5

6.5

6 5.5 5 4.5 AUC - 0.6 AUC - 0.7 AUC - 0.8 AUC - 0.9

4 40

40

60

80 100 Energy (keV)

120

140

(b) 3AFC

7

3.5

AUC - 0.6 AUC - 0.7 AUC - 0.8 AUC - 0.9

4.5

3.5

Log-fluence (m–2)

Log-fluence (m–2)

5.5

4

140

(a) SKE/BKE

6

60

80

100

120

140

6 5.5 5 4.5

AUC - 0.6 AUC - 0.7 AUC - 0.8 AUC - 0.9

4 3.5

40

Energy (keV) (c) 4AFC

60

80

100

120

140

Energy (keV) (d) SKEV (M = 3)

FIGURE 23.8 The logarithm of the optimal fluence (m−2) in log scale versus energy using Laguerre–Gauss Hotelling observer for SKE/BKE, 2AFC, 3AFC, and SKEV (with M = 3) tasks. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

To summarize, although the optimal effective dose has a clear positive correlation with the complexity of the task, the energy level at which the optimal effective dose value is achieved shows no such relationship in our simulations. Neither the increase in the number of background variations nor the lesion variations have predictable effect on the energy level at which the optimal effective dose is achieved.

23.6 Conclusions In conclusion, we examined the technique of determining the optimal dose administered to a patient using a chain of computer simulations. An x-ray simulator with computerized observers can be utilized in radiography dose optimization. In the examples

–6.5

–7.5

–6

–7

–7.5

–8 AUC - 0.8 AUC - 0.9 AUC - 0.95 AUC - 0.99

–8.5 –9

40 (a) SKE/BKE Log-eﬀective dose per unit area

543

–6.5

–7

60

80 100 Energy (keV)

120

–8 AUC - 0.8 AUC - 0.9 AUC - 0.95 AUC - 0.99

–8.5 –9

140

40

60

(b) 3AFC

–6

80 100 Energy (keV)

120

140

–6

–6.5

–6.5 –7

–7

–7.5

–7.5 –8 AUC - 0.8 AUC - 0.9 AUC - 0.95 AUC - 0.99

–8.5 –9 40

(c) 4AFC

Log-eﬀective does per unit area

–6

Log-eﬀective dose per unit area

Log-eﬀective dose per unit area

Applications to Optimization of X-Ray Radiographic Imaging

60

80 100 Energy (keV)

120

140

–8 AUC - 0.8 AUC - 0.9 AUC - 0.95 AUC - 0.99

–8.5 –9 40

60

80 100 Energy (keV)

120

140

(d) SKEV (M = 3)

FIGURE 23.9 Logarithm of effective dose (mSv) versus energy level for SKE/BKE, 2AFC, 3AFC, and SKEV (with M = 3) tasks using Hotelling observer. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

presented in this chapter, the optimal x-ray energy ranges from 45 to 65 keV. The entire optimization process is digitized, where every component was modeled in a computer simulation. This has the advantage of reducing human labor, as well as being costeffective and time-efficient. Previous studies have demonstrated that computerized observer computational phantoms such as the channelized Hotelling and the NPWE perform comparably to human observers. Modifications to the basic SKE/BKE observer task in the form of location and lesion variations also provide more clinically realistic scenarios. Future generations of the system can incorporate improvements to the current limitation and provide a more rigorous data processing and image generation algorithms: 1. Apply more advanced computational phantom and/or computational phantoms of several different of several different individuals with different body types. 2. The computerized observer package may be expanded to include observer computational phantoms with more accurate representation of the human eye, such as the channelized Hotelling computational phantom with Gabor filters29 and nonprewhitening eye filter.4

–6.5 –7 –7.5 –8 AUC - 0.6 AUC - 0.7 AUC - 0.8 AUC - 0.9

–8.5 –9

Log-effective dose per unit area

40 (a) SKE/BKE

60

80 100 Energy (keV)

120

–6.5 –7 –7.5 –8 AUC - 0.6 AUC - 0.7 AUC - 0.8 AUC - 0.9

–8.5 –9

(c) 4AFC

60

80 100 Energy (keV)

–6.5 –7 –7.5 –8 AUC - 0.6 AUC - 0.7 AUC - 0.8 AUC - 0.9

–8.5 –9 40

60

80 100 Energy (keV)

(b) 3AFC

–6

40

–6

140

120

Log-effective dose per unit area

Log-effective dose per unit area

–6

Log-effective dose per unit area

Handbook of Anatomical Models for Radiation Dosimetry

544

120

140

–6 –6.5 –7 –7.5 –8 AUC - 0.6 AUC - 0.7 AUC - 0.8 AUC - 0.9

–8.5 –9

140

40

60

80 100 Energy (keV)

(d) SKEV (M = 3)

120

140

FIGURE 23.10 Logarithm of effective dose (mSv) versus energy level for SKE/BKE, 2AFC, 3AFC, and SKEV (with M = 3) tasks using Laguerre–Gauss Hotelling observer. (From Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission.)

TABLE 23.3 Fluence and the Energy Level at the Optimal Dose for Various IQTs IQT 0.6

Observer HOT

CHOT

Task

Effective Dose (mSv)

Energy (keV)

Fluence (m−2)

SKE/BKE

1.13E−10

10

1,000

2AFC

1.17E−10

45

1,000

3AFC

2.925E−10

45

2,500

4AFC

4.27E−10

50

3,500

SKEV (M = 2)

1.755E−10

50

2,000

SKEV (M = 3)

3.51E−10

45

4,000

SKEV (M = 4)

5.85E−10

45

5,000

SKE/BKE

2.44E−8

50

200,000

2AFC

1.33E−8

50

100,000

3AFC

5.49E−8

50

450,000

4AFC

4.37E−7

60

3,000,000

SKEV (M = 2)

3.71E−8

45

350,000

SKEV (M = 3)

1.16E−7

50

950,000

Applications to Optimization of X-Ray Radiographic Imaging

545

TABLE 23.3 (continued) Fluence and the Energy Level at the Optimal Dose for Various IQTs IQT

0.7

Observer

HOT

Task

Effective Dose (mSv)

SKEV (M = 4)

4.68E−7

45

4,000,000

SKE/BKE

1.13E−10

40

1,000

2AFC

4.68E−10

45

4,000

3AFC

1.08E−9

55

8,500

4AFC

1.97E−9

60

15,000

SKEV (M = 2)

7.61E−10

45

5,500

SKEV (M = 3)

2.04E−9

50

15,000 35,000

SKEV (M = 4) CHOT

4.10E−9

SKE/BKE

8.0E−8

45

700,000

2AFC

4.49E−8

45

450,000

3AFC

3.67E−7 N/A

50

3,000,000

N/A 50

N/A 2,000,000

55

4,000,000

1.13E−10

N/A 40

N/A 1,000

2AFC

2.73E−9

45

20,000

3AFC

3.18E−9

55

25,000

4AFC

5.66E−9

45

55,000

SKEV (M = 2)

3.18E−9

55

25,000

SKEV (M = 3)

7.22E−9

45

60,000

SKEV (M = 4)

1.05E−8

45

90,000

SKE/BKE

8.0E−8

45

2,500,000

2AFC

2.97E−7

55

2,000,000

3AFC

3.67E−7 N/A

50

6,500,000

N/A 45

N/A 4,500,000 N/A N/A 25,000

SKEV (M = 3) HOT

CHOT

SKEV (M = 4) SKE/BKE

4AFC SKEV (M = 2)

0.9

HOT

CHOT

Fluence (m−2)

45

4AFC SKEV (M = 2)

0.8

Energy (keV)

SKEV (M = 3) SKEV (M = 4) SKE/BKE

2.22E−7 5.29E−7 N/A

5.27E−7 N/A N/A 3.05E−9

N/A N/A 50

2AFC

5.93E−9

55

55,000

3AFC

1.05E−8

45

90,000

4AFC

2.15E−8

45

200,000

SKEV (M = 2)

1.10E−8

50

90,000

SKEV (M = 3)

2.64E−8

50

200,000

SKEV (M = 4)

4.07E−8

50

350,000

SKE/BKE

5.27E−7

45

4,500,000

5.85E−7 N/A N/A N/A N/A N/A

45

5,500,000

N/A N/A N/A N/A N/A

N/A N/A N/A N/A N/A

2AFC 3AFC 4AFC SKEV (M = 2) SKEV (M = 3) SKEV (M = 4)

Source: Son, Y. et al., Phys. Med. Biol., 51, 4289, 2006. With permission. The value, N/A, indicates that the optimal dose was off the range of fluence that were investigated. HOT, Hotelling; CHOT, Laguerre–Gauss Hotelling.

546

Handbook of Anatomical Models for Radiation Dosimetry

3. It has been noted previously, that even under noise-free conditions, human observer response varies.30–32 This variation is attributed to the internal noise in human system. The internal noise may be modeled by further randomizing the observer’s response around the chosen threshold point.33

Acknowledgments We wish to thank Dr. Walter Huda, Dr. Kent Ogden, and Dr. Ernest Scalzetti of the Department of Radiology, State University of New York, Upstate Medical University for their guidance and input on the clinical realism of the simulator. Various portions of this research were supported by the Center for Subsurface Sensing and Imaging Systems, under the Engineering Research Center Program of the National Science Foundation Award Number EEC-9986821, and Rensselaer Polytechnic Institute. In addition, Dr. X.G. Xu and his former student, Dr. Mark Winslow, were supported during this project by grants from the National Science Foundation (BES-9875532) and the National Institutes of Health (1 R03 LM007964-01).

References 1. United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR), UNSCEAR 2000 Report: Sources and effects on ionizing radiation, United Nations Publications, New York, 2000. 2. Winslow, M., Xu, X.G., and Yazici, B. Development of a simulator for radiographic image optimization, Computer Methods and Programs in Biomedicine, 78, 179, 2005. 3. Spitzer, V.M. and Whitlock, D.G. Atlas of the Visible Human Male, Jones & Bartlett Publishers, Sudbury, MA, 1998. 4. Eckstein, M.P., Abbey, C.K., and Bochud, F.O. In Handbook of Medical Imaging Vol. I: Physics and Psychophysics, SPIE Press, Bellingham, p. 593, 2000. 5. Pineda, A.R. and Barrett, H.H. Figures of merit for detectors in digital radiography. I. Flat background and deterministic blurring, Medical Physics, 31, 348, 2004. 6. Pineda, A.R. and Barrett, H.H. Figures of merit for detectors in digital radiography. II. Finite number of secondaries and structured backgrounds, Medical Physics, 31, 359, 2004. 7. Winslow, M.P. Simulation of radiographic images for quality and dose analysis, PhD dissertation, Rensselaer Polytechnic Institute, Troy, NY, 2005. 8. Xu, X.G., Chao, T.C., and Bozkurt, A. VIP-Man: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations, Health Physics, 78, 476, 2000. 9. ICRP. Report of the Task Group on Reference Man, ICRP Publication 23, Pergamon Press, Oxford, 1975. 10. Box, G.E.P. and Muller, M.E. A note on the generation of random normal deviates, Annals of Mathematical Statistics, 29, 610, 1958. 11. ICRP. 1990 Recommendations of the International Commission on Radiological Protection, ICRP Publication 60, Pergamon Press, Oxford, 1991. 12. Winslow, M. et al. Use of the VIP-Man model to calculate energy imparted and effective dose for x-ray examinations, Health Physics, 86, 174, 2004.

Applications to Optimization of X-Ray Radiographic Imaging

547

13. Myers, K.J. and Barrett, H.H. Addition of a channel mechanism to the ideal-observer model, Journal of the Optical Society of America A: Optics Image Science and Vision, 4, 2447, 1987. 14. Yao, J. and Barrett, H.H. Predicting human performance by a channelized Hotelling observer model, Proceedings of the SPIE, Mathematical Methods in Medical Imaging, 1768, 161, 1992. 15. Barrett, H.H. et al. Stabilized estimates of Hotelling-observer detection performance in patientstructured noise, Proceedings of the SPIE, 3340, 27, 1998. 16. Abbey, C.K., Barrett, H.H., and Eckstein, M.P. Practical issues and methodology in assessment of image quality using model observers, Proceedings of SPIE, 3032, 182, 1997. 17. Burgess, A.E. and Ghandeharian, H. Visual signal-detection. 2. Signal-location identification, Journal of the Optical Society of America A: Optics Image Science and Vision, 1, 906, 1984. 18. Eckstein, M.P. and Abbey, C.K. Model observers for signal known statistically task, Proceedings of SPIE, 4324, 91, 2001. 19. Eckstein, M.P. et al. Automated computer evaluation and optimization of image compression of x-ray coronary angiograms for signal known exactly detection tasks, Optics Express, 11, 460, 2003. 20. Agarwal, S. et al. Generalization bounds for the area under the ROC curve, Journal of Machine Learning Research, 6, 393, 2005. 21. Chan, H.P. and Doi, K. Studies of x-ray-energy absorption and quantum noise properties of x-ray screens by use of Monte-Carlo simulation, Medical Physics, 11, 37, 1984. 22. Erbel, R. et al. Understanding the pathophysiology of the arterial wall: Which method should we choose? Electron beam computed tomography, European Heart Journal Supplements, 4, F47, 2002. 23. Fayad, Z.A. and Fuster, V. Clinical imaging of the high-risk or vulnerable atherosclerotic plaque, Circulation Research, 89, 305, 2001. 24. Bochud, F.O., Abbey, C.K., and Eckstein, M.P. Search for lesions in mammograms: Statistical characterization of observer responses, Medical Physics, 31, 24, 2004. 25. Burgess, A.E. Statistically defined backgrounds: Performance of a modified nonprewhitening observer model, Journal of the Optical Society of America A: Optics Image Science and Vision, 11, 1237, 1994. 26. Eckstein, M.P. and Whiting, J.S. Visual signal detection in structured backgrounds. 1. Effect of number of possible spatial locations and signal contrast, Journal of the Optical Society of America A: Optics Image Science and Vision, 13, 1777, 1996. 27. Kijewski, M.F., Mueller, S.P., and Moore, S.C. The barankin bound: A model of detection with location uncertainty, Proceedings of the SPIE, Mathematical Methods in Medical Imaging, 1768, 153, 1992. 28. Eckstein, M.P. et al. Optimization of model observer performance for signal known exactly but variable tasks leads to optimized performance in signal known statistically tasks, Proceedings of the SPIE, Medical Imaging 2003: Image Perception, Observer Performance, and Technology Assessment, 5034, 123, 2003. 29. Zhang, Y.I., Pham, B.T., and Eckstein, M.P. Automated optimization of JPEG 2000 encoder options based on model observer performance for detecting variable signals in x-ray coronary angiograms, IEEE Transactions on Medical Imaging, 23, 459, 2004. 30. Burgess, A.E. and Colborne, B. Visual signal-detection. 4. Observer inconsistency, Journal of the Optical Society of America A: Optics Image Science and Vision, 5, 617, 1988. 31. Burgess, A.E. and Humphrey, K. Density dependence of signal-detection in radiographs, Medical Physics, 8, 646, 1981. 32. Burgess, A.E., Humphrey, K., and Wagner, R.F. Detection of bars and discs in quantum noise, Proceedings of the Society Photo Optical Instrumentation Engineers: Application of Optical Instrumentation in Medicine VII, 173, 34, 1979. 33. Ahumada, A.J., Jr. Classification image weights and internal noise level estimation, Journal of Vision, 2, 121, 2002. 34. Son, Y. et al. X-ray imaging optimization using virtual phantoms and computerized observer modeling, Physics in Medicine and Biology, 51, 4289, 2006.

24 Applications to Nuclear Medicine Imaging and Dosimetry Involving MCAT, NCAT, and MOBY Phantoms Benjamin M.W. Tsui and W. Paul Segars

CONTENTS 24.1 Introduction ............................................................................................................... 549 24.2 Simulation Tools ....................................................................................................... 550 24.3 Applications of Computer-Generated Phantoms................................................. 554 24.3.1 Effects of Image Degrading Factors ......................................................... 554 24.3.2 Evaluation of New Image Reconstruction and Processing Methods ............................................................................ 559 24.3.3 Dosimetry Calculations ............................................................................. 561 24.4 Summary ...................................................................................................................564 References .............................................................................................................................564

24.1 Introduction Recent advances in computer-generated phantoms, especially those described in Chapter 5, have found many important applications in medical imaging and dosimetry calculations. In medical imaging, a computer-generated phantom that realistically models the anatomical structures and physiological functions of human or a small animal, when combined with methods that computational phantom the physics involved in the imaging process and characteristics of the imaging system, form a set of unique simulation tools that provides imaging data that closely mimic clinical and experimental image data. Most importantly, the ability to insert known features and abnormalities in the computer-generated phantom and simulated data offers “true” status of subject that is difficult to obtain from clinical and experimental animal image data. The knowledge of the “truth” allows rigorous evaluation of the effects of image degradation factors in the imaging process, the imaging system, and image reconstruction and processing methods on medical images. Computer-generated phantoms with their realistic anatomy allow accurate calculations of dose to various tissue organs from external beams of radiation in radiation oncology and diagnostic radiology and internally distributed radioactivity found in nuclear medicine, brachytherapy, and radioimmunotherapy. In this chapter, we present examples of the use of computer-generated phantoms, specifically the MCAT, NCAT, and MOBY phantoms, in the investigation of the effects of image 549

Handbook of Anatomical Models for Radiation Dosimetry

550

degrading factors and image evaluation studies using realistic simulation data. The purposes of the image evaluation studies include optimization of design parameters of instrumentation, and image acquisition, reconstruction, and processing methods. Also, examples of the use of computer-generated phantoms to dosimetry calculations are presented to demonstrate their important role in diagnostic radiology, radiation oncology, and health physics.

24.2 Simulation Tools A computer-generated phantom is one of the major tools necessary in a realistic simulation study. It provides a realistic computational phantom of the anatomical structure and physiological functions of an imaging subject, be it a human or small animal. The other important simulation tool is a method that generates image data that mimics that acquired from the imaging process and instrumentation in a clinical or experimental setting. The most accurate methods for generating x-ray and nuclear medicine image data are Monte Carlo simulation methods that require intensive computations. In order to conduct Monte Carlo simulation, a fast computing facility is often required as the third major tool required in a realistic simulation study. Figure 24.1 shows the use of the three-dimensional (3D) NCAT phantom1 to simulate twodimensional (2D) radiography and x-ray CT image data. The 3D NCAT phantom provides x-ray source

3D NCAT phantom

(a)

(d)

(b)

Array of detectors

Relative counts, η(Ei)

6.00E+05 5.00E+05 4.00E+05 3.00E+05 2.00E+05 1.00E+05 0.00E+00

(c)

0

20

40

60

80

100

120

140

Energy, Ei in (keV)

(e)

FIGURE 24.1 (See color insert following page 524.) High-resolution 2D x-ray planar and 3D CT image simulation using the 3D NCAT phantom. (a) The head and torso portion of the 3D NCAT phantom. (b) A typical x-ray imaging configuration with the 3D NCAT phantom replacing a patient. (c) A typical 120 kVp x-ray spectrum. (d) Sample 2D frontal planar images with, from left to right, 30, 60, and 90 kVp monoenergetic x-ray energy beam. (e) Sample transaxial slices from the 3D CT image reconstructed from multiple 2D planar projections of different views around the phantom.

Applications to Imaging and Dosimetry

551

3D surfaces that define the size and shape of the organs and anatomical structures realistically. Each specific structure is assigned its own characteristic attenuation properties of the x-ray photons. For high-resolution x-ray imaging, the geometry of the x-ray beam, e.g., fan beam or cone beam is modeled by an array of narrow x-ray beams. The transmission of each of the narrow beams is calculated from its direction, its intersection of the surface of each organ, and the interaction of the x-ray photons with the tissue materials. Also, the polychromatic x-ray spectrum is modeled by a weighted sum of discrete bins of monoenergetic x-ray photons. The simulation can be calculated using analytical methods for computational efficiency. To accurately model the interactions of the x-ray photons with the tissue materials, Monte Carlo simulation methods can be used at the expense of higher computational time. The 2D x-ray projections from multiple views are then used in image reconstruction to generate the CT images. As shown in Figure 24.2, in simulating nuclear medicine image data, each structure in the 3D NCAT phantom is assigned a given amount of radioactivity or radioactivity concentration in addition to the attenuation and scattering properties of the tissue for the gamma-ray photons emitted from the radioactivity. The distribution of radioactivity among all organs represents the biodistribution of the administered radiopharmaceutical labeled with a given isotope and gamma-ray photon emissions. Since the nuclear medicine images are of relatively poor resolution, the 3D NCAT phantom is usually voxelized for use in the simulation. The attenuation and scattering properties of the organs are used in simulating the photon transport through the body. 3D radioactivity distribution

3D attenuation coeffient distribution

(a)

(b)

(c)

(d)

FIGURE 24.2 SPECT/CT image simulation using the 3D NCAT phantom. (a) The 3D NCAT phantom. (b) A representative 3D radioactivity distribution of a Tc-99m Sestamibi myocardial perfusion study and 3D attenuation coefficient distribution of the 3D NCAT phantom. (c) Sample simulated noise-free 2D projection data from different views around the phantom that include the effect of photon attenuation and scatter, and CDR. (d) Sample transaxial slices of the 3D SPECT image reconstructed from the multiple 2D planar projections of different views around the phantom.

Handbook of Anatomical Models for Radiation Dosimetry

552

The 2D planar nuclear medicine image data, or the 2D projection data in SPECT and PET imaging, are simulated by modeling the physics of photon transport of the gamma-ray emissions from the internally distributed radioactivity distribution and the characteristics of the imaging system. For example, for a scintillation camera fitted with a collimator, the point response function of the collimator–detector defines the region of the radioactivity distribution that will be integrated to form a data point on the simulated image. Monte Carlo simulation methods provide accurate means to model the absorption and scatter of photons while being transported through the organs. With a realistic computer-generated phantom and accurate modeling of the imaging system and imaging process, the simulated image data will closely mimic clinically or experimentally acquired data. In SPECT and PET, the simulated 2D projection data will then be used in image reconstruction to obtain the 3D reconstructed image dataset. In Figure 24.3, we show examples of comparison between simulated images obtained using the 3D NCAT phantom and targeted clinical CT, SPECT, and PET images for the simulations. The comparisons show the similarities between the simulated and clinical images indicating the effectiveness of the simulation methods. Similar simulation techniques can be applied to small animal imaging using the computer-generated MOBY phantom2 described in Chapter 5. Figure 24.4 shows simulated SPECT/CT modeling images acquired from a small animal SPECT/CT system where the SPECT system is fitted with pinhole collimators with 1.0 mm diameter pinhole apertures. Known abnormalities can be modeled and inserted in the computer-generated phantoms. Figure 24.5 shows examples of a nodule in the lung, a perfusion defect along the

22

39

3D NCAT phantom

66

3D NCAT phantom 61 83

Patient

(a)

3D NCAT phantom (b)

44

(c)

Patient

3D NCAT phantom (d)

Patient

Patient

FIGURE 24.3 Comparison of sample simulated images using the 3D NCAT phantom and clinical images. (a) CT images of the chest. (b) Tc-99m Sestamibi myocardial perfusion SPECT images. (c) In-111 ProstaScient ® SPECT images. (d) PET brain receptor images.

Applications to Imaging and Dosimetry

553

Tc-99m Sestamibi activity distribution

Corresponding attenuation distribution

(c)

4 (a)

(b)

8

(d)

16 32 Number of iteration

64

FIGURE 24.4 SPECT/CT image simulation using the 3D MOBY phantom. (a) The 3D MOBY phantom. (b) A representative 3D radioactivity distribution of a Tc-99m Sestamibi myocardial perfusion study and 3D attenuation coefficient distribution of the 3D MOBY phantom. (c) Sample simulated noisy 2D projection data from different views around the phantom obtained using a 1 mm pinhole collimator and include the effect of photon attenuation and scatter, and CDR. (d) Sample transaxial slices of the 3D SPECT image reconstructed from the multiple 2D planar pinhole projections of different views around the phantom.

Iliac vessels

Hypogastric vessels

Lymph nodes Vas deferens

Anterior (a)

Bladder

Seminal vesicles

Prostate

Posterior

(c)

Plaque

Defec

(b)

(d)

FIGURE 24.5 (See color insert following page 524.) Example of simulated abnormalities. (a) Lung with trachea tree and a lung nodule, (b) an atherosclerotic plaque in the aorta region of the 3D MOBY phantom, (c) a perfusion defect modeled as a wedge in the left ventricular myocardial wall, and (d) detailed anatomy of the pelvic area with lymph nodes where metastases of prostate cancer first appear in the 3D NCAT phantom.

554

Handbook of Anatomical Models for Radiation Dosimetry

myocardium and lymph nodes in the pelvic area of the 3D NCAT phantom. Also shown is an atherosclerotic plaque in the aorta of the 3D MOBY phantom. These abnormalities can be simulated with specific dimensions, shapes, and characteristics. For example, a lung nodule can be assigned with high radioactivity concentration to simulate lung cancer in PET imaging. The atherosclerotic plaque can be assigned with high calcium content in CT imaging or high radioactivity concentration in SPECT or PET imaging to simulate different stages of atherosclerotic plaque development. The known abnormalities allow for evaluation of the quantitative accuracy of image reconstruction and processing methods.

24.3 Applications of Computer-Generated Phantoms Computer-generated phantoms and simulation studies have found many applications in diagnostic imaging, radiation oncology, and dosimetry calculations. New applications are continuously being explored and investigated. The main advantages of the new generation of computer-generated phantoms include realistic modeling of the anatomical structures and physiological functions of humans and small animals, ability to model normal and abnormal cases, and the known information and “true” status of the phantom. When combined with Monte Carlo simulation techniques, the simulation tools allow the study of image degradation factors individually and in different combinations, the evaluation of new image reconstruction and processing methods, and the evaluation of image quality in many medical imaging applications. In the following, we present examples of these applications in CT, SPECT, and PET with the 3D and four-dimensional (4D) MCAT, NCAT, and MOBY computer-generated phantoms. 24.3.1 Effects of Image Degrading Factors X-ray photons from external x-ray sources in radiography and CT imaging and gammaray photons from internally distributed radioactivity in nuclear medicine imaging travel through layers of tissues in the body before reaching the detector. The amount and properties of the exiting photons depend on the photon energy, and composition and density of the different tissues. The major photon interactions involved in medical imaging are photoelectric absorption and Compton scattering. They result in photon attenuation that reduces the fractional amount of photons exiting the body especially from thicker and denser layers of traversed tissues and scattered photons that contribute to lower image contrast. The radiation detector used in image formation contributes to degradation in spatial resolution. Other image degrading effects include image noise from statistical noise in radiation detection within a fi nite acquisition time, image artifacts such as breast and diaphragm shadows due to photon attenuation anatomical structures, and patient voluntary and involuntary motions such as cardiac and respiratory motions.3,4 In Figure 24.6, we show the effects of photon attenuation and additional photon scatter, collimator–detector response (CDR), and Poisson noise in myocardial perfusion SPECT images obtained using the conventional filtered backprojection (FBP) image reconstruction algorithm without any compensation. The image degradation and artifacts caused by

Applications to Imaging and Dosimetry

555

Activity distribution

Attenuation distribution

(a)

Ideal data (b)

Attenuation

Attenuation and CDR

Attenuation, CDR and scatter

Attenuation, CDR, scatter and Poisson noise

With post smoothing filter

FIGURE 24.6 Effects of image degrading factors in myocardial perfusion SPECT. (a) A computer-generated phantom based on: (left) a segmented activity distribution in a clinical myocardial perfusion SPECT study and (right) corresponding attenuation distribution. (b) FBP reconstructed images obtained from simulated projection data that include the effects of (from left to right) no image degrading effect, with attenuation effect only, with additional CDR and Poission noise fluctuations, and with postprocessing using a smoothing filter.

the individual and combinations of image degrading factors are demonstrated by comparing the FBP reconstructed and phantom images. Postprocessing using a smoothing filter reduces the image noise and smoothes out the image artifacts but the images remain substantially degraded. To demonstrate the effects of the image degrading factors in clinical myocardial SPECT, the flexibility of the 3D MCAT phantom5 was used in a simulation study to create anatomical variations including the shape of the diaphragm and larger breast size in female patients. Figure 24.7 shows reconstructed images using the conventional FBP image reconstruction method without any compensation. The figure illustrates how the detection of myocardial defects at different locations may be affected resulting in a possible effect on clinical diagnosis. For example, a raised diaphragm in male patients and larger breasts in female patients may generate attenuation artifacts in the inferior and superior wall, respectively, that may be mistakenly interpreted as myocardial defects in normal patients. On the other hand, actual myocardial defects in the superior and inferior wall may be interpreted as normal in male patient with a raised diaphragm and female with larger breasts, respectively. The 4D NCAT phantom with realistic models of cardiac and respiratory motions6,7 allows demonstration of the effects of cardiac and respiratory motion in medical images.8,9 In Figure 24.8, we show the effect of the number of gated frames used in gated myocardial SPECT studies. For the same imaging time, the higher number of gated frames per cardiac cycle provides higher temporal resolution but higher image noise per gated frame. Also, we demonstrate the performance of the newer statistical (orderedsubsets expectation-maximization) OS-EM and MAP-RBI image reconstruction methods in improving the gated myocardial SPECT images as compared to the conventional FBP algorithm.10

Handbook of Anatomical Models for Radiation Dosimetry

556

Phantom

Male flat diaphragm

Male raised Female large diaphragm breasts

No defect

Anterior defect

Inferior defect

Septal defect

FIGURE 24.7 Demonstration of the effects of photon attenuation in myocardial SPECT imaging using the 3D MCAT phantom in a simulation study. All images are from a sample short-axis slice from the center of the heart phantom with and without myocardial defect. (Leftmost column) A sample short-axis slice through the center of the heart of the 3D MCAT phantom. The FBP reconstructed images without any compensation of (2nd column from left) the male MCAT phantom with a flat diaphragm, (2nd column from right) the male MCAT phantom with a raised diaphragm, and (rightmost column) the female MCAT phantom with large breasts. The corresponding phantom slice has (top row) no myocardial defect, (2nd row from top) a anterior defect, (2nd row from bottom) an inferior defect, (bottom row) a septal defect.

2D FBP ramp filter

3DOS-EM 4D MAP-RBI w/2 iteration and w/2 iterations 3D OS-EM compensation and compensation w/o 2 iterations

8 gates

16 gates

24 gates

36 gates

FIGURE 24.8 Sample simulated gated myocardial short-axis SPECT images from the 4D NCAT phantom with a realistic beating heart model with a septal-interior wall motion abnormality. From the top to the bottom row are sample single frame images from dividing the beating heart cycle with the same total imaging time into 8, 16, 24, and 36 gates, respectively. They show the increased image noise with higher number of gated frames although the temporal resolution is higher. From the left to right column, the single frame images are obtained from using the 2D FBP and 2 iterations and 8 subsets of OS-EM algorithms without any compensation, and 2 iterations and 8 subsets 3D OS-EM and the 4D MAP-RBI algorithms with compensation for CDR.

Applications to Imaging and Dosimetry

(a)

w/o respiratory motion

w/ respiratory motion

557

(b)

(c) FIGURE 24.9 (a) Sample SPECT images at the same transaxial slice through the center of the lung nodule in the 3D NCAT phantom shown in Figure 24.5 (a) with and without respiratory motion. They demonstrate the blurring effect of the respiratory motion. (b) Two coronary slice of a liver lesion in the NCAT phantom (left column) without and (right column) with a realistic model of respiratory motion. (c) Coronal images from simulated respiratorygated PET images at nine different coronary slices from simulated gated projection data using the combined SimSET and GATE Monte Carlo simulation codes.

Respiratory motion affects medical images in several ways.8,9,11 Figure 24.9 demonstrates the blurring effect of respiration motion in the detection of lung nodules and liver cancers in simulated SPECT and PET studies. In Figure 24.9a, sample SPECT images at the same transaxial slice through the center of the lung nodule modeled in the 3D NCAT phantom shown in Figure 24.5a with and without respiratory motion. The images demonstrate the blurring effect of the respiratory motion. Figure 24.9b shows a liver lesion in the 3D NCAT phantom without and with a realistic computational phantom of respiratory motion. The 4D NCAT phantoms with the liver lesion were used to generate respiratory-gated PET data using a combined SimSET and GATE Monte Carlo simulation codes as shown in Figure 24.9c. The noisy PET images were due to the extensive computational time required in the Monte Carlo simulations despite the use of an efficient combined SimSET12 and GATE13 Monte Carlo codes to speed up the computational time.14 Respiratory motion-generated unique artifacts in myocardial SPECT and PET images.15–17 In Figure 24.10, we show results from a simulation study that used the 4D NCAT phantom with a computational phantom of the respiratory motion to demonstrate its effect on myocardial perfusion SPECT images. The degree of respiratory motion was adjusted to give maximum diaphragm amplitudes of 0, 2, and 4 cm modeling no respiration, slightly enhanced and deep breathing, respectively. Projection data were simulated using three

Handbook of Anatomical Models for Radiation Dosimetry

558

(a)

(b)

(c)

FIGURE 24.10 Sample (top row) short-axis myocardial SPECT images and (bottom row) the corresponding bull’s-eye plots showing the myocardial perfusion from the apex (center) to the base (outer ring) of the heart. The projection data were simulated using three 4D NCAT phantoms with Tc-99m Sestamibi uptake distribution modeling that of a typical normal subject with (a) 0 cm, (b) 2 cm, and (c) 4 cm maximum diaphragm amplitudes. The projection data were reconstructed using the iterative image reconstruction method with attenuation correction. The concurrent decrease in image intensity in the superior and inferior regions of the myocardium is indicative of the effect of respiratory motion in myocardial SPECT images.

4D NCAT phantoms with the three models of respiratory motions and Tc-99m Sestamibi uptake distribution. The projection data were reconstructed using the iterative OS-EM image reconstruction method with attenuation correction. The sample short-axis images and bull’s-eye plots demonstrate the concurrent decrease in image intensity in the superior and inferior regions of the myocardium indicative of the effect of respiratory motion in myocardial perfusion SPECT images. In simulation using PET image data, similar respiratory motion-induced image artifacts are found but at less maximum diaphragm amplitude due to higher system resolution. The image artifacts are confirmed in clinical SPECT and PET studies. Simulation techniques has been found useful in identifying image artifacts due to respiratory motion in the low dose and slow CT subunit used in first-generation SPECT/CT systems. In Figure 24.11, we show simulated chest and abdominal CT images obtained with the 4D NCAT phantom without and with respiratory motion. The projection data model those acquired using a slow CT subunit with 14 s per 360° rotation time. The typical 5 s respiratory breathing cycle together with the 14 s/rotation acquisition time resulted in distinct image artifacts in the simulated images that have great similarity with those found in clinical images. The simulation has been found useful in identifying the cause of the image artifacts. In Figure 24.12, we show an example of the use of the 4D MOBY phantom in a molecular imaging application. The uptake of Tc-99m labeled Annexin V in an atherosclerotic plaque in the aorta region of the ApoE −/− mouse computational phantom was simulated using the 4D MOBY phantom without and with respiratory motion.18 Also shown are the signalto-noise ratio (SNR) of the plaques determined from the SPECT images as a function of the amplitude of the respiration motion and the Tc-99m Annexin V uptake ratio of the plaque. The simulation study provides a guide for the instrumentation and experimental design in the detection of atherosclerotic plaques in molecular SPECT/CT imaging of mice.

Applications to Imaging and Dosimetry

559

Sample simulated CT images without respiration motion

Sample simulated CT images with respiration motion eﬀect

Sample simulated CT images with respiration motion eﬀect (a)

(b)

FIGURE 24.11 Simulation study to identify image artifacts due to respiratory motion in low dose and slow CT imaging found in first-generation SPECT/CT systems. Sample CT images from (a) the chest and (b) abdominal regions. (Top row) Sample simulated transaxial CT images from the 4D NCAT phantom without respiratory motion. (Middle row) Simulated CT images from the 4D NCAT phantom with respiratory motion. The simulated projection data model the slow 14 s/rotation used in the clinical CT data acquisition. (Bottom row) Sample clinical CT images with matching transaxial slices.

8 7 6 SNR

5 4 3 2 1 0 (a)

(b)

9-1

18-1 36-1 Plaque uptake ratio

No motion 0.5 mm

1.0 mm 2.0 mm

FIGURE 24.12 (a) Sample simulated fused (from left to right) transaxial, coronal and MIP fused SPECT/CT images from the 4D MOBY phantom shown in Figure 24.4 and with an atherosclerotic plaque in the aorta region as shown in Figure 24.5. The images in the top and bottom rows are from the 4D MOBY phantom without and with respiratory motion, respectively. A ~40% drop in lesion contrast is found in the plaque images. (b) SNR of plaque detection in the mouse model as a function of respiration motion amplitude and the radioactivity uptake ratio of plaque.

24.3.2 Evaluation of New Image Reconstruction and Processing Methods Computer-generated phantoms have found useful application in simulation studies to evaluate new image reconstruction and processing methods. For example, an important goal of quantitative image reconstruction and processing methods in medical imaging is

Handbook of Anatomical Models for Radiation Dosimetry

560

to provide images that accurately and precisely represent the “true” status of the subject being imaged. Without the known “truth,” it is difficult to evaluate the performance of these methods and subjective judgment is often used instead. In simulation studies using computer-generated phantom, the “truth” is known. With knowledge of the known truth, quantitative evaluation can be achieved by comparing the reconstructed and processed image with the phantom image. In Figures 24.6 and 24.7, we demonstrate the effects of various image degrading factors on myocardial perfusion SPECT images. An important goal of any new image reconstruction and processing method is to compensate for these effects and to generate reconstructed and processed images that closely resemble the true status. Since attenuation has the greatest effect on image quality and quantitative accuracy, image reconstruction methods with accurate attenuation compensation (AC) will be important to improve the quality and quantitative accuracy of the reconstructed images that will potentially lead to more accurate clinical diagnosis.19,20 Figure 24.13 shows the effectiveness of an iterative maximum-likelihood expectation-maximization (ML-EM) image reconstruction method with AC in a simulation study using the 3D NCAT phantom. The same corresponding sample images as those shown in Figure 24.7 are presented except that the images are obtained with the ML-EM image reconstruction methods with AC. When compared to the FBP images obtained without AC in Figure 24.7, the effectiveness of the image artifacts caused by photon attenuation and different anatomical structures are substantially reduced, and the reconstructed images approach that of the corresponding phantom images on the left column of Figure 24.13. The simulated images clearly demonstrate the effectiveness of the

Phantom

Male flat diaphragm

Male raised diaphragm

Female large breasts

No defect

Anterior defect

Inferior defect

Septal defect

FIGURE 24.13 The sample simulated short-axis myocardial perfusion SPECT images are similar to those in Figure 24.6 except that the ML-EM image reconstruction method with AC was used. Comparing the corresponding images in Figure 24.6 demonstrate the reduction in image artifacts due to photon attenuation caused by variations in patient anatomy and defect locations and the attenuation compensated images are much closer to the corresponding phantom images.

Applications to Imaging and Dosimetry

561

iterative image reconstruction methods with AC in generating quantitatively accurate reconstructed images that approach the true perfusion distribution in the myocardium. Image reconstruction and compensation methods that provide compensation for other image degrading factors have been developed for SPECT for further improvement in image quality and quantitative accuracy.3,4,21,22 In Figure 24.14a, we show the 3D NCAT phantom that models the distribution of Tc-99m Sestamibi in different organs in a typical clinical myocardial perfusion SPECT study and a myocardial defect. Together with a corresponding 3D NCAT phantom with known attenuation coefficient distribution and the imaging characteristics of a typical camera fitted with a high-resolution collimator, SPECT projection data that include the effects of the CDR and photon attenuation and scatter were simulated. In Figure 24.14b, all the SPECT images correspond to the same phantom image slice marked in Figure 24.14a. From the top to the bottom row, they are obtained from the iterative 3D OS-EM image reconstruction methods without any compensation, with compensation of the collimator–detector blur only, and with additional attenuation and scatter compensation after 1, 2, 3, 5, 10, and 20 iterations. When compared with the marked phantom image in Figure 24.14a, the reconstructed images demonstrate the ability of the image reconstruction and compensation methods to reduce the effects of the image degrading factors for improved image quality and quantitative accuracy. In particular, the images demonstrate the effect of the individual compensation method and the best images quality and highest quantitative accuracy that can be achieved when a combination of all the compensation methods are used. In Figures 24.9 and 24.12, the 4D NCAT and MOBY computer-generated phantoms were used in simulation studies to demonstrate potential image degradation and artifacts caused by respiratory motion in various biomedical imaging applications. The same computer-generated phantoms, acting as the known “truth,” can also be used to study the effectiveness of the different respiratory motion compensation methods. For example, respiratory gating is one of the effective means to compensate the blurring effect in lung nodule and atherosclerotic plaques detection as shown in Figures 24.9 and 24.12, respectively.23–25 It can also reduce the image artifact found in myocardial perfusion SPECT and PET images as shown in Figure 24.10. In Figure 24.15, we demonstrate the application of the 4D NCAT phantom with its computational phantom of the respiratory motion to evaluate the effectiveness of respiratory gating to improve the quality and quantitative accuracy of myocardial perfusion SPECT. Projection data were simulated from the 4D NCAT phantom that models a typical normal myocardial perfusion SPECT study with and without respiratory gating. Figure 24.15a shows the image artifacts caused by respiratory motion with and without AC as compared to the known uniform myocardial perfusion distribution. In Figure 24.15b the SPECT images from all respiratory gates are transformed to the gated frame at expiration. When attenuation correction is applied, the summed image shows uniform myocardial perfusion distribution, especially with respiratory-gated attenuation map, indicating the effectiveness of respiratory motion compensation using the respiratory gating method. 24.3.3 Dosimetry Calculations As discussed in previous chapters, much work is being done to apply more advanced computational phantoms to the field of radiation dosimetry. Based on patient imaging data, these phantoms are far more realistic than the generic stylized phantoms used in the past. As a result, dose calculations performed using them are closer to those from actual patients. With its high level of realism and flexibility, the 3D NCAT phantom has found application to the field of radiation dosimetry as discussed in Chapter 12. The flexibility of

Handbook of Anatomical Models for Radiation Dosimetry

562

(a) w/o compensation w/ CDR (geo.) compensation w/ CDR and attenuation compensation w/ CDR, attenuation and scatter compensation 1 (b)

2

3

5

10

20

Number of iterations

FIGURE 24.14 (a) Transaxial images of the 3D NCAT phantom that model the distribution of Tc-99m Sestamibi in various organs of a typical clinical myocardial perfusion SPECT study. A myocardial perfusion defect is located at the lateral wall of the myocardium. The transaxial image slice that is used in the evaluation study shown in (a) is marked with a thick border. (b) Sample SPECT images obtained from the simulated projection data of the 3D NCAT phantom shown in (a) including the effects of collimator–detector blur and photon attenuation and scatter. The reconstructed images correspond to the same phantom image marked in (a) and are obtained from the iterative 3D OS-EM image reconstruction methods (top row) without any compensation, (second row), with compensation of the collimator–detector blur, (third row) with additional compensation of photon attenuation, and (bottom row) photon scatter, after 1, 2, 3, 5, 10 and 20 iterations.

Applications to Imaging and Dosimetry

563

No respiratory gating

Respiratory gating (6 respiratory gates, gated frame at expiration)

w/o AC

w/o AC

w/AC (a)

w/AC using averaged attenuation map (b)

w/AC using gated attenuation map

FIGURE 24.15 (a) Bull’s-eye plots from simulated myocardial perfusion short-axis SPECT images of the entire heart from the apex (center) to the base (outer ring). The SPECT images were obtained from simulated projections of the 4D NCAT phantom with normal and uniform myocardial perfusion and respiratory motion. However, no respiratory gating was used and the SPECT images were obtained (top row) without and (bottom row) with AC. (b) Bull’s-eye plots similar to those in (a) except the respiratory gating with 6 gates was used. The SPECT images from all respiratory gates are transformed to the gated frame at expiration. The SPECT images are reconstructed (top row) without and (bottom row) with AC using (left image) the averaged and (right image) respiratory-gated attenuation maps.

the phantom allows it to model any number of anatomical variations. Anatomical variations are simply created by applying transforms to the surface primitives that define the organs and structures. With this ability, the 3D NCAT phantom is being used as a template from which to create a series of computational phantoms realistically representing adults and children of different ages, Figure 24.16. Current work is underway investigating

FIGURE 24.16 Anatomical variations at different ages (adult to 1 year old) produced by morphing the 3D NCAT phantom to patient CT data.

564

Handbook of Anatomical Models for Radiation Dosimetry

techniques to rapidly transform the 3D NCAT phantom to match patient imaging data so as to create patient-specific phantoms. These computational phantoms can provide a vital tool for patient-specific dose calculation, enabling more accurate radiation risk estimation and more informed protocol design for patients undergoing medical examination or treatment. The 3D MOBY phantom is similarly being investigated for use in dosimetry for animal studies. Like the 3D NCAT phantom, it can be easily transformed to model many different anatomies.

24.4 Summary In this chapter, we have presented examples of the application of computer-generated phantoms in simulation studies to investigate the effects of image degrading factors and to evaluate image reconstruction and compensation methods to improve the quality and quantitative accuracy of medical images. Also, examples of the application of computergenerated phantoms in dosimetry calculations are also presented. The success of the recent generation of computer-generated phantoms in medical imaging and dosimetry calculations can be attributed to several factors. They include realistic modeling of anatomical structures, and cardiac and respiratory motions of both human and small animals. When combined with advanced simulation techniques, such as Monte Carlo simulation methods that accurately model the physics of the imaging process and imaging characteristics of the imaging system, simulated image data that closely mimic those from clinical and experimental animal studies can be obtained. Furthermore, knowledge of the true normal or abnormal anatomy and physiological motions allow evaluation of imaging systems and image reconstruction and processing methods that is not possible using clinical data where the true status of the patient is often unknown. We anticipate continuous and growing use of computer-generated phantoms and simulation studies in an increasing number of applications in medical imaging and dosimetry calculations. In particular, populations of the computer-generated phantoms with variations in anatomy, parameters of physiological functions, and abnormalities provide realistic computational phantoms of patient and animal populations have been generated and used in image evaluation studies. New generation of Monte Carlo simulation methods allows accurate modeling of complex imaging instrumentation and imaging process. The use of populations of realistic phantoms and accurate simulated image data with known truth will have great potential to complement or even substitute expensive clinical trials in the evaluation of medical imaging systems, data acquisition and image processing methods in quantitative terms with improved precision and accuracy.

References 1. Segars, W.P. Development and application of the new dynamic NURBS-based Cardiac-Torso (NCAT) phantom, PhD thesis, University of North Carolina at Chapel Hill, 2001. 2. Segars, W.P. et al. Development of a 4-D digital mouse phantom for molecular imaging research, Molecular Imaging and Biology, 6, 149, 2004.

Applications to Imaging and Dosimetry

565

3. Jaszczak, R.J. and Tsui, B.M.W. In Principles of Nuclear Medicine, H.N. Wagner, Z. Szabo, J.W. Buchanan, eds., Philadephia, PA: W.B. Saunders Company, p. 317, 1995. 4. Tsui, B.M.W. In Nuclear Medicine, R.E. Henkin, M.A. Boles, G.L. Dillehay, J.R. Halama, R.H. Wagner, A.M. Zimmer, S.M. Karesh, eds., Philadephia, PA: Mosby Elsevier, p. 223, 2006. 5. Terry, J.A. et al. In Biomedical Engineering: Opening New Doors, D.C. Mikulecky, and A.M. Clarke, eds., New York: New York University Press, p. 185, 1990. 6. Segars, W.P., Lalush, D.S., and Tsui, B.M.W. A realistic spline-based dynamic heart phantom, Nuclear Science Symposium, 1998. Conference Record. 1998 IEEE, 1998. 7. Segars, W.P., Lalush, D.S., and Tsui, B.M.W. Modeling respiratory mechanics in the MCAT and spline-based MCAT phantoms, IEEE Transactions on Nuclear Science, 48, 89, 2001. 8. Smyczynski, M. et al. Impact of respiratory motion on the detection of solitary pulmonary nodules with SPECT imaging of NeoTect, IEEE Medical Imaging Conference and Nuclear Science Symposium, Norfolk, VA, November 10–16, 2002. 9. Tsui, B.M.W., Segars, W.P., and Lalush, D.S. Effects of upward creep and respiratory motion in myocardial SPECT, IEEE Transactions on Nuclear Science, 47, 1192, 2000. 10. Lee, T.S., Segars, W.P., and Tsui, B.M.W. Development and optimization of a 4D MAP-RBI-EM reconstruction algorithm with space-time gibbs priors for application to gated myocardial perfusion SPECT, Proceedings of the Fully 3D Image Reconstruction Meeting in Radiology and Nuclear Medicine, Salt Lake City, Utah, July 6–9, 2005. 11. Wang, J. et al. Evaluation of amplitude-based sorting algorithm to reduce lung tumor blurring in PET images using 4D NCAT phantom, Computer Methods and Programs in Biomedicine, 87, 112, 2007. 12. Harrison, R.L. et al. A public-domain simulation system for emission tomography—photon tracking through heterogeneous attenuation using importance sampling, Journal of Nuclear Medicine, 34, P60, 1993. 13. Jan, S. et al. GATE: A simulation toolkit for PET and SPECT, Physics in Medicine and Biology, 49, 4543, 2004. 14. Shilov, M.A. et al. Improved Monte-Carlo simulations for dynamic PET, Journal of Nuclear Medicine Meeting Abstracts, 47, 197, 2006. 15. Kovalski, G. et al. Correction of heart motion due to respiration in clinical myocardial perfusion SPECT scans using respiratory gating, Journal of Nuclear Medicine, 48, 630, 2007. 16. Le Meunier, L. et al. PET/CT imaging: Effect of respiratory motion on apparent myocardial uptake, Journal of Nuclear Cardiology, 13, 821, 2006. 17. Segars, W.P. and Tsui, B.M.W. Study of the efficacy of respiratory gating in myocardial SPECT using the new 4-D NCAT phantom, IEEE Transactions on Nuclear Science, 49, 675, 2002. 18. Tsui, B.M.W. and Wang, Y.C. High-resolution molecular imaging techniques for cardiovascular research, Journal of Nuclear Cardiology, 12, 261, 2005. 19. Tsui, B.M.W. et al. Correction of nonuniform attenuation in cardiac SPECT imaging, Journal of Nuclear Medicine, 30, 497, 1989. 20. Tsui, B.M.W. et al. Implementation of simultaneous attenuation and detector response correction in SPECT, IEEE Transactions on Nuclear Science, 35, 778, 1988. 21. Frey, E.C. and Tsui, B.M. In Quantitative Analysis of Nuclear Medicine Images, Zaidi, H, ed., New York: Kluwer Academic/Plenum Publishers, p. 141, 2005. 22. Glick, S.J. et al. Noniterative compensation for the distance-dependent detector response and photon attenuation in SPECT imaging, IEEE Transactions on Medical Imaging, 13, 363, 1994. 23. Boucher, L. et al. Respiratory gating for 3-dimensional PET of the thorax: Feasibility and initial results, Journal of Nuclear Medicine, 45, 214, 2004. 24. Nehmeh, S.A. et al. Effect of respiratory gating on reducing lung motion artifacts in PET imaging of lung cancer, Medical Physics, 29, 366, 2002. 25. Sang-Keun, W. et al. Development of a motion correction system for respiratory-gated PET study, Nuclear Science Symposium Conference Record, 2004 IEEE, Rome, Italy, 2004.

25 Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy Harald Paganetti

CONTENTS 25.1 Introduction ............................................................................................................... 567 25.1.1 The Concern for Scattered Doses in External Beam Radiation Therapy ....................................................................................... 567 25.1.2 Studies on Scattered Doses in External Beam Radiation Therapy....... 568 25.1.2.1 Photons ......................................................................................... 568 25.1.2.2 Protons .......................................................................................... 570 25.1.3 The Need for Computational Phantoms to Study Scattered Doses in External Beam Radiation Therapy .......................... 570 25.2 Methods ...................................................................................................................... 571 25.2.1 Implementation of Patient Computational Phantoms for Organ Dose Calculation into a Monte Carlo Environment ............................... 571 25.2.1.1 Computational Phantom Types ................................................ 571 25.2.2 Monte Carlo Implementation ..................................................................... 573 25.2.3 Radiation Quality Factors .......................................................................... 576 25.2.4 Dose Scoring ................................................................................................ 577 25.2.5 A Few Words on Computational Phantoms to Estimate the Risk for Second Malignancies ............................................................. 578 25.3 Results ......................................................................................................................... 579 25.4 Conclusion.................................................................................................................. 582 Acknowledgments ............................................................................................................... 583 References ............................................................................................................................. 583

25.1 Introduction 25.1.1 The Concern for Scattered Doses in External Beam Radiation Therapy Radiation oncology has seen the introduction of many improvements in patient treatment, e.g., advanced imaging, new chemotherapy agents, and new delivery techniques for radiation therapy. Together with early cancer detection, this has caused a steady increase in the probability of cure for cancer patients treated by radiation alone, or in combination with other therapies. The longer life expectancy posttreatment results in a growing concern with respect to long-term side effects, like radiation-induced cancer. 567

568

Handbook of Anatomical Models for Radiation Dosimetry

The most popular form of external beam radiation treatment are linear accelerators that produce x-ray beams leading to what is called conventional or 3D conformal therapy (3DCRT). Conventional radiation therapy has been evolved into intensity modulation radiation therapy (IMRT). In IMRT, a multi leaf collimator (MLC) is used to block part of the beam at any given time during the beam delivery to allow inhomogeneous irradiation of the target from each beam direction. The advantage is the potential of higher dose gradients between target structure(s) and organs at risk (OAR). Other radiation therapy modalities are, for example, electron, neutron, and proton therapy. The latter has recently gained momentum with various new facilities worldwide. Proton beam therapy causes less integral dose deposited in the patient compared to photon therapy while maintaining the prescribed dose to the target. Dose as a function of depth is deposited in a “Bragg curve” with the maximum being deep in the tissue.1 Due to its highly conformal dose distribution, proton therapy is advantageous for many treatment sites.2 Proton beam therapy comes basically in two flavors, using either passive scattering or beam scanning.3 Volumes in the patient receiving dose during treatment with radiation can be classified into three regions. There is the target (tumor), characterized by the gross tumor volume, clinical target volume, and planning target volume, treated with the therapeutic dose. In addition, there are OARs typically defined adjacent to the tumor. These may or may not intersect directly with the beam path and are allowed to receive low to intermediate doses according to dose constraints specified in a treatment plan. Such constraints are mostly based on clinical experience. Both the target structures as well as the OAR structures are identified based on tomographic images obtained prior to treatment. Finally, there is the rest of the patient body, which is typically not covered by any imaging procedure, which may receive low scattered doses.4 In the tumor and along the path of the therapeutic radiation field, one may have to accept a small risk for second malignancies because of the therapeutic benefit of radiation therapy. Further, since the goal is to kill tumor cells without leaving behind cells with the potential for mutation, the risk for radiation-induced cancer directly within the target area might be negligible. The dose to areas not directly irradiated carries a risk, however, that in some cases may not be justified. Of particular concern are pediatric patients because they have a long life expectancy after treatment and because their organs show a higher natural risk for developing second malignancies.5 Scattered dose in photon treatments is typically caused by photons from treatment head leakage or patient scatter. Only if the beam energy is above ~6–10 MV (depending on the material), is the threshold for nuclear interactions reached, resulting in the release of neutrons. In proton beams, scattered dose is caused mainly by neutrons and the particles generated by them. These neutrons originate either in the patient, or in the treatment head from which they can potentially reach the patient. The biological effectiveness of neutrons is less known than that of photons.6 Consequently, there is a concern about the carcinogenic effect of neutron dose in proton beam therapy. It has been discussed whether the use of proton therapy (and also IMRT) could result in a higher incidence of radiation-associated second cancers compared to conventional (3DCRT) radiotherapy.7–10 25.1.2 Studies on Scattered Doses in External Beam Radiation Therapy 25.1.2.1 Photons Studies on scattered doses in photon therapy, 3DCRT as well as IMRT, have been reviewed by Xu et al.4 Assessments of scattered doses have been done for 3DCRT11 but scattered doses are of concern in particular for IMRT treatments. Due to the partial blocking of

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

569

the beam with an MLC, the beam-on time in IMRT is higher than in 3DCRT. This causes an increase in delivered fluence (monitor units) to achieve the prescribed dose in the target. Because of this, IMRT fields have been studied extensively, for example by Sharma et al.12,13 Klein et al.14 measured the dose outside the treatment volume for different field sizes for 6 and 18 MV beams in IMRT. They analyzed the dose as a function of distance to the target showing that a 6 MV beam causes a greater dose at close distances while the dose from an 18 MV beam was higher at far distances due to greater leakage. Mansur et al.15 measured dose outside the treated volume for IMRT and 3DCRT in anthropomorphic computational phantoms. They found an increase in scattered dose when moving from 3DCRT to IMRT due to an increase in monitor units. Measurements to assess scattered dose using anthropomorphic computational phantoms were also done by Shepherd et al.16 Thermoluminescence detectors (TLDs) are useful to measure dose in vivo at various positions.17 Measurements using TLD were used by Verellen and Vanhavere.18 They estimated the whole-body dose equivalent in 6 MV photon beams to be 1.2 × 10−2 milli-Sievert per monitor unit (mSv/MU) and 1.6 × 10−2 mSv/MU for 3DCRT and IMRT, respectively. In addition to these types of measurements, Monte Carlo simulations have been used to assess scattered doses in photon beams. Mazonakis et al.19 used Monte Carlo to study the scattered dose and found that for a 6 MV photon beam, the scattered dose may vary from 2.6 to 4.4 cGy per treatment Gy depending on the patient. Kry et al.20 also used Monte Carlo to assess the scattered dose in 6 MV beams for different field characteristics. Most patient treated with photons will be exposed to 6 MV beams, thus no secondary neutrons can be expected. However, for some tumor locations, 18 MV beams are the method of choice. Several investigators focused in particular on neutron contamination in photon beams. Lin et al.21 as well as Reft et al.22 measured the neutron contamination in clinical photon beams. Kry et al.23 measured scattered photon and neutron doses using TLDs. They confirmed that the photon dose equivalent decreased with increasing distance from the central axis, independent of depth in the patient, whereas the neutron dose equivalent decreased with increasing depth in the patient but was not related to distance from the central axis. Vanhavere et al.24 estimated the whole-body effective dose for 6 MV IMRT and 18 MV 3DCRT using TLD measurements for photons and bubble detectors for neutrons. Neutron doses only dominated far away from the beam axis. Roy and Sandison25,26 estimated photon and neutrons doses in 18 MV beams using measurements in anthropomorphic RANDO and box polystyrene computational phantoms. Howell et al.27,28 investigated scattered doses in 6, 15, and 18 MV 3DCRT and IMRT beams for various different field sizes using TLD and Bonner Sphere detectors and an anthropomorphic computational phantom. Howell et al.27,28 also used Monte Carlo to model the treatment head and calculated neutron fluences from various elements in the delivery system. Equivalent doses were generated using quality conversion factors and depth correction factors recommended by the ICRP.29 The 6 MV beam resulted in a significantly lower effective dose than the 18 MV beam. Kry et al.30 used a Monte Carlo computational phantom for calculating dose outside conventional treatment fields from an 18 MV beam. The authors determined neutron fluences for various field sizes and locations and concluded that neutrons contribute the predominate component of dose equivalent outside the treatment field at large distances from the central axis. Based on Monte Carlo simulations of neutron spectra, Ongaro et al.31 showed neutron dose equivalent values of between 1 and 4.8 mSv/Gy. Other Monte Carlo simulations include those by Difilippo et al.32. They simulated a humanoid computational phantom exposed to 18 and 25 MV beams. Close to the main field, the contribution of dose equivalent from neutrons was found to be at least one order of magnitude lower than the dose equivalent from photons. Chibani and Ma33 considered the effects of field size, primaryelectron energy, and machine type on the dose for high-energy photon beams using a

570

Handbook of Anatomical Models for Radiation Dosimetry

tissue equivalent computational phantom. Again, the dose from neutrons was orders of magnitude lower than the photon dose. In other Monte Carlo studies, Zanini et al.34 and Barquero et al.35 determined the neutron background for different positions relative to isocenter and different field parameters produced by a clinical 18 MV beam. 25.1.2.2 Protons In proton therapy, the amount of neutrons generated in the patient increases with the proton beam range (i.e., the beam energy) and treatment volume. The neutron yield generated in the treatment head depends on various geometrical and physical parameters. The treatment head is typically the dominant neutron source compared to the patient contribution.36 Passive scattering uses various scatterers, beam-flattening devices, collimators, and energy modulation devices. Further, for each patient, individual apertures and compensators are required. Neutrons are produced primarily in the field-specific aperture because passive scattered proton beam delivery systems typically offer only a limited set of different field sizes impinging on the final aperture, thus influencing the neutron yield due to the fraction of the proton beam blocked by the aperture.10,37 The fact that neutrons are generated in proton beams has led several investigators to study scattered radiation in proton beam therapy, in particular in passive scattering systems. One of the first were Binns and Hough38 who measured the secondary dose in a 200 MeV proton beam under design for clinical use. Neutron equivalent doses were measured by Yan et al.39 in a 160 MeV proton beam with a passive scattering beam delivery system. They found neutron equivalent doses of 1–15 mSv per treatment dose. Anthropomorphic computational phantoms were used by Roy and Sandison26 in a passive scattered proton beam. They deduced neutron equivalent dose between 0.1 and 0.26 mSv/Gy for a beam energy of 198 MeV. In another study, Mesoloras et al.40 assessed the dependence of equivalent neutron doses on several beam parameters, e.g., aperture size and air gap. The neutron equivalent dose varied from 0.03 to 0.87 mSv/Gy. Other measurements using anthropomorphic computational phantoms and microdosimetric detectors found equivalent doses from 3.9 to 0.18 mSv/Gy.41 Tayama et al.42 measured neutron equivalent doses up to 2 mSv/Gy outside of the primary radiation field in a 200 MeV proton beam. Like in photon beams, Monte Carlo simulations have been used. Agosteo et al.43 used Monte Carlo to study scattered doses in passive beam delivery systems. The absorbed dose due to neutrons was in the range from 3.7 × 10−7 to 7.1 × 10−2 Gy per treatment Gy depending on the distance to the field and the chosen beam parameters. Further, Polf and Newhauser44 simulated neutron doses for a passive scattering delivery system and found that the neutron dose decreased from 6.3 to 0.63 mSv/Gy with increasing distances to isocenter from 50 to 150 cm. For proton beam scanning, a proton pencil beam is magnetically scanned over the target volume without the need for scattering, flattening, or compensating devices. Scanned proton beams minimize interaction with devices in the treatment head and can therefore be expected to cause negligible neutron contamination. This was confirmed in several studies.43,45 It is thus undisputed that scanned proton beams can offer a lower second cancer risk than passive scattered protons or even photons.45,46 25.1.3 The Need for Computational Phantoms to Study Scattered Doses in External Beam Radiation Therapy Measurements or simulations of scattered doses under various conditions like the ones mentioned above are useful to understand the relative differences among treatment

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

571

modalities or beam conditions. Ultimately, however, one is interested in estimating the actual risk for developing a second malignancy due to scattered radiation. Based on measurements of whole-body effective doses, Followill et al.47 compared IMRT and 3DCRT with respect to the relative risk for radiation-induced cancers. The whole-body dose equivalent for 25 MV beams was found to be eight times greater than that for the 6 MV beams. Consequently, the risk of any fatal secondary cancer was estimated to be 0.4% and 24.4% for the 6 and 25 MV cases, respectively. Kry et al.9 used measured dose equivalents to calculate the risk of fatal second malignancies from IMRT applying risk coefficients recommended by the NCRP.48 They concluded that the risk of a radiation-induced fatal second malignancy was significantly lower for 3DCRT compared to IMRT treatments. Along the same lines, others18,24 used probability coefficients for a lifetime risk of excess cancers given by the ICRP49 and found that IMRT might increase the second cancer risk by a factor of 8 compared to conventional treatments. The lifetime risk for an excess fatal cancer was calculated to be 5%.24 Like the dosimetric studies discussed above, these risk analysis are based on wholebody effective doses and organ weighting factors. While this certainly gives an impression about the scattered dose to be expected, in particular if different modalities are compared, epidemiological risk assessments have to be based on organ-specific equivalent dose estimations. Organ equivalent doses are cumbersome to measure. Monte Carlo simulations, alternately, allow a realistic consideration of arbitrary patient-specific geometries. Because of the concern of excessive radiation with most imaging techniques, whole body scans are rarely available and patient geometry is obtained using computed tomography for treatment planning purposes only for volumes directly irradiated with the treatment beams. Thus, in order to perform epidemiological meaningful studies, the use of computational phantoms in combination with Monte Carlo transport calculations is a valuable option. The following sections describe the use of computational phantoms to determine organspecific scattered equivalent doses in radiation therapy. Monte Carlo simulations, together with patient-sculpted whole-body computational phantoms, provide information for proper analysis of the risk for developing second malignancies due to scattered radiation. These kinds of simulations could potentially provide dosimetric information to improve risk computational phantoms based on long-term follow-up of radiation therapy patients and the knowledge about the organ doses they received during the course of their treatment for the primary cancer.

25.2 Methods 25.2.1 Implementation of Patient Computational Phantoms for Organ Dose Calculation into a Monte Carlo Environment 25.2.1.1 Computational Phantom Types The simpler the geometry, the faster a Monte Carlo simulation typically is. Consequently, initially, dose calculations were based on stylized computational phantoms.50–55 These computational phantoms are built out of simple geometrical shapes and a distinction is drawn only between bone, soft tissue, and lung. Stylized computational phantoms have been applied in a variety of simulations for radiation protection, nuclear medicine, and medical imaging.49,54,56–59 Work has been done on organ doses from medical exposures

572

Handbook of Anatomical Models for Radiation Dosimetry

using stylized computational phantoms60,61 and to derive dose–response relationships for patients in epidemiological studies. Because human anatomy is more complex than those modeled with stylized representations, results based on such computational phantom calculations are controversial and uncertainties may be significant.62,63 For example, simulated organ and marrow doses based on stylized computational phantoms have not produced strong correlations with radiotoxicity.62 A better representation of the human body is possible by using voxel computational phantoms, where each voxel is identified in terms of tissue type (soft tissue, hard bone, etc.) and organ identification (lungs, skin, etc.).64 Lee et al.65 analyzed the differences between the use of stylized computational phantoms and the use of voxel computational phantoms and found dosimetric differences of up to 150% in some organs. Others have compared doses based on stylized and voxelized showing differences in organ doses to be as high as 100%.65–68 The discrepancies were explained by the geometrical considerations in the stylized computational phantom, namely relative positions of organs as well as individual organ shapes. Obviously, in order to accurately assess scattered doses to different organs, voxel computational phantoms considering a variety of tissues and a variety of organs are required. Figure 25.1 illustrates the main differences between a stylized and a voxel computational phantom. There are numerous whole-body voxel computational phantoms available. The many different flavors of voxel computational phantoms are reviewed by Zaidi and Xu.64 One popular voxel computational phantoms is the adult male computational phantom, VIP-Man,69 developed based on anatomical color images of the Visible Man from the Visible Human Project by the National Library of Medicine.69,70 It has a resolution of 0.33 × 0.33 × 1 mm3.54 The VIP-Man computational phantom has been used for Monte Carlo studies of organ doses for photons, electrons, neutrons, and protons.36,68,69,71–79 Researchers have recognized that scattered doses in radiology and radiation therapy are of particular concern for the pediatric patient population. Thus, there was a need for pediatric computational phantoms. Such computational phantoms cannot be obtained

FIGURE 25.1 Torso of a stylized model (left) and a voxel model (VIP-Man model, right).

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

573

by scaling an adult computational phantom because of the differences in relative organ position, relative organ sizes, and even organ composition as a function of a person’s age. A series of five computational phantoms of different ages were constructed from CT images of live patients for use in medical dosimetry.80–82 The computational phantoms approximate the bodies of a 9 month, 4 year, 8 year, 11 year, and 14 year old, with resolutions between 0.43 × 0.43 × 3.0 mm3 and 0.625 × 0.625 × 6.0 mm3. Age-interpolated reference body masses, body heights, sitting heights and internal organ masses as well as changes in geometry and material composition as a function of age and gender were assigned based on ICRP references.83 These voxel computational phantoms are state-of-the-art whole-body representations for use in Monte Carlo dosimetry. Nevertheless, new developments are under way because voxel computational phantoms also have shortcomings. They are based on CT images of live patients or, sometimes, cadavers. Organ contours are manually segmented in these images. Uncertainties are introduced because of image noise and because some representations of mobile organs may be blurred. Further, in order to match a particular patient as closely as possible, one might have to interpolate between two different computational phantoms of a specific age. In voxel computational phantoms, body dimensions can only be modified by changing the voxel resolution, which generally limits the modification to uniform scaling. For example, creating a non-50th percentile individual from a reference 50th-percentile cannot be done realistically using voxel computational phantoms, where the distribution of subcutaneous fat can be highly unique to individual patients. To overcome these limitations, voxel data can be combined with surface equations to design hybrid computational phantoms. In these computational phantoms, one can adjust the boundary of an organ to the desired shape and volume using patient-specific images and deformable image registration. A series of reference (i.e., 50th height/weight percentile) pediatric hybrid computational phantoms based on non uniform B-spline fits (NURBS) surfaces has been developed.84 One of the first pediatric hybrid-computational phantoms was that of a newborn child.85 One of the most sophisticated hybrid approaches to model construction has been made in nuclear imaging.86 Conventional mathematical descriptions were replaced by equations with NURBS.87 Segars et al.88–90 developed the 4D NURBS-based cardiac-torso computational phantom that is used as a realistic and deformable computational phantom to simulate SPECT images and respiratory motion.91 25.2.2 Monte Carlo Implementation Initially, computational phantoms have been used in combination with analytical dose computational phantoms. Diallo et al.92 estimated the dose to areas outside the target volume using a whole-body computational phantom with 151 control points. However, Monte Carlo methods are typically the method of choice when it comes to calculating scattered (or secondary) radiation doses. Besides realistic computational phantom geometry, the irradiation field has to be known accurately as well. The scattered radiation generated in the patient (computational phantom), as well as the scattered radiation generated in the treatment head and then reaching the patient, needs to be modeled. Figure 25.2 shows an example of a proton therapy treatment head with many different beam shaping and beam monitoring devices. Monte Carlo simulations of treatment heads have been reported extensively; for example, for photons93–111 and protons.112–115 The beam entering the treatment head typically has to be characterized based on parameterizations obtained from measurements.93,108,110,115,116 Such treatment head simulations typically characterize a radiation field using a phase

574

Handbook of Anatomical Models for Radiation Dosimetry

FIGURE 25.2 (See color insert following page 524.) Proton therapy treatment head (at Massachusetts General Hospital) to illustrate the complexity of the geometry. Two particle tracks are drawn as well.

space file. A phase space typically records the spatial, energy, and angular information of the particles exiting the treatment head at a predefined scoring plane. However, different information can be scored depending on the scope of the study. For example, one might be interested to separate the scattered dose that a patient would receive from the treatment head and the scattered dose produced in the patient in passive scattering proton therapy (see Figure 25.3).36,117 In order to assess the relative contribution of these two neutron sources to the organ equivalent dose, one can either divide the phase space into a proton component and a neutron component or define a flag in the file that codes the information about the particle type. Similar definitions can be made for photon simulations dealing with either external neutrons or scattered photon radiation. Flags can also be set to identify particles that originate from specific devices in the treatment head. Dose calculations are conducted with source particles read from the phase space files. This allows multiple simulations based on the same field characteristics but different computational phantom geometries without having to track all particles through the treatment head again. Several codes have been used to study low scattered dose in radiation therapy. These are mainly EGSnrc118 and MCNP119 for assessing scattered photon doses,19,120 MCNPX121 for assessing neutron and photon doses in photon beams20,30,32,33,35 and proton beams,44,122 as well as FLUKA19,120 and Geant4123 to assess scattered doses in proton beams.36,43,117 When it comes to implementing computational phantoms, most Monte Carlo codes can easily handle simple geometrical structures, like boxes, spheres or tubes. Thus, the implementation of stylized computational phantoms into a Monte Carlo environment is typically straightforward. Mazonakis et al.19 used the MCNP code to study the scattered dose to the thyroid during 6 MV photon cranial irradiation for individuals at ages 3–18 years old. The stylized

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

575

FIGURE 25.3 External beam impinging on a patient (simulated with a whole-body model). In the patient, the light grey arrow corresponds to an internally created secondary particle, while the dark grey arrow shows an example for external radiation scattered in the patient.

anthropomorphic computational phantoms were based on description by Cristy and Eckerman53,124 and scaled depending on the assumed age, weight, and height of the patient. A different stylized computational phantom was used to assess the risk for a second malignancy from various radiation therapy treatment modalities.120 Dose to the prostate, bladder, and rectum was analyzed. Another example of computational phantoms in a Monte Carlo environment is the work by Barquero et al.35 They used MCNPX to estimate neutron doses to critical organs during 18 MV photon radiation therapy treatments. Many organs and anatomical regions, including the skin, the muscle, and the bone, were considered. The stylized computational phantom125 did approximate the human body using cylindroids of circular and elliptical cross sections of the outer body and ellipsoids, spheres, semispheres, rhomboids, and cylinders for various organs. Difilippo et al.32 studied scattered doses in various organs for high energy photon beams and secondary neutrons using a humanoid computational phantom within the MCNPX code. The computational phantom was an adult male stylized computational phantom.126 Scattered doses for other organs were simulated assuming a brain irradiation with 18 and 25 MV photons. Besides these photon studies, neutron doses from proton therapy were calculated by Agosteo et al.43 using FLUKA. Others have also implemented the Alderson–Rando computational phantom geometry into the Geant4 code.127 All these studies were done with stylized computational phantoms of various detail implemented in a Monte Carlo code. In order to use whole-body computational voxel computational phantoms in Monte Carlo codes, these either have to be able to handle voxelized geometries, i.e., a large amount of individual voxels, or to incorporate contoured organ shapes via surface equations. For dose calculations involving real patients, the information stored for each CT voxel is a Hounsfield number, which reflects the attenuation coefficient of tissues to diagnostic x-rays. In contrast, for computational phantom simulations each voxel is usually tagged with a specific material composition and density. The VIP-Man was implemented into four Monte Carlo codes, EGS4,68,73,74 MCNP,72 MCNPX,71 and Geant436,117 to calculate organ doses for internal electrons,74 external photons,68 external electrons,73 external neutrons,71,72 and external protons.36,117 Recently,

576

Handbook of Anatomical Models for Radiation Dosimetry

Xu et al.128 implemented a pregnant female computational phantom based on voxelization of a boundary representation into the Monte Carlo codes EGS4 and MCNPX. Pediatric voxel computational phantoms have been used within Geant4 as well.117 There are quite a few studies on scattered doses using the VIP-Man. Some of the work focuses on calculating fluence-to-dose conversion factors.68,71,73 The VIP-Man was also incorporated into a proton therapy Monte Carlo system (based on Geant4) as “virtual patients” to assess organ-specific doses.36,117 The composition of VIP-Man tissues/materials was done according to the ICRU.129 The work on neutron dose contamination in proton therapy based on the VIP-Man computational phantom was subsequently extended in considering pediatric patient anatomies117 based on five computational phantoms.81,82 For each organ and computational phantom, age- and gender-dependent densities as well as age-dependent material compositions were adopted based on ICRU.57 Thus, the computational phantoms did not only take into account changes in geometry as a function of age, but also changes in organ-specific material composition as a function of age based on individuals at the ICRP reference ages.83 25.2.3 Radiation Quality Factors When analyzing the dose to different organs with the aim of identifying their biological effect, radiation specific weighting factors have to be used, taking into account the contributions from different particles. Different particles or different radiations are often classified using their linear energy transfer (LET). There is not a one-on-one relationship between LET and biological effect because the latter is influenced by the track structure, i.e., the microscopic energy deposition patterns. The parameter used to compare different radiations in radiation therapy is the relative biological effectiveness (RBE). The RBE is defined as the ratio of the doses required by two radiations to cause the same level of effect. Thus, the RBE depends on the dose (typically increases with decreasing dose) and the biological endpoint. It is typically defined relative to 60Co for a specific effect (e.g., surviving fraction or mutation yield). RBE values are often based on cell survival data because this is the main endpoint of interest in radiation therapy. One can expect differences in RBE for cell survival compared to cell mutation.130 For radiation protection purposes, one is interested in defining a parameter that is mostly independent of dose and biological endpoint (e.g., a maximum RBE). This is for three reasons. First, dose levels of interest in radiation protection are typically low; second, recommendations for the general public should be easy to understand; and third, a radiation protection recommendation does not aim at accuracy but at a conservative guideline. For low doses, the ICRP defines the “radiation weighting factor,”49 which superseded the quantity of “quality factor,” as a conservative and simple measure of radiation effect. The ICRP also defines the “equivalent dose” as the average absorbed dose in an organ or tissue, modified by the radiation weighting factor for the type, and sometimes the energy, of the radiation.49 These weighting factors convert absorbed dose in Gy to Sievert (Sv). In turn, the radiation protection quantity, the “effective dose,” normalizes partial-body exposures to various internal and external irradiations in terms of whole-body stochastic risk.49 The effective dose is used to recommend an occupational dose limit for radiation protection and is not measurable or additive because it depends on not only radiation weighting factors but also tissue-weighting factors for specific organs. The effective dose concept should not be used to indicate risk for specific individuals.49 For organ dose calculations aiming at a detailed analysis of biological effects, like for example the risk for developing a second cancer, equivalent doses are more meaningful.

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

577

For γ-rays, fast electrons, and x-rays a radiation weighting factor of 1 can be assumed49 (although there is evidence based on chromosomal aberration data and on biophysical considerations that, at low doses, the effectiveness per unit absorbed dose of standard x-rays may be about twice that of high-energy photons). The ICRP recommends for photons and electrons a radiation weighting factor of 1, for protons a weighting factor of 2, and for alpha particles a weighting factor of 20.131 Note that in proton radiation therapy, a constant RBE for protons of 1.1 is used.130 Based on experimental data, one can assume that the RBE of protons depends on dose and on the proton energy.130 Although the BEIR report5 recommends a neutron radiation weighting factor of 10 in all situations, the neutron weighting factor clearly depends on neutron energy. The ICRP defines an energy-dependent bell-shaped curve with a maximum weighting factor of 20 at around 1 MeV.49,131 It has been discussed that the neutron quality factors are subject to significant uncertainties, which could affect risk estimations.6,132,133 The ICRP quality factors may not be accurate for extremely low doses.134 Energy averaged neutron quality factors in the human body based on the ICRP curve are typically between 2 and 11.36,39,41 However, much higher neutron RBE values have been found for various endpoints in vivo and in vitro.135–137 The NCRP has shown neutron quality factors of more than 80 for fission neutrons considering several radiation endpoint in the energy range of 1–2 MeV, where the ICRP recommendation assumes a weighting factor of 20.137 Dennis135 has reviewed experimental neutron RBE data and found maximum values at low doses of up to 71 in vivo. Ambiguities in weighting factor assignments exist for uncharged particles. For example, neutrons deposit their energy mostly via secondary protons. Nevertheless, the maximum quality factor recommendation for neutrons is 20, while the factor for protons is constant and equals 2. When estimating equivalent doses under various conditions, e.g., in the case of a patient treated with radiation therapy, the dose rate has to be taken into account. Radiation therapy is typically delivered in multiple fractions, e.g., in 30 consecutive days (typically excluding weekends). Most risk models are valid for a single irradiation. The difference in effect between a single fraction and a multiple fraction irradiation with the same dose is due to the difference in repair capability of the tissues. In order to account for this effect, a dose and dose rate effectiveness factor (DDREF) has to be applied. For example, the BEIR committee5 recommends the use of an average correction factor of 1.5 to take into account fractionation when using dosimetric data for risk analysis for solid tumors and linear dose–response relationships. While this is appropriate for photon radiation, equivalent doses from high-LET radiation, like neutrons, should not be scaled using DDREF when dealing with low dose exposure because of the different biological mechanisms with which neutrons interact with tissues.133 25.2.4 Dose Scoring There are different ways to determine equivalent doses using Monte Carlo simulations, as discussed by the ICRU.138 One possible strategy is to calculate the average absorbed dose for the organ under consideration and scale the dose with an average radiation-weighting factor. Another approach frequently used44,122 is to calculate the particle fluences at the surface of a region of interest (organ) and then use energy-dependent fluence-to-equivalent dose conversion coefficients.68,71–73,139,140 In this case, dose deposition events are not explicitly simulated. The latter method has some advantages when dealing with neutron radiation. Neutrons interact infrequently with materials. Fast neutrons lose most of their kinetic energy in the initial relatively small number of interactions. In the low/thermal energy region, there is a decreasing probability for neutrons to slow down causing a large

578

Handbook of Anatomical Models for Radiation Dosimetry

number of elastic scatterings in soft tissues. This is why neutron energy distributions in the patient are dominated by low energy neutrons.36 Thus, Monte Carlo simulations are typically quite time consuming (in order to achieve a reasonable statistical accuracy) when based on the dose actually deposited via neutrons. The method of simulating all energy deposition events can be further divided into whether the radiation weighting factor is applied on an average basis or during the simulation at each energy deposition event.117,141 Application of radiation weighting factors on a step-by-step basis should, by definition, be the most accurate method. Each energy deposition event is classified by the particle type and particle origin. For example, in a simulation of organ-specific neutron equivalent doses in proton beam therapy,117 each history of an event was searched for a neutron in the interaction history of the dose depositing particle. If a neutron was found, the dose deposition was considered to be due to a neutron and a neutron radiation-weighting factor was then assigned. Similarly, if a proton from a proton chain deposited the absorbed dose, the dose depositions would be classified as proton induced. For each interaction chain history, a division into different groups can be done depending on particle energy in order to apply energy-dependent quality factors. Different dose scoring methods were compared by Zacharatou-Jarlskog and Paganetti.141 A classification of dose deposition events by particle type, particle history, and particle energy was used to select the radiation-weighting factor on a step-by-step basis. Alternatively, the calculation was done by averaging the deposited dose over the total number of simulated events irrespective of particle type, origin, and energy. The average organ dose was then scaled by a mean neutron radiation-weighting factor. This factor was calculated by weighing the fluence distribution of the respective particles entering the organ. For neutron equivalent doses in proton beam therapy, it was found that the averaged values seem to underestimate the neutron equivalent dose in comparison to those calculated on a step-by-step basis. The difference was found to be around 25% depending on organ and field specifications.141 25.2.5 A Few Words on Computational Phantoms to Estimate the Risk for Second Malignancies Quite often, risk estimates are based on whole-body effective doses and organ weighting factors.48 The NCRP defines probabilities of fatal cancer for the bladder, the bone marrow, the bone surface, the breast, the esophagus, the colon, the liver, the lung, the ovary, the skin, the stomach, the thyroid, and the remainder of the body.48 The ICRP as well defines a whole body effective dose with organ-specific weighting factors.49 Tissue-weighting factors employed by the NCRP and ICRP for the effective dose represent sex and age averaged values applying a radiation independent dose rate correction. Thus, these computational phantoms are rough approximations, which yield a nominal risk value of 5 × 10−2 Sv−1. Effective doses are suited for radiation protection studies but not for risk models for secondary cancer, which are site specific. For detailed risk estimations of epidemiological value, the BEIR5 report provides formalisms to calculate risks of cancer incidence and mortality for a variety of organs. Dose– response relationships are typically defined as a function of age, gender, and site. The cancer incidence rate at a given point in time is defined as the ratio of diagnosed individuals in a time interval divided by the interval duration and the total number of unaffected individuals at the beginning of this interval. Cancer risk, on the other hand, is defined as the probability for disease occurrence in the population under observation, i.e., risk equals the ratio of diagnosed to total number of individuals in the given time interval. In order to

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

579

obtain a measure of the relation between the incidence rate in the exposed population and the incidence rate in the unexposed population, one can use either their difference or their ratio. Excess relative risk models express the risk as relative to the background risk. The excess risk can be calculated as a function of attained age of the individual, age at exposure, dose received, sex index, and an index denoting population characteristics. Several different risk models have been applied by researchers to estimate the risk of second malignancies induced by radiation. The computational phantoms in use today are largely based on the atomic bomb survivor data. Both the BEIR VII Committee5 and the ICRP49 recommend a linear dose–response relationship without a low-dose threshold based on the epidemiological data obtained from Japanese atomic bomb survivors for doses below 0.1 Gy. This population was exposed to a single fraction of between 0.1 and 2.5 Sv. The radiation field, dose, and dose rate were certainly much different than the radiation fields in radiation therapy. However, extracting dose–response relationships from patient data is associated with large statistical uncertainties.142 At low doses, none of the epidemiological data are sufficient to establish the shape of the dose–response relation and larger studies are required to quantify the risk to a useful degree of precision.143 One reason for considerable uncertainties in risk models is the fact that actual second cancer incidences are difficult to interpret because of the lack of accurate dosimetric information.

25.3 Results This section summarizes some of the results on Monte Carlo simulations for proton therapy using whole-body computational voxel (non stylized) computational phantoms to assess organ equivalent doses. Such data are very valuable to understand the impact of scattered doses on cancer risk as well as to compare different radiation therapy modalities. In proton therapy, scattered neutron radiation can potentially be produced in the treatment head (external neutrons) and in the patient (internal neutrons). Jiang et al.36 simulated organ doses caused by neutrons generated in a therapeutic proton beam assuming treatments of a tumor in the head and neck region and a tumor in the lung. The applied voxel computational phantom was the whole-body VIP-Man computational phantom. The simulations were based on the Geant4 Monte Carlo code123 and its proton therapy implementation.115,144,145 This system uses a parameterization of the proton beam at the entrance to the treatment room and a detailed computational phantom of the treatment head.115 The treatment head incorporates the different settings (combinations of scatterers, variable jaws, etc.) necessary to simulate hardware configurations for each treatment field. Other than in 3DCRT photon treatments, proton therapy requires field-dependent adjustment of treatment head parameters. As described above, the first step of the simulation is to generate phase space distributions at the exit plane of the treatment head. These are then used to “irradiate” different computational phantom geometries. Based on different phase space files, a detailed analysis of the origin of scattered radiation can be done. For example, it can be analyzed if scattered neutron radiation depositing dose in a specific organ was caused by neutrons produced in the patient via interactions of primary protons with tissue, and the later generations of neutrons originated from them, or by neutrons generated in the treatment head and also the next generations of neutrons generated by them in the patient. This allows identifying areas where additional shielding or design changes have to be looked at.

580

Handbook of Anatomical Models for Radiation Dosimetry

The quantities of interest to determine organ equivalent doses were the yields of neutrons on the surface of various organs, the absorbed doses to these organs, and the energy distribution of neutrons. The latter is important to calculate the radiationweighting factor. The result was that the whole body effective dose for a patient treatment based on a lung tumor plan was about six times higher than that for a head and neck plan. This is because of the higher dose prescription, the bigger target volume and the fact that many organs with a large organ-weighting factor were close to the target. The average neutron dose equivalent in mSv per treatment Gy for organs in the abdomen was 1.9 and 0.2 mSv/Gy for the lung and head and neck treatment plan, respectively. Also simulated were the doses to the red bone marrow, resulting in ~20 Sv to about 10% of the bone marrow volume. The study revealed that externally produced neutrons dominated the scattered neutron dose equivalent for most organs. The dosimetric data on organ doses were later used to estimate the potential risk of a second malignancy and to assess the impact of different scenarios with respect to neutron radiation weighting factor assignments.6 The work by Jiang et al.36 focused on doses to an adult patient (using the VIP-Man computational phantom). The work was later extended by Zacharatou-Jarlskog et al.117 They simulated proton beam therapy for pediatric patients and considered several proton fields of varying field size, beam range, and modulation width for the treatment of tumors in the intracranial region. To simulate age and organ-specific equivalent doses, one adult computational phantom and five pediatric computational phantoms (a 9 month old, a 4 year old, an 8 year old, an 11 year old, and a 14 year old) were considered. Organ doses were presented as a function of organ index for up to 48 different organs and structures. The results showed how the organ-specific neutron equivalent doses vary as a function of various treatment field parameters. Further, the variation of dose among different organs is caused by differences in volume, differences in their distance to the target and by differences in their elemental composition. For example, a bigger range in tissue requires a higher beam energy and thus more material (tissue) is needed to slow down the proton beam. Consequently, simulations based on the 4 year old voxel computational phantom resulted in about 1.3 mSv/Gy in the lungs for low-range fields and about 2.7 mSv/Gy for fields with a high beam range. Further, neutron equivalent doses to organs increased with treatment volume simply because the number of protons necessary to deposit the prescription dose in the target increases with target volume. For all considered treatment fields, external neutrons dominated the neutron dose because of protons hitting the patientspecific aperture. Thus, the neutron equivalent dose due to external neutrons increases with decreasing field size.10,37 It was found that for a small field the contribution of neutrons from the treatment head can be close to 99%, while for a large volume being treated it can go down to ~60%. Other than stylized computational phantoms, age-dependent voxel computational phantoms allow a detailed analysis of scattered doses considering patient’s age or trunk height.117 The data show a considerable dependency of the organ dose on patient’s age. For example, younger patients are exposed to a higher neutron contribution from the treatment head because of their smaller body. With increasing distance to the target, doses vary more significantly with patient age. For example, simulation based on a 9 month old computational phantom showed a ~50% higher dose to the thyroid as simulations based on an adult computational phantom. In the case of esophagus, the relative increase in dose between the adult and the 9 month old computational phantoms was roughly a factor of 4. The maximum neutron equivalent dose delivered to an organ was simulated as ~10 mSv

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

581

3.5

Neutron equivalent dose per treatment Gy (mSv/Gy)

3.0 2.5 2.0 1.5 1.0 0.5

Liver

Stomach Gall bladder Lower humerus Aorta Kidneys Spleen Pancreas Adrenals Colon L-vertebrae Small intestine Radii, ulnae Rectosigmoid Sacrum Os coxae Upper femur Hand Bladder Ovaries Patellae Lower femur Uterus Fibula Ankle, feet

Ribs

Heart Breast

Pharynx Lenses Tonsil Larynx C-vertebrae Salivary glands Trachea Thyroid Tongue Scapulae Mandible Esophagus Upper humerus Thymus Clavicles Spinal cord Lungs T-vertebrae Sternum Bronchi

0.0

FIGURE 25.4 Scattered dose when applying a brain treatment field in proton beam therapy simulated for various organs with age-dependent voxel models and Monte Carlo transport calculations. (Data from Zacharatou–Jarlskog, C. et al., Phys. Med. Biol., 53, 693, 2008.)

per treatment Gy.117 Figure 25.4 shows the neutron equivalent doses for various organs for a typical brain treatment field. The field was used to treat a 9 year old patient at the Francis H Burr Proton Therapy Center in Boston. The organ doses were simulated for the closest matched voxel computational phantom, i.e., the 8 year old. The data are drawn based on the results published by Zacharatou-Jarlskog et al.117 Obviously, the ultimate goal is to estimate the potential risk for developing a second malignancy for patients undergoing radiation treatment. Based on the organ-specific neutron equivalent dose reported by Zacharatou-Jarlskog et al.,117 the risk for developing a second malignancy posttreatment was estimated based on risk computational phantoms published in the BEIR report.5 Figure 25.5 shows an example of the resulting data for the thyroid146 as based on simulations of the proton treatments of two 9 year old patients with brain tumors. The diagram gives the risk and the baseline risk (risk for the non irradiated population). The figure also illustrates what the risk analysis would have shown if an adult computational phantom (the VIP-Man) had been used for the Monte Carlo dose calculation. The discrepancy between adult and pediatric patient is due to age-dependent cancer risks and due to geometric differences in the anatomy. Thus, it reveals that for proper risk analysis not only voxel computational phantoms, but also age-dependent voxel computational phantoms are desirable. Thyroid cancer is the second most frequent second cancer. Ronckers et al.147 analyzed thyroid cancer cases in childhood cancer survivors. A trend for decreasing risk with increasing age at irradiation has been reported.148 This is due to a greater radiation effect in humans during the period of rapid cell proliferation, i.e., during development of the thyroid gland.149

Handbook of Anatomical Models for Radiation Dosimetry

582

Lifetime attributable risk (%)

0.8

0.6

0.4

0.2

2 se lin ec hi ld Ba

ﬁe ld Ch

ild

ﬁe ld

ea

ild Ch

se lin

1

lt du

2 ﬁe ld Ba

lt Ad u

Ad u

lt

ﬁe ld

1

0.0

FIGURE 25.5 Lifetime attributable risk (LAR) for second cancer incidence in the thyroid for a 9 year old treated for a brain tumor with a prescribed therapeutic dose of 70 Gy.146 Two fields that were used clinically are shown. The fi rst two columns show the risk for an adult patient. The forth and fi fth column give the estimated risk for a pediatric patient. Shown are also the respective baseline risks. Calculations are based on dose simulations using voxel models of an adult and an 8 year old117 and applying risk models published in the BEIR report. 5

25.4 Conclusion This chapter discussed the origin and magnitude of scattered doses in external beam radiation therapy. Further, it illustrated the value of whole-body computational phantoms in understanding scattered doses and in providing organ-specific dosimetric information necessary for risk analysis. Monte Carlo transport calculations based on whole-body computational phantoms in combination with epidemiological risk computational phantoms can give valuable information about the risks of different radiation modalities as a function of various beam parameters and even the patient’s age. Risk computational phantoms show considerable uncertainties because second cancer incidences are difficult to interpret due to the lack of accurate dosimetric information. The BEIR VII report5 states that, “Epidemiologic studies, in general, have limited ability to defi ne the shape of the radiation dose–response curve and to provide quantitative estimates of risk in relation to radiation dose, especially for relatively low doses. To even attempt to do so, a study should (1) be based on accurate, individual dose estimates, preferably to the organ of interest. Although progressively larger epidemiological studies are required to quantify the risk to a useful degree of precision,143 the work on whole-body computational phantoms and organ doses could potentially provide organ-specific dosimetric information that could help to improve risk computational phantoms.

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

583

Acknowledgments The author would like to thank the following individuals for many fruitful discussions regarding the subject: X. George Xu and Bryan Bednarz (Rensselaer Polytechnic Institute), Christina Zacharatou-Jarlskog and Herman D. Suit (Massachusetts General Hospital), Bernard Gottschalk (Harvard University), Wesley Bolch (University of Florida, Gainesville), and David Brenner (Columbia University). Furthermore, the author would like to acknowledge the support of NIH grants RO1 CA116743-01 “Virtual Patients for Computing Radiation Doses” and PO1 CA21239-26 “Proton Radiation Therapy Research”.

References 1. Paganetti, H. and T. Bortfeld, Proton therapy. In: New Technologies in Radiation Oncology; Eds. Schlegel, W.; Bortfeld, T.; Grosu, A.L.; Springer-Verlag, Heidelberg, ISBN 3-540-00321-5, 2005. 2. Wilson, V.C., J. McDonough, and Z. Tochner, Proton beam irradiation in pediatric oncology: An overview. J Pediatr Hematol Oncol, 2005. 27: 444–448. 3. Blattmann, H., Beam delivery systems for charged particles. Radiation and Environ Biophys, 1992. 31: 219–231. 4. Xu, X.G., B. Bednarz, and H. Paganetti, A review of dosimetry studies on external-beam radiation treatment with respect to second cancer induction. Phys Med Biol, 2008. 53: R193–R241. 5. BEIR, Health Risks from Exposure to Low Levels of Ionizing Radiation, BEIR VII, Phase 2. National Research Council, National Academy of Science, The National Academies Press, Washington, DC, 2006. 6. Brenner, D.J. and E.J. Hall, Secondary neutrons in clinical proton radiotherapy: A charged issue. Radiother Oncol, 2008. 86(2): 165–170. 7. Hall, E.J., Intensity-modulated radiation therapy, protons, and the risk of second cancers. Int J Radiat Oncol Biol Phys, 2006. 65(1): 1–7. 8. Hall, E.J. and C.-S. Wuu, Radiation-induced second cancers: The impact of 3D-CRT and IMRT. Int. J. Radiat. Oncol. Biol. Phys., 2003. 56: 83–88. 9. Kry, S.F. et al., The calculated risk of fatal secondary malignancies from intensity-modulated radiation therapy. Int J Radiat Oncol Biol Phys, 2005. 62(4): 1195–203. 10. Paganetti, H., T. Bortfeld, and T.F. Delaney, Neutron dose in proton radiation therapy: in regard to Eric J. Hall (Int J Radiat Oncol Biol Phys 2006; 65:1–7). Int J Radiat Oncol Biol Phys, 2006. 66(5): 1594–1595; author reply 1595. 11. Stovall, M. et al., Fetal dose from radiotherapy with photon beams: Report of AAPM radiation therapy committee task group no. 36. [Erratum in Med Phys 1995, 22, 1353–1354]. Med. Phys., 1995. 22: 63–82. 12. Sharma, D.S. et al., Peripheral dose from uniform dynamic multileaf collimation fields: implications for sliding window intensity-modulated radiotherapy. Br J Radiol, 2006. 79(940): 331–335. 13. Sharma, S.D., R.R. Upreti, and D.D. Deshpande, Use of peripheral dose data from uniform dynamic multileaf collimation fields to estimate out-of-field organ dose in patients treated employing sliding window intensity-modulated radiotherapy. Phys Med Biol, 2006. 51(11): 2987–2995. 14. Klein, E.E. et al., Peripheral doses from pediatric IMRT. Med Phys, 2006. 33(7): 2525–3251.

584

Handbook of Anatomical Models for Radiation Dosimetry

15. Mansur, D.B., E.E. Klein, and B.P. Maserang, Measured peripheral dose in pediatric radiation therapy: A comparison of intensity-modulated and conformal techniques. Radiother Oncol, 2007. 82(2): 179–184. 16. Shepherd, S.F. et al., Whole body doses from linear accelerator-based stereotactic radiotherapy. Int J Radiat Oncol Biol Phys, 1997. 38(3): 657–665. 17. Koshy, M. et al., Extra-target doses in children receiving multileaf collimator (MLC) based intensity modulated radiation therapy (IMRT). Pediatr Blood Cancer, 2004. 42(7): 626–630. 18. Verellen, D. and F. Vanhavere, Risk assessment of radiation-induced malignancies based on whole-body equivalent dose estimates for IMRT treatment in the head and neck region. Radiother Oncol, 1999. 53: 199–203. 19. Mazonakis, M. et al., Scattered dose to thyroid from prophylactic cranial irradiation during childhood: A Monte Carlo study. Phys Med Biol, 2006. 51(8): N139–N145. 20. Kry, S.F. et al., A Monte Carlo model for out-of-field dose calculation from high-energy photon therapy. Med Phys, 2007. 34(9): 3489–3499. 21. Lin, J.P. et al., The measurement of photoneutrons in the vicinity of a Siemens Primus linear accelerator. Appl Radiat Isot, 2001. 55(3): 315–321. 22. Reft, C.S., R. Runkel-Muller, and L. Myrianthopoulos, In vivo and phantom measurements of the secondary photon and neutron doses for prostate patients undergoing 18 MV IMRT. Med Phys, 2006. 33(10): 3734–3742. 23. Kry, S.F. et al., Out-of-field photon and neutron dose equivalents from step-and-shoot intensitymodulated radiation therapy. Int J Radiat Oncol Biol Phys, 2005. 62(4): 1204–1216. 24. Vanhavere, F., D. Huyskens, and L. Struelens, Peripheral neutron and gamma doses in radiotherapy with an 18 MV linear accelerator. Radiat Prot Dosim, 2004. 110(1–4): 607–612. 25. Roy, S.C. and G.A. Sandison, Shielding for neutron scattered dose to the fetus in patients treated with 18 MV x-ray beams. Med Phys, 2000. 27(8): 1800–1803. 26. Roy, S.C. and G.A. Sandison, Scattered neutron dose equivalent to a fetus from proton therapy of the mother. Radiat Phys Chem, 2004. 71: 997–998. 27. Howell, R.M. et al., Investigation of secondary neutron dose for 18 MV dynamic MLC IMRT delivery. Med Phys, 2005. 32(3): 786–793. 28. Howell, R.M. et al., Calculation of effective dose from measurements of secondary neutron spectra and scattered photon dose from dynamic MLC IMRT for 6 MV, 15 MV, and 18 MV beam energies. Med Phys, 2006. 33(2): 360–368. 29. ICRP, Conversion Coefficients for use in Radiological Protection against External Radiation. International Commission on Radiological Protection (Pergamon Press), 1996. 74 (Annals of the ICRP Volume 26/3). 30. Kry, S.F. et al., A Monte Carlo model for calculating out-of-field dose from a Varian 6 MV beam. Med Phys, 2006. 33(11): 4405–4413. 31. Ongaro, C. et al., Analysis of photoneutron spectra produced in medical accelerators. Phys Med Biol, 2000. 45(12): L55–L561. 32. Difilippo, F. et al., Contamination dose from photoneutron processes in bodily tissues during therapeutic radiation delivery. Med Phys, 2003. 30(10): 2849–2854. 33. Chibani, O. and C.-M.C. Ma, Photonuclear dose calculations for high-energy photon beams from Siemens and Varian Linacs. Med Phys, 2003. 30: 1990–2000. 34. Zanini, A. et al., Monte Carlo simulation of the photoneutron field in linac radiotherapy treatments with different collimation systems. Phys Med Biol, 2004. 49(4): 571–582. 35. Barquero, R. et al., Monte Carlo simulation estimates of neutron doses to critical organs of a patient undergoing 18 MV x-ray LINAC-based radiotherapy. Med Phys, 2005. 32(12): 3579–3588. 36. Jiang, H. et al., Simulation of organ specific patient effective dose due to secondary neutrons in proton radiation treatment. Phys Med Biol, 2005. 50: 4337–4353. 37. Gottschalk, B., Neutron dose in scattered and scanned proton beams: In regard to Eric J. Hall (Int J Radiat Oncol Biol Phys 2006; 65:1–7). Int J Radiat Oncol Biol Phys, 2006. 66(5): 1594; author reply 1595.

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

585

38. Binns, P.J. and J.H. Hough, Secondary dose exposures during 200 MeV proton therapy. Radiat Protect Dosim, 1997. 70: 441–444. 39. Yan, X. et al., Measurement of neutron dose equivalent to proton therapy patients outside of the proton radiation field. Nucl Instrum Methods Phys Res, 2002. A 476: 429–434. 40. Mesoloras, G. et al., Neutron scattered dose equivalent to a fetus from proton radiotherapy of the mother. Med Phys, 2006. 33(7): 2479–2490. 41. Wroe, A., A. Rosenfeld, and R. Schulte, Out-of-field dose equivalents delivered by proton therapy of prostate cancer. Med Phys, 2007. 34: 3449–3456. 42. Tayama, R. et al., Measurement of neutron dose distribution for a passive scattering nozzle at the Proton Medical Research Center (PMRC). Nucl Instrum Methods Phys Res A, 2006. 564: 532–536. 43. Agosteo, S. et al., Secondary neutron and photon dose in proton therapy. Radiother Oncol, 1998. 48: 293–305. 44. Polf, J.C. and W.D. Newhauser, Calculations of neutron dose equivalent exposures from rangemodulated proton therapy beams. Phys Med Biol, 2005. 50(16): 3859–3873. 45. Schneider, U. et al., Secondary neutron dose during proton therapy using spot scanning. Int J Radiat Oncol Biol Phys, 2002. 53: 244–251. 46. Miralbell, R. et al., Potential reduction of the incidence of radiation-induced second cancers by using proton beams in the treatment of pediatric tumors. Int J Radiat Oncol Biol Phys, 2002. 54: 824–829. 47. Followill, D., P. Geis, and A. Boyer, Estimates of whole-body dose equivalent produced by beam intensity modulated conformal therapy. Int J Radiat Oncol Biol Phys, 1997. 38: 667–672. 48. NCRP, Limitation of Exposure to Ionizing Radiation (Supersedes NCRP Report No. 91). National Council on Radiation Protection and Measurements Report, Bethesda, MD, 1993. 116. 49. ICRP, Recommendations of the International Commission on Radiological Protection. International Commission on Radiological Protection (ICRP), 1991. 60 (Annals of the ICRP Volume 21/1–3). 50. Snyder, W.S. et al., Estimates of absorbed fractions for monoenergetic photon sources uniformly distributed in various organs of a heterogeneous phantom. J Nucl Med, 1969. 10(Suppl 3): 7–52. 51. Kramer, R. et al., The calculation of dose from external photon exposures using reference human phantoms and Monte Carlo methods. Part I: The male (ADAM) and female (EVA) adult mathematical phantoms. Gesellschaft fuer Strahlen- und Umweltforschung, 1982. GSFBericht-S-885. 52. Stabin, M.G. et al., Mathematical models and specific absorbed fractions of photon energy in the nonpregnant adult female and at the end of each trimester of pregnancy, 1995. ORNL/ TM-12907. 53. Cristy, M. and K.F. Eckerman, Specific absorbed fractions of energy at various ages from internal photon sources, 1987. ORNL/TM-8381/I-VII. 54. ICRP, Reference Man: Anatomical, Physiological and Metabolic Characteristics. International Commission on Radiological Protection (Pergamon Press), 1975. 23. 55. Bouchet, L.G. et al., A revised dosimetric model of the adult head and brain. J Nucl Med, 1996. 37(7): 1226–1236. 56. ICRP, Radiation Dose to Patients from Radiopharmaceuticals. International Commission on Radiological Protection (Pergamon Press), 1998. 80(Annals of the ICRP Volume 28/3). 57. ICRU, Photon, Electron, Proton and Neutron Interaction Data for Body Tissues. International Commission on Radiation Units and Measurements, Bethesda, MD, 1992. Report No. 46. 58. ICRU, Measurement of Dose Equivalents from External Photon and Electron Radiations. International Commission on Radiation Units and Measurements, Bethesda, MD, 1992. Report No. 47. 59. NCRP, Dosimetry of X-Ray and Gamma-Ray Beams for Radiation Therapy in the Energy Range 10 keV to 50 MeV. National Council on Radiation Protection and Measurements Report, Bethesda, MD, 1996. 69. 60. Stovall, M. et al., Genetic effects of radiotherapy for childhood cancer: Gonadal dose reconstruction. Int J Radiat Oncol Biol Phys, 2004. 60(2): 542–552.

586

Handbook of Anatomical Models for Radiation Dosimetry

61. Stovall, M., S.A. Smith, and M. Rosenstein, Tissue doses from radiotherapy of cancer of the uterine cervix. Med Phys, 1989. 16(5): 726–733. 62. Lim, S.M. et al., Prediction of myelotoxicity using radiation doses to marrow from body, blood and marrow sources. J Nucl Med, 1997. 38(9): 1374–1378. 63. Ron, H., F.O., Uncertainty in radiation dosimetry and their impact on dose–response analysis. National Cancer Institute. National Institute of Health. Workshop Proceedings, 1997. 99-4541. 64. Zaidi, H. and X.G. Xu, Computational anthropomorphic models of the human anatomy: The path to realistic Monte Carlo modeling in radiological sciences. Annu Rev Biomed Eng, 2007. 9: 471–500. 65. Lee, C., C. Lee, and W.E. Bolch, Age-dependent organ and effective dose coefficients for external photons: A comparison of stylized and voxel-based paediatric phantoms. Phys Med Biol, 2006. 51(18): 4663–4688. 66. Jones, D.G., A realistic anthropomorhpic phantom for calculating specific absorbed fractions of energy deposited from internal gamma emitters. Radiat Protect Dosim, 1998. 79: 411–414. 67. Petoussi-Henss, N. et al., The GSF family of voxel phantoms. Phys Med Biol, 2002. 47(1): 89–106. 68. Chao, T.C., A. Bozkurt, and X.G. Xu, Conversion coefficients based on the VIP-Man anatomical model and GS4-VLSI code for external monoenergetic photons from 10 keV to 10 MeV. Health Phys, 2001. 81: 163–183. 69. Xu, X.G., T.C. Chao, and A. Bozkurt, VIP-MAN: An image-based whole-body adult male model constructed from color photographs of the Visible Human Project for multi-particle Monte Carlo calculations. Health Phys, 2000. 78: 476–485. 70. Spitzer, V.M. and D.G. Whitlock, The visible human dataset: The anatomical platform for human simulation. Anat Rec, 1998. 253(2): 49–57. 71. Bozkurt, A., T.C. Chao, and X.G. Xu, Fluence-to-dose conversion coefficients based on the VIP-Man anatomical model and MCNPX code for monoenergetic neutrons above 20 MeV. Health Phys, 2001. 81: 184–202. 72. Bozkurt, A., T.C. Chao, and X.G. Xu, Fluence-to-dose conversion coefficients from monoenergetic neutrons below 20 MeV based on the VIP-Man anatomical model. Phys Med Biol, 2000. 45: 3059–3079. 73. Chao, T.C., A. Bozkurt, and X.G. Xu, Organ dose conversion coefficients for 0.1–10 MeV external electrons calculated for the VIP-Man anatomical model. Health Phys., 2001. 81: 203–214. 74. Chao, T.C. and X.G. Xu, Specific absorbed fractions from the image-based VIP-Man body model and EGS4-VLSI Monte Carlo code: Internal electron emitters. Phys Med Biol, 2001. 46: 901–927. 75. Chao, T.C. and X.G. Xu, S-values calculated from a tomographic head/brain model for brain imaging. Phys Med Biol, 2004. 49(21): 4971–4984. 76. Winslow, M. et al., Use of the VIP-Man model to calculate energy imparted and effective dose for x-ray examinations. Health Phys, 2004. 86(2): 174–182. 77. Winslow, M. et al., Monte Carlo simulations of patient x-ray images. Am Nucl Soc Trans, 2004. 90: 459–460. 78. Xu, X.G., Dose conversion coefficients for 0.1–10 MeV electrons calculated for the VIPMan tomographic model—Response to Zankl and Petoussi-Henss. Health Phys, 2002. 82: 255–256. 79. Xu, X.G., T.C. Chao, and A. Bozkurt, A male radiation worker model developed from transverse color images of the visible human project. Proceedings of the Fourth Visible Human Conference, Keystone, CO, 2002. 80. Lee, C. and W. Bolch, Construction of a tomographic computational model of a 9-mo-old and its Monte Carlo calculation time comparison between the MCNP4C and MCNPX codes. Health Phys, 2003. 84: S259. 81. Lee, C. et al., Whole-body voxel phantoms of paediatric patients—UF series B. Phys Med Biol, 2006. 51(18): 4649–4661.

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

587

82. Lee, C. et al., The UF series of tomographic computational phantoms of pediatric patients. Med Phys, 2005. 32(12): 3537–3548. 83. ICRP, Basic Anatomical and Physiological Data for Use in Radiological Protection: Reference Values. International Commission on Radiological Protection (Pergamon Press), 2003. 89. 84. Lee, C. et al., Hybrid computational phantoms of the 15-year male and female adolescent: Applications to CT organ dosimetry for patients of variable morphometry. Med Phys, 2008. 35: 2366–2382. 85. Lee, C. et al., Hybrid computational phantoms of the male and female newborn patient: NURBSbased whole-body models. Phys Med Biol, 2007. 52: 3309–3333. 86. Tsui, B.M.W. et al., Quantitative cardiac SPECT reconstruction with reduced image degradation due to patient anatomy. IEEE Trans Nucl Sci, 1994. 41: 2838–2844. 87. Piegl, L., On NURBS: A survey. IEEE Comput Graph Appl, 1991. 11: 55–71. 88. Garrity, J.M. et al., Development of a dynamic model for the lung lobes and airway tree in the NCAT phantom. IEEE Trans Nucl Sci, 2003. 50: 378–383. 89. Segars, W.P., Development and application of the new dynamic NURBS-based cardiac-torso (NCAT) phantom. Dissertation in Biomedical Engineering, University of North Carolina, Chapel Hill, NC, 2001. 90. Segars, W.P., D.S. Lalush, and B.M.W. Tsui, A realistic spline-based dynamic heart phantom. IEEE Trans Nucl Sci, 1999. 46: 503–506. 91. Segars, W.P. and B.M.W. Tsui, Study of the efficacy of respiratory gating in myocardial SPECT using the new 4-D NCAT phantom. IEEE Trans Nucl Sci, 2002. 49: 675–679. 92. Diallo, I. et al., Estimation of the radiation dose delivered to any point outside the target volume per patient treated with external beam radiotherapy. Radiother Oncol, 1996. 38(3): 269–271. 93. Keall, P.J. et al., Determining the incident electron fluence for Monte Carlo-based photon treatment planning using a standard measured data set. Med Phys, 2003. 30(4): 574–582. 94. Aaronson, R.F. et al., A Monte Carlo based phase space model for quality assurance of intensity modulated radiotherapy incorporating leaf specific characteristics. Med Phys, 2002. 29(12): 2952–2958. 95. Ding, G.X., Energy spectra, angular spread, fluence profiles and dose distributions of 6 and 18 MV photon beams: Results of Monte Carlo simulations for a Varian 2100EX accelerator. Phys Med Biol, 2002. 47: 1025–1046. 96. Fix, M.K. et al., Simple beam models for Monte Carlo photon beam dose calculations in radiotherapy. Med Phys, 2000. 27: 2739–2747. 97. Fix, M.K. et al., A multiple source model for 6 MV photon beam dose calculations using Monte Carlo. Phys Med Biol, 2001. 46: 1407–1427. 98. Deng, J. et al., Photon beam characterization and modelling for Monte Carlo treatment planning. Phys Med Biol, 2000. 45(2): 411–427. 99. Chetty, I., J.J. DeMarco, and T.D. Solberg, A virtual source model for Monte Carlo modeling of arbitrary intensity distributions. Med Phys, 2000. 27(1): 166–172. 100. Siebers, J.V. et al., Comparison of EGS4 and MCNP4b Monte Carlo codes for generation of photon phase space distributions for a Varian 2100C. Phys Med Biol, 1999. 44(12): 3009–3026. 101. Ma, C.M. et al., Clinical implementation of a Monte Carlo treatment planning system. Med Phys, 1999. 26(10): 2133–2143. 102. van der Zee, W. and J. Welleweerd, Calculating photon beam characteristics with Monte Carlo techniques. Med Phys, 1999. 26(9): 1883–1892. 103. Fix, M.K. et al., An efficient framework for photon Monte Carlo treatment planning. Phys Med Biol, 2007. 52(19): N425–N437. 104. Panettieri, V. et al., SBRT of lung tumours: Monte Carlo simulation with PENELOPE of dose distributions including respiratory motion and comparison with different treatment planning systems. Phys Med Biol, 2007. 52(14): 4265–4281. 105. Kawrakow, I. and B.R. Walters, Efficient photon beam dose calculations using DOSXYZnrc with BEAMnrc. Med Phys, 2006. 33(8): 3046–3056.

588

Handbook of Anatomical Models for Radiation Dosimetry

106. Aljarrah, K. et al., Determination of the initial beam parameters in Monte Carlo Linac simulation. Med Phys, 2006. 33(4): 850–858. 107. Naqvi, S.A. et al., Using a photon phase-space source for convolution/superposition dose calculations in radiation therapy. Phys Med Biol, 2005. 50(17): 4111–4124. 108. Fix, M.K., P.J. Keall, and J.V. Siebers, Photon-beam subsource sensitivity to the initial electronbeam parameters. Med Phys, 2005. 32(4): 1164–1175. 109. Fix, M.K. et al., Monte Carlo source model for photon beam radiotherapy: Photon source characteristics. Med Phys, 2004. 31(11): 3106–3121. 110. Cho, S.H. et al., Reference photon dosimetry data and reference phase space data for the 6 MV photon beam from Varian Clinac 2100 series linear accelerators. Med Phys, 2005. 32(1): 137–148. 111. Deng, J. et al., Modelling 6 MV photon beams of a stereotactic radiosurgery system for Monte Carlo treatment planning. Phys Med Biol, 2004. 49(9): 1689–1704. 112. Newhauser, W. et al., Monte Carlo simulations of a nozzle for the treatment of ocular tumours with high-energy proton beams. Phys Med Biol, 2005. 50: 5229–5249. 113. Paganetti, H., Monte Carlo method to study the proton fluence for treatment planning. Med Phys, 1998. 25: 2370–2375. 114. Paganetti, H., Monte Carlo calculations for absolute dosimetry to determine output factors for proton therapy treatments. Phys Med Biol, 2006. 51: 2801–2812. 115. Paganetti, H. et al., Accurate Monte Carlo for nozzle design, commissioning, and quality assurance in proton therapy. Med Phys, 2004. 31: 2107–2118. 116. Janssen, J.J. et al., A model to determine the initial phase space of a clinical electron beam from measured beam data. Phys Med Biol, 2001. 46(2): 269–286. 117. Zacharatou-Jarlskog, C. et al., Assessment of organ specific neutron doses in proton therapy using whole-body age-dependent voxel phantoms. Phys Med Biol, 2008. 53: 693–714. 118. Kawrakow, I., Accurate condensed history Monte Carlo simulation of electron transport: I. EGSnrc, the new EGS4 version. Med Phys, 2000. 27: 485–498. 119. Briesmeister, J.F., MCNP—A general Monte Carlo n-particle transport code, Version 4C. Los Alamos National Laboratory Technical Report, 2000. LA-13079-M. 120. Rijkee, A.G. et al., Assessment of induction of secondary tumours due to various radiotherapy modalities. Radiat Prot Dosim, 2006. 118(2): 219–226. 121. Waters, L., MCNPX User’s Manual. Los Alamos National Laboratory, LA-CP-02-408, 2002. 122. Zheng, Y. et al., Monte Carlo study of neutron dose equivalent during passive scattering proton therapy. Phys Med Biol, 2007. 52(15): 4481–4496. 123. Agostinelli, S. et al., GEANT4—A simulation toolkit. Nucl Instrum Methods Phys Res, 2003. A 506: 250–303. 124. Eckerman, K.F., M. Cristy, and J.C. Ryman, The ORNL mathematical phantom series. Informal Paper, Oak Ridge National Laboratory, Oak. Ridge, TN, 1996. 125. ANS, Specifications for the Bottle Mannequin Absorption Phantom. ANSI/HPS, 1999. N13.35. 126. Cristy, M., Mathematical phantoms representing children of various ages for use in estimates of internal dose. U.S. Nuclear Regulatory Commission, 1980. Report No. NUREG/CR-1159. 127. Rodrigues, P. et al., Application of GEANT4radiation transport toolkit to dose calculations in anthropomorphic phantoms. Appl Radiat Isot, 2004. 61: 1451–1461. 128. Xu, X.G. et al., A boundary-representation method for designing whole-body radiation dosimetry models: Pregnant females at the ends of three gestational periods—RPI-P3, -P6 and -P9. Phys Med Biol, 2007. 52: 7023–7044. 129. ICRU, Tissue Substitutes in Radiation Dosimetry and Measurement. International Commission on Radiation Units and Measurements, Bethesda, MD, 1989. Report No. 44. 130. Paganetti, H. et al., Relative biological effectiveness (RBE) values for proton beam therapy. Int J Radiat Oncol Biol Phys, 2002. 53: 407–421. 131. ICRP, Relative Biological Effectiveness (RBE), Quality Factor (Q), and Radiation Weighting Factor (wR). International Commission on Radiological Protection (Pergamon Press), 2003. 92. 132. Hall, E.J., The impact of protons on the incidence of second malignancies in radiotherapy. Technol Cancer Res Treat, 2007. 6(4 Suppl): 31–34.

Applications to Secondary Radiation Dosimetry in External Beam Radiation Therapy

589

133. Kocher, D.C., A.I. Apostoaei, and F.O. Hoffman, Radiation effectiveness factors for use in calculating probability of causation of radiogenic cancers. Health Phys, 2005. 89(1): 3–32. 134. Kellerer, A.M., Risk estimates for radiation-induced cancer—The epidemiological evidence. Radiat Environ Biophys, 2000. 39: 17–24. 135. Dennis, J.A., The relative biological effectiveness of neutron radiation and its implications for quality factor and dose limitation. Prog Nucl Ener, 1987. 20: 133–149. 136. Edwards, A.A., Neutron RBE values and their relationship to judgements in radiological protection. J Radiol Prot, 1999. 19(2): 93–105. 137. NCRP, The Relative Biological Effectiveness of Radiations of Different Quality. National Council on Radiation Protection and Measurements Report, Bethesda, MD, 1990. 104. 138. ICRU, Conversion Coefficients for Use in Radiological Protection against External Radiation. International Commission on Radiation Units and Measurements, Bethesda, MD, 1998. 57. 139. Boag, J.W., The statistical treatment of cell survival data. In: Proceedings of the Sixth L.H. Gray Conference: Cell Survival after Low Doses of Radiation; Ed. T. Alper; The Institute of Physics, Wiley, New York, 1975: pp. 40–53. 140. NCRP, Protection against Neutron Radiation. National Council on Radiation Protection and Measurements Report, Bethesda, MD, 1973. 38. 141. Zacharatou-Jarlskog, C. and H. Paganetti, Sensitivity of different dose scoring methods on organ specific neutron doses calculations in proton therapy. Phys Med Biol, 2008. 53: 4523–4532. 142. Suit, H. et al., Secondary carcinogenesis in patients treated with radiation: A review of data on radiationinduced cancers in human, non-human primate, canine and rodent subjects. Radiat Res, 2007. 167: 12–42. 143. Brenner, D.J. et al., Cancer risks attributable to low doses of ionizing radiation: assessing what we really know. Proc Natl Acad Sci USA, 2003. 100(24): 13761–1376. 144. Jiang, H. and H. Paganetti, Adaptation of GEANT4 to Monte Carlo dose calculations based on CT data. Med Phys, 2004. 31: 2811–2818. 145. Jiang, H., J. Seco, and H. Paganetti, Effects of Hounsfield number conversions on patient CT based Monte Carlo proton dose calculation. Med Phys, 2007. 34: 1439–1449. 146. Zacharatou-Jarlskog, C. and H. Paganetti, The risk of developing second cancer due to neutron dose in proton therapy as a function of field characteristics, organ, and patient age. Int J Radiat Oncol Biol Phys, 2008. 69: 228–235. 147. Ronckers, C.M. et al., Thyroid cancer in childhood cancer survivors: A detailed evaluation of radiation dose response and its modifiers. Radiat Res, 2006. 166(4): 618–628. 148. Shore, R.E., Issues and epidemiological evidence regarding radiation-induced thyroid cancer. Radiat Res, 1992. 131(1): 98–111. 149. Ron, E. et al., Thyroid cancer after exposure to external radiation: A pooled analysis of seven studies. Radiat Res, 1995. 141: 259–277.

26 Applications to Image-Guided Radiation Treatment Planning Chengyu Shi, Martin Fuss, Niko Papanikolaou, and X. George Xu

CONTENTS 26.1 Introduction ............................................................................................................... 591 26.1.1 Review of Current Clinical Strategies of 4D Treatment ........................ 591 26.1.2 Review of Current Existing 4D Physical Computational Phantoms ...................................................................................................... 597 26.1.2.1 Dynamic Thorax Computational Phantom ............................ 597 26.1.2.2 Quality Assurance System for Advanced Radiotherapy (QUASAR)™ Computational Phantom ............ 598 26.1.3 Review of Current Existing 4D Virtual Computational Phantoms ..... 598 26.1.3.1 NCAT Computational Phantom ............................................... 599 26.1.3.2 4D VIP-Man ................................................................................. 599 26.2 Methods ...................................................................................................................... 599 26.3 Discussion .................................................................................................................. 603 26.3.1 Ionizing Imaging vs. Nonionizing Imaging ........................................... 603 26.3.2 Tracking vs. Predicting .............................................................................. 603 26.3.3 Preplanning, Live Planning, and Postplanning .....................................604 26.3.4 Future Trends of IGRT ................................................................................604 26.4 Conclusion..................................................................................................................604 References .............................................................................................................................604

26.1 Introduction 26.1.1 Review of Current Clinical Strategies of 4D Treatment With the development of imaging technology, researchers can use radiation therapy to treat tumors more precisely in this image-guided radiation therapy (IGRT) era. Currently, there are at least two trends in the use of radiation therapy: it is more timely and can be used more efficiently than conventional treatment. “More timely” means the real-time imaging, or the four-dimensional (4D) imaging, make the treatment more be able to account for intrafraction motion, which is mainly caused by the patient’s respiratory motion.1 “More efficiently” means on-board imaging systems make the treatment delivery closer to planning, by accounting for interfraction motion, such as the patient’s setup uncertainty, using 591

Handbook of Anatomical Models for Radiation Dosimetry

592

an adaptive radiation therapy algorithm.2 In both trends, medical imaging is the foundational work. Without medical imaging technology development, it would be difficult to achieve the goals of those two trends. Medical imaging plays an important role in radiation therapy through the entire treatment process, especially in treatment planning and delivery. Treatment planning and delivery is a proverbial “chicken-and-egg” dilemma. Even though planning will normally happen before delivery, delivery can also affect planning by updating the patient’s current status (termed real-time planning) or his/her treatment history (termed adaptive planning). To solve this dilemma, researchers require the acquisition of more patient’s status information through medical imaging systems. They also require the processing of that imaging information quickly and accurately. Therefore, several challenges have been proposed, and need solutions, in order to achieve the twin goals of “more timely” and “more efficiently.” In theory, any kind of imaging method can be used for IGRT. Figure 26.1 illustrates a way of classifying imaging methods. Researchers have classified the imaging methods into two general groups: ionizing radiation and nonionizing radiation. For ionizing radiation, researchers can use computed tomography (CT) and x-ray imaging in real time for imageguided radiation treatment. Researchers have also classified other imaging methods, such as positron emission tomography (PET) and single photon emission computed tomography (SPECT), to assistant treatment planning. For nonionizing radiation, an ultrasound is used commonly for treatment positioning and planning, such as prostate cancer treatment. Radiofrequency (RF) marks can be used for patient alignment, such as the Calypso marker (Calypso Medical Technologies Inc, Seattle, WA). Magnetic resonance imaging (MRI) imaging-guided treatment is still under development. Some laser-guided images can assist for treatment planning to avoid collision and patient setup.3 In the future, fast laser scans may be used to assist treatment in real time.4 Other kinds of ionizing and nonionizing radiation can be also developed and used for IGRT. For the ionizing radiation imaging modality, Figure 26.2 shows a CT modality called CT on-rail.5 A CT and treatment linac sharing the same couch minimizes the positioning error caused by transferring the patient from one coordinate system to another coordinate system.

Imaging

Ionizing

Nonionizing

CT

X-ray

Ultrasound

MRI

SPECT

PET

RF

Laser

Others FIGURE 26.1 Classification of imaging methods for IGRT.

Others

Applications to Image-Guided Radiation Treatment Planning

593

Linac

CT

Couch

FIGURE 26.2 CT on-rail. (Photograph courtesy of Jiajin Fan, PhD)

Figure 26.3 shows a nonionizing imaging modality by using Calypso position system for patient’s setup. To illustrate the use of those images in 4D treatment, Figure 26.4 depicts a diagram for the IGRT process, including the treatment planning and delivery. For an external photon beam, a CT scan is essential for treatment planning. Based on the preplan CT images (MRI, PET, etc., which can assist the planning using image registration technology), a preplan can be developed and patient-specific quality assurance (QA), (predelivery) can be done with the same settings as in the preplan. The current approach of scientists is to treat the patient (live delivery) without replanning. With the development of on-board imaging systems, however, live-plan images can be generated before or during treatment. With the help of the live-plan images, a researcher may develop a live plan based on preplan

FIGURE 26.3 Calypso system for patient’s setup. (a) A flat panel sensor to acquire signal sending out from implanted seed; (b) console interface to show the shifts in lateral, longitudinal, and vertical direction. (Photograph courtesy of Chuan Wu, PhD)

Handbook of Anatomical Models for Radiation Dosimetry

594

Preplan images

Preplan

Predelivery

Live-plan images

Live-plan

Live-delivery

Postplan images

Postplan

Postdelivery

FIGURE 26.4 Diagram for IGRT process of treatment planning and delivery.

information, and may modify the preplan to do live delivery (4D delivery).6–8 Reseachers can acquire postplan images during or after treatment. Based on the postplan images, a researcher can develop a postplan and evaluate the live-delivered dose information. If it is necessary, a new preplan can be generated to account for the dose difference between the live delivery and the preplan (adaptive).2 Figure 26.4 provides general information for IGRT. Another applied example is the prostate implant, where an ultrasound can be used for preplan, live plan, and postplan image modality. After generating a preplan, radioactive seed information can be derived for ordering radiation isotopes (such as Pd-103). Live delivery is now possible (such as the FIRST System of Nucletron Corp., AX Veenendaal, The Netherlands). A researcher can use a postplan to evaluate the delivery results and take any necessary steps if the result does not look good. In Figure 26.4, we gave two examples of the applications of IGRT in the clinical. In addition, Figure 26.5 shows the comparison of a prostate implant guided by ultrasound and CT. Before the actual treatment day, the patient comes in and we study the patient volume. The volume study has at least two purposes: a qualification for implant and a seed order. By examining the ultrasound images, a physician can draw the contour of the target and determine the patient’s qualification for seed implant, based on the prostate volume size. If the patient is qualified, a preplan can be done based on the ultrasound images. From the preplan, the seed number and the needle number can be estimated. On the day of the actual implant, the doctor can scan the live ultrasound images and enact the live plan. Through the live delivery, the live plan can be updated. The delivery can also be done, however, based solely on the live-plan information. After all the seeds implanted, the researcher will get either ultrasound or CT postplan images. The doctor will develop a postplan based on those images, and will confirm the delivery information. Figure 26.5 shows the dose volume histogram (DVH) and isodose plots for the preplan, the live plan, and the postplan. Due to the image quality and the swelling of the prostate, the volume and dose information may be different among plans; however, the images play an important role in the treatment for each step. Figure 26.6 depicts another example of external photon beam treatment using HiArt tomotherapy unit. A researcher can use kVCT for treatment planning, by using a tomotherapy planning station. After the plan has been accomplished, and before each

Applications to Image-Guided Radiation Treatment Planning

595

treatment, a researcher an use an megavoltage computed tomography (MVCT) scan for the patient alignment by fusing the kilovoltage computed tomography (kVCT) and MVCT images. The researcher can also use the MVCT images for a postplan, using the adaptive function provided by TomoTherapy Inc. Figure 26.6a illustrates the preplan kVCT and Figure 26.6b depicts one of the MVCT images before the treatment. 120.00 Gy (100.0x) 180.00 Gy (150.0x) 240.00 Gy (200.0x) Volume: implant Type: cumulative

V (%)

D: 102.0 V: 100.0

100.0 80.0 60.0 40.0 20.0 0.0 102.0

PD NPD

272.7

443.5

614.2

785.0

955.7

D (Gy)

–7.5 [8]

(a) 120.00 Gy (100.0x) 180.00 Gy (150.0x) 240.00 Gy (200.0x) Volume: implant Type: cumulative

V (%) 100.0

D: 98.4 V: 100.0

80.0 60.0 40.0 20.0 0.0 98.4

PD NPD

269.8

441.3

612.8

784.2 955.7 D (Gy)

–17.5 (14)

(b) FIGURE 26.5 Preplan (a), live plan (b), and postplan (c) for prostate implant. Preplan and live plan are based on ultrasound images, and postplan is based on CT images. DVH (left) and one-slice image is shown. The image slices are not at the same position. (continued)

Handbook of Anatomical Models for Radiation Dosimetry

596

144.00 Gy (100.0x) 216.00 Gy (150.0x) 288.00 Gy (200.0x)

Volume: implant Type: cumulative

V (%) 100.0

D: 66.6 V: 100.0

80.0 60.0 40.0 20.0 0.0 66.6

NPD

282.6

498.6

714.6

930.6 1146.6 D (Gy)

–45.0 (19)

(c) FIGURE 26.5 (continued)

(a)

(b)

(c)

(d)

(e)

(f)

(g) FIGURE 26.6 (See color insert following page 524.) (a) kVCT; (b) MVCT; (c) correlated images axial view; (d) correlated images coronal view; (e) correlated images sagittal view; (f) DVH comparison; and (g) isodose comparison.

Applications to Image-Guided Radiation Treatment Planning

597

Figure 26.6c through e shows the correlated images in axial view, in coronal view, and in sagittal view. Figure 26.6f shows the adaptive plan using DVH comparison while Figure 26.6g illustrates the isodose comparison for the adaptive plan and for the original plan. If the adaptive will be performed, the researcher can generate an adaptive treatment plan for the future treatment to account for the interfraction dosage differences. For the full description of imaging and delivery technologies, Dong and Mohan9 published a book chapter and an AAPM TG 76 report.10 In this chapter, we focus on the IGRT treatment planning for the external photon beam treatment, and discuss especially the use of virtual computational phantoms for patient’s 4D treatment. Treatment planning and delivery are linked. Therefore, we propose potential treatment delivery technology linked with the planning. Virtual 4D computational phantoms will be used here for reference images to help guide treatment. 26.1.2 Review of Current Existing 4D Physical Computational Phantoms Researchers have developed several commercial and research 4D physical computational phantoms for previous 4D studies. It is useful to use a 4D physical computational phantom to evaluate the 4D planning, and to study the characteristics of the 4D treatment before the actual clinical application occurs. The following paragraph will introduce two commercial 4D physical computational phantoms. 26.1.2.1 Dynamic Thorax Computational Phantom Computerized Imaging Reference Systems (CIRS), Inc. developed a dynamic thorax computational phantom. CIRS is a world renowned company recognized for making physical computational phantoms (http://www.cirsinc.com/index.html). They make a product called dynamic computational phantom (Computational phantom 008, U.S. Patent 7151253B2) for a radiation therapy application computational phantom as shown in Figure 26.7. The system includes a tissue-equivalent thorax computational phantom, a precision motion actuator, and a controller with preset motion profiles, such as a real patient’s breathing curve shown in Figure 26.7b. By applying synchronized linear and rotational motion to a moving rod, researchers can generate 3D motion with achievable submillimeter accuracy and reproducibility CIRS made mimic tissues within 1% from 50 keV to 25 MeV for the 0.3

Amplitude (cm)

0.2 0.1 0 –0.1 –0.2 –0.3

(a)

–0.4 550 (b)

600

FIGURE 26.7 (a) Dynamic thorax model by CIRS, Inc.; (b) a real patient’s breathing curve.

650 Time (s)

700

750

Handbook of Anatomical Models for Radiation Dosimetry

598

Lung equivalent insert Platform House-made target Film cassette Icon chamber holder Phantom body Motion controller with remote (a)

(b)

FIGURE 26.8 QUASAR model and inters.

computational phantom. Different inters are available to be positioned within the lungs and detectors can be placed directly within the tumor volume. 26.1.2.2 Quality Assurance System for Advanced Radiotherapy (QUASAR) ‘ Computational Phantom Jake Van Dyk at the London Regional Cancer Centre, Ontario, Canada developed the Modus Medical Devices Inc. (http://www.modusmed.com/), another dynamic computational phantom, as shown in Figure 26.8. Scientists have claimed that the computational phantom design enables many of the nondosimetric QA tests recommended by TG 53.11 The QUASAR Body Computational phantom has the dimensions 20 cm high, by 30 cm wide and 8 cm long. Different kinds of inters are available for different QA purposes. Recently updated software can import a patient’s breathing curve data and can control the intermotion. 26.1.3 Review of Current Existing 4D Virtual Computational Phantoms Even though physical dynamic computational phantoms are available for different kinds of purposes, the physical dynamic computational phantoms have several disadvantages: 1. Once the computational phantom is made, it is difficult to change the shape. 2. The computational phantom is different from real human body, although the outline shape is similar to a person’s body shape. 3. The material is close to human body composition, however, it cannot represent a real patient. 4. The motor motion drive can import a patient’s breathing profile, but it may have a delay or a limit on the range. 5. The insert location is not easily changed. With recent developments in computer science, it is possible to simulate a patient’s motion in a virtual world. Several research groups have also developed the so-called 4D virtual computational phantoms. In this next section, we summarize two existing 4D virtual computational phantoms.

Applications to Image-Guided Radiation Treatment Planning

599

26.1.3.1 NCAT Computational Phantom Dr. William Paul Segar developed a dynamic virtual computational phantom designed for diagnosis imaging purpose as his dissertation in 2001.12 As reviewed in detail in Chapter 5, this NURBS-based Cardiac-Torso (NCAT) is defined using the NURBS control points to shape the surface geometry in a time-dependent manner to model the organ motion. The computational phantom has been used for research in cardiac imaging, such as SPECT, and to study the motion effect on SPECT. 26.1.3.2 4D VIP-Man In 2005, based on the 3D VIP-Man image set, researchers developed a dynamic computational phantom called 4D VIP-Man which is used in the Monte Carlo simulation.13 The purpose of developing the 4D VIP-Man was to study the organ motion effect on external photon beam treatment. The development of 4D VIP-Man computational phantom referred to the same algorithm as Dr. Segar’s. The image set and the application, however, are different. Furthermore, the developed computational phantom was applied to intensity modulated radiation therapy (IMRT) treatment simulation using the Monte Carlo code, EGS4.14 In this chapter, we describe the 4D VIP-Man as applied to external photon beam treatment planning and as a Monte Carlo simulation study using the phantom. We also discuss potential future applications of 4D VIP-Man in IGRT.

26.2 Methods The development of 4D VIP-Man has been discussed in Chapter 6 (Section 6.1.6). Here we focus on the steps to apply this 4D virtual computational phantom to IGRT as illustrated in the flowchart in Figure 26.9. Step 1: Since 4D CT has more (10 or 20 times) images than the 3D CT, autocontouring is needed. There are at least three solutions for contouring: manual, semiautomatic, and fully automatic. To manually contour 10 or 20 times is a big burden for clinical application. One realistic solution is the semiautomatic contour. Dosimetrists/ physicians still contour one phase of the 4D CT, but map the contoured phase to other phases. Modification may be needed, depending on the mapping effect. A fully automatic contour is not available now, except with a research approach. Currently, Pinnacle3 planning station version 8.0 has the ability to use computational phantom-based segmentation.15 Users can build their own organ library, and can use the organs as a computational phantom to segment the new image set. Users need auto adaptation, which uses an internal force (such as edge vectors) and an external force (such as feature points) to adapt the computational phantom into the new image set. Step 2: Based on the contours from each phase, a 4D dynamic virtual computational phantom (A) can be built according to the approach detailed in Chapter 6. The 4D dynamic computational phantom is a patient-specific computational phantom that can represent the fundamental features of that patient. Step 3: From the 4D CT images and the patient’s respiratory experiment, researchers may extract some mechanical properties (such as elastic properties). Currently,

Handbook of Anatomical Models for Radiation Dosimetry

600

1. Contour

2. Model

A 3. Mech. properties

4. Link

5. Control

B 6. Predict

7. Adjust

8. Image

C FIGURE 26.9 Flowchart for 4D virtual model-guided radiation therapy.

Step 4: Step 5:

Step 6: Step 7:

Step 8:

researcher practice several approaches to computational phantom lung functions. For example, Low et al.16 modeled lung motion as independent tidal volume and airflow displacement vectors with the position of the object at time. Al-Mayah et al.17 developed a finite element computational phantom to study the mechanical behavior of human lungs with the effect of contact surfaces and hyperelastic material properties. The extracted mechanical properties may be linked with the patient’s breathing pattern, which is recorded during 4D CT scan and monitored during treatment. Using the breathing pattern with mechanical properties, the developed 4D dynamic virtual computational phantom (A) can be controlled and evolved into a dynamic computational phantom with mechanical features (B). Researchers can also use the breathing pattern to predict the patient’s respiratory motion using a mathematical algorithm, such as linear regression algorithm.18 Doctors can use the prediction and real-time feedback about breathing patterns to control the B computational phantom and the extended B computational phantom into the C computational phantom, which is a computational phantom representing the patient’s current status. If the C computational phantom is accurate enough, it becomes a useful computational phantom for IGRT. For example, the C computational phantom can be used to generate images for a patient’s setup. The C computational phantom can also be used to guide radiation beam motion in real time and for dose calculations, as well as to predict the delivered dose in real time.

The flowchart in Figure 26.9 proposes a potential solution to the application of 4D dynamic computational phantom in IGRT. Since IRGT is still a very new modality, each of the above

Applications to Image-Guided Radiation Treatment Planning

601

1

Normalized volume

0.8

0.6 Phase 1 Phase 2* Phase 3* Phase 4* Phase 5* Phase 6* Phase 7* Phase 8* 4D average 4D total • Mismatched phase

0.4

0.2

0 0

1000

2000

3000

4000

5000

Absolute dose FIGURE 26.10 DVH plot for 4D simulation (Phase 1 º middle inhale, Phase 2 º late inhale, Phase 3 º peak inhale, Phase 4 º early exhale, Phase 5 º middle exhale, Phase 6 º late exhale, Phase 7 º peak exhale, Phase 8 º early inhale).

steps will require further refinement. In the following paragraphs, we summarize a preliminary study performed recently. Three scenarios were considered: 3D conformal, gating, and 4D treatment. “3D conformal” means that the treatment beam is conformal to one of the breathing phases and it has a potential problem in missing the target that is moving. “Gating” is used to represent the treatment beam that follows the 3D conformal phase throughout the whole treatment; however, since the delay and other uncontrollable reasons, it may too deliver the dose to a wrong phase. Finally, “4D treatment” is a case when the beam follows and change according to each breathing phase in sequence. We first compared the gating treatment delivery19 and 4D image-guided treatment delivery methods for a 3D conformal radiation treatment (CRT) plan. Figure 26.10 summarizes the DVHs for the target resulting from the gating treatment and the 4D image-guided treatment delivery, respectively. In the CRT plan delivered with a gating method, the radiation beam is made to conform with the lesion according to its location in Phase #1 (middle inhalation) of the respiratory cycle. Figure 26.10 establishes that a more similar spatial and temporal distribution between a specific phase and Phase #1 at which the beam is aimed would result in the better conformity in the DVH curve. Ideally, researchers only turn the beam on at the Phase #1. During the gating treatment delivery, however, the gating window may mismatch a phase and the conformity would suffer when this happens. To simulate this gating mismatching, Phases #1–8 were modeled as various mismatched gating moments and the results were averaged as the “4D-average” by summing the doses from all individual phases, divided by the total number of phases. This result was used to show the average dose distribution when the gating is added in the whole respiratory cycle in Figure 26.10. In contrast, the “4D-total” DVH curve represents the case when the photon beam was planned for Phase #1 but was delivered in an ungated way to represent the dose distribution in the

Handbook of Anatomical Models for Radiation Dosimetry

602

entire 4D free respiratory cycle. From the DVH curves for “4D-average” and “4D-total,” we observe that, although the gating technology may sometime happen at a wrong phase (i.e., the 4D-average curve), the fi nal dose result is still better than that obtained without gating at all (i.e., the 4D-total curve). Therefore, the simulation results presented in the Figure 26.9 confirms that the use of gating technology, as expected, can improve the dose distributions. Figure 26.11a shows the DVH curves of the IMRT gating delivery. In this case, the opening density matrix (ODM) was aligned to the lesion whose location is defi ned in Phase #1. In this case, the DVH for “4D-average” showed a better conformity than that of the Phase #1. This improved conformity is due to the way the final dose for the IMRT is integrated. The data in Figure 26.11 demonstrates that the lesion treated under Phase #3 and Phase #4 were obviously underdosed because of the respiratory motions, while the data in other phases exhibited a varying degree of discrepancies. We derived the “4D-average” curve by averaging these eight respiratory phases. The absolute dose of “4D-average” at 90% is 3855 cGy. Figure 26.11b shows the DVH curves for the lesion at various phases for the 4D imageguided treatment delivery. In this case, the center of the ODM always conforms to the center of the lesion in each of the eight respiratory phases. This setting simulates an ideal 4D “image-guided” treatment plan where the beam moves precisely with the lesion throughout all eight respiratory phases. The data shows an overall more uniform dose distribution for each of the eight phases. In addition, the averaged dose distribution (i.e., 4D-average) is still more uniform than that of any one single phase. The absolute dose of “4D-average” at 90% is 4060 cGy. Comparing the DVH curves in Figures 26.10 and 26.11, we can observe the different effects of organ motion on the final lesion dose distributions. Both Figures 26.10 and 26.11 showed that gating method improves the dose distributions, even though for a mismatched phase (as in the Figure 26.11a). For free-breathing treatment, if the beam can follow the breathing pattern, the dose distributions would be more accurate, according to the “4D-average” curve in the Figure 26.11b. If the beam fails to follow the breathing pattern, however, the dose distribution could be much worse as shown in the “4D total” curve in the Figure 26.10 discussed previously. This result confirms that the monitoring of patient breathing pattern is critical during a 4D treatment delivery. Relative technologies for monitoring of patient

1

0.8

0.8

0.6 0.4 0.2 0 0

(a)

Normalized volume

Normalize volume

Dose volume histogram plot for 4D simulation 1

Phase 1• Phase 2• Phase 3• Phase 4• Phase 5• Phase 6• Phase 7• Phase 8• 4D average • Mismatched phase

1000

3000 4000 2000 Absolute dose (cGy)

0.6 0.4 0.2 0 0

5000 (b)

Dose volume histogram plot for 4D simulation

Phase 1 Phase 2 Phase 3 Phase 4 Phase 5 Phase 6 Phase 7 Phase 8 4D average

1000

3000 4000 2000 Absolute dose (cGy)

FIGURE 26.11 (a) DVH results from IMRT gating simulation; (b) DVH results from IMRT 4D hybrid simulation.

5000

Applications to Image-Guided Radiation Treatment Planning

603

breathing pattern, such as synchronized moving aperture radiation therapy (SMART) method, have been reported elsewhere.20 4D radiation treatment planning and delivery is one of the most challenging tasks in modern radiation treatment. A motion-simulating computational phantom as reported in this study can be used to further investigate various issues in a controlled laboratory environment. For example, although a 4D treatment plan can conform more to the tumor than a 3D plan, this conformity relies on increased precision in timing and positioning of the radiation beams. Despite of approaches for dose delivery using dynamic multileaf collimators (DMLC),21–23 practical implementation is yet to demonstrate the efficacy of the methods. Image registration is generally based on either matching geometric image features, or on voxel similarity measures.24,25 Currently, studies on 4D image registration are mainly based on voxel similarity.26–32 Geometric image features, however, may be more useful to derive information such as DVH curves, since the organ contours are available from the treatment planning system. In this study, we used surface equations to deform the organ contours and reversed the equations to obtain integrated voxel-to-voxel dose results. A deformable registration method, although not necessary for this study, has been reported by Vedam et al.33 Tissue density changes during 4D treatment has also been discussed by Heath and Seuntjens.34 The organ density changes for different phases are important for nonrigid organs such as the lung. For nonrigid organs, the DVHs may be derived from total energy deposited instead of absorbed dose. For tumors and organs that are more rigid, on the other hand, the density can be assumed to be identical for the entire respiratory cycle.

26.3 Discussion 26.3.1 Ionizing Imaging vs. Nonionizing Imaging Current on-board imaging systems, such as cone-beam CT and megavoltage CT, are based on x-rays due to their favorable image quality and real-time response. However, there are increasing concerns about the extra dose to the patient.35 A traditional CT scan delivers a dose about 2–4 cGy per scan, depending on the anatomical site. 4D CT can give 5 or 10 times more. The dose from the MVCT can range from 2 to 4 cGy per scan, however, considering daily usage in patient setup involving multiple scans. The overall dose to the patient will be therefore about tens or hundreds of cGy. Adverse effects of such low level but uniform exposures on the patient have not been adequately studied.35 Therefore, the analysis of the risk versus benefit of good image quality, motion visualization ability and small setup uncertainty is yet to be performed. Conversely, nonionizing imaging devices, such as ultrasound or MRI, may achieve similar results without the use of extra radiation. For setup uncertainties, studies show that there is no much benefit using ionizing radiation devices.36 26.3.2 Tracking vs. Predicting Tracking the treatment of a moving target is better studied than predicting the treatment of moving targets. With a gating ability, the PTV margin can be reduced with similar treatment results. Tracking, however, has a problem with delayed responses. Conversely, predicting can address the issue but the prediction algorithms are known to bear some degree of uncertainty. Several studies have been reported about developing computational

604

Handbook of Anatomical Models for Radiation Dosimetry

phantoms for predicting treatment. For example, Low et al.37 hypothesized a function of five degrees of freedom: the position of the tissues at a user-specified reference breathing phase, tidal volume and its temporal derivative airflow (tidal volume phase space) can predict the motion of lung and lung tumor tissues; Al-Mayah et al.17 reported a finite element lung computational phantom with subcentimeter uncertainties. Those practices are useful and may be applied in clinical in the near future. 26.3.3 Preplanning, Live Planning, and Postplanning If time permits, a close loop of preplanning, live planning, and postplanning as in Figure 26.3 can be a better solution for IGRT. With preplanning, researchers can establish an initial guess or a baseline of the whole treatment. Live planning can improve the quality of preplanning. Postplanning can make up preplanning and live-planning errors. Therefore, the whole close-loop will improve the overall quality of IGRT. If a live planning is not available, the combination of preplanning and postplanning is also a good choice, which can also reduce uncertainty with preplanning only. 26.3.4 Future Trends of IGRT Based on the discussions above, we think that the ideal IGRT technology should meet the following conditions: lower or nonextra dose for imaging, prediction ability, and closeloop treatment planning. Since nothing is perfect, the final solution will depend on the application and may involve a compromise in these features.

26.4 Conclusion In this chapter, we summarized challenges and opportunities in IGRT. We then demonstrated the use of a virtual computational phantom and discussed various considerations in establishing an IGRT solution. The IMRT/IGRT technological development has come a long way but we appear to be still at an early stage of research and implementation. IGRT is a very nice idea, but the outcome will only be as good as the solution. No matter what new IGRT modalities and procedures may be developed in the future, it is prudent to carefully evaluate the benefit and risk associated with every aspect of the radiation delivery. A motion-simulating, phantom is a unique tool to investigate extremely complex spatial and temporal distributions of radiation in the patient body. But the next-generation patient phantom must be physics-based, capable of simulating multiorgan deformation and of predicting the treatment outcome as part of the close-loop IGRT planning.

References 1. Vedam, S.S. et al., Acquiring a four-dimensional computed tomography dataset using an external respiratory signal, Phys. Med. Biol., 48, 45, 2003. 2. Yan, D. et al., Adaptive radiation therapy, Phys. Med. Biol., 42, 123, 1997. 3. Xing, L., Chang, J., and Orton, C.G., Kilovoltage imaging is more suitable than megavoltage imaging for guiding radiation therapy, Med. Phys., 34, 4563, 2007.

Applications to Image-Guided Radiation Treatment Planning

605

4. Gong, X. et al., Evaluation of a volumetric display for radiation therapy treatment planning, Med. Phys., 33, 2209, 2006. 5. Ma, C. and Paskalev, K., In-room CT techniques for image-guided radiation therapy, Med. Dosim., 31, 30, 2005. 6. Wijesooriya, K. et al., Determination of maximum leaf velocity and acceleration of a dynamic multileaf collimator: Implications for 4D radiotherapy, Med. Phys., 32, 932, 2005. 7. Papiez, L., The leaf sweep algorithm for an immobile and moving target as an optimal control problem in radiotherapy delivery, Math. Comput. Model., 37, 735, 2003. 8. Webb, S., The effect on IMRT conformality of elastic tissue movement and a practical suggestion for movement compensation via the modified dynamic multileaf collimator (dMLC) technique, Phys. Med. Biol., 50, 1163, 2005. 9. Dong, L. and Mohan, R., Treatment Planning in Radiation Oncology, 2nd ed., Lippincott Williams & Wilkins, Philadelphia, 2007, p. 12. 10. Murphy, M.J. et al., The management of imaging dose during image-guided radiotherapy: Report of the AAPM Task Group 75, Med. Phys., 34, 4041, 2007. 11. Fraass, B. et al., AAPM radiation therapy committee task group 53: Quality assurance for clinical radiotherapy treatment planning, Med. Phys., 25, 1773, 1998. 12. Segars, W.P., Development and application of the new dynamic NURBS-based cardiac-torso (NCAT) phantom, PhD thesis, University of North Carolina, NC, 2001. 13. Xu, X.G. and Shi, C.Y., Preliminary development of a 4D anatomical model for Monte Carlo simulations, on CD-ROM of Monte Carlo 2005 Topical Meeting. The Monte Carlo method: versatility unbounded in a dynamic computing world, Chattanooga (TN), American Nuclear Society, LaGrange Park, IL, 2005. 14. Zhang, J.Y. et al., Development of a geometry-based respiratory-motion-simulating patient model for radiation treatment dosimetry, J. Appl. Clin. Med. Phys., 9, 1, 2008. 15. Kaus, M.R. et al., Model-based segmentation for treatment planning with Pinnacle3, white paper, Phillips Inc., 2007. 16. Low, D.A. et al., Novel breathing motion model for radiotherapy, Int. J. Radiat. Oncol. Biol. Phys., 63, 921, 2005. 17. Al-Mayah, A., Moseley, J., and Brock, K.K., Contact surface and material nonlinearity modeling of human lungs, Phys. Med. Biol., 53, 305, 2008. 18. Sharp, G.C. et al., Prediction of respiratory tumor motion for real-time image-guided radiotherapy, Phys. Med. Biol., 49, 425, 2004. 19. Shi, C.Y. and Papanikolaou, N., Tracking versus gating in the treatment of moving targets, Eur. Oncol. Dis., 5, 83, 2007. 20. Neicu, T. et al., Synchronized moving aperture radiation therapy (SMART): Improvement of breathing pattern reproducibility using respiratory coaching, Phys. Med. Biol., 51, 617, 2006. 21. Paganetti, H., Jiang, H., and Trofimov, A., 4D Monte Carlo simulation of proton beam scanning: Modeling of variations in time and space to study the interplay between scanning pattern and time-dependent patient geometry, Phys. Med. Biol., 50, 983, 2005. 22. Vedam, S. et al., Dosimetric impact of geometric errors due to respiratory motion prediction on dynamic multileaf collimator-based four-dimensional radiation delivery, Med. Phys., 32, 1607–1620, 2005. 23. Wijesooriya, K. et al., Determination of maximum leaf velocity and acceleration of a dynamic multileaf collimator: Implications for 4D radiotherapy, Med. Phys., 32, 932–941, 2005. 24. Mäkelä, T. et al., A review of cardiac image registration methods, IEEE Trans. Med. Imaging, 21, 1011, 2002. 25. Chmielewski, L. and Kozinska, D., Image registration, in Proceedings of the 3rd Polish Conference on Computer Pattern Recognition Systems, MiBków (Poland), KOSYR, 2003, p. 163. 26. Bookstein, F.L., Principal warps-thin-plate splines and the decomposition of deformations, IEEE Trans. Pattern Anal. Mach. Intell., 11, 567, 1989. 27. Christensen, G.E. et al., Image-based dose planning of intracavitary brachytherapy: Registration of serial-imaging studies using deformable anatomic templates, Int. J. Radiat. Oncol. Biol. Phys., 51, 227, 2001.

606

Handbook of Anatomical Models for Radiation Dosimetry

28. Brock, K.K. et al., Creating a four-dimensional model of the liver using finite element analysis, Med Phys., 29, 1403, 2002. 29. Hartkens, T., Measuring, analyzing, and visualizing brain deformation using non-rigid registration, PhD thesis, University of London, London, U.K., 2003. 30. Crum, W.R., Hartkens, T., and Hill, D.L., Non-rigid image registration: theory and practice, Br. J. Radiol., 77, S140, 2004. 31. Guerrero, T. et al., Intrathoracic tumour motion estimation from CT imaging using the 3D optical flow method, Phys. Med. Biol., 49, 4147, 2004. 32. Guerrero, T. et al., Elastic image mapping for 4-D dose estimation in thoracic radiotherapy, Radiat. Prot. Dosim., 115, 497, 2005. 33. Vedam, S.S. et al., Acquiring a four-dimensional computed tomography dataset using an external respiratory signal, Phys. Med. Biol., 48, 45, 2003. 34. Heath, E. and Seuntjens, J., A direct voxel tracking method for four-dimensional Monte Carlo dose calculations in deforming anatomy, Med. Phys., 33, 434, 2006. 35. Brenner, D.J. and Hall E.J., Computed tomography—An increasing source of radiation exposure, N. Engl. J. Med., 357, 2277, 2007. 36. Orton, N.P. and Jaradat, H.A., Clinical assessment of three-dimensional ultrasound prostate localization for external beam radiotherapy, Med. Phys., 33, 4710, 2006. 37. Low, D.A. et al., Novel breathing motion model for radiotherapy, Int. J. Radiat. Oncol. Biol. Phys., 63, 921, 2005.

27 Dose Calculations in Radiation Therapy Based on Patient Models Using the Geant4 Monte Carlo Code Harald Paganetti

CONTENTS 27.1 Introduction ............................................................................................................... 607 27.2 Methods ...................................................................................................................... 609 27.2.1 The Geant4 Monte Carlo Package in Radiation Therapy ....................... 609 27.2.1.1 Geant4 Philosophy ....................................................................... 609 27.2.1.2 Geant4 Physics .............................................................................. 610 27.2.1.3 Benchmarking .............................................................................. 613 27.2.2 Modeling the Treatment Head ................................................................... 614 27.2.2.1 Photon Beam Therapy ................................................................. 614 27.2.2.2 Proton Beam Therapy .................................................................. 614 27.2.3 Modeling the Patient for Static Dose Calculation (3D) ........................... 615 27.2.3.1 Voxelized Geometries .................................................................. 615 27.2.3.2 CT Conversion .............................................................................. 616 27.2.3.3 Dose Scoring ................................................................................. 618 27.2.3.4 Dose-to-Water versus Dose-to-Tissue........................................ 618 27.2.3.5 Clinical Use ................................................................................... 619 27.2.4 Modeling the Patient for Dynamic Dose Calculation (4D) .................... 620 27.2.4.1 Modeling Dynamic Geometries in Geant4 .............................. 620 27.2.4.2 Monte Carlo Dose Calculation for Moving Targets ................ 622 27.2.4.3 Deforming Anatomy.................................................................... 623 27.3 Results ......................................................................................................................... 623 27.4 Discussion and Conclusion ..................................................................................... 626 Acknowledgments ............................................................................................................... 626 References ............................................................................................................................. 627

27.1 Introduction Monte Carlo methods are considered to be the most accurate methods used to calculate the absorbed dose in the patient in radiation therapy. Many treatment-planning studies have shown that there are significant differences between the results from Monte Carlo dose 607

Handbook of Anatomical Models for Radiation Dosimetry

608

calculation and the results from conventional methods, such as pencil beam and convolution/superposition algorithms.1,2 Monte Carlo simulations are generally more accurate than analytical algorithms because they take into account the physics of particle interactions on a particle-by-particle basis. This is done using theoretical models or experimental cross-section data for electromagnetic and nuclear interactions. One has to keep in mind that the reliability of Monte Carlo transport calculations depends on the accuracy of the underlying physics data. Monte Carlo simulations consider tissue inhomogeneities by using specific material properties, e.g., elemental composition, electron density, mass density, or ionization potential. The appropriate position of inhomogeneities along the beam path and its scattering effects are modeled. Secondary particles can be tracked, which, for example, allows the appropriate consideration of nuclear fragments. For some treatment scenarios, the difference between dose distributions obtained with pencil-beam algorithms and those from Monte Carlo can be significant. Thus, Monte Carlo methods can play an increasingly important role in helping to improve the accuracy of radiation treatments for cancer patients.3 The delivery of dose to a designated target in radiation therapy has become more precise with the advent of various modalities like intensity-modulated radiation therapy (IMRT) and proton beam therapy, as well with sophisticated imaging techniques (image guided therapy). In order to take full advantage of the capability to precisely deliver the dose, accurate predictions of the dose distribution delivered to the patient are necessary. Consequently, interest in Monte Carlo dose calculation methods has increased and at the same time, computers have reached the capacity to perform such calculations in a reasonable time frame. Proton beam therapy is becoming more popular. There are various facilities already operating or under construction worldwide which will offer proton therapy to more patients in the years to come. Proton beam therapy has the physical advantage in delivering lessintegral dose to the human body than x-ray techniques for the same prescribed dose to the target. Protons, being positively charged, show different energy deposition characteristics than photons (Figure 27.1). Dose as a function of depth is deposited in a “Bragg curve,” with the maximum being deep in the tissue (depending on the proton beam energy) and negligible dose downstream of the designated target area.4 This chapter describes the implementation and use of Monte Carlo dose calculation for patients using the Geant4 Monte Carlo code.5

100

Dose (%)

80 60 40 20 0 0

50

100 150 Depth (mm)

200

FIGURE 27.1 Example depth–dose characteristics of a photon beam (sold line) and a proton beam (dashed line).

Dose Calculations in Radiation Therapy Based on Patient Models

609

27.2 Methods 27.2.1 The Geant4 Monte Carlo Package in Radiation Therapy 27.2.1.1 Geant4 Philosophy Several Monte Carlo codes such as MCNP,6 MCNPX,7 BEAM/EGS/EGSnrc,8,9 Geant4,5 PENELOPE,10 PEREGRINE,11 and DPM12 are used in radiation therapy. This is not a complete list, as many other codes exist. Furthermore, several groups have developed many flavors or developments of these codes. Monte Carlo methods have been applied to verify the results of the approximate dose calculation algorithms implemented in commercial treatment-planning tools.13,14 Due to the availability of fast computers, Monte Carlo dose calculations methods have been implemented in commercial treatment-planning systems for photon/electron dose calculations. Most of these implementations are based on the VMC Monte Carlo code.15,16 The code Geant4 is based on C++ object-oriented architecture5 and is available for UNIX, LINUX, and MS-Windows environment. Unlike Geant3, which was designed for highenergy physics applications and later extended to lower energies, Geant4 was designed explicitly to serve the medical physics community as well. The code has found its wide applications in brachytherapy,17,18 external photon/electron,19–21 and external proton22–26 dose calculations. Geant4 has been validated for use in medical physics, in particular for proton therapy applications.24,27–29 Most Monte Carlo codes consist of an executable, for which the user has to write an appropriate input fi le depending on the specific problem and geometry. On the other hand, Geant4 is not a stand-alone executable but an object-oriented toolkit of libraries. Thus, it provides a toolkit with the functionality to simulate different process organized in different functions within a C++ class structure. These functions and classes allow the specification of all aspects of a simulation geometry, the materials involved, the particles of interest, the generation of primary events, the tracking of particles through materials and electromagnetic fields, the physics processes governing particle interactions, the response of sensitive detector components, the generation of event data, the storage of events and tracks, the visualization of the detector and particle trajectories, and the analysis of simulation data at different levels of detail. The user has to write a stand-alone application to take advantage of the architecture provided by the Geant4 libraries. The user code typically deals with descriptions of geometry, materials, particles of interest, physics processes, and user actions regarding input and output of data. The user code is compiled and linked to the precompiled Geant4 class libraries to create the problem-specific stand-alone executable. In the Geant4 code, particle transport is controlled under several levels. These are “run” (dealing with the initialization, termination, and overall control of running a predefined number of histories), “event” (taking care of the simulation of one particle history), “track” (an intermediate level between “event” and “step”), and “step” (dealing with one step on the trajectory of a particle in the Monte Carlo). The corresponding object classes are “G4Run,” “G4Event,” “G4Track,” and “G4Step.” The managing classes are “G4RunManager,” “G4EventManager,” “G4TrackingManager,” and “G4SteppingManager.” Further, user action classes provide an interface to the user via “G4UserRunAction,” “G4UserEventAction,” “G4UserTrackingAction,” and “G4UserSteppingAction.” The “Run” class category basically describes a collection of events and is invoked by the “beamOn()” method of the “G4RunManager,” which starts the simulation (particle tracking).

610

Handbook of Anatomical Models for Radiation Dosimetry

The “Event” class category maintains the information obtained by the interaction of a single particle in the geometry. The “Tracking” class invokes physics processes. Obviously, an important class category is the one needed to define the geometry. The highest volume in the Geant4 geometry hierarchy is the “world volume.” A particular geometry within the “world volume” is defined in a C++ class and related to the“world volume” through a hierarchy of parent–child relationships. Each relationship between a parent and child contains the relative geometric orientation of the child relative to its parent. Each part of the geometry is defined in the “G4VUserDetectorConstruction” class. The Material class category describes the material compositions or mixture of elements, the density of the material and percent compositions. Then there are “logical volumes” and “physical volumes” used to describe different shapes and materials. The “physical volume” points to the position in the geometrical structure to place the “logical volume.” Classes dealing with the geometry allow for the proper propagation of particles by providing the boundary constraints for the tracking process. The track class category contains information of the track and steps a particle takes in the defined geometry. For example, “G4Step” stores temporary information about the step currently being simulated and includes information about the starting position (“PreStepPoint”) and ending position (“PostStepPoint”) of the step, as well as energy transfers along the way. This information is important in dose calculation, for example, because it defi ned the position where the energy lost by the particle along the step is transferred to the medium. The “G4Track” stores information on the final status of the particle at the “PostStepPoint.” Thus, some information is accessible via both, “G4Step” and “G4Track.” In order to extract user-defined information from the simulation, e.g., energy deposited in a volume, energy of a particle crossing a plane, histograms or other output files can be generated through a data interface not necessarily part of the Geant4 environment. This can be, for example, the Abstract Interface for Data Analysis (AIDA). AIDA defines abstract interfaces for common physics analysis objects such as histograms and n-tuples. The user can interact with the Geant4 environment via messengers, which are software entities used to transfer information. They also provide a way for the interaction between class categories that do not depend on each other. For debugging the geometrical setup and to visualize the geometry, parts of the geometry, or particle tracks, a visualization class category manages visualization supporting various graphical formats. 27.2.1.2 Geant4 Physics Geant4 allows the tracking of leptons, bosons, mesons, baryons, ions, and short-lived particles. Obviously, the modeling of the appropriate physics is a prerequisite for accurate results. Geant4, being used for many different applications and over a wide range of particle energies, offers many different options for defining the underlying physics. Each physics interaction is implemented with a process class. The user has to decide which interactions are of importance for the specific application because Geant4 only takes interactions into account that were explicitly registered to the process manager. The process “G4Transportation” handles the propagation of particles in the geometric space by considering boundary constraints. Processes determine the behavior of particles in a “Step” (refers to the “G4SteppingManager” class) via two methods. One is the GPIL (“GetPhysicalInteractionLength”) method, which determines the length of each step based on the underlying physics. The second is the “DoIt” method, which models the actual actions; for example, energy transfer. The instances of “DoIt” may include “AlongStepDoIt” (for continuous processes), “PostStepDoIt” (for discrete processes), and “AtRestDoIt,”

Dose Calculations in Radiation Therapy Based on Patient Models

611

depending on the nature of the physics process. The spatial position, energy, direction, spin, traveling time, etc of the particle are updated during and at the end of each step. Although not a physics model parameter per se, results from Monte Carlo simulations heavily depend on the chosen step size. The step size should be small, so that the difference of the cross sections at the beginning and the end of the step is small. On the other hand, a large step size increases the computing time. In Geant4, for continuous energy loss, the variable “dRoverRange” controls the maximum percent of the range reduction a particle can undergo during a step. Geant4 is mainly used for high-energy physics applications and therefore spans a wide range of particles and energy domains. It offers many options for registering physics settings for specific interactions. Thus, it may not be obvious which physics setting is appropriate for a specific application. Some models or data sets may be more accurate for high energies, others for lower energies. The choice may also be particle dependent. The physics of particle interactions (i.e., the probability density function) can be well described based on experimentally determined cross sections. However, such cross sections may not be available for all particles of interest and over the entire energy region of interest. In these cases models, parameterizations or a combination of models, parameterizations, and experimental data have to be used. The specific choice may depend not only on the particle or energy region, but also on the required accuracy (versus efficiency) of a particular application. Geant4 is, for the better and worse, flexible. The user has to provide the appropriate physics in a subroutine typically called the “Physics List” for each application. It contains all registered processes. The Monte Carlo implementation of electromagnetic interactions is straightforward with only a few different options. The implementation of nuclear interactions is more complex. Nuclear interactions have to describe a variety of exit channels. Thus, nuclear interaction processes may require the combination of several models. A “Physics List” has been constructed as recommendation for proton therapy applications.30 It is constructed in a modular way and has four modules each addressing the following types of interactions: (a) electromagnetic, (b) elastic scattering, (c) inelastic scattering of protons and neutrons, and (d) inelastic scattering of heavier hadrons. For electromagnetic hadronic energy loss, Geant4 has two different models, called the “standard” model and the “low-energy parameterized” model. The standard model is an analytical model describing interactions of photons and all-charged particles down to 1 keV. The energy loss of hadrons is calculated by the Bethe–Bloch equation down to 2 MeV. Below 2 MeV a parameterization based on stopping power formalism of the ICRU31 is used. The accuracy in the energy region above 2 MeV is presumably in the order of 1%. The parameterizations used at lower energies are believed to be accurate in the order of about 5%.31 The low-energy extension to the model can be used for electrons and photons down to 250 eV using data libraries from Lawrence Livermore National Laboratory. In contrast to the standard model, it is mainly parameterization driven. As in the standard model, hadron ionizations are described by the Bethe–Bloch and ICRU formalism above 2 MeV and below 2 MeV, respectively. Below 1 keV, the free electron gas model is applied. GEANT4 does offer alternative energy loss and stopping power tables to the one published by the ICRU,31 e.g., the ones by Andersen and Ziegler32 and Ziegler.33 Medin and Andreo34 compared the stopping power ratios for different tables and found discrepancies of up to 4.5%. The ionization process is simulated as a discrete production of γ-rays and, at the same time, as a continuous energy loss of the particle. Three different energy straggling models are implemented (a Gaussian distribution with Bohr’s variance for large step lengths, a Vavilov distribution for intermediate step lengths, and a Landau distribution for

612

Handbook of Anatomical Models for Radiation Dosimetry

small step lengths). Geant4 calculates material-dependent ionization potentials internally but it also allows user-defined settings, for example, based on the ICRU. A question connected to the definition of the correct physics is which particles have to be tracked for a specific application. Particles that are not tracked should deposit their energy locally. In proton therapy dose calculation protons, electrons, and photons need to be considered. However, for most protons in the patient (<150 MeV) the largest possible energy transferred to an electron is about 300 keV (the energy of most electrons is much less than 300 keV), which corresponds to a range of 1 mm in water, unless there is an air cavity, which happens frequently in particular in head and neck treatments. The tracking of secondary electrons is certainly required for microscopic simulations, e.g., to study radiobiological properties, for microdosimetry, or for absolute dose simulations.29 To describe multiple scattering, Geant4 uses a condensed history algorithm that utilizes functions to calculate the angular and spatial distributions of the scattered particle implementations (“G4UrbanMscModel”).35 The implementation differs slightly from Molière theory.36,37 The dose deposited via nuclear interaction products, i.e., secondary protons, can be significant in proton therapy.38,39 Neutrons, deuterons, tritons, helium ions, and heavier recoil nuclei do not have to be tracked (the energy can be deposited locally), except for studies on scattered doses.40 Geant4 treats nuclear interactions by using theory-driven, parameterization-driven, and data-driven models. The latter are considered where experimental data are available with sufficient coverage. The Geant4 implementation of nuclear interactions offers a variety of options for the user to choose from. There are no modular packages, but individual models or parameterizations for each interaction group can be combined to a complete physics setup. An interaction between the projectile and the nucleus is modeled as an intranuclear cascade with the probability of secondary particle emission. For the energy region of interest in medical physics applications, three models to simulate the intranuclear cascade can be used: the binary cascade model, the Bertini cascade model, and the low-energy parameterized model. The explicit cascade models are more theoretically motivated as compared to the low-energy parameterized model. The latter has been designed for fast simulations and conserves energy and momentum only on average and not on an event-by-event basis. Further, the Bertini model is a simplification compared to the binary cascade model. Different model settings in Geant4 have been described in more detail and also compared for proton therapy simulations.30 Once the energy of the particles in a cascade has reached a lower limit, a preequilibrium model is applied. Geant4 offers a precompound model (“G4PreCompoundModel”), invoked by the binary cascade model and a preequilibrium model, invoked by the Bertini cascade model. The precompound model of Geant4 applies to incoming particle energies below 100 MeV. The nucleus is viewed as a collection of exciton states. When the nucleus has reached statistical equilibrium the simulation invokes the equilibrium models. The equilibrium models of Geant4 describe the emission of photons, nucleons, and light fragments from this residual state.30 For proton therapy dose calculations, only the secondary protons need to be tracked, unless one is interested in studying the effect of secondary neutrons.41 For simulating elastic interactions, the low-energy parameterized model is the only complete implementation in Geant4. The model offers two user-defined options, “G4LElastic” and “G4HadronElastic.” In addition, there are two implementations of the underlying cross sections. The combination of the two elastic cross sections was implemented as a new elastic process, “G4UHadronElasticProcess” which was combined with “G4HadronElastic” model into the “UHElastic” module.30

Dose Calculations in Radiation Therapy Based on Patient Models

613

27.2.1.3 Benchmarking In a clinical setting a dose calculation algorithm should be able to predict the dose in a patient geometry with an accuracy of about ±2.5%. Because Monte Carlo is a statistical method there has to be a trade-off between accuracy of the simulated results and calculation speed. Before a Monte Carlo system can be used for interpreting clinical data, it has to be benchmarked specifically for the type of radiation and the energy region for which it is intended. It is generally accepted that analytical models result in dose distributions that are an approximation of the true dose distribution, and that the true dose distribution can only be obtained from Monte Carlo simulations. However, one has to keep in mind that Monte Carlo codes are usually based on nuclear and atomic physics models or on interpolations of experimental cross sections. Both may have significant uncertainties. There is an extensive amount of literature on benchmarking Monte Carlo methods and codes for radiation therapy applications. Benchmarking studies are typically done using heterogeneous models consisting of various materials.27,42,43 To name just two examples for benchmarking studies, Sheikh-Bagheri et al.44 have benchmarked photon beam simulations with the BEAM code.9 In order to assess possible differences down to the percentage level, the Monte Carlo simulation of the linac included the electron exit window, target, flattening filter, monitor chambers, collimators, as well as the polymethylmethacrylate (PMMA) walls of the water phantom used for the measurements. Further Heath et al.45 have benchmarked the PEREGRINE code11 for clinical use in IMRT treatment planning. Dose profiles calculated in homogeneous and heterogeneous phantoms using this system were compared to both measurements and simulations using the EGSnrc8 Monte Carlo code. Most benchmarking studies refer to photon doses, thus electromagnetic interactions. For proton therapy, similar studies can be and have been done by comparing measured and simulated dose distributions.24,46 For proton beams, the comparison of dose distributions is a valuable benchmark for electromagnetic interactions. However, nuclear interactions play only a small role when it comes to interpreting dose distributions for treatment planning in proton beam therapy. Nevertheless, for percentage accuracy in Monte Carlo dose calculations, e.g., if Monte Carlo is used for absolute, and not only relative dosimetry, their correct modeling is important.29 There are other areas in proton therapy where a precise modeling of nuclear interactions is vital. These are, for example, Monte Carlo simulations of scattered neutron doses to assess potential risks for patients40,41 or Monte Carlo simulations to support PET imaging in proton beam therapy.47–49 The International Atomic Energy Agency (IAEA) has recognized the need for standardizing the use and modeling of nuclear interactions in proton and heavy ion radiation therapy.50 Some benchmarking experiments are not only tailored to a specific radiation but also to a specific interaction. An experimental tool particularly useful for testing proton nuclear interaction data is the multilayer Faraday Cup.28,51 It is a device sensitive to electromagnetic and nuclear inelastic reactions that measures the longitudinal charge distribution of primary and secondary particles. Two versions of the detector exist. One consists of 65 polyethylene (CH2) absorbers of dimensions 15 cm × 15 cm × 0.317 cm interspaced with 66 brass charge collectors (64 of which are active) arranged orthogonally to the main direction of the beam. Another one was built out of 66 copper sheets separated by insulating material. As a proton beam propagates in the detector, the developed charge in the plateau region depends on nuclear interactions, whereas the charge distribution near the end of range of the beam is due to protons that underwent solely electromagnetic processes. Measurements were done with a 160 MeV beam and

614

Handbook of Anatomical Models for Radiation Dosimetry

the charge distributions were simulated using Geant4.28,51 Both measured and simulated distributions count charge as a function of number of protons in the beam (which, in the experiment, was determined by the beam-on time and the current of the beam), which allows the result to be in absolute units without normalization. Various Geant4 models for the simulation of electromagnetic and nuclear interactions were validated against the measured charge distribution from the Faraday cup.30 As a result, the standard electromagnetic model was found to be more suitable than the low-energy parameterized. Further, the binary cascade model was most successful in describing the experimental data for nuclear interactions. 27.2.2 Modeling the Treatment Head 27.2.2.1 Photon Beam Therapy Because photon beams are primarily used in radiation oncology, most Monte Carlo simulations deal with modeling photon beam delivery. A comprehensive review of Monte Carlo simulations for clinical photon/electron beam is given elsewhere.52 Geant4 has been introduced later into the radiation therapy community than other codes. Furthermore, it has been shown to be less time efficient than the standard codes for photon or electron calculations, for example, EGSnrc.8 Nevertheless, Geant4 has been used and validated for use in photon and electron beam therapy.27 Because the code was not designed for radiation therapy applications, problems have been reported with the electron transport algorithm that have been taken care of in subsequent releases.53 Nevertheless, the application of Geant4 to photon therapy is less frequent than, for example, BEAM/ EGS/EGSnrc8,9 and DPM.12 27.2.2.2 Proton Beam Therapy Initially, the characteristics of the beam at the entrance to the simulated geometry have to be known. Typically, this would be at the entrance of the treatment head in the treatment room. The beam parameters are the beam energy and energy spread, the angular spread, and the geometrical beam spot. Note that, although the beam energy is usually a known parameter, the beam is typically characterized by its range in water, a clinically more meaningful term. The range, defined as the 90% or 80% dose level of the distal fall-off of the Bragg curve, can be measured accurately. Using the nominal beam energy it was shown that Monte Carlo simulations can predict the range of the proton beam with accuracy better than 1 mm compared with measurements.24 The energy spread of the beam may not be known to a high precision but can be determined by analyzing the measured width of the Bragg peak and the ratio of the peak dose and the entrance dose. The distribution can be assumed a Gaussian distribution, although there is usually a low energy tail. The shape of the beam spot is typically close to a Gaussian distribution. It can be measured easily by segmented transmission ionization chambers and is typically in the order of a few mm at treatment head entrance. The beam’s angular spread at the nozzle entrance is a parameter that cannot be determined easily. It can be simulated considering the beam optics. The Monte Carlo geometry has to model all beam-influencing devices in the treatment head. A treatment head consists of many different objects of different shape. Geant4 offers a variety of standard shapes (tubes, boxes, cones, etc.). However, some of the objects in a treatment head may be impossible to model easily out of regular, predefined shapes.

Dose Calculations in Radiation Therapy Based on Patient Models

615

Geant4 offers a variety of solutions in these cases, for example, modeling a contoured device (contoured scatterer in proton therapy) by combining stacks of cones resulting in the so-called polycones. There are other irregular shaped objects; for example, the patientspecific aperture and compensator whose contour depends on the contour of the tumor volume. The geometry of the patient-specific hardware can be imported into the Monte Carlo by taking advantage of the format used by a milling machine. For example, the aperture opening can be approximated by a 2D triangulation reducing the problem to the simulation of polyhedrons with a triangular base.24 In proton therapy, the treatment head configuration is different for each treatment field and the position of some devices is even being modified during beam delivery. Proton beam therapy comes in two flavors, using either passive scattering or beam scanning.54 For beam scanning, the beam at the treatment head entrance varies, and the magnetic field settings of the scanning magnets are modified during beam delivery. In passive scattered proton beam therapy, the beam at treatment head entrance, the settings of the doublescattering system and the position of a rotating modulator are variables. Thus, the Monte Carlo does not only need the information about the actual treatment head setting at the start of the simulation, but also needs information about the changes of this configuration that have to be applied during simulation. The geometrical setup within the Monte Carlo thus has to have parameters to vary the treatment head geometry that will be imported via an input file. For example, during a simulation the rotational position of the wheel has to be changed. The geometry variation in the treatment head can be considered by performing multiple Monte Carlo simulations. An alternative is to do a four-dimensional (4D) Monte Carlo simulation.24,55 Taking advantage of the object-oriented design of Geant4, one can, for example, literally rotate the absorber in the treatment head during a simulation. The geometry is defined in a concrete C++ class and the rotation vector of the physical volume is updated during a run break. A run is defined as the simulation of a predefined number of histories. In this technique of 4D Monte Carlo simulation,55 the geometry is changed without requiring recompilation of the source code. 27.2.3 Modeling the Patient for Static Dose Calculation (3D) 27.2.3.1 Voxelized Geometries The patient geometry is obtained for treatment-planning purposes by computer tomography (CT) scans (magnetic resonance or positron emission tomography imaging may be used in addition). The data are typically available as a digital imaging and communications in medicine56 (DICOM) stream. The grid under which CT data are stored can be regular or with a variable, nonequidistant, slice spacing. During CT scanning, regions of greater interest are often scanned with a smaller slice thickness. The Monte Carlo can either use this irregular grid,57 or a resampling to a regular, more course grid. Monte Carlo dose calculation is typically used to recalculate dose distributions that have been the result of a commercial planning system. This is simply because Monte Carlo techniques are typically not fast enough to perform a full-scale plan optimization. An interface between the Monte Carlo system and the planning system is required, which may make it convenient to use a CT format specific to the planning system instead of DICOM.57 To use DICOM directly, different versions of a DICOM interface have been developed for Geant4.58,59 The first implementation of a DICOM interface was previously included as one of the extended examples of the Geant4 package. The implementation did not handle

616

Handbook of Anatomical Models for Radiation Dosimetry

all the manufacturers’ DICOM data stream formats. In an improved solution by Kimura et al.59 Geant4 was connected to a DICOM data stream compliant with many manufacturers’ DICOM implementations. It not only allows the import of DICOM data into Geant4 but also the output of the simulation results in DICOM format. Furthermore, they incorporated dose reporting under the DICOM-RT specification and also developed a visualization tool (“GRAPE”) with capabilities of volume rendering and fusion of patient image and calculated dose distributions.60 The original DICOM CT grid is resampled to a user-defined larger grid size by using trilinear interpolation of Hounsfield units. The main concern when using CT geometry with millions of voxels is memory consumption. If a specific volume is used repeatedly, e.g., a voxel, Geant4 supports the use of a “parameterized volume.” Such a volume contains multiple copies of a volume that can even be different in solid type, size, position in space, and material composition. Each subvolume is given an exclusive index number as identifier (in “G4VPVParameterisation:: ComputeTransformation”). An advantage of parameterized volumes is the high efficiency in memory usage. For each CT voxel, only 2 bytes are used for storing a Hounsfield number. This is because at any given time during tracking, there exist only one volume object (the one where the tracked particle is positioned). This volume is used to represent all the daughter volumes by dynamically changing its properties of volume type, dimensions, location, and material. Even though there are relatively straightforward solutions to incorporate voxelized geometries into Geant4, particle tracking in a voxelized geometry is quite slow in the standard Geant4 package. In order to use Geant4 dose calculations on a patient geometry in a clinical environment, part of the Geant4 source code was recoded and optimized.57 Some of the modifications are relatively obvious and may not even be Geant4 specific. For example, the data information stored virtually during a Geant4 run may not be of interest for simulations in radiation therapy, e.g., traveling time and particle spin, and can thus be avoided. Some other changes are more complex and Geant4 specific. The main changes to the Geant4 source code deal with a quick voxel search algorithm, fast volume optimization, and dynamic assignment of material density applicable to photons, electrons and heavy charged particles. For example, in Geant4, for tracking a particle in a “parameterized volume,” the particle’s next (target) voxel has to be identified relative to particle’s current voxel. This search, performed by a “nearest neighbor algorithm” turned out to be very inefficient in Geant4, because more of the maximum possible six neighboring voxels were considered in the search. The algorithm in the member function “LevelLocate” of the “G4ParametrisedNavigation” class was therefore replaced. Further, the “G4VPVParameterisation” class was extended.57 27.2.3.2 CT Conversion A CT image shows a distribution of Hounsfield numbers, representing the radiographic densities in the imaged volume. Analytical dose calculation algorithms typically calculate dose by using the Hounsfiled unit information, converted to either electron density61 or mass density62 to characterize energy loss in tissues. Using model materials, CT scanners can be calibrated to create a relationship between Hounsfield numbers and electron density. Such a calibration method is valid for photon or electron beams because the dominant energy loss process is interaction with electrons (Compton scattering for x-rays,61 and ionization for charged particles.62) There can be significant dosimetric effects when materials are misassigned.63 Protons lose energy by ionizations, multiple Coulomb scattering and nonelastic nuclear reactions. Thus, for proton therapy, a conversion from Housnfield

Dose Calculations in Radiation Therapy Based on Patient Models

617

numbers to relative stopping power has to be used to accurately define water equivalent tissue properties. Relative stopping power depends on electron density and a Z-dependent kinetic term. Monte Carlo dose calculation does not rely solely on density information. This is because a Monte Carlo system can calculate dose-to-tissue, as compared to dose-to-water in an analytical algorithm. Thus, in addition to density information, a conversion from Hounsfield numbers to material compositions of tissues is required for each voxel. The accuracy of Monte Carlo dose calculations is affected by the ability to precisely define tissues.64 There is not a one-to-one correspondence between Hounsfield numbers and tissue materials. A Hounsfield number reflects the attenuation coefficient of human tissues to diagnostic x-rays, and may be identical for a variety of combinations of elemental compositions, elemental weights, and mass densities. Models made out of tissue materials or animal tissues have been used to determine the correspondence between Hounsfield numbers and human tissues.65 A frequently used conversion method covering tissue materials is based on the work by Schneider et al.,66 which is based on a stoichiometric calibration of Hounsfield numbers with mass density and elemental weights. A conversion table can be extended to higher Hounsfield units in order to deal with high-Z materials in the patient (e.g., Titanium implants).48 Based on such conversion algorithms, materials/tissues have been grouped into one group for air, one group for lung tissue, seven groups for soft tissues, fifteen groups for skeletal tissues, and three groups for high-Z materials.48,57,64 Jiang et al.64 studied the effects of three Hounsfield number conversion methods on Monte Carlo dose simulations of proton beams. One method was the one by Schneider et al.66 who determined Hounsfield numbers for 71 human tissues. The Hounsfield numbers were grouped into 24 bins, each with a distinct elemental composition and weight. The second method was the one implemented in the BEAMnrc/DOSXYZnrc code.67,68 Here, only four major types of tissues are defined: air, lung tissue, soft tissue, and bone. The method is very similar to others used in previous studies.2,69–71 The third method considered each and materials as water by applying a conversion from Hounsfield numbers to relative stopping powers. The difference between the method by Schneider et al. and the simple Water equivalent methods was as much as 8% in the range of skeletal tissues, and up to 4.7% for soft tissues. Thus, the method chosen to assign tissue materials can affect the absorbed dose distribution significantly. In geometries with considerable tissue inhomogeneities, e.g., in the head and neck region, treatment-planning decisions could be affected by these differences. The impact of the chosen Hounsfield unit conversion method is more significant in organs at risk compared to the target volume because of large dose gradients in the beam penumbra or at the distal end of the high-dose region when using proton beams. One advantage of proton beam therapy is the sharp distal dose fall-off. Mass density assignments affect the proton beam range in particular for deep proton fields because the beam has to penetrate several centimeters of tissue. Thus, different conversion methods could cause discrepancies in range predictions of a few millimeters. It was shown that regions where more adipose is in the proton path are more significantly affected, compared to regions with less or no adipose in the proton path.64 The optimum number of tissue groups may not have to be as high as 24 (as in the method by Schneider et al.). Nevertheless, more than 24 entities need to be considered when the material mass densities are assigned. In Geant4 a material is defined by providing density, elemental composition, and relative weights or atomic fractions. The definition of a density is linked to a material definition. Thus, even though there might be 24 materials, one would have to define more in order to take into account density variations. This is inefficient in

618

Handbook of Anatomical Models for Radiation Dosimetry

terms of computer memory or simulation run-time because of frequent changes in cross sections and other associate tables. Depending on the particles to be tracked and the registered physics, Geant4 will create tables of energy loss, mean free path, range, etc. It is computation-intensive and time-consuming to build these tables. An alternative strategy was developed to dynamically assign mass densities during particle transport so that one group (defined as elemental composition and their weights) can share one material even though the mass density continuously increases as the Hounsfield number increases.57 The difficulty in Geant4 lies in the consideration of the appropriate physics once a density change occurs without a change in material. The problem was solved by having one material defined for each Hounsfield number group with a density corresponding to the Hounsfield number in the middle of that group. Physics tables would then be computed in the standard way. The ratio of the actual density of the voxel material to the density of the reference material is defined (a parameter added to the Geant4 “G4Material class”) to allow density-related adjustments of the particle physics characteristics.57 27.2.3.3 Dose Scoring The storing of information from particle interactions, for example, the energy deposited, occurs in the “G4UserSteppingAction” class. The history of particle transportation and the copy number in a “parameterized volume” is accessible via the “G4TouchableHistory” class. The most important parameter when calculating dose is the energy deposited in a voxel. A Monte Carlo system typically provides information about the status of a tracked particle either at the beginning or at the end of an individual step. In Geant4 each step consists of a “prestep point” and a “poststep point.” Whether dose is deposited prestep or poststep has no consequence if the particle does not stop at a boundary. Geant4 defines the poststep point as within the volume to be entered within the “G4Transportation” class. Thus, one has to make sure that energy is deposited prestep to avoid large dosimetric effects at boundary crossings, for example, between soft tissue and bone. 27.2.3.4 Dose-to-Water versus Dose-to-Tissue Historically, dose distributions in radiation therapy are reported as dose-to-water. Thus, clinicians are used to looking at and making their decisions based upon dose distributions that show dose-to-water. The reason is that analytical dose calculation methods typically do not calculate dose-to-tissue because they model the human body as a structure composed of voxels of water with varying mass density. Monte Carlo algorithms, however, do calculate dose-to-tissue, which draws a more accurate picture of the deposited energy in tissue. For soft tissues the difference between the two metrices may be only between 1% and 2%. However, for higher density materials, such as cortical bone, the difference can be as large as 15% even for photon beams.72 Similar effects can be expected in proton beams.73 The effect can be, in part, corrected for by applying unrestricted waterto-medium mass collision stopping power ratios. Converting dose-to-tissue into doseto-water, or vice versa, is not necessarily straightforward for proton beams due to the involvement of not only ionization, but also multiple Coulomb scattering and nuclear interactions with which protons can deposit energy. With the clinical implementation of more Monte Carlo dose calculation routines, the future may bring a paradigm shift from the use of dose-to-water to dose-to-tissue. Right now there is an open debate about the pros and cons of using these two metrices.74 Arguments in favor of using dose-to-water are the clinical experience being based on dose-to-water,

Dose Calculations in Radiation Therapy Based on Patient Models

619

the fact that quality assurance and absolute dose measurements are done in water and the fact that tumor cells in the human body consist mostly of water. 27.2.3.5 Clinical Use The sections above have described how the beam information can be simulated and how the patient geometry can be implemented and handled within a Monte Carlo. Clinical implementation requires more steps that go beyond this. Geant4-based Monte Carlo dose calculation is currently in use at the Francis H. Burr Proton Therapy Center at Massachusetts General Hospital. The following paragraphs described its implementation.46 Treatment planning is done using a commercial planning system with a dose calculation based on a pencil-beam algorithm. For a specific treatment plan, this system provides the number of fields and for each field, the gantry angle, the patient couch angle, the air gap between patient and treatment head, the dose, the range of the beam (e.g., the distal 90% dose level of the spread-out Bragg peak [SOBP]), and the modulation with of the SOBP (e.g., distance between the proximal and distal 90% dose level in a SOBP). When treating a patient, this information is transferred to the hardware control software of the treatment facility. The Geant4-based system uses the treatment information in a similar way as the control software, i.e., it imports all necessary parameters from the planning system. To facilitate data flow between the planning system and the Monte Carlo, a user interface was developed. The user may select a specific patient, plan, and treatment field. The CT information and planning information is then imported from the planning system and input files for the Monte Carlo phase space and dose calculation are created. The Monte Carlo treatment head geometry is updated according to the prescription. The input file for simulating a phase space distribution contains information about incident protons at the nozzle entrance (mean energy, energy spread, angular spread, and beam spot size), information about the database location of the milling machine file for the aperture geometry, the database location of the milling machine file for the compensator, and the information about beam range and modulation width. The latter is then translated into the appropriate beam and treatment head configuration. Milling machine files are imported into the Monte Carlo to create virtual devices. Once the treatment head is set up, the Monte Carlo calculates a phase space distribution. A phase space typically records the spatial, energy, and angular information of the particles exiting the treatment head at a predefined scoring plane (typically downstream of the range compensator, which is the last device the beam passes through prior to reaching the patient). Other information can be scored depending on the scope of the study. The second step of the simulation, starting with the precomputed phase space distribution for each field, is the dose calculation in the patient. Separating dose calculation into simulations of phase space and dose calculation is common practice when using Monte Carlo in radiation therapy.22,75 Typically, this is done in order to reuse phase space distributions for different patients and thus save computation time. In proton therapy however, this is impractical because treatment head settings are patient field specific. However, the separation of the phase space calculation from the dose calculation is done because the treatment head may overlap with the patient CT cube (depending on the size of the air gap). Overlapping geometries can cause ambivalent situations in a Monte Carlo transport environment and should be avoided. Thus, the phase space concept allows the particles that are “stored” in the phase space file to be started within the CT geometry.

620

Handbook of Anatomical Models for Radiation Dosimetry

For the patient dose calculation the input file has the information about the database location of the CT data, the gantry angle, the patient couch angle, the isocenter position of the CT in the coordinate system of the planning program, the number of voxels and slice dimensions in the CT coordinate system, the size of the air gap between treatment head and patient, and the prescribed dose. The different coordinate systems (treatment control system, treatment head, planning system, and CT system) have to be converted into a common simulation coordinate system. Also, couch and gantry rotations have to be applied with respect to different axes in a predefined order. In clinical use, the patient geometry (CT data) is transferred from the departmental patient database to the planning system. From there, the CT is imported into the Monte Carlo using a file format unique to the planning system, preserving the (variable) slice spacing and the voxel grid size. The planning system does not calculate dose on the CT grid, but rather on a course-planning grid, which is typically in the order of 2 × 2 × 2.5 mm3. In order to avoid uncertainties due to interpolation between different tissues, we decided to calculate dose in the Monte Carlo based on the actual CT grid, which can be as small as 0.5 × 0.5 × 1.25 mm3. The dose is stored in a file having the same format as the imported CT data. In order to compare the Monte Carlo results with the ones from the planning system, the Monte Carlo dose can be resampled to the planning grid. The resulting file, including the appropriate format including the file header, can potentially be imported into the planning system for analysis or comparison with the pencil-beam results. In dose calculation for radiation therapy, one typically aims at 2.5% statistical accuracy in the target volume. The required number of particle histories depends on the efficiency of the treatment head, the required geometrical resolution, and on beam range and modulation width. A typical number of proton histories for a given patient field is ~25 million at nozzle entrance if sufficient statistical accuracy is requested for a single field and if the dose is reported on the treatment-planning grid resolution. If an analysis is done for the entire plan, and it is not field specific, fewer protons per field may be sufficient. To ensure that the statistical accuracy for each field is sufficient before starting the simulation, a look-up table of required histories as a function of beam and treatment head parameters can be used. Note that a certain statistical accuracy in the target volume does not guarantee the same accuracy in the organs at risk. However, the impact of statistical deficiencies is less for dose– volume analysis in organs at risk because the dose distribution is less homogeneous.76,77 Whether a dose calculation engine is capable of reporting dose relative to a reference dose or as absolute dose delivered per monitor unit depends on several factors. A monitor unit is a defined charge collected in a transmission ionization chamber in the treatment head. Thus, for a specific treatment head configuration (i.e., field parameters) a specific charge corresponds to a specific dose at a point in the target. In order to predict absolute doses, a planning system has to have a measure of the treatment head geometry. Otherwise, the machine output has to be either calculated separately29,78 or measured prior to treatment.79 If the ionization chamber reading is simulated as part of the geometry, the Monte Carlo code can predict absolute doses without any empirical normalization factors. Thus, the Monte Carlo generated dose distributions can be specified absolute, in cGy per monitor unit.29 27.2.4 Modeling the Patient for Dynamic Dose Calculation (4D) 27.2.4.1 Modeling Dynamic Geometries in Geant4 Some radiation therapy techniques involve electro mechanical motions in the treatment head during beam delivery. This is the case for leaf motion in intensity-modulated photon therapy (IMRT), modulator wheel rotation in conventional proton therapy80 and

Dose Calculations in Radiation Therapy Based on Patient Models

621

variable magnet settings in proton beam scanning.81 In order to consider these geometry changes, the results of several independent simulations are usually combined. In IMRT separate dose calculations can be done for individual leaf segments which are then integrated.82 Further, stochastic techniques to simulate dynamic multileaf motion for IMRT were suggested which randomize the leaf position during the simulation to mimic the planned photon fluence map.83,84 Another method modifies the beam phase space based on the probability of beam attenuation by an MLC.85 Such an approach does not take the time sequence of events into account, which may be desirable for certain studies. Difficulties also arise when high time resolution is required, e.g., when investigating IMRT delivery using a sliding window technique. In passive scattered proton therapy, the beam has to penetrate through absorbers of variable thickness, i.e., a modulator wheel, resulting in a variation of the beam energy. This creates a set of individual Bragg curves that are combined to a SOBP. In order to simulate this motion, many different geometrical incidences have to be considered, because the wheel is spinning continuously, thus at any given time the beam spot may overlap with more than one absorber step. Geant4 allows the geometry to be changed in the idle period between two consecutive runs (defined as the simulation of a given number of histories). The result is a time-dependent geometry, or a “4D Monte Carlo simulation.”23,55,86 During the simulation the change of the geometry model is performed according to parameters imported from the input file via a “Messenger” class method. The “G4UImessenger” class delivers the user command to the destination class object. “G4UIcommand” derived classes can use “SetNewValue()” methods to set geometrical parameters. The geometry change takes place via member functions of the “G4RunManger” class. The command “Defi neWorldVolume” sets the pointer of the world physical volume. The “GeometryHasBeenModified” command is used to notify the run manager that the geometry model has been changed (the variable “geometryNeedsToBeClosed” is set to be true, which permits the execution of “G4GeometryManager::BuildOptimisations()” to perform reoptimization of the geometry). Next, “ResetNavigator” resets the navigator for tracking and to implement the geometry optimization. The “G4Navigator” class of Geant4 is used to locate points in the geometry and compute distances to geometry boundaries. Finally, the state of geometry has to be set close, which denies any request for geometry change before the next run is fi nished. If the geometry is changed frequently one has to make sure that the previous geometry is discarded from memory. Thus, the new geometry should not be reconstructed via a “Construct()” method, but should only be updated. Typically, one is interested in changing only part of the simulated geometry (e.g., a leaf in a multileaf collimator). Thus, it is not necessary to rearrange the entire geometry within Geant4. However, if the run manger is informed that the geometry needs to be reoptimized, no information is processed which part of the geometry was changed, leading to recreation of the whole geometry. To reduce the time spent on optimization during geometry change, the method in Geant4 can be modified so that the optimization process is only imposed on those volumes that contain a modified daughter volume.57 In this method, the run manager is no longer informed about a geometry change. Instead, the geometry change is processed at the close state of geometry. When considering motion in 4D Monte Carlo, the resolution of the simulation has usually little influence on the speed of the calculation, since motion is handled quickly by simply changing pointers within memory. Since each simulation step can be as small as considering only one particle history, it can be virtually continuous.

622

Handbook of Anatomical Models for Radiation Dosimetry

27.2.4.2 Monte Carlo Dose Calculation for Moving Targets Some of the geometry changes described above handle changes of the treatment head geometry. Geometry changes can also occur, however, in the patient. In particular for tumors in the lung and liver, one has to consider organ motion during treatments. Motion typically takes place along the cranio-caudal and anterior–posterior axes during respiration.87,88 The phase and the amplitude is patient dependent, determined by tumor position in the lung, patient breathing pattern, lung capacity, etc. The motion period is typically around 4 s. Intrathoracic tumor movement with respiratory and cardiac motion has been analyzed by Ross et al.89–93 Tumor displacement due to breathing can be as large as 2 cm. The deformation of the lung changes the volume by about 25% during respiration. Organ motion during irradiation can have a significant impact on tumor coverage and cause dose perturbations.89,94–97 Thus, treatment planning or beam delivery have to take into account the time dependence of the patient geometry, in particular for highly conformal techniques (IMRT or proton therapy) because of the sharp dose fall-offs. In the treatment of moving tumors, different strategies may be applied to mitigate the effects of motion. These include the irradiation of the internal target volume,92 the respiratory gating98 and the tumor tracking.99,100 4D planning based on pencil-beam algorithms has been discussed in several publications.101,102 Research is being done to understand the implications of different motion mitigation strategies. Analytical dose calculation can incorporate organ motion effects via convolution methods.94,103–106 Although the assumption of a rigid motion may provide useful information, the inclusion of organ deformation is necessary for a detailed analysis of the dosimetric impact of motion in particular for proton therapy. Because proton beams stop in the patient, motion of low or high density materials in and out of the beam can result in significant shifts in the range, i.e., the sharp distal dose fall-off. The patient-specific information about organ motion can be obtained from four-dimensional computer tomography (4DCT). It creates several patient geometry snapshots from the patient’s breathing cycle, an information that can be incorporated into the treatment planning process.107,108 One problem with 4DCT data are motion artifacts due to irregular breathing and inaccuracies in sorting the images (obtained in scanning cine-mode) into different breathing phases.109,110 Dose calculation for motion analysis consists of individual dose calculations performed on each CT data set obtained via 4DCT. Typically between 5 and 10 phases out of the breathing cycle are used. The analysis is done by mapping the dose distributions from each phase to a reference phase. Various dose remapping methods have been proposed.23,111–115 Monte Carlo emerges as the best solution to study motion effects in lung dosimetry, given its accuracy, especially for low-density materials and in presence of heterogeneities.116 An extension of the 4D Monte Carlo technique described above for the treatment head has been proposed for use in organ motion studies.23,117 It avoids summing or statistically combining multiple 3D calculations from multiple runs and allows the study of 4D effects in any arbitrary time scale and includes the temporal sequence of events. The temporal sequence of events becomes important when interplay effects between two independently moving systems, e.g., treatment head motion and organ motion, are to be considered.86,118 4D Monte Carlo dose calculations to study variations in patient anatomy due to organ motion have been done to analyze dosimetric effects of motion.23,117 In 4D Monte Carlo, the time parameter is translated into the number of histories because the geometry is updated based on a certain number of histories in between two incidents in time.

Dose Calculations in Radiation Therapy Based on Patient Models

623

The advantage of 4D Monte Carlo compared to considering multiple 3D dose calculations obviously depends on the required temporal resolution, and whether interplay effects are of interest. If interplay effects can be neglected, treatment head and patient calculations can be separated. Further, because the patient geometry as obtained from 4DCT may be known only for a few incidents in time during the breathing phase, multiple 3D calculations may be as efficient as 4D. In addition to 4D dose calculation studies on patient data, computational patient models may play a role in providing insights into action thresholds. Sophisticated computational models incorporating motion have been developed.119–121 Such geometries can be implemented into a 4D Monte Carlo dose calculation environment.122 Segars et al.119,120,123 developed a 4D NURBS-based Cardiac-Torso (NCAT) computational phantom to simulate respiratory motion.121 The computational phantom has a spline-based description of the human anatomy. 27.2.4.3 Deforming Anatomy In this case of a rigid motion, the movement of each volume is well defined, and dose can be accumulated during the Monte Carlo simulation for each subvolume. In order to simulate dose deposition over the respiratory cycle in a patient anatomy, the position of each voxel as a function of time has to be known. Thus, dose cannot be computed as a function of position in a fixed coordinate system but has to be accumulated as a function of each anatomical voxel. A parameterized volume in Geant4, with its voxel identifier, is wellsuited for this task.23 Nevertheless, the tracking of voxel positions over time is nontrivial because of anatomical deformations. To correlate different respiratory phases, the so-called voxel displacement maps are required, which can be obtained using deformable image registration algorithms.124 Image registration methods can be based on, for example, BSpline125 and demons algorithms.126 4D Monte Carlo dose calculation using different implementations of voxel displacement information has been done for protons using Geant4,23 and for photons using other codes.112,118,127 Each of these simulations is done on a regular voxel grid with the dose being registered based on a deformable image registration map. An alternative method of 4D Monte Carlo dose calculation was introduced where each individual voxel is allowed to deform.128

27.3 Results This section shows an example of a dose calculation for a patient treated with proton beam therapy.46 First, a phase space distribution at a plane at the exit of the treatment head was simulated. Figure 27.2 shows the plane and the particle (proton) tracks as simulated. The effect of the double-scattering system, on the beam, starting from a small pencil beam to a homogenous field at the treatment head exit is illustrated. The figure also clearly shows the limitations of a passive scattered proton beam facility, in that a large amount of protons are “lost” in the treatment head reducing the efficiency of the delivery system. The phase space generated at the exit of the treatment head is used as input into the patient dose calculation. The origin of the phase space particles in the patient dose calculation may

624

Handbook of Anatomical Models for Radiation Dosimetry

FIGURE 27.2 With Geant4 simulated treatment head for passive scattered proton beam therapy at the Francis H. Burr Proton Therapy Center at Massachusetts General Hospital.24 Shown are 1 track (upper), 100 tracks (middle), and 1000 tracks (lower) as they are transported through the treatment head. The phase space scoring plane at treatment head exit can be seen as well.

be inside of the CT data cube because of the clinically used air gap (typically in the order of only a few cm). Figure 27.3 shows some dose distributions for a patient with a paraspinal tumor. The Monte Carlo dose calculation was done on a CT gird with 176 × 147 × 126 slices with voxel dimensions of 0.932 × 0.932 × (2.5 or 3.75) mm3. This was a relatively simple case in terms of the tumor geometry. Thus, the treatment could be done with only three coplanar fields. Each field has its specific gantry angle, couch angle, beam range, field modulation width, aperture, and compensator. Such simulations can be used to benchmark analytical dose calculation methods or to study the effect of different treatment head settings on the dose distribution in the patient.

Dose Calculations in Radiation Therapy Based on Patient Models

625

1 Gy 3 Gy 5 Gy 7 Gy 9 Gy 11 Gy 13 Gy 15 Gy 17 Gy

FIGURE 27.3 (See color insert following page 524.) Axial, coronal, and sagittal views of dose distributions calculated using the Geant4 Monte Carlo system at Massachusetts General Hospital.46 The patient was treated with three fields (columns 1–3) for a paraspinal tumor. The treatment plan was done with a commercial planning system.

Another research aspect that can be studied using Monte Carlo dose calculation is the importance of nuclear interactions when it comes to proton beam energy loss in tissue. This has implications for the analytical modeling of dose delivery and for understanding areas with increased biological effectiveness due to high-LET components.38 It has been shown that the percentage of dose deposited via nuclear interactions decreases with penetration depth in water,38 which can be confirmed in the patient (Figure 27.4). The maximum is about 4.5% in dose at about 50% of the beam range in this particular case. Thus, although the consideration of nuclear interactions is important in Monte Carlo dose calculation to accurately model the dose in the entrance region and the fluence reduction due to inelastic interactions, the contribution of dose due to nuclear secondaries seems to be negligible at the distal end and thus has negligible impact on the beam range.

626

Handbook of Anatomical Models for Radiation Dosimetry

10 Gy 20 Gy 30 Gy 35 Gy 40 Gy 42 Gy 44 Gy 46 Gy 48 Gy

1% 2% 3% 4% 5% 6%

FIGURE 27.4 Same patient as in Figure 27.3. The upper figure shows the total dose (sum of the three fields shown in Figure 27.3). The two lower figures show the contribution of dose from secondary protons (protons generated in nuclear interactions) for all three fields combined (left) and for just one of the fields (right) given as percent of the prescribed target dose (for all three fields and for one field, respectively).

27.4 Discussion and Conclusion This chapter illustrated the use of the Geant4 Monte Carlo code for dose calculations in a clinical environment. Some of the solutions are specific to Geant4. However, the overall philosophy and the different tasks to be solved on the route to a clinical implementation of a Monte Carlo dose calculation engine are code independent.

Acknowledgments The author would like to thank the following individuals for many fruitful discussions regarding the subject: Christina Zacharatou Jarlskog and Hongyu Jiang (Massachusetts General Hospital), and Joseph Perl (SLAC, Stanford University). Furthermore, the author would like to acknowledge the support of NIH grants RO1 CA1077640-01 “Four Dimensional Monte Carlo Dose Calculation” and PO1 CA21239-26 “Proton Radiation Therapy Research.”

Dose Calculations in Radiation Therapy Based on Patient Models

627

References 1. Knoos, T. et al., Comparison of dose calculation algorithms for treatment planning in external photon beam therapy for clinical situations. Phys Med Biol, 2006. 51(22): 5785–5807. 2. Ma, C.M. et al., Clinical implementation of a Monte Carlo treatment planning system. Med Phys, 1999. 26(10): 2133–2143. 3. Chetty, I.J. et al., Report of the AAPM Task Group No. 105: Issues associated with clinical implementation of Monte Carlo-based photon and electron external beam treatment planning. Med Phys, 2007. 34(12): 4818–53. 4. Paganetti, H. and T. Bortfeld, Proton therapy. In: New Technologies in Radiation Oncology; Eds. Schlegel, W.; Bortfeld, T.; Grosu, A.L.; Springer, ISBN 3-540-00321-5, 2005. 5. Agostinelli, S. et al., GEANT4—A simulation toolkit. Nucl Instrum Methods Phys Res, 2003. A506: 250–303. 6. Briesmeister, J.F., MCNP—A general Monte Carlo n-particle transport code, Version 4C. Los Alamos National Laboratory Technical Report, 2000. LA-13079-M. 7. Waters, L., MCNPX User’s Manual. Los Alamos National Laboratory, Los Alamos, NM, 2002. 8. Kawrakow, I., Accurate condensed history Monte Carlo simulation of electron transport: I. EGSnrc, the new EGS4 version. Med Phys, 2000. 27: 485–498. 9. Rogers, D.W.O. et al., BEAM: A Monte Carlo code to simulate radiotherapy treatment units. Med Phys, 1995. 22: 503–524. 10. Salvat, F. et al., PENELOPE, A code system for Monte Carlo simulation of electron and photon transport. Proceedings of a Workshop/Training Course, OECD/NEA, November 5–7, 2001, NEA/ NSC/DOC(2001)19, ISBN: 92-64-18475-9, 2001. 11. Hartmann Siantar, C.L. et al., Description and dosimetric verification of the PEREGRINE Monte Carlo dose calculation system for photon beams incident on a water phantom. Med Phys, 2001. 28(7): 1322–1337. 12. Sempau, J., S.J. Wilderman, and A.F. Bielajew, DPM, a fast, accurate Monte Carlo code optimized for photon and electron radiotherapy treatment planning dose calculations. Phys Med Biol, 2000. 45: 2263–2291. 13. Carlsson, A.K., P. Andreo, and A. Brahme, Monte Carlo and analytical calculation of proton pencil beams for computerized treatment plan optimization. Phys Med Biol, 1997. 42: 1033–1053. 14. Ma, C.-M. et al., Monte Carlo verification of IMRT dose distributions from a commercial treatment planning optimization system. Phys Med Biol, 2000. 45: 2483–2495. 15. Fippel, M., Fast Monte Carlo dose calculation for photon beams based on the VMC electron algorithm. Med Phys, 1999. 26(8): 1466–75. 16. Kawrakov, I. and M. Fippel, Investigation of variance reduction techniques for Monte Carlo photon dose calculation using XVMC. Phys Med Biol, 2000. 45: 2163–2183. 17. Perez-Calatayud, J., D. Granero, and F. Ballester, Phantom size in brachytherapy source dosimetric studies. Med Phys, 2004. 31: 2075–2081. 18. Torres, J. et al., Dosimetry characterization of 32P intravascular brachytherapy source wires using Monte Carlo codes PENELOPE and GEANT4. Med Phys, 2004. 31: 296–304. 19. Fix, M.K. et al., Simple beam models for Monte Carlo photon beam dose calculations in radiotherapy. Med Phys, 2000. 27: 2739–2747. 20. Fix, M.K. et al., Monte Carlo simulation of a dynamic MLC based on a multiple source model. Phys Med Biol, 2001. 46: 3241–3257. 21. Fix, M.K. et al., A multiple source model for 6 MV photon beam dose calculations using Monte Carlo. Phys Med Biol, 2001. 46: 1407–1427. 22. Paganetti, H., Monte Carlo method to study the proton fluence for treatment planning. Med Phys, 1998. 25: 2370–2375. 23. Paganetti, H. et al., Monte Carlo simulations with time-dependent geometries to investigate organ motion with high temporal resolution. Int J Radiat Oncol, Biol, Phys, 2004. 60: 942–950.

628

Handbook of Anatomical Models for Radiation Dosimetry

24. Paganetti, H. et al., Accurate Monte Carlo for nozzle design, commissioning, and quality assurance in proton therapy. Med Phys, 2004. 31: 2107–2118. 25. Paganetti, H. and T. Schmitz, The influence of the beam modulation method on dose and RBE in proton radiation therapy. Phys Med Biol, 1996. 41: 1649–1663. 26. Szymanowski, H. and U. Oelfke, Two-dimensional pencil beam scaling: An improved proton dose algorithm for heterogeneous media. Phys Med Biol, 2002. 47: 3313–3330. 27. Carrier, J.F. et al., Validation of GEANT4, an object-oriented Monte Carlo toolkit, for simulations in medical physics. Med Phys, 2004. 31: 484–492. 28. Paganetti, H. and B. Gottschalk, Test of Geant3 and Geant4 nuclear models for 160 MeV protons stopping in CH2. Med Phys, 2003. 30: 1926–1931. 29. Paganetti, H., Monte Carlo calculations for absolute dosimetry to determine output factors for proton therapy treatments. Phys Med Biol, 2006. 51: 2801–2812. 30. Zacharatou-Jarlskog, C. and H. Paganetti, Physics settings for using the Geant4 toolkit in proton therapy. IEEE Trans Nucl Sci, 2008. 55: 1018–1025. 31. ICRU, Stopping Powers and Ranges for Protons and Alpha Particles. International Commission on Radiation Units and Measurements, Bethesda, MD, 1993. Report No. 49. 32. Andersen, H.H. and J.F. Ziegler, The Stopping and Ranges of Ions in Matter, Vol. 3. Pergamon Press, Oxford, 1977. 33. Ziegler, J.F., The Stopping and Ranges of Ions in Solids. Pergamon Press, Oxford, 1985. 34. Medin, J. and P. Andreo, Stopping powers for the ion-chamber dosimetry of radiotherapeutic heavy-particle beams. Nucl Instrum Methods Phys Res, 1992. B69: 64–75. 35. Urban, L., Multiple scattering model in Geant4. CERN report, 2002. CERN-OPEN-2002-070. 36. Andreo, P., J. Medin, and A.F. Bielajew, Constraints of the multiple-scattering theory of Moliere in Monte Carlo simulations of the transport of charged particles. Med Phys, 1993. 20: 1315–1325. 37. Gottschalk, B. et al., Multiple Coulomb scattering of 160 MeV protons. Nucl Instrum Methods Phys Res, 1993. B74: 467–490. 38. Paganetti, H., Nuclear interactions in proton therapy: Dose and relative biological effect distributions originating from primary and secondary particles. Phys Med Biol, 2002. 47: 747–764. 39. Medin, J. and P. Andreo, Monte Carlo calculated stopping-power ratios, water/air, for clinical proton dosimetry (50–250 MeV). Phys Med Biol, 1997. 42: 89–105. 40. Zacharatou-Jarlskog, C. and H. Paganetti, Sensitivity of different dose scoring methods on organ specific neutron doses calculations in proton therapy. Phys Med Biol, 2008. 53: 4523–4532. 41. Zacharatou-Jarlskog, C. et al., Assessment of organ specific neutron doses in proton therapy using whole-body age-dependent voxel phantoms. Phys Med Biol, 2008. 53: 693–714. 42. Doucet, R. et al., Comparison of measured and Monte Carlo calculated dose distributions in inhomogeneous phantoms in clinical electron beams. Phys Med Biol, 2003. 48(15): 2339–54. 43. Wang, L., M. Lovelock, and C.-S. Chui, Experimental verification of a CT-based Monte Carlo dose-calculation method in heterogeneous phantoms. Med Phys, 1999. 26: 2626–2634. 44. Sheikh-Bagheri, D. et al., Comparison of measured and Monte Carlo calculated dose distributions from the NRC linac. Med Phys, 2000. 27(10): 2256–66. 45. Heath, E., J. Seuntjens, and D. Sheikh-Bagheri, Dosimetric evaluation of the clinical implementation of the first commercial IMRT Monte Carlo treatment planning system at 6 MV. Med Phys, 2004. 31(10): 2771–9. 46. Paganetti, H. et al., Clinical implementation of proton Monte Carlo dose calculation in proton beam therapy. Phys Med Biol, 2008 (submitted). 47. Parodi, K. et al., Clinical CT-based calculations of dose and positron emitter distributions in proton therapy using the FLUKA Monte Carlo code. Phys Med Biol, 2007. 52(12): 3369–87. 48. Parodi, K. et al., PET/CT imaging for treatment verification after proton therapy: A study with plastic phantoms and metallic implants. Med Phys, 2007. 34(2): 419–35. 49. Parodi, K. et al., Patient study of in vivo verification of beam delivery and range, using positron emission tomography and computed tomography imaging after proton therapy. Int J Radiat Oncol Biol Phys, 2007. 68(3): 920–34.

Dose Calculations in Radiation Therapy Based on Patient Models

629

50. Paganetti, H. et al., The IAEA initiative to standardize nuclear data for heavy charged-particle radiotherapy (abstract). Med Phys, 2007. 34: 2495. 51. Gottschalk, B., R. Platais, and H. Paganetti, Nuclear interactions of 160 MeV protons stopping in copper: A test of Monte Carlo nuclear models. Med Phys, 1999. 26: 2597–2601. 52. Verhaegen, F. and J. Seuntjens, Monte Carlo modelling of external radiotherapy photon beams. Phys Med Biol, 2003. 48(21): R107–64. 53. Poon, E., J. Seuntjens, and F. Verhaegen, Consistency test of the electron transport algorithm in the GEANT4 Monte Carlo code. Phys Med Biol, 2005. 50: 681–694. 54. Blattmann, H., Beam delivery systems for charged particles. Radiat Environ Biophys, 1992. 31: 219–231. 55. Paganetti, H., Four-dimensional Monte Carlo simulation of time dependent geometries. Phys Med Biol, 2004. 49: N75–N81. 56. ACR-NEMA, Digital Imaging and Communications. ACR-NEMA Standards Publication, National Electrical Manufacturer’s Association, Washington, DC, 1985. 300. 57. Jiang, H. and H. Paganetti, Adaptation of GEANT4 to Monte Carlo dose calculations based on CT data. Med Phys, 2004. 31: 2811–2818. 58. Kimura, A. et al., DICOM interface and visualization tool for Geant4-based dose calculation. IEEE Nucl Sci Symp Conf Rec, 2005. 2: 981–984. 59. Kimura, A. et al., DICOM data handling for Geant4-based medical physics application. IEEE Nucl Sci Symp Conf Rec, 2004. 4: 2124–2127. 60. Kimura, A. et al., GRAPE: A visualization tool for Geant4-based medical physics applications. IEEE Nucl Sci Symp Conf Rec, 2006. 2: 793–795. 61. Seco, J. and P.M. Evans, Assessing the effect of electron density in photon dose calculations. 2005: 1–20. 62. Ahnesjoe, A. and M.M. Aspradakis, Dose calculations for external photon beams in radiotherapy. Phys Med Biol, 1999. 44: R99–R155. 63. Verhaegen, F. and S. Devic, Sensitivity study for CT image use in Monte Carlo treatment planning. Phys Med Biol, 2005. 50(5): 937–46. 64. Jiang, H., J. Seco, and H. Paganetti, Effects of Hounsfield number conversions on patient CT based Monte Carlo proton dose calculation. Med Phys, 2007. 34: 1439–1449. 65. Schaffner, B. and E. Pedroni, The precision of proton range calculations in proton radiotherapy treatment planning: Experimental verification of the relation between CT-HU and proton stopping power. Phys Med Biol, 1998. 43: 1579–1592. 66. Schneider, W., T. Bortfeld, and W. Schlegel, Correlation between CT numbers and tissue parameters needed for Monte Carlo simulations of clinical dose distributions. Phys Med Biol, 2000. 45: 459–478. 67. Rogers, D.W.O. et al., BEAMnrc User Manual. NRCC Report PIRS-0509, 2002. 68. Walters, B.R.B. and D.W.O. Rogers, DOSXYZnrc User Manual. NRCC Report, 2002. PIRS-794. 69. DeMarco, J.J., T.D. Solberg, and J.B. Smathers, A CT-based Monte Carlo simulation tool for dosimetry planning and analysis. Med Phys, 1998. 25(1): 1–11. 70. du Plessis, F.C.P. et al., The indirect use of CT numbers to establish material properties needed for Monte Carlo calculation of dose distributions in patients. Med Phys, 1998. 25: 1195–1201. 71. Zamenhof, R. et al., Monte Carlo-based treatment planning for boron neutron capture therapy using custom designed models automatically generated from CT data. Int J Radiat Oncol Biol Phys, 1996. 35(2): 383–97. 72. Siebers, J.V. et al., Converting absorbed dose to medium to absorbed dose to water for Monte Carlo based photon beam dose calculations. Phys Med Biol, 2000. 45: 983–995. 73. Palmans, H. and F. Verhaegen, Assigning nonelastic nuclear interaction cross sections to Hounsfield units for Monte Carlo treatment planning of proton beams. Phys Med Biol, 2005. 50: 991–1000. 74. Liu, H.H. and P. Keall, Dm rather than Dw should be used in Monte Carlo treatment planning. Med Phys, 2002. 29: 922–924. 75. Siebers, J.V. et al., Comparison of EGS4 and MCNP4b Monte Carlo codes for generation of photon phase space distributions for a Varian 2100C. Phys Med Biol, 1999. 44(12): 3009–26.

630

Handbook of Anatomical Models for Radiation Dosimetry

76. Jiang, S.B., T. Pawlicki, and C.-M. Ma, Removing the effect of statistical uncertainty on dosevolume histograms from Monte Carlo dose calculations. Phys Med Biol, 2000. 45: 2151–2162. 77. Keall, P.J. et al., The effect of dose calculation uncertainty on the evaluation of radiotherapy plans. Med Phys, 2000. 27(3): 478–84. 78. Kooy, H. et al., Monitor unit calculations for range-modulated spread-out Bragg peak fields. Phys Med Biol, 2003. 48: 2797–2808. 79. Kooy, H.M. et al., The prediction of output factors for spread-out proton Bragg peak fields in clinical practice. Phys Med Biol, 2005. 50: 5847–5856. 80. Koehler, A.M., R.J. Schneider, and J.M. Sisterson, Range modulators for protons and heavy ions. Nucl Instrum Methods, 1975. 131: 437–440. 81. Lomax, A., Intensity modulation methods for proton radiotherapy. Phys Med Biol, 1999. 44: 185–205. 82. Verhaegen, F. and I.J. Das, Monte Carlo modelling of a virtual wedge. Phys Med Biol, 1999. 44: N251–259. 83. Liu, H.H., F. Verhaegen, and L. Dong, A method of simulating dynamic multileaf collimators using Monte Carlo techniques for intensity-modulated radiation therapy. Phys Med Biol, 2001. 46: 2283–2298. 84. Verhaegen, F. and H.H. Liu, Incorporating dynamic collimator motion in Monte Carlo simulations: An application in modelling a dynamic wedge. Phys Med Biol, 2001. 46: 287–296. 85. Keall, P.J. et al., Monte Carlo dose calculations for dynamic IMRT treatments. Phys Med Biol, 2001. 46: 929–941. 86. Paganetti, H., H. Jiang, and A. Trofimov, 4D Monte Carlo simulation of proton beam scanning: Modeling of variations in time and space to study the interplay between scanning pattern and time-dependent patient geometry. Phys Med Biol, 2005. 50: 983–990. 87. Booth, J.T. and S.F. Zavgorodni, Set-up error & organ motion uncertainty: A review. Aust Phys Eng Sci Med, 1999. 22: 29–47. 88. Langen, K.M. and D.T.L. Jones, Organ motion and its management. Int J Radiat Oncol Biol Phys, 2001. 50: 265–278. 89. Ross, C.S. et al., Analysis of movement of intrathoracic neoplasms using ultrafast computerized tomography. Int J Radiat Oncol Biol Phys, 1990. 18: 671–677. 90. van Soernsen de Koste, J.R. et al., Tumor location cannot predict the mobility of lung tumors: A 3D analysis of data generated from multiple CT scans. Int J Radiat Oncol Biol Phys, 2003. 56: 348–354. 91. Hanley, J. et al., Deep inspiration breath-hold technique for lung tumors: The potential value of target immobilization and reduced lung density in dose escalation. Int J Radiat Oncol Biol Phys, 1999. 45: 603–611. 92. Liu, H.H. et al., Assessing respiration-induced tumor motion and internal target volume using four-dimensional computed tomography for radiotherapy of lung cancer. Int J Radiat Oncol Biol Phys, 2007. 68(2): 531–40. 93. Rosu, M. et al., The impact of breathing motion versus heterogeneity effects in lung cancer treatment planning. Med Phys, 2007. 34(4): 1462–73. 94. Bel, A., M. van Herk, and J.V. Lebesque, Target margins for random geometrical treatment uncertainties in conformal radiotherapy. Med Phys, 1996. 23: 1537–1545. 95. Killoran, J.H. et al., A numerical simulation of organ motion and daily setup uncertainties: Implications for radiation therapy. Int J Radiat Oncol Biol Phys, 1997. 37: 213–221. 96. Rudat, V. et al., Combined error of patient positioning variability and prostate motion uncertainty in 3D conformal radiotherapy of localized prostate cancer. Int J Radiat Oncol Biol Phys, 1996. 35: 1027–1034. 97. Shirato, H. et al., Four-dimensional treatment planning and fluoroscopic real-time tumor tracking radiotherapy for moving tumor. Int J Radiat Oncol Biol Phys, 2000. 48: 435–442. 98. Keall, P. et al., The clinical implementation of respiratory-gated intensity-modulated radiotherapy. Med Dosim, 2006. 31(2): 152–62. 99. Neicu, T. et al., Synchronized moving aperture radiation therapy (SMART): Average tumour trajectory for lung patients. Phys Med Biol, 2003. 48: 587–598.

Dose Calculations in Radiation Therapy Based on Patient Models

631

100. Shirato, H. et al., Organ motion in image-guided radiotherapy: Lessons from real-time tumortracking radiotherapy. Int J Clin Oncol, 2007. 12(1): 8–16. 101. Rietzel, E. et al., Four-dimensional image based treatment planning: Target volume segmentation and dose calculation in the presence of respiratory motion. Int J Radiat Oncol Biol Phys, 2005. 61: 1535–1550. 102. Keall, P.J. et al., Four-dimensional radiotherapy planning for DMLC-based respiratory motion tracking. Med Phys, 2005. 32(4): 942–51. 103. Chetty, I.J. et al., A fluence convolution method to account for respiratory motion in three-dimensional dose calculations of the liver: A Monte Carlo study. Med Phys, 2003. 30: 1776–1780. 104. Li, J.G. and L. Xing, Inverse planning incorporating organ motion. Med Phys, 2000. 27: 1573–1578. 105. Lujan, A.E. et al., A method for incorporating organ motion due to breathing into 3D dose calculations. Med Phys, 1999. 26: 715–720. 106. Unkelbach, J. and U. Oelfke, Inclusion of organ movements in IMRT treatment planning via inverse planning based on probability distributions. Phys Med Biol, 2004. 49: 4005–4030. 107. Rietzel, E. et al., Design of 4D treatment planning target volumes. Int J Radiat Oncol Biol Phys, 2006. 66: 287–295. 108. Kang, Y. et al., 4D Proton treatment planning strategy for mobile lung tumors. Int J Radiat Oncol Biol Phys, 2007. 67(3): 906–14. 109. Chen, G.T., J.H. Kung, and K.P. Beaudette, Artifacts in computed tomography scanning of moving objects. Semin Radiat Oncol, 2004. 14(1): 19–26. 110. Mutaf, Y.D., J.A. Antolak, and D.H. Brinkmann, The impact of temporal inaccuracies on 4DCT image quality. Med Phys, 2007. 34(5): 1615–22. 111. Brock, K.K. et al., Inclusion of organ deformation in dose calculations. Med Phys, 2003. 30(3): 290–5. 112. Flampouri, S. et al., Estimation of the delivered patient dose in lung IMRT treatment based on deformable registration of 4D-CT data and Monte Carlo simulations. Phys Med Biol, 2006. 51(11): 2763–79. 113. Guerrero, T. et al., Elastic image mapping for 4-D dose estimation in thoracic radiotherapy. Radiat Prot Dosim, 2005. 115(1–4): 497–502. 114. Rosu, M. et al., Dose reconstruction in deforming anatomy: Dose grid size effects and clinical implications. Med Phys, 2004. 32: 2487–2495. 115. Schaly, B. et al., Tracking the dose distribution in radiation therapy by accounting for variable anatomy. Phys Med Biol, 2004. 49(5): 791–805. 116. Vanderstraeten, B. et al., Accuracy of patient dose calculation for lung IMRT: A comparison of Monte Carlo, convolution/superposition, and pencil beam computations. Med Phys, 2006. 33(9): 3149–58. 117. Keall, P.J. et al., Monte Carlo as a four-dimensional radiotherapy treatment-planning tool to account for respiratory motion. Phys Med Biol, 2004. 49: 3639–3648. 118. Seco, J. et al., Dosimetric impact of motion in free-breathing and gated lung radiotherapy: A 4D Monte Carlo study of intra/inter fraction effects. Med Phys, 2008. 35: 356–366. 119. Garrity, J.M. et al., Development of a dynamic model for the lung lobes and airway tree in the NCAT phantom. IEEE Trans Nucl Sci, 2003. 50: 378–383. 120. Segars, W.P., D.S. Lalush, and B.M.W. Tsui, A realistic spline-based dynamic heart phantom. IEEE Trans Nucl Sci, 1999. 46: 503–506. 121. Segars, W.P. and B.M.W. Tsui, Study of the efficacy of respiratory gating in myocardial SPECT using the new 4-D NCAT phantom. IEEE Trans Nucl Sci, 2002. 49: 675–679. 122. Riboldi, M. et al., Design and testing of a simulation framework for dosimetric motion studies integrating an anthropomorphic computational phantom into four-dimensional Monte Carlo. Technol Cancer Res Treat, 2008. 7: 449–456. 123. Segars, W.P., Development and application of the new dynamic NURBS-based cardiac-torso (NCAT) phantom. Dissertation in Biomedical Engineering, University of North Carolina, Chapel Hill, NC, 2001. 124. Crum, W.R., T. Hartkens, and D.L.G. Hill, Non-rigid image registration: Theory and practice. Br J Radiol, 2004. 77: S140–S153.

632

Handbook of Anatomical Models for Radiation Dosimetry

125. Rueckert, D. et al., Non-rigid registration using free-form deformations: Application to breast MR images. IEEE Trans Med Imaging, 1999. 18: 712–721. 126. Thirion, J.P., Image matching as a diffusion process: An analogy with Maxwell’s demons. Med Image Anal, 1998. 2: 243–260. 127. Chetty, I.J. et al., Accounting for center-of-mass target motion using convolution methods in Monte Carlo-based dose calculations of the lung. Med Phys, 2004. 31: 925–932. 128. Heath, E. and J. Seuntjens, A direct voxel tracking method for four-dimensional Monte Carlo dose calculations in deforming anatomy. Med Phys, 2006. 33(2): 434–45.

28 Applications to Patient-Speciﬁc Voxel Computational Phantoms in EGS Monte Carlo Codes for Radiation Treatment Involving Photons and Electrons C.-M. Charlie Ma

CONTENTS 28.1 Introduction ...............................................................................................................633 28.1.1 History of the EGS System..........................................................................633 28.1.2 An Overview of the EGS Shower Processes ............................................ 635 28.1.3 Implementation of the EGS System ........................................................... 637 28.1.4 EGS User Codes............................................................................................ 638 28.1.5 EGS Simulation Parameters........................................................................ 639 28.2 Methods ......................................................................................................................640 28.2.1 3D Rectilinear Geometry for Radiation Therapy Simulation................640 28.2.2 Simulation Geometry for Beam Modifiers ............................................... 641 28.2.3 Conversion of CT Data for Radiotherapy Simulation .............................642 28.3 Discussion ..................................................................................................................644 28.3.1 Effect of Voxel Size .......................................................................................644 28.3.2 Effect of Energy Cutoff ................................................................................645 28.3.3 Dose to Water or Dose to Medium? .......................................................... 647 28.3.4 Negative USTEP ........................................................................................... 649 Appendix 28.A......................................................................................................................650 Appendix 28.B ......................................................................................................................650 References ............................................................................................................................. 651

28.1 Introduction 28.1.1 History of the EGS System Butcher and Messel1,2 and Varfolomeev and Svetlolobov3 were the first two groups that conducted a Monte Carlo simulation of high-energy cascades on electronic digital computers. Their collaboration resulted in an extensive set of tables describing the shower distribution functions, which were used by the later codes. Zerby and Moran4,5 developed a Monte Carlo computer code for high-energy electromagnetic cascade simulation, which was used by Alsmiller and others6–8 for a number of studies. Although the code was not readily available to the public, nor was it maintained, some useful ideas and 633

634

Handbook of Anatomical Models for Radiation Dosimetry

experiences were reflected in the Electron Gamma Shower (EGS) code system. Nagel’s code SHOWER9,10 represented a practical tool for the experimental physicists during the middle 1960s—and one version of his code eventually became the EGS3 Code System. The applications of SHOWER, however, were limited because of its unwieldiness and its limitations on scope. For example, its built-in geometry only allowed the use of a single cylinder of Pb and radiation transport could only be initiated for monoenergetic electrons with energies up to 1 GeV. Except for annihilation, positrons and electrons were treated alike and were followed down to 1 MeV (kinetic energy), and photons were followed down to 0.25 MeV (but this still represented cutoff energies that were, at the time, as low as or lower than those used by Messel and Crawford and Zerby and Moran). Continuing efforts were made to modify SHOWER so as to make it versatile, upward compatible, and userfriendly. As a result of the joint effort of Ford and Nelson,11 the Electron Gamma Shower version 3 (EGS3) computer code system was formally introduced. Researchers designed EGS3 to simulate electromagnetic cascades in various geometries and at energies up to a few thousand GeV and down to cutoff kinetic energies of 0.1 MeV (photons) and 1 MeV (electrons and positrons). With the PEGS (Preprocessor for EGS) auxiliary code, radiation transport could be simulated for 100 elements, any compound, or any mixture of those elements. EGS3 also included some processes that provided more efficient sampling schemes. Researchers carried out a wide range of benchmark comparisons and applications after the release of EGS3, especially in the field of high-energy particle physics and in the design of the electromagnetic shower counters. After the release of EGS3, more applications were found in the simulation of low-energy photon and electron transport for a variety of problems in addition to those normally associated with electromagnetic cascade showers. The requirement for the extension of the lower energy limits—i.e., down to 1 and 10 keV for photons and electrons, respectively, was growing fast. As a result of a joint effort by Nelson, Hirayama, Rogers, and other colleagues, the flexibility of EGS was further improved together with an important benchmarking effort in the low-energy particle transport.12 The EGS4 code system was finally introduced in 1985.13 In the following years, the EGS4 system was widely used in medical physics and dosimetry research, especially for low-energy radiation transport problems, though it was still heavily used in high-energy particle physics. Researchers made further improvements to the EGS4 code system, including an angular distribution for the emission of photoelectrons and an electron transport algorithm called PRESTA.14 The latter simplified many of the cumbersome problems related to user-defined restrictions on electron step-sizes and the limitations on electron multiple-scattering. Hirayama and Namito reported a general treatment of photoelectric-related phenomena for compounds or mixtures in EGS4.15 They introduced the energy-dependent branching ratio of each sub shell (L1-, L2-, and L3-) in this improvement by fitting to a quadratic function in log–log from data provided for limited materials (Ag, Pb, and U). This addition is further improved using a generalized treatment,16 in which K-, L1-, L2-, L3-, and other sub shell photoelectric cross sections taken from the PHOTX data base (see below) are fitted to cubic functions in log–log form and the fitted coefficients and other associated data are initialized in a BLOCK DATA subprogram of EGS. It thus becomes possible to calculate the branching ratios for each element of compounds and mixtures inside EGS, negating the need to use approximate piece-wise linear-fitted data from PEGS. Verhaegen et al. reported an accurate simulation for kilovoltage x-ray units employing these developments.17 The EGS4 electron transport was based on the Moliere multiple-scattering theory. Moliere considered his theory to be valid for a number of electron interactions ½ > 20 whereas the default EGS4 uses ½ = 1. In PRESTA, ½ was increased to e, which is the mathematical limit of the Moliere theory. It still needed improvements

Applications to Patient-Speciﬁc Voxel Computational Phantoms

635

to deal properly with electron backscatter,18 and also minimum electron step-sizes when different energy cut offs were used in a simulation, or if a different number of media were used in the same simulation. More recently, researchers incorporated the sampling of bremsstrahlung photon angles into EGS4.19 Duane et al.20 implemented the up-to-date stopping powers recommended by ICRU.21 Ma and Nahum implemented a new algorithm for the EGS4 low-energy electron transport to account for the change in the discrete interaction cross-section with electron energy.22 A large number of applications were reported based on the EGS4 system in radiation detector simulations,18,23,24 radiotherapy treatment machine simulations,25–28 radiotherapy treatment planning dose calculations, and plan verifications.29–34 The results showed that the EGS4 system was sufficiently accurate for radiation therapy treatment planning dose calculations. The most recent development of the EGS code system was the significant improvement in electron multiple scattering simulation by Kawrakow,35 which has led to the release of the NRCC version of the EGS code system, EGSnrc, in 2000.36 EGSnrc also incorporated improvements in the implementation of photon and electron transport, better low energy photon physics, more efficient sampling algorithms, and various bug fixes compared to EGS4. In parallel, researchers made a significant effort to develop the next version of EGS; this EGS5 was announced in 2005.37 EGS5 incorporated many improvements in photon and electron transport improvements. In particular, the PEGS and EGS codes were combined in EGS5 so that cross-sectional data could be generated for every simulation. EGS5 also adopted an advanced electron transport algorithm, which uses a dual random hinge approach.38 The primary advantages of this technique lie in that the random multiple scattering hinge preserves near secondorder spatial moments of the transport equation over long step lengths,39 and that the hinge mechanics can be formulated so as to permit transport across boundaries between regions of differing media. The PRESTA boundary crossing logic was abandoned. Despite many new developments for the EGS system that often associate with somewhat longer computer simulation time, little has been reported on the necessity of these later EGS versions for external beam radiotherapy treatment planning and radiation protection dose calculations. Therefore, the materials presented in this chapter are mainly based on the radiation transport algorithms, geometrical packages and clinical applications of the EGS4 system and their user codes, which are still widely used for radiotherapy and radiation protection applications. 28.1.2 An Overview of the EGS Shower Processes The Monte Carlo technique provides a good way for solving the shower generation problem, not only because all the fundamental processes can be included, but also because arbitrary geometries can be simulated. In addition, any significant radiation interaction processes can be added as a generalization or for various intended applications. One can also switch on or off some events during the shower, and even “force” some rare events to occur more frequently by means of variance reduction techniques. Another important reason for using the Monte Carlo method in shower simulation, however, is that of the intrinsically random nature of the radiation transport, which cannot be treated by solving the radiation transport equation. Although various analytical solutions are available for a limited number of applications, such as the use of asymptotic formulae to describe pair production and bremsstrahlung (under the condition of severe approximation to other processes), many problems of interest today cannot be dealt with by analytic approaches. Analytic treatments, however, may play a significant role in the future. The EGS code system is a Monte Carlo computer simulation program. It simulates the random trajectories of

636

Handbook of Anatomical Models for Radiation Dosimetry

individual particles by using machine-generated pseudorandom numbers to sample from the probability distributions governing the physical processes involved in particle transport and radiation interactions. By simulating a large number of particle histories, information can be obtained about the average values of macroscopic quantities, such as the energy deposited and the particle fluence in a given volume. Due to the statistical nature of the Monte Carlo method, the accuracy of the results depends on the number of histories run (the statistical uncertainty of a Monte Carlo calculation is proportional to the inverse square root of the number of histories), supposing the correctness of the cross-section data used and the coding of the program. In general, the computing time needed for a shower history for a given energy cut off increases slightly more than the increase of the energy of the incident particle. It also takes about 10 times more CPU time to generate an electron history than it does to generate a photon history for a given energy cut off. To achieve an uncertainty of about 1%, or even 0.1%, in Monte Carlo calculations, hundreds of millions of particle histories must be simulated. This is why analog Monte Carlo simulations are so time-consuming. To save the computing time, the EGS4 code system has divided the calculation task into two parts. First a preprocessing code (PEGS4) uses theoretical (and sometimes empirical) formulas to compute the various physical quantities needed, and then prepares them in form for fast numerical evaluation. The user then runs his/her own user code, which has been connected with the EGS4 code, along with these data and his/ her own data to carry out the actual simulation. The EGSnrc and EGS5 follow similar procedures. The PEGS code (Preprocessor for EGS) is a stand-alone utility program originally written in Mortran.40 Although the primary use of PEGS is to produce material data for the EGS itself, it also provides other services for the EGS user to study electromagnetic interactions. In order to do so, PEGS was written in a modular form with over 95 subroutines. These include functions to evaluate the physical quantities needed by PEGS4 and the routines necessary for the operation of EGS4, such as the fitting routines and the routines to write the data for a given material onto a data set. Other routines included in PEGS4 are used to plot the functions on the line printer, or a graphics device, or a comparison of the theoretical final-state density functions with sampled final-state distributions on a line printer plot. It is necessary to check the cross-section data generated using PEGS before any actual calculations are made; at least, the electron stopping-powers and the photon attenuation coefficients, by means of the PEGS examining facilities. Researchers have found that cross-sectional data generated with PEGS for different materials may not be consistent. The stopping-power data for materials not included in the table of default Sternheimer density effect coefficients in PEGS4,13 for example, are calculated based on the general prescription of Sternheimer and Peierls41 for the density effect correction, which might be different from that included in the PEGS4 default table. A good example is the restricted mass stopping powers for water and for the standard Fricke ferrous sulfate solution; water is one of the materials listed in the PEGS4 default table, while the Fricke solution is not. This difference resulted in a 1% lower stopping-power ratio for water to the Fricke solution calculated from the PEGS4 default data than that calculated from the data recommended by ICRU.21 This resulted in the Fricke-to-water dose conversion factor being about 1% lower than the real values for high-energy photon beams (where the stopping-power ratios are dominant in the Fricke-to-water dose conversion factor) and a much larger shifting effect correction factor for high-energy electron beam Fricke dosimetry.22 The problems were resolved by implementing the ICRU-37 stopping-powers in PEGS4 using the method by Duane et al.20 The default version of EGS4 is an analog Monte Carlo program that has been designed to simulate the actual physical processes. Researchers did not introduce any inherent variance reduction techniques into the default version of EGS4, in order that the fluctuations

Applications to Patient-Speciﬁc Voxel Computational Phantoms

637

in the Monte Carlo results can truly be representative of real-life fluctuations. To make some calculations more efficient, however, it is necessary to implement various variance reduction techniques. This has been done by adding Mortran3 templates in key positions in EGS4 and the respective user codes to facilitate the inclusion of macros for the required implementation. The simulation of an electromagnetic cascade shower can be decomposed into a simulation of the transport and interactions of a single particle, along with some necessary bookkeeping. A last-in/first-out stack is used to store the properties of particles that have yet to be simulated. The basic strategy is to transport the top particle on the stack until an interaction takes place, or until its energy drops below a predetermined cut off energy, or until it enters a particular region of space. In the latter two cases, the particle is taken off the stack and the simulation resumes with the new top particle. If an interaction occurs, and if there is more than one product particle, the particle with the lowest energy is put on the top of the stack. Initially, a shower history will start with only the incident particle on the stack. When a particle is removed from the stack and none remain, the simulation of the history is complete. 28.1.3 Implementation of the EGS System The EGS code itself consists of two user-callable subroutines, HATCH, and SHOWER, which in turn call the other subroutines in the EGS code system, including two userwritten subroutines, HOWFAR and AUSGAB, which determine the geometry and scoring, respectively. To use the EGS, the user must write a user code, which consists of a MAIN program and subroutines HOWFAR and AUSGAB. Additional auxiliary subroutines might be included in the user code to facilitate the initiation of the simulation and data processing. Various COMMON variables can be used in the user code for communications with the EGS subroutines. Usually, MAIN will perform the initialization needed for the geometry subroutine, HOWFAR, and will set the values of certain EGS COMMON variables, which specify the medium names to be used, the desired energy cut off values, and the unit of distance. MAIN then calls the HATCH subroutine to read in the crosssection data from the data sets prepared by PEGS for the materials requested. After this initialization, MAIN may then call SHOWER when desired. Each call to SHOWER results in the generation of one particle history.13 Figure 28.1 shows the flow of control and data in an EGS program after a user code is connected to the EGS code system. An EGS user code can be used as a complete “black box” for new EGS users as long as the rules for setting up the EGS parameters are obeyed. Even experienced users can easily make mistakes, however, in running an EGS user code (and all the other Monte Carlo computer codes), such as errors in setting up the geometry to be simulated, inappropriate choice of the energy cutoff values, and most of all, inappropriate modeling of the physical quantities to be calculated. Conversely, advanced EGS users may take advantages of the flexibility of the EGS code system to go deep into the code and modify the treatments for individual processes or add new variance reduction techniques to an existing user code. This can be done by including macrodefinitions in MAIN to control or override various functions in EGS as well as in the user-written codes. In summary, an EGS user code operates EGS by means of 1. COMMON blocks—to change values of COMMON variables 2. Macrodefinitions—to redefine the predefined features 3. EGS4 or user-written subroutines

Handbook of Anatomical Models for Radiation Dosimetry

638

User code Output files

User control data

MAIN

HATCH

PEGS4 DATA

HOWFAR

SHOWER

ELECTRON

AUSGAB

PHOTON

EGS4 code

FIGURE 28.1 The implementation of EGS4. The thick horizontal line is the interface between the two components of EGS4— the user code and the major EGS4 subroutines. The arrows indicate the flow of information when running the EGS4 code system.

HATCH—to establish media cross-section data SHOWER—to initiate the cascade HOWFAR—to specify the simulation geometry AUSGAB—to score and output the results 28.1.4 EGS User Codes An EGS user code is usually referred to as the MAIN/HOWFAR/AUSGAB combination (plus auxiliary subroutines and macros). In this section, several EGS user codes used in radiotherapy treatment planning dose calculation will be discussed, all of which have used similar geometric specifications for patient-specific voxel (volume elements) models. Many other EGS user codes developed for other general applications, such as detector response, shielding design, and radiation treatment unit simulations, will not be discussed because of their problem-specific geometries. DOSXYZ is a general-purpose Monte Carlo EGS442 user code for three-dimensional (3D) dose calculations. DOSXYZ simulates the transport of photons and electrons in a Cartesian volume and scores the energy deposition in the designated voxels. DOSXYZ is “stand alone” in the usual EGS sense in that it is controlled by scripts, which explicitly refers to the .pegs4dat and .egs4inp files, and is capable of writing out ASCII formatted dose distribution arrays. The code uses the standard National Research Council of Canada (NRCC) system of scripts. The geometry is a rectilinear volume with the X–Y plane on the page, X to the right, Y down the page and the Z-axis into the page. Voxel dimensions are completely variable in all three directions, although only uniform voxel sizes are allowed for radiotherapy dose calculation with CT data. Individual voxels can have different materials and mass densities (for use with CT data). DOSXYZ can use simplified monoenergetic diverging or parallel beams, phase-space data generated by a BEAM simulation,43 or a model-based beam reconstruction produced by BEAMDP.44 CT to phantom conversion can be performed using a CTCREATE program, which converts CT data sets into the voxelized geometry needed for DOSXYZ to do a simulation. It handles ADAC Pinnacle,

Applications to Patient-Speciﬁc Voxel Computational Phantoms

639

AAPM, CADPLAN, and DICOM formats for the CT fi les.42 Researchers adopted the concise specification and efficient boundary tracking for the voxelized patient geometry in DOSXYZ from XYZDOS, which was written by Dave Rogers in 1986 to show Ralph Nelson that the special purpose coding of rectilinear voxels was faster than using Ralph’s more general macros. Commonly used combinatory geometries suffer from the same reasons when simulating voxelized geometries. XYZDOS was used to estimate the time required to do a full Monte Carlo treatment planning calculation, and the results were published 3 years later in a book chapter.18 It was also used in a timing benchmark study.45 With the development of EGSnrc, DOSXYZ was ported to the EGSnrc system by Blake Walters to give DOSXYZnrc.46,47 A significant effort was made to expend the DOSXYZ functionalities, especially to facilitate the dose calculation for advanced radiotherapy treatments at Stanford University and at the Fox Chase Cancer Center. As a result, researchers developed the EGS4 user codes MCDOSE and MCSIM.48,49 Developed as a response to the need for a clinical dose calculation, MCDOSE and MCSIM can also be used as a treatment verification tool for radiotherapy. One advantage of an independent dose calculation and verification tool is that one can compare the dose distributions from different treatment modalities, such as stereotactic radiosurgery (SRS) and intensity modulated radiation therapy (IMRT), which are often planned using different treatment planning systems (TPS). Another advantage is that one can use it to develop new treatment techniques, which are not commercially available. Of course, one can also use this tool to verify the results from any TPS before or after a treatment is delivered, or to perform clinical research on dosimetric quantities that cannot be calculated by an existing TPS. MCDOSE and MCSIM have included many features from earlier EGS user codes and added new functionalities to facilitate clinical applications. A major improvement with the geometry simulation in MCDOSE and MCSIM is the combination of patient-specific geometry and beam modifiers in the same simulation. This advancement has greatly improved the simulation efficiency for radiotherapy treatment planning dose calculation. Another major improvement is the inclusion of treatment target and critical structure information in the geometry package so that dose–volume histograms (DVH) can be calculated, which is an essential feature for clinical radiotherapy plan comparison and evaluation. Many variance reduction techniques were implemented in the MCDOSE and MCSIM codes to improve the simulation efficiency; the CPU time for typical photon and electron treatment plans calculated by MCDOSE and MCSIM is 10–30 times less than that by DOSXYZ.50 For clarity, all the geometrical descriptions in this chapter are based on the implementation and coding in the MCDOSE and MCSIM codes. 28.1.5 EGS Simulation Parameters According to the classification by Berger,51 the electron transport scheme in the EGS code system belongs to a Class II code category. This means that the production of δ-ray particles and bremsstrahlung photons is treated individually. As a consequence, two of the requirements to run the system are to define the threshold energies for such interactions, and to prepare cross-section data for required threshold energies (using PEGS), which will in general vary for different simulation applications. The threshold energy for δ-ray production is called AE; electron “hard” collisions are simulated by discrete interactions creating δ-ray particles with energy above AE and by continuous collisional energy loss to electrons with energies below AE. The continuous part of the energy loss, which consists of the δ-ray events with energies below AE and other “soft” collisions and lowenergy-loss events, is modeled using the restricted stopping-power. The threshold energy

640

Handbook of Anatomical Models for Radiation Dosimetry

for bremsstrahlung photon production is called AP; the radiative energy loss to electrons is accounted for in either a discrete or continuous manner, depending on the energy of the bremsstrahlung photon above or below the photon energy threshold AP. Two further parameters required for the Monte Carlo transport of electrons and photons have been termed ECUT and PCUT, which are used in both Class I and Class II algorithms, respectively. Electrons and photons with energies below these energy cutoffs will be terminated and their energies deposited locally. The real purpose of the energy cut offs is to terminate the simulation of a particle (e.g., electron) transport so that it will not “go on forever.” Obviously, one can also use these two parameters to stop some electron and photon histories if the particles cannot make any further contribution to the final results. However, caution must be exercised in choosing these parameters, as high values of ECUT and PCUT may seriously bias the results in some cases, such as in interface simulation and in the calculations of ionization chamber correction factors.52 During recent years, the low-energy electron transport in EGS4 has been improved, mainly by the inclusion of user-defined restrictions in electron step-sizes12 through two parameters ESTEPE and TMAX. ESTEPE is the maximum fractional energy loss per electron step, which varies between 0 and 1. TMAX is the upper limit (in cm) on the electron step-size. The choice of the two parameters depends on many factors such as cross-section data, the simulation geometry and the various limitations set by the Moliere theory, which is used for accounting for electron multiple scattering in EGS. In general, TMAX is confined by ESTEPE or the simulation geometry and using smaller ESTEPE will result in a more accurate simulation of the electron transport, as long as the electron step-size is not too small to invalidate the Moliere multiple-scattering formalism. As a tradeoff, the computing time increases dramatically with decreasing ESTEPE. This was greatly improved, however, by the introduction of the Parameter Reduced Electron-Step Algorithm (PRESTA) in late 1980s.14 Dose distributions calculated using the EGS4 system based on the Moliere theory was shown to be accurate for radiotherapy treatment planning applications.30,31,48–50 Further improvements were made by adopting more accurate electron multiple scattering theories as in EGSnrc and EGS5, which are necessary for accurate dose calculations in complex detectors involving thin geometries and high-Z materials.35–38 In summary, successful simulations need the correct choice of the Monte Carlo code system and the associated transport parameters. This is usually done according to a user’s experience, and by careful comparisons and analyses of the different options available. The choice of the transport parameters must be made according to the individual situations.18

28.2 Methods 28.2.1 3D Rectilinear Geometry for Radiation Therapy Simulation The MCDOSE and MCSIM codes were developed for accurate and efficient dose calculation in clinical radiotherapy applications, including treatment planning and dosimetry verification.48,49 Both EGS4 user codes employ original EGS4 photon and electron transport algorithms. Their simulation geometry is an integrative geometry consisting of several beam modifiers and a 3D rectilinear voxel (volume elements) patient geometry model. The voxel geometry is used to model the patient heterogeneous anatomy. The voxel dimensions are variable in all three directions. Every voxel can be assigned to a different

Applications to Patient-Speciﬁc Voxel Computational Phantoms

641

material, and the cross-section data for the materials used are available in a preprocessed PEGS4 cross-section data file. The density of the material defaults to that in the PEGS4 data file, but can also be varied in a patient-specific dose calculation for use with the patient’s CT data, although the density effect corrections for the stopping powers of the material remain unchanged. The voxel dimensions and materials were defined in an input file, together with the geometry input for the beam modifiers. A sample of the MCDOSE/ MCSIM input data file describing the 3D simulation geometry is given in Appendix 28.A. The beam and patient setup parameters are also input through the input file together with the transport parameters such as the energy thresholds for secondary particle creation (AE, AP), energy cutoffs for particle transport (ECUT, PCUT), maximum electron step length (SMAX) or fractional energy loss per electron step (ESTEPE), and the parameters required by PRESTA. A special program was developed to convert the patient’s CT data (or electron density data) to desired dimensions, material types, and mass densities. The program can read directly the patient CT, the beam setup, and the target and contour data from the clinical treatment planning computers. A tag called REGFLAG is used to label each voxel in order to associate it with the structures of interest. Every bit of REGFLAG is associated with a structure, and a voxel can be associated with up to 32 structures using REGFLAG. For example, bit 0 is used to tell whether the voxel is inside the patient external contour (set to 1 if it is inside and 0 otherwise). Based on the patient external contour, all the voxels outside the patient are set to air with density equal to 0.001 g cm−3. Bit 1 is often used to mark the gross tumor volume (GTV) and the other bits for any other structures or organs. MCDOSE and MCSIM compute these dose volume histograms (DVHs) for the target and any structures of interest if the patient contour information is available. They also produce a data file that contains geometry specifications, such as the number of voxels in all the three directions and their boundaries as well as the dose values and the associated (1σ) statistical uncertainties in the individual voxels. A sample of the 3D dose data file is given in Appendix 28.B. The dose distributions can be analyzed using a program called MCSHOW to plot or compare isodose distributions on patient’s CT images with target and critical organ contours shown. The MCSHOW program can also be used to smooth the isodose distributions due to statistical uncertainties using simple linear spatial average or different distributions such as a Gaussian distribution with variable σ values. 28.2.2 Simulation Geometry for Beam Modifiers Beam modifiers are important components of medical accelerators and treatment aids that can shape the radiation fields (such as jaws, blocks, multileaf collimators (MLC), and cutouts) and can modulate the beam intensities or ranges (such as wedges, compensators, and boli). Since the use of beam modifiers is patient-specific, the simulation of beam modifiers has to be combined with the patient geometry built from CT data. In the MCDOSE and MCSIM implementation,48,49 the beam modifiers were simulated as a series of 3D objects stacked one above another with each modifier occupying a slab in space. Researchers performed a full particle transport for beams modifiers including jaws, wedges, compensators, blocks, and cutouts. The effect of a bolus can be simulated by adding the bolus material to the patient geometry. A dynamic beam delivery using an MLCs and dynamic wedges were simulated by changing the particle weighting factors during a simulation using an intensity distribution derived from the segment treatment table (STT) or the MLC leaf sequence file. For IMRT dose verification, MLC leaf sequence files were used to reconstruct the photon intensity distribution for each treatment field (including the effect of MLC leaf leakage). The implementation of the dynamic beam delivery simulation was described in

642

Handbook of Anatomical Models for Radiation Dosimetry

detail31 and the accuracy of the physical beam modifier simulation was verified with comparisons to the EGS4/BEAM/DOSXYZ results.50 The simulation geometry for each of the physical beam modifiers is described below. Photon jaws are modeled as two pairs: the x jaws and y jaws. The inner edges of the bars focus to the target. The dimensions and material for the jaws are the same as the manufacturer’s specifications (the default values). For the patient dose calculations, only four parameters are required in the input file to specify the jaw opening, i.e., the locations of the inner edges of the upper and lower jaws (i.e., x1, x2, y1, y2). These values correspond to the light field size defined by the jaws at the isocenter plane (100 cm away from the x-ray source). For different accelerators, the jaw locations and dimensions for the upper and lower jaws are constructed properly based on the given jaw opening values. The photon blocks and the supporting tray are simulated as one beam modifier. The location, thickness, material of the blocks, and the tray are required. The location is defined as the distance from the bottom surface of the block tray to the isocenter plane (at 100 cm SSD). The block opening can be of any shape. The opening is specified by the coordinates of the vertices as projected on the isocenter plane. The user needs to input the points continuously around the perimeter of the opening. The planes defining the inner surfaces of the opening are angled with respect to the Z-axis (beam axis) toward the target through a single focus point. This module can also be used to simulate a static MLC by setting the material of tray to air. The physical wedge and the supporting tray are simulated as one beam modifier using the geometry information provided in the input file. The parameters required are the orientation, the location, the material and the two-dimensional point coordinates for the wedge and the materials and thickness for the tray. A compensator (including the tray) requires the input for the location, the materials and the tray thickness. The compensator is divided into different bins in two directions. The bin (pixel) boundaries and thickness for each bin are required. An electron cutout is simulated in a way similar to a photon block except that it does not have a tray. Depending on the applications, it is optional to have diverging planes or straight and parallel inner planes to define the opening. The location, material, and the coordinates of each vertex used to specify the opening (as projected on the isocenter plane) are required. The user needs to input the points continuously around the perimeter of the opening. The bolus is simulated by adding an extra layer of the bolus material to the patient’s geometry (outside the patient external contour) according to the bolus’ material and thickness. This can be done by changing the voxel material from air to the bolus material for all the voxels with their centers inside the bolus contour. This change introduces an uncertainty in the bolus thickness up to one-half of the voxel dimension. 28.2.3 Conversion of CT Data for Radiotherapy Simulation Ma et al.30 discussed in detail the conversion of the CT numbers to materials and mass densities for radiotherapy treatment planning dose calculation. A similar conversion process is used by the MCDOSE and MCSIM codes48,49 and a program was developed to convert files directly from commercial TPS in DICOM RT format or RTOG format. The essential information required for the conversion process includes the CT format, the CT data file name, and voxel dimensions for the simulation geometry. For any material, the mass density and CT number limits can be set by the user. The density for a given voxel is assigned by a linear interpolation of a mass density versus a CT number curve. Table 28.1 gives the

Applications to Patient-Speciﬁc Voxel Computational Phantoms

643

TABLE 28.1 Density Range for the Four Materials Used to Build the Monte Carlo Simulation Phantoms (a); Only Air, Tissue, and Bone Were Used if a Lung Was not Present in the Regions of Interest (b) (a) Phantom with a lung Medium CT # range Mass density range

Air

Lung

Tissue

Bone

0–70 0.001–0.07

70–800 0.07–0.80

Air

Tissue

Bone

0–70 0.001–0.07

70–1250 0.07–1.28

1250–4000 1.28–2.88

800–1250 0.8–1.28

1250–4000 1.28–2.88

(b) Phantom without a lung Medium CT # range Mass density range

Source: Data from Ma, C.-M. et al., Med. Phys., 26, 2133, 1999.

CT number range and density range for each of the four materials used in the implementation (Table 28.1a). These are consistent with the recommendations in an ICRU report.21 The density-effect corrections of the stopping powers for air and tissue were from another ICRU report.53 The lung material is used only when a lung is present. Otherwise, the same CT number range 70–1280 is replaced by tissue with the same density limits (Table 28.1b). The CT slices are usually scanned at 3 or 5 mm intervals. A CT image typically contains 512 × 512 pixels with the side of a pixel equal to about 0.94 mm for a 48 cm field of view (FOV). A subsection of the original CT data can be included in the CT simulation geometry that only includes volumes of interest. This enables the simulation to be performed at a resolution higher than if the calculation used the entire CT volume, and enabled the user to trim some of the air surrounding the patient on a typical CT image, which becomes more significant for CT data from a large bore CT that has a 75–80 CM FOV. It is sometimes more convenient to assign the electron density ranges to materials of different mass density ranges, since the CT number to electron density conversion is already calibrated if the TPS has been properly commissioned. Table 28.2 shows the electron density range and the density range for air, soft tissue, and bone. The effect of the composition differences among various soft tissue types and between soft bone and compact bone is ignored, due to the negligible clinical impact. MCDOSE/MCSIM can also compute dose in a water-only phantom with a mass density varying according to the electron density in each voxel. The dose is given as the energy deposited in the material in the voxel by default. One can also give the dose as dose-to-water, if so required. This can be done by converting the dose-to-materials to the dose-to-water, during the simulation, using the materials-to-water

TABLE 28.2 Electron Density Values and the Corresponding Mass Density Values to Build the Piece-Wise Linear Conversion Curve for the Three Materials Used to Build the CT Simulation Phantoms for Monte Carlo Dose Calculations Material

Air

Electron density Mass density Source:

Tissue

Bone

0.001–0.700

0.700–1.000 1.000–1.207

1.207–1.707 1.707–2.639

0.001–0.700

0.700–1.000 1.000–1.253

1.253–1.839 1.839–2.880

Data from Ma, C.-M. et al., Phys. Med. Biol., 47, 1671, 2002.

644

Handbook of Anatomical Models for Radiation Dosimetry

stopping power ratios.54 This conversion may alter the dose values by 3%–10% for soft bone or compact bone but it has little effect on soft tissues (generally smaller than 2%). Caution must be exercised in converting the beam setups for the Monte Carlo dose calculation. Although both commercial TPS and EGS user codes employ a right-hand coordinate system, a 270° rotation with respect to the x-axis is often needed to convert the TPS/ Linac coordinates to the MCDOSE/MCSIM coordinates for the CT data. Furthermore, the field shape on a TPS/Linac is often defined in a beam’s eye view (BEV) while the EGS4 BEAM43 and the MCBEAM55 systems and MCDOSE/MCSIM codes use the patient’s eye view. This required another conversion (180° rotation with respect to the x-axis for the phase space) when researchers simulated the BLOCK and the MLC. Researchers in MCSIM also developed additional coding to handle field shaping by electron cutouts and the MLC. The cutout shape or the MLC leaf positions were coded in BEV so that the latter conversion could be omitted. This was further facilitated by the symmetry of the phase-space data with respect to the x- and y-axes because they were either scored at a plane immediately above the lowest scraper for electrons or immediately above the secondary collimators for photons. Other considerations included gantry rotation, collimator rotation, couch rotation, and translation.

28.3 Discussion 28.3.1 Effect of Voxel Size In radiotherapy treatment planning, voxel size plays an important role in each step of the planning process; from the structure contouring, the field size/beamlet definition, the dose calculation, the plan optimization, to the plan analysis. Larger voxel sizes may affect the accuracy of the dose planning and smaller voxel sizes will increase computation time significantly. The choice of voxel size is a compromise between the accuracy and the computational speed. For treatment sites with a small target and critical structures, and in inhomogeneous regions, a smaller voxel size is expected to ensure the accuracy in dose calculation. The effects of voxel size on the accuracy and the computational speed of Monte Carlo-based radiotherapy treatment planning have been reported by a number of investigators.56–58 Monte Carlo dose calculations rely on patient CT data to provide the 3D information of the anatomical structures. A typical CT image has a spatial resolution of 1 mm. In order to achieve the same accuracy as the CT images, ideally the voxel size used in Monte Carlo calculations should be the same as CT. Due to the computing speed and the memory limitations, patient CT data may be converted into Monte Carlo simulation geometries with larger voxels. Patient CT used for radiotherapy treatment planning generally has a 0.94 mm resolution with a 512 × 512 matrix (for a 48 cm FOV). The patient CT data were converted into 256 × 256, 128 × 128 and 64 × 64 matrix CT geometries, respectively, to investigate the voxel size effect.56–58 Ma et al. performed Monte Carlo dose calculations for four prostate cases using these three voxel sizes and the same MLC leaf sequences as generated by CORVUS. They compared targets and critical structure volumes. The maximum volume change between different voxel sizes was about 4% for the GTV. There are less than 1% volume differences, however, in GTV between 256 × 256 and 128 × 128 matrices. They also compared isotope distributions and DVHs for the target and critical structure and found no significant differences (<3%) for DVH among the three voxel sizes. Also, the

Applications to Patient-Speciﬁc Voxel Computational Phantoms

645

changes in DVH between 256 × 256 and 128 × 128 matrices were negligible. The changes in isocenter position were studied and their effects on IMRT plans were clinically insignificant. In Monte Carlo IMRT planning, preoptimization beamlet dose calculations using different voxel sizes may vary with beamlet sizes, but their effects were usually small: between 4 × 4 and 2 × 2 mm2. However, the results from the optimizations showed that different voxel sizes may generate different beamlet weights, and could cause differences in MU calculations and generate different plans, and the effects could be clinically significant. The effect on optimization is still under investigation. The effects of different voxel sizes on the dose and DVH calculations depend on the treatment site, inhomogeneity and the volumes of the targets/organs, which could be up to 10% between using 4 × 4 mm2 and 2 × 2 mm2 voxel sizes. For the pelvic region, the effects on the dose calculations for the same plan (calculated using the same leaf sequences generated on Corvus) were not clinically significant between 4 × 4 mm2 and 2 × 2 mm2; 4 × 4 mm2 voxel size can ensure both the accuracy and computational speed. For head and neck treatment with an air cavity close to the treatment area, however, a 2 × 2 mm2 voxel size is needed specially for small structures like optical nerves. This is also true for small treatment target in SRS and stereotactic radiotherapy (SRT); some small (1–10 cm3) intracranial and extracranial (lung) targets may require 1 × 1 mm2 voxels. Large (>2 × 2 mm2) voxel sizes may affect the isodose distributions significantly for electron beams due to the spatial averaging effect of the air and tissue/bone voxels.57 28.3.2 Effect of Energy Cutoff The energy cutoff values for particle transport, especially for the electron multiple-scattering simulation, play an important role in computer efficiency and accuracy; smaller cutoffs will generally result in a more accurate simulation of the low energy particles, but this means more CPU time to track these low energy particles. Ma et al.49 investigated the effect of energy cutoffs for electron transport on the dose calculation accuracy and the CPU time. Figure 28.2a shows the dose distributions calculated with different electron cutoff energies (total energy) with other transport parameter values unchanged. This is a 9-field IMRT plan for a paraspinal treatment using 15 MV photon beams (the mass density in the target and surrounding regions is unity or higher). The difference between the dose distributions calculated using 700 keV and 1 MeV was negligible (not shown). The simulation time for the 1 MeV run, however, was about 30% less compared to the 700 keV run. The isodose curves calculated using the 1.5 MeV ECUT were within 0.1–0.2 cm of those calculated using lower ECUT values, although the CPU time for this run was further reduced by another 40% compared to the 1 MeV run. The shift in the isodose lines for the 2.5 MeV case (more than 0.2 cm sometimes), however, may be considered to be unacceptable clinically. The electron continuous slowing-down (CSDA) range for 1 MeV kinetic energy is about 4 mm in soft tissue, which is about the same as the voxel size used in the dose calculations. The shift in the isodose curves was expected to be within a voxel for the 1.5 MeV ECUT case. For ECUT smaller than 1 MeV, the electron CSDA range in soft tissue is smaller than half of the voxel size used. The changes in the isodose lines should be generally negligible. Figure 28.2b and c shows dose distributions for a mantle field plan involving a low density lung for 0.7, 1.0, and 1.5 MeV ECUT. There is little difference in the isodose lines for the regions inside the treatment field and outside the lung for the three different ECUT values. The 10% isodose line, however, shifted by more than 0.5 cm in the lung region outside the treatment field when 1.5 MeV ECUT was used compared to ECUT = 0.7 MeV (Figure 28.2c). This result is not surprising since the residual range of a 1 MeV electron is 1–2 cm in lung

Handbook of Anatomical Models for Radiation Dosimetry

646

(a)

(b)

(c)

(d)

(e)

(f)

FIGURE 28.2 (See color insert following page 524.) Isodose distributions calculated using different AE and ECUT values: (a) a paraspinal treatment plan calculated with AE = 0.7 MeV, ECUT = 0.7 MeV (thick lines) and ECUT = 2.5 MeV (thin lines); (b) a mantle field plan calculated with AE = 0.7 MeV, ECUT = 0.7 MeV (thick lines) and ECUT = 1.0 MeV (thin lines); (c) a mantle field plan for AE = 0.7 MeV, ECUT = 0.7 MeV (thick lines), and ECUT = 1.5 MeV (thin lines); (d) a nasocavity plan with a 6 MeV electron beam calculated using ECUT = AE = 0.521 MeV (thin lines), AE = 0.7 MeV, and ECUT= 1.0 MeV (thick lines). The isodose lines in (d) are 4%, 10%, 30%, 50%, and 90% of prescribed dose; (e) same nasocavity geometry as in (d) irradiated by a 1 cm × 1 cm 4 MV photon beam. Thick lines are for AE = ECUT = 0.521 MeV and thin lines are for AE= 0.7 MeV and ECUT = 1.0 MeV; (f) same as in (e) with thick lines representing AE = ECUT = 0.521 MeV and thin lines representing AE = ECUT = 0.7 MeV. The isodose lines shown in (e) and (f) are 1%, 5%, 10%, 30%, and 50%, normalized to the maximum dose for AE = ECUT = 0.7 MeV. (From Ma, C.-M. et al., Phys. Med. Biol., 47, 1671, 2002. With permission.)

Applications to Patient-Speciﬁc Voxel Computational Phantoms

647

(the density varies between 0.2 and 0.4 g cm−3). The difference in isodose lines between ECUT = 0.7 MeV and ECUT = 1.0 MeV was less than 0.2 cm in general (Figure 28.2b). The residual range of a 0.5 MeV electron is 0.3–0.6 cm and this should only have a limited effect on the low isodose lines outside the treatment field. The residual range of a 0.2 MeV electron is 0.1–0.2 cm in lung and the 0.7 MeV ECUT value should be sufficient for the simulation in lung. The final case dealt with the dose calculation for a head and neck plan. Figure 28.2d shows the isodose distributions for a nasocavity calculation with a 4 cm × 4 cm 6 MeV electron beam. The isodose lines are 4%, 10%, 30%, 50%, and 90% of prescribed dose. The thin lines represent calculations with ECUT = AE = 0.521 MeV and the thick lines represent calculations with AE = 0.7 MeV and ECUT = 1.0 MeV. The shifts in the 4% and 10% isodose lines exceeded 3 mm in some locations but in general the shifts were within 0.2 cm for higher isodose lines. These observations confirmed that the “early” termination of the low energy electrons with higher ECUT shortens the effective electron range. Similar results were found for 12 and 20 MeV electron beams. Further dose calculations were performed for the same nasocavity geometry using 1 cm × 1 cm 4 MV photons. Figure 28.2e shows the isodose distributions calculated using AE = ECUT = 0.521 MeV (thick lines) and AE = 0.7 MeV and ECUT = 1.0 MeV. The 1%, 5%, 10%, 30%, and 50% isodose lines (normalized to the same dose maximum for AE = ECUT = 0.7 MeV) are shown. There are significant (more than 0.3 cm) shifts in the low and high isodose lines. The difference in the dose is more than 10% of the maximum dose. In Figure 28.2f, we can see the isodose distributions are for the same case as in Figure 28.2e with AE = ECUT = 0.521 MeV (thick lines) and AE = ECUT = 0.7 MeV (thin lines). Although the dose difference can be as high as 4%–6% of the maximum dose in some locations, the shifts in isodose lines are negligible in the high isodose lines (above 5%), again indicating that a 0.7 MeV value is adequate for both ECUT and AE for air cavity simulations. 28.3.3 Dose to Water or Dose to Medium? Several important factors may affect the Monte Carlo calculated dose distributions and the DVH curves. First, the isodose lines and the DVH curves are affected by the materials used in the patient CT geometry, i.e., whether we plot dose to tissue only or dose to any material (such as air, tissue, or bone). It seems reasonable that previous radiotherapy experience was based on dose to water (or dose to tissue, the difference between the two is generally within 1%) and therefore the dose values should be expressed as dose to water/tissue. It can also be argued, however, that the real dose to the biological material, such as bone, should be given whenever possible. Only in this way can the relationship between the “old” practice and “new” experience be established. The AAPM TG105 report recommended that both dose to water and dose to medium be reported in the fi nal calculation results. To understand the effect of the conversion of the dose to different materials, Figure 28.3 shows the dose distributions calculated using Monte Carlo with different materials and density configurations for an IMRT case.31 In Figure 28.3a, we can see that we calculated the dose distribution using tissue with unit density and air. This phantom should not show any heterogeneity effect due to the change in tissue (or bone) densities. In Figure 28.3b, the dose distribution was calculated using air and tissue with variable density converted from the CT data. This is similar to the model used by the Corvus system. In Figure 28.4c, the dose distribution was calculated with air, unit density tissue, and 10 g cm−3 density bone. For comparison, the dose distribution calculated using air, tissue, and bone with proper densities converted from the CT data is shown in each figure. The isodose curves were computed by normalizing the dose values to the prescribed target dose.

Handbook of Anatomical Models for Radiation Dosimetry

648

(a)

(b)

(c)

FIGURE 28.3 Dose distributions for different tissue types and material densities: (a) tissue and bone with variable density (thick line) and tissue with unity density (thin line); (b) tissue and bone with variable density (thick line) and tissue with variable density (thin line); and (c) tissue and bone with variable density (thick line) and tissue with unity density and bone with 10 g cm−3 density (thin line). The phantom geometry and beam arrangements are the same as in Figure 28.2a. The isodose lines are given as 10%, 30%, 50%, 70%, and 80% of the prescribed target dose. (From Ma, C.M. et al., Phys. Med. Biol., 45, 2483, 2000. With permission.)

100

Volume (%)

80

60

40

20

0 0.0

5.0

10.0

15.0

20.0

25.0

Dose (Gy) FIGURE 28.4 DVHs for the GTV in Figure 28.3. Full circles are obtained using CT geometry consisting of air, tissue, and bone of variable density with dose calculated for each material. Open circles are obtained using the same geometry with dose converted to tissue using the stopping power ratios of tissue to bone or tissue to air. Open squares are obtained using CT geometry consisting of only tissue with variable density. (From Ma, C.M. et al., Phys. Med. Biol., 45, 2483, 2000. With permission.)

The differences in the isodose lines in Figures 28.3a and 28.4b are small (2%–3%). This difference is also clearly shown in Figure 28.4 for the target DVHs. The difference between dose to tissue and dose to bone would be about 3.5% for soft bone and about 10% for hard bone,54 if converted using the stopping power ratios for tissue to bone (or soft bone) at these beam qualities, assuming the same electron energy fluence. Clearly, such conversion is not equivalent to performing the Monte Carlo simulation using a water (or tissue) phantom with variable mass (or electron) density. In fact, Figure 28.4 indicates that the previous dose to water calculation with variable electron density is similar to the Monte Carlo

Applications to Patient-Speciﬁc Voxel Computational Phantoms

649

dose to medium calculation in terms of the numerical values. This may imply that dose to medium is a better quality for Monte Carlo dose calculations, which provides a direct link to the conventional dose to water calculation and such dose values can be directly used for outcome analysis. It is known that electron backscattering from the high atomic number materials may disturb the dose in tissue near the tissue-bone interface. This effect is less significant when dose values are averaged over course scoring volumes (0.3–0.4 cm voxels). In Figure 28.3c, the density of bone was artificially increased to 10 g cm−3, which caused significant attenuation of the beams and therefore altered the doses behind the bones. The effects on the surface doses are smaller than for the high dose regions. The maximum dose for the artificially high-density bone geometry is about 15% lower than that for the phantom with normal material density. 28.3.4 Negative USTEP When a particle crosses a boundary of two regions, the EGS will stop the particle at the boundary and then find the new region number to continue the particle transport. This stopping is necessary with EGS, at least for two regions of different media or mass densities, because only one medium/mass density can be considered at a time for the particle transport and interactions. The negative USTEP arises simply because of computational errors such as truncation or round-off errors; after a particle is transported from one location to another the actual coordinates of the particle may not correspond to the exact location intended. As shown in Figure 28.5, a particle is transported near a boundary. Due to truncation or round-off errors, the particle may end up either on the other side of the boundary (an overshoot) even if the particle is supposed to move in the same region, or is still in the same region (an undershoot) even if the particle is supposed to stop at the boundary. For charged particles, EGS applies a multiple-scattering angular deflection at the end of the step (some other Monte Carlo codes also do it this way), which can cause an electron or positron to scatter away from the boundary that it is assumed to have crossed, leading to a negative USTEP. Figure 28.5 illustrates how a negative USTEP occurs. In the first example, a particle approaches a boundary with an “overshoot” due to a computational error and then moves into the new region. When EGS places the particle on the boundary of the new region, the distance from the particle position to the boundary becomes negative. In the

Forward, approaching a boundary + an overshoot Negative USTEP Accurate steps Backward, leaving a boundary + an undershoot Negative USTEP Accurate steps

FIGURE 28.5 Negative USTEPs are caused by truncation or round-off errors. Shown in the schematic are two typical situations, where a negative USTEP occurs when a particle approaches a boundary with an overshoot and when a particle is scattered back from a boundary with an undershoot.

650

Handbook of Anatomical Models for Radiation Dosimetry

second example, a particle arrives at a boundary with an “undershoot” due to a computational error and then is scattered back into the same region. When EGS places the particle on the boundary, the calculated step length becomes negative. In both cases, EGS will set the USTEP to zero and continue the particle transport from the current position. If one gets a few negative USTEP warnings, it is safe to simply ignore them. The warning is telling the user that this occurred with a magnitude greater than 10−4, which was a completely arbitrary limit originally set in the EGS codes. If one gets many negative USTEP warnings, it is likely that there is an error in the geometry input; such as, the plane boundaries out of order, or the geometry contains too many very thin layers with low-density media. Many EGS user codes will abort after 1000 of these warnings. One can increase this limit to a greater number if the cause for these warnings is known and if it is safe to do so.

Appendix 28.A The following is the partial content of a same MCDOSE/MCSIM geometry input data file for a 4 × 4 × 4 CT phantom. The phantom starts at −2 cm and ends at 2 cm in x- and y-directions, and starts at 0 and ends at 4 cm along z. This defines 64 1 cm voxels. There are two media, air and tissue, and three contours, external contour (skin), the GTV and the lung. 2 Air Tissue 0.1, 0.1 4, 4, 4 −2.., −1., 0., 1., 2. −2.., −1., 0., 1., 2. 0.., 1., 2., 3., 4. 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1… 0.001, 0.001, 0.2, 0.3, 0.8, 1., 1., 1., 1., … 0, 0, 1, 1, 1, 3, 5, 5, 7, 7, 7, 5, 3, 3, 3, 1, 1, 1, …

Number of media Name for medium 1, MED(1) Name for medium 2, MED(2) ESTEPE values for the two media Nx, Ny, Nz x boundaries x(i), Nx + 1 values y boundaries y(j), Ny + 1 values z boundaries z(k), Nz + 1 values Medium for each voxel, MED(i), Nx × Ny × Nz values Density for each voxel, RHO(i), Nx × Ny × Nz values Contour specifier, IREGFLAG(i), Nx × Ny × Nz values

1. Skin external contour 2. GTV structure name 3. Lung structure name

Appendix 28.B The following is the content of a sample 3D dose data file for a 4 × 4 × 4 phantom. The phantom starts at −2 cm and ends at 2 cm in x- and y-directions, and starts at 0 and ends at 4 cm along z. This defines 64 1 cm voxels.

Applications to Patient-Speciﬁc Voxel Computational Phantoms

4, 4, 4 −2., −1., 0., 1., 2. −2., −1., 0., 1., 2. 0., 1., 2., 3., 4. 1., 2., 2., 1., 2., 8., 8., 2., 2., 8., 1., 4., … .1, .1, .1, .1, .1, .1, .2, .1, .2, .1, .1, .2, …

651

Nx, Ny, Nz x boundaries x(i), Nx + 1 values y boundaries y(j), Ny + 1 values z boundaries z(k), Nz + 1 values Dose D(i,j,k), Nx˙Ny˙Nz values Uncertainty δD(i,j,k)/D(i,j,k), Nx ∙ Ny ∙ Nz values

References 1. Butcher, J.C. and Messel, H. Electron number distribution in electron-photon showers, Physical Review, 112, 2096, 1958. 2. Butcher, J.C. and Messel, H. Electron number distribution in electron-photon showers in air and aluminium absorbers, Nuclear Physics, 20, 15, 1960. 3. Varfolomeev, A.A. and Svetlolobov, I.A. Monte-Carlo calculations of electromagnetic cascades with account of the influence of the medium on bremsstrahlung, Soviet Physics JETP-USSR, 9, 1263, 1959. 4. Zerby, C.D. and Moran, H.S. Studies of longitudinal development of electron-photon cascade showers, Journal of Applied Physics, 34, 2445, 1963. 5. Zerby, D. and Moran, H.S. Studies of the Longitudinal Development of High-Energy Electron Photon Cascade Showers in Copper, Report ORNL-3329 (Oak Ridge, TN: Oak Ridge National Laboratory), 1962. 6. Alsmiller, R.G., Barish, J., and Dodge, S.R. Energy deposition by high-energy electrons (50 to 200 Mev) in water, Nuclear Instruments & Methods, 121, 161, 1974. 7. Alsmiller, R.G. and Moran, H.S. Electron-Photon Cascade Calculations and neutron Yields from Electrons in Thick Targets, Report ORNL-TM-1502 (Oak Ridge, TN: Oak Ridge National Laboratory), 1966. 8. Alsmiller, R.G. and Moran, H.S. Calculation of energy deposited in thick targets by high energy (1-Gev) electron-photon cascades and comparison with experiment, Nuclear Science and Engineering, 38, 131, 1969. 9. Nagel, H.H. Die Berechnung von Elektron-Photon-Kaskaden in Blei mit Hilfe der Monte-Carlo Methode, Inaugural-Dissertation zur Erlangung des Doktorgrades der Hohen MathematichNaturwissenschaftlichen Fakultät der Rheinischen Freidrich-Wilhelms-Universität, Bonn, 1964. 10. Nagel, H.H. and Schlier, C. Berechnung Von Elektron-Photon-Kaskaden in Blei Fur Eine Primarenergie Von 200 Mev, Zeitschrift Fur Physik, 174, 464, 1963. 11. Ford, R.L. and Nelson, W.R. The EGS Code System: Computer Programs for the Monte Carlo Simulation of Electromagnetic Cascade Showers (Version 3), Report SLAC-210, Stanford Linear Accelerator Center, 1978. 12. Rogers, D.W.O. Low-energy electron-transport with Egs, Nuclear Instruments & Methods in Physics Research Section a-Accelerators Spectrometers Detectors and Associated Equipment, 227, 535, 1984. 13. Nelson, W.R., Hirayama, H., and Rogers, D.W.O. The EGS4 Code System, SLAC-265 (Stanford, CA: Stanford Linear Accelerator Centre, Stanford University), 1985. 14. Bielajew, A.F. and Rogers, D.W.O. Presta—The parameter reduced electron-step transport algorithm for electron Monte-Carlo transport, Nuclear Instruments & Methods in Physics Research Section B-Beam Interactions with Materials and Atoms, 18, 165, 1987. 15. Hirayama, H. and Namito, Y. Implementation of a General Treatment of Photoelectric-Related Phenomena for Compounds or Mixtures in EGS4, KEK Internal 2000–3 (Japan: High Energy Accelerator Research Organization), 2000.

652

Handbook of Anatomical Models for Radiation Dosimetry

16. Hirayama, H. and Namito, Y. Implementation of a General Treatment of Photoelectric-Related Phenomena for Compounds or Mixtures in EGS4 (Revised Version), KEK Internal 2004–6 (Japan: High Energy Accelerator Research Organization), 2004. 17. Verhaegen, F. et al. Monte Carlo modelling of radiotherapy kV x-ray units, Physics in Medicine and Biology, 44, 1767, 1999. 18. Rogers, D.W.O. and Bielajew, A.F. Monte Carlo techniques of electron and photon transport for radiation dosimetry, in The Dosimetry of Ionizing Radiation, 3rd vol., Kase, K.R., Bj∙∙arngard, B.E., Attix, F.H., eds. (New York: Academic Press), 1990, pp. 427. 19. Bielajew, F., Mohan, R., and Chui, C.S. Improved Bremsstrahlung Photon Angular Sampling in the EGS4 Code System, Report PIRS-0203, National Research Council of Canada, 1989. 20. Duane, S., Bielajew, A.F., and Rogers, D.W.O. Use of ICRU-37/NBS Collision Stopping Powers in the EGS4 System, Report PIRS-0173, National Research Council of Canada, 1989. 21. ICRU. Stopping Powers for Electrons and Positrons, ICRU Report 37, ICRU, Washington, DC, 1984. 22. Ma, C.M. and Nahum, A.E. A new algorithm for Egs4 low-energy electron-transport to account for the change in discrete interaction cross-section with energy, Nuclear Instruments & Methods in Physics Research Section B-Beam Interactions with Materials and Atoms, 72, 319, 1992. 23. Andreo, P. Monte-Carlo techniques in medical radiation physics, Physics in Medicine and Biology, 36, 861, 1991. 24. Rogers, D.W.O. How accurately can Egs4/Presta calculate ion-chamber response, Medical Physics, 20, 319, 1993. 25. Ma, C.M. et al. Accurate characterization of Monte Carlo calculated electron beams for radiotherapy, Medical Physics, 24, 401, 1997. 26. Ma, C.M. and Jiang, S.B. Monte Carlo modelling of electron beams from medical accelerators, Physics in Medicine and Biology, 44, R157, 1999. 27. Rogers, D.W.O. et al. Beam—A Monte-Carlo code to simulate radiotherapy treatment units, Medical Physics, 22, 503, 1995. 28. Verhaegen, F. and Seuntjens, J. Monte Carlo modelling of external radiotherapy photon beams, Physics in Medicine and Biology, 48, R107, 2003. 29. Chetty, I.J. et al. Report of the AAPM task group No. 105: Issues associated with clinical implementation of Monte Carlo-based photon and electron external beam treatment planning, Medical Physics, 34, 4818, 2007. 30. Ma, C.-M. et al. Clinical implementation of a Monte Carlo treatment planning system, Medical Physics, 26, 2133, 1999. 31. Ma, C.M. et al. Monte Carlo verification of IMRT dose distributions from a commercial treatment planning optimization system, Physics in Medicine and Biology, 45, 2483, 2000. 32. Mackie, T.R. Applications of the Monte Carlo method in radiotherapy, in Dosimetry of Ionizing Radiation, 3rd vol., Kase, K., Bjarngard, B., Attix, F.H., eds. (New York: Academic Press), 1990, pp. 541. 33. Pawlicki, T. and Ma, C.M. Monte Carlo simulation for MLC-based intensity-modulated radiotherapy, Medical Dosimetry, 26, 157–168, 2001. 34. Reynaert, N. et al. Monte Carlo treatment planning for photon and electron beams, Radiation Physics and Chemistry, 76, 643, 2007. 35. Kawrakow, I. Accurate condensed history Monte Carlo simulation of electron transport: I. EGSnrc, the new EGS4 version, Medical Physics, 27, 485, 2000. 36. Kawrakow, I. and Rogers, D.W.O. The EGSnrc code system: Monte Carlo simulation of electron and photon transport, NRC Report PIRS-701, 2006. 37. Herayama, H. et al. The EGS5 Code System, SLAC-R-730 (Stanford, CA: Linear Accelerator Center), 2005. 38. Bielajew, A.F. and Wilderman, S.J. Innovative electron transport methods in EGS5, in Proceedings of the Second International Workshop on EGS4 (Japan: KEK), 2000. 39. Kawrakow, I. and Bielajew, A.F. On the condensed history technique for electron transport, Nuclear Instruments & Methods in Physics Research Section B-Beam Interactions with Materials and Atoms, 142, 253, 1998.

Applications to Patient-Speciﬁc Voxel Computational Phantoms

653

40. Cook, A.J. Mortran3 user’s guide, Technical Memorandum CGTM 209, SLAC Computation Research Group, 1983. 41. Sternhei, R.M. and Peierls, R.F. General expression for density effect for ionization loss of charged particles, Physical Review B, 3, 3681, 1971. 42. Ma, C.M. et al. DOSXYZ Users Manual, NRC Report PIRS 746, 1995. 43. Rogers, D.W.O. et al. BEAM Users Manual, NRC Report PIRS 509a, 1995. 44. Ma, C.M. and Rogers, D.W.O. Beam characterization: A multiple-source model, NRC Report PIRS 509d, 1995. 45. Bielajew, A.F. and Rogers, D.W.O. A Standard timing benchmark for Egs4 Monte-Carlo calculations, Medical Physics, 19, 303, 1992. 46. Ma, C.M., Rogers, D.W.O., and Walters, B. DOSXYZnrc Users Manual, NRC Report PIRS 509b(revF), 2001. 47. Walters, B.R.B. and Rogers, D.W.O. DOSXYZnrc Users Manual, NRC Report PIRS 746, 2002. 48. Ma, C.M. et al. MCSIM: A Monte Carlo dose calculation tool for radiation therapy, in Proceedings of the 14th International Conference on the Use of Computer in Radiation Therapy (ICCR), Seoul, 2004. 49. Ma, C.-M. et al. A Monte Carlo dose calculation tool for radiation therapy treatment planning, Physics in Medicine and Biology, 47, 1671, 2002. 50. Li, J.S. et al. Validation of a Monte Carlo dose calculation tool for radiotherapy treatment planning, Physics in Medicine and Biology, 45, 2969, 2000. 51. Berger, M.J. Methods in Computational Physics, vol. 1, Fernbach, S., Alder, B., Rothenberg, M., eds. (New York: Academic Press), 1963. 52. Ma, C.M. and Nahum, A.E. Effect of size and composition of the central electrode on the response of cylindrical ionization chambers in high-energy photon and electron-beams, Physics in Medicine and Biology, 38, 267, 1993. 53. ICRU. Tissue Substitutes in Radiation Dosimetry and Measurement, ICRU Report 44, Bethesda, MD, 1989. 54. Siebers, J.V. et al. Converting absorbed dose to medium to absorbed dose to water for Monte Carlo based photon beam dose calculations, Physics in Medicine and Biology, 45, 983, 2000. 55. Li, J.S. et al. MCBEAM—A Monte Carlo tool for beam simulation, in Proceedings of the 15th International Conference on the Use of Computer in Radiation Therapy (ICCR), Oakville, 2007. 56. Chen, Z. et al. Effect of voxel size on Monte Carlo dose calculation for intensity modulated radiotherapy treatment planning, Medical Physics, 33, 2095, 2006. 57. Mora, G. et al. Effect of voxel size on Monte Carlo dose calculation for IMRT treatment planning, Medical Physics, 30, 1451, 2003. 58. Qin, L. et al. Effect of voxel size on Monte Carlo dose calculations, Medical Physics, 31, 1883, 2004.

29 Applications to Nonionizing Radiation Protection Ji Chen, Wolfgang Kainz, and Dagang Wu

CONTENTS 29.1 Induced Current within Pregnant Woman Computational Phantoms due to Walk-through Metal Detector Emissions ............................... 655 29.1.1 Method .......................................................................................................... 656 29.1.1.1 Measurement of Magnetic Field Distribution .......................... 656 29.1.1.2 Calculation of the Equivalent Current Source ......................... 658 29.1.1.3 The Impedance Method .............................................................. 659 29.1.2 Nine-Month Pregnant Woman Computational Phantoms .................... 662 29.1.3 Results and Discussions .............................................................................664 29.2 SAR and Temperature within Pregnant Woman Computational Phantoms due to Magnetic Resonance Imaging (MRI) Radiofrequency (RF) Transmitting Coil ................................................................ 668 29.2.1 Method .......................................................................................................... 668 29.2.2 Birdcage Coil Computational Phantom .................................................... 669 29.2.3 Results and Discussions ............................................................................. 671 29.2.3.1 Specific Absorption Rate and Temperature Rise for 64 MHz Birdcage Coil.................................................... 671 29.2.3.2 Specific Absorption Rate and Temperature 128 MHz Birdcage Coil ................................................................ 673 29.3 Conclusions ................................................................................................................ 676 References ............................................................................................................................. 676

29.1 Induced Current within Pregnant Woman Computational Phantoms due to Walk-through Metal Detector Emissions Walk-through metal detectors (WTMDs) are integral parts of the security surveillance systems found in airports and government buildings. Most of these metal detectors use electromagnetic signal variations to detect metallic objects. A pair or pairs of transmitter/ receiver coils is embedded in the frames of the WTMDs. A metal object between the transmitter/receiver pairs varies the electromagnetic signals, and if the signal variation exceeds a certain threshold, a metal object is detected. In principle, a stronger electromagnetic emission will lead to a higher signal-to-noise ratio and consequently produce more accurate results in metal detection. However, the electromagnetic emission used to detect metal objects also causes undesired electromagnetic energy depositions within humans. 655

656

Handbook of Anatomical Models for Radiation Dosimetry

To ensure public safety, exposure safety reference levels and basic restrictions have been developed by the International Commission on Nonionizing Radiation Protection (ICNIRP)1,2 and IEEE C95.6.3 These guidelines define the maximum energy deposition within humans at different exposure frequencies. WTMD emissions are typically low frequencies (i.e., below 10 kHz). At these frequencies, the induced current density and electric field are used as basic restrictions for exposure safety assessments.2–4 Both ICNIRP and IEEE provide, in addition to the basic restrictions, reference levels or maximum permissible exposure (MPE) levels. In the following sections we will use only MPE when referring to reference levels or MPEs. The MPEs were derived from the basic restrictions using simple human anatomy based on adult males. An exposure situation below the MPEs is expected also to meet the basic restrictions. If this is true for all human beings and for any exposure situations, then the safety standard can be considered consistent. With more anatomical computational phantoms available in the future, new dosimetric studies will show if the ICNIRP guidelines, the IEEE Standards C95.6, and the C95.1-20054 are indeed consistent exposure safety standards. Adult males make poor computational phantoms for pregnant women whose fetuses are surrounded by amniotic fluid, which is highly conductive. In a body region close to this kind of liquid, we can expect higher induced currents than in an adult male computational phantom. This section describes the application of using a numerical method to perform accurate safety assessment for WTMDs. Using the method developed in this study we assessed the exposure of pregnant woman computational phantoms to one particular WTMD. The procedure consists of three steps: (1) the measurement of the magnetic field distribution emitted by WTMD; (2) the extraction of equivalent current source based on the measured magnetic field distribution; and (3) the use of the equivalent current source to calculate the induced fields within pregnant woman computational phantoms using the impedance method. In the following paragraphs, we first describe the methodology applied and then present the results of our dosimetric study for one particular WTMD. Then we apply the results of this dosimetric study to evaluate the consistency of the ICNIRP guidelines and the IEEE Standard C95.6. 29.1.1 Method Since it is difficult to perform direct induced current measurement within human computational phantoms, we employee numerical simulations to perform the evaluation of induced currents within human computational phantoms. One efficient technique for such low-frequency application is the impedance method. Using the impedance method to calculate the induced fields within human computational phantoms requires the knowledge of magnetic field emissions. Due to the complex operating modes of WTMDs, the current distribution inside the WTMD pylons, and the magnetic field emission into the surrounding environment, are not directly available. The first step in calculating the induced fields within human computational phantoms will be to develop an equivalent current source by measuring the magnetic field distribution in the absence of the human subject. For the particular WTMD used in this study, we examined the worst-case conditions to the dosimetric study by choosing a WTMD operating mode which maximizes the emitted magnetic field strength. 29.1.1.1 Measurement of Magnetic Field Distribution A typical WTMD consists of two pylons, as shown in Figure 29.1. Depending on different types of WTMD, pulse or continuous wave signals are emitted from the pylons. For the

Applications to Nonionizing Radiation Protection

657

Walk-through metal detector x

R

Measurement planes r

r΄ z

Measurement probe y Probe holder

Reference probe Point of origin FIGURE 29.1 Configuration of the measurement system and the reference coordinate system used in this study. r refers to measurement location and the r¢ refers to the source location. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

particular WTMD used in this study, we used a pulsed system with the main frequency component at 940 Hz. The bandwidth of the signal is around 3000 Hz with most of the energy located below 1000 Hz. In later simulations, an equivalent magnitude for 940 Hz signal was used. This equivalent magnitude is calculated by equating the energy of the sinusoidal signal to that of the pulsed signal. To measure the magnetic field distribution between the WTMD pylons, we used an automatic three-axis measurement system. The system consists of three-axis magnetic field probes including a combination of the following instruments (1) Wandel Goltermann (Eningen, Germany) Computational phantom EFA-2 Magnetic Field Analyzer, (2) Holaday Industries (Eden Park, MN) Computational phantom 3627 ELF Magnetic Field Meter, (3) Holaday Industries Computational phantom 3637 VLF Magnetic Field Meter, (4) Holaday Industries HI-4433 LFH Broadband Isotropic Field Probe, and (5) Electric Research Management—ERM (State College, PA) Computational phantom 1678.002 Magnetic Field Sensor (with a sensor volume of 2 × 2 × 2 cm). For the whole system, the minimum sensitivity for magnetic field strength measurement was about 1.0 A/m or 1.26 μT. This measurement system can simultaneously measure all three components of the magnetic field strength. A personal computer with customized software controls those three motors to move magnetic field probes automatically to any desired position. The dimension of each magnetic probe is about 1 cm in all directions. This system can measure magnetic field strengths in four channels; three channels are used to collect magnetic field strength in the x-, y-, and z-directions, and the fourth channel is used to determine the relative phase information for the three channels. This fourth channel is referred to as the reference channel, and is connected to the reference probe, as shown in Figure 29.1. The reference probe needs to be placed at a position where a strong and stable magnetic field value can be measured. Such a location can typically be found on one of the pylons. The dimension of the WTMD used here is 70 cm in width, 50 cm in depth, and 220 cm in height.

Handbook of Anatomical Models for Radiation Dosimetry

100

40

50

20

Hz (A/m)

Hx (A/m)

658

0

–20

–50

–100 2

(a)

0

z (m 1 )

0 –0.5

0

x

0.5 ) (m

–40 2

1

z (m 1 )

(b)

0 –0.5

0

0.5

1

x (m)

FIGURE 29.2 Measured magnetic field strength in a plane that is 5 cm away from the WTMD pylon. (a) x-Component and (b) z-component. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

We measured the magnetic field distribution in four vertical planes (xz planes as illustrated in Figure 29.1), located at 5, 10, 15, and 20 cm parallel to the pylons. Each plane has a size of 120 cm horizontally and 180 cm vertically with a measurement resolution of 5 cm in both directions. Waveforms from each channel and the peak-to-peak signal strengths for x-, y-, and z-directions were recorded. The relative phase information was obtained by correlating the reference probe signal with the x, y, and z signals. Figure 29.2 shows the magnetic field distributions (x- and z-components) in the plane that is 5 cm away from the source pylon. 29.1.1.2 Calculation of the Equivalent Current Source Knowing the magnetic field distributions at four vertical planes, we now can calculate an equivalent current source. This equivalent source would not be the exact coil configuration embedded in the pylon, but would be able to produce the same magnetic field distribution as the actual coil configuration.5 Since the wavelength of the electromagnetic fields emitted from the WTMD is larger than the physical dimensions of the WTMDs, quasistatic approximations can be used. As a consequence, we use the Bio-Savart law and the leastsquare method to extract the equivalent current source based on the measured magnetic field distribution. To calculate the equivalent current distribution on the pylons, i.e., the equivalent current source, we first divide one side of the pylon into a two-dimensional (2D) grid. At each grid location, electric currents in both vertical (x) and horizontal (z) directions are assumed, T as shown in Figure 29.3. Denote J x = [J x _ 1 , J x _ 2 ,..., J x _ l ] as the currents in x-direction, T

J y = ⎡⎣ J y _ 1 , J y _ 2 ,..., J y _ m ⎤⎦ as the currents in y-direction (which are zero for our coordinate T system as shown in Figure 29.3), and J z = [J z _ 1 , J z _ 2 ,..., J z _ n ] as the currents in z-direction. The magnetic fields strength generated by these currents are given by ⎡ Hx ⎤ ⎡ 0 ⎢ ⎥ ⎢ H = ⎢ H y ⎥ = ⎢m yx ⎢⎣ H z ⎥⎦ ⎢⎣ m zx

m xy 0 m zy

m xz ⎤ ⎡ J x ⎤ ⎥⎢ ⎥ m yz ⎥ ⎢ J y ⎥ , 0 ⎥⎦ ⎢⎣ J z ⎥⎦

(29.1)

Applications to Nonionizing Radiation Protection

659

R

r r΄ x z y

FIGURE 29.3 Illustration of the discretized representation of the equivalent current and the reference coordinate system used in this study. r refers to measurement location and the r¢ refers to the source location. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

m xy is a matrix that relates the magnetic field strengths in x-direction generated by the currents flowing in y-directions following Biot–Savart’s law. The dimension of this matrix depends on the number of magnetic fields measured, as well as the number of equivalent current source points. The exact value of this term can be found.6 Equation 29.1 can also be written into H = M × J; where the left side of the equation H corresponds the measured magnetic fields distribution. The right-hand side term J is the unknown equivalent current source. Since the number of the measurement points may not be the exact number of the equivalent source points, the matrix itself may not be a square matrix. To determine the equivalent current source which best matches the measured current distribution, we use the least-square method to minimize the value of |H − M × J|. Figure 29.4 shows the extracted equivalent source-generated magnetic field strength distribution as well as its relative error (in absolute value) to the measured magnetic field strength in the plane that is 5 cm away (the same plane as the measured magnetic field in Figure 29.2). Figure 29.4 demonstrates that the magnetic field distribution generated by the equivalent current source is very close to that from the actual measurement. 29.1.1.3 The Impedance Method Knowing the equivalent current source, we now can perform the dosimetric assessments for the particular WTMDs using the impedance method.7–10 The impedance method represents the anatomical computational phantom by a 3D impedance network as shown in Figure 29.5. Each cell has a shape of a parallelepiped. We assign an impedance element, which connects to two nodes, at each edge. The value of each impedance element is related to the electrical property of the four cells surrounding the impedance element. For example, the impedance value for the x-directional impedance Zx is given by

Handbook of Anatomical Models for Radiation Dosimetry

100

20

50

0 –20 –40 2

(a)

Relative error of Hx (%)

Hx (A/m)

40

0 –50

–100 2

1.5 1 z (m 0.5 )

0 –0.5

0

z (m 1 0.5 )

5

0 2

1

m

1 0 –0.5

0

0 0 –0.5

x (m)

(c)

)

x (m

15 10 5 0 2

1

z(

0.5

)

0.5

(b)

10

z(

1

1.5

1 0.5 x (m)

Relative error of Hz (%)

Hz (A/m)

660

m)

1

0.5 0

0 –0.5

x (m)

(d)

FIGURE 29.4 Equivalent source generated magnetic field distributions and their relative errors to the original measured magnetic field in a plane that is 5 cm away from the WTMD pylon. (a) x-Component magnetic field distribution; (b) z-component magnetic field distribution; (c) relative error (in absolute value) of the x-Component; and (d) relative error (in absolute value) in the z-component. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

(σ, ε)

Ix

Iy z

y

Zz

Zy Iz

x Zx

FIGURE 29.5 Illustration of the impedance network in the impedance method. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

Applications to Nonionizing Radiation Protection

Zx =

661

Δx , Δy Δz(σ + jωε)

(29.2)

where σ and ε are the conductivity and permittivity along the edge Δx, Δy, Δz are the size of the voxels in the x-, y-, and z-directions Similar expressions can be derived for all y- and z-directional impedances. We apply Kirchhoff’s voltage equation, as illustrated in the Figure 29.5, to each closed loop, and then relate the excitation voltage to the electromotive force from the external magnetic fields by emf = −

∂ ∂t

∫∫

μH × d s

(29.3)

The s in Equation 29.3 is the surface area of each closed loop. The direction of the s lies perpendicular to the surface. For example, each loop in the yz plane provides an area for each closed loop, and is represented by Δy × Δz. The direction for s is in the x-direction. Applying Kirchhoff voltage equations to each loop, we derive the following system of equations: 3

∑a

i , j ,k mn

x n (i , j , k ) = emfm ; 1 ≤ m ≤ 3,

(29.4)

n =1

where x1,2,3 stands for the loop current in x-, y-, and z-direction i, j,k a mn is the coefficient from Kirchhoff’s voltage equations We can solve this system of equations using the successive over-relaxation method. With the obtained loop currents from the impedance method, the induced current densities and electric fields along x-, y-, and z-directions within the biological computational phantom can be obtained. The resultant total current densities and electric field strengths can be then calculated by the following equations: J=

J x2 + J y2 + J z2 ,

(29.5)

E = Ex2 + Ey2 + Ez2 .

(29.6)

According to the ICNIRP guidelines, basic restrictions on the computational phantom are defined by the current density calculated as an average value over an area of 1 cm2, whereas the IEEE C95.6 standard averages the electric field strength over a straight line segment of 5 mm length oriented in any direction. During the postprocessing we evaluated the 1 cm2 averaged current density and the 5 mm averaged electric field using an interpolation technique described.10,11 In our simulations we used a grid size of 2.5 mm in all directions. The applied interpolation technique is shown in Figure 29.6a and b. We calculate the corresponding 1 cm2 averaged current density as the average of the current density at the four corners of a 1 cm2 square area that is perpendicular to the direction of current flow:

Handbook of Anatomical Models for Radiation Dosimetry

662

2.5 mm

E8

E7

C

B E5

E6

2.5 mm

E3 1

E4

E1

A

E2

1 (cm)

5 mm

(a)

(b)

FIGURE 29.6 (a) Calculation of 1 cm2 averaged current density. (b) Calculation of the 5 mm averaged electric field strength. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

J avg =

1 (J1 + J A + J B + J C ), 4

(29.7)

where JI, JA, JB, and JC are the resultant current densities whose direction is perpendicular to the surface at node 1, A, B, and C, respectively. It should be noted that there are other techniques to calculate the 1 cm 2 averaged current density. For example, one can average the current density at all 25 grid locations in the Figure 29.6a. In this case, we followed the technique descried.10,12 This technique is valid as long as the direction of the current density at the four corners of the 1 cm 2 square varies insignificantly, which is the case in our simulation. To find the average electric field strength accordingly to the IEEE C95.6 standard, we first calculate the values of electric field components in the center of a 2.5 mm cube, based on the electric field components on the cube edges. Second, we calculate the averaged x-component of the electric field strength for a 5 mm cube. The technique is shown in Figure 29.6b and the averaged electric field x-component is calculated using Exavg =

1⎛ ⎜ 8⎝

8

∑E

n, x

n=1

⎞ ⎟. ⎠

(29.8)

A similar procedure can be used to evaluate the averaged electric field components in the other two directions. Once all three components of the electric field strength are obtained in the center of the 5 mm cube, the resultant 5 mm averaged induced electric field strength is calculated using Equation 29.6. 29.1.2 Nine-Month Pregnant Woman Computational Phantoms The 9-month pregnant woman computer-aided design (CAD) computational phantoms used in this study were codeveloped by the Food Drug Administration (FDA) and the University of Houston.13 Nine pregnant woman computational phantoms, representing each month of pregnant stage, are shown in Figure 29.7. Two body computational phantoms

Applications to Nonionizing Radiation Protection

663

FIGURE 29.7 (See color insert following page 524.) Nine-month pregnant woman models. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

and one fetus computational phantom were used in the computational phantom development. For the first four computational phantoms (month 1–4), a regular female body computational phantom was used while for the next five computational phantoms (month 5–9), a pregnant female body computational phantom was used. The height of those body computational phantoms is about 90 cm. Both female body computational phantoms are in a standing position. These computational phantoms include the body, the placenta, the embryonic fluid, the bladder, the bone, the fetus, and the uterus. In our simulations, these nine woman computational phantoms are discretized with a resolution of 2.5 mm in all three directions. The electrical properties of the tissues are assigned according to frequency-dependent material database13–16 and listed for 940 Hz in Table 29.1.

TABLE 29.1 Dielectric Properties of Various Tissues at 940 Hz Conductivity (S/m) Bladder Body Bone Fetus Liquid Placenta Uterus

2.07E − 01 2.30E − 01 2.02E − 02 1.86E − 01 1.50E + 00 7.00E − 01 4.90E − 01

Relative Permittivity 5.40E + 04 3.47E + 05 2.83E + 03 2.78E + 05 9.90E + 01 5.26E + 03 1.06E + 06

Handbook of Anatomical Models for Radiation Dosimetry

664

29.1.3 Results and Discussions Using the equivalent current source, we can evaluate the induced current densities within the nine pregnant woman computational phantoms using the impedance method. For the WTMD considered in this study, only one pylon emits magnetic signals, and therefore, the equivalent current source was calculated only for one pylon. Figure 29.8 shows the current density in x- and z-directions of the calculated equivalent current source. The pregnant woman computational phantoms are always placed at 5.5 cm away from the equivalent current source. Since our computational phantoms are truncated, the bottom of the computational phantoms are placed 50 cm above the ground. The configuration for the simulation setup is shown in Figure 29.9. The displacement of fetus region is also summarized in Table 29.2. The calculated layer-averaged current densities of each month are depicted in Figure 29.10. As indicated in the figure, the layer-averaged current densities surrounding the fetus region increase as the stage of pregnancy increases. This could be attributed to the increased volume of conducting fluid around the fetus. In general, it is noticed that at late stages of the pregnancy, the induced current densities could be twice as high as those for early stages. We compared results for all nine computational phantoms with two international safety standards: the ICNIRP guidelines and IEEE C95.6 safety standards. Both standards define the exposure reference levels (in terms of magnetic flux density) for the MPEs and the basic restrictions.1–3 Since the dominant components of electromagnetic emissions from the WTMD are mainly the magnetic flux density, MPE comparisons were made for this quantity. Table 29.3 summarizes the whole body averaged (ICNIRP requires an MPE average over whole body), the magnetic flux density inside the nine computational phantoms, the ICNIRP reference level, the maximum magnetic flux density inside the computational phantoms, and the MPE reference level given by IEEE C95.6. Since our system measured the magnetic field strength, the equivalent magnetic field strength was also given in the

20 2

15

2

10 10 5

1.5 0

1

z (m)

z (m)

1.5

0 –5

1

–10

–10 0.5

–15

0.5 –20

–20

0

0 –0.5

(a)

0 x (m)

0.5

–25 –0.5

(b)

0 x (m)

0.5

FIGURE 29.8 Extracted equivalent current sources (A) in the pylon. (a) x-Direction equivalent current and (b) z-direction equivalent current. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

Applications to Nonionizing Radiation Protection

665

WTMD pylon

5.5 cm Main direction of the emitted magnetic field

Equivalent current source plane

50 cm

FIGURE 29.9 Simulation setup for pregnant woman models in WTMD. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

TABLE 29.2 Vertical Placement of Fetus Inside WTMD Pylon Fetus Placement (cm)

Month 1 Month 2 Month 3 Month 4 Month 5 Month 6 Month 7 Month 8 Month 9

Bottom

Top

83.125 82.625 82.875 81.125 82.375 82.375 82.125 80.875 80.625

83.625 84.375 87.375 91.375 98.875 101.125 104.875 107.875 107.875

table. As indicated in Table 29.3, the magnetic flux densities inside these pregnant woman computational phantoms are found to exceed the ICNIRP reference level but are well below the IEEE C95.6 reference level. Therefore the ICNIRP reference level is more conservative at this particular operating frequency (940 Hz).

Handbook of Anatomical Models for Radiation Dosimetry

666

1.5 1.4 1.3

Height (m)

1.2 1.1 1 0.9

Month 1 Month 2 Month 3 Month 4 Month 5 Month 6 Month 7 Month 8 Month 9

0.8 0.7 0.6 0.5 0.1

0.2

0.3

0.4

0.6

0.5

Current density (mA/m

0.7

0.8

2)

FIGURE 29.10 Layer-averaged current density along the vertical direction. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

TABLE 29.3 Calculated, Averaged, and Maximum Magnetic Flux Density B inside the Nine Pregnant Woman Models Compared to ICNIRP and IEEE C95.6 Reference Levels

Month 1 Month 2 Month 3 Month 4 Month 5 Month 6 Month 7 Month 8 Month 9

Whole-Body Averaged B/H, Bavg(mT)/H (A/m)

ICNIRP Reference Level, B (mT)/H (A/m)

Maximum B/H, Bmax(mT)/H (A/m)

IEEE C95.6 MPE, B (mT)/H (A/m)

9.92/7.89 9.92/7.89 9.92/7.89 9.92/7.89 9.13/7.27 9.07/7.22 9.02/7.18 9.02/7.18 9.02/7.18

6.25/4.97 6.25/4.97 6.25/4.97 6.25/4.97 6.25/4.97 6.25/4.97 6.25/4.97 6.25/4.97 6.25/4.97

29.4/23.4 29.4/23.4 29.4/23.4 29.4/23.4 28.3/22.5 28.0/22.3 28.0/22.3 28.3/22.5 28.3/22.5

730/581 730/581 730/581 730/581 730/581 730/581 730/581 730/581 730/581

Since these pregnant woman computational phantoms are exposed to an external magnetic field strength that exceeds the maximum permissible reference level according to ICNIRP, the induced current densities within these pregnant woman computational phantoms shall be compared to the basic restriction. To verify the consistency of ICNIRP and C95.6 we also calculate the electric field strength within these computational phantoms and compare them against the IEEE C95.6 basic restrictions. The IEEE C95.6 gives different

Applications to Nonionizing Radiation Protection

667

basic restriction values for brain and nonbrain tissues. We compare the calculated electric field strength within the fetus against the basic restriction value for brain tissue, and the electric field strength in other regions against that for nonbrain tissues. At 940 Hz the ICNIRP basic restriction for induced current densities is 2 mA/m2 while the IEEE C95.6 electric field strength basic restriction is 277 mV/m for brain tissue and 701 mV/m for other tissues. Table 29.4 summarizes the calculated induced current densities and electric field strength within the nine pregnant woman computational phantoms against these two standards. As indicated in Table 29.4, the values of electric field strength within the nine pregnant woman computational phantoms are all below the IEEE C95.6 basic restrictions. However, the induced current densities within the computational phantoms can exceed the ICNIRP basic restrictions. In particular, starting from month 6, the maximum induced current density exceeds the ICNIRP restrictions of 2 mA/m2. From our results, we observe that the basic restriction for the ICNIRP is more conservative than that of the IEEE C95.6 for our applications. In all cases the exposures exceed the MPE value of 6.25 μT (or 4.97 A/m), and in some cases fail to comply with the basic restriction. Therefore ICNIRP and C95.6 can be considered consistent safety standards for these particular computational phantoms and exposure scenarios. The basic restrictions and MPE values stated in ICNIRP and C95.6 show that ICNIRP is more conservative than C95.6 for this particular application. The particular WTMD evaluated in this study passes the C95.6 MPEs and the basic restrictions, but fails to comply with the ICNIRP reference levels for all nine computational phantoms. Comparing the induced current density to ICNIRP’s basic restrictions shows that this particular WTMD complies with the ICNIRP basic restrictions for month 1–5 computational phantoms, but leads to both fetus and pregnant woman overexposure for month 6–9 computational phantoms. It is expected that with different placement of these computational phantoms, the induced current density within these computational phantoms could be even higher. The deployment of security systems, such as WTMDs and electronic article surveillance systems, and an increase in emissions levels to raise the detection sensitivity, makes it

TABLE 29.4 Calculated Maximum 1 cm2 Current Density (J) and Electric Field Strength (E) for the Nine Pregnant Woman Models Exposed to the WTMD System Calculated Jmax (mA/m2) (Fetus) Month 1 Month 2 Month 3 Month 4 Month 5 Month 6 Month 7 Month 8 Month 9

0.31 0.64 1.14 1.16 1.90 2.35 2.79 3.09 3.06

IEEE C95 ICNIRP

Jmax (mA/m2) (Other Tissue)

E (mV/m) (Fetus)

E (mV/m) (Other Tissue)

Jmax (mA/m2)

E (mV/m) (Fetus)

E (mV/m) (Other Tissue)

1.22 1.22 1.35 1.57 2.22 2.86 3.34 3.64 3.69

1.65 3.42 6.10 6.22 10.23 12.65 15.01 16.61 16.45

27 26 26 27 30 32 34 33 34

2 2 2 2 2 2 2 2 2

277 277 277 277 277 277 277 277 277

701 701 701 701 701 701 701 701 701

Handbook of Anatomical Models for Radiation Dosimetry

668

necessary to carefully evaluate the exposure of humans inside, or close to, such systems. Before being deployed, every new security system should be carefully assessed to avoid overexposure on people. If the MPE are exceeded, the equivalent source can be used to calculate internal fields within a variety of anatomical computational phantoms and compare them to related basic restrictions. The most conservative exposure restrictions should be applied in cases where pregnant women or children could be exposed to a security system.

29.2 SAR and Temperature within Pregnant Woman Computational Phantoms due to Magnetic Resonance Imaging (MRI) Radiofrequency (RF) Transmitting Coil MRI exams are widely used in the medical field to obtain topographic images of patients’ internal structures. This procedure can produce images with a resolution of a millimeter. Physicians use these images to diagnose defects in ligaments, muscles, brain tissue, and other soft tissues. During the imaging, patients are subjected to several kinds of magnetic fields produced by MRI coils.17 While these magnetic fields are required for the purpose of imaging, they also deposit electromagnetic energy within patients. Due to such energy deposition, temperature rises within patients have been observed.18,19 In this section, we examine the electromagnetic energy emitted by the MRI RF coil only, even though the gradient coils also emit electromagnetic energy at lower frequency. Since typical MRI machines operate at 64 and 128 MHz, the quantitatives used for safety evaluation are the specific absorption rate (SAR) and the temperature rise. We present a detailed numerical study of SAR and temperature rise within pregnant woman computational phantoms exposed to MRI coil radiations in this section. SAR and temperature rises within and around fetuses at different stages were evaluated for two MRI operating modes (normal mode and first-level controlled mode) at 64 and 128 MHz. 29.2.1 Method The simulation methods used in this study are described elsewhere.20 The finite difference time-domain (FDTD) scheme was applied to both Maxwell’s equations and bioheat equation to obtain SAR distributions and temperature rises within our computational phantoms.21–24 SAR, used as a fundamental metric for RF heating, is defi ned as σ|E|2/ρ.25–27 In this expression, σ is the conductivity (S/m), ρ is the mass density (kg/m3), and E is the peak value of electric fields (V/m) within the biological subjects. The strength of the electric field at any location within the biological tissue can be directly obtained from the results of electromagnetic FDTD simulations.22,24 To evaluate the temperature rise within biological tissues, the finite difference technique is applied to the bioheat equation.28 With electric fields obtained from previous electro magnetic simulations, the temperature distribution (T) inside biological tissues can then be evaluated via cρ

∂T 2 = ∇ ⋅ (K ∇T ) + σ E − B0 (T − Tb ) + A0 , ∂t

(29.9)

Applications to Nonionizing Radiation Protection

669

where c is the specific heat capacity (J/kg/°C) K is the thermal conductivity (W/m/°C) B0 is the blood perfusion coefficient (W/m3/°C) A0 is the metabolic heat production rate (W/m3) Tb is the blood temperature At the external surfaces between biological tissue and surrounding air, the following convective boundary condition is applied. K

∂T = − H a (T − Ta ), ∂n

(29.10)

where Ha is the convection coefficient (W/m2/°C) Ta is the ambient temperature n is the normal direction to the surface The simulation procedure starts with using the electromagnetic FDTD method to evaluate the electric/magnetic field within human computational phantoms due to MRI RF coil radiations. Once the electric field distributions are obtained, we can predict the temperature rise using Equations 29.9 and 29.10. It should be noted that a basal temperature distribution within the biological tissue should be evaluated fi rst. This is evaluated by setting the value of electric field equal to zero in Equation 29.9. In the following studies, the temperature of the blood Tb is set at 37°C, and the ambient temperature Ta is assumed to be 24°C. The convection coefficient of 10.5 W/m 2/°C is used in all simulations.21 29.2.2 Birdcage Coil Computational Phantom The RF coil used in this study is the commonly used birdcage coil, which was preliminary designed using the birdcage builder software package developed by Pennsylvania State University.28 Like most birdcage coils, this coil is composed of two end-rings connected by 16 equally spaced legs, as shown in Figure 29.11. The legs and rings of the birdcage coil are constructed using a perfect electric conductor (PEC). Each leg is split, and a capacitor is placed in the gap. The parasitic capacitance between gaps must also be considered in the design, since it also affects the coil resonant frequency. To reduce RF interference, the coil is surrounded by an RF shield, which is modeled as PEC. This shield also affects the tuning of capacitors because the coil would induce eddy currents in the shield, thereby affecting the resonant frequency.29 This preliminary designed structure needs to be tuned in order to achieve exact MRI resonant frequencies. The detailed dimensions for the coil are shown in Figure 29.11. In this study, a quadrature excitation is used for birdcage coils. In the quadrature excitation, two voltage sources are used to excite the 16-leg coil. These two sources need to be on one end-ring, to be spatially separated by 90°, and to be 90° out of phase with the excitation waveform. Theoretically, having such an arrangement would quadruple the effective power of the coil. The resulting magnetic field is highly homogenous, leading to a larger operating region.17

Handbook of Anatomical Models for Radiation Dosimetry

670

Z Y

X

Y

74.3 cm X

Z

67.0 cm

X Z Z

X

Y

Y

FIGURE 29.11 Constructed birdcage coil used in this study. Top left: side view; top right: top view; bottom left: 3D view; and bottom right: coil with shielding. (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.)

Once the excitation scheme is determined, the next step is to tune those capacitors to achieve either 64 or 128 MHz coil resonance. To determine correct capacitor values, we used Gaussian pulses as excitations and recorded time-domain signals at various locations within the coil. The capacitor value was altered until the highest response is detected at 64 or 128 MHz. After this tuning, we can obtain the capacitor value for an empty coil without any computational phantom placed within it (unloaded coil). The frequency would slightly shift if a computational phantom is placed within the coil (loaded coil). Figure 29.12 shows the positioning of the computational phantom in the coil. Each computational phantom had to be placed within the coil consistently, while considering the changes in the size of the computational phantoms. The back of the computational phantom is always placed 22.7 cm away from the inner edge of the coil to simulate the effect of a woman supine on a bed while being scanned. The navel of the computational phantom is always placed on the xy plane, as shown in the figure. This positioning ensured that the computational phantoms would be in the same location for each simulation. The resonant frequency may shift due to the coupling between the RF coil and the loaded computational phantoms once the coil is loaded. This is an interactive process with several iterations necessary before the correct capacitor value was found. The tuning procedure needs to be repeated in order to bring the resonant frequency back to 64 or 128 MHz.

Applications to Nonionizing Radiation Protection

671

Z Y X

Y Z

X

22.7 (cm)

FIGURE 29.12 Positioning of the ninth-month model within the MRI RF coil.20 (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.)

This procedure must be repeated for each of the nine pregnant woman computational phantoms because each computational phantom had a different effect on the coil loading, which would cause various shifting in the resonant frequency. After setting up the simulation for each month of the pregnant computational phantom, and the final capacitor tuning, the simulations were executed in series. Computation time for each month typically takes around 1 day, depending on the exact mesh used in simulations. After the simulations were complete, a postprocessing was used to extract the distributions of temperature rise and energy deposition. 29.2.3 Results and Discussions In this section, the calculated maximum 10 g averaged SAR and the maximum temperature rises within various tissues are given. These values are obtained for the birdcage coils operating at both 64 and 128 MHz. All of these 10 g averaged SARs are normalized against the whole-body averaged energy limit developed by the International Electrotechnical Commission. The whole-body averaged safety limits for energy deposition are 2 W/kg and 4 W/kg for normal mode and first-level controlled mode, respectively. The local 10 g averaged SAR is 10 W/kg for both modes. The temperature rise limits are 0.5°C for the normal mode and 1°C for the first-level controlled mode. 29.2.3.1 Speciﬁc Absorption Rate and Temperature Rise for 64 MHz Birdcage Coil For the normal mode of operation, the 10 g averaged SARs for each tissue at different stages of pregnancy are shown in Figure 29.13. From the figure, we can see that the maximum SARs for fetus are all below the limit of 10 W/kg. However, the temperature rises within some tissues exceed the limit of 0.5°C starting at the fifth month of pregnancy, as depicted in Figure 29.14. The maximum energy deposition was found in the body of the pregnant woman computational phantoms, however, while the maximum temperature rise was found either in the placenta or fluids. It was also noticed that, for later stages of pregnancy, the temperature rises within the fetus could be higher than those in the body tissue.

Handbook of Anatomical Models for Radiation Dosimetry

672

20 18 Normalized maximum specific absorption rate10g (W/kg)

16 Safety limit (normal mode) Bladder Body Bone Fetus Liquid Placenta Uterus

14 12 10 8 6 4 2 0

2

1

3

4

5 Month

6

7

8

9

FIGURE 29.13 Normalized maximum 10 g averaged SAR (64 MHz normal mode).20 (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.)

1.8

Maximum temperature rise (°C)

1.6 1.4 Safety limit (normal mode) Bladder Body Bone Fetus Liquid Placenta Uterus

1.2 1 0.8 0.6 0.4 0.2 0

1

2

3

4

5 Month

6

7

8

9

FIGURE 29.14 Maximum temperature rises (64 MHz normal mode). (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.)

For the first-level controlled mode, the maximum 10 g averaged SARs and maximum temperature rises for each tissue at different pregnant stages are shown in Figures 29.15 and 29.16. Due to increased RF power, the maximum energy depositions within the fetus exceed the limit beyond the fifth month of pregnancy. The temperature rises are also above the limit starting the fourth month. We found that in many tissues, such as the uterus, the placenta, and the liquid, either the SAR or the temperature rise will exceed the limits.

Applications to Nonionizing Radiation Protection

673

40

Normalized maximum specific absorption rate10g (W/kg)

35 30 Safety limit (first level controlled mode) Bladder Body Bone Fetus Liquid Placenta Uterus

25 20 15 10 5 0

2

1

3

4

5 Month

6

7

8

9

FIGURE 29.15 Normalized maximum 10 g averaged SAR (64 MHz first-level controlled mode).21 (Extracted from Wu, D.G. et al., IEEE Trans. Microw., Theory Tech., 54, 4472, 2006.)

3.5

Maximum temperature rise (°C)

3 2.5

Safety limit (first level controlled mode) Bladder Body Bone Fetus Liquid Placenta Uterus

2 1.5 1 0.5 0

1

2

3

4

5 Month

6

7

8

9

FIGURE 29.16 Maximum temperature rises (64 MHz first-level controlled mode). (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.)

29.2.3.2 Speciﬁc Absorption Rate and Temperature 128 MHz Birdcage Coil For a normal mode operation of the 128 MHz RF coil, maximum 10 g averaged SAR and maximum temperature rises for each tissue are plotted in Figures 29.17 and 29.18. It is found that SAR for the body tissue exceeds the limit. The maximum temperatures rises within the body, the liquid, and the placenta exceed the limit at some stages of pregnancy.

Handbook of Anatomical Models for Radiation Dosimetry

674

18

Normalized maximum specific absorption rate10g (W/kg)

16 14 Safety limit (normal mode) Bladder Body Bone Fetus Liquid Placenta Uterus

12 10 8 6 4 2 0

1

2

3

4

6

5 Month

7

8

9

Maximum temperature rise (°C)

FIGURE 29.17 Normalized maximum 10 g averaged SAR (128 MHz normal mode). (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.)

1 Safety limit (normal mode) Bladder Body Bone Fetus Liquid Placenta Uterus

0.8 0.6 0.4 0.2 0

1

2

3

4

5 Month

6

7

8

9

FIGURE 29.18 Maximum temperature rises (128 MHz normal mode). (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.)

Figure 29.19 depicts the maximum 10 g averaged SAR for each tissue under the firstlevel controlled mode. Again, due to the increased RF power, the energy deposition within body, bone, uterus, and liquid exceed the limit at some stages of pregnancy. The energy depositions within the fetus, however, are below the limit for all stages of pregnancy. The maximum temperature rises of each tissue are depicted in Figure 29.20. Although a higher temperature rise within the fetus is observed for later stages of pregnancy, temperature rises within the fetus are all below the limit. Similar to the results for normal mode, temperature rises within body, liquid, and placenta exceed the limit at some stages of pregnancy.

Applications to Nonionizing Radiation Protection

675

35

Normalized maximum specific absorption rate10g (W/kg)

30 25

Safety limit (first level controlled mode) Bladder Body Bone Fetus Liquid Placenta Uterus

20 15 10 5 0

2

1

3

4

5 Month

6

7

8

9

FIGURE 29.19 Normalized maximum 10 g averaged SAR (128 MHz first-level controlled mode). (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.) 2

Maximum temperature rise (°C)

1.8 1.6 Safety limit (first level controlled mode) Bladder Body Bone Fetus Liquid Placenta Uterus

1.4 1.2 1 0.8 0.6 0.4 0.2 0

1

2

3

4

5 Month

6

7

8

9

FIGURE 29.20 Maximum temperature rises (128 MHz first-level controlled mode). (Extracted from Wu, D.G. et al., IEEE Trans. Microw. Theory Tech., 54, 4472, 2006.)

From these results, we find that the maximum 10 g averaged SARs within the fetus computational phantoms are within the international electrotechnical commission limit under normal mode operation for the 64 MHz system, whereas these values could exceed the limit under the first-level controlled mode operation. The core temperature rises within tissues at first-level controlled mode are also larger than those at normal mode operation. For the 128 MHz system, the maximum 10 g averaged SAR and temperature rises inside the fetus computational phantoms are below the limits for both operating modes.

676

Handbook of Anatomical Models for Radiation Dosimetry

It is observed that the distributions of energy deposition and temperature rise are quite different at two MRI operating frequencies. Our simulation results suggest a higher maximum SAR and temperature rises within tissues when they are exposed to 64 MHz RF coils. This can be explained by the fact that the actual input power for the 128 MHz RF coil is lower than that of the 64 MHz RF coil, when the 10 g averaged SAR is normalized against the whole-body averaged SAR. In this study, detailed electromagnetic and thermal simulations were performed to investigate SAR and temperature rise within nine pregnant woman computational phantoms exposed to MRI RF coils. Two MRI coils were constructed and tuned for 64 and 128 MHz MRI systems. Energy depositions and temperature rises within pregnant woman computational phantoms exposed to these two coils were calculated. We show that, for the 64 MHz MRI coil an increased trend in both SAR and temperature rise is observed as the stage of pregnancy increases. SAR within some tissues can exceed the international electrotechnical commission limit for the first-level controlled mode, while the temperature rises within the fetus exceed the limits for both operating modes. For the 128 MHz MRI coil, although maximum local SARs and temperature rises around the fetus are all within the limits, their values are approaching the limits as the stage of pregnancy increases. Based on the results of this study, we recommend not performing MRI exams for pregnant women at the first-level controlled mode.

29.3 Conclusions Detailed electromagnetic modeling works were performed to evaluate the safety issues related to pregnant women exposed to both active WTMD and MRI RF coils. It was found that the induced electric fields/currents within these computational phantoms are close to or exceed safety levels constructed for different frequency ranges. However, it is still too early to make any final conclusions as to safety issues arising from WTMD and MRI RF coils. More research is needed, and better pregnant woman computational phantoms are desired.

References 1. ICNRP. Guidelines for limiting exposure to time-varying electric, magnetic, and electromagnetic fields (up to 300 GHz), Health Phys, 74, 494, 1998. 2. Matthes, R. Response to questions and comments on ICNIRP, Health Phys, 75, 438, 1998. 3. IEEE Std C95.6-2002. IEEE standard for safety levels with respect to human exposure to electromagnetic fields, 0–3 kHz, IEEE Std C95.6-2002, Institute of Electrical and Electronics Engineers, 01 September 2002. 4. IEEE Std C95.1–2005. IEEE Standard for Safety Levels with Respect to Human Exposure to Radio Frequency Electromagnetic Fields, 3 kHz to 300 GHz, IEEE Std C95.1–2005 (Revision of IEEE Std C95.1–1991), 01 September 2006. 5. Nishizawa, S. et al. Numerical study on an equivalent source model for inhomogeneous magnetic field dosimetry in the low-frequency range, IEEE Trans Biomed Eng, 51, 612, 2004. 6. Wu, D.G. et al. Possible exposure of pregnant women in the third trimester due to scanning in one model of walk through metal detector, Phys Med Biol, 52, 5735, 2007.

Applications to Nonionizing Radiation Protection

677

7. Gandhi, O.P. Some numerical methods for dosimetry: Extremely low frequencies to microwave frequencies, Radio Sci, 30, 161, 1995. 8. Gandhi, O.P. and DeFord, J.F. Calculation of EM power deposition for operator exposure to RF induction heaters, Electromagn Compatibility, IEEE Trans, 30, 63, 1988. 9. Gandhi, O.P., Deford, J.F., and Kanai, H. Impedence method for calculation of power deposition patterns in magnetically induced hyperthermia, Biomed Eng IEEE Trans BME-31, 644, 1984. 10. Gandhi, O.P. and Kang, G. Calculation of induced current densities for humans by magnetic fields from electronic article surveillance devices, Phys Med Biol, 46, 2759, 2001. 11. Gandhi, O.P. et al. Currents induced in anatomic models of the human for uniform and nonuniform power frequency magnetic fields, Bioelectromagnetics, 22, 112, 2001. 12. Wu, D.G. et al. Evaluations of specific absorption rate and temperature increase within pregnant female models in magnetic resonance imaging birdcage coils, IEEE Trans Microw Theory Tech, 54, 4472, 2006. 13. Gabriel, S., Lau, R.W., and Gabriel, C. The dielectric properties of biological tissues: III. Parametric models for the dielectric spectrum of tissues, Phys Med Biol, 41, 2271, 1996. 14. Gabriel, C. Compilation of the dielectric properties of body tissues at RF and microwave frequencies, AL/OE-TR-1996–0037, Armstrong Laboratory (AFMC), Brooks AFB, TX, 1996. 15. Gabriel, C., Gabriel, S., and Corthout, E. The dielectric properties of biological tissues: I. Literature survey, Phys Med Biol, 41, 2231, 1996. 16. Gabriel, S., Lau, R.W., and Gabriel, C. The dielectric properties of biological tissues: II. Measurements in the frequency range 10 Hz to 20 GHz, Phys Med Biol, 41, 2251, 1996. 17. Jin, J. Electromagnetic Analysis and Design in Magnetic Resonance Imaging, CRC Press, New York, 1999. 18. Schaefer, D.J. Safety aspects of radiofrequency power deposition in magnetic resonance, Magn Reson Imaging Clin N Am, 6, 775, 1998. 19. Shellock, F.G. Radiofrequency energy-induced heating during MR procedures: A review, J Magn Reson Imaging, 12, 30, 2000. 20. Shamsi, S. et al. Specific absorption rate evaluation in pregnant woman models radiated from MRI birdcage coil, IEEE MTT-S International Microwave Symposium, San Francisco, CA, June, 2006. 21. Bernardi, P. et al. Specific absorption rate and temperature increases in the head of a cellularphone user, Microw Theory Techniq IEEE Trans, 48, 1118, 2000. 22. Kunz, K.S. and Luebbers, R.J. The Finite Difference Time Domain Method for Electromagnetic, CRC, Boca Raton, FL, 1993. 23. Pennes, H.H. Analysis of tissue and arterial blood temperatures in the resting human forearm, J Appl Physiol, 1, 93, 1948. 24. Taflove, A. Computational Electrodynamics: The Finite-Difference Time-Domain Method, Artech House, Norwood, MA, 1995. 25. Collins, C.M. et al. Temperature and SAR calculations for a human head within volume and surface coils at 64 and 300 MHz, J Magn Reson Imaging, 19, 650, 2004. 26. Gandhi, O.P. and Chen, X.B. Specific absorption rates and induced current densities for an anatomy-based model of the human for exposure to time-varying magnetic fields of MRI, Magn Reson Med, 41, 816, 1999. 27. Nguyen, U.D. et al. Numerical evaluation of heating of the human head due to magnetic resonance imaging, Biomed Eng, IEEE Trans, 51, 1301, 2004. 28. Birdcage Builder. http://psunmr.hmc.psu.edu/birdcage/index.htm. 29. Renhart, W. et al. Investigation of the resonance behavior of a MR-birdcage applying a 3-D-FEM code, Magn, IEEE Trans, 37, 3688, 2001.

30 Summary and Future Needs Related to Computational Phantoms X. George Xu, Michael G. Stabin, Wesley E. Bolch, and W. Paul Segars

A PhD candidate is often asked to include in the dissertation a section called “future work.” This task requires the student to go beyond the data-based scientific process and predict what future developments might follow. “Crystal ball gazing,” however, is unscientific. History has shown that forecasts of the future of science and technology, even by the most experienced individuals, are often proven near-sighted or wrong. Who had foreseen in the 1960s that voxel phantoms would 30 years later gain such popularity in the field of radiation protection dosimetry? In fact, prophetic remarks are more often sources for amusement. One of the most frequently cited examples is the prediction by Popular Mechanics in 1949 that “Computers in the future may weigh no more than 1.5 tons.” Legend has it that the chairman of IBM at the time stated, “I think there is a world market for maybe five computers.” The famous prognosticator Peter Drucker said, “The only thing we know about the future is that it will be different” and “The best way to predict the future is to create it.” Nevertheless, a consensus-based statement on “future needs” may serve as a guide in steering researchers and funding agencies toward problems that are too significant to tackle without a concerted effort. This is the rationale for the concluding chapter of this handbook. The preceding chapters have provided detailed accounts for a number of historically important phantoms and their applications in radiation protection, radiotherapy, medical imaging, as well as nonionizing radiation studies. Each of us has contributed chapters that summarize our own research activities in the last decade—a period that was extremely exciting and fruitful. We are grateful for the supports of individual and collaborative grants from the National Institutes of Health including a multicenter Virtual Patients Project since 2005. Computational phantoms have come a remarkably long way. As discussed in Chapter 1, the first generation of computational phantoms which originated from groundbreaking work performed at Oak Ridge National Laboratory in the 1960s, were designed to represent “population-average” 50th-percentile anatomies. Based on anatomically simplified shapes, these “stylized” computational phantoms allowed the use of Monte Carlo simulations during times when computers were much less powerful. For nearly 40 years, a family of these stylized phantoms, plus versions developed elsewhere (particularly at the GSF in Germany), have been used as the de facto “standard” representation of the so-called ICRP Reference Man methodology. As such, stylized phantoms have been applied to a wide range of internal and external dosimetric studies in the fields of health and medical physics.

679

680

Handbook of Anatomical Models for Radiation Dosimetry

It is clear from Chapter 2 that the original developers understood the obvious shortcomings of stylized phantoms; human anatomy was obviously far too complex to be realistically modeled with a limited set of quadric equations. This inherent limitation in stylized modeling was, forgiven by health physicists who were accustomed to working with the large margins of error inherent to operational radiation protection dosimetry. However, for research purposes and for medical procedures, more precise and accurate dosimetry is often required. As discussed in Chapter 20 for nuclear medicine involving internally administered radionuclides, calculated average organ and marrow doses from the stylized phantoms have not produced strong correlations with observed marrow toxicity. Consequently, most nuclear medicine physicians tend to administer lower-than-optimal amounts of radioactivity to avoid toxicity. With more precise personnel monitoring devices and increasingly restrictive regulations in recent years, radiation protection dosimetry has been motivated to further reduce the uncertainty in various dose estimates that were derived from stylized phantoms. With the advent of tomographic imaging and more powerful computers in the late 1980s and early 1990s, it was natural that anatomically realistic voxel phantoms gained rapid acceptance. Voxel phantoms consist of segmented patient datasets, typically computed tomography (CT) or magnetic resonance imaging (MRI), where each organ or structure is assigned a unique identifier. With their basis on patient data, voxel phantoms are much more realistic than stylized phantoms. Therefore, it was not surprising that these voxel phantoms have been widely adopted in studies involving medical procedures (Chapters 21 through 26). The ICRP has supported the shift away from the stylized phantoms and, as described in Chapter 15, the next-generation reference adult male and reference adult female phantoms have been finalized and will soon be released to the public. Chapters 3 through 14 describe a number of additional voxels phantoms. Like their predecessors, many of these anatomically realistic and person-specific phantoms were adjusted to match, within 1% or better, the 50th-percentile body and organ weights recommended by the ICRP. Despite the success with phantoms constructed from tomographic images, experiences by various research groups suggested that voxels were rather inconvenient to deform. As a result, modeling of anatomical variations and patient motion is severely limited. Looking for better methods, authors of Chapters 5, 8, 12, 13, and 14 found that the boundary representation (BREP) primitives—nonuniform rational B-splines (NURBS) surfaces and efficient polygon meshes—that are widely used in computer games and modeling were ideal building blocks for future 3D and 4D computational phantoms. The BREP surfaces are derived from anatomical image data defi ned in voxels, thereby providing a realistic representation of the anatomy. At the same time, their NURBS or polygon mesh primitives afford the necessary flexibility to model anatomical variations and cardiac and respiratory motions. Looking forward, we envision that the following critical needs should and can be addressed by the research community in the next 5–10 years: 1. Intercomparison of ICRP-89 compatible phantoms Although there are several sets of anatomically realistic and ICRP-compatible phantoms, they are all based on person-specific anatomical data. Consequently, organ locations and shapes differ from phantom to phantom, leading to potential discrepancies in the derived dose estimates. It seems that the recommended use of a pair of reference adult male and female voxel phantoms by the ICRP will work against the original motivation to improve the accuracy of dose estimates. What is needed in the near future is a systematic intercomparison of these phantoms to quantify the

Summary and Future Needs Related to Computational Phantoms

681

significance of dosimetric uncertainty due to variations in organ topology. A phantom data center hosted on a web-based server, such as www.virtualphantoms.org, needs to be established to facilitate this effort. Software tools, with a standardized phantom format and organ identifying system, need to be developed to archive and process a large amount of data efficiently. Advanced image registration and segmentation algorithms need to be adopted to streamline future phantom development and application efforts. 2. A fundamental change in the Reference Man-based paradigm Research and development in radiation protection dosimetry has been predominately driven by the “Reference Man” paradigm. This approach required a computational phantom to match approximately the 50th-percentile values in terms of body height and weight, often at the organ and tissue levels, for a specific gender and age group. The two major variables contributing to overall uncertainty in a dose estimate for a “reference” individual applied to the nuclear medicine population are variability in biokinetics and organ mass. In radiation protection dosimetry, anatomical variations associated with body size are known to account for a significant portion of the overall uncertainty in the internal and external dose estimates derived from 50th-percentile phantoms. Therefore, we believe that a fundamental paradigm change in the ICRP Reference Man methodology needs to take place by expanding the 50th-percentile phantoms into a much larger set of phantoms ranging from the 10th percentile to the 90th percentile in each of the gender and age groups. Anthropometric data from population surveys can be used to evaluate trends in body and organ distributions. Both body weight index and trunk height can be used to match a specific individual with this expanded library of reference phantoms. Such phantoms, which encompass many anatomical variations, are also needed for clinical studies or trials for assessing the efficacy of medical imaging devices, patient dose from modern CT examinations, and the second-cancer effects after radiotherapy. 3. Physics-based methods for deformation modeling Motion-simulating 4D computational phantoms will play an increasingly vital role in the understanding and management of organ motion in many radiotherapy and medical imaging modalities. As discussed in Chapters 5 and 6, work has been done to model cardiac and respiratory motions using BREP-based methods. These phantoms are based on gated patient imaging data and therefore provide a realistic simulation of the motion. A limitation to these models is that they provide only one realization of the motion, specifically that observed in the patient data. They do not have the ability to realistically simulate motion variations that may occur within the same individual or within the population at large. To model variations in motion would require accurate handling of the interactions between the organs. Finite element methods have the advantage of being physicsbased and can be used in single- and multiple-organ deformable registration. The major limitations of current physics-based multiorgan mechanical models are (1) the organ material properties are assumed to be linear and are culled from literature for unrelated experiments. (2) They do not take into account the full-organ contact mechanics, but rather rely on the motion of interlinked organs using prescribed boundary conditions. (3) Commercially available finite element software tools are computationally inefficient. For real-time simulation of cardiac and respiratory motions including organ interactions, more powerful physics-based algorithms are needed. With truly physics-based deformation techniques, users may change a

682

Handbook of Anatomical Models for Radiation Dosimetry

host of parameters (left and right diaphragm movement, amount of chest expansion, material properties of the organs, etc.) and the phantoms will realistically simulate and predict the deformation of multiple interlinked organs under user-specified conditions. Such motion-simulating phantoms will find novel applications in both medical imaging and radiotherapy modalities involving the heart, lung, liver, and kidneys. Furthermore, phantoms that account for aerodynamic flow of radioactive particulates in various parts of a deformable respiratory system need to be developed to improve internal dose models for radiation protection that are currently based on a static 3D lung anatomy. 4. Monte Carlo simulations with advanced geometries Monte Carlo codes for radiation transport simulations were originally designed for nuclear engineering and high-energy physics applications. Although these codes contain excellent radiation physics algorithms, they suffer from poor software engineering design and are only able to handle constructive solid geometries (CSG) defined by quadric equations or voxels, as described in Chapter 1. These deficiencies have resulted in three problems that compromise current efforts on phantom research: (1) the implementation of various anatomically complex phantoms in these codes require cumbersome manual processes and the MCNP/MCNPX codes, for example, fail to handle phantoms that consist of more than 25 millions of voxels. (2) Currently the NURBS-based or mesh-based phantoms must be converted to voxels before using with Monte Carlo programs. (3) Existing Monte Carlo codes are unable to handle a “moving” target such as the dynamic heart or lung. Research to convert an object defined in computer-aided design (CAD) software to CSG surfaces that are acceptable by a Monte Carlo code has been on going in the nuclear engineering community for years. However, in this approach, the final geometries are still defined by planes, spheres, and cones which are probably satisfactory for nuclear reactor applications, but fail to define the details in a human anatomy. Future research is needed to develop new and computationally efficient ray-tracing algorithms that can directly process NURBS- and mesh-based geometric objects in a general-purpose Monte Carlo radiation transport engine. A major challenge is to gain necessary support from government-funded and mission-driven Monte Carlo code developers. A trend in the user community toward shifting to open-source codes has been observed in recent years and is expected to accelerate in future work related to advanced phantoms. To conclude, it is interesting to note that a paper which appeared in 2000 about the VIP-Man phantom (see Chapter 6) predicted that the advantages afforded by both the BREP-type of surface geometries and anatomically realistic voxels would be eventually combined: “For the purposes of setting radiation protection standards, it may be possible to eventually bridge these two types of models, leading to a new generation of hybrid ‘standard’ model(s) that will be acceptable to the radiation protection community. Such a new generation of models for radiation protection should be realistic enough to accurately represent major radiosensitive tissues and organs, and flexible enough to represent different populations by scaling. Computers are going to be so powerful that very complex models can be handled without a problem.” Impressively, these so-called hybrid reference phantoms were realized only several years later (e.g., see Chapters 5, 6, 8, 12, and 13). In the next 10 years, advances in computational phantom research will be mostly driven by the advanced research, perhaps related to the fi nal crusade against cancer.

Summary and Future Needs Related to Computational Phantoms

683

Increasingly personalized whole-body computational phantoms will be developed and applied for various clinical applications. Such phantoms will contain deformable anatomies that are physics-based and are, therefore, biomechanically realistic in depicting realtime and multiorgan deformation associated with cardiac and respiratory motions. Using a multiscale and virtual reality enabled computational framework of the human body, these phantoms will also possess physiological and functional information of the human body at the organ and cellular levels. Breakthroughs in computational radiobiology, in the context of cancer radiotherapy, are expected to bring a new horizon to the personalized radiation medicine by understanding and harnessing the massive power of genomic data. The 40-year history reflected in this book shows that coordinated and cooperative efforts among radiological engineers, computer scientists, biologists, and clinicians are the key to the success of future research endeavors.

About the Contributors Manuel Bardiès supervises the Radionuclide Dosimetry group within the Oncology Research Department (INSERM Unit 892, Nantes, France). He joined the French National Institute of Health and Medical Research (INSERM) in 1992 after earning his PhD in radiological physics from the University Paul Sabatier, Toulouse, France. The development and supervision of the group is made in tight coordination between the other research groups within the department (chemistry, radiopharmacy, radiobiology) and with the clinical departments in both the Nantes Cancer Center and the University Hospital. Research activities are being carried out in different institutions, including Nantes National Veterinary School for preclinical imaging experiments. Radiopharmaceutical dosimetry is addressed at different scales, including cell microdosimetry for alpha radioimmunotherapy, small animal imaging and dosimetry, and participation to targeted radiotherapy clinical trials. Achievements in radiopharmaceutical dosimetry include the development of several dosimetry clinical trials, the setup of a modeling computing platform (48 nodes computer cluster) for MonteCarlo-based dosimetry (Oedipe), and camera modeling (Gate). The group is involved in the development of Gate and Manuel Bardiès is a member of the openGate collaboration steering committee. He has also been a member of the European Association of Nuclear Medicine (EANM) dosimetry committee since 2001 and is a member of several other scientific societies. Dr. Bardiés has authored more than 30 publications and has given many invited lectures at national and international meetings (EANM, International Atomic Energy Agency [IAEA], European Society for Therapeutic Radiology and Oncology [ESTRO]). Giovanni Bibbo has extensive work experience in different scientific fields. He graduated in physics at the University of Rome, Italy, majoring in astrophysics and then completed his PhD in atmospheric physics at the University of Adelaide. He continued his research in space science at the Mullard Space Science Laboratory, University College London, United Kingdom, and as a visiting scientist at the European Space Research and Technology Centre, the Netherlands. On his return to Australia, he worked as a senior scientist in the South Australian Environment Protection Authority. For the last 18 years, Giovanni has worked at the Women’s and Children’s Hospital as a principal medical scientist and a radiation safety officer. He has an adjunct research position with the University of South Australia, where he supervises the work of research students; he has also been actively involved with a number of professional societies.

685

686

About the Contributors

Wesley E. Bolch is currently a professor of radiological and biomedical engineering at the University of Florida and the director of the Advanced Laboratory for Radiation Dosimetry Studies at the University of Florida. Dr. Bolch manages a broad research program including (1) National Cancer Institute (NCI)- and Department of Energy (DOE)funded projects to construct high-resolution models of the skeleton to support dose–response studies in radionuclide therapy and radiation epidemiology; (2) National Institute for Biomedical Imaging and Bioengineering (NIBIB)-funded projects to develop scalable voxel-based computational phantoms of pediatric patients for organ dose assessment in Computed Tomography (CT), fluoroscopic imaging, and radiotherapy; and (3) Center for Disease Control and Prevention (CDC)-funded projects in stochastic modeling of worker inhalation and gamma-ray exposures from radiological accidents and potential terrorist events. He is a member of the Society of Nuclear Medicine’s Medical Internal Radiation Dose (MIRD) Committee since 1993, a member of Committee 2 of the International Commission on Radiological Protection (ICRP) since 2005, the chair of ICRP Committee 2’s Task Group on Dose Calculation, and a member of the National Council on Radiation Protection and Measurements (NCRP), also since 2005. At present, Dr. Bolch is an active member on two NCRP scientific committees: NCRP SC 4-1 on Management of Persons Contaminated with Radionuclides and NCRP SC 6-3 on Uncertainties in Internal Radiation Dose Assessment. Ahmet Bozkurt attended the Middle East Technical University, Gaziantep, Turkey, and obtained his BS in engineering physics in 1991. In 1993, he received a national scholarship from the Higher Education Council and came to the United States to pursue graduate studies. He graduated from the University of Missouri-Rolla, Rolla, Missouri, in 1996 with an MS in nuclear engineering. He then came to the Rensselaer Polytechnic Institute, Troy, New York, and received his PhD in nuclear engineering and science in 2000. Dr. Bozkurt joined the Department of Physics at Harran University, Sanliurfa, Turkey, in 2001 where he now serves as an associate professor of nuclear physics. His current research interests span environmental and operational health physics, radiation protection and dosimetry, and medical and industrial applications of nuclear techniques. Martin Caon has been a senior lecturer in biophysical science at Flinders University School of Nursing and Midwifery, Adelaide, South Australia, for 20 years. His research interest is in the dosimetry of diagnostic radiology procedures in children—in particular CT. He is an editor of the journal Australasian Physical & Engineering Sciences in Medicine (http://apesm.acpsem.org.au) and serves on the editorial board of Biomedical Imaging and Intervention Journal (biij). He is actively involved in the Australasian Radiation Protection Society.

About the Contributors

687

Tsi-Chian Ephraim Chao is an assistant professor in the Department of Medical Imaging and Radiological Sciences at Chang Gung University (CGU), Taiwan. Dr. Chao received his PhD in nuclear engineering at the Rensselaer Polytechnic Institute in 2001. His research interests include Monte Carlo simulation and radiation dosimetry/instrumentation. Dr. Chao codirects the Radiation Therapy Laboratory at CGU, including responsibilities in quality assurance (QA) for radiation therapy, small field dosimetry, transmission dosimetry based on electronic portal imaging devices (EPID), proton therapy, microdosimetry, and nanodosimetry. Ji Chen is an associate professor in the Department of Electrical and Computer Engineering at the University of Houston, Houston, Texas. Dr. Chen received his PhD in electrical and computer engineering from the University of Illinois at Urbana-Champaign in 1998. His research interests include safety and effectiveness of medical devices and the interaction between electromagnetic field with human subjects. In 2004, Dr. Ji Chen collaborated with Food and Drug Administration (FDA) and Foundation for Research on Information Technologies in Society (IT’IS) on the Virtual Family Project. John J. DeMarco is a medical physicist and an associate clinical professor at the University of California, Los Angeles (UCLA) Department of Radiation Oncology, and a faculty member in the UCLA Biomedical Physics graduate program. Dr. DeMarco received his PhD in biomedical physics from the UCLA Biomedical Physics program in 1996. His current research interests include radiation dosimetry and Monte Carlo modeling as applied to the therapeutic and diagnostic use of radiation.

Aiping Ding is currently pursuing his PhD in nuclear engineering at the Rensselaer Polytechnic Institute, Troy, New York. He is interested in software development related to medical image visualization and analysis, Monte Carlo simulation, geometrical modeling and visualization, and automated phantom deformation. Aiping received his BS in computer science and engineering from Hebei University of Technology, Tianjin, China, and his MS in nuclear science and engineering from the Institute of Plasma Physics, Chinese Academy of Sciences, Hefei, China.

688

About the Contributors

Mary Ann Emmons-Keenan is a diagnostic medical physicist in the Department of Radiology and Radiological Sciences at Vanderbilt University Medical Center, Nashville, Tennessee. She received her BSc in chemistry from Athens State University, Athens, Alabama, and her MSc in medical physics from Vanderbilt University School of Medicine, Nashville, Tennessee. She is a registered radiological inspector with the state of Tennessee. She is also a member of the American Association of Physicists in Medicine and of the Society of Nuclear Medicine.

Michael J. Fernald is a research assistant for RADAR Inc. in Nashville, Tennessee. He received his BS in physics and mathematics from the University of Idaho in 2006 and his MS in medical physics from the Vanderbilt University in 2008.

Martin Fuss is a professor in the Department of Radiation Oncology/Radiation Medicine at the Oregon Health & Science University (OHSU). He specializes in image-guided radiation therapy, intensity-modulated radiation therapy, stereotactic (intracranial and extracranial/stereotactic body) radiosurgery, and fractionated treatments for various solid tumors. Dr. Fuss did his residency at the University of Heidelberg and the German Cancer Research Center, and his fellowship at the Loma Linda University Proton Center, and rapidly progressed to the rank of associate professor at the University of Texas Health Science Center at San Antonio. He joined the medical staff at OHSU in 2006. He is the director for the imageguided radiation therapy program. David E. Hintenlang is a professor of nuclear and radiological engineering and biomedical engineering at the University of Florida. Dr. Hintenlang’s research programs focus largely on the accurate determination of ionizing radiation doses delivered to patients as a result of medical procedures. This includes the development of measurement devices, techniques, and anthropomorphic phantoms that can be readily applied in clinical environments. Areas of continuing research include CT dose to pediatric and adult patients, for both Multi-Detector CT

About the Contributors

689

and Cone Beam CT, as well as mammography doses delivered from digital mammography or tomosynthesis techniques. Dr. Hintenlang is the program director of the Commission on Accredition of Medical Physics Educational Programs (CAMPEP) Accredited Graduate Medical Physics Program at the University of Florida, a diplomate of the American Board of Radiology with certification in diagnostic radiological physics, and an active clinical physicist licensed in the State of Florida. He is an active member of the AAPM, the past president of the Florida Chapter American Association of Physicists in Medicine (FLAAPM), and a fellow of the American College of Medical Physics.

Gabriela Hoff is a professor of physics at the Catholic Pontifical University of Rio Grande do Sul, Rio Grande do Sul, Porto Alegre, Brazil. Dr. Hoff received her PhD in nuclear bioscience and medical physics from the State University of Rio de Janeiro in 2005. Her current research interests include image segmentation and computational dosimetry, especially in diagnostic imaging. Dr. Hoff directs the Group of Experimentation and Computational Simulation in Medical Physics.

Jorge Hurtado is currently a PhD student in the Department of Nuclear and Radiological Engineering at the University of Florida and a research for the Advanced Laboratory for Radiation Dosimetry Studies (ALRADS) also at the University of Florida. Jorge received his bachelors of engineering in nuclear engineering and his MSc in nuclear engineering sciences at the University of Florida. His dissertation studies focus on the development of computational adult hybrid phantoms as well as applications including CT dosimetry and homeland security guidelines. He is currently a member of the American Association of Physicists in Medicine.

An Jin is currently working toward her PhD in biomedical engineering at the Rensselaer Polytechnic Institute in Troy, New York. Her research interests include signal modeling and reconstruction, and medical imaging analysis and understanding. She received her BS in biomedical engineering from Zhejiang University, Hangzhou, China.

690

About the Contributors

Wolfgang Kainz is a senior principal scientist at the U.S. Food and Drug Administration in the Center for Devices and Radiological Health. Dr. Kainz received his MS in electrical engineering from the Technical University of Vienna, Austria, in 1997 and his PhD in technical science from the same university in 2000. He is the chairman of the IEEE, International Committee on Electromagnetic Safety (ICES), Technical Committee 34, which develops compliance techniques for wireless devices, and he is also a member of the Administrative Committee of the ICES. After working for the Austrian Research Center Seibersdorf (ARCS), he joined the Foundation for Research on Information Technologies in Society (IT’IS) in Zurich, Switzerland, as an associate director. At IT’IS, Dr. Kainz worked on the development of in-vivo and in-vitro exposure setups for bio-experiments. His research interest is currently focused on the safety and effectiveness of medical devices and safety of humans in electromagnetic fields. This includes computational electrodynamics for safety and effectiveness evaluations using anatomical models of the human anatomy; magnetic resonance imaging safety; performance and safety of wireless technology used in medical devices; electromagnetic compatibility of medical devices; and dosimetric exposure assessments. In 2004, Dr. Kainz initiated the Virtual Family Project in cooperation with IT’IS and the University of Houston. Iwan Kawrakow is a senior research officer at the Ionizing Radiation Standards (IRS) group, National Research Council of Canada, and an adjunct professor in the Department of Physics, Carleton University, both in Ottawa. He received his doctoral degree in high-energy physics from the University of Leipzig, Germany, in 1994. After obtaining his degree, Dr. Kawrakow worked for two years at the Clinic for Radiation Therapy and Oncology, University of Leipzig, before joining the IRS group in Ottawa. Dr. Kawrakow is the principal developer of the EGSnrc system, a widely used Monte Carlo simulation package for coupled electron and photon transport. His research focuses on transport theory, theoretical dosimetry, radiation physics, and Monte Carlo techniques and their use in radiation therapy, radiation dosimetry, and medical imaging. Helen Jamil Khoury is a full professor in the Nuclear Energy Department at Federal University of Pernambuco (DEN/ UFPE), Recife, Brazil. She received her PhD in nuclear instrumentation from the Pontifícia Universidade Católica, São Paulo, Brazil, in 1981. Her current research interests include radiation dosimetry and nuclear instrumentation related to health physics, diagnostic imaging, and radiotherapy. Dr. Khoury is the head of the Radiation Dosimetry and Instrumentation Research Group at DEN/UFPE (http://www.grupodoin. com) and is the coordinator of the Laboratory of Radiation Metrology (LMR)–DEN/UFPE. She serves on several editorial boards and is involved in various technical committees.

About the Contributors

691

She was the president of the Brazilian Society of the Radiation Protection during the period 2000–2001 and 2004–2006. Chan Hyeong Kim is a professor of nuclear engineering at Hanyang University in Seoul, South Korea. Dr. Kim received his PhD in nuclear engineering from Texas A&M University in 1998. His current research interests include Monte Carlo particle transport and detector simulations, high-precision radiation detection and nuclear imaging, and advanced computational models for precision dose calculation in the human body. He is a member of the American Association of Physicists in Medicine (AAPM), the American Nuclear Society (ANS), the Health Physics Society (HPS), and several Korean societies in nuclear engineering and medical physics. He is currently serving the societies as the managing editor of the Journal of Radiation Protection, as an editorial advisory board member of The Open Nuclear & Particle Physics Journal, and as a board member of the Korean Society of Medical Physics. Sakae Kinase is a senior scientist at the Japan Atomic Energy Agency. Dr. Kinase received his PhD from Tohoku University in 2002. He has been working on radiation dosimetry, in vivo counting, and tissue substitute developments. He is involved in the technical committees of various organizations such as the Atomic Energy Society of Japan (AESJ) and the Japan Health Physics Society (JHPS). He has been given an AESJ award, a radioisotopes award, and a IM2005 poster award.

Gary H. Kramer is the head of the National Internal Radiation Assessment Section (NIRAS) in Health Canada’s Radiation Protection Bureau, Ottawa, Ontario, Canada. Dr. Kramer received both degrees (BSc and PhD) from the University of Sussex in 1971 and 1974, respectively. NIRAS operates the Canadian National Calibration Reference Centre (NCRC) for bioassay and in vivo monitoring. The NCRC carries out performance testing of the Canadian Regulator’s licensees for bioassay measurements, in vivo monitoring, and internal dosimetry. As a senior research scientist in Health Canada, his current research interests include rapid bioassay and in vivo methods for emergency response, and in vivo methods using experimental and Monte Carlo techniques. The latter makes use of voxel phantoms. His section has memoranda of understanding with many other international agencies for the purpose of furthering work and many of his publications are a result of this collaborative work with countries such as South Korea, Romania, the United States, Spain, and Brazil. Dr. Kramer is involved in various technical committees such as the Health Physics Society, the International Commission on Radiological Protection, and the Canadian Radiation Protection Association (CRPA); he was also the past president of the CRPA. He is currently the Canadian elected member of International Radiation Protection Association’s Executive Council (2004–2012).

692

About the Contributors

Richard Kramer is a professor in the Departamento de Energia Nuclear of the Universidade Federal de Pernambuco (DEN/UFPE) in Recife, Brazil. Dr. Kramer received his doctoral degree in nuclear physics from the Technische Universität München, Germany, in 1978. He worked at the former National Research Center for Environment and Health (GSF) research center near Munich, where he developed the sex-specific MIRD5-type mathematical phantoms, called Adult Male (ADAM) and Adult Female (EVA), which have been widely used for Monte Carlo calculations of organ and tissue equivalent doses in the human body from external exposures. Currently, Dr. Kramer is the head of the computational dosimetry group at the DEN/UFPE, where he and his colleagues developed the MAX06 and the FAX06 phantoms, the first adult human voxel models based on anatomical reference data specified by the International Commission on Radiological Protection in Report No. 89. Dr. Kramer’s group also developed a method for using microCT images of trabecular human bone for the purposes of dosimetry in skeletal soft tissues at risk. Choonik Lee graduated from the medical physics program at the University of Florida (UF). Under Dr. Wesley Bolch’s supervision, Dr. Lee has developed the UF series of pediatric voxel phantoms. He also performed a series of Monte Carlo simulation to assess organ doses under various radiation exposure conditions, including helical CT. After graduation, Dr. Lee joined the radiation oncology department at the M. D. Anderson Cancer Center Orlando, Orlando, Florida, as a postdoctoral researcher to investigate the dosimetric effect of organ deformation during a fractionated radiation treatment. At present, Dr. Lee is involved in a medical physics residency program where he conducts clinical training in various aspects of radiation oncology physics. Choonsik Lee is currently a research associate of radiological and biomedical engineering at the University of Florida. Dr. Lee has a bachelor of science and a master of engineering degree in nuclear engineering and received his PhD in health physics from Hanyang University in South Korea. He is a member of the Health Physics Society and the American Association of Physicists in Medicine. His research interests include computational radiation dosimetry using Monte Carlo transport technology, development of computational stylized, voxel and hybrid human phantoms, and pediatric dosimetry in diagnostic and therapeutic radiation exposure.

About the Contributors

693

Junli Li is a professor and the director of the Radiation Protection Division and an associate head of the Department of Engineering Physics of Tsinghua University in Beijing, China. Dr. Li obtained his BS, MS, and PhD from Tsinghua University in 1991, 1992, and 1995, respectively. From 1995 to 1997, he worked as a postdoctor at the Institute of Nuclear and New Energy Technology of Tsinghua University and became an associate professor in 1997. From 1998 onward he served as the director of Radiation Protection and Environment Protection Laboratory, Department of Engineering Physics of Tsinghua University. He is currently serving on the following committees: vice-chair of the Committee of Monte Carlo Method & Application, Chinese Computational Physics Society; member and director of the Chinese Computational Physics Society; member of the Infective Disease & Outburst Health Accident Commission in Beijing; member of the Committee of Radiation Health Standardization Technology, Ministry of Health, China; and a member of the 4th Editing Board of the Chinese Journal of Radiological Health. His research interests cover various aspects of radioprotection and environmental protection, including shield technology, radiation protection optimization, radiation dosimetry, and Monte Carlo methods. Vanildo Júnior de Melo Lima is a professor of anatomy in the Department of Anatomy and a PhD student in the Department of Nuclear Energy at the Universidade Federal de Pernambuco, Recife, Pernambuco, Brazil. His current research interests include anatomical models to be used in radiation dosimetry.

Qian Liu is an associate professor from the Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology (HUST), Wuhan, China. Dr. Liu received his PhD in biomedical engineering at HUST and worked as a visiting scholar at the University of Pennsylvania for one year with the support of Johnson Research Foundation in 2002. As the leader of the Visible Chinese Human research group (http://www.vch.org.cn), his research interests include computational physiome study, biomedical informatics, and biomedical photonics. He is involved in several projects funded by the National High-Tech Research and Development Program of China (863 program) and the National Natural Science Foundation of China.

694

About the Contributors

Daniel L. Lodwick is currently a student of the College of Medicine at the University of Florida. After receiving his bachelor’s degree in nuclear engineering and master’s degree in nuclear engineering sciences, Lodwick has maintained research interests in pediatric modeling of both reference and patient-specific individuals for use in radiation dosimetry calculations of organ absorbed dose in CT imaging, fluoroscopic imaging, nuclear medicine, and radiation therapy. He is a member of the American Association of Physicists in Medicine, the American Medical Student Association, and a student member of the American Medical Association.

Eduardo César de Miranda Loureiro is a professor in the mechanical engineering department at the Escola Politécnica da Universidade de Pernambuco, Recife, Brazil. Dr. Loureiro received his DSc in nuclear engineering/numerical dosimetry from the Universidade Federal de Pernambuco, Recife, Brazil, in 2002. His current research interests include computational dosimetry, radiological imaging, image processing, and radioprotection.

C.-M. Charlie Ma is a professor and director of radiation physics in the Department of Radiation Oncology, Fox Chase Cancer Center, Philadelphia, Pennsylvania. Dr. Ma received his PhD in medical physics from the University of London, London, U.K., in 1992. His research interests include radiation dosimetry and instrumentation related to radiation physics and radiotherapy, Monte Carlo simulation techniques for radiation therapy and protection, treatment optimization and delivery techniques for advanced radiotherapy employing novel beam modalities, and fractionation and dose schemes. Dr. Ma has published extensively in these areas. He is a reviewer for a number of scientific journals and serves on several editorial boards. He is also a scientific reviewer for National Institutes of Health (NIH), Department of Defense (DOD), DOE, National Science Foundation (NSF), American Society for Therapeutic Radiology and Oncology (ASTRO), Radiological Society of North America (RSNA), and several other funding organizations. Dr. Ma is a fellow of the AAPM and Institute of Physics (IOP) and has served on many committees and task groups of AAPM, ASTRO, American College of Medical Physics (ACMP), and North American Chinese Medical Physicists Association (NACMPA).

About the Contributors

695

Jizeng Ma is a professor of radiation protection and environmental protection at the China Institute of Atomic Energy (CIAE), Beijing, China. He graduated from the physics department at Nanjing University in 1982 majoring in nuclear physics. Dr. Ma received his PhD in experimental fundamental radiology/health physics in Shiga University of Medical Science of Japan in 1996. His research interests include radiation dosimetry and occupational radiation protection, indoor radon measurement, and radiation detection technique. Dr. Ma is the deputy director of the Department of Radiation Safety, CIAE. He led a cooperative project with IAEA on improvement of internal exposure monitoring technique for radiation workers. He was previously a correspondent member of the ICRP/ICRU task group for radiation protection of aircrew and a member of the Radiation Safety Standard Committee of the IAEA. Dr. Ma is the vice chairman of the Chinese Radiological Health Society. He is also the editor of several Chinese radiation protection journals. Michael McNitt-Gray, PhD, is a professor of radiological sciences in the David Geffen School of Medicine at UCLA. He is also the director of the biomedical physics graduate program there. He received his PhD in biomedical physics from UCLA in 1993, his MSEE from Carnegie-Mellon University in 1980, and his BSEE from Washington University in St. Louis in 1979. He is ABR certified in diagnostic medical physics since 1998. Dr. McNitt-Gray’s current research interests include investigations into x-ray computed tomography (CT) imaging with the goal of maximizing the information that can be extracted from the resulting image data. These activities include research into (a) the physics of CT image acquisition including estimating radiation dose and assessing image quality and (b) image processing techniques to analyze and extract information from the resulting image data, including computer-aided detection and diagnosis. Dr. McNitt-Gray currently serves on the ICRU Committee on Image Quality and Patient Dose in Computed Tomography, the ACR CT Accreditation Program Physics subcommittee, as well as various AAPM, RSNA, and SPIE technical and scientific program committees. He also serves as a reviewer for several medical physics and radiology journals such as Medical Physics; Physics in Medicine and Biology; and others. William E. Moloney is a practicing medical physicist at Petrone Associates, LLC, in New York City. His responsibilities include diagnostic quality control in computed tomography, radiography, and fluoroscopy, as well as treatment planning and quality assurance in IMRT, tomotherapy, and brachytherapy. Moloney received his BSc in physics at Saint Joseph’s University in Philadelphia, Pennsylvania. He received his MSc in medical physics from the University of Florida in Gainesville, Florida, where his research entailed the development and application of advanced radiation dosimetry in diagnostic imaging, specifically, computed tomography.

696

About the Contributors

Yong Hum Na is currently a PhD candidate in the Department of Biomedical Engineering at the Rensselaer Polytechnic Institute, Troy, New York. His current research involves the development of multidimensional mesh deformation algorithms and computational tools for size-adjustable anatomy models for radiological studies. He previously received his MS in electrical engineering from Washington University in St. Louis, Missouri.

Tomoaki Nagaoka is an expert researcher for the Electromagnetic Compatibility Group, Applied Electromagnetic Research Center at National Institute of Information and Communications Technology, Tokyo, Japan. Dr. Nagaoko received his PhD in medical science from Kitasato University, Kanagawa, Japan, in 2004. His current research interests include human phantom development and electromagnetic dosimetry related to biomedical electromagnetic compatibility. He was the recipient of several awards, including the 2004 Best Paper Award of Physics in Medicine and Biology; the 2007 Young Researcher Award of the Institute of Electronics, Information and Communication Engineers; and the 2008 International Scientific Radio Union Young Scientist Award. Dr. Nagaoka is a member of the Consortium of Computational Human Phantoms.

Deanna Hasenauer Pafundi is currently a PhD candidate in the Department of Nuclear and Radiological Engineering at the University of Florida and a researcher for the Advanced Laboratory for Radiation Dosimetry Studies at the University of Florida (UF). Her dissertation studies focus on the development of detailed three-dimensional image-based skeletal dosimetry models for the ICRP pediatric age series. She has also taught lectures in radiation dosimetry and radiation protection at UF. Pafundi received her BSc summa cum laude in nuclear engineering and her MSc in nuclear engineering sciences in the Department of Nuclear and Radiological Engineering at UF. She is the recipient of the four-year Department of Energy Health Physics Fellowship. Pafundi held a three-month DOE practicum position at the United States Transuranium and Uranium Registries under Dr. Anthony James constructing 3D voxel-based phantoms for detector calibration and activity concentration studies.

About the Contributors

697

She is a member of the HPS, the ANS, the AAPM, the Alpha Nu Sigma, the Florida Chapter of the Health Physics Society (FL-HPS), and the Florida Chapter American Association of Physicists in Medicine (FL-AAPM). Pafundi has also served as a member of the Engineering Student Advisory Council at UF. She has served as the president of the Society of Health and Medical Physics Students at UF, along with committee chairs of Engineering Fair, Fundraising, and Community Service and Outreach. In addition to research and academics, Pafundi is certified by the Aerobics and Fitness Association of America and instructs both recreational and academic group fitness classes in strength training, aerobics, nutrition, anatomy, physiology, and kinesiology. Harald Paganetti is a biophysicist at Massachusetts General Hospital in Boston and an associate professor of radiation oncology at Harvard Medical School. Dr. Paganetti received his PhD in experimental nuclear physics at the Rheinische-Friedrich-Wilhelms University, Bonn, Germany, in 1992. He is currently the Director of Physics Research for the Department of Radiation Oncology at Massachusetts General Hospital. His research interests include “Monte Carlo” calculations in proton therapy and intensity-modulated photon therapy, dosimetric effects of breathing motion, radiation-induced cancer risks, positron emission tomography imaging in proton therapy, adaptive treatment planning, and biological effects of proton beams. Dr. Paganetti is the author of more than 60 peer-reviewed publications and has been awarded many research grants. He serves on several editorial boards and is a member of various task groups and committees within the American Association of Physicists in Medicine, the International Commission of Radiation Units & Measurements, the International Atomic Energy Agency, and the National Institutes of Health/National Cancer Institute. Niko Papanikolaou is a professor of radiology and radiation oncology at the University of Texas Health Sciences Center in San Antonio and the director of medical physics at the Cancer Therapy and Research Center. He currently teaches medical physics to residents, graduate students, and dosimetry students and his research interests include IGRT, dose calculation algorithms, and biological optimization.

John E. Pattison is a senior lecturer in applied physics and university radiation safety officer at the University of South Australia, Adelaide, Australia. He received his MSc in radiation dosimetry from the University of Adelaide in 1971, after completing his research in the Department of Medical Physics at the Royal Adelaide Hospital. His current research interests include Monte Carlo modeling for radiation dosimetry related to health physics, and population genetics. Pattison is the codirector of the knowledge-based Intelligent Engineering Systems Centre, and the leader of

698

About the Contributors

its Intelligent Medical Systems and Health Physics Group. He has recently served as an editor of the journal Australasian Physical and Engineering Sciences in Medicine, and is a fellow of the Australian Institute of Physics and a companion of the College of Biomedical Engineering of Engineers Australia.

Nina Petoussi-Henss is currently a research scientist in the group Medical Physics of the Helmholtz Zentrum München, German Research Center for Environmental Health. Dr. Petoussi-Henss obtained her PhD in physics at the University of Birmingham, U.K. Dr. Petoussi-Henss has been working in the field of radiation protection; numerical dosimetry with Monte Carlo methods and anthropomorphic phantoms for internal and external exposures; organ dose calculations due to occupational, environmental, and medical exposures; and patient dosimetry after incorporation of radionuclides. She is a member of the ICRP Committee 2 Task Group DOCAL (dose calculations).

John W. Poston, Sr. has been a professor in the Department of Nuclear Engineering at Texas A&M University in College Station, Texas since 1985. During this time, he served for 10 years as head of this department. Before coming to Texas A&M, he was on the faculty at the Georgia Institute of Technology and prior to that was a section head in the Health Physics Division at the Oak Ridge National Laboratory. Dr. Poston received his PhD in nuclear engineering from the Georgia Institute of Technology in 1971. He has served as the president of the Health Physics Society (HPS) and was elected as a fellow of this society in 1987. He received the Founders Award from the HPS in 1994 and the Robley D. Evans Commemorative Medal from the HPS in 2005. He was elected fellow of the American Nuclear Society in 1996 and, in the fall of 2001, fellow of the American Association for the Advancement of Science, the oldest scientific organization in the United States having been formed in the mid-1800s. He served as an elected member of the National Council on Radiation Protection and Measurements (NCRP) for 12 years and in 2002 was elected a distinguished emeritus member (lifetime) of the council. He received the Glen Murphy Award from the American Society for Engineering Education and the Loevinger-Berman Award from the Society of Nuclear Medicine. In 1995, he was elected to the Georgia Tech Academy of Distinguished Engineering Graduates. His areas of expertise include internal dose assessment as well as external dosimetry of mixed radiation fields. He currently serves as the vice president of the NCRP responsible for the nuclear and radiological security and safety program area.

About the Contributors

699

Rui Qiu is a postdoctoral associate in the Department of Engineering Physics, Tsinghua University, Beijing, China. Dr. Qiu obtained her BS in nuclear science and technology in 2002 and her PhD in health physics both from Tsinghua University. From October 2006 to October 2007, she worked at Radiation Safety Division of Pohang Accelerator Laboratory (Korea) as a researcher. Dr Qiu’s research interests cover various aspects of the Monte Carlo method, radiation dosimetry, radiation effect, and radiation protection.

Warren Dan Reece is a professor of nuclear engineering at Texas A&M University, College Station, Texas. Dr. Reece received his PhD in nuclear engineering from Georgia Institute of Technology in 1988. His current research interests include Monte Carlo transport calculations, detector simulations, precision dose measurements, operation and use of research reactors, and technology-aided teaching methods. Dr. Reece is also the director of the Nuclear Science Center and is involved in many activities dealing with nuclear sciences. He is best known for graduating very bright students such as Drs. George Xu and Chan Kim, among many others. Kimiaki Saito is a senior principal researcher of the Nuclear Science and Engineering Directorate, Japan Atomic Energy Agency. Saito directs the Radiation Effect Analysis Group and is also the manager of the Research Unit for Quantum Beam Life Science Initiative (http://yayoi.kansai.jaea.go.jp/ biou/en/index.html). He started his work with dose evaluation for environmental radiations. His main research interests include dosimetry using human models and Monte Carlo simulations for radiation protection purposes. In recent years, he has conducted simulation researches on biological radiation effects, and has also extended his research into medical fields. Saito is a member of the Consortium of Computational Human Phantoms. Kaoru Sato is a researcher of the Research Group for Radiation Protection, Division of Environment and Radiation Sciences, Nuclear Science and Engineering Directorate, Japan Atomic Energy Agency. His latest research is the development of reliable dose evaluation methods in Japanese public members or workers using voxel phantoms (http://www. jaea.go.jp/04/nsed/ers/radiation/rpro/index-e.htm). He is interested in studying the application of voxel phantoms in dose assessment against various radiation exposures.

700

About the Contributors

W. Paul Segars is an assistant professor of radiology and biomedical engineering and a member of the Duke Advanced Imaging Laboratories at Duke University, Durham, North Carolina. Dr. Segars received his PhD in biomedical engineering from the University of North Carolina in 2001. Dr. Segars is among the leaders in the development of simulation tools for medical imaging research where he has applied state-of-the-art computer graphics techniques to develop realistic anatomical and physiological models. Foremost among these are the extended 4D NURBS-based Cardiac-Torso (XCAT) phantom, a computational model for the human body, and the 4D Mouse Whole-Body (MOBY) phantom—a model for the laboratory mouse. These phantoms are widely used to evaluate and improve imaging devices and techniques.

Chengyu Shi is currently an assistant professor in the radiation oncology department at the University of Texas Health Science Center at San Antonio (UTHSCSA), Texas. Dr. Shi received his BS and MS in engineering physics from Tsinghua University, Beijing, China, in 1997 and 2000, respectively. He then attended the Rensselaer Polytechnic Institute where he received his PhD in nuclear engineering in 2004. After spending two years as a clinical medical physics resident at the University of Arkansas for Medical Sciences, he joined the medical physicist group at Cancer Therapy & Research Center at UTHSCSA in December 2005. Dr. Shi is an active member of the American Association of Physicists in Medicine, the American College of Medical Physics, and the American Society for Therapeutic Radiology and Oncology. He is board certified by the American Board of Radiology. His research interests cover various aspects in radiation detection, radiation dosimetry, the Monte Carlo method, medical image segmentation, registration, and 4D simulation.

Il-Young Son is currently working toward his PhD in electrical engineering at the University of California, San Diego, California. He received his MS in electrical engineering at Rensselaer Polytechnic Institute in Troy, New York, under the tutelage of Dr. Birsen Yazici. His previous research work was in the field of image reconstruction and statistical signal processing. His current research interest is in machine learning.

About the Contributors

701

Michael G. Stabin is an associate professor in the Department of Radiology and Radiological Sciences at Vanderbilt University, Nashville, Tennessee. He is a member of the Health Physics Society, the Society of Nuclear Medicine, and the Sigma Xi scientific research and Tau Beta Pi engineering honor societies. He has developed several models, methods, and tools that have become widely used in the nuclear medicine community, including the MIRDOSE and OLINDA/EXM personal computer software codes for internal dose calculations.

Valery Taranenko is currently a postdoctoral scholar at the University of California, San Francisco. Dr. Taranenko received his PhD in health physics from the Russian Academy of Science in 2006. His doctoral thesis focused on radiation dose reconstruction for the general public affected by Mayak—a major weapon-grade plutonium production facility. Dr. Taranenko conducted research in the field of Monte Carlo radiation transport modeling for various European projects while at GSF German National Research Center. As a postdoctoral associate from 2006 to 2008 at the Rensselaer Polytechnic Institute, he was part of a team that developed a novel BREP method for pregnant female phantoms. His research expertise encompasses retrospective dosimetry, nonhuman biota dose estimates, dose calculations in voxel phantoms, and design of semiconductor detectors. Dr. Taranenko presently works in the field of medical physics on the improvement of linac beam model. Benjamin M.W. Tsui is currently the director of the Division of Medical Imaging Physics in the Russell H. Morgan Department of Radiology and Radiological Science and a professor of radiology, electrical and computer engineering, and environment health sciences and biomedical engineering at the Johns Hopkins University. Dr. Tsui received his PhD in medical physics from the University of Chicago in 1977. His research interests include medical imaging physics and instrumentation especially in nuclear and x-ray imaging, quantitative image reconstruction methods, and image quality evaluation. He pioneered the development and application of simulation tools including computer-generated phantoms and Monte Carlo simulation methods to medical imaging. He is the author and coauthor of over 300 scientific papers, review articles, and book chapters. He is a fellow of the IEEE, the IOP, and the AIMBE and is an active member of seven other professional societies including the AAPM, the SNM, the ASNC, the SMRM, the BMES, the AMI, and the SMI. He has also served on the editorial board of several professional journals and many scientific review committees of the U.S. government agencies including the NIH, the DOE and the DOD, U.S. state government agencies, and private foundations.

702

About the Contributors

José Wilson Vieira is a professor at the Escola Politécnica de Pernambuco and at the Centro de Educação Tecnológica de Pernambuco in Recife, Brazil. Dr. Vieira received his DSc in energy and nuclear engineering from the Universidade Federal de Pernambuco in Recife, Brazil, in 2004. His current research interests include digital image processing and computational dosimetry. Dr. Vieira is a member of the Grupo de Dosimetria Numérica.

Scott R. Whalen is pursuing his PhD in medical physics at the University of Florida. He received his BSc cum laude in nuclear engineering and his MSc in nuclear engineering sciences in the Department of Nuclear and Radiological Engineering at the University of Florida. His research interests include radiation dosimetry calculations for IMRT and proton-based radiation therapy utilizing reference and patient-specific models. Whalen is a member of the ANS, the AAPM, and the Alpha Nu Sigma. He is also a medical physicist at the Community Cancer Center of North Florida. Jonathan L. Williams is the director of pediatric radiology at the University of Florida. Dr. Williams is an associate chair of the Department of Radiology and serves as the associate program director for the Residency Training Program. His nonclinical research efforts have been directed to a team approach to defi ning organ-specific dose in CT and fluoroscopy. His current funded project deals with scalable pediatric phantoms for dose assessment. He is a member of the American College of Radiology, the Roentgen Ray Society, the Radiological Society of North America, and the Society for Pediatric Radiology. James Winslow is a PhD candidate in the Department of Nuclear and Radiological Engineering at the University of Florida. His current research interests include the fabrication of detailed anthropomorphic phantoms and advanced dosimeters. James received his bachelor of arts degrees in physics and in speech communication from the State University of New York at Potsdam. He received his MSc in physics from the College of Arts and Sciences and his MSc in engineering science from the College of Engineering at the University of South Florida, where he was a recipient of the Lucent Technologies Fellowship. James was admitted into Sigma Pi Sigma—the National

About the Contributors

703

Physics Honor Society, and Lambda Pi Eta—the National Communication Association Honor Society. He is a member of the ANS, the AAPM, the HPS, and the ACMP. Dagang Wu is currently a research scientist with Halliburton Energy Services, Houston, Texas. Dr. Wu received his BSc and MSc from Southeast University, Nanjing, China, and his PhD from the University of Houston, Houston, Texas, in 1999, 2002, and 2006, respectively, all in electrical engineering. From 2006 to 2007, he was a postdoctoral fellow with the Well Logging Laboratory, University of Houston, Houston, Texas. His current research interests include forward/inverse electromagnetic evaluations of wireline logging, logging while drilling/measure while drilling, and marine electromagnetic logging. Birsen Yazıcı received her BS in electrical engineering and mathematics in 1988 from Bogazici University, Istanbul, Turkey, and her MS and PhD in mathematics and electrical engineering in 1990 and 1994, respectively, both from Purdue University, West Lafayette, Indiana. From September 1994 until 2000, Dr. Yazici was a research engineer at the General Electric Company Global Research Center, Schenectady, New York. During her tenure in industry, she worked on radar, transportation, and industrial and medical imaging systems. Her work on industrial systems received the second best paper award in 1997 given by IEEE Transactions in Industrial Applications. In 2003, she joined the Rensselaer Polytechnic Institute where she is currently an associate professor in the Department of Electrical, Computer and Systems Engineering and in the Department of Biomedical Engineering. Her research interests span the areas of statistical signal processing, inverse problems in imaging, biomedical optics, and radar. She is the recipient of the Rensselaer Polytechnic Institute 2007 School of Engineering Research Excellence Award. She holds 11 U.S. patents. Maria Zankl is a research associate in the Department Medical Physics of the Institute of Radiation Protection at the Helmholtz Zentrum München—German Research Center for Environmental Health (formerly GSF— National Research Center for Environment and Health), in Neuherberg near Munich, Germany. She received her MSc in mathematics from the Technical University Munich in 1981. Her current research interests include the simulation of photon transport in matter using Monte Carlo methods, the calculation of organ dose conversion coefficients for occupational and medical radiation exposures, and the construction of models of the human body for radiation transport calculations. Ms. Zankl is the principal developer of the GSF family of voxel computational phantoms and of the phantoms adopted by the ICRP as the official computational models representing the adult ICRP Reference Male and Reference Female.

704

About the Contributors

Ms. Zankl is member of a committee of the German Radiation Protection Commission and a member of the Task Group DOCAL of ICRP Committee 2. Binquan Zhang received his bachelor’s degree in 2002 and doctorate in 2007 from Tsinghua University, Beijing, China. Dr. Zhang’s PhD dissertation topic was on “Construction of a Chinese Visible Human (CVH) voxel phantom and its application in internal exposure monitoring.” For several months in 2006, Dr. Zhang took part in the IAEA fellowship program of “Occupational Exposure Control” at Institut De Radioprotection et de Sûreté Nucléaire (IRSN), France. From 2007 to 2008, Dr. Zhang was a visiting scholar at the Rensselaer Polytechnic Institute from the China Institute for Radiation Protection, China. His research interests include human virtual models, radiation detection, in vivo measurement, and Monte Carlo simulation. Guozhi Zhang received his BS in bioinformatics from the Huazhong University of Science and Technology, Wuhan, China in 2008. As an undergraduate student, he participated in research at Britton Chance Center for Biomedical Photonics in Wuhan National Laboratory for Optoelectronics for two years and was involved in the Visible Chinese Human research group. Zhang is currently enrolled in a PhD program in medical physics at Katholieke Universiteit Leuven, Belgium.

Juying Zhang has been a PhD student for the past four years at the Rensselaer Polytechnic Institute (RPI), Troy, New York. His doctoral research is related to the development of a pair of adult male and female models using mesh-based geometry modeling techniques. Prior to coming to RPI, Zhang received his BS and MS in engineering physics from Tsinghua University, Beijing, China.

Index A Abstract interface for data analysis (AIDA), 610 AC, see Attenuation compensation ADELAIDE phantom body size radiological protection, 97 young Australians, 92–93 construction of CT image constraints in, 88–89 purpose for, 88 segmentation process, 89–92 effective dose for CT examinations of, 101 EGS4 ready, 99–101 limitation of, 102 organ mass in autopsy data of, 97 formula for, 93 lung, 97 muscle and soft tissue, 95 proportional, 95–96 14 year old torso vs. 15 year old trunk, 94–95 organ shape and position in, 97 human rib cage, 98 intestines, 99 skeleton, 98–99 segmented anatomy images conversion, 99–101 ADELAIDE torso lung mass, 97 mass of organs in, 96–97 modeling, 93 Adipose tissue, 381 segmentation of, 172 Adult human phantom adult female phantom breasts in, 53–54 changes to, 57, 307 uterus in, 54, 57–58, 307–308 adult male phantom breasts in, 53–54 head and trunk, 51–52 idealized model of skeleton, 52–53 major sections, 50–51 modifications to, 49–50 uterus in, 54

development of, 48 for dose estimates, 46 organ dose conversion coefficients for, 76–77 Adult reference male, characteristics of, 382 ALGAM code, 48 Alimentary systems of ORNL newborn, 215 of UF voxel newborn, 215 Anatomical computational phantoms adult female phantom, 307–308 human skeletal tissues, use of, 393 physical feature of, 400 radiological parameters of, 400 Reference Man, 307 stylized computational phantoms, 306–307 Anatomically adjustable adult phantoms applications lung counter, virtual calibration of, 370–371 lung counting efficiency, virtual calibration of, 371–373 Monte Carlo simulations, 371 phoswich scintillation detectors, 371 materials and methods for anatomical references, 352 automatic adjustable phantom, software development, 360–370 candidate anatomical organ models, 350–352 collision detection, organ overlaps, 355–356 deformable phantom, 348–349 mesh deformation algorithm, 354–355 mesh deformation operations, 357–358 mesh preprocessing, 352–353 mesh volume calculation, 353–354 Monte Carlo N-particle eXtended, 360 voxelization procedures, 359–360 Anatomical physical phantoms, 390, 393 Angular response factor, 439 Anthropomorphic phantoms, 489 construction, 165 Fisher–Snyder computational phantom, 474 marrow dose computational phantoms, 475 NURBS computational phantoms (see NURBS computational phantoms) soft, bone, and lung tissue, 401 705

706

ARF, see Angular response factor Asian adult voxel computational phantom, 256 Asian organ density, 258 Attenuation coefficient phantom, 4D MCAT as, 108 Attenuation compensation, 560 Azimuthal angle, 430–431 B Beam-flattening devices, 570 Beam modifiers, simulation geometry for, 641–642 BED, see Biologically effective dose Bertini model, 612 Beta-emitting radionuclides, absorbed dose (Db) for, 44 Beta-particle-emitting substances, 459 BG, see Bust Girth Biacromial breadth, 504 Bioassay, 449, 453–454, 467 Biologically effective dose, 483 Birdcage coil computational phantom, 669–671 specific absorption rate of, 671–676 Bodybuilder software, 491 Boltzmann radiation transport problems, 30 BOMAB, see Bottle Manikin Absorber BOMAB phantom family, 463 Bone counters, 467 Boron neutron capture therapy (BNCT), 250 Bottle Manikin Absorber, 463 Boundary representation, 315, 680 geometry, 348 information in, 5 lung modeling by, 7 modeling anatomy in, 315–316 flowchart for, 316 Bragg peak, 614 Bremsstrahlung photon production, 640 BREP, see Boundary representation BREP-based CAD, 5 BREP-based phantoms, 24–25 Bust Girth, 358 C Calibration phantom, 464 Calypso marker, 592 Calypso system, 593 CAM phantom, 11–12 computer program, 59 surface plots of, 60

Index

Cancer liver, 557 lung, 23, 149, 158, 554 patients, 484 pregnant women, 306 prostate, 157, 405, 553 risk, 567–579, 581 screening, 256 second effects, 32, 681 skin, 155 thyroid, 571 Cardiac computational phantom based on ellipsoids, 109 of 4D NCAT, 115–117 functionality of, 108–109 heart twisting motion modeling by, 109–110 perfusion defects modeling by, 110 Cardiac imaging, 34 Cardiac twisting motion, 121, 127 Caucasian computational phantoms, 222, 256 CDF, see Cumulative distribution function CDR, see Collimator–detector response Cellularity factor, 176 Charged particle equilibrium, 440, 512 Chest–abdomen–pelvis, CT scans of, 490 Chest CT examination, anatomical regions irradiated, 503 Chest phantoms, 406 Chinese Man phantom, see CNMAN phantom Chinese phantoms CNMAN (see CNMAN phantom) CVP phantom 3D anterior views of, 292 organ/tissue masses of, 291 VCH phantom alimentary system, 289 brain, 290 cadaver sectioning and image acquisition, 284 circulatory system, 289 3D reconstruction, 286–287 image registration and segmentation, 284–286 integumentary system, 289 organ masses, 287–288 respiratory system, 289–290 skeletal system, 290 specimen selection and evaluation, 282–283 urogenital system, 290 visualization, 287 Chinese voxel phantoms, see CVP phantom Chromosomal aberration data, 577

Index

CIRS, see Computerized Imaging Reference Systems, Inc. Claycamp, 427 CNMAN phantom construction of, 280–281 3D reconstruction of, 281–282 image processing in, 281 for radiation protection conversion coefficients, 293 lung counting system, 294 MCNP code, 293 whole-body counting system, 294–295 radiation protection using, 296–300 voxel size of, 282 Collimator–detector response, 554 Collision detection algorithm, 356 Collision-kerma, 513 COM, see Component object model Commercial MDCT scanner, 517 Compartment model, 450 Component object model, 360 Compton scattering, 554 Computational anatomical female (CAF) phantom, 59 Computational anatomical male phantom, see CAM phantom Computational anatomical models, 352 Computational exposure model, 164 Computational phantoms, 164, 398, 418, 428, 431, 474, 488, 530, 579, 663 definition, 4 effective dose calculations monoenergetic photons, 530 VIP-Man phantom, 529 ViPRIS background, 526–527 X-ray image generation, 527–529 evolution of, 106 first-generation, 43, 66 future work, 679–683 historical development BREP phantoms, 23–26 stylized phantoms, 7–12 technological advancement and, 33–34 voxel phantoms, 12–23 image quality analysis, 530 internal dose calculations, 472–474 Korean (see Korean computational phantoms) Monte Carlo codes used with EGS, 31 FLUKA, 32 GEANT4, 32 MCNP and MCNPX, 31 PENELOPE, 32

707

purpose of, 27 radiation source terms, 30 random numbers in, 30 needs, 680–683 pediatric microCT-based skeletal, 217 pregnant female (see Pregnant female computational phantoms) pregnant woman computer-aided design, 662–663 purpose of, 65–66 radiation dosimetry, 561 work at NICT on (see NICT computational phantoms) Computed tomography, 370, 487; see also Kilovoltage computed tomography; Megavoltage computed tomography; Multidetector computed tomography; Single photon emission computed tomography; Volume computed tomography adult patients MCNP lattice, 513–514 MCNP, source modeling, 514 MCNPX™ Monte Carlo Code, 512–513 mesh dose scoring, 513–514 normalization and simulation benchmarks, 514–516 voxelized patient models, 513 simulation benchmarks, 517 voxelized patient data, experiments, 519–522 voxelized phantom data, experiments, 517–519 Computed tomography dosimetry computational human phantoms, 489 physical phantoms, 488–489 UF hybrid pediatric phantoms material and methods, 504 results and discussion, 504–508 UF voxel phantoms material and methods, 490–493 results and discussion, 493–504 voxel phantoms, 489–490 Computer-based medical imaging simulation, 106 Computer-generated phantoms advantages, 105 applications of dosimetry calculations, 561–564 image degrading factors, 554–559 image reconstruction and processing methods, 559–561 MCAT, NCAT, and MOBY phantoms, 549 simulation tools, 550–554

708

Computerized Imaging Reference Systems, Inc. ATOM phantoms, 402 synthetic bone material, 401 Computer tomography, see Computed tomography Cone beam CT imaging organ doses calculation from, 158 organ models, 224 pelvic scans, 405 Conformal radiation treatment, 601 Connective tissue, segmentation of, 172 Constructed birdcage coil, 670 Constructive solid geometry of left lung, 6–7 solid-geometry models, 5 Cortical and trabecular bone, ratios between, 176 Corvus system, 647 CPE, see Charged particle equilibrium Cristy computational phantom lungs, 97 organ mass in computation of, 96 muscle and soft tissue, 95 Cristy–Eckerman phantoms, 10 CRT, see Conformal radiation treatment CSG, see Constructive solid geometry CT, see Computed tomography CTCREATE program, 638 CTDI body phantom contiguous axial scan, 519 helical scan, 519 pencil chamber exposure measurements vs. MCNPX simulation, 518 CT imaging cancer risk, radiation epidemiological studies, 509 CT on-rail, 592–593 Cumulative distribution function, 514 CVP phantom 3D anterior views of, 292 development of, 280 organ/tissue masses of, 291 radiation protection using irradiation geometries for, 296 organ dose conversion coefficients for, 296–300 D DCFs, see Dose conversion factors 3D conformal therapy, 568 3DCRT, see 3D conformal therapy DDE, see Deep dose equivalent

Index

DDREF, see Dose and dose rate effectiveness factor Deep dose equivalent, 425 Deformable breast phantoms, anthropometric data of, 358 Deformable phantoms components, 348 postures of, 373 process, flowchart of, 349 type of, 348 Deformation modeling, physics-based methods for, 681–682 Deformation process, liver and right lung, 356 Deformed child computational phantoms, 231–232 Detector quantum efficiency, 403 Diagnostic radiology, 549 DICOM, see Digital imaging and communications in medicine Diethylene triamine pentaacetic acid, 451 Different dose scoring methods, 578 Digital imaging and communications in medicine, 615 DLP, see Dose length product DMLC, see Dynamic multileaf collimators 4D NCAT phantom, 561 3D NURBS, 319 Dose and dose rate effectiveness factor, 577 Dose calculation system basic concept of, 247 elemental compositions–dose accuracy relation, 248–250 schematic diagram of, 247–248 Dose conversion coefficients, 76–77 Dose conversion factors AFs, 475 radionuclides, 476 standard organ masses, 476 Dose estimation, 4 adult phantom of use in, 46 anthropomorphic computational phantoms for, 337 application in radiotherapy, 247–250 CT, 21 derived from stylized phantoms, 12 external (see External dose calculation) improvement in, 32, 34 internal (see Internal dose calculation) Monte Carlo code systems for UCPIXEL, 234 UCRTP, 236 UCSAF, 234–235 UCWBC, 235

Index

two-dosimeter algorithms for computational phantom, 427 idealized photon beams, 430–434 MIRD-type mathematical model, 427 optimal algorithm, development of, 428–430 steam generator channel head, 434–435 whole-body counting peak efficiency to photon energy, 246–247 water models, 245–246 Dose length product, 488 Dose–volume histograms, 594, 639 Dosimetric database structure, 438 4D physical computational phantoms dynamic thorax computational phantom, 597–598 quality assurance system for advanced radiotherapy, 598 DQE, see Detector quantum efficiency 4D radiation treatment, 603 3D radionuclide distribution models, 108 3D rendering, types of, 144, 147 DTPA, see Diethylene triamine pentaacetic acid DVH, see Dose–volume histograms DVH curves, 601, 647 4D VIP-Man development, 599 4D virtual computational phantoms, 598 4D VIP-Man, 599 intensity modulated radiation therapy, 599 NCAT computational phantom, 599 4D virtual model-guided radiation therapy, 600 Dynamic multileaf collimators, 603 E ECUT values, isodose distributions, 646 EDE, see Effective dose equivalent EDE calculator, graphical user interface of, 443–444 Effective dose definition of, 164 ICRP recommendations, 164 Effective dose equivalent, 425 Effective half-life, physical/biological, 451–452 EGS3, see Electron Gamma Shower version 5 EGS code system, see Electron gamma shower code system EGS4 code system, 100, 636 EGS external photons simulations, 325 EGS4 implementation of, 638 low-energy electron transport, 640 photon, 640

709

EGSnrc, 418 EGS system, history of, 633 EGS user codes, 638 Electromagnetic hadronic energy, 611 Electromagnetic interactions, Monte Carlo implementation of, 611 Electron energy, 477 Electron gamma shower code system, 634 BLOCK DATA subprogram, 634 computer code system, 634 electron transport, 634 Monte Carlo computer simulation program, 635 negative USTEP, 649–650 simulation parameters, 639–640 user code, 637 Electron Gamma Shower version 3, 634 Electron transport methods, 477 ENDF/B-VI photoatomic interaction data, 513 Energy cutoff effects, 645 Environmental dosimetry, 78 Environmental exposures gamma ray fields, calculation of, 414–415 Helmholtz–Zentrum München, 414 human body, 414 Monte Carlo codes, organ calculations, 417–421 radionuclides distribution, 413 simulation of, 415–417 voxel computational phantoms, organ calculations, 417–421 Environmental photon three-step method, 414 Epidemiological value, 578 Epoxy-resin mixtures, 393 EPRI EDE calculator, 442–445 Exposure model, 164 4D Extended cardiac-torso phantom, see 4D XCAT phantom External beam radiation therapy scattered doses, 567–568 computational phantoms, 570–571 photons, 568–570 proton therapy, 570 External organ dose calculation external electron exposure, 239–240 external photons bladder wall, 236, 238 thyroids, 236–238 External electron dosimetry, 153 External neutron dosimetry, 153–154 External photon dosimetry, 152–153 External proton dosimetry, 154

Index

710

Extra thoracic airways FAX06 phantom of, 183 segmentation of, 172 F FAX06 (Female Adult voXel) phantoms; see also Female computational phantoms absorbed/equivalent dose calculations in, 165 anatomical corrections, 170 database, 168 equivalent dose to RBM and BSC in for external exposure to photons, 189–194 spongiosa, 187–188 head and arms, 170 and ICP Reference, 183–184, 186 and MAX06 phantoms ICRP103 recommendations, 170 segmentation of new organs and tissues, 171–176 segmented organs and tissues, 170–171 RBM equivalent dose in (see RBM dosimetry) segmentation of images, 168–170 skeletal tissue segmented for, 173–175, 179, 181–182 two-step development of, 164 FBP, see Filtered backprojection FDA, see Food Drug Administration FDTD, see Finite difference time-domain Female adult computational phantoms, 377 Female breast phantoms, 360 Female computational phantoms, 504 breast sizes of, 358, 360 deformable phantom operations, software GUI for, 369 deformation of abdomen, 231 3D front and back views, 369 3D images of, 229 doses for, 236 high-resolution human voxel model using, 234 organ masses of, 223, 230 organ/tissue masses, densities, and volumes, 366–368 ovary dose, 498 Female GSF voxel phantoms, 71–72 Female reference voxel computational phantom, 379 Fiber optic coupled dosimeter systems, 395 Filtered backprojection, 554 Finite difference time-domain simulations, 668 Fisher–Snyer adult phantom, 7

FLUKA, 32 FOC dosimeter systems, see Fiber optic coupled dosimeter systems Food Drug Administration, 662 Frank (GSF male voxel phantom), 71 Fricke-to-water dose conversion factor, 636 Full width at half maximum (FWHM), 371 G Gallbladder, segmentation of, 172 Gamma radiation, absorbed dose for, 44 Gaussian assumption, 539 Gaussian energy broadening, 371 Gaussian pulses, 670 GEANT4, 32 Geant4-based system, 619 Geant4 models, 614 Geant4 Monte Carlo code, 625 C++ object-oriented architecture5, 609 dose characteristics of, 608 implementation, 612 Geant4 simulated treatment head, 624 GEB, see Gaussian energy broadening G4HadronElastic, 612 GI tract of mathematical model, 147 of MIRD adult phantom, 56 of VIP-Man computational phantom, 141–142 Golem, 71, 380 G4ParametrisedNavigation, 616 G4PreCompoundModel, 612 Graphical user interface, 360 Graves’ disease, 483 Gross tumor volume, 641, 648 GSF models key organs, radiation doses, 522 normalized effective dose plot, 520–521 GSF voxel phantoms adult voxel computational phantom, 69 applications in radiation dosimetry bone dosimetry, 76 environmental geometries, 78 idealized geometries, 76–77 for medical imaging, 78–80 Monte Carlo codes, 76 organ cross-fire, 80 specific absorbed fraction, 80–81 “Baby,” 69 construction of bone marrow distribution estimation, 68–69 CT images, 66–67

Index

drawing of borderline, 68 image segmentation, 67 manual drawing of organs, 68 morphological methods for, 67–68 organ identification number, 68 region growing procedure, 67 threshold method for, 67 Donna and Irene, 71 family characteristics of members of, 70 masses of main organs of, 73–75 female, 72 male, 71–72 pediatric, 71 VHP image, 71 G4Track, 610 G4Transportation, 610 GTV, see Gross tumor volume GUI, see Graphical user interface G4UserSteppingAction, 618 G4VPVParameterisation, 616 H HDRK-Man construction of, 268 3D views of, 273 organ masses of, 273–274 skeleton of, 272 Head-torso model, 20 Health physics dosimetry, VIP-Man phantom, 152 Heart computational phantom, see Cardiac computational phantom Heart motion effects, cardiac imaging, 108 Heart walls, segmentation of, 172 Helium nucleus, 454 Hermaphrodite phantoms, 49 Heterogeneous phantoms idealizations, 47 improvements in, 47 MIRD phantoms (see MIRD phantoms) High-definition Reference Korean-Man, see HDRK-Man High-resolution human voxel model, 234 Homogenous test phantom, 515 Hounsfield number, 575, 616–618 Human alimentary tract model, 378 Human body computational models of, 348 computational phantoms, radiation transport codes, 377 Human computational phantoms, 656

711

Human respiratory tract model, 378 Hybrid computational phantoms development of, 106 flexibility of, 508 I ICNIRP, see International Commission on Nonionizing Radiation Protection ICRP, see International Commission on Radiological Protection ICRP-89 compatible phantoms, intercomparison of, 680–681 ICRP Reference Man organ masses for, 143, 145–146 organs and tissue masses for, 166–167 skeleton and RBM mass distributions from, 140–142 vs. VIP-Man computational phantom, 151 ICRU, see International Commission on Radiation Units and Measurements IDL, see Interactive Data Language IGRT, see Image-guided radiation therapy Image generation algorithms, 543 Image-guided radiation therapy, 591 4D virtual computational phantom, 599 external photon beam treatment, 597 future trends of, 604 imaging method, 592 ionizing imaging vs. nonionizing imaging, 603 preplanning, live planning, and postplanning, 604 tracking vs. predicting, 603–604 treatment planning and delivery, 594 Imaging performance assessment of CT, 511, 520 IMBA, see Integrated molecules for bioassay analysis ImPACT, see Imaging performance assessment of CT IMRT, see Intensity modulated radiation therapy Integrated molecules for bioassay analysis calibration alpha counting, 462 beta counting, 462 gama counting, 463 lung counting, 464–466 organ counting, 466–467 quality assurance, 467 radiation, measure of, 461 whole-body counting, 463–464

Index

712

wound counting, 467 direct methods dose estimates, 461 lung counters, 459–460 organ counting, 460–461 radionuclide, 457 whole-body counting, 458–459 wound counting, 461 indirect methods alpha emitters, 454–455 beta emitters, 455–456 dose estimate, 457 X-ray and gamma emitters, 456–457 lung absorption model, 451 Intensity modulated radiation therapy, 568, 599, 602, 639 Interactive Data Language, 381 Internal dose calculation SAFs for photons, 240–243 S values for positron emitters, 242, 244 to urinary bladder wall, 245 Internal electron dosimetry using VIP-Man phantom, 154 Internal photon dosimetry using VIP-Man phantom, 155 International Commission on Nonionizing Radiation Protection, 656 International Commission on Radiation Units and Measurements, 360 International Commission on Radiological Protection, 348, 425, 474 adult reference computational phantoms, 418 adult reference male/female, description of general features, 381–384 limitations, 385–386 skeleton, special features of, 384–385 female computational phantom, 419 phantom properties, 378 radiation dosimetry, applications, 386–387 reference adult voxel phantoms, 382 reference mass values, 380 skeleton volumes of, 380 Sv/Gy, effective dose in, 420 voxel computational phantoms, 378 voxel models, conversion coefficients, 419 Ionizing radiation imaging, 592 phantoms for, 4 IQTs, optimal dose, 544–545

J Japanese voxel computational phantoms dose calculation external (see External dose calculation) internal (see Internal dose calculation) JAEA computational phantoms body thicknesses and widths, 223–224, 227 CT pictures, 224 distances between organs, 228 organ masses, 223, 225, 227 physical characteristics, 223 voxel size, 222 NICT computational phantoms, 226 applications in electromagnetic dosimetry, 233 of arbitrary posture, 232 deformed child computational phantoms, 231–232 3D images of, 229 higher resolution, 233 male and female models, 229–230 whole-body pregnant woman computational phantom, 230–231 K KATS, see Korean Agency for Technology and Standards Kidney model, 55 Kilovoltage computed tomography, 595 King–Spiers factors, 185 Korean Agency for Technology and Standards, 394 Korean computational phantoms HDRK-Man, 268, 272–274 Korean reference organ data, 256–260 KORMAN and KORWOMAN, 262–265 KTMAN-1 and KTMAN-2, 265–268 stylized computational phantoms, 260–262 Korean male phantom tissue-equivalent materials, 400 Korean Reference Man Project, 256 Korean reference organ data, 256–260 Korean stylized phantom adult male and female organ mass, 261 3D frontal and rear views, 262 Korean reference organ volume and, 260 skeleton of, 261 Korean voxel computational phantoms KORMAN and KORWOMAN 3D reconstruction of, 264

Index

internal organ masses, 264–265 MR tomographic images, 262–263 KTMAN-1 and KTMAN-2 3D frontal views of, 272 ET1 region surfaces, 267 MR and CT image acquisition, 265–266 organ mass calculation, 268–271 segmentation of images, 266–268 kVCT, see Kilovoltage computed tomography L Laguerre–Gauss Hotelling observer, 539–540, 542 LAR, see Lifetime attributable risk Lawrence Livermore National Laboratory’s phantom internal organs, 465 versions, 464 Left lung BREP modeling of, 7 CSG modeling of, 6–7 LET, see Linear energy transfer Lifetime attributable risk, 582 Linear energy transfer, 576 Lung counter nuclide and photon energy, 372 in vivo measuring device, 370 Lung counting detectors, 459 Lung counting system, 294 Lung nodules, respiration motion of, 557 Lymphatic nodes, segmentation of, 173 M Magnetic field emission, 656 Magnetic resonance imaging, 592 Major airway, model of, 58 Male computational phantom child figure computational phantoms, 231–232 3D front and back views, 369 ICRP Caucasian, 236 organ/tissue masses, densities, and volumes, 230, 363–365 Male reference computational phantom, 379 Male skull visual inspection, 359 Mandible mesh preprocessing, 353 MANTISSUE3-6 voxel phantom, 166 Marinelli beaker, 463 Mass spectrometry techniques, 456 Mathematical Cardiac Torso phantom, 10

713

Mathematical computational phantoms, 66, 478 MATLAB 7.4®, 356 Maximum-likelihood expectationmaximization, 560 Maximum permissible exposure, 656 MAX06 (Male Adult voXel) phantom absorbed/equivalent dose calculations in, 165 database, 166–167 equivalent dose to RBM and BSC in, 187 and FAX06 phantom ICRP103 recommendations, 170 segmentation of new organs and tissues, 171–176 segmented organs and tissues, 170–171 heart region, 179 and ICP Reference bone volumes between, 177 organ and tissue masses between, 168, 178, 181, 186 salivary glands, 178 skeleton tissues of, segmented, 177–178 two-step development of, 164 MCAT phantom 3D MCAT phantom large breasts, 556 myocardial SPECT imaging, 556 4D MCAT phantom anatomy(ies), 107, 113 as attenuation coefficient/transmission phantom, 108 breathing motion model, 110–113 cardiac computational phantom in, 108–109 3D surface renderings of, 107–108 imaging simulation, 107 myocardial lesion in, 111 organs modeled in, 107 as radiopharmaceutical uptake/emission phantom, 109 volume curve for LV of, 110 MCDOSE/MCSIM geometry input data file, 650 MCNP transport code system, 31 MCNPX, see Monte Carlo N-Particle eXtended MCNPX simulation, 325, 516 MCNPX transport code system, 31 MDCT, see Multidetector computed tomography MDCT colonography protocols, 405 Medical imaging simulation, 105

714

Medical internal radiation dose, 377, 390, 511 adult model exterior and cutaway views of, 337–338 kidneys in, 55 body models, 483 committee age-specific models of head and brain, 54 kidney model, 55 mathematical model, 427 phantoms computational tool, 390 development, 46, 255 idealizations, 47 intestines of, 99 mathematical organs shape computation by, 98 MIRD-5 phantom, 7, 59 organ mass in, 95–96 Megavoltage computed tomography, 595 Mesh-based adult male/female phantom collision detection, 355–356 Mesh-based BREP phantoms, 26 Mesh deformation operations, 357 Microsoft Word document, automatic report generation, 444 MIRD, see Medical internal radiation dose MLC, see Multileaf collimators ML-EM, see Maximum-likelihood expectationmaximization 3D MOBY phantom atherosclerotic plaque, 553 patient imaging data, 564 SPECT/CT image simulation, 553 4D MOBY phantom, 558 bb-MRI cardiac data set of, 127 cone-beam x-ray CT images generated from, 131 creation of, 126 inspiratory and expiratory motions in, 129 respiratory motion in, 128 small animal imaging research using, 131–132 SPECT coronal image generated from, 130 Moliere multiple-scattering theory, 634 Moliere theory, 640 Monoenergetic photons conversion coefficients, 421 Monte Carlo-based CT scanner model, 512 Monte Carlo systems, 350, 435, 516, 575, 617–618, 633 computational phantom, 569 computer programs, 474

Index

for dose calculation, 402, 426, 607–608, 643, 649 UCPIXEL, 234 UCRTP, 236 UCSAF, 234–235 UCWBC, 235 EGS, 31 FLUKA, 32 GEANT4, 32 geometry, 614 implementation of VIP-Man into, 148 IMRT planning, 645 internal dose assessment, 47–48 MCNP and MCNPX, 31 PENELOPE, 32 purposes of, 27, 43 radiation, 30 radiotherapy, 644 random numbers in, 30 for RBM dosimetry, 76 transport calculations, 608 for voxel computational phantoms, 76 Monte Carlo N-Particle eXtended, 360 Monte Carlo particle transport method, 426 Monte Carlo radiation, 358 transport codes, 493 transport methods, 489 transport simulations, 370 Monte Carlo simulation, 435, 462, 464, 466, 512, 554, 571, 577, 608 advanced geometries, 682 anthropomorphic phantoms, 489 4D Monte Carlo simulation, 621, 623 patient computational phantoms, 571–573 phantoms, density range, 643 for proton therapy, 579 of radiation transport, 473 tissue-absorbed dose, estimate of, 493 treatment, beam entering, 573 MOSFET dosimetry system, 395 Mouse phantom, see 4D MOBY phantom MPE, see Maximum permissible exposure MRI, see Magnetic resonance imaging MRI RF coil ninth-month model, positioning of, 671 MSCT examinations, 497 Multicompartment kidney model, 55 Multidetector computed tomography, 392 Multileaf collimators, 568, 641 MVCT, see Megavoltage computed tomography Myocardial perfusion SPECT, 555, 560 Myocardial SPECT images, 558

Index

N National Institute of Information and Communication Technology (NICT), 222 National Radiological Protection Board, 405, 511 National Research Council of Canada system, 638 NCAT phantom coronary slice, 557 3D NCAT phantom anatomical variations, 563 CT image simulation, 550 pelvic area, 554 sample simulated images, comparison of, 552 SPECT/CT image simulation, 551 transaxial images of, 562 4D NCAT phantom anatomical variations modeling in, 119 anterior and posterior views of, 115 beating heart computational phantom of, 115–117 cardiac model of, 118 vs. geometry-based MCAT, 121 organs modeled in, 115–116 patient motions modeling using, 115–116 perfusion defects, 117 respiration modeling in, 118–119 lung nodule, 557 NCAT torso anatomy, 115 NCRP, optimization process, 428 Neutron RBE data, 577 NICT computational phantoms, 226 applications in electromagnetic dosimetry, 233 of arbitrary posture, 232 deformed child, 231–232 3D images of, 229 higher resolution, 233 male and female models, 229–230 whole-body pregnant woman, 230–231 Nine-month pregnant woman models, 663 Nonionizing radiation biological effects on fetuses, 306 phantoms for, 4 protection electric field strength, 667 equivalent current source calculation, 658–659 impedance method, 659–662 magnetic field distribution measurement, 656–658

715

nine-month pregnant woman models, 663 pregnant woman computational phantoms, 655–656, 664, 666 pylon, extracted equivalent current sources, 664 tissue, electrical properties of, 663 vertical direction, layer-averaged current density, 666 radiofrequency-based devices, 306 Non prewhitening matched filter with eye filter, 526 Nonuniform rational B-splines, 149, 348, 488, 680 NORMAN phantom, 23 NPWE, see Non prewhitening matched filter with eye filter NRCC system, see National Research Council of Canada system NRPB, see National Radiological Protection Board Nuclear medicine applications, 471 dose calculational approaches, 482–484 dosimetry calculations, 477–481 tomographic data, quantification of, 481–482 Nuclides, movement of, 450 NURBS, see Nonuniform rational B-splines NURBS-based BREP phantoms, 26 4D NURBS-based cardiac-torso phantom, 114 NURBS-based hybrid newborn phantom, 206 NURBS-based mouse beating heart model, 128 NURBS computational phantoms adult male and female, 338 anterior views of, 338 methods to create, 339–340 organ masses of, 340–342 photon SAFs for, 340, 343–345 radiation transport calculations in, 340, 344 screen image from, 339 NURBS surface for branching structures, 321 shape modification, 114 O Optically stimulated luminescence, 394 Organ-absorbed dose; see also Organ dose calculations effective dose, comparison of, 499–502 normalized values of, 506–507 Organ dose calculations from ADELAIDE and EGS4, 101 from cone-beam CT imaging for IGRT, 158

716

from interventional cardiological examinations, 157 Monte Carlo simulations for, 138 patient computational phantoms computational phantom types, 571–573 dose scoring, 577–578 Monte Carlo implementation, 573–576 radiation quality factors, 576–577 second malignancies, risk estimates, 578–579 from proton radiation treatments, 158 from respiration management in IGRT, 158 from SPECT and PET brain imaging, 156 from X-ray radiographs, 156 Organ IDs, list of, 350–351 Organ mesh, triangular face, 352 Organ self-absorption, SAFs for, 80 Organ-specific mass fractions, 361–362 Organ surfaces defined using NURBS, 149 deformation of, 149–150 ORNL phantoms age-dependent effective doses, 494–496 with arm bones, 492 kidneys in, 55 male and female, 496 organs of, 261–262 UF phantoms, 496 ORNL stylized phantoms, 56 OSL, see Optically stimulated luminescence Outer bremsstrahlung, 459 Overlapping regions, 480 P Paired-image radiation transport, 185 Parameter Reduced Electron-Step Algorithm, 640 Patient’s respiratory motion, 591 PEC, see Perfect electric conductor Pediatric phantoms, 10 computational, 490 development of, 48 external features of, 51 microCT-based skeletal computational, 217 motivation for producing, 87 ORNL series, 491 UF series B, 490–491 whole-body phantoms based on NURBS surfaces (see UF hybrid newborn phantoms) Pediatric voxel phantoms, 21–22 Pencil ion chamber, 517

Index

PENELOPE, 32 Penelope simulations, 326 Perfect electric conductor, 669 Peritoneal cavity model, 58 Personnel radiation dose calculation, 426 effective/radiation doses calculation, 435 input frame, 440 Monte Carlo simulations, 438 output–energy spectra at dosimeters, 441 output frame, 440 photon energy spectra, 441 software package for, 435–442 PET, see Positron emission tomography Phantoms manufacturers, 395 Photon beam, 426, 430 incident beam directions, distributions of, 430–431 single-direction, 433 Photon energies, 429, 435–437 Photon entrance, 426 Photon exposure dose equivalent rates, 422–423 Photon radiations, 421 Photon radiation therapy, 575 Photon simulations, 574 Photon transport model, 512 Physical anatomical phantoms, 391 commercially produced phantoms CIRS—atom phantom series, 401–403 CIRS—3D sectional torso phantom, 403 Kyoto Kagaki phantoms, 406 RANDO phantom, 403–405 university-based phantoms Korean male phantom, 399–400 University of Florida, 396–399 Physical anthropomorphic phantoms, 517 Physical exposure model, see Exposure model Physical phantoms, 4, 26–29 attributes of design and development, 393–394 dosimetry, integration of, 394–396 tissue-equivalent materials, 392–393 CT scans, 394 dosimetry measurements, 398 human doses, accurate determination of, 389–390 image quality/radiation dosimetry, 389 Korean male, 399 pediatric tissues, age-dependent density, 398 quality assurance, 389 University of Florida, 397 Piece-wise linear conversion curve, 643 PIRT, see Paired-image radiation transport Planar imaging, 479

Index

Planchet counting, 462 PMMA, see Polymethylmethacrylate Polygonal meshes, 349 Polygonal models, 34 Polymethylmethacrylate, 393, 613 Polyurethane, 399 Positron emission tomography, 478, 592 Posture model, variable, 233 PRDC, see Personnel radiation dose calculation Pregnant female computational phantoms, 476, 668, 671 anatomical data for, 317 breasts, 318 BREP modeling method for, 316–317 2D and 3D anatomical information from, 318–319 CT images, 309–310 fetus, anatomical components of, 317 “hybrid” approach to, 308 Monte Carlo modeling of external photons simulations, 325 Monte Carlo codes, 323 Penelope simulations, 326 segmented CT data of, 319–320 SILVY, 308–309 skin surface model of fetus model in, 321 stylized, 308 surface-geometry based modeling approach, 315 tissue elemental compositions, 323 tomographic (see Tomographic pregnant woman model) uterus and placenta, 317–318 voxelization procedure for, 322–323 Pregnant women model, see Nine-month pregnant woman models; Tomographic pregnant woman model; WTMD pregnant woman models Pregnant worker, radiation risk to, 306 PRESTA, see Parameter Reduced Electron-Step Algorithm Prostate radiation treatment, 157–158 Prostate, segmentation of, 173 Proton beam, 570 propagatation, 613 scanning, 570 therapy, 581, 614 Proton therapy, 579 Proton therapy treatment, 574 Pulmonary vessels, 483

717

Q QA, see quality assurance Quality assurance system for advanced radiotherapy, 598 Quantitative scintigraphic imaging, 480 QUASAR, see Quality assurance system for advanced radiotherapy Quattro Pro, 427

R Radiation dosimetry anatomical computational phantoms for (see Anatomical computational phantoms) applications in, 391–392 challenges in, 4 for medical imaging, 78–80 in nuclear medicine, 44 objectives of, 3 Radiation oncology, 549, 567 Radiation protection, 450 equivalent dose quantity in, 164 in high-energy physics accelerator facility, 27 of pregnant patients and embryo/fetus, 306 principles of, 450 Radiation therapy; see also External beam radiation therapy; Image-guided radiation therapy; Intensity modulated radiation therapy; Monte Carlobased radiotherapy; Photon radiation therapy Geant4 Monte Carlo package benchmarking, 613–614 dynamic dose calculation, 620–623 Geant4 philosophy, 609–610 Geant4 physics, 610–612 photon beam therapy, 614 proton beam therapy, 614–615 static dose calculation, 615–620 treatment head modeling, 614–615 objectives of, 3 simulation, CT data conversion, 642–644 simulation, 3D rectilinear geometry, 640–641 treatment planning, 644 Radioimmunotherapy, 549 Radionuclides, 421–423, 451, 458 Radiopharmaceutical activity, 472 Radiopharmaceuticals standard dose estimates computational phantoms, 476 image-based approaches, 477 Radiotherapy, see Radiation therapy RANDO phantoms, 390, 402–404

718

RANDO soft tissue-equivalent material, 404 Rapid Prototyping and Manufacturing technique, 399 RBM dosimetry, 76 for external exposure to photons 3CF/FDR data, 189–190 exposure models, 191–192 REX/REGINA effective dose, 192–194 8 SP cluster method, 189 using VIP-Man phantom, 155–156 Red bone marrow segmentation, 33 Reference adult male/female phantoms, 383–384 Reference Asian Man, anthropomorphic parameters of, 257 Reference computational phantoms, 379 Reference female computational phantoms, 382 Regulatory information summary, 426 Relative biological effectiveness, 576 Respiratory motion, 110 anatomical variations, 113 diaphragm movement modeling, 111 in 4D MOBY phantom, 128 in 4D NCAT phantom, 118–119 in 4D VIP-Man phantom, 150–151 in 4D XCAT phantom, 121, 123–124 in MCAT phantom, 111–112 simulating time-dependent deformation by, 149–150 Rhinoceros software, 320 RIS, see Regulatory information summary Risk computational phantoms, 582 ROI, see Region of interest RPI-P computational phantom series anatomical parameters of organ centroids, 330–332 organs volumes vs. reference data, 326–330 visual inspections, 330 EGS4-VLSI for, 325 implementation in MCNPX, 325 RP&M technique, see Rapid Prototyping and Manufacturing technique S SAF, see Specific absorbed fraction Salivary glands, segmentation of, 173 SAR, see Specific absorption rate Segmentation of pixels cerebrum, 139 RBM distribution, 139–141 VHP-cryosection images, 138–139 Segment treatment table, 641

Index

Shielding, 458 Signal-to-noise ratio, 558 Similitude phantoms, see Pediatric phantoms Single photon emission computed tomography, 592 Size-and posture-adjustable phantoms, 348 Skeletal dosimetry development of, 184 methods of, 182–185 skeleton tissue segmentation for, 173–175 Skeleton tissue segmentation for skeletal dosimetry, 173–175 special features of, 384–385 theoretical volume distribution of, 175 Skin voxels, 380 SNR, see Signal-to-noise ratio Snyder adult phantom, 49 SOBP, see Spread-out Bragg peak Software interfaces, 445 Software program, 349 Solid-geometry modeling techniques BREP method, 5, 7 CSG method, 5–6 8 SP cluster method based on vertebral bone sample, 189–190 graphical display of, 188 for skeletal dosimetry, 185–188 Specific absorbed fraction, 378 for organ cross-fire, 80 for organ self-absorption, 80 for self-irradiation of thyroid, 81 Specific absorption rate, 668 SPECT, see Single photon emission computed tomography SPECT/CT systems CT imaging, 559 image reconstruction, 561 MOBY phantom, 559 phantom image, 561 Spiers factors, 185 Spread-out Bragg peak, 619 SRS, see Stereotactic radiosurgery Steam generator channel head, 434 Stereotactic radiosurgery, 639 STT, see Segment treatment table Stylized model, torso of, 572 Stylized phantoms, 8–9 ADAM and EVA, 58–59 CAM phantom, 11–12, 59–60 historical developments of family phantoms, 49–54 MIRD committee and, 54–56 MIRD-5 phantom, 45–48, 56 models of trunk of adult human, 45

Index

ORNL stylized model, 56 pediatric phantoms, 48 photon sources, 45 pregnant female phantom, 57–58 Korean, 260–262 of lower abdomen, 56–57 of major airway, 58 for medical applications, 10–11 MIRD-5 phantom, 7, 10 of nasal cavity, 58 of pregnant female, 57–58 SURFdriver surface reconstruction program, 115 T Teeth mesh quality, comparison of, 353 Therapeutic radiopharmaceuticals, 472 Therapy Body phantoms, 406 Thermoluminescence detectors, 569 Thermoluminescence dosimeters, 394, 427, 489 Thoracic scan protocol/simulation, 520–522 Thoracic tissues, classification of, 108 Three-dimensional medical imaging techniques, 137 Thyroid cancer, 581–582 Thyroid counters, 466 Tissue-absorbed dose, estimate of, 493 Tissue-equivalent materials, 393 Tissue kerma, energy spectrum of, 440 Tissues, dielectric properties of, 663 Tissue-weighting factors, 578 TLD, see Thermoluminescence dosimeters Tomographic phantoms, 13 Tomographic pregnant woman model compression of, 313 coordinates and voxel IDs for, 313 SAF values for, 314 tissues/organs in, 311–312 Tomotherapy planning station, 594 Torso phantom, 21, 403 TPS, see Treatment planning systems Transmission phantom, 108 Treatment planning systems, 639 Triangular mesh, 354 Triple energy window scatter correction, 479 U UF hybrid newborn phantoms advantages of, 214–215 development procedure extended, 206 NURBS modeling, 201–202

719

polygonization of voxel phantom, 200–201 voxelization of polygon computational phantom, 202–203 for newborn male and female, 216 3D rendering, 208–209 organs and tissues, 208, 210–213 vs. ICRP89, 214 older members of, 216–217 segmentation of, 200 standardization of alimentary and respiratory systems., 204–205 anthropometric data sets, 203 ICRP publication 89 values, 204 opposite sex using original female phantom, 205–206 voxelization algorithm, 206–207 UF newborn voxel phantom, 200 Ultrasound images, postplan/live plan/ preplan, 594–596 Urinary excretion, 453 Uterine model, 57–58 V Variable tube current simulation, 514–515 VCH, see Visible Chinese Human VCH phantom alimentary system, 289 brain, 290 cadaver sectioning and image acquisition, 284 circulatory system, 289 3D reconstruction, 286–287 image registration and segmentation, 284–286 integumentary system, 289 organ masses, 287–288 radiation protection using LAHET code system, 295 organ dose conversion coefficients for, 296–300 photons, neutrons, and protons, 295–296 special considerations for, 296 respiratory system, 289–290 skeletal system, 290 specimen selection and evaluation, 282–283 urogenital system, 290 visualization, 287 VCT, see Volume computed tomography Vertebral trabecular bone, 3D mCT image of, 187 Vertex normal modification, 355 VHP, see Visible Human Project

Index

720

VHP-cryosection color photographs properties in, 139 segmentation of, 138 VIP-Man Chest Phantom, 348 VIP-Man computational phantom, 572, 576 anterior 2D visualization for, 142 applications in external electron dosimetry, 153 external neutron dosimetry, 153–154 external photon dosimetry, 152–154 health physics dosimetry, 152 internal electron dosimetry, 154 internal photon dosimetry, 155 radiological imaging, 156–157 radiotherapy, 157–158 RBM dosimetry, 155–156 development of 3D visualization of, 144, 147 labeling, 144 Monte Carlo codes implementation, 148 segmentation of pixels, 138–144 selection of original images, 137–138 4D VIP-Man deformation of organ surfaces, 149–150 revoxelization, 150–151 voxel data conversion, 149 3D visualization for RBM of, 142 GI tract in, 141–142 vs. ICRP Reference Man, 151 organ and tissue masses of, 143, 145–146 RBM mass distributions from, 140–142 skeleton from, 142 VIP-Man model, 572 VIP-Man phantom, 350, 682 VIP-Man voxel phantoms, 21 ViPRIS, see Virtual photographic and radiographic imaging simulator Virtual photographic and radiographic imaging simulator, 526 Visible Chinese Human, 280 Visible Human Project, 71, 137–138, 526 Visible photographic man computational phantom, see VIP-Man computational phantom Visualization Toolkit, 360 VMC Monte Carlo code, 609 Volume computed tomography, 392 Voxel computational phantoms, 573, 581 Voxel geometry, 640 Voxelization process of newborn left lung, 206–207 Voxelization tool, 358

VOXELMAN (torso voxel phantom), 166 Voxel model, torso of, 572 Voxel phantoms, 12 adult Chinese male, 22 anatomical differences between, 33 based on CT images, 66 development history, 33, 66 development of, 137 family of, 13–19 GOLEM and LAURA, 20 GSF (see GSF voxel phantoms) Japanese (see Japanese voxel computational phantoms) NORMAN, 20–21 pediatric, 21–22 segmented, 166 software to automatically construct, 250 from tomographic images, 13 torso phantom, 21 VIP-Man, 21 VOXTISS8 phantom linear voxel dimensions of, 167 organs and tissue masses for, 166–167 VTK, see Visualization Toolkit W Walk-through metal detectors, 655 Water equivalent methods, 617 Whole-body counting system, 294–295 Whole-body pregnant woman computational phantom, 230–231 Whole body retention, urinary and fecal excretion rates, 452 WTMD pregnant woman models, 665 WTMDs, see Walk-through metal detectors X 4D XCAT phantom anatomical variations modeling by, 124–125 cardiac motions of, 121 imaging simulations using, 125 male and female anatomies of, 122 NURBS basis for, 124 organs modeled in, 123 respiratory motion of, 121, 123–124 X-ray beam, 551 X-ray image quality analysis, 526 X-ray photons, 551 X-ray radiographic imaging, 525

Index

data generation, 535–538 image quality analysis observer computational phantoms, 530–532 performance measurement, 534–535 task descriptions, 532–534

721

Laguerre–Gauss Hotelling observer, 538–539 optimization, 525–526 Z Zubal phantoms, 20, 33

Identification of organs in each slice of a 2D pixel map

Registration of all slices

Finished 3D voxel phantom

FIGURE 1.3 Steps to create a voxel phantom illustrated using the Visible Human cadaver image data set.

(b) FIGURE 2.6b Surface plots of the CAM phantom. (b) The close-up view of the facial details.

FIGURE 3.2 GSF male voxel phantoms; Golem (left), Frank (middle), and Visible Human (right).

FIGURE 3.3 GSF female voxel phantoms: Donna, Helga, Irene, and Laura (from left to right).

XCAT

Evolution of computerized phantoms...

MOBY

MIRD

MCAT

NCAT

FIGURE 5.2 Original MIRD phantom1 and phantoms developed in our laboratory that approach more ideal computerized phantoms.

FIGURE 5.23 Male (left) and female (right) anatomies of the 4D XCAT phantom.

(a)

(b)

(c)

FIGURE 6.2 VIP-Man in 3D views showing (a) whole-body skin and skeletal structure; (b) details of internal organs with lungs in red, stomach in gold, ULI in purple, kidney in red, liver in maroon, LLI in brown, etc.; and (c) details of the head and brain containing skull in gold, white matter in white, gray matter in gray, nerve in blue, spinal cord in gold, thyroid in red, skin in white, etc. VTK was used in the surface rendering of the voxelized images.

FIGURE 7.11 Frontal and lateral views of the skeletons and internal organs of the FAX06 and the MAX06 phantoms (adipose and muscle tissues removed).

1-year male

1-year female

5-year male

5-year female

10-year male

10-year female

15-year male

15-year female

FIGURE 8.5 Series of UF hybrid pediatric male (left) and female (right) phantoms: newborn, 1, 5, 10, and 15 year old phantoms.

Otoko

Onago

FIGURE 9.1 Japanese voxel models developed at JAEA.

JM

JF

SAR(W/kg) 10–4

10–3

10–2

10–1

100

101

FIGURE 9.11 Example of SAR distribution in a model exposed to plane wave (80 MHz). Incident power density is 1 mW cm−2.

(a) KTMAN-1

(b) KTMAN-2

FIGURE 10.4 The 3D frontal views of the MR-based (a) KTMAN-1 and CT-based (b) KTMAN-2 where skin and residual soft tissue was made semitransparent to visualize internal organs and skeleton.

Eye Brain Oral mucosa

Salivary glands

Thyroid

Esophagus Bone

Thymus Breast Heart

Lung

Liver Stomach

Spleen

Small intestine

Kidneys

Bladder Gonads

Colon

FIGURE 10.6 The 3D whole-body frontal view with semitransparent skin (left) and major organs and tissues (right) of the HDRK-Man. (From Kim, C.-H., Phys. Med. Biol., 53, 4093, 2008. With permission.)

(a) FIGURE 11.3 3D rendering of CNMAN: (a) whole body; (b) skeleton.

(b)

(a)

(b)

FIGURE 11.5 Illustration of the computed distribution of the red and yellow bone marrows within a sample color image of the VCH phantom: (a) before segmentation; (b) after segmentation.

(a)

(b)

(c)

FIGURE 11.7 (a) Internal organs, (b) whole-body skeleton, and (c) vascular system of the VCH phantom.

(a) Updated

(b)

(c)

FIGURE 12.10 The finalized RPI-P9 models. (a) Rendering of 3D models of the RPI-P3, P6 and P9 (from left to right) plotted from Rhinoceros. (b) Rendering of the voxelized RPI-P9 model before translated into the MCNPX. (c) A direct MCNPX geometry plot showing a cross section view of the 3 mm voxel model of the RPI-P9 implemented for Monte Carlo radiation transport calculations. Visual inspections allow the anatomical geometries to be verified.

(a)

(b)

(c)

(d)

FIGURE 13.4 Sample images of NURBS standardized models, scaled to the ICRP 89 recommended reference organ masses, developed for internal and external dose assessment: (a) newborn female model, (b) 5-year-old male model, (c) 10-year-old female model, and (d) 15-year-old male model.

(a)

(b)

FIGURE 14.11 3D front and back views of the phantoms: (a) the male; (b) the female.

FIGURE 15.1 Frontal view of the ICRP-AM (left) and ICRP-AF (right), the voxel models representing the ICRP adult Reference Male and Reference Female.

10th

50th

90th

(a)

10th

50th

90th

(b)

FIGURE 21.8 Lateral views of 3D rendering of the 10th (left), 50th (middle), and 90th (right) weight percentile UFH-NURBS 15 year old (a) male and (b) female phantoms. (From Lee, C., Med. Phys., 35, 2366, 2008. With permission.)

x-ray source

3D NCAT phantom

Relative counts, η(Ei)

6.00E+05 5.00E+05 4.00E+05 3.00E+05 2.00E+05 1.00E+05 0.00E+00

(a)

(d)

(b)

Array of detectors

(c)

0

20

40

60

80

100

120

140

Energy, Ei, in (keV)

(e)

FIGURE 24.1 High-resolution 2D x-ray planar and 3D CT image simulation using the 3D NCAT phantom. (a) The head and torso portion of the 3D NCAT phantom. (b) A typical x-ray imaging configuration with the 3D NCAT phantom replacing a patient. (c) A typical 120 kVp x-ray spectrum. (d) Sample 2D frontal planar images with, from left to right, 30, 60, and 90 kVp monoenergetic x-ray energy beam. (e) Sample transaxial slices from the 3D CT image reconstructed from multiple 2D planar projections of different views around the phantom.

IIiac vessels

Hypogastric vessels

Lymph nodes Vas deferens

Anterior (a)

Bladder

Seminal vesicles

Prostate

Posterior

(c)

Plaque

Defec

(b)

(d)

FIGURE 24.5 Example of simulated abnormalities. (a) Lung with trachea tree and a lung nodule, (b) an atherosclerotic plaque in the aorta region of the 3D MOBY phantom, (c) a perfusion defect modeled as a wedge in the left ventricular myocardial wall, and (d) detailed anatomy of the pelvic area with lymph nodes where metastases of prostate cancer first appear in the 3D NCAT phantom.

FIGURE 25.2 Proton therapy treatment head (at Massachusetts General Hospital) to illustrate the complexity of the geometry. Two particle tracks are drawn as well.

(a)

(b)

(c)

(d)

(e)

(f)

(g) FIGURE 26.6 (a) kVCT; (b) MVCT; (c) correlated images axial view; (d) correlated images coronal view; (e) correlated images sagittal view; (f) DVH comparison; and (g) isodose comparison.

1 Gy 3 Gy 5 Gy 7 Gy 9 Gy 11 Gy 13 Gy 15 Gy 17 Gy

FIGURE 27.3 Axial, coronal, and sagittal views of dose distributions calculated using the Geant4 Monte Carlo system at Massachusetts General Hospital.46 The patient was treated with three fields (columns 1–3) for a paraspinal tumor. The treatment plan was done with a commercial planning system.

(a)

(b)

(c)

(d)

(e)

(f)

FIGURE 28.2 Isodose distributions calculated using different AE and ECUT values: (a) a paraspinal treatment plan calculated with AE = 0.7 MeV, ECUT = 0.7 MeV (thick lines) and ECUT = 2.5 MeV (thin lines); (b) a mantle field plan calculated with AE = 0.7 MeV, ECUT = 0.7 MeV (thick lines) and ECUT = 1.0 MeV (thin lines); (c) a mantle field plan for AE = 0.7 MeV, ECUT = 0.7 MeV (thick lines), and ECUT = 1.5 MeV (thin lines); (d) a nasocavity plan with a 6 MeV electron beam calculated using ECUT = AE = 0.521 MeV (thin lines), AE = 0.7 MeV, and ECUT = 1.0 MeV (thick lines). The isodose lines in (d) are 4%, 10%, 30%, 50%, and 90% of prescribed dose; (e) same nasocavity geometry as in (d) irradiated by a 1 cm × 1 cm 4 MV photon beam. Thick lines are for AE = ECUT = 0.521 MeV and thin lines are for AE= 0.7 MeV and ECUT = 1.0 MeV; (f) same as in (e) with thick lines representing AE = ECUT = 0.521 MeV and thin lines representing AE = ECUT = 0.7 MeV. The isodose lines shown in (e) and (f) are 1%, 5%, 10%, 30%, and 50%, normalized to the maximum dose for AE = ECUT = 0.7 MeV. (From Ma, C.-M. et al., Phys. Med. Biol., 47, 1671, 2002. With permission.)

FIGURE 29.7 Nine-month pregnant woman models. (From Wu, D.G., Phys. Med. Biol., 52(19), 5735, 2007. With permission.)

Handbook of Anatomical Models for Radiation Dosimetry (Series in Medical Physics and Biomedical Engineering)

Medical physics and biomedical engineering

Handbook of Photonics for Biomedical Science (Series in Medical Physics and Biomedical Engineering)

Proton Therapy Physics (Series in Medical Physics and Biomedical Engineering)

Nuclear Medicine Physics (Series in Medical Physics and Biomedical Engineering)

Radiation Physics for Medical Physicists (Biological and Medical Physics, Biomedical Engineering)

Radiation Physics for Medical Physicists (Biological and Medical Physics, Biomedical Engineering)

Biophotonics (Biological and Medical Physics, Biomedical Engineering)

The Physics of Modern Brachytherapy for Oncology (Series in Medical Physics and Biomedical Engineering)

Intensity-Modulated Radiation Therapy (Series in Medical Physics and Biomedical Engineering)

Introduction to Radiological Physics Radiation Dosimetry

Radiation Protection in Medical Physics

Advanced Materials and Techniques for Radiation Dosimetry

Topology in Molecular Biology (Biological and Medical Physics, Biomedical Engineering)

Topology in Molecular Biology (Biological and Medical Physics, Biomedical Engineering)

E. coli in Motion (Biological and Medical Physics, Biomedical Engineering)

Practical Biomedical Signal Analysis Using MATLAB® (Series in Medical Physics and Biomedical Engineering)

Topology in Molecular Biology (Biological and Medical Physics, Biomedical Engineering)

Contemporary IMRT: Developing Physics and Clinical Implementation (Series in Medical Physics and Biomedical Engineering)

Biomechanics of the Brain (Biological and Medical Physics, Biomedical Engineering)

Handbook of Medical Imaging: Processing and Analysis (Biomedical Engineering)

Handbook of Optical Sensing of Glucose in Biological Fluids and Tissues (Series in Medical Physics and Biomedical Engineering)

Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine (Series in Medical Physics and Biomedical Engineering)

Clinical Engineering Handbook (Biomedical Engineering)

Physics and Engineering of Radiation Detection

Clinical Engineering Handbook (Biomedical Engineering)

Fundamentals of Cryobiology: Physical Phenomena and Mathematical Models (Biological and Medical Physics, Biomedical Engineering)

Intelligent and Adaptive Systems in Medicine (Series in Medical Physics and Biomedical Engineering)

Biomedical Engineering Handbook

Artificial Sight: Basic Research, Biomedical Engineering, and Clinical Advances (Biological and Medical Physics, Biomedical Engineering)

Emergent Computation: Emphasizing Bioinformatics (Biological and Medical Physics, Biomedical Engineering)

Handbook of Anatomical Models for Radiation Dosimetry (Series in Medical Physics and Biomedical Engineering)

Medical physics and biomedical engineering

Handbook of Photonics for Biomedical Science (Series in Medical Physics and Biomedical Engineering)

Proton Therapy Physics (Series in Medical Physics and Biomedical Engineering)

Nuclear Medicine Physics (Series in Medical Physics and Biomedical Engineering)

Radiation Physics for Medical Physicists (Biological and Medical Physics, Biomedical Engineering)

Radiation Physics for Medical Physicists (Biological and Medical Physics, Biomedical Engineering)

Biophotonics (Biological and Medical Physics, Biomedical Engineering)

The Physics of Modern Brachytherapy for Oncology (Series in Medical Physics and Biomedical Engineering)

Intensity-Modulated Radiation Therapy (Series in Medical Physics and Biomedical Engineering)

Introduction to Radiological Physics Radiation Dosimetry

Radiation Protection in Medical Physics

Advanced Materials and Techniques for Radiation Dosimetry

Topology in Molecular Biology (Biological and Medical Physics, Biomedical Engineering)

Topology in Molecular Biology (Biological and Medical Physics, Biomedical Engineering)

E. coli in Motion (Biological and Medical Physics, Biomedical Engineering)

Practical Biomedical Signal Analysis Using MATLAB® (Series in Medical Physics and Biomedical Engineering)

Topology in Molecular Biology (Biological and Medical Physics, Biomedical Engineering)

Contemporary IMRT: Developing Physics and Clinical Implementation (Series in Medical Physics and Biomedical Engineering)

Biomechanics of the Brain (Biological and Medical Physics, Biomedical Engineering)

Handbook of Medical Imaging: Processing and Analysis (Biomedical Engineering)

Handbook of Optical Sensing of Glucose in Biological Fluids and Tissues (Series in Medical Physics and Biomedical Engineering)

Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine (Series in Medical Physics and Biomedical Engineering)

Clinical Engineering Handbook (Biomedical Engineering)

Physics and Engineering of Radiation Detection

Clinical Engineering Handbook (Biomedical Engineering)

Fundamentals of Cryobiology: Physical Phenomena and Mathematical Models (Biological and Medical Physics, Biomedical Engineering)

Intelligent and Adaptive Systems in Medicine (Series in Medical Physics and Biomedical Engineering)

Biomedical Engineering Handbook

Artificial Sight: Basic Research, Biomedical Engineering, and Clinical Advances (Biological and Medical Physics, Biomedical Engineering)

Emergent Computation: Emphasizing Bioinformatics (Biological and Medical Physics, Biomedical Engineering)

Recommend Documents