Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine Series in Medical Physics and Biomedical Engineering (IoP, 2003)

THERAPEUTIC APPLICATIONS OF MONTE CARLO
CALCULATIONS IN NUCLEAR MEDICINE

Series in Medical Physics and Biomedical Engineering
Series Editors:
C G Orton, Karmanos Cancer Institute and Wayne State University, Detroit,
USA
J A E Spaan, University of Amsterdam, The Netherlands
J G Webster, University of Wisconsin-Madison, USA
Other books in the series

Minimally Invasive Medical Technology
J G Webster (ed)
Intensity-Modulated Radiation Therapy
S Webb
Physics for Diagnostic Radiology
P Dendy and B Heaton
Achieving Quality in Brachytherapy
B R Thomadsen
Medical Physics and Biomedical Engineering
B H Brown, R H Smallwood, D C Barber, P V Lawford and D R Hose
Monte Carlo Calculations in Nuclear Medicine: Applications in Diagnostic Imaging
M Ljungberg, S-E Strand and M A King (eds)
Introductory Medical Statistics, third edition
R F Mould
Ultrasound in Medicine
F A Duck, A C Barber and H C Starritt (eds)
Design of Pulse Oximeters
J G Webster (ed)
The Physics of Medical Imaging
S Webb
Forthcoming titles in the series

The Physical Measurement of Bone
C Langton and C Njeh (eds)
The Physics of MRI
W T Sobol
The Physics of High Dose Rate Brachytherapy
D Baltas, H Kreiger and N Zamboglou
Series in Medical Physics and Biomedical Engineering
THERAPEUTIC APPLICATIONS
OF MONTE CARLO CALCULATIONS
IN NUCLEAR MEDICINE
Edited by
Habib Zaidi, PhD
Division of Nuclear Medicine,
Geneva University Hospital,
Switzerland
George Sgouros, PhD

Memorial Sloan-Ketting Cancer Center,
New York,
USA
Institute of Physics Publishing

Bristol and Philadelphia
# IOP Publishing Ltd 2003
All rights reserved. No part of this publication may be reproduced, stored in

a retrieval system or transmitted in any form or by any means, electronic,
mechanical, photocopying, recording or otherwise, without the prior permis-
sion of the publisher. Multiple copying is permitted in accordance with the
terms of licences issued by the Copyright Licensing Agency under the
terms of its agreement with Universities UK (UUK).
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
ISBN 0 7503 816 8
Library of Congress Cataloging-in-Publication Data are available
Series Editors:
C G Orton, Karmanos Cancer Institute and Wayne State University, Detroit,
USA
J A E Spaan, University of Amsterdam, The Netherlands
J G Webster, University of Wisconsin-Madison, USA
Commissioning Editor: John Navas

Production Editor: Simon Laurenson
Production Control: Sarah Plenty
Cover Design: Victoria Le Billon
Marketing: Nicola Newey and Verity Cooke
Published by Institute of Physics Publishing, wholly owned by The Institute

of Physics, London
Institute of Physics Publishing, Dirac House, Temple Back, Bristol BS1 6BE, UK
US Office: Institute of Physics Publishing, The Public Ledger Building, Suite
929, 150 South Independence Mall West, Philadelphia, PA 19106, USA
Typeset by Academic+Technical Typesetting, Bristol

Printed in the UK by MPG Books Ltd, Bodmin, Cornwall
The Series in Medical Physics and Biomedical Engineering is the official book
series of the International Federation for Medical and Biological Engineering
(IFMBE) and the International Organization for Medical Physics (IOMP)
IFMBE
The IFMBE was established in 1959 to provide medical and biological
engineering with an international presence. The Federation has a long history
of encouraging and promoting international cooperation and collaboration in
the use of technology for improving the health and life quality of man.
The IFMBE is an organization that is mostly an affiliation of national
societies. Transnational organizations can also obtain membership. At present
there are 42 national members and one transnational member with a total
membership in excess of 15 000. An observer category is provided to give
personal status to groups or organizations considering formal affiliation.
Objectives
. To reflect the interests of the affiliated organizations.
. To generate and disseminate information of interest to the medical and
biological engineering community and international organizations.
. To provide an international forum for the exchange of ideas and concepts.
. To encourage and foster research and application of medical and bio-
logical engineering knowledge and techniques in support of life quality
and cost-effective health care.
. To stimulate international cooperation and collaboration on medical
and biological engineering matters.
. To encourage educational programmes which develop scientific and
technical expertise in medical and biological engineering.
Activities
The IFMBE has published the journal Medical and Biological Engineering
and Computing for over 35 years. A new journal Cellular Engineering was
established in 1996 in order to stimulate this emerging field in biomedical
engineering. In IFMBE News members are kept informed of the develop-
ments in the Federation. Clinical Engineering Update is a publication of
our division of Clinical Engineering. The Federation also has a division for
Technology Assessment in Health Care.
Every three years the IFMBE holds a World Congress on Medical Physics
and Biomedical Engineering, organized in cooperation with the IOMP and
the IUPESM. In addition, annual, milestone and regional conferences are
organized in different regions of the world, such as Asia Pacific, Baltic,
Mediterranean, African and South American regions.
The administrative council of the IFMBE meets once or twice a year and is
the steering body for the IFMBE. The council is subject to the rulings of the
General Assembly which meets every three years.
For further information on the activities of the IFMBE, please contact Jos
A E Spaan, Professor of Medical Physics, Academic Medical Centre,
University of Amsterdam, PO Box 22660, Meibergdreef 9, 1105 AZ, Amster-
dam, The Netherlands. Tel: 31 (0) 20 566 5200. Fax: 31 (0) 20 691 7233. E-
mail: [email protected]. WWW: http://www.ifmbe.org/.
IOMP
The IOMP was founded in 1963. The membership includes 64 national
societies, two international organizations and 12 000 individuals. Member-
ship of IOMP consists of individual members of the Adhering National
Organizations. Two other forms of membership are available, namely
Affiliated Regional Organization and Corporate members. The IOMP is
administered by a Council, which consists of delegates from each of the
Adhering National Organizations; regular meetings of Council are held
every three years at the International Conference on Medical Physics
(ICMP). The Officers of the Council are the President, the Vice-President
and the Secretary-General. IOMP committees include: developing countries,
education and training; nominating; and publications.
Objectives
. To organize international cooperation of medical physics in all its
aspects, especially in developing countries.
. To encourage and advise on the formation of national organizations of
medical physics in those countries which lack such organizations.
Activities
Official publications of the IOMP are Physiological Measurement, Physics
in Medicine and Biology and the Series in Medical Physics and Biomedical
Engineering, all published by the Institute of Physics Publishing. The
IOMP publishes a bulletin Medical Physics World twice a year.
Two council meetings and one General Assembly are held every three years at
the ICMP. The most recent ICMPs were held in Kyoto, Japan (1991), Rio de
Janeiro, Brazil (1994), Nice, France (1997) and Chicago, USA (2000). These
conferences are normally held in collaboration with the IFMBE to form the
World Congress on Medical Physics and Biomedical Engineering. The IOMP
also sponsors occasional international conferences, workshops and courses.
For further information contact: Hans Svensson, PhD, DSc, Professor,
Radiation Physics Department, University Hospital, 90185 Umeå, Sweden.
Tel: (46) 90 785 3891. Fax: (46) 90 785 1588. E-mail: Hans.Svensson@radfys.
umu.se. WWW: http://www.iomp.org.
Contents
LIST OF CONTRIBUTORS xv
PREFACE xviii
1 THE MONTE CARLO METHOD: THEORY AND

COMPUTATIONAL ISSUES 1
1.1. Introduction 1
1.1.1. Early approaches to Monte Carlo simulations 1
1.1.2. Conceptual role of Monte Carlo simulations 2
1.2. Random number generation 3
1.2.1. Linear congruential generators 5
1.2.2. Lagged-Fibonacci generators 6
1.3. Photon transport 6
1.3.1. Photon cross section data 7
1.3.2. Simulating interaction of photons with matter 12
1.4. Electron transport 13
1.4.1. Stopping power data 15
1.5. Analogue sampling 16
1.5.1. Direct method 17
1.5.2. Rejection method 17
1.5.3. Mixed methods 17
1.6. Non-analogue sampling ‘variance reduction’ 18
1.6.1. Photon-specific methods 18
1.6.2. Electron-specific methods 20
1.6.3. General methods 21
1.7. Summary 22
Acknowledgements 23
References 23
2 MONTE CARLO TECHNIQUES IN NUCLEAR

MEDICINE DOSIMETRY 28
2.1. Introduction 28
vii
viii Contents
2.2. Monte Carlo techniques in medical radiation physics 29

2.3. Applications of Monte Carlo techniques in nuclear
medicine imaging 31
2.3.1. Diagnostic nuclear medicine imaging 31
2.3.2. Therapeutic nuclear medicine imaging 36
2.4. Monte Carlo techniques in nuclear medicine dosimetry 38
2.4.1. Calculation of absorbed fractions 38
2.4.2. Derivation of dose-point kernels 39
2.4.3. Pharmacokinetic modelling 41
2.5. Monte Carlo techniques in radiation protection 42
2.5.1. Shielding calculations 43
2.5.2. Characterization of detectors and radiation
monitoring instruments 43
2.5.3. Radiation dose calculations to staff 44
2.5.4. Revisiting release criteria for patients administered
therapeutic doses 44
2.6. Future applications of Monte Carlo 45
2.6.1. Patient-specific dosimetry and treatment planning 45
2.6.2. On-line PET monitoring of radiotherapy beams 46
References 48
3 MEDICAL IMAGING TECHNIQUES FOR RADIATION

DOSIMETRY 55
3.2. Determination of the total mass. Determination of spatial
distribution of attenuation coefficient, and of mass 56
3.3. Segmentation of patient data into VoI 57
3.4. Determination of the distribution of radioactivity in both
space and time 58
3.4.1. Quantitative whole-body imaging 59
3.4.2. Quantitative imaging with conjugate views 60
3.4.3. Quantitative imaging with SPECT 62
3.4.4. Effect of reconstruction algorithm on SPECT
imaging 66
3.4.5. Effect of attenuation, scatter and blurring
correction on SPECT imaging 68
3.4.6. Time series of images 71
3.4.7. Methods used to register multimodality images 72
3.4.8. Quantitative imaging with PET 74
3.4.9. Effect of attenuation and scatter correction on PET
imaging 76
3.4.10. Reconstruction algorithms for fully-three-
dimensional PET 76
3.5. Auxiliary contribution from PET 77
Contents ix
3.6. Treatment planning 78

Acknowledgements 79
References 79
4 COMPUTATIONAL METHODS IN INTERNAL RADIATION

DOSIMETRY 84
4.2. Radiation quantities and units 85
4.2.1. Stochastic versus deterministic quantities 85
4.2.2. Definitions of dosimetric quantities 85
4.3. The MIRD schema 89
4.4. Patient- and position-specific dosimetry 95
4.4.1. Patient-specific dosimetry: case study—radioiodine
therapy of metastatic thyroid cancer 95
4.4.2. Adaptation of the MIRD schema to patient-specific
dosimetry 98
4.4.3. Position-specific dosimetry: calculation of
non-uniform dose distributions 100
4.5. Summary 101
References 102
5 MATHEMATICAL MODELS OF THE HUMAN ANATOMY 108

5.1.1. Early approaches to dose assessment 108
5.1.2. Need for mathematical models in dose assessment 109
5.1.3. Conceptual role of mathematical phantoms 109
5.2. Historical developments 110
5.2.1. Simple models of Brownell, Ellett and Reddy 110
5.2.2. Early models developed by Snyder 110
5.2.3. The Snyder–Fisher phantom 111
5.2.4. The MIRD-5 Phantom 112
5.2.5. Photon and electron transport 113
5.2.6. Similitude and paediatric phantoms 114
5.2.7. Development of MIRDOSE codes 115
5.3. The current stylized models 115
5.3.1. The ORNL phantom series 115
5.3.2. The MIRD stylized models 116
5.3.3. Stylized models of the lower abdomen 118
5.3.4. Other stylized models of the human anatomy 120
5.3.5. Use of stylized models in therapeutic nuclear
medicine 120
5.4. Tomographic models 121
5.4.1. Methods of construction 122
5.4.2. Review of representative models 123
x Contents
5.4.3. Comparisons with stylized mathematical models 124

5.5. Summary 126
References 126
6 MONTE CARLO CODES FOR USE IN THERAPEUTIC

NUCLEAR MEDICINE 133
6.2. Historical developments 134
6.2.1. The ‘Reference Man’ phantoms and the ALGAMP
code 136
6.2.2. Development of MIRDOSE codes 138
6.3. Public domain Monte Carlo codes 139
6.3.1. The EGS code 139
6.3.2. The MCNP code 139
6.3.3. The ETRAN code 140
6.3.4. ITS 141
6.3.5. The GEANT code 141
6.3.6. Other Monte Carlo codes 141
6.4. Limitations of current nuclear medicine dose calculations 142
6.4.1. Introduction 142
6.5. Improvements in models for nuclear medicine therapy 144
6.5.1. Current approaches to patient-specific dose
calculations 144
6.5.2. Innovation in software development tools 149
6.5.3. Parallel computing aspects 150
6.5.4. Towards clinical applications of ‘on-the-fly’ Monte
Carlo-based dosimetry calculations 152
6.6. Summary 153
Acknowledgements 153
References 153
7 DOSE POINT-KERNELS FOR RADIONUCLIDE

DOSIMETRY 158
7.2. Methods used to generate dose point-kernels 160
7.3. Review of dose point-kernels available for radionuclide
dosimetry 162
7.3.1. Photons 162
7.3.2. Electrons 163
7.3.3. Combined approaches 165
7.3.4. Impact of the medium 166
7.4. A general approach for use of dose point-kernels in
calculations 168
7.4.1. Absorbed fractions on a cellular scale 169
Contents xi
7.4.2. Absorbed fractions on a millimetre scale 169

7.4.3. Absorbed fractions on an organ scale 170
7.5. Conclusions 170
References 171
8 RADIOBIOLOGY ASPECTS AND RADIONUCLIDE

SELECTION CRITERIA IN CANCER THERAPY 175
8.2. Radiobiological effects 175
8.2.1. Molecular lesions 175
8.2.2. Cellular responses 176
8.2.3. Tissue responses 179
8.2.4. Radiation quality 181
8.3. Targeting principles in radionuclide therapy 182
8.3.1. Choice of radionuclide 182
8.3.2. Half-life 185
8.3.3. Choice of vector or ligand 186
8.3.4. Properties of targets 187
8.4. Experimental therapeutics 187
8.4.1. particle emitters 187
8.4.2. particle emitters 189
8.4.3. Auger electron emitters 191
References 194
9 MICRODOSIMETRY OF TARGETED RADIONUCLIDES 202

9.2. Alpha emitters 204
9.2.1 Monte Carlo simulation of energy deposition by
particle emitters 205
9.2.2. Applications of microdosimetry to particle
emitters 209
9.3. Auger electron emitters 213
9.3.1. Monte Carlo simulation of Auger decays and their
energy deposition 215
9.3.2. Targeted therapy with Auger electron emitters 218
9.4. Future directions 219
References 221
10 THE MABDOSE PROGRAM FOR INTERNAL

RADIONUCLIDE DOSIMETRY 228
10.1. Introduction—the need for better dosimetry 228
10.2. Dosimetry software design considerations 229
10.2.1. Dosimetry considerations 230
10.2.2. Dosimetry approaches 232
xii Contents
10.3. Historic development of MABDOSE 233

10.3.1. Version 1 233
10.3.2. Version 2 234
10.3.3. Version 3 235
10.3.4. Version 4 235
10.4. MABDOSE: from start to finish 236
10.4.1. Overview 236
10.4.2. The target lattice 236
10.4.3. Declaring source volumes 237
10.4.4. Time-activity data and the generation of cumulated
activities 239
10.4.5. Selection of a radionuclide 240
10.4.6. Simulation of radioactive decay 241
10.4.7. Simulation—cross section lookup tables 241
10.4.8. Simulation—radiation transport 242
10.4.9. Dose display 243
10.5. Future developments 244
10.5.1. Automated or semi-automated image segmentation 244
10.5.2. Generalization of the atomic composition of target
voxels 245
10.5.3. Incorporation of electron simulation 245
10.5.4 Incorporation of nonlinear models 245
10.6. Summary 246
References 246
11 THE THREE-DIMENSIONAL INTERNAL DOSIMETRY

SOFTWARE PACKAGE, 3D-ID 249
11.2. Background to 3D-ID development 250
11.3. Programming aspects/development philosophy 250
11.4. Monte Carlo-based implementation 251
11.5. Point-kernel based implementation 252
11.6. Individual modules 252
11.7. Considerations in clinical implementations 257
11.8. Future potential uses 258
References 258
12 EVALUATION AND VALIDATION OF DOSE

CALCULATION PROCEDURES IN PATIENT-SPECIFIC
RADIONUCLIDE THERAPY 262
12.2. Dose–volume information versus mean–average
organ/tumour dose 262
Contents xiii
12.3. Monte Carlo simulation of imaging systems 263

12.3.1. Monte Carlo programs for radionuclide imaging 264
12.3.2. Software phantoms 265
12.4. Evaluation of image quality degradation by Monte Carlo
methods 266
12.4.1. Evaluation of planar 131 I imaging 267
12.4.2. Evaluation of SPECT 131 I imaging 269
12.4.3. Other radionuclides 272
12.5. Evaluation of absorbed dose calculation procedures by
Monte Carlo methods 273
12.6. Evaluations based on direct measurements 275
12.6.1. Thermoluminescence dosimeters 275
12.6.2. Biological dosimeters 276
12.6.3. Other dosimetry methods 278
12.6.4. Phantoms for experimental measurements 279
12.6.5. In vivo verifications—experimental animals 279
12.6.6. In vivo verifications—humans 280
12.6.7. Stack phantom for realistic RNT dosimetry
verifications 280
References 281
13 MONTE CARLO METHODS AND MATHEMATICAL

MODELS FOR THE DOSIMETRY OF SKELETON AND
BONE MARROW 286
13.1.1. Anatomy of the skeletal system 286
13.1.2. Dosimetry of the skeletal system 288
13.2. Trabecular bone models 290
13.2.1. Research at the university of Leeds 290
13.2.2. Model of trabecular bone in MIRD Pamphlet 11 290
13.2.3. Skeletal dosimetry in ICRP Publication 30 291
13.2.4. Spherical model of trabecular bone 291
13.2.5. The Eckerman trabecular bone model 292
13.2.6. The Bouchet et al. trabecular bone model 293
13.2.7. Differences between the three-dimensional model
of Bouchet et al. and the one-dimensonal model of
Eckerman 293
13.2.8. Model of a trabecular bone lesion 295
13.2.9. High-resolution imaging for trabecular bone
dosimetry 295
13.3. Cortical bone models 296
13.3.1. Research at the University of Leeds 296
13.3.2. The ICRP-30 cortical bone model 297
xiv Contents
13.3.3. The Eckerman cortical bone model 297

13.3.4. The Bouchet et al. cortical bone model 298
13.3.5. The research of Akabani 298
13.4. Improving the models: patient-specific dosimetry 299
13.4.1. The clinical experience—ability to predict toxicity
with current dosimetric models 299
13.4.2. Correcting the models using patient-specific data 301
13.4.3. In-vivo high-resolution imaging 303
13.5. Summary 304
References 305
14 MONTE CARLO MODELLING OF DOSE DISTRIBUTIONS

IN INTRAVASCULAR RADIATION THERAPY 310
14.2. Candidate radionuclides and technologies 312
14.3. Radiation dosimetry studies 313
14.3.1. Vessel wall dose—calculations 313
14.3.2. Vessel wall dose—measurements 317
14.3.3. Dose to body organs in case of a balloon rupture 317
14.4. Radiation effects—artery walls 321
References 322
15 THE MONTE CARLO METHOD AS A DESIGN TOOL IN

BORON NEUTRON CAPTURE SYNOVECTOMY 324
15.1.1. Rheumatoid arthritis and current methods of
treatment 325
15.2. Boron neutron capture synovectomy 326
15.3. Neutron beam design calculations for BNCS 328
15.3.1. Optimal beam energy 329
15.3.2. Optimal beam design 329
15.4. Experimental characterization of the BNCS beam 334
15.5. Experimental investigation of BNCS in an animal model 338
15.6. Whole-body dosimetry for BNCS 340
15.7. Summary and conclusions 343
References 344
16 SUMMARY 348
BIOSKETCHES 350
INDEX 361
List of contributors
Habib Zaidi, PhD

Division of Nuclear Medicine, Geneva University Hospital, CH-1211
Geneva, Switzerland
George Sgouros, PhD

Memorial Sloan-Kettering Cancer Center, Department of Medical Physics,
1275 York Avenue, New York 10021, USA
Pedro Andreo, PhD, DSc, Professor

Department of Medical Radiation Physics, University of Stockholm–
Karolinska Institute, PO Box 260, SE-17176 Stockholm, Sweden
Manuel Bardies, PhD

Institut National de la Santé et de la Recherche Médicale (INSERM), Unité
463, 9 quai Moncousu, Nantes 44035, France
Wesley Bolch, PhD

Department of Nuclear and Radiological Engineering, University of Florida,
Gainesville, FL 32611-8300, USA
Lionel Bouchet, PhD

University of Florida Brain Institute, Department of Neurological Surgery,
100 S Newell Dr., Gainesville, FL 32610-0265, USA
A Bertrand Brill, PhD

Department of Radiology and Radiological Sciences, Vanderbilt University,
1161 21st Avenue South, Nashville, TN 37232-2675, USA
Yuni Dewaraja, PhD

Radiology Department, The University of Michigan Medical Center, Ann
Arbor, MI 48109-0552, USA
xv
xvi List of contributors
Dennis Duggan, PhD

Department of Radiation Oncology, B902 The Vanderbilt Clinic, Vanderbilt
University Medical Center, Nashville, TN 37232-5671, USA
Keith Eckerman, PhD

Life Sciences Division, Oak Ridge National Laboratory, 1060 Commerce
Park, MS 6480, TN 37831, USA
David Gierga, PhD

Department of Nuclear Engineering, Massachusetts Institute of Technology,
Room NW14-2207, 150 Albany Street, Cambridge, MA 02139-4307, USA
John Humm, PhD

Timothy K Johnson, PhD

Anschutz Cancer Pavilion, Department of Radiation Oncology, Box 6510,
Mail Stop F706, Aurora, CO 80010-0510, USA
Amin I Kassis, PhD

Department of Radiology, Harvard Medical School, Boston, MA 02115-
5729, USA
Katherine S Kolbert, MS
Kenneth F Koral, PhD

Internal Medicine Department, The University of Michigan Medical Center,
Ann 3480 Kresge III, 204 Zina Pitcher Place, Ann Arbor, MI 48109-0552, USA
Cheuk S Kwok, PhD

Hong Kong Polytechnic University, Department of Optometry and Radio-
graphy, Hung Hom, Kowloon, Hong Kong
Michael Ljungberg, PhD

Lund University, Department of Radiation Physics, S-22185 Lund, Sweden
John W. Poston Sr, PhD, Professor

Department of Nuclear Engineering, Texas A&M University, 3133 TAMU,
College Station, TX 77843-3133, USA
John Roeske, PhD

Department of Radiation and Cellular Oncology, University of Chicago,
Chicago, IL 60637, USA
List of contributors xvii
Ruth Shefer, PhD

Newton Scientific, Inc., 245 Bent St, Winchester, MA 01890, USA
Michael Stabin, PhD

Department of Radiology and Radiological Sciences, Vanderbilt University,
1161 21st Avenue South, Nashville, TN 37232-2675, USA
Sven-Erik Strand, PhD, Professor

Lund University, Department of Radiation Physics, S-22185 Lund, Sweden
Jacquelyn Yanch, PhD

Department of Nuclear Engineering, Massachusetts Institute of Technology,
Room NW14-2207, 150 Albany Street, Cambridge, MA 02139-4307, USA
Pat Zanzonico, PhD

Memorial Sloan-Kettering Cancer Center, Nuclear Medicine Service, 1275
York Avenue, New York 10021, USA
Preface
This book provides a review of the Monte Carlo method as it is applied in the
field of therapeutic nuclear medicine. Driven in part by the remarkable
increase in computing power and its ready and inexpensive availability,
this is a relatively new yet rapidly expanding field. Likewise, although the
use of radionuclides for therapy has origins dating back almost to the
discovery of natural radioactivity itself, radionuclide therapy and, in
particular, targeted radionuclide therapy has only recently emerged as a
promising approach for therapy of cancer and, to a lesser extent, other
diseases. An effort has, therefore, been made to place the reviews provided
in this book in a broader context. The effort to do this is reflected by the
inclusion of chapters that do not directly address Monte Carlo techniques,
but rather provide an overview of issues that are closely related to therapeutic
nuclear medicine and to the potential role of Monte Carlo in this field. A
brief overview of each chapter is provided below.
Chapters 1, 2 and 6 review the fundamental theory of Monte Carlo,
techniques that have been used in nuclear medicine dosimetry and Monte
Carlo codes currently available for carrying out absorbed dose calculations.
Unlike external radiotherapy or brachytherapy, the kinetics and biodistribu-
tion of internally administered radionuclides must be measured. Imaging is,
therefore, an integral part of nuclear medicine dosimetry. As such, imaging
considerations are incorporated in almost every chapter of the book. The
contributions of Monte Carlo in nuclear medicine imaging are specifically
reviewed in chapter 3. The anatomical input required for Monte Carlo-
based dose calculations may be provided either as idealized geometric
representations of anatomy or as discrete, voxel-based representations that
reflect the actual shape and composition of a human. Again, these
considerations are present throughout the book, but chapters 5 and 13
focus on this specifically for the whole body and for the bone marrow,
respectively, the latter being a particularly important and challenging portion
of the anatomy. The role of Monte Carlo in providing a means for validating
the overall scheme of imaging-based, patient-specific absorbed dose
calculations is reviewed in chapter 12. Chapters 9, 14 and 15 examine specific
xviii
Preface xix
implementations of the Monte Carlo approach in the special cases of micro-

dosimetry, intravascular therapy and boron neutron capture synovectomy.
Microdosimetry is becoming increasingly important as high linear energy
transfer (LET) (see chapter 8) emissions are being considered for therapeutic
use. In alpha-emitter therapy, for example, a small number of decays can
substantially influence cell survival; the Monte Carlo method is therefore
indispensable in simulating such a stochastic process and thereby under-
standing experimental cell survival studies and potentially translating such
results into the clinic. Intravascular radiation therapy using radionuclides
is a relatively new area of radionuclide therapy. The potential efficacy of
this approach is heavily dependent upon the spatial distribution of absorbed
dose relative to the anatomical distribution of the cells involved in restenosis.
The absorbed dose distribution is, in turn, dependent upon the radionuclide
emissions, the configuration of emitters and the heterogeneous environment
that is traversed in going from the source region to the target regions. These
aspects are best addressed by Monte Carlo calculations. Boron neutron
capture therapy incorporates aspects of both external radiotherapy (the
neutron beam) and systemic radionuclide therapy (the biodistribution and
kinetics of boron-enriched compounds). In chapter 15, Monte Carlo is
used to investigate the design of a neutron beam configuration that is
optimized for synovectomy.
Conventional, non-Monte Carlo-based approaches to radionuclide
dosimetry are described in chapters 4 and 7. Chapter 4, in particular, provides
a review of the dose schema, developed by the Medical Internal Radionuclide
Dose (MIRD) Committee, that is most widely implemented for radionuclide
dosimetry. Chapter 7 describes the point-kernel methodology which has
been used extensively in patient-specific absorbed dose calculations and
which may be thought of as a precursor to a full Monte Carlo implementation
of imaging-based patient-specific dosimetry. Chapters 10 and 11 briefly review
two software packages that have been developed for carrying out radionuclide
dosimetry in therapeutic nuclear medicine. The first, MABDOSE, facilitates
implementation of the MIRD schema while also allowing for on-line Monte
Carlo calculations that make it possible to incorporate idealized tumour
geometries within the MIRD formalism. The second, 3D-ID, describes an
imaging-based approach to patient-specific dosimetry.
Chapter 8 provides a comprehensive review of the radiobiology relevant
to therapeutic nuclear medicine. This chapter is essential in understanding
the biological aspects critical to the design of successful radionuclide therapy
for cancer. It serves as a reminder that regardless of how precisely the
distribution of ionizing energy density (i.e., absorbed dose) is estimated,
this information will predict biologic response only if the pertinent biologic
considerations are also incorporated.
Finally, we would like to thank all of the contributors for their
invaluable hard work and for keeping to a very tight schedule. The topic
xx Preface
of this book is rapidly evolving and the editors felt it important to minimize
the time required to get this book into press. That this problem was so very
well overcome speaks of the commitment and dedication of the contributors.
We found compilation of this book to be a rewarding and educational
experience and hope that the reader is left with the same experience.
Chapter 1
The Monte Carlo method: theory and

computational issues
Habib Zaidi
1.1. INTRODUCTION
Usually, physicists and mathematicians like to give funny names to things

that are otherwise boring. There is a gaming aspect to Monte Carlo cal-
culations. Every simulation is based upon events that happen randomly,
and so the outcome of a calculation is not always absolutely predictable.
This element of chance reminds one of gambling and so the originators of
the Monte Carlo technique, Ulam and von Neumann, both respectable
scientists, called the technique Monte Carlo to emphasize its gaming aspect
[1, 2].
This introductory chapter briefly reviews the historical developments
and conceptual role of the Monte Carlo method, then summarizes its
derivation and methodological basis. Emphasis is given to applications
where photon and/or electron transport in matter are simulated. Some
computational aspects of the Monte Carlo method, mainly related to
random number generation, sampling and variance reduction, are discussed.
A general presentation of potential applications of Monte Carlo techniques
in different areas of nuclear medicine is given in chapter 2. Widely used
Monte Carlo codes in connection with computing facilities, vectorized and
parallel implementations are described in chapter 6 of this book.
1.1.1. Early approaches to Monte Carlo simulations

The Monte Carlo method describes a very broad area of science, in which
many processes, physical systems and phenomena are simulated by statistical
methods employing random numbers. The history of using random or
chance events to ascertain truths hidden from common perception may
date back to the documented practice of the ancient Israelites of a
1
2 The Monte Carlo method: theory and computational issues
dice-throwing method (the Urim and Thummim) to divine the will of God in
a given situation [3]. Dr de Buffon showed that by randomly tossing needles
on to a table on which parallel lines were drawn, one could determine experi-
mentally the value of pi [4] (it is said that this experiment started by de
Buffon throwing French stick loaves over his shoulder on to a tiled floor
and counting the number of times the loaves fell across the lines between
tiles on the floor). In 1899, Lord Rayleigh showed how the use of a one-
dimensional random walk could provide an approximate solution to a
parabolic differential equation [5]. In 1931 Kolmogorov showed the
relationship between Markov stochastic processes and certain integro-
differential equations [6].
Ulam in his autobiography [7] writes that shortly after his return to Los
Alamos, probably in 1946–1947, he gave two seminars that ‘had good or
lucky ideas and led to successful further development’. One was on what
became known as the Monte Carlo method. He describes an earlier discus-
sion of the basic idea with von Neumann as they drove from Los Alamos
to Lamy. This part of the book is vague on chronology—the seminar may
have been as late as 1948. The abbreviated bibliography lists one paper by
Ulam on Monte Carlo [8]. About 1948 Fermi, Metropolis and Ulam
obtained Monte Carlo estimates for the eigenvalues of the Schrödinger
equation [1] (the reference does not include Fermi, but he was involved in
the derivation). Ulam and von Neumann coined the name ‘Monte Carlo’
in about 1944 in the Manhattan Project during World War II, taken from
the name of the well-known city in Monaco famous for its games of
chance. Both of them led the modern Monte Carlo age and were pioneers
in the development of Monte Carlo techniques and their realizations on
digital computers.
The earliest application of the Monte Carlo method to radiation trans-
port problems appears to be a study by Spencer [9] of the effects of polariza-
tion on multiple (successive) Compton scatterings, a problem further
examined later on using the Boltzmann equation and diffusion theory [10].
This is further emphasized in the pioneering work of Kahn [11] for predicting
neutron fluxes associated with nuclear reactors designed and brought on line
in the 1940s. Mayer used one of the first automatic computers, the ENIAC,
as early as 1949 for neutron transport Monte Carlo calculations as reported
at a 1949 symposium on the Monte Carlo method [12].
1.1.2. Conceptual role of Monte Carlo simulations

The general idea of Monte Carlo analysis is to create a model which is as
similar as possible to the real system of interest, and to create interactions
within that system based on known probabilities of occurrence, with
random sampling of the probability density functions (pdfs). As the
number of individual events (called ‘histories’) is increased, the quality of
Random number generation 3
the reported average behaviour of the system improves, meaning that the
statistical uncertainty decreases. Almost any complex system in principle
can be modelled; perhaps we wish to model the number of cars passing a
particular intersection during certain times of the day, to optimize traffic
management, or the number of people that will make transactions in a
bank, to evaluate the advantages of different queuing systems. If we know
from experience the distribution of events that occur in our system, we can
generate a pdf and sample it randomly to simulate the real system. Detailed
descriptions of the general principles of the Monte Carlo method are given in
a number of publications [13, 14] and will not be repeated here. The major
components of a Monte Carlo method are briefly described below. These
components comprise the foundation of most Monte Carlo applications.
The following sections will explore them in more detail. An understanding
of these major components will provide a sound foundation for the developer
to construct his own Monte Carlo method. The primary components of a
Monte Carlo simulation method include the following:
(i) Probability density functions (pdfs): the physical system must be
described by a set of pdfs.
(ii) Random number generator: a source of random numbers uniformly
distributed on the unit interval must be available.
(iii) Sampling rule: a prescription for sampling from the specified pdfs.
(iv) Scoring: the outcomes must be accumulated into overall tallies or scores
for the quantities of interest.
(v) Error estimation: an estimate of the statistical error (variance) as a func-
tion of the number of trials and other quantities must be determined.
(vi) Variance reduction techniques: methods for reducing the variance in the
estimated solution to reduce the computational time for Monte Carlo
simulation.
(vii) Parallelization and vectorization algorithms to allow Monte
Carlo methods to be implemented efficiently on advanced computer
architectures.
1.2. RANDOM NUMBER GENERATION
Computational studies requiring the generation of random numbers are

becoming increasingly common. All random number generators (RNGs)
are based upon specific mathematical algorithms, which are repeatable. As
such, the numbers are just pseudo-random. Here, for simplicity, we shall
term them just ‘random’ numbers. Formally, random is defined as exhibiting
‘true’ randomness, such as the time between ‘tics’ from a Geiger counter
exposed to a radioactive element. Pseudo-random is defined as having the
appearance of randomness, but nevertheless exhibiting a specific, repeatable
pattern. Quasi-random is defined as filling the solution space sequentially (in

fact, these sequences are not at all random; they are just comprehensive at a
preset level of granularity). Monte Carlo methods make extensive use of
random numbers to control the decision-making process when a physical
event has a number of possible results. The RNG is always one of the
most crucial subroutines in any Monte Carlo-based simulation code. A
large number of generators are readily available [15], and many of these
are suitable for implementation on any computer system since today there
is no significant distinction in floating-point processing capabilities between
a modern desktop and a mainframe computer. A typical simulation uses
from 107 to 1012 random numbers, and subtle correlations between these
numbers could lead to significant errors [16]. The largest uncertainties are
typically due more to approximations arising in the formulation of the
model than those caused by lack of randomness in the RNG. Mathematically
speaking, the sequence of random numbers used to effect a Monte Carlo
model should possess the following properties [17]:
(i) Uncorrelated sequences: the sequences of random numbers should be
serially uncorrelated. Most especially, n-tuples of random numbers
should be independent of one another.
(ii) Long period: ideally, the generator should not repeat; practically, the
repetition should occur only after the generation of a very large set of
random numbers.
(iii) Uniformity: the sequence of random numbers should be uniform, and
unbiased. That is, suppose we define n-tuples uni ¼ ðui þ 1 ; . . . ; ui þ n Þ and
divide the n-dimensional unit hypercube into many equal sub-volumes.
A sequence is uniform if, in the limit of an infinite sequence, all the
sub-volumes have an equal number of occurrences of random n-tuples.
(iv) Reproducibility: when debugging programs, it is necessary to repeat the
calculations to find out how the errors occurred. The feature of reprodu-
cibility is also helpful while porting the program to a different machine.
(v) Speed: it is of course desirable to generate the random numbers quickly.
(vi) Parallelization: the generator used on vector machines should be vector-
izable, with low overhead. On massively parallel architectures, the
processors should not have to communicate among themselves,
except perhaps during initialization.
Although powerful RNGs have been suggested including shift register, inver-
sive congruentional, combinatorial and ‘intelligent’ methods such as those
implemented in the MCNP code [18], the most commonly used generator
is the linear congruential RNG (LCRNG) [19]. Recently, Monte Carlo
researchers have become aware of the advantages of lagged-Fibonacci
series (LFRNG). With extremely long periods, they are generally faster
than LCRNG and have excellent statistical properties [20]. Those generators
are briefly described below.
Random number generation 5
1.2.1. Linear congruential generators

The LCRNG has the form [19]
un þ 1 ¼ aðun þ cÞ modðmÞ; ð1:1Þ
where m is the modulus, a the multiplier, and c the additive constant or
addend. The size of the modulus constrains the period, and is usually
chosen to be either prime or a power of 2 [21]. An important subset of
LCRNG is obtained by setting c ¼ 0 in equation (1.1), which defines the
multiplicative linear congruential RNG (MLCRNG). This generator (with
m a power of 2 and c ¼ 0) is the de facto standard included with FORTRAN
and C compilers [22]. One of the biggest disadvantages to using a power of 2
modulus is that the least significant bits of the integers produced by these
LCRNGs have extremely short periods. For example, un modð2 j Þ will have
a period of 2 j [21]. In particular, this means the least-significant bit of the
LCRNG will alternate between 0 and 1. Some cautions to the programmer
are in order: (i) the bits of un should not be partitioned to make several
random numbers since the higher-order bits are much more random than
the lower-order bits; (ii) the power of 2 modulus in batches of powers of 2
should be avoided; (iii) RNGs with large modulus are preferable to ones
with small modulus. Not only is the period longer, but also the correlations
are lower. In particular, one should not use 32 bit modulus for applications
requiring a high resolution in the random numbers. In spite of this known
defect of power of 2 LCRNGs, 48 bit multipliers (and higher) have passed
many very stringent randomness tests.
The initial seed should be set to a constant initial value, such as a large
prime number (it should be odd, as this will satisfy period conditions for any
modulus). Otherwise, the initial seed should be set to a ‘random’ odd value.
Anderson [23] recommends setting the initial seed to the following integer:
u0 ¼ i yr þ 100 ði month 1 þ 12 ði day 1 þ 31
ði hour þ 24 ði min þ 60 i secÞÞÞÞ; ð1:2Þ
where the variables on the right-hand side are the integer values of the date
and time. Note that the year is 2 digits long, i.e., the domain of i yr is [0–99].
However, it may be preferable to introduce the maximum variation in the
seed into the least significant bits by using the second of this century,
rather than the most significant bits. The following equation is preferable:
u0 ¼ i sec þ 60 ði min þ 60 ði hour þ 24 ði day 1 þ 31
ði month 1 þ 12 i yrÞÞÞÞ: ð1:3Þ
Generally parametrizing the iteration process, either through the multiplier
or the additive constant, best parallelizes LCRNGs. Based on the modulus,
different parametrizations have been tried [23].
1.2.2. Lagged-Fibonacci generators

The lagged-Fibonacci series RNG (LFRNG) have the following general
form [15]
un ¼ un l un k ðmÞ; l > k; ð1:4Þ
where may be one of the binary arithmetic operators þ, or ; l and k are
the lags, and m is a power of 2 ðm ¼ 2p Þ. In recent years the additive lagged-
Fibonacci RNG (ALFRNG) has become a popular generator for serial as
well as scalable parallel machines [24] because it is easy to implement, it is
cheap to compute and it does well on standard statistical tests, especially
when the lag k is sufficiently high (such as k ¼ 1279). The maximal period
of the ALFRNG is ð2k 1Þ2p 1 and has 2ðk 1Þð p 1Þ different full-period
cycles [25]. Another advantage of the ALFRNG is that one can implement
these generators directly in floating-point to avoid the conversion from
integer to floating-point that accompanies the use of other generators.
However, some care should be taken in the implementation to avoid float-
ing-point round-off errors.
Instead the ALFRNG can be parametrized through its initial values
because of the tremendous number of different cycles. Assigning each
stream a different cycle produces different streams. An elegant seeding
algorithm that accomplishes this is described by Mascagni et al. [24]. An
interesting cousin of the ALFRNG is the multiplicative lagged-Fibonacci
RNG (MLFRNG). While this generator has a maximal period of
ð2k 1Þ2p 3 , which is a quarter the length of the corresponding ALFRNG,
it has empirical properties considered to be superior to ALFRNGs [15]. Of
interest for parallel computing is that a parametrization analogous to that
of the ALFRNG exists for the MLFRNG. This latter algorithm was used
for generating uniformly distributed random numbers on a parallel computer
based on the MIMD principle [26]. The sequence of 24 bit random numbers
has a period of about 2144 and has passed stringent statistical tests for random-
ness and independence [20].
1.3. PHOTON TRANSPORT
For radiation transport problems, the computational model includes geom-

etry and material specifications. Every computer code contains a database of
experimentally obtained quantities, known as cross sections, that determine
the probability of a particle interacting with the medium through which it is
transported. Every cross section is peculiar to the type and energy of the
incident particle and to the kind of interaction it undergoes. These partial
cross sections are summed to form the total cross section; the ratio of the
partial cross section to the total cross section gives the probability of this
particular interaction occurring. Cross section data for the interaction
Photon transport 7
types of interest must be supplied for each material present. The model also
consists of algorithms used to compute the result of interactions (changes in
particle energy, direction, etc.) based on the physical principles that describe
the interaction of radiation with matter and the cross section data provided.
Therefore, it is extremely important to use an accurate transport model, as
the Monte Carlo result is only as valid as the data supplied.
1.3.1. Photon cross section data

When a photon (having an energy below 1 MeV for isotopes of interest in
nuclear medicine) passes through matter, any of the three interaction
processes (photoelectric, incoherent scattering, coherent scattering) may
occur. The probability of a photon of a given energy E undergoing absorp-
tion or scattering when traversing a layer of material Z can be expressed
quantitatively in terms of a linear attenuation coefficient (cm1 ) which is
dependent on the material’s density, (g cm3 ):
¼ photo þ incoh þ coh : ð1:5Þ
Many approximations are made in Monte Carlo simulation packages to
either simplify the computational model or improve the speed of operation.
For example, when simulating positron emission tomography systems, it is
common to neglect coherent scattering in Monte Carlo simulation of
photon transport because of its low contribution to the total cross section at
511 keV. In the following examples, the relative importance of the various
processes involved in the energy range of interest (below 1 MeV) are
considered for some compounds and mixtures used in nuclear medicine to
justify some of the approximations made in Monte Carlo codes. Figure 1.1
illustrates the relative strengths of the photon interactions versus energy for
water (H2 O) and bismuth germanate (BGO), respectively. For water, a moder-
ately low-Z material, we note two distinct regions of single interaction domi-
nance: photoelectric below and incoherent above 20 keV. The almost order of
magnitude depression of the coherent contribution is some justification for the
approximations discussed. The coherent contribution to the total cross section
is less than 1% for energies above 250 keV. However, this contribution is in the
order of 7% for high-Z materials like BGO. Therefore, efforts should be made
to treat the coherent scattering process adequately for detector materials.
Accurate Monte Carlo simulations rely on detailed understanding and
modelling of radiation transport and on the availability of reliable, physically
consistent databases or cross section libraries used for photon/electron trans-
port calculations [27]. As discussed and historically reviewed in some detail
by Hubbell [28], there exist many compilations of photon cross section
data. The discrepancies and envelope of uncertainty of available interaction
data have been examined from time to time, including the effects of molecular
and ionic chemical binding, particularly in the vicinity of absorption edges.
Figure 1.1. Components of photon cross-sections for H2 O and BGO of interest

in PET imaging illustrating the relative contribution of each process.
Photon transport 9
Calculations of photon interaction data are generally in terms of atomic

cross sections, in units of cm2 /atom, customarily in units of barns/atom
(or b/atom) where 1 barn ¼ 1024 cm2 .
The Storm and Israel [29] photon cross section data have been used
extensively in medical physics. This is a 1970 compilation of data for
elements 1–100 and energies 1 keV to 100 MeV, and contains mass attenua-
tion coefficients, mass energy transfer coefficients, and mass energy absorp-
tion coefficients, presented in units of barns/atom. The medical physics
community makes extensive use of these coefficients in different applications
including Monte Carlo modelling. The various Monte Carlo codes developed
to simulate nuclear medicine imaging systems use different photon cross
section libraries. In the MCNP Monte Carlo code [30], the photon interac-
tion tables for Z ¼ 84, 85, 87, 88, 89, 91 and 93 are based on the compilation
of Storm and Israel from 1 keV to 15 MeV. For all other elements Z ¼ 1–94
the photon interaction tables are based on data from Evaluated Nuclear
Data Files (ENDF) [31] from 1 keV to 100 MeV. Data above 15 MeV for
the Storm and Israel data and above 100 MeV for the ENDF data come
from adaptation of the Livermore Evaluated Photon Data Library
(EPDL89) [32] and go up to 100 GeV. The original EGS4 system [33] also
uses compilation by Storm and Israel for the photoelectric and pair produc-
tion cross sections. Recently, cross section data for this code, based on the
PHOTX library [34], was created by Sakamoto [35]. ITS [36] includes cross
sections for bound electrons, the effect of which is ignored in the default
EGS4 package. For the photon energy range over 1 keV to 50 MeV, of
most interest to medical physicists, particular attention is called to a recently
assembled electronic database, including values of energy absorption co-
efficients, developed by Boone and Chavez [37]. In addition to attenuation
coefficients, other useful data such as the density, atomic weight, K, L1 , L2 ,
L3 , M and N edges, and numerous characteristic emission energies are
output from the program, depending on a single input variable.
In a recent investigation, photon cross section libraries (XCOM [38],
PHOTX [34]) and parametrizations implemented in simulation packages
(GEANT [39], PETSIM [40]) were compared with the recent library provided
by the Lawrence Livermore National Laboratory (EPDL97) [41] for energies
from 1 keV to 1 MeV for a few human tissues and detector materials of
interest in nuclear imaging [42]. The cross section data for mixtures and
compounds were obtained from the equation
X
¼ wi ð=Þi ; ð1:6Þ
i
where is the density of the material, wi the fraction by weight of the ith atomic
constituent as specified in ICRU report 44 [43], and ð=Þi the mass attenuation
coefficients. Different photon cross section libraries show quite large variations
as compared to the most recent EPDL97 data files (figure 1.2). The EPDL97
Figure 1.2. Comparisons (percentage differences) between the different

libraries and the EPDL97 database for water (H2 O). The coefficients shown
are: (a) total, (b) photoelectric, (c) incoherent and (d) coherent. The compar-
isons were calculated at energies given to the keV resolution between 1 and
1000 keV for both XCOM and PHOTX, between 10 and 1000 keV for
GEANT, and between 15 and 511 keV for PETSIM. Reprinted from [42]
# 2000 IEEE, with permission.
Photon transport 11
Figure 1.3. (a) Integral profile through a two-dimensional projection (single

angular view) of a simulated Hoffman three-dimensional brain phantom gener-
ated using EPDL97 and XCOM photon cross section libraries. (b) Same as (a)
for EPDL97 and PETSIM. Error bars are shown on the plots. Reprinted from
[42] # 2000 IEEE, with permission.
library is the most up-to-date, complete and consistent library available and is
already a standard in the nuclear reactor industry. It is recommended that
Monte Carlo developers only use the most recent version of this library. Its
use as a standard in the simulation of medical imaging systems will help to elim-
inate potential differences between the results obtained with different codes.
Further evaluation of the effect of the photon cross section library on
actual simulation of PET data was performed by generating data sets for the
three-dimensional Hoffman brain phantom contained in a cylindrical homo-
geneous water phantom (20 cm diameter, 17.5 cm height) with apparent relative
concentrations of 4, 1 and 0 for grey matter, white matter and the ventricles,
respectively, simulating the activity distributions found in normal human
brain for cerebral blood flow and metabolism studies currently employed in
PET [44]. Comparisons between the profiles through a simulated two-dimen-
sional projection of the three-dimensional brain phantom estimated using
different libraries are illustrated in figure 1.3. The statistical analysis of the
data sets revealed the existence of a statistically significant difference between
the resulting projections generated using different libraries ( p < 0:025) [42].
1.3.2. Simulating interaction of photons with matter

For radiation transport problems, the computational model includes geometry
and material specifications. Object modelling is fundamental to perform
photon transport efficiently using the Monte Carlo method. It consists of a
description of the geometry and material characteristics for the object. The
material characteristics of interest are density and energy-dependent cross
sections. Different steps are followed when tracing the photon in both the phan-
tom/patient and the detector volume and depending on the application sought
relevant parameters are stored. The relative ratios of the cross sections for
photoelectric effect, incoherent and coherent scattering to the total cross section
are used to choose randomly which process occurs at each interaction vertex.
In the case of photoelectric absorption, the total photon energy is trans-
ferred to an atomic electron and the random walk is terminated. In an
incoherent photon interaction, a fraction of the photon energy is transferred
to the atomic electron. The direction of the scattered photon is changed to
conserve the total momentum of the interaction. The Klein–Nishina expres-
sion for the differential cross section per electron for an incoherent interaction
is used to sample the energy and polar angle of the incoherently scattered
photon taking into account the incoherent scattering factor [45]. The coherent
scattering only results in a change in the direction of the scattered photon since
the momentum change is transferred to the whole atom. The random number
composition and rejection technique [46] is used to sample the momentum of
the scattered photon and the scattering angle according to the form-factor
distributions. Coherent scatter distributions are sharply forward peaked and
vary considerably with atomic number and energy [47]. The pathlength of
Electron transport 13
the interacting photon is randomly generated according to the exponential

attenuation based on the interaction length. The total cross section at the
energy of the interacting photon determines the interaction length of the
exponential distribution. Calculation of the distances between interactions in
a medium are performed by sampling from the exponential attenuation distri-
bution (see equation 1.12). Different techniques have been proposed to improve
the computation speed when sampling from the probability distributions. They
are described in more detail in sections 1.5 and 1.6.
For example when simulating data acquisition for a PET scanner oper-
ating in three-dimensional mode, the tomograph is simulated as a number of
detection rings, each ring consisting of a number of scintillator crystals [48].
Grouping crystals in matrices simulates the detection block, typical of
current generation PET scanners. Photon history is tracked within a crystal,
across crystals within a block and across blocks. Crystals are considered as
adjacent in the transaxial plane and in the axial direction. Two external
shields are simulated as tungsten rings located at the two axial edges of the
tomograph, partly shielding radiation coming from outside the scanner
field of view. In the detector blocks, at each interaction vertex, the local
energy deposition is recorded. Tracking is stopped either by photoelectric
absorption, escape of the photon from the block volume, or by Compton
scattering leaving less than 5 keV to the recoil photon. The energies of all
interaction vertices are summed to yield the total absorbed energy in the scin-
tillation detector. This total energy is assumed to be converted to scintillation
light using Gaussian random smearing to account for the combined energy
resolution, E=E, of the scintillator
pffiffiffiffi and its photomultipliers. E=E is
assumed to be proportional to 1= E . The position blurring step then calcu-
lates the mean detection coordinates (X; Y; Z) of each incident photon. This
is done by computing the centroid of all interaction vertices, each weighted
by the ratio of its individual energy to the total energy. The mean X and Y
coordinates of each photon are smeared to account for spatial resolution
degradation due to positron range effect, annihilation photon accolinearity
and position miscoding in the detector block.
1.4. ELECTRON TRANSPORT
The Monte Carlo simulation of electron transport, as opposed to the local

absorption of electrons in their site of production, should be included in
all calculations where the complete electromagnetic cascade is to be consid-
ered. The interest in medical radiation physics focuses on situations where
electron ranges are comparable with the dimensions or the spatial resolution
of the problem to be solved, or the contribution of bremsstrahlung plays a
significant role in the amount of energy locally deposited or transported
away. Because of the low energies involved in nuclear medicine, secondary
electrons are generally assumed to deposit all their energy at the point of
interaction and their bremsstrahlung production is negligible. Thus, electron
transport has not received particular attention in nuclear imaging appli-
cations of the Monte Carlo method. However, a number of investigators
in the field have considered the simulation of electron transport, mainly
for dosimetry calculations [49–53], and for this reason a brief description
of electron transport techniques is given in this section.
Strictly, all the different electron interaction types should be considered
in detail in a Monte Carlo simulation, much in the same way as it is done with
photon interactions. This is a technique called microscopic simulation [54],
which is used at low electron energies, for example in electron microscopy.
The complexity of the procedures used in microscopic techniques varies
considerably, although a common approach is to neglect bremsstrahlung
interactions due to the low energies involved. The most common simple
models used are based on the simulation of all the single-scattering events,
calculating the electron step-length between consecutive collisions with the
elastic mean-free-path. Energy losses are determined from the Bethe theory
of stopping power [55] and energy-loss straggling is accounted for approxi-
mately. These models have been improved by taking inelastic collisions
into account.
When the electron energy increases above a few tens of keV, the large
number of interactions that occur during electron slowing-down makes it
unrealistic to simulate all the single physical interactions. Instead, multiple
collisions are taken into account in what resembles a series of ‘snapshots’ of
an electron history taken at certain time or energy intervals; single inter-
actions of different kinds are considered between two consecutive
snapshots. This grouping scheme constitutes the macroscopic [54] or
condensed-history techniques [56]. In a seminal work, Berger [56] estab-
lished the basis of these techniques for charged-particle transport, where
algorithms are based on multiple scattering theory (Moliere, Goudsmit–
Saunderson) and stopping power calculations to describe the grouped
angular deflections and energy losses, respectively. He divided algorithms
into two broad classes, class I and class II, which are distinguished
mainly by the way in which energy losses are grouped and, more specifi-
cally, by how the individual interactions leading to a large energy transfer,
namely the production of energetic knock-on electrons and bremsstrahlung
photons, are described.
The Monte Carlo system ITS3 [57] with its physics based on the original
ETRAN code by Berger and Seltzer [58], and EGS4 [33] are, respectively, the
most widely used examples of class I and class II algorithms in medical
physics. Unbiased comparisons of the two systems for a variety of situations
and energies have been made by Andreo [59]. In recent years, the most
advanced electron transport algorithms, with physics that has overcome
that of older Monte Carlo systems, combine macroscopic and microscopic
Electron transport 15
techniques, where the latter are used in situations that require sophisticated
boundary-crossing algorithms [60, 61].
1.4.1. Stopping power data

As is well known, the stopping power of an electron in a material gives the
average rate at which the electron loses energy, be it transferred to the
medium, or to secondary electrons and/or bremsstrahlung photons that
transport the energy away from the interaction site; its integral provides
the average distance an electron can travel in the material, that is, the electron
range. Although a deterministic quantity, stopping powers are the founda-
tion stones of the Monte Carlo simulation of electron transport. They are
also key quantities in dosimetry, as they allow the determination of the
energy absorbed in a medium based on the knowledge of the energy absorbed
in a different medium, the detector (cavity theory). Stopping powers are
calculated according to Bethe theory [55], applicable to electrons and
heavy charged particles. ICRU Report 37 [62] provides details on how the
key quantity mean ionization potential, I, has been derived for elements
and compounds, and on the evaluation of the density-effect correction; it
also describes the limitations of Bethe theory at energies below 1–10 keV
approximately. A similar ICRU Report 49 has been published for protons
and alpha particles [63]. A set of computer programs, ESTAR, PSTAR
and ASTAR [64], calculate stopping-powers and ranges for electrons,
protons and alpha particles, respectively, in different elements and
compounds from a pre-computed set of data, which form the basis of the
two mentioned ICRU reports. These data are not only necessary for
transport calculations but have become standard reference data for the
radiological sciences.
In the case of dosimetry, the interest rests on the calculation of stopping-
power ratios, medium to detector, where the use of the Monte Carlo
method to derive electron spectra, combined with analytical cavity theories,
has enabled the determination of stopping-power ratios that include all
generations of electrons and photons produced during the simulation of
the transport of incident electrons or photons. There is extensive literature
on the use of the Monte Carlo method for producing such data, mainly for
radiotherapy dosimetry [65–70], and for the set of values in current use, the
agreement between the various calculations is at the impressive 0.1% level,
even when different Monte Carlo codes have been used. It is worth
emphasizing that, in the evolution of the use of the Monte Carlo method
in radiotherapy dosimetry, the most important changes (of the order of
several per cent), as well as a dramatic improvement in our knowledge
and interpretation of physical phenomena, occurred during the 1980s as a
result of the extended use of Monte Carlo calculations by Berger and
Seltzer. The 1990s have mainly been characterized by a refinement of the
data, at the level of 1% or so, associated with the computer revolution that
has improved so dramatically many other aspects of medical radiation
physics.
1.5. ANALOGUE SAMPLING
Analogue Monte Carlo attempts to simulate the full statistic development of

the electromagnetic cascade. If we assume that a large number of particle
histories, N, are included in a batch, the individual batch estimates can be
considered as drawn from a normal distribution. For a given calculation,
the estimated uncertainty is proportional to the inverse of the square root
of the number of histories simulated. The efficiency " of a Monte Carlo
calculation can therefore be defined as [71]
1
"¼ ; ð1:7Þ
2 T
where T is the calculation time required to obtain a variance estimate 2 . For
large N, " should be constant as long as the calculation technique remains the
same.
As discussed earlier, the imaging system and/or source distribution can
be described in terms of probability density functions (pdfs). These pdfs,
supplemented by additional computations, describe the evolution of the
overall system, whether in space, energy, time, or even some higher-
dimensional phase space. The goal of the Monte Carlo method is to simulate
the imaging system by random sampling from these pdfs and by performing
the necessary supplementary computations needed to describe the system
evolution. In essence, the physics and mathematics are replaced by random
sampling of possible states from pdfs that describe the system. Thus, it is
frequently necessary to sample some physical event, the probability of
which is described by a known pdf. Examples include the distance to the
next interaction and the energy of a scattered photon. Let x be the physical
quantity to be selected and f ðxÞ the pdf. Among the properties of the pdf is
that it is integrable and non-negative. Assume that the domain of f ðxÞ is the
interval [xmin ; xmax ] and that it is normalized to unit area. The cumulative
distribution function FðxÞ of the frequency function f ðxÞ gives the prob-
ability that the random variable is less than or equal to x. It is defined as
ðx
FðxÞ probabilityð xÞ ¼ f ðÞ d: ð1:8Þ
xmin
A stochastic variable can be sampled by the use of uniformly distributed

random numbers R in the range [0–1] using one of the techniques described
below.
Analogue sampling 17
1.5.1. Direct method

This method can be used if the inverse of the cumulative distribution function
F 1 ðxÞ is easily obtainable. Since FðxÞ is uniformly distributed in [0–1], the
sampled value of x could be obtained by substituting FðxÞ in equation
(1.8) by a uniform random number R, that is x ¼ F 1 ðRÞ. A practical exam-
ple of using this technique is the calculation of the distance to the next inter-
action vertex. The inversion is not always possible, but in many important
cases the inverse is readily obtained.
1.5.2. Rejection method

Another method of performing this when it is too complicated to obtain the
inverse of the distribution function is to use the rejection technique [46],
which follows the following steps:
(i) Define a normalized function f 0 ðxÞ ¼ f ðxÞ=fmax ðxÞ, where fmax ðxÞ is the
maximum value of f ðxÞ.
(ii) Sample two uniformly distributed random numbers R1 and R2 .
(iii) Calculate x using the equation x ¼ xmin þ R1 ðxmax xmin Þ.
(iv) If R2 is less than or equal to f 0 ðxÞ then x is accepted as a sampled value,
otherwise a new value of x is sampled.
Over a large number of samples, this technique will yield a set of values of x
within the required distribution. It does, however, require two random
numbers per trial and many trials may be required depending on the area
under of the curve of f ðxÞ. A typical example of using this technique is the
sampling of the photon energy and scattering angle resulting from incoherent
scattering.
1.5.3. Mixed methods

When the previous two methods are impractical, the mixed method, which
combines the two, may be used [71]. Assume that the pdf can be factored
as follows:
f ðxÞ ¼ hðxÞ gðxÞ ð1:9Þ
where hðxÞ is an invertible function and gðxÞ is relatively flat but contains
most of the mathematical complexity. The method consists in the following
steps:
Ð max 0
(i) Normalize hðxÞ producing h0 ðxÞ such that xxmin h ðxÞ dx ¼ 1.
0
(ii) Normalize gðxÞ producing g ðxÞ such that g0 ðxÞ 1 for x in [xmin ; xmax ].
(iii) Use the direct method to select an x using h0 ðxÞ as the pdf.
(iv) Use x and apply the rejection method using g0 ðxÞ, i.e., choose a random
number R; if g0 ðxÞ R, accept x, otherwise go back to step (iii).
1.6. NON-ANALOGUE SAMPLING ‘VARIANCE REDUCTION’
A direct Monte Carlo simulation using true probability functions may require
an unacceptably long time to produce statistically relevant results. Emission of
photons is isotropic, so directional parameters may be sampled uniformly
within their individual ranges. Nuclear imaging systems have a low geometri-
cal efficiency because of the small solid angle defined by the collimator and/or
the small axial aperture. Therefore, the calculation would be very ineffective
in terms of required computing time [72]. It is thus desirable to bias the
sampling (non-analogue sampling) by introducing different types of impor-
tance sampling and other variance reduction techniques to improve the
computational efficiency of the Monte Carlo method [73]. The results obtained
by non-analogue simulation are, however, biased by the variance reduction
technique and a correction for this is required. A particle history weight, W,
is introduced, which describes the probability of the particle following the
current path. This weight is calculated for each particle history, and used
in the calculation of the results. If an event occurs, the weight W is added
to the counter rather than incrementing the counter by one unit. Bielajew
and Rogers [71] divided variance reduction techniques into three categories:
those that concern photon transport only, those that concern electron trans-
port only, and other more general methods. The most useful techniques are
described below. It is worth noting that concerning the Poisson nature of the
activity distributions in nuclear imaging, these variance reduction approxi-
mations may result in statistically relevant deviations from an otherwise
unbiased simulation result since Monte Carlo simulated projection data
are (as a result of variance reduction) not count but probability data.
1.6.1. Photon-specific methods

Interaction forcing. In an analogue Monte Carlo simulation, photons are
tracked through the object until they either escape the object, are absorbed,
or their energy drops below a selected threshold. The probability function for
a photon interaction is given by
pðxÞ ¼ e x : ð1:10Þ
The probability that a photon will travel a distance d or less is given by
ðd
pðdÞ ¼ e x dx ¼ 1 e d : ð1:11Þ
0
To sample the pathlength, a uniform random number R is substituted for

pðdÞ and the problem is solved for d:
logð1 RÞ
d¼ : ð1:12Þ

Non-analogue sampling ‘variance reduction’ 19
Since the maximum distance, dmax , the photon travels before interaction is
infinite and the number of mean free paths across the geometry in any
practical situation is finite, there is a large probability that photons leave
the geometry of interest without interacting. To increase the statistical
accuracy in the imparted energy calculation, we force the photons to interact
by assigning dmax a finite distance, e.g., the thickness of the detector being
simulated. A true distributed photon pathlength d within dmax can be
sampled from the equation
1
d ¼ lnð1 R½1 e dmax Þ: ð1:13Þ

The photon’s weight must be multiplied by the interaction probability
Wn þ 1 ¼ Wn ½1 e dmax : ð1:14Þ
In emission computed tomography, the photon is allowed to interact through
coherent or incoherent interactions only within the phantom since photo-
absorption does not contribute to energy imparted in the crystal. The
weight is then multiplied by the probability of the photon being scattered:

þ coh
Wn þ 1 ¼ Wn incoh ; ð1:15Þ

where incoh and coh are the cross section data for incoherent and coherent
scattering, respectively, and is the total linear attenuation coefficient.
Stratification. Stratification refers to the process of determining the

frequencies with which the various regions of state space are used to start
a particle [74]. The solid angle of acceptance of the detector array, max , is
small due to collimation and to the size of the detector array itself. This
results in significant computational inefficiencies with analogue Monte
Carlo simulation, because only a few per cent of the photons generated
and tracked will actually be detected. The goal of stratification is to simulate
only photons that are emitted in directions within the solid angle which can
be calculated from the maximum acceptance angle, max , which in turn can be
estimated from the dimensions of the phantom and the detection system. The
solid angle does not change in magnitude when simulating source locations
off-centre. The photon escaping from the phantom is either primary or scat-
tered. If the photon happens to be a primary photon, its direction within the
solid angle could be sampled from
cosðÞ ¼ 1 R½1 cos max : ð1:16Þ
In this case, the weight is multiplied by the probability of escape without
interaction in the solid angle max
½1 cos max
Wn þ 1 ¼ W n : ð1:17Þ
2
Exponential transform, Russian roulette and particle splitting. The exponen-

tial transform is a variance reduction technique used to bias the sampling
procedure to give more interactions in the regions of interest and thus
improve the efficiency of the calculation for those regions. To implement
this method, the distance to the next interaction in number of mean free
paths, d , should be sampled from [71]
lnðRÞ
d ¼ ; ð1:18Þ
ð1 C cos Þ
where C is a parameter that adjusts the magnitude of the scaling and the
angle of the photon with respect to the direction of interest. The new weight-
ing factor is given by
expðd C cos Þ
Wn þ 1 ¼ W n : ð1:19Þ
ð1 C cos Þ
Note that the new weighting factor is dependent on d . If 0 < C < 1, the
particle pathlength is stretched in the forward direction, which is used for
shielding problems. For 1 < C < 0, the average distance to the next
interaction is shortened in the forward direction, which is used for surface
problems. For C ¼ 0, we recover the unbiased sampling. The optimal
choice of this parameter is dependent on the problem to be solved. The
general guideline is to avoid to use large weighting factors because they
may increase the variance.
Russian roulette and splitting are often used together with the
exponential transform, although they are still effective when used indepen-
dently. In Russian roulette, a random number is selected and compared
with a threshold, . If the random number turns out to be smaller than ,
the particle is allowed to survive but the weight should be updated accord-
ingly: Wn þ 1 ¼ Wn =. In particle splitting, a particle coming from a region
of interest can be divided into N particles, each having a new weighting:
Wn þ 1 ¼ Wn =N.
1.6.2. Electron-specific methods

Electron range rejection. A fundamental difference between the transport of
photons and electrons in a condensed-history simulation code is that photons
travel relatively long distances before interacting while electron tracks
are interrupted not only by geometrical boundaries but also by multiple
scattering ‘steps’. A large amount of simulation time is spent on checking
boundaries and selecting deflection angles and so on. Electron range rejec-
tion means that electrons with their residual range smaller than the distance
to the nearest boundary or to the region of interest in the simulation will be
terminated to save computing time. Different methods have been suggested
Non-analogue sampling ‘variance reduction’ 21
for electron range rejection. The reduced interrogation of geometry (RIG)

method calculates the distance to the nearest boundary and compares it
with the maximum multiple-scattering step length. If the electron cannot
reach any of the boundaries during this step, the boundary checking routine
will not be called and this will save computing time. Another method called
‘disregard within a zone’ is usually used with RIG to further speed up the
simulation. It consists of disregarding electrons whose energies are so low
that they cannot reach the nearest boundary. Those methods are, however,
inefficient for simulations involving curved surfaces where the time required
to calculate the distance to the closest boundary may be considerable [71]. An
alternative way is to use a range-related ‘region rejection’ technique. In this
method, different energy cut-offs are chosen for the regions surrounding the
region where energy deposition is to be scored, each energy cut-off being
chosen according to the distance to the nearest boundary of the region of
interest.
Parameter-reduced electron step. This algorithm allows using small elec-

tron steps in the vicinity of interfaces and boundaries and large steps
elsewhere [75]. Its components are: a path-length correction algorithm
which is based on the multiple scattering theory of Moliere and which
takes into account the differences between the straight path length and
the total curved path length for each electron step; a lateral correlation
algorithm which takes into account lateral transport; and a boundary
crossing algorithm which ensures that electrons are transported accurately
in the vicinity of interfaces. The algorithm has been implemented in the
EGS4 Monte Carlo system and proved that substantial savings in comput-
ing time may be realized when using this method. The most recent version
of the code system called EGSnrc [61] incorporates more accurate electron
physics and transport algorithms which include an improved electron-
step algorithm, a correct implementation of the fictitious cross section
method for sampling distances between discrete interactions, a more
accurate evaluation of energy loss, as well as an exact boundary crossing
algorithm.
1.6.3. General methods

Correlated sampling. The correlated sampling technique can be used in
transport of both photons and electrons. It is especially effective for calculat-
ing ratios or differences of two quantities which are nearly equal. The basic
idea is that the simulations of the geometries of interest are kept as closely
correlated as possible so that most of the statistical fluctuations will cancel
in the ratios and differences. The real difference between the two geometries
will be better reflected in the ratios and the differences obtained. The calcula-
tional uncertainties in the ratios and the differences obtained with correlated
sampling are in general smaller than those obtained from uncorrelated

simulations.
There are several ways of doing correlated sampling in radiation trans-
port. In coupled photon–electron transport, a simple method has been used
in which random number seeds of the particle histories, for which a primary
particle or any of the secondaries has deposited energy in the region of
interest for one geometry, is stored and used for the simulations of the
alternative geometry [71]. A correlated sampling method for the transport
of electrons and photons has been developed in which a main particle
history is split up whenever a particle meets the boundary of the region
where the medium differs between two or more cases [76]. This particle is
then followed separately for each case until it and all its descendants termi-
nate. Holmes et al. [77] described a correlated sampling technique which
forces histories to have the same energy, position, direction and random
number seed as incident on both a heterogeneous and homogeneous water
phantom. This ensures that a history that has, by chance, travelled through
only water in the heterogeneous phantom will have the same path as it
would have through the homogeneous phantom, resulting in a reduced
variance when a ratio of heterogeneous dose to homogeneous dose is
formed.
Use of geometry symmetry. The use of some of the inherent symmetry of the
geometry may realize considerable increase in efficiency. If both the source
and target configurations contain cylindrical planar or spherical–conical
simulation geometries, the use of symmetries is more obvious. Other uses
of symmetry are less obvious but the saving in computing time is worth
the extra care and coding.
1.7. SUMMARY
Today’s applications of Monte Carlo techniques include diagnostic imaging

and radiation therapy, traffic flow, Dow-Jones forecasting, and oil well
exploration, as well as more traditional physics applications like stellar
evolution, reactor design, and quantum chromo-dynamics. Monte Carlo
methods are widely used in modelling of materials and chemicals, from
grain growth modelling in metallic alloys, to behaviour of nanostructures
and polymers, and protein structure predictions.
The application of Monte Carlo modelling in medical physics is an
everlasting enthusiastic topic and an area of considerable research interest.
Monte Carlo modelling has contributed to a better understanding of
the physics of photon and electron transport by clinical physicists. The
large number of applications of the Monte Carlo method attests to its
usefulness as a research tool in different areas of nuclear medicine.
References 23
Recent developments in diagnostic and therapeutic nuclear medicine

combined with advanced computing facilities and physics support have
created new opportunities for Monte Carlo simulation in this exciting
field. Important developments and applications are summarized in the
following chapters of the book.
ACKNOWLEDGMENTS
The author would like to thank Professor Pedro Andreo and Dr Mike Stabin
for fruitful discussions. Their comments and suggestions on this chapter are
gratefully acknowledged.
REFERENCES
[1] Metropolis N and Ulam S 1949 The Monte Carlo method J. Am. Stat. Assoc. 44 335–
41
[2] McCracken D D 1955 The Monte Carlo method Sci. Am. 192 90–6
[3] Van Dam C 2000 The Urim and Thummim: A Means of Revelation in Ancient Israel ed
I Eisenbrauns (Winona Lake, IN)
[4] de Buffon G 1777 Essai d’arithmétique morale Supple´ment à l’Histoire Naturelle
vol 4
[5] Rayleigh J W S 1899 On the theory of resonance Collected Scientific Papers 1
33–75
[6] Kolmogorov A N 1933, 1956 Foundations of the Theory of Probability (2nd English
edition) transl ed N Morrison (New York: Chelsea)
[7] Ulam S M 1976 Adventures of a Mathematician (New York: Scribners)
[8] Ulam S M 1951, 1949 On the Monte Carlo Method Symposium on Large-Scale Digital
Calculating Machines (Cambridge, MA: Harvard University Press)
[9] L V Spencer 1948 ‘Polarization of multiply scattered gamma rays’ (unpublished)
(National Bureau of Standards Interim Report on Contract with US Office of
Naval Research)
[10] Spencer L V and Wolff C 1953 Penetration and diffusion of hard X-rays: Polarization
effects Phys. Rev. 90 510–14
[11] Kahn H 1949 ‘Stochastic (Monte Carlo) attenuation analysis’ Report p-88
(RAND)
[12] Mayer M 1951 Report on a Monte Carlo calculation performed with the ENIAC
Monte Carlo Method eds A S Householder, G E Forsythe and H H Germond US
National Bureau of Standards Applied Math Series No 12, pp 19–20
[13] Raeside D E 1976 Monte Carlo principles and applications Phys. Med. Biol. 21
18197
[14] Turner J E, Wright H A and Hamm R N 1985 A Monte Carlo primer for health
physicists Health Phys. 48 717–33
[15] Marsaglia G and Zaman A 1994 Some portable very-long-period random number
generators Comput. Phys. 8 117–21
[16] Ferrenberg A M, Landau D P and Wong Y J 1992 Monte Carlo simulations: hidden
errors from ‘good’ random number generators Phys. Rev. Lett. 69 3382–4
[17] Vattulainen I, Kankaala K, Saarinen J and Ala-Nissila T 1995 A comparative
study of some pseudorandom number generators Comp. Phys. Comm. 86 209–
26
[18] Booth T E 1998 The intelligent random number technique in MCNP Nucl. Sci. Eng.
100 248–54
[19] Lehmer D H 1949 Mathematical methods in large-scale computing units 2nd
Symposium on Large-Scale Digital Calculating Machinery, Cambridge (Cambridge,
MA: Harvard University Press) pp 141–6
[20] Marsaglia G and Zaman A 1993 Monkey tests for random number generators Comp.
Math. Appl. 23 1–10
[21] Knuth D E 1981 The Art of Computer Programming, Seminumerical Algorithms 2nd
edition (Reading, MA: Addison-Wesley) vol 2
[22] Press W H 1992 Numerical Recipes in C, The Art of Scientific Computing (Cambridge:
Cambridge University Press)
[23] Anderson S L 1990 Random number generators on vector supercomputers and other
advanced architectures SIAM Rev. 32 221–51
[24] Mascagni M, Cuccaro S A, Pryor D V and M L Robinson 1995 A fast, high-quality,
and reproducible lagged-Fibonacci pseudorandom number generator J. Comput.
Phys. 15 211–19
[25] Brent R P 1994 On the periods of generalized Fibonacci recurrences Math. Comput.
63 389-401
[26] Zaidi H, Labbé C and Morel C 1998 Implementation of an environment for Monte
Carlo simulation of fully 3D positron tomography on a high-performance parallel
platform Parallel Comput. 24 1523–36
[27] Zaidi H 1999 Relevance of accurate Monte Carlo modeling in nuclear medical
imaging Med. Phys. 26 574–608
[28] Hubbell J H 1999 Review of photon interaction cross section data in the medical and
biological context Phys. Med. Biol. 44 R1–22
[29] Storm E and Israel H I 1970 Photon cross sections from 1 keV to 100 MeV for
elements Z ¼ 1 to Z ¼ 100 Nucl. Data Tables A7 565–681
[30] Briesmeister J F 1997 (Los Alamos, NM: Los Alamos National Laboratory)
[31] Hubbell J H et al. 1975 Atomic form factors, incoherent scattering functions, and
photon scattering cross sections J. Phys. Chem. Ref. Data 4 471–538
[32] Cullen D E et al. 1989 ‘Tables and graphs of photon interaction cross sections from
10 eV to 100 GeV derived from the LLNL Evaluated Photon Data Library (EPDL)’
UCRL-50400 (Lawrence Livermore National Laboratory, CA)
[33] Nelson W R, Hirayama H and Rogers D W O 1985 (Stanford, CA: Stanford Linear
Accelerator Center)
[34] Trubey D K, Berger M J and Hubbell J H 1989 Photon cross sections for ENDF/B-
IV. American Nuclear Society Topical Meeting, Advances in Nuclear Engineering
Computation and Radiation Shielding, Santa Fe, NM
[35] Sakamoto Y 1993 Photon cross section data PHOTX for PEGS4 code. EGS4 users’
meeting, Japan (KEK Proceedings) pp 77–82
[36] Halbleib J A, Kensek R P, Valdez G D, Seltzer S M and Berger M J 1992 ITS: The
Integrated TIGER Series of electron/photon transport codes—version 30 IEEE
Trans. Nucl. Sci. 39 1025–30
References 25
[37] Boone J M and Chavez A E 1996 Comparison of x-ray cross sections for diagnostic
and therapeutic medical physics Med. Phys. 23 1997–2005
[38] Berger M J and Hubbell J H 1987 ‘XCOM: Photon cross sections on a personal
computer’ NBSIR 87-3597 (US National Bureau of Standards)
[39] Brun R, Bruyant F, Maire M, McPherson A C and Zanarini P 1994 ‘GEANT detec-
tor description and simulation tool’ DD/EE/84-1 (CERN)
[40] Picard Y, Thompson C J and Marrett S 1993 Improving the precision and accuracy of
Monte Carlo simulation in positron emission tomography IEEE Trans. Nucl. Sci. 39
1111–16
[41] Cullen D E, Hubbell J H and Kissel L 1997 ‘EPDL97: the Evaluated Photon Data
Library’ UCRL-50400 (Lawrence Livermore National Laboratory, CA)
[42] Zaidi H 2000 Comparative evaluation of photon cross section libraries for materials
of interest in PET Monte Carlo simulations IEEE Trans. Nucl. Sci. 47 2722–35
[43] ICRU 1989 Tissue Substitutes in Radiation Dosimetry and Measurement vol 44
(International Commission on Radiological Units and Measurements)
[44] Hoffman E J, Cutler P D, Digby W M and Mazziotta J C 1990 3-D phantom to
simulate cerebral blood flow and metabolic images for PET IEEE Trans. Nucl. Sci.
37 616–20
[45] Klein O and Nishina Y 1929 Über die Streuung von Strahlung durch freie Elektronen
nach der neuen relativistischen Quantendynamik von Dirac Z. Phys. 52 853–68
[46] Kahn H 1956 Use of Different Monte Carlo Sampling Techniques in Monte Carlo
Methods ed H A Meyer (New York: Wiley)
[47] Kaplan M S, Harrison R L and Vannoy S D 1998 Coherent scatter implementation
for SimSET IEEE Trans. Nucl. Sci. 45 3064–8
[48] Zaidi H, Scheurer A H and Morel C 1999 An object-oriented Monte Carlo simulator
for 3D cylindrical positron tomographs Comput. Methods Programs Biomed. 58
13345
[49] Stabin M and Konijnenberg M W 2000 Re-evaluation of absorbed fractions for
photons and electrons in spheres of various sizes J. Nucl. Med. 41 149–60
[50] Bouchet L G and Bolch W E 1999 A three-dimensional transport model for deter-
mining absorbed fractions of energy for electrons within cortical bone J. Nucl.
Med. 40 2115–24
[51] Bouchet L G, Jokisch D W and Bolch W E 1999 A three-dimensional transport model
for determining absorbed fractions of energy for electrons within trabecular bone
J. Nucl. Med. 40 1947–66
[52] Furhang E E, Chui C S, Kolbert K S, Larson S M and Sgouros G 1997 Implementa-
tion of a Monte Carlo dosimetry method for patient-specific internal emitter therapy
Med. Phys. 24 1163–72
[53] Yoriyaz H, Stabin M and dos Santos A 2001 Monte Carlo MCNP-4B-based
absorbed dose distribution estimates for patient-specific dosimetry J. Nucl. Med. 42
662–9
[54] Andreo A 1991 Monte Carlo techniques in medical radiation physics Phys. Med. Biol.
36 861–920
[55] Bethe H A 1930 Zur Theorie des Durchgangs schneller Korpuskularstrahlen durch
Materie Ann. Phys., Lpz. 5 325
[56] Berger M J 1963 Monte Carlo calculation of the penetration of diffusion of fast
charged particles Methods in Computational Physics ed B Alder, S Feinbach and
M Rotenberg (New York: Academic) pp 135–215
[57] J A Halbleib et al. 1992 ‘ITS version 3.0: The integrated TIGER Series of coupled
electron/photon Monte Carlo transport codes’ Report SAND91-1634 (Sandia
National Laboratories)
[58] Berger M J and Seltzer S M 1968 ‘Electron and photon transport programs: I. Intro-
duction and notes on program DATAPAC 4; II. Notes on program ETRAN-15’
Reports NBS 9836 and 9837. Report ORNL-RSIC CCC-107 (Oak Ridge: Oak
Ridge National Laboratory; National Bureau of Standards)
[59] Andreo P 1995 A Comparison Between the ITS3 and EGS4 Monte Carlo Codes in
Radiation Physics, Advanced Monte Carlo Computer Programs for Radiation Trans-
port (Saclay: Publications Service OECD)
[60] Salvat F, Fernandez-Varea J M, Sempau J and Mazurier J 1999 Practical aspects of
Monte Carlo simulation of charged particle transport: mixed algorithms and variance
reduction techniques Radiat. Environ. Biophys. 38 15–22
[61] Kawrakow I 2000 Accurate condensed history Monte Carlo simulation of electron
transport. I. EGSnrc, the new EGS4 version Med. Phys. 27 485–98
[62] ICRU 1984 ‘Stopping powers for electrons and positrons’ Report 37 (International
Commission on Radiological Units and Measurements)
[63] ICRU 1993 ‘Stopping powers and ranges for protons and alpha particles’ Report 49
(International Commission on Radiological Units and Measurements)
[64] Berger M J 1993 ‘ESTAR, PSTAR, and ASTAR: computer programs for calculating
stopping-power and range tables for electrons, protons, and helium ions’ NIST
Report NISTIR 4999 (National Institute of Standards and Technology)
[65] Berger M J, Seltzer S M, Domen S R and Lamperti P J 1975 Stopping-power ratios
for electron dosimetry with ionization chambers. Biomedical Dosimetry (Vienna:
IAEA) pp 589–609
[66] Nahum A E 1978 Water/air mass stopping power ratios for megavoltage photon and
electron beams Phys. Med. Biol. 23 24–38
[67] Andreo P and Brahme A 1986 Stopping power data for high-energy photon beams
Phys. Med. Biol. 31 839–58
[68] Ding G X, Rogers D W and Mackie T R 1995 Calculation of stopping-power ratios
using realistic clinical electron beams Med. Phys. 22 489–501
[69] Malamut C, Rogers D W and Bielajew A F 1991 Calculation of water/air stopping-
power ratios using EGS4 with explicit treatment of electron–positron differences
Med. Phys. 18 1222–8
[70] ICRU 1984 ‘Radiation dosimetry: electron beams with energies between 1 and 50 MeV’
Report 35 (International Commission on Radiological Units and Measurements)
[71] Bielajew A F and Rogers D W O 1989 Variance-reduction techniques Monte Carlo
Transport of Electrons and Photons ed T Jenkins, A Rindi, A Nahum and D W O
Rogers (New York: Plenum) pp 407–19
[72] Haynor D R, Harrison R L and Lewellen T K 1990 Improving the efficiency of
emission tomography using variance reduction techniques IEEE Trans. Nucl. Sci.
37 749–53
[73] Zubal I G and Harrell C R 1992 Voxel-based Monte Carlo calculations of nuclear
medicine images and applied variance reduction techniques Image Vision Computing
10 342–8
[74] Haynor D R, Harrison R L and Lewellen T K 1991 The use of importance sampling
techniques to improve the efficiency of photon tracking in emission tomography
simulations Med. Phys. 18 990–1001
References 27
[75] Bielajew A F and Rogers D W O 1987 PRESTA: The parameter reduced electron-step
transport algorithm for electron Monte Carlo transport Nucl. Instrum. Methods Phys.
Res. B18 165–81
[76] Ma C M and Nahum A E 1993 Calculation of absorbed dose ratios using correlated
Monte Carlo sampling Med. Phys. 20 1189–99
[77] Holmes M A et al. 1993 The application of correlated sampling to the computation of
electron beam dose distributions in heterogeneous phantoms using the Monte Carlo
method Phys. Med. Biol. 38 675–88
Chapter 2
Monte Carlo techniques in nuclear

medicine dosimetry
Habib Zaidi and Pedro Andreo
2.1. INTRODUCTION
The use of the Monte Carlo method to simulate radiation transport has
become the most accurate means of predicting absorbed dose distributions
and other quantities of interest in the radiation treatment of cancer patients
using either external or radionuclide radiotherapy. The same trend has
occurred for the estimation of the absorbed dose in diagnostic procedures
using radionuclides. There is broad consensus in accepting that the earliest
Monte Carlo calculations in medical radiation physics have been made in
the area of nuclear medicine, where the technique has been used for
dosimetry modelling and computations. Formalism and data based on
Monte Carlo calculations, developed by the Medical Internal Radiation
Dose (MIRD) Committee of the Society of Nuclear Medicine, have been
published in a series of supplements to the Journal of Nuclear Medicine,
the first one being released in 1968 [1]. Some of these pamphlets made
extensive use of Monte Carlo calculations to derive specific absorbed frac-
tions for electron and photon sources uniformly distributed in organs of
mathematical phantoms. Interest in Monte Carlo-based dose calculations
with and emitters has been revived with the application of radiolabelled
monoclonal antibodies to radioimmunotherapy (RIT). As a consequence of
this generalized use, many questions are being raised, primarily about the
need and potential of Monte Carlo techniques, but also about how accurate
it really is, and what it would take to apply it clinically and make it available
widely to the nuclear medicine community at large.
In this chapter, we try to answer many of these questions, sometimes
without explicitly stating them. Many other questions will be answered
when Monte Carlo techniques are implemented and used for more routine
calculations and for in-depth investigations. A survey is presented of the
28
Monte Carlo techniques in medical radiation physics 29
different applications of the Monte Carlo method in medical physics in

general, and diagnostic and therapeutic nuclear medicine in particular,
with special emphasis on internal dose calculations in radionuclide therapy.
Interested readers are encouraged to consult the detailed and comprehensive
reviews [2, 3] and books [4, 5] published on the subject. The main purpose of
this chapter is to present a framework for applying Monte Carlo simulations
for a wide range of problems in nuclear medicine dosimetry. Emphasis is
given to applications where photon and/or electron transport in matter are
simulated. Some computational aspects of the Monte Carlo method,
mainly related to the generation of random numbers, sampling and variance
reduction techniques are presented in chapter 1.
2.2. MONTE CARLO TECHNIQUES IN MEDICAL RADIATION

PHYSICS
Monte Carlo techniques have become one of the most popular tools in
different areas of medical physics following the development and subsequent
implementation of powerful computing systems for clinical use. In particu-
lar, they have been extensively applied to simulate processes involving
random behaviour and to quantify physical parameters that are difficult or
even impossible to calculate analytically or to determine by experimental
measurements.
Following the review by Andreo [2], the applications of the Monte Carlo
method in medical radiation physics cover almost all topics including radia-
tion protection, nuclear medicine, diagnostic radiology and radiotherapy,
with an increasing enthusiastic interest in exotic and exciting new applica-
tions such as intravascular radiation therapy and boron neutron capture
therapy and synovectomy. The use of Monte Carlo techniques in the last
two areas falls outside the scope of this survey whereas applications in
radiation protection and nuclear medicine imaging and dosimetry are
covered in the following sections. The present authors are not aware of a
comprehensive review on applications of the Monte Carlo method in diag-
nostic radiology and would recommend a recent paper by Boone et al. [6]
for an up-to-date validation study providing a general overview of the
field. Interested readers in radiotherapy physics are encouraged to check
the reviews by Ma and Jiang [7] for electron beams, and by Ahnesjo and
Aspradakis [8] for photon beams. With the rapid development of computer
technology, Monte Carlo-based treatment planning for radiation therapy is
becoming practical.
An area which has received considerable attention during the past
few years is intravascular brachytherapy (see chapter 14). The irradiation
of the coronary arteries is being studied by a number of researchers world-
wide, with the hope of reducing the rate of restenosis. Currently, there is
30 Monte Carlo techniques in nuclear medicine dosimetry
considerable discussion about the best radionuclides and radiation delivery

systems that will be effective. Stabin et al. [9] used both the MCNP-4B and
EGS4 Monte Carlo systems to simulate radiation sources, and results have
been crosschecked between the two systems using 186 Re and 188 Re balloon
sources as well as an 192 Ir wire source. Three-dimensional dose distributions
have also been calculated for a 32 P impregnated stent and a 198 Au stent using
EGS4 [10].
The promise of a new experimental cancer therapy with some indication
of its potential efficacy has led many scientists from around the world to
work on an approach called neutron capture therapy (NCT), where the
use of boron (BNCT) has dominated this type of application. The idea
behind BNCT is straightforward. A tumour-seeking compound containing
boron-10 (10 B), a non-radioactive isotope, is introduced into the brain and
given time to accumulate in the tumour. The patient is then exposed to a
beam of neutrons, usually in a nuclear reactor facility, which are ‘captured’
or absorbed by the 10 B. Capturing neutrons causes the boron nuclei to break
apart, resulting in the emission of radiation and lithium nuclei. Both
particles and lithium are high in energy but short in range, which means
they potentially can destroy the malignant cells in which 10 B is imbedded
without hurting the adjacent healthy cells. Advanced methods of BNCT
use an epithermal neutron beam in conjunction with tumour-targeting
boron compounds for irradiation of glioblastomas and metastatic melano-
mas. Recently, a systematic Monte Carlo investigation of the dosimetric
properties of near-threshold neutron beams has been performed indicating
that accelerator proton energies between 1.93 and 1.99 MeV, using 5 cm of
H2 O moderator followed by thin 6 Li and Pb shields, can provide therapeu-
tically useful beams with treatment times less than one hour and accelerator
currents less than 5 mA [11]. A comprehensive review of the state of the art in
NCT has been published recently by the IAEA [12].
On the other hand, boron neutron capture synovectomy (BNCS) is a
potential application of the 10 B(n,)7 Li nuclear reaction for the treatment
of rheumatoid arthritis [13]. In BNCS, a stable 10 B labelled compound is
injected directly into the joint, and the joint is irradiated with an epithermal
neutron beam. BNCS is currently being investigated at the laboratory for
accelerator beam applications at the MIT in collaboration with Newton
Scientific, Inc. and Brigham and Women’s Hospital (see chapter 15). A dedi-
cated BNCS beamline, consisting of an accelerator target and cooling system
mounted in a neutron moderator and reflector assembly, has been
constructed, installed, and is now operational on a high-current tandem
electrostatic accelerator. Monte Carlo simulations have indicated that treat-
ment times range from 4 to 25 min-mA (e.g. 4 minutes for a 1 mA beam
current) depending on the size of the joint and the neutron source reaction
used in the irradiation. The feasibility of using 157 Gd as an alternative to
boron as a neutron capture agent for NCS, termed gadolinium neutron
Applications of Monte Carlo techniques 31
capture synovectomy (GNCS), has also been conducted using Monte Carlo
simulations where 10 B and 157 Gd have been compared as isotopes for
accelerator-based NCS [14].
2.3. APPLICATIONS OF MONTE CARLO TECHNIQUES IN

NUCLEAR MEDICINE IMAGING
Recent nuclear medical imaging innovations such as single-photon emission

computed tomography (SPECT), positron emission tomography (PET), and
multiple emission tomography (MET) as well as combined functional and
structural imaging (see chapter 3) are ideal for Monte Carlo modelling tech-
niques because of the stochastic nature of radiation emission, transport and
detection processes. Factors which have contributed to the wider use of
Monte Carlo techniques include improved models of radiation transport
processes, the practicality of its application following the development of
acceleration schemes and the improved speed of computers. This section
summarizes potential areas of application of Monte Carlo techniques in
nuclear imaging. An overview of existing simulation programs is presented
in [3] together with examples of useful features of such sophisticated tools
in connection with common computing facilities and more powerful
multiple-processor parallel processing systems.
Recent developments in nuclear medicine instrumentation and multiple-
processor parallel processing systems have created new opportunities for
Monte Carlo simulation in nuclear medicine imaging. One of the aims of
the medical physicist involved in nuclear medical imaging research is to
optimize the design of imaging systems and to improve the quality and quan-
titative accuracy of reconstructed images. Several factors affect the image
quality and the accuracy of the data obtained from a nuclear medicine
scan. These include the physical properties of the detectors, collimator and
gantry design, attenuation and scatter compensation, and reconstruction
algorithms [15]. Integrating improvements in these topics with current
tracers, and sensitive and specific tracers under development, will provide
major advantages to the general nuclear medicine clinician and research
investigator (figure 2.1). Mathematical modelling is necessary for the
assessment of various parameters in nuclear medical imaging systems since
no analytical solution is possible when the radiation transport equation,
describing the interaction of photons with non-uniformly attenuating body
structures and complex detector geometries, is solved.
2.3.1. Diagnostic nuclear medicine imaging

There has been an enormous increase and interest in the use of Monte Carlo
techniques in all aspects of nuclear imaging, including planar imaging [16],
Figure 2.1. Scientific and technical strategy for recording accurate functional
images. In bold, the parts where Monte Carlo simulation plays an important
role. Reprinted with permission from AAPM [3].
SPECT [17, 18], PET [19, 20] and MET [21]. However, due to computer
limitations, the method has not yet fully lived up to its potential. With the
advent of high-speed supercomputers, the field has received increased attention,
particularly with parallel algorithms, which have much higher execution rates.
Figure 2.2. Principles of Monte Carlo simulation of a nuclear medical imaging

system. Adapted and reproduced with permission from AAPM [3].
Figure 2.2 illustrates the idea of Monte Carlo or statistical simulation as

applied to an imaging system. Assuming that the behaviour of the imaging
system can be described by probability density functions (pdfs), then the
Monte Carlo simulation can proceed by sampling from these pdfs, which
necessitates a fast and effective way to generate random numbers uniformly
distributed. Photon emissions are generated within the phantom and are
transported by sampling from pdfs through the scattering medium and detec-
tion system until they are absorbed or escape the volume of interest without
hitting the crystal. The outcomes of these random samplings, or trials, must
be accumulated or tallied in an appropriate manner to produce the desired
result, but the essential characteristic of Monte Carlo is the use of random
sampling techniques to arrive at a solution of the physical problem.
The Monte Carlo method is a widely used research tool in different areas
of nuclear imaging such as detector modelling and systems design, image
correction and reconstruction techniques, internal dosimetry and pharma-
cokinetic modelling [3]. The method has proven to be very useful for solving
complex problems that cannot be modelled by computer codes using deter-
ministic methods or when experimental measurements may be impractical.
Monte Carlo simulation of detector responses and determination of their
efficiencies is one of the areas which has received considerable attention.
Many detector modelling applications have been developed in the PET
field including the pioneering work of Derenzo [22] who simulated arrays
of detectors of different materials and sizes to study the effect of the inter-
crystal septa and later on to optimize the optical coupling between BGO
crystals and PMTs [23] by taking into account the reflection and scattering
along the detection system. The search for an appropriate detector for this
imaging modality has been conducted in a comparative study of several

crystals including BGO, CsF and NaI(Tl) [24], BaF2 used in time-of-flight
PET [25], and liquid xenon [26].
The design of SPECT and PET systems using the Monte Carlo method
has received considerable attention and a large number of applications were
the result of such investigations. During the past two decades, simulation of
scintillation camera imaging using both deterministic and Monte Carlo
methods has been developed to assess qualitatively and quantitatively
the image formation process and interpretation [27] and to assist in the
development of collimators [28]. Several researchers have used Monte
Carlo simulation methods to study potential designs of dedicated small
animal positron tomographs [29, 30]. An important conclusion drawn
from these studies is that, unlike human imaging where both sensitivity
and spatial resolution limitations significantly affect the quantitative imaging
performance of a tomograph, the imaging performance of dedicated animal
tomographs is almost solely based upon its spatial resolution limitations.
Simulation of transmission scanning allowed study of the effect of down-
scatter from the emission (99m Tc) into the transmission (153 Gd) energy window
in SPECT [32] and investigation of detected scattered photons in single-
photon transmission measurements using 137 Cs single-photon sources for
PET [33]. A comparative evaluation of pre- and post-processing attenuation
correction methods in lung SPECT has been also conducted using Monte
Carlo simulations of a digital thorax phantom containing a homogeneous
activity distribution in the lung [34].
Another promising application of Monte Carlo calculations is quantifi-
cation and correction for photon attenuation and scattering in nuclear
medicine imaging, since the user has the ability to separate the detected
photons into their components, e.g., primary events, scatter events, contribu-
tion of down-scatter events. Monte Carlo modelling thus allows a detailed
investigation of the spatial and energy distribution of Compton scatter,
which would be difficult to perform using present experimental techniques,
even with very good energy resolution detectors [35]. A Monte Carlo study
of the acceptance of scattered events in a depth-encoding large-aperture
camera made of position-encoding blocks modified to resolve the depth of
interaction through a variation in the photopeak pulse-height has been
performed by Moison et al. [31]. Figure 2.3 shows a comparison between
measured and simulated single-energy spectra of the ECAT-953B PET
scanner as well as measured and simulated scatter fractions as a function
of the lower energy threshold. An energy resolution of 23% FWHM has
been assumed since this is a typical value for BGO block detectors. Monte
Carlo simulations have also been extensively used to evaluate and compare
scatter correction schemes in both SPECT [36] and PET [37].
Monte Carlo simulations have been shown to be very useful for
validation and comparative evaluation of image reconstruction techniques.
Figure 2.3. (a) Comparison between measured and simulated single-energy

spectra and (b) measured and simulated scatter fractions for the ECAT
953B PET scanner. Reprinted from [31] # 1996 IEEE, with permission.
Smith et al. [38] used Monte Carlo modelling to study photon detection
kernels, which characterize the probabilities that photons emitted by radio-
isotopes in different parts of the source region will be detected at particular
projection pixels of the projection images for the case of parallel-hole
collimators. The authors also proposed a reconstruction method using the
three-dimensional kernels in which projection measurements in three
adjacent planes are used simultaneously to estimate the source activity of
the centre plane. The search for unified reconstruction algorithms led to
the development of inverse Monte Carlo (IMC) reconstruction techniques.
The principal merits of IMC are that, like direct Monte Carlo, the method
can be applied to complex and multivariable problems, and variance reduc-
tion procedures can be applied [39]. Floyd et al. [40] used IMC to perform
tomographic reconstruction for SPECT with simultaneous compensation
for attenuation, scatter, and distance-dependent collimator resolution.
The Monte Carlo code Eidolon [20], for example, has been developed
using modern software development tools. This simulator is an efficient
tool that can be used to generate data sets in a controllable manner in
order to assess different reconstruction algorithms and scatter correction
methods [37]. As the ‘best’ algorithm can only be selected with respect to a
certain task, different ‘basic’ performance measures can be used. Image
degrading effects are illustrated using simulated projections of the digitized
three-dimensional Hoffman brain phantom [41]. A slice of this phantom is
shown in figure 2.4(A). The ratio between the activity in white, grey matter
and ventricles has been chosen as 1 :4 :0, respectively. The projections of
this phantom at different levels of fidelity are generated. The strengths of
the image degrading factors are characteristic of an [18 F]-FDG-brain
study. Figure 2.4(F–I) shows the effects of different aspects of image degra-
dation on filtered backprojection reconstructions. The loss of resolution
caused by detector blurring (FWHM ¼ 4 mm) on projection data and FBP
reconstructions is shown in figures 2.4(C) and 2.4(G), respectively, while in
D and H, effects of detector blurring, attenuation and scatter are included
in the simulation and no corrections performed on the simulated data sets.
Finally, in E and I, effects of detector blurring, attenuation and scatter are
included and appropriate corrections for attenuation and scatter applied.
2.3.2. Therapeutic nuclear medicine imaging

For internal radiation dose estimates, the biodistribution of a trace 131 I-
labelled monoclonal antibody is generally used to predict the biodistribution
of a high-dose administration for therapy. Imaging therapeutic doses would
further confirm the hypothesis that the biodistribution is similar; however,
current generation scintillation cameras are unable to handle accurately
the corresponding high counting rate. Monte Carlo calculations have been
used in the development of a method for imaging therapeutic doses of 131 I
Figure 2.4. A. Transaxial slice of the digital three-dimensional Hoffman brain

phantom. Monte Carlo simulation of the three-dimensional Hoffman brain
phantom in different imaging situations showing two-dimensional projections
(middle row) and their filtered backprojection reconstructions (bottom row).
In case B and F, a two-dimensional projection and reconstructed image neglect-
ing attenuation, scatter and detector blurring. In C and G, effects of detector
blurring are included, while in D and H, effects of detector blurring, attenuation
and scatter are included in the simulation and no corrections performed on the
simulated data sets. Finally, in E and I, effects of detector blurring, attenuation
and scatter are included and appropriate corrections for attenuation and scatter
are applied.
by using thick Pb sheets placed on the front surface of a high-energy parallel-

hole collimator [42]. The accuracy of 131 I tumour quantification after RIT
has been further investigated with an ultra-high-energy (UHE) collimator
designed for imaging 511 keV photons [43]. It has been shown that the
difference in tumour size, relative to the size of a calibration sphere, has
the biggest effect on accuracy, and recovery coefficients are needed to
improve quantification of small tumours.
There has been renewed interest in pinhole collimation for high-
resolution imaging of small organs and structures (e.g., the thyroid) and
for regionally administered monoclonal antibody imaging since it provides
an improved spatial resolution and an increase in sensitivity as the distance
between the source and the pinhole aperture decreases [44]. Huili et al. [45]
simulated point response functions for pinhole apertures with various
aperture span angle, hole size and materials. The point responses have been
parametrized using radially circular symmetric two-dimensional exponential

functions, which can be incorporated into image reconstruction algorithms
that compensate for the penetration effect. The effect of pinhole aperture
design parameters on angle-dependent sensitivity for high-resolution pinhole
imaging has been also investigated using Monte Carlo modelling [46].
Simulated 131 I SPECT studies for uniform cylinders showed that activity
concentrations have been underestimated towards the outside of the
cylinders when a sin3 rather than the correct sinx sensitivity correction
has been applied in image reconstruction, where x is a parameter describing
the sensitivity and is the angle of the incident ray with the surface of the
detector crystal.
Different strategies are being developed to improve image quality and
quantitative accuracy in tumour SPECT imaging including collimator–
detector response compensation and high-energy scatter correction tech-
niques. In an elegant study by Dewaraja et al. [47], Monte Carlo simulations
have been used to evaluate how object shape influences ‘spill-out’ and
‘spill-in’, which are major sources of quantification errors associated with
the poor spatial resolution of 131 I SPECT and to characterize energy and
spatial distributions of scatter and penetration [48].
2.4. MONTE CARLO TECHNIQUES IN NUCLEAR MEDICINE

DOSIMETRY
2.4.1. Calculation of absorbed fractions

The origins of the absorbed fraction concept adopted by the MIRD commit-
tee for internal dose calculations trace back to the earlier work by Ellet et al.
[49] in the 1960s. Many of the pamphlets published by the MIRD committee
made extensive use of Monte Carlo calculations to derive specific absorbed
fractions for photon sources uniformly distributed in organs of mathematical
phantoms [50]. This has been extended later on to electrons and particles
[51, 52].
Cristy [53] demonstrated that the reciprocity theorem which states that,
for any pair of regions in a uniform isotropic or uniform scatterless model,
the specific absorbed fraction is independent of which region is designated
source and which is designated target may also be valid for heterogeneous
phantoms for certain conditions. Other approaches using the MIRD formal-
ism have also been proposed [54]. Poston et al. [55] calculated photon specific
absorbed fractions for both the Cristy and Eckerman gastrointestinal tract
and their revised model and reported differences between electron absorbed
fraction values with and without electron tracking. Previously calculated
absorbed fractions for unit density spheres in an infinite unit density
medium for photon and electron emitters have been recently re-evaluated
Monte Carlo techniques in nuclear medicine dosimetry 39
using both the EGS4 and MCNP4B Monte Carlo codes [56]. Calculation of
absorbed fractions for positron emitters relevant to neurologic studies has
also been reported [57].
The application of the Monte Carlo method to internal radiation dosi-
metry is further emphasized in two recent MIRD Pamphlets. In MIRD
Pamphlet No. 15 [58], the EGS4 Monte Carlo radiation transport code has
been used to revise substantially the dosimetric model of the adult head
and brain originally published in MIRD Pamphlet No. 5. Pamphlet No.
17 [59] crystallizes the utility of the MIRD formalism for the calculation of
the non-uniform distribution of radiation-absorbed dose in different
organs through the use of radionuclide-specific S values defined at the
voxel level. Skeletal S values and absorbed fractions estimates to both
marrow and endosteum in trabecular and cortical bone have been further
improved through the use of electron transport models [60].
Mathematical anthropomorphic phantoms are continuously being
improved. Current developments are aimed at computer phantoms that
are flexible while providing accurate modelling of patient populations. An
important contribution came from Bouchet and Bolch [61], making available
a series of five dosimetric head and brain models developed to allow more
precise dosimetry in paediatric neuroimaging procedures. A new rectal
model [62] and dynamic urinary bladder model [63] have also been proposed
more recently. To develop more patient-specific dosimetry, new mathemati-
cal models for adults of different height have been developed using anthropo-
metric data [64]. The use of dynamic anthropomorphic phantoms in Monte
Carlo simulations is becoming possible due to the increasing availability of
computing power. This includes the development of appropriate primitives
that allow accurate modelling of anatomical variations and patient
motion, like superquadrics [65] and non-uniform rational B-spline surfaces
[66]. More recently, an image-based whole-body model, called VIP-Man,
has been developed using high-resolution transversal colour photographic
images obtained from the National Library of Medicine’s Visible Human
Project [67]. The phantom is shown in figures 2.5(b) and 2.5(c) in comparison
with the conventional anthropomorphic mathematical models (figure 2.5(a)).
The EGS4 code has been also used to estimate specific absorbed fractions
from internal electron emitters with energies from 100 keV to 4 MeV [68].
2.4.2. Derivation of dose-point kernels

In most cases, Monte Carlo calculations are used to simulate the random
distribution of sources or targets whereas the actual dosimetric calculation
is performed using the so-called dose-point kernels. Such kernels, usually
spherical and calculated for mono-energetic photon or electron sources,
describe the pattern of energy deposited at various radial distances from
photon [1, 69, 70] and electron or [71, 72] point sources. Dose-point kernels
Figure 2.5. Anterior view of (a) the mathematical model and the VIP-Man
model in (b) two-dimensions and (c) three-dimensions. Reprinted with
permission from the Institute of Physics [68]. (See plate 1 for colour version.)
can be calculated using analytical or Monte Carlo methods. Hybrid

approaches (analytical calculations using Monte Carlo data) have also
been considered to decrease the computation time [73]. Three Monte Carlo
systems have mainly been used for this purpose, namely, ETRAN [71], the
ACCEPT code of the ITS system [74] used by Sgouros et al. [75], and
EGS4 [72, 76]. Limitations and constraints of some of these codes have
been reported in the literature whose impact on the calculated kernels is
difficult to evaluate. ETRAN, for instance, had an incorrect sampling of
the electron energy-loss straggling that has been corrected for in the ITS3
system (based on ETRAN); EGS4 did not include the accurate electron
physics and transport algorithms which have been incorporated in the
recent EGSnrc system [77]. The EGS4 system has also been used to charac-
terize the spatial and energy distribution of bremsstrahlung radiation from
Monte Carlo techniques in nuclear medicine dosimetry 41
point sources in water of relevance to RIT [78]. This study provided the
initial data required for modelling and analysing the scatter, attenuation,
and image formation processes in quantitative imaging of bremsstrahlung
for RIT dosimetry. Furhang et al. [70] generated photon point dose kernels
and absorbed fractions in water for the full photon emission spectrum of
radionuclides of interest in nuclear medicine, by simulating the transport
of particles using Monte Carlo techniques. The kernels have been then
fitted to a mathematical expression.
A unified approach for photon and particle dosimetry has been
proposed by Leichner [79] by fitting Berger’s tables for photons and electrons
to generate an empirical function that is valid for both photons and
particles. Therefore both point-kernel and Monte Carlo techniques can be
effectively employed to calculate absorbed dose to tissue from radionuclides
that emit photons or electrons. The latter are computationally much more
intensive; however, point-kernel methods are restricted to homogeneous
tissue regions that can be mathematically described by analytical geometries
whereas Monte Carlo methods have the advantage of being able to accom-
modate heterogeneous tissue regions with complex geometric shapes.
2.4.3. Pharmacokinetic modelling

Pharmacokinetic modelling is a useful component for estimating the
cumulated activity in various source organs in the body. A few applications
of Monte Carlo techniques have been reported in the field. In particular,
Casciari et al. [80] developed a compartmental model of [F-18] fluoromisoni-
dazole transport and metabolism to compute the volume average kappa in
tissue regions from [F-18] fluoromisonidazole PET time–activity data and
characterized it using Monte Carlo simulations. This model has been able
to accurately determine kappa for a variety of computer generated time–
activity curves including those for hypothetical heterogeneous tissue regions
and poorly perfused tissue regions. Compartmental models allow also the
in vivo analysis of radioligand binding to receptor sites in the human
brain. Benzodiazepine receptor binding has been studied using a three-
compartmental model [81]. The validity of the results of the coefficient of
variation of each parameter has been verified with statistical results provided
by Monte Carlo simulation. Burger [82] examined the possibility of using a
mathematical metabolite correction, which might obviate the need for
actual metabolite measurements. Mathematical metabolite correction has
been implemented by estimating the input curve together with kinetic
tissue parameters. The general feasibility of the approach has been evaluated
in a Monte Carlo simulation using a two-tissue compartment model. A
simplified approach involving linear-regression straight-line parameter
fitting of dynamic scan data has been developed for both specific and non-
specific models [83]. Monte Carlo simulations have been used to evaluate
parameter standard deviations, due to data noise, and much smaller noise-
induced biases. The authors reported good agreement between regression
and traditional methods.
Welch [84] investigated and quantified the effect of typical SPECT
system resolution and photon counting statistics on the bias and precision
of dynamic cardiac SPECT parameters. Simulation of dynamic SPECT
projection data has been performed using a realistic human torso phantom
assuming both perfect system resolution and a system resolution typical of
a clinical SPECT system. The results showed that the rate constant charac-
terizing the washing of activity into the myocardium is more sensitive to
the position of the region of interest than the washout rate constant and
that the main effect of increased photon noise in the projection data is to
decrease the precision of the estimated parameters.
Computer simulations demonstrate that estimation of the kinetic
parameters directly from the projections is more accurate than the estimation
from the reconstructed images [85]. A strategy for joint estimation of
physiological parameters and myocardial boundaries has been proposed
and evaluated by simulated myocardial perfusion studies based on a simpli-
fied heart model [86]. A method allowing the estimation of kinetic parameters
directly from SPECT cone-beam projections has also been proposed and
validated with a simulated chest phantom [87]. The results showed that
myocardial uptake and washout parameters estimated by conventional
analysis of noiseless simulated cone-beam data had biases ranging between
3 and 26% and between 0 and 28%, respectively, while uncertainties of
parameter estimates with this method ranged between 0.2 and 9% for the
uptake parameters and between 0.3 and 6% for the washout parameters.
2.5. MONTE CARLO TECHNIQUES IN RADIATION PROTECTION
There is a considerable debate between scientists working in medical

radiation physics and in the field of radiation safety on the term ‘radiation
protection’ as applied to radiation medicine, and in particular to patients
undergoing diagnostic medical procedures or radiation treatment. The
determination of the dose delivered to different organs, for example, pertains
to the standard clinical procedure or quality assurance process performed in
each medical specialty, be it a diagnostic nuclear medicine or radiology
technique, or a radiotherapy treatment. For a hospital physicist, dosimetry
is not radiation protection but clinical practice. The radiation safety officer
calls the task radiation protection. As the authors of this survey are members
of the first group, it has been a natural decision not to consider dosimetry
applications within this section, but as a clinical speciality in nuclear medi-
cine (see section 2.4). Other applications of the Monte Carlo method are
described below which can be considered more closely related to radiation
Monte Carlo techniques in radiation protection 43
protection aspects. It should be emphasized, however, that considering the

large uncertainties allowed in radiation protection, and the considerable
man-power and time required to perform Monte Carlo simulations, many
of the problems analysed by the existing applications could have been
solved using rather simple analytical procedures. The use of Monte Carlo
techniques in radiation protection should only be justified in cases of
complex situations, either due to the presence of complicated mixed radiation
sources or geometries.
2.5.1. Shielding calculations

Monte Carlo techniques have been used to determine detector calibration
factors and to predict -ray spectra for well defined measurements. The
techniques have sometimes been used to predict shielding requirements
and critical structural specifications that must conform to radiation pro-
tection rules. Different Monte Carlo systems have been used to evaluate
shielding including EGS4 [88] and MCNP for diagnostic X-ray rooms [89],
and ITS to calculate door thicknesses for megavoltage radiotherapy [90].
More sophisticated applications include quantifying the effect of lead
shielding on dose distributions during radiotherapy treatment planning
with electron beams [91] and designing the moderator–collimator–shielding
system of a neutron beam optimized for radiography that utilizes a 252 Cf
source [92]. Monte Carlo methods have also been used to estimate the
neutron and photon dose rates in a variety of locations in the vicinity of
an accelerator-based neutron capture therapy facility, as well as the shielding
configuration required when the device is run at maximum current [93].
2.5.2. Characterization of detectors and radiation monitoring instruments

Monte Carlo simulations have been shown to be useful in the evaluation of
the design of collimated detectors used to measure 125 I or 131 I uptake in the
thyroid gland [94]. Two detector sizes have been simulated for each radio-
isotope, and activity has been placed both in the gland and in the remainder
of the body in varying amounts to assess the efficacy of the collimation. This
study has been followed by a detailed report where the authors assessed the
uncertainty introduced into an activity estimate of radioiodine in the thyroid
when the size and shape of the gland differs from that of the calibration
phantom [95]. The depth-dose distribution in TLD dosimeters irradiated
with rays from a 90 Sr–90 Y source has also been investigated using the
Monte Carlo method [96]. A pilot study has been conducted at the Lawrence
Livermore National Laboratory to demonstrate the applicability of Monte
Carlo techniques to calibrate in vivo measurement systems and investigate
the effects of the source geometry and the detector size on the measured
efficiency [97]. The development of a calibration phantom for in vivo
measurements of actinides using the MCNP Monte Carlo system to deter-

mine the point source distribution that closely approximates a homogeneous
bone contamination has also been reported [98].
2.5.3. Radiation dose calculations to staff

A few studies have reported the use of the Monte Carlo method to calculate
the dose received by medical and technical staff in a clinical environment. A
recent publication estimated the dose due to radiation scattered from the roof
of an adjacent radiotherapy facility where only the primary X-ray radiation
had been accounted for in the shielding calculations [99]. Another group has
calculated coefficients for converting dosimeter readings to equivalent dose
to the foetus using the equivalent dose to the uterus and Monte Carlo model-
ling to simulate the equivalent dose to the foetus during the first two months
of pregnancy [100]. Pattison [101] used -ray and particle dose rates
modelled using distributed point sources and empirically based Monte
Carlo approaches, respectively, to determine the dose received by the skin
of the fingers of clinical and laboratory staff during injections of 153 Sm.
Conversion coefficients per air-kerma free-in-air for the personal dose
equivalent, Hp (10) have also been calculated using Monte Carlo methods
for various dosimeter positions in the trunk of a voxel model of an adult
male versus the direction of incidence of broad parallel photon beams with
energies between 10 keV and 10 MeV [102]. Fluence-to-effective dose
equivalent conversion factors as a function of radiation angles and sex for
monoenergetic photon beams of 0.08, 0.3 and 1.0 MeV have also been
reported elsewhere [103].
2.5.4. Revisiting release criteria for patients administered therapeutic doses

Application of the Monte Carlo method in this area has been the subject of
an interesting debate between supporters [104] and detractors [105] of the
utility of the technique to solve a problem that can be easily resolved other-
wise. Johnson and Barnhart [105] suggested that accurate results could have
been derived by using ‘pen-and-paper’ instead of performing an expensive
Monte Carlo simulation. Siegel et al. [106], on the other hand, argued that
it would be difficult to show that the approximations involved in the calcula-
tion by hand are accurate enough without having the Monte Carlo or other
‘gold standard’ results (such as a phantom study) for comparison. It should
be recognized that holding a Geiger–Muller counter at 1 m from the
umbilicus of a patient who has been administered a known amount of 131 I
allows checking if the model predicts the readings or not. The results
obtained by Johnson and Barnhart [105] indicated that their analytical
model accurately predicted the reading without having to resort to a
Monte Carlo simulation. Moreover, the limits of integration of their most
Future applications of Monte Carlo 45
accurate model (model #3) could be changed to accommodate an individual

with different torso dimensions, and obtain an answer in a fraction of the
time that it would take to adapt the target geometry and run a Monte
Carlo simulation.
2.6. FUTURE APPLICATIONS OF MONTE CARLO
2.6.1. Patient-specific dosimetry and treatment planning

Currently, the preferred strategy with radiolabelled antibodies is to use
personalized patient dosimetry and this approach may become routinely
employed clinically. The dose distribution pattern is often calculated by
generalizing a point source dose distribution [75, 107], but direct calculation
by Monte Carlo techniques is also frequently reported because it allows
media of inhomogeneous density to be considered [108–110].
The development of a three-dimensional treatment planner based on
SPECT/PET imaging is an area of considerable research interest and several
dose calculation algorithms have been developed. Figure 2.6 lists the essential
Figure 2.6. Diagram showing the essential steps required in developing a three-
dimensional internal dosimetry program for treatment planning with RIT based
on quantitative emission computed tomography where Monte Carlo simulations
play a crucial role. Adapted and reproduced with permission from AAPM [3].
steps required in developing a three-dimensional treatment planning

program for RIT. Projection data acquired from an emission tomographic
imaging system are processed to reconstruct transverse section images
which yields a count density map of source regions in the body. This count
density is converted to an activity map using the sensitivity derived from a
calibration phantom. In the final step, this activity distribution is converted
to a dose rate or dose map either by convolving the activity distribution with
dose point kernels or by direct Monte Carlo calculations. To elaborate a
treatment plan for an individual patient, prospective dose estimates can be
made by using a tracer activity of radiolabelled antibody to obtain biodistri-
bution information prior to administration of a larger therapeutic activity.
The clinical implementability of treatment planning algorithms will depend
to a significant extent on the time required to generate absorbed dose
estimates for a particular patient. For more details, the reader is referred
to chapters 6 and 10–12, which fully cover the development steps and clinical
experience gained using these tools.
2.6.2. On-line PET monitoring of radiotherapy beams

On-line monitoring of the positron-emitting activity created in patient tissues
undergoing radiotherapy treatment has been a goal pursued by several
investigators since the 1990s. Researchers at the Research Centre of Rossen-
dorf in Dresden [111–113] and at the University of Michigan [114] have made
important contributions to this technique in radiotherapy treatments with
photons, protons and light and heavy ions. Whereas the clinical application
of on-line PET monitoring in photon radiotherapy has so far been limited by
the reduced activity produced in a patient using today’s clinical accelerators,
its use in heavy-particle radiotherapy has become a useful technique to
visualize þ activity distributions that help to determine the effective range
of heavy particles in the patient, as well as to evaluate blood flow in some
organs [115].
The rationale for the preferred use of light (Z 10) and heavy ions over
conventional radiation beams is twofold. First, the dose distribution
increases along the penetration depth of the beam ending with a sharp
maximum at the end of the particle range (the Bragg peak), at the time
that the very low scattering properties produce a very narrow penumbra in
the beam. Second, there is an intense ionization pattern along the particle
path, and notably at the end of its range, which results in localized bursts
of energy deposition at microscopic level yielding increased cell killing and
thus a radiobiological effect largely superior to that of conventional
radiotherapy beams, even in conditions of hypoxia (cf. [115–117]). The
resolution achieved is of 2–3 mm, which poses strict demands in target
localization. Because of their lower production costs, the use of protons in
radiotherapy has become well established and today the number of patients
Future applications of Monte Carlo 47
treated with this modality is close to 30 000 [118]. From the two properties
mentioned above the physical dose distribution of protons, enabling accurate
dose conformation delivery, poses a clear advantage over conventional
radiotherapy beams; however, their radiobiological effect is only about
810% higher than that of photons or electrons. The radiobiological
superiority of ions heavier than protons, with an RBE augmentation of the
order of three to four times higher than that of conventional beams, led to
an increased interest worldwide that resulted in the construction of the
Japanese HIMAC clinical facility in Chiba, near Tokyo. HIMAC started
the treatment of patients mainly with carbon ions in 1994 [119], and more
than 1000 patients have been treated in this facility. It had become clear
that whereas heavy ions like the neon beams used in the early 1970s in
Berkeley have the largest radiobiological effects, these also appear in regions
close to the beam entrance and in the plateau region, where usually normal
tissue is situated. In addition, due to the large penetration of the fragmenta-
tion products released by the incident ions, the tail of the dose distribution
beyond the Bragg peak may be too high for sparing normal tissue beyond
the primary ion range. These two aspects suggest that the ideal ions for
radiotherapy are the light ions, and for this reason carbon has dominated
the clinical applications at HIMAC. The GSI heavy-ion physics research
facility in Darmstadt, Germany, initiated clinical treatments also with
carbon ions in 1997 [115]. GSI uses advanced three-dimensional beam
scanning techniques and PET imaging to monitor the delivery of dose by
visualizing the positrons emitted by 10 C and 11 C ions produced in the nuclear
reactions of the incident carbon ions with tissue, together with radio-
biologically optimized treatment planning. Plans for building additional
radiotherapy light-ion facilities have followed in Europe, and feasibility studies
have been conducted for such facilities in Germany, Austria, Italy, France,
Sweden and Switzerland, among others [120].
Several groups have reported the applicability of PET to in vivo dosi-
metry for proton radiotherapy using the Monte Carlo method. Different
Monte Carlo codes have been used to investigate this challenging field
including Berger’s PTRAN [121], used by Del Guerra et al. [122], FLUKA
[123], and specially designed software [114]. Parodi and Enghardt [123], in
particular, made Monte Carlo simulations of the number and spatial distri-
bution of the positron emitters created by therapeutic protons in PMMA.
Since proton therapy requires a particle fluence up to 20 times higher than
that required in carbon therapy for the same physical dose level, the authors
concluded that the total activity produced within the irradiated volume
should be twice as intense as that actually used at the carbon ion facility at
GSI. Moreover, the considered ratio between the fluence of protons and
carbon ions could increase considerably due to the higher relative biological
effectiveness of carbon ions, leading to a lower physical dose than for protons
at the same level of biological dose.
REFERENCES
[1] Berger M J 1968 MIRD pamphlet 2: energy deposition in water by photons from
point isotropic sources J. Nucl. Med. 9 15–25
[2] Andreo A 1991 Monte Carlo techniques in medical radiation physics Phys. Med.
Biol. 36 861–920
[4] Jenkins T M, Nelson W R and Rindi A 1998 Monte Carlo Transport of Electrons and
Photons (New York: Plenum)
[5] Ljungberg M, Strand S-E and King M A 1998 Monte Carlo Calculations in Nuclear
Medicine: Applications in Diagnostic Imaging (London: IOP Publishing)
[6] Boone J M, Buonocore M H and Cooper V N 2000 Monte Carlo validation in
diagnostic radiological imaging Med. Phys. 27 1294–304
[7] Ma C M and Jiang S B 1999 Monte Carlo modelling of electron beams from medical
accelerators Phys. Med. Biol. 44 R157–89
[8] Ahnesjo A and Aspradakis M M 1999 Dose calculations for external photon beams
in radiotherapy Phys. Med. Biol. 44 R99–155
[9] Stabin M , Konijnenberg M, Knapp F F and Spencer R H 2000 Monte Carlo
modeling of radiation dose distributions in intravascular radiation therapy Med.
Phys. 27 1086–92
[10] Reynaert N, Verhaegen F, Taeymans Y, Van Eijkeren M and Thierens H 1999
Monte Carlo calculations of dose distributions around 32P and 198Au stents for
intravascular brachytherapy Med. Phys. 26 1484–91
[11] Lee C L, Zhou X L, Kudchadker R J, Harmon F and Harker Y D 2000 A Monte
Carlo dosimetry-based evaluation of the 7Li(p,n)7Be reaction near threshold for
accelerator boron neutron capture therapy Med. Phys. 27 192–202
[12] Levin V, Andreo P and Dodd B eds 2001 Current Status of Neutron Capture Therapy.
IAEA Tec-Doc 1223 (Vienna: IAEA, International Atomic Energy Agency)
[13] Gierga D P, Yanch J C and Shefer R E 2000 Development and construction of a
neutron beam line for accelerator-based boron neutron capture synovectomy
Med. Phys. 27 203–14
[14] Gierga D P, Yanch J C and Shefer R E 2000 An investigation of the feasibility of
gadolinium for neutron capture synovectomy Med. Phys. 27 1685–92
[15] Zaidi H 1996 Quantitative SPECT: recent developments in detector response,
attenuation and scatter correction techniques Physica Medica 12 101–17
[16] Webb S, Binnie D M, Flower M A and Ott R J 1992 Monte Carlo modelling of the
performance of a rotating slit-collimator for improved planar gamma-camera
imaging Phys. Med. Biol. 37 1095–108
[17] Ljungberg M and Strand S-E 1989 A Monte Carlo program for the simulation of
scintillation camera characteristics Comput. Meth. Programs Biomed. 29 257–72
[18] de Vries D J et al. 1990 Development and validation of a Monte Carlo simulation of
photon transport in an Auger camera IEEE Trans. Med. Imaging 9 430–8
[19] Dahlbom M, Rosenquist G, Eriksson L and Bohm C 1989 A study of the possibility
of using multi-slice PET systems for 3D imaging IEEE Trans. Nucl. Sci. 36 1066–71
[20] Zaidi H, Scheurer A H and Morel C 1999 An object-oriented Monte Carlo simulator
for 3D cylindrical positron tomographs Comput. Meth. Programs Biomed. 58
13345
References 49
[21] Dahlbom M et al. 1998 A YSO/LSO phoswich array detector for single and
coincidence photon imaging IEEE Trans. Nucl. Sci. 45 1128–32
[22] Derenzo S E 1981 Monte Carlo calculations of the detection efficiency of NaI(Tl),
BGO, CsF, Ge and plastic detectors for 511 keV photons IEEE Trans. Nucl. Sci.
28 131–6
[23] Derenzo S E and Riles J K 1982 Monte Carlo calculations of the optical coupling
between bismuth germanate crystals and photomultiplier tubes IEEE Trans. Nucl.
Sci. 29 191–5
[24] Bottigli U et al. 1985 Monte Carlo simulation and experimental tests on BGO, CsF
and NaI(Tl) crystals for positron emission tomography J. Nucl. Med. Allied Sci. 29
221–7
[25] Bice A N et al. 1990 Monte Carlo simulation of BaF2 detectors used in time-of-flight
positron emission tomography IEEE Trans. Nucl. Sci. 37 696–701
[26] Lopes M I, Chepel V, Carvalho J C, Ferreira Marques R and Policarpo A J P 1995
Performance analysis based on a Monte Carlo simulation of a liquid xenon PET
detector IEEE Trans. Nucl. Sci. 42 2298–302
[27] Bradshaw J, Burnham C and Correia J 1985 Application of Monte Carlo methods to
the design of SPECT detector systems IEEE Trans. Nucl. Sci. 32 753–7
[28] Kimiaei S, Ljungberg M and Larsson S A 1997 Evaluation of optimally designed
planar-concave collimators in single-photon emission tomography Eur. J. Nucl.
Med. 24 1398–404
[29] Pavlopoulos S and Tzanakos G 1996 Design and performance evaluation of a high-
resolution small animal positron tomograph IEEE Trans. Nucl. Sci. 43 3249–55
[30] Moses W W, Virador P R G, Derenzo S E, Huesman R H and Budinger T F 1997
Design of a high-resolution, high-sensitivity PET camera for human brains and
small animals IEEE Trans. Nucl. Sci. 41 1487–91
[31] Moisan C, Tupper P, Rogers J G and de Jong J K 1996 A Monte Carlo study of the
acceptance to scattered events in a depth encoding PET camera IEEE Trans. Nucl.
Sci. 43 1974–80
[32] Ljungberg M, Strand S-E, Rajeevan N and King M A 1994 Monte Carlo simulation
of transmission studies using a planar source with a parallel collimator and a line
source with a fan-beam collimator IEEE Trans. Nucl. Sci. 41 1577–84
[33] Wegmann K et al. 1999 Investigation of the scatter contribution in single photon
transmission measurements by means of Monte Carlo simulations IEEE Trans.
Nucl. Sci. 46 1184–90
[34] Gustafsson A et al. 1998 Evaluation of attenuation corrections using Monte Carlo
simulated lung SPECT Phys. Med. Biol. 43 2325–36
[35] Zaidi H 2001 Scatter modelling and correction strategies in fully 3-D PET Nucl.
Med. Commun. 22 1181–4
[36] El Fakhri G N et al. 1999 Respective roles of scatter, attenuation, depth-dependent
collimator response and finite spatial resolution in cardiac single-photon emission
tomography quantitation: a Monte Carlo study Eur. J. Nucl. Med. 26 437–46
[37] Zaidi H 2000 Comparative evaluation of scatter correction techniques in 3D
positron emission tomography Eur. J. Nucl. Med. 27 1813–26
[38] Smith M F, Floyd C E, Jaszczak R J and Coleman R E 1992 Three-dimensional
photon detection kernels and their application to SPECT reconstruction Phys.
Med. Biol. 37 605–22
[39] Dunn W L 1981 Inverse Monte Carlo analysis J. Comput. Phys. 41 154–66
[40] Floyd C E, Jaszczak R J, Greer K L and Coleman R E 1986 Inverse Monte Carlo as
a unified reconstruction algorithm for ECT J. Nucl. Med. 27 1577–85
[41] Hoffman E J, Cutler P D, Digby W M and Mazziotta J C 1990 3-D phantom to
simulate cerebral blood flow and metabolic images for PET IEEE Trans. Nucl.
Sci. 37 616–20
[42] Pollard K R, Bice A N, Eary J F, Durack L D and Lewellen T K 1992 A method for
imaging therapeutic doses of iodine-131 with a clinical gamma camera J. Nucl. Med.
33 771–6
[43] Dewaraja Y K, Ljungberg M and Koral K F 2000 Accuracy of 131I tumor
quantification in radioimmunotherapy using SPECT imaging with an ultra-high-
energy collimator: Monte Carlo study J. Nucl. Med. 41 1760–7
[44] Zaidi H 1996 Assessment of thyroid volume with pinhole emission computed
tomography Physica Medica XII 97–100
[45] Huili W, Jaszczak R J and Coleman R E 1996 Monte Carlo modeling of penetration
effect for iodine-131 pinhole imaging IEEE Trans. Nucl. Sci. 43 3272–7
[46] Smith M F and Jaszczak R J 1997 The effect of gamma ray penetration on angle-
dependent sensitivity for pinhole collimation in nuclear medicine Med. Phys. 24
1701–9
[47] Dewaraja Y K, Ljungberg M and Koral K F 2001 Monte Carlo evaluation of object
shape effects in iodine-131 SPECT tumor activity quantification Eur. J. Nucl. Med.
28 900–6
[48] Dewaraja Y K, Ljungberg M and Koral K F 2000 Characterization of scatter and
penetration using Monte Carlo simulation in 131I imaging J. Nucl. Med. 41 123–30
[49] Ellett W H and Humes R M 1971 Absorbed fractions for small volumes containing
photon-emitting radioactivity J. Nucl. Med. Suppl. 5 25–32
[50] Snyder W, Ford M R and Warner G 1978 Estimates of Specific Absorbed Fractions
for Photon Sources Uniformly Distributed in Various Organs of a Heterogeneous
Phantom, NM/MIRD (Society of Nuclear Medicine Publication, New York,
Pamphlet No 5)
[51] Akabani G, Poston J W and Bolch W E 1991 Estimates of beta absorbed fractions in
small tissue volumes for selected radionuclides J. Nucl. Med. 32 835–9
[52] Siegel J A and Stabin M 1994 Absorbed fractions for electrons and beta particles in
spheres of various sizes J. Nucl. Med. 35 152–6
[53] Cristy M 1983 Applying the reciprocal dose principle to heterogeneous phantoms:
practical experience from Monte Carlo studies Phys. Med. Biol. 28 1289–303
[54] Cristy M and Eckerman K F 1987 Specific Absorbed Fractions of Energy at Various
Ages from Internal Photon Sources I Methods, II One Year Old, III Five Year Old, IV
Ten Year Old, V Fifteen Year Old Male and Adult Female, VI New-Born and VII
Adult Male ORNL/TM 8381/V1-V7 Oak Ridge National Laboratory.
[55] Poston J W, Kodimer K A and Bolch W E 1996 A revised model for the calculation
of absorbed energy in the gastrointestinal tract Health Phys. 71 307–14
[57] Bice A N, Links J M, Wong D F and Wagner H N 1985 Absorbed fractions for dose
calculations of neuroreceptor PET studies Eur. J. Nucl. Med. 11 127–31
[58] Bouchet L G, Bolch W E, Weber D A, Atkins H L and Poston J W 1999 MIRD
Pamphlet No 15: Radionuclide S values in a revised dosimetric model of the adult
head and brain. Medical Internal Radiation Dose J. Nucl. Med. 40 62S–101S
References 51
[59] Bolch W E et al. 1999 MIRD pamphlet No 17: the dosimetry of nonuniform activity
distributions—radionuclide S values at the voxel level. Medical Internal Radiation
Dose Committee J. Nucl. Med. 40 11S–36S
[60] Bouchet L G, Bolch W E, Howell R W and Rao D V 2000 S values for radionuclides
localized within the skeleton J. Nucl. Med. 41 189–212
[61] Bouchet L G and Bolch W E 1999 Five pediatric head and brain mathematical
models for use in internal dosimetry J. Nucl. Med. 40 1327–36
[62] Mardirossian G et al. 1999 A new rectal model for dosimetry applications J. Nucl.
Med. 40 1524–31
[63] Thomas S R, Stabin M, Chen C T and Samaratunga R C 1999 MIRD Pamphlet No
14 revised: A dynamic urinary bladder model for radiation dose calculations Task
Group of the MIRD Committee, Society of Nuclear Medicine J. Nucl. Med. 40
102S–123S
[64] Clairand I et al. 2000 Improvement of internal dose calculations using mathematical
models of different adult heights Phys. Med. Biol. 45 2771–85
[65] Peter J, Gilland D R, Jaszczak R J and Coleman R E 1999 Four-dimensional
superquadric-based cardiac phantom for Monte Carlo simulation of radiological
imaging systems IEEE Trans. Nucl. Sci. 46 2211–17
[66] Segars W P, Lalush D S and Tsui B M W 2001 Modeling respiratory mechanics in
the MCAT and spline-based MCAT phantoms IEEE Trans. Nucl. Sci. 48 89–97
[67] Xu X G, Chao T C and Bozkurt A 2000 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. 78 476–86
[68] Chao T C and Xu X G 2001 Specific absorbed fractions from the image-based VIP-
Man body model and EGS4-VLSI Monte Carlo code internal electron emitters
[69] Mackie T R, Bielajew A F, Rogers D W and Battista J J 1988 Generation of
photon energy deposition kernels using the EGS Monte Carlo code Phys. Med.
Biol. 33 1–20
[70] Furhang E E, Sgouros G and Chui C S 1996 Radionuclide photon dose kernels for
internal emitter dosimetry Med. Phys. 23 759–64
[71] Berger M J 1971 MIRD Pamphlet 7: Distribution of absorbed doses around point
sources of electrons and beta particles in water and other media J. Nucl. Med. 12
5–23
[72] Simpkin D J and Mackie T R 1990 EGS4 Monte Carlo determination of the beta
dose kernel in water Med. Phys. 17 179–186
[73] Bardies M and Myers M J 1996 Computational methods in radionuclide dosimetry
[74] Halbleib J A, Kensek R P, Valdez G D, Seltzer S M and Berger M J 1992 ITS: the
Integrated TIGER Series of electron/photon transport codes—version 3.0 IEEE
[75] Sgouros G et al. 1993 Three-dimensional dosimetry for radioimmunotherapy
treatment planning J. Nucl. Med. 34 1595–601
[76] Johnson J C, Langhorst S M, Loyalka S K, Volkert W A and Ketring A R 1992
Calculation of radiation dose at a bone-to-marrow interface using Monte Carlo
modeling techniques (EGS4) J. Nucl. Med. 33 623–8
transport. I. EGSnrc, the new EGS4 version Med. Phys. 27 485–98
[78] Simpkin D J, Cullom S J and Mackie T R 1992 The spatial and energy dependence of
bremsstrahlung production about beta point sources in H2O Med. Phys. 19 105–14
[79] Leichner P K 1994 A unified approach to photon and beta particle dosimetry J. Nucl.
Med. 35 1721–9
[80] Casciari J J, Graham M M and Rasey J S 1995 A modeling approach for quantifying
tumor hypoxia with [F-18]fluoromisonidazole PET time–activity data Med. Phys. 22
1127–39
[81] Millet P, Delforge J, Pappata S, Syrota A and Cinotti L 1996 Error analysis on
parameter estimates in the ligand-receptor model application to parameter
imaging using PET data Phys. Med. Biol. 41 2739–56
[82] Burger C and Buck A 1996 Tracer kinetic modelling of receptor data with
mathematical metabolite correction Eur. J. Nucl. Med. 23 539–45
[83] Thie J A, Smith G T and Hubner K F 1997 Linear least squares compartmental-
model-independent parameter identification in PET IEEE Trans. Med. Imaging 16
11–16
[84] Welch A, Smith A M and Gullberg G T 1995 An investigation of the effect of finite
system resolution and photon noise on the bias and precision of dynamic cardiac
SPECT parameters Med. Phys. 22 1829–36
[85] Zeng G L, Gullberg G T and Huesman R H 1995 Using linear time-invariant system
theory to estimate kinetic parameters directly from projection measurements IEEE
[86] Ping-Chun C, Rogers W L, Clinthorne N H, Fessler J A and Hero A O 1994 Model-
based estimation for dynamic cardiac studies using ECT IEEE Trans. Med. Imaging
13 217–26
[87] Huesman R H, Reutter B W, Zeng G L and Gullberg G T 1998 Kinetic parameter
estimation from SPECT cone-beam projection measurements Phys. Med. Biol. 43
973–82
[88] Simpkin D J 1989 Shielding requirements for constant-potential diagnostic x-ray
beams determined by a Monte Carlo calculation Health Phys. 56 151–64
[89] Metzger R, Richardson R and Van Riper K A 1993 A Monte Carlo model for
retrospective analysis of shield design in a diagnostic x-ray room Health Phys. 65
164–71
[90] Biggs P J 1991 Calculation of shielding door thicknesses for radiation therapy
facilities using the ITS Monte Carlo program Health Phys. 61 465–72
[91] Verhaegen F, Buffa F M and Deehan C 2001 Quantifying effects of lead shielding in
electron beams: a Monte Carlo study Phys. Med. Biol. 46 757–69
[92] da Silva A X and Crispim V R 2001 Moderator–collimator–shielding design for
neutron radiography systems using 252Cf Appl. Radiat. Isot. 54 217–25
[93] Howard W B and Yanch J C 1995 Shielding design and dose assessment for
accelerator based neutron capture therapy Health Phys. 68 723–30
[94] Kramer G H, Chamberlain M J and Yiu S 1997 A study of thyroid radioiodine
monitoring by Monte Carlo simulations: implications for equipment design Phys.
Med. Biol. 42 2175–82
[95] Kramer G H and Crowley P 2000 The assessment of the effect of thyroid size and
shape on the activity estimate using Monte Carlo simulation Health Phys. 78 727–38
[96] Rodriguez Villafuerte M, Gamboa deBuen I and Brandan M E 1997 Monte Carlo
simulation of depth-dose distributions in TLD-100 under 90Sr–90Y irradiation
Health Phys. 72 574–8
References 53
[97] Hickman D P, Kruchten D A, Fisher S K and Anderson A L 1994 Calibration of a

241
Am wound-monitoring system using Monte Carlo techniques Health Phys. 66
400–6
[98] Gualdrini G et al. 2000 Development and characterisation of a head calibration
phantom for in vivo measurements of actinides Appl. Radiat. Isot. 53 387–93
[99] Zavgorodni S F 2001 A method for calculating the dose to a multi-storey building
due to radiation scattered from the roof of an adjacent radiotherapy facility Med.
Phys. 28 1926–30
[100] Osei E K and Kotre C J 2001 Equivalent dose to the fetus from occupational
exposure of pregnant staff in diagnostic radiology Br. J. Radiol. 74 629–37
[101] Pattison J E 1999 Finger doses received during 153Sm injections Health Phys. 77
5305
[102] Zankl M 1999 Personal dose equivalent for photons and its variation with dosimeter
position Health Phys. 76 162–70
[103] Xu X G, Reece W D and Poston J W 1995 A study of the angular dependence
problem in effective dose equivalent assessment Health Phys. 68 214–24
[104] Sparks R B, Siegel J A and Wahl R L 1998 The need for better methods to
determine release criteria for patients administered radioactive material Health
Phys. 75 385–8
[105] Johnson T K and Barnhart T 1999 An alternative to Monte Carlo in determining
release criteria for patients administered radioactive material Health Phys. 77
724–6
[106] Siegel J A, Sparks R B and Wahl R L 1999 Response to Johnson and Barnhart
[107] Akabani G, Hawkins W G, Eckblade M B and Leichner P K 1997 Patient-specific
dosimetry using quantitative SPECT imaging and three-dimensional discrete
Fourier transform convolution J. Nucl. Med. 38 308–14
[108] Furhang E E, Chui C S, Kolbert K S, Larson S M and Sgouros G 1997
Implementation of a Monte Carlo dosimetry method for patient-specific internal
emitter therapy Med. Phys. 24 1163–72
[109] Tagesson M, Ljungberg M and Strand S E 1996 A Monte-Carlo program converting
activity distributions to absorbed dose distributions in a radionuclide treatment
planning system Acta Oncol. 35 367–72
absorbed dose distribution estimates for patient-specific dosimetry J. Nucl. Med.
42 662–9
[111] Enghardt W, Fromm W D, Manfrass P and Schardt D 1992 Limited-angle 3D
reconstruction of PET images for dose localization in light ion tumour therapy
[112] Enghardt W et al. 1992 The spatial distribution of positron-emitting nuclei generated
by relativistic light ion beams in organic matter Phys. Med. Biol. 37 2127–31
[113] Enghardt W et al. 1999 The application of PET to quality assurance of heavy-ion
tumor therapy Strahlenther. Onkol. 175 Suppl 2 33–6
[114] Litzenberg D W, Becchetti F D and Roberts D A 1997 On-line PET monitoring of
radiotherapy beams: image reconstruction and Monte Carlo simulations of detector
geometries IEEE Trans. Nucl. Sci. 44 1646–57
[115] Kraft G 2000 Tumor therapy with heavy charged particles Prog. Part. Nucl. Phys. 45
S473–S544
[116] Amaldi U 1999 Cancer therapy with particle accelerators Nucl. Phys. A 654 375–99
[117] Chu W T 1999 Hadron therapy Biomedical Uses of Radiation—Therapeutic
Applications ed W R Hendee (Weinheim: VCH–Wiley) Part B, pp 1057–131
[118] Sisterson J 2001 Particles PTCOG Newsletter 27 January
[119] Kanai T et al. 1999 Biophysical characteristics of HIMAC clinical irradiation system
for heavy-ion radiation therapy Int. J. Radiat. Oncol. Biol. Phys. 44 201–10
[120] Rosso D 2000 A boost for hadron therapy Phys. World
[121] Berger M J 1993 Proton Monte Carlo Transport Program PTRAN Report NISTIR
5113 (Bethesda, MD: NIST)
[122] Del Guerra A, Di Domenico G and Mukhopadhayay D 1997 PET dosimetry in
proton radiotherapy: a Monte Carlo Study Appl. Radiat. Isot. 18 1617–24
[123] Parodi K and Enghardt W 2000 Potential application of PET in quality assurance of
proton therapy Phys. Med. Biol. 45 N151–6
Chapter 3
Medical imaging techniques for radiation

dosimetry
Kenneth F Koral, Habib Zaidi and Michael Ljungberg
3.1. INTRODUCTION
Radiation dose itself cannot be imaged non-invasively. However, medical

imaging can non-invasively provide information so that estimates of radia-
tion dose can be calculated. These estimates can include those for average
dose, dose spatial distribution, dose rate, etc. The status of such non-invasive
imaging was reviewed comprehensively in 1993 [1] and again in 1999 [2]. At
this time, the imaging has three distinct areas in which it contributes.
1. Determination of the spatial distribution of mass and of attenuation
coefficient.
2. Segmentation of the patient into volumes of interest (VoIs) that repre-
sent specific organs, tumours, etc. for which the estimate of radiation
dose is desired.
3. Determination of the distribution of radioactivity (MBq) in space and
time. (Radioactivity will be shortened to activity below.)
We will cover each of these areas in the next three sections. Before proceeding,
we note that medical imaging can be carried out for two broad dosimetric
purposes:
1. To monitor radiation dose during therapy.
2. To predict radiation dose during therapy from a pre-therapy evaluation,
often given the name of treatment planning.
It is clear that if monitoring is the purpose then imaging during therapy may
be involved, while if treatment planning is the purpose then imaging during a
pre-therapy evaluation will be necessary. However, the accuracy of predict-
ing results during therapy from measurements carried out during evaluation
can be questioned. The matter will be discussed further at the end of this
55
56 Medical imaging techniques for radiation dosimetry
chapter. Also, monitoring radiation dose during therapy may be the purpose
of interest, but you may still choose to do it from evaluative imaging rather
than from intra-therapy imaging. This choice has the advantage of lower
radiation exposure to imaging personnel, and of having no worries about
camera deadtime.
3.2. DETERMINATION OF THE TOTAL MASS. DETERMINATION

OF SPATIAL DISTRIBUTION OF ATTENUATION
COEFFICIENT, AND OF MASS
If one is using traditional techniques based on Medical Internal Radiation

Dose (MIRD) methods, one may employ a standard reference man for a
macrodose calculation. Here macrodose is defined as the average dose for
a large-size object such as a liver or a tumour. In this case, only the time
integral of the total activity (MBq h) within the organ is needed. This time
integral must be measured or inferred for a particular patient and so the
resultant dose is patient specific. This statement is made using the phrase
in its most rigorous sense. The usual use of the phrase ‘patient specific’
implies including additional input about the patient. If one wishes to
carry out a patient-specific dose calculation in this more general sense, it is
important to also estimate at least the total mass of the organ or tumour
of interest.
The total mass of a tumour or organ is sometimes determined from VoIs
that are directly calculated using the emission image. Another method is to
outline specific organs and tumours on an anatomic image. For some
organs, like the liver or spleen, the outlining on an anatomic image is
fairly straightforward. For the kidney, where the shape is not simple, or
for some tumours where the edge is not clear, the outlining is harder. Once
the volume in pixels is known, it is converted to cm3 by using the volume
of a single pixel, which has been determined by a previous calibration.
Except for the lungs, or bone, one usually assumes a density of 1 g/cm3 to
calculate the mass from the volume.
Once the total mass is estimated for an organ, it can be ratioed with the
mass in the reference-man phantom to provide a mass-based correction to the
first-order MIRD dose (that from electrons). Or the mass can be used in a
more-complicated calculation of the radiation macrodose that includes the
dose from -rays [3]. For tumours, the total mass can be used in a calculation
of the radiation macrodose that usually goes back to basic considerations.
Photon (or -ray) attenuation is the common name for photon absorp-
tion or scattering with or without energy degradation that occurs during
acquisition of projection data. Many reconstruction algorithms include
compensation for photon attenuation that is based on imaging. Attenuation
correction based on imaging is discussed further in sections 3.4.5 and 3.4.9.
Segmentation of patient data into VoI 57
Three-dimensional mass distributions (and perhaps material distribu-

tions) are necessary to compute three-dimensional dose distributions if one
is not going to assume a water-like medium within the body skin surface.
These distributions can be obtained from the X-ray computed tomography
(CT) or a nuclear magnetic resonance image (MRI) if image registration
has been carried out [4]. Empirical approaches are required, however.
3.3. SEGMENTATION OF PATIENT DATA INTO VoI
With conjugate-view imaging, if an anatomical image set such as one from CT

or one from MRI is available, then tumours can be outlined in the anatomical
space. These outlines are then available for use in the projection space with
techniques such as computer assisted matrix inversion [5] discussed in section
3.4.2. Functional volume estimation from planar imaging has shown its
limitations when compared with three-dimensional imaging [6].
With tomographic imaging without registration of image sets, target
segmentation in the image space is needed. This statement is true for both
SPECT and positron emission tomography (PET). With SPECT, many of
the segmentation publications do not specifically deal with the imaging of
131
I or other higher-energy, single-photon emitters; they will not be reviewed
here. In their early work on segmentation with 131 I, Fleming and colleagues
[7, 8] use one of two related methods to choose a threshold percentage. This
percentage multiplies the maximum pixel to give a threshold pixel value.
Pixels containing a value greater than or equal to this value are included in
the volume of interest. Both methods require an initial box of interest that
is manually chosen to surround the object. The first method is called maximi-
zation of interclass variance [9]. For this, the threshold is calculated assuming
there are two levels within the box, one for the object, and one for the
surround. The weighted sum of the two squares of the difference between
the mean value of the level and the mean value of the entire box is calculated.
The threshold is chosen to make this weighted sum a maximum. The idea is
that when the threshold is correct, both the upper level and the lower level
will be far from the fixed mean of the entire box and so the above two-
component sum will be large. The second method replaces the manually
chosen box with the VoI chosen by the method above. It then expands this
VoI in all directions by a set number of pixels. Next, it maximizes the
same weighted sum referred to above to find a new threshold. This procedure
is repeated, yielding an iterative approach. The stopping criterion was not
specified [7]. Recently, Fleming and colleagues have proposed case-specific
volume-of-interest segmentation rules [8], but they have not yet applied
them to radionuclides such as 131 I.
With PET Erdi et al. use a method wherein the threshold percentage that
determines the target outline is chosen so as to fit the situation [10], which is
similar to the use of case-specific rules referred to above. That is, Erdi et al.
propose an initial segmentation, after which an approximate target to back-
ground ratio is calculated. In addition, the target volume is estimated from
CT. Then, depending on the ratio and the volume, a threshold is chosen
with which to carry out the final segmentation. Another approach, which
is receiving considerable attention, is the use of fuzzy segmentation techni-
ques. They demonstrate excellent performance and produce good results as
automated, unsupervised tools for segmenting noisy images in a robust
manner. Based on the Fuzzy C-Means (FCM) algorithm, Boudraa et al.
segmented PET images into a given number of clusters to extract specific
areas of differing activity concentration (MBq cm3 ) for brain lesion quanti-
fication purposes [11].
With SPECT or PET imaging, if a registration between the radio-
pharmaceutical tomographic image set and CT or MRI is available, the
VoI outlines from CT or MRI can simply be transferred to the three-
dimensional space of the reconstructed activity distribution [12]. If a
multimodality imager is employed, a phantom calibration provides the
transformation to place the CT outline into the activity space [13]. In that
space, the outlines provide the VoI over which the total activity is found,
or, if pixel-by-pixel dose, or pixel-by-pixel dose rate, is calculated, the
volume over which to investigate the dose parameter. Image registration is
discussed further in section 3.4.7.
3.4. DETERMINATION OF THE DISTRIBUTION OF

RADIOACTIVITY IN BOTH SPACE AND TIME
In the accurate estimation of dose from imaging, it is very important to attain

an accurate measurement for the amount of activity. Determining an activity
value that is correct in the absolute sense is ideal and the ultimate goal.
However, it should be pointed out that, within a single study, the activity
value can be self-consistent but incorrect by a constant multiplier. That is,
for example, such an error does not disturb the correlation of calculated
radiation dose with response. When one wants to compare radiation-dose
results from different groups, however, or combine them, then the impor-
tance of estimating absolute activity is clear.
Similarly, in a given patient, an accurate estimate of the distribution of
activity is ideal and frequently the goal. However, if one is interested in only
an estimate of the radiation dose to a tumour or organ as a whole, then one
needs only the total activity and may sacrifice knowing the distribution
within that tumour or organ.
Finally, to calculate radiation dose, one must know the time distribution
of activity for the target. If a certain amount is present within a sealed, fixed-
size object, then the amount will change at the rate of physical decay of the
Distribution of radioactivity in both space and time 59
radionuclide. However, even if a tumour or organ is a fixed size, or assumed

to be so, usually the amount of activity within that fixed size is changing by
the radiopharmaceutical entering and/or leaving. Therefore, one must
measure or estimate the dependence of the activity on time. If the tumour
size is changing as well, then one may define another fixed-size element,
such as a voxel within the tumour, determine the time–activity curve for
that voxel, and eventually calculate the dosimetry for it. Obtaining a time
distribution of activity is discussed in section 3.4.6.
Unlike CT or MRI, which look at anatomy or body morphology,
nuclear medicine imaging determines metabolic activity and tissue function.
We therefore next look at nuclear medicine imaging. The potential for
quantification of activity with whole-body imaging, spot conjugate views,
SPECT or PET is examined.
3.4.1. Quantitative whole-body imaging

Whole body imaging for the determination of the amount of activity present
in the body, without quantitative description of where, is of interest for the
estimation of whole-body radiation dose. This radiation dose is of interest,
in turn, because it has been shown to correlate with bone-marrow toxicity.
Bone-marrow toxicity is of great importance in all therapies that do not
employ bone-marrow support.
Whole-body imaging in this application competes with a probe survey of
the patient. With both methods, the end goal is the whole-body activity as a
function of time. In the dose calculation, the measured activity is assumed to
be uniformly distributed at each time point. The reason is partly to simplify
the calculation and partly because methods to calculate bone-marrow dose
rather than whole-body dose are difficult to devise and so no use can be
made of the spatial distribution.
Whole-body imaging is usually carried out with a scanning scintillation
camera. A set amount of time is used for the passage of the camera from the
head to the feet of the patient. The total number of counts in images acquired
at different times after the initial activity administration is then plotted as a
function of the start time of each image. Since the first image is taken
immediately after the injection or infusion of a known amount of activity,
the total measured counts correspond to that activity, yielding a conversion
factor. (The rationale is that at that time all of the administered activity must
be located within the patient and very little of it has yet decayed.) The conver-
sion factor between image counts and estimated total activity is then applied
to the count total in each subsequent image. The method and assumptions
for the calculation of whole-body dose from total activity can be found in
[14]. In 131 I metaiodobenzylguanidine therapy of neuroblastoma, the result-
ing whole-body dose has the largest coefficient for the correlation with
marrow toxicity compared with other predictors [15]. In 131 I tositumomab
therapy of lymphoma, the correlation is best for treated patients who had
not previously undergone bone-marrow transplantation when the nadir in
platelet counts was used [16].
Advantages. This imaging is simple and self-calibrating.
Disadvantages. There is no resolution in space.
3.4.2. Quantitative imaging with conjugate views

For total activity within an isolated organ or tumour, one can use an imaging
method that is simpler than three-dimensional imaging with SPECT or PET.
That is, one can get the total activity using only two opposed views of the
patient, the so-called conjugate views. Note, however, that by sacrificing
the knowledge about the three-dimensional spatial distribution of the
activity, one gives up accurately estimating the three-dimensional spatial
distribution of the radiation dose within the organ or tumour. However, if
the added information on the dose distribution does not have an application,
and/or if there is no added accuracy in estimating total activity, then perhaps
one should save the extra effort and estimate the total activity with spot
conjugate views.
The methods for classical quantitative conjugate views [17] have been
summarized in detail [2]. The well known equation for activity, A, is:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
IA IP f
A¼ ð3:1Þ
T C
where IA is the count within the RoI in the anterior view, IP is the count in the
posterior view, T is the transmission through the body at the tumour location
from anterior to posterior, f is a correction that is 1 for small objects, and C is a
conversion factor that converts counts to activity. In this case, a transmission
measurement is taken explicitly and the value for C is derived from a measure-
ment in air, usually with a point source. Note that with this equation two
objects which are located behind each other along the anterior–posterior
direction, are completely unresolved. Thus, the resolution problem here is so
great that one obtains a sum of two activities. This sum activity may be
ascribed to one object whose volume is the sum of two objects. When radiation
dose is calculated for that object, one obtains a dose which is unfortunately not
even the average of the true doses to the two objects. Phantom tests of the
method using 131 I have been carried out [18, 19].
The Siegel reference above outlines how situations more complicated than
a single source in a non-radioactive attenuating medium can be handled [2].
One extension to a more complicated situation requires taking a lateral view
to obtain needed information. However, it is clear that there is error involved
in obtaining the needed extra information, at least from resolution problems.
Perhaps for this reason the use of a third view has not been widespread.
The original method assumes narrow beam geometry for attenuation,

while in the usual clinical case scattered photons are corrupting the
measurement. A pseudo-extrapolation number is one of the ways introduced
to counteract the corruption from inclusion of -rays that have been
scattered by the patient and accepted by the camera. This number (2.48 for
67
Ga with a medium-energy collimator, for example) is determined from
an experimentally measured curve of transmission versus water-phantom
thickness. It multiplies the T value in equation (3.1), thus reducing the
apparent activity by compensating for having counted scattered -rays [2].
Pixel-by-pixel scatter correction methods [20] can also be employed. For
131
I and usual collimators, the methods based on three or more energy
windows actually compensate for patient scatter of the primary -ray
(364 keV) and also for septal penetration of higher-energy -rays that deposit
only part of their initial energy in the scintillation crystal. Another approach
is one that employs buildup factors [2]. There are two versions of this
method. Common to both versions is the need to take calibration measure-
ments. The main problem is that the results from these measurements
depend on parameters such as the shape and size of the source. Therefore,
it is not clear how well they apply to a given target source in a given clinical
case.
Concern over complications from the transmission measurement or
from the calibration has led other investigators to related, but perhaps
more robust, methods. With one 131 I technique, a calibrated source is
included during the patient imaging [21]. This method has recently been
used for neuroblastoma patients undergoing therapy with 131 I metaiodo-
benzylguanidine. The resulting tumour self-absorbed dose had a statistically
significant ( p < 0:01) correlation with degree of overall tumour response for
27 patients [22].
Another conjugate view method with 131 I involves a transmission
measurement with a large-diameter 57 Co disc source [14]. It is based on
earlier research using 111 In [23]. It requires the acquisition of two phantom
calibration curves. One gives transmission versus thickness of water using
an 46 cm (18 inch) square cross section water bath. The other yields
efficiency as a function of thickness, where efficiency is geometric mean
counts/activity. It is determined by successively imaging a syringe source
of known activity that is centred in the bath filled with changing depths of
water. The data for each calibration are fitted with a second-order poly-
nomial and then thickness is eliminated between the two fitted curves to
yield a final working curve.
Disadvantages stemming from lack of three-dimensional resolution are
partially solved by the addition of a three-dimensional anatomical image in
computer-assisted matrix inversion [5]. A required condition, however, is
that you assume the activity is uniformly distributed in each of the targets
and within the background region that separates the targets. The method
also requires that a CT image set, registered with the projection image, is
available.
Advantages. These related methods are simple and lead directly to
macrodose. The simplicity is an aid for making the repeated measurements
needed for time–activity curves. Newer methods can employ information
from an anatomical image.
Disadvantages. Outlining each target region in one of the projections is
subject to error. Resolution in the anterior-to-posterior direction is either
non-existent or not of the highest quality. In some cases, results are obtained
for only one large target when there are actually two or more smaller targets.
3.4.3. Quantitative imaging with SPECT

Both SPECT and PET provide an estimate of the activity distribution in
three dimensions. However, this estimate is not totally accurate due to the
finite nature of the system spatial resolution. The lack of activity accuracy
causes errors in dosimetric estimation. Quantitative SPECT imaging itself
was well summarized in three review articles published between 1994 and
1996 [24, 25, 26]. We will describe the problem of activity quantification
with SPECT for radiopharmaceuticals of most interest in therapy (we usually
will use 131 I, since it has the highest familiarity to the authors). We will also
include some of the more recent developments in the field.
First, let us note that for 131 I and the usual high-energy general-purpose
collimator, the detector response to a point source at a distance in front of the
collimator features a considerable pedestal of counts due to septal penetration
from the primary (364 keV) -rays. Some of the events in this pedestal come
from 364 keV -rays that both penetrate the collimator septa without energy
loss and deposit their full energy in the camera crystal. Such events cannot
be eliminated by scatter-compensation methods such as the triple-energy-
window method [20] because their spectrum is identical to that of the correctly
collimated 364 keV photons. Quantification complications from these events
depend on how the counts in the reconstructed image are converted into
image activity. This conversion is sometimes called camera calibration.
Readers are therefore reminded that accuracy in activity quantification in
this case is intimately connected with the method of camera calibration.
For 131 I and a collimator with thicker septa, such as a commercial colli-
mator designed for positron imaging, or a special purpose, high-resolution,
low-efficiency rotating collimator, the detector response does not have the
pedestal but features a low-intensity hole pattern [27]. This hole pattern
would itself be expected to cause problems, but incorporating such a pattern
into a reconstruction program has recently been accomplished for 111 In
[28] and, with their rotating collimator, Gonzales-Trotter et al. argue that
the intensity of the hole pattern is so small that it can be neglected [29].
With standard high-energy collimator imaging of a single target volume

of interest in a non-radioactive background, activity spills out into other
areas of the three-dimensional image. However, this spill-out is not a
problem in this case: one simply uses a volume of interest that is larger
than the true target volume [8].
On the other hand, usually there are other targets or structures fairly
nearby that contain activity, and a low level of activity in the regions between
these targets and structures. In this more general case, two things happen:
1. One must use a restricted VoI for the target of interest in order to not
include other targets or structures. Then some target activity spills out
of that restricted VoI.
2. There is spill into the target VoI from the other targets and structures
and from the low-activity-level background.
A solution to these problems has not yet been found. Below, we outline
current attempts to minimize them.
Links et al. [30] studied the problems specifically for brain imaging using
simulations with specified three-dimensional Gaussian spatial resolution.
Their reference list includes possible solutions for particular regions of
interest within the brain.
Gonzalez-Trotter et al. [29, 31] use a camera calibration that yields a
conversion factor between planar counts and true activity, plus a reconstruc-
tion program that effectively distributes one planar count from a projection
fractionally into many voxels in reconstruction space.
Their conversion factor depends on how many planar counts are
obtained by the camera–collimator system from a small in-air 131 I source
that has a known activity. They employ both attenuation and scatter–
penetration compensations and take collimator characteristics into account
in their Ordered-Subsets Expectation Maximization (OSEM) algorithm
[32]. For a sphere inside a water bath that contains a uniform activity concen-
tration, they compute activity as a function of the radius, r, of the volume of
interest employed. The computed activity is the difference between the
estimate of a large activity and that of a small underlying pedestal. For a
phantom measurement with a small sphere (1.1 cm radius) in a water bath
without background activity, the estimate of sphere activity approaches an
almost flat plateau at the correct activity value, as the radius increases
beyond that for the physical size of the sphere. At the radius of the sphere,
the true activity is slightly underestimated (by about 20%) by the region of
interest. These facts are shown in figure 3.1. Therefore, it appears that an
accurate estimate (within 5%) can be found from the plateau value.
Fleming and colleagues have advocated quantification that uses a
calibration factor proportional to that from a planar measurement of a
point source of known activity but also sensitive to the details of the
reconstruction algorithm [33]. To this calibration is added other procedures
Figure 3.1. Plot of activity as a function of the radius ri of the spherical volume
of interest. OSEM is defined in the text. Two other methods of reconstruction
are also presented for comparison. FBP ¼ standard filtered backprojection.
MAP/ICD ¼ maximum a posteriori/iterative coordinate descent. Reprinted
from [29] # 2001 IEEE, with permission.
that adapt the quantification to the particular case. These other procedures
include the use of regression equations [8]. In cases where there is no uniform
surrounding activity and no nearby objects that contain activity, a volume of
interest that exactly fits the object is used but only after expansion in all direc-
tions by one pixel to account for finite SPECT resolution [33]. The regression
equations relate the true activity to the measured activity with a constant
greater than 1. The constant is smaller when VoI expansion has been
employed, as is to be expected [33]. Also for 131 I, Fleming and colleagues
measured the activity values for variously shaped objects in a large cylindri-
cal water bath experimentally. Activity values ranged from 0.8 to 29.3 MBq.
Using segmentation based on maximization of interclass variance (see
section 3.3 above), the measured values compared with the true activities
fall quite close to the line of identity [8]. With some of the procedures of
Fleming and colleagues, it appears that the segmentation of the object of
interest and the determination of its activity are intimately connected,
rather than being two distinct steps.
Tang et al. have developed a set of methods for SPECT quantification
using a research multimodality CT-SPECT imager. They employ attenuation
correction and model the collimator but do not include scatter–penetration
correction. For three neuroblastoma patients being treated with 131 I meta-
iodobenzylguanidine, each with a large and a small tumour, they list activity
concentration results from eight methods, including conjugate views [13].
Table 3.1. Ratio indicating the variation in activity concentration with the
evaluation method as a function of tumour size and patient number.
Patient Tumour Ratio of maximum activity over minimum activity

number size
For all 7 methods For 5 methods
1 large 1.9 1.3

1 small 13 7.2
2 large 1.8 1.3
2 small 18 2.0
3 large 4.2 2.3
3 small 21 4.2
Herein, we exclude results from one method in a baseline mode that did not
include compensation for collimator effects. Then table 3.1 shows the ratio of
the largest estimate of activity concentration to the smallest by patient and
tumour size.
For the large tumours, the ratio ranges from 1.9 to 4.2. For the harder-
to-estimate small tumours, it ranges from 13 to 21. The authors explain that
one of these methods is likely to yield a lower bound on the activity while at
least one other is likely to yield an upper bound. Excluding the largest and
smallest result from the seven methods, one still gets a large variation as
shown in the fourth column of the table: the ratio is never less than a
factor of 1.3 for the large tumours and not less than a factor of 2.0 for the
small tumours. Clearly, for a given patient tumour, results can vary by a
considerable factor depending on the method of quantification employed.
So, the choice of method is important and, perhaps, difficult to optimize
for all targets.
Koral et al. [14] employ a conversion factor for focal activity, C, that
varies both with the radius of rotation, r, of the circular orbit of the
acquisition (which affects resolution) and also with a measured parameter,
b, that indicates the relative magnitude of count spill-in from a uniform
background concentration compared with count spill-out from the target.
Phantom measurements of the functional dependence of C on r and b are
made using a 200 cm3 sphere located inside an elliptical water-bath phantom.
In contrast to some of the methods of Tang [34], no explicit subtraction for
the effects of spill-in of counts from the background is needed. Presumably,
the noise-amplification of a subtraction is thus not present. The authors
emphasize that the parameters for the SAGE reconstruction [35] of their
calibration phantom are the same as for their target measurements, and
that the identity of the method for phantoms and patients is also true for
the attenuation compensation, the outlining of the target of interest, and
the scatter–penetration compensation. The attenuation compensation is
based on a fused CT image as the target outline is transferred between

SPECT and CT (such a possibility was discussed in section 3.3). In addition,
the authors correct for camera dead time. However, at day 2 after the therapy
administration it is usually less than 10%. Finally, the activity for a target
volume that is not equal to the sphere volume is corrected by using an empiri-
cal recovery coefficient. This coefficient is dependent on volume, and, for one
of two cameras, also on radius of rotation. The correction factor is 1 for a
volume equal to that of the calibration sphere, greater than 1 for a smaller
volume, and less than 1 for larger volumes. To date, target shapes, which
are much different from spherical, have a bias [36]. However, resultant
dose values tend to correspond with the degree of response in patients [37].
Advantages. One can obtain an estimate of the three-dimensional

activity distribution at a given time and one can use detailed models for
the imaging system to correct for problems from finite resolution.
Disadvantages. Since the data processing is usually complicated,
choosing this imaging can involve a large time commitment.
3.4.4. Effect of reconstruction algorithm on SPECT imaging

The accuracy of a SPECT reconstruction plays an important role in a three-
dimensional dosimetry calculation. Most methods for three-dimensional
absorbed dose calculation assume that the input, the SPECT image set,
reflects the actual activity distribution. That is, the methods simply calculate
from these images the absorbed dose, either by direct Monte Carlo methods
tracing individual particles or by utilizing dose-kernel convolution methods.
However, if the reconstruction method has inherent limitations, they will be
transferred to the absorbed dose images. Also, it is important to recognize
that a SPECT image set almost always includes a degree of blurring due to
the limited spatial resolution of the imaging system, as mentioned in the
immediately previous section.
The simplest reconstruction method is basic backprojection. This
method simply re-projects a measured projection data set into the image
space. The basic backprojection method is fast but generates severe streak
artefacts as well as image blurring. The streak artefacts can cause problems
for quantification in two ways. First, they can create counts in parts of the
image that may be outside the patient or in interior regions where there
may be no activity at all. The latter result can affect the detectability of
small, low-contrast lesions. Secondly, it may be that the normalization
procedure in the reconstruction methods preserves the measured counts,
that is, the number of counts in all projections is equal to the number of
counts in the reconstruction volume. If streak artefacts are present outside
the object then the count level inside the object will be lower in order to
fulfil the normalization criterion. This result may then affect the activity
Figure 3.2. Images of a 99m Tc SPECT simulation of six spheres of different

activity concentration. The left image shows the result from using FBP with a
ramp-filter. The right image shows the result using 12 iterations of MLEM
after the initial guess of a uniform image. The projection data were simulated
without scatter or attenuation. A non-linear display has been employed to
better visualize the low-intensity parts of the image.
determination for an organ or tumour. Therefore, basic backprojection

reconstruction has seen limited use in SPECT.
In the past and still today, projection data are often first filtered with a
ramp filter (sinc filter in the spatial domain) before reconstruction. This filter
theoretically only removes counts produced by the backprojection method-
ology and not those from measured projection data. Streak artefacts can
still occur, such as shown on the left of figure 3.2. (The projection data set
has been simulated with no scatter and attenuation in order to visualize
the effects generated only by the reconstruction process.) Generally, the
ramp filter needs to be rolled over with a smoothing low-pass filter to
reduce amplification of high-frequency noise present in the projection data.
The resulting procedure is called filtered backprojection (FBP). An alter-
native to using the rollover of the filter is to utilize post-reconstruction
three-dimensional filtering. Both these low-pass filtering methods unfortu-
nately degrade spatial resolution.
Recently, the emergence of fast computers has allowed the wide and
routine use of iterative reconstruction methods. These methods are gener-
ally based on comparing measured projections with projections calculated
from an initial guess of the source distribution. This guess can either be a
uniform image where all pixels are set equal or can be the results from an
FBP algorithm. The difference between the calculated and measured
projections is then used to change the pixels in the estimated image. A
common algorithm is the maximum-likelihood-expectation-maximization
(MLEM) algorithm [38], which is a statistical method. A given image esti-
mate is obtained from all the projection data. Then the image-estimation
procedure is repeated (iterated) a fixed number of times, or until a specified
stopping criterion is reached. Unfortunately, a quantitative stopping
criterion is yet to be discovered. A sample result is shown on the right of
figure 3.2.
Figure 3.3. These images show a slice reconstructed by iterative MLEM after
1, 5, 10 and 50 iterations. The images show a 131 I-labelled monoclonal antibody
in a patient at a slice location just above the bladder. Note the increase in noise
as the number of iterations becomes large.
As seen in figure 3.3, the number of iterations has an effect on the result
and so must be chosen carefully.
Newer algorithms use only sub-sets in a certain order before the image is
updated. This procedure has been shown to decrease the number of needed
iterations dramatically and is therefore one of a set of methods called accel-
erated methods. OSEM is such a commonly used method [32]. The iterative
methods do not have the streak artefact and are generally expected to be
more quantitatively accurate, especially in typical oncology ‘low-statistics’
studies.
3.4.5. Effect of attenuation, scatter and blurring correction on SPECT

imaging
In SPECT, photon attenuation can introduce errors in the reconstructed
image if the attenuation varies with the projection view. For example, in
cardiac imaging, photon attenuation can cause false ischemic regions in
the inferior walls. Photon attenuation also limits the accurate quantification
of measured data in terms of activity and activity concentration. Correcting
for this attenuation is very important. The attenuation distribution should
generally be measured for each patient to get the patient outline and the
internal variation in his/her density and composition. For absorbed dose
calculation, the technique with an external radioactive source (transmission
computed tomography or TCT) may be inaccurate because of the relatively
high noise level present and the poor spatial resolution of the attenuation
map. It is therefore somewhat more common to register a CT study of the
patient to the SPECT study.
One method for employing a registered CT to estimate the spatial
distribution (or map) of attenuation coefficients is based on a technique
introduced by Nickoloff et al. [39]. The variation [40] yields the following
piecewise linear equation:

CT
¼ 1:131 102 þ1 CT 25
868:6
¼ 5:744 106 CT þ 1:113 102 25 CT 485 ð3:2Þ

3 CT
¼ 8:929 10 þ1 CT 485
868:6
where is the attenuation coefficient for 364 keV gamma rays in mm1 and
CT is the X-ray CT number in Hounsfield units for a scanner with an
effective energy of 76 keV. Figure 3.4 shows a plot of the relationship. The
Nickoloff reference specifies the means of experimentally determining the
effective energy of a particular CT scanner.
Tang et al. [41] suggest that the above method is too simplistic if an
iodine contrast agent is employed in the CT scan and the SPECT image is
obtained immediately after the CT scan. In this case, they say that the
iodinated regions and the bone regions in the image should be separated
and treated differently. The reason is that these two regions effectively
have a different slope in the relationship for linear attenuation coefficient
versus CT. For their scanner at 140 kVp, they present results from a
Figure 3.4. Plot of linear attenuation coefficient for gamma rays from 131 I
versus the pixel CT number from an X-ray CT scanner with an effective
energy of 76 keV. It is assumed that above CT ¼ 485 HU (Hounsfield units)
the material is bonelike and below CT ¼ 25 HU tissue-like. Between
25 HU and 485 HU, it is assumed there is a transition region.
calibration experiment that show a slope difference of a factor of 10 when

converting from CT to the that is appropriate for the 364 keV -rays of
131
I, for example [34]. Their final relationship is also piecewise linear, but
with a change in slope at CT ¼ 0. Moreover, the slope for the higher CT
values is larger if the region has been designated bone than if it has been
designated soft tissue containing contrast material.
Attenuation correction can be applied (a) prior to reconstruction (an
example is the geometric mean of opposed projections method [42]), (b) post
reconstruction (an example is the Chang algorithm [43]) or (c) as part of an
iterative reconstruction algorithm (see for example [26]). As part of an iterative
reconstruction, the attenuation to the boundary is calculated for each pixel
along the ray before summation. The comparison is then between an attenu-
ated projection and a measured projection. This procedure will eventually
make the estimated SPECT image more accurate since attenuation is also
present in the measured projection. If a registered attenuation map is available,
then non-homogeneous attenuation can be easily implemented [4].
Scatter corruption of images originates from the poor resolution of the
NaI(Tl) crystal (about 9–10% full width half maximum at 140 keV). A rela-
tively large energy window is needed to acquire a sufficient number of
photons. This large window then includes small-angle scattered photons
that pass the lower-energy-window discriminator. The crystal location of
these events may not be consistent with a ray from their emission point
perpendicular to the crystal face. Thus, they should be eliminated by a
method such as the dual- [44] or triple-energy window scatter correction
[20] methods or by some other means.
The scatter problem described above is less acute as the photon energy
increases. However, for emitters that have both a high-energy primary
photon and also super-high-energy emissions (such as 131 I), scatter in the
crystal can create the addition of counts in a different way. If a super-high-
energy photon partly deposits its energy into the crystal and then escapes
the crystal, the event can increment the count in the main-energy photo-
peak window. If the photon has escaped the object without Compton
scattering, the position in the image of the partly absorbed event is correct
even though the energy is below the full-absorption peak of that photon.
For energy-window-based scatter correction methods, such as the triple-
energy window method, subtraction of these events may result in a decrease
in the number of photons that are correctly positioned in space. The result
will then be a higher noise level in the corrected image. Effects on the bias
of the activity estimate will be negative if the calibration has included such
events, but neutral if they have also been excluded there. In either case,
fortunately, the number of such events is usually small.
Scatter correction methods that are included in reconstruction programs
[45] have recently been developed by some research groups. Frey and Tsui
[46] have been developing a method where depth-dependent scatter functions
are used to model a spatially variant scatter distribution. This so called ESSE
method utilizes an attenuation map to compensate for non-homogeneous
attenuation. A fast Monte Carlo-based scatter compensation method has
also been developed for SPECT [47–49]. This method uses a rotation-
based Monte Carlo simulation that calculates scatter using various variance
reduction methods. An application is to calculate down-scatter in dual-
isotope SPECT. With this method, only a fraction of all SPECT projections
have to be simulated in a standard manner. The other projections can be
estimated rapidly using the results of these standard calculations. For
efficiency, approximations have to be made in the method with regard to
the final scatter angle of the detected photons. Further speed-up is obtained
by combining it with convolution-based forced detection instead of standard
forced detection, which is the more common variance reduction technique.
Rotation-based Monte Carlo with convolution-based forced detection was
found to be about three orders of magnitude faster than standard Monte
Carlo with forced detection. Even though SPECT has great potential, a
major drawback is its limited spatial resolution [50], mainly caused by the
inherent design of collimators. It is rare that the reconstructed spatial
resolution using commercial collimators is less than 10 mm for commonly
used radioisotopes in radionuclide therapy. Several groups have developed
methods for compensation for distance-dependent blurring. These methods
are basically attacking the previously mentioned activity spill-in and spill-
out. The methods can be based on Fourier-based methods, such as the
energy–distance principle [51] and Wiener filtering [52], but collimator-blur
compensation is mostly incorporated in the iterative reconstruction [47,
49]. The advantage with incorporation in the iterative algorithm is that
pre-calculated response functions can be placed into lookup tables and
selected for use based on the corresponding pixel location.
In many cases in nuclear medicine, a low count-rate produces images
with disturbing noise. The common way to reduce noise is to apply two-
or three-dimensional low-pass filters. However, it is important to recognize
that low-pass filtering of projections for noise reduction also will result in
a degradation of image resolution. For small objects, blurring due to low-
pass filtering reduces image contrast and the ability to detect lesions. The
blurring also affects, for example, segmentation of volumes of interest,
which may incorrectly increase the determined volume. Furthermore, the
limited spatial resolution will produce ‘spill-in’ and ‘spill-out’ of activity
from one region to another.
3.4.6. Time series of images

The most straightforward way to obtain the time series of information
needed for dosimetry is to image the patient repeatedly over time by conju-
gate views or by a tomographic method (SPECT of PET). The tomographic
imaging is potentially more accurate, but is probably more complicated and

possibly more time consuming. Registration of multiple tomographic
images, one three-dimensional image set to another, allows the use of
identical evaluation regions of interest for all time points, and/or reduces
noise in the plots of bio-distribution versus time.
An alternative to a time-series of tomographic images has been
employed by one research group in the imaging of lymphoma patients [37].
It consists of the use of a time–activity curve from conjugate views for a
larger composite tumour to provide an estimate of time–activity curves for
the smaller individual tumours of which it is composed. These time–activity
curve estimates then have their ordinate axes scaled by individual corrections
based upon the smaller-tumour activities determined by a SPECT image
taken at a single time point.
3.4.7. Methods used to register multimodality images

At the very outset, let us note that achieving a high degree of accuracy for a
spatial transformation between image sets can be quite complicated. Physical
factors, such as noise, limited spatial resolution, scatter, and septal penetra-
tion, and biological factors such as persistent activity in the blood pool and
non-specific uptake, may decrease the contrast and blur the image. There-
fore, it can be difficult to locate consistent landmarks. Furthermore, diagnos-
tic CT images are usually taken using breath-holding techniques while many
radionuclide investigations involve imaging of the thorax, abdomen or
pelvis, where organ motion exists. That is, the SPECT or PET data are
acquired during a relatively long time period so the resultant reconstructed
image set is an average of all phases of respiration. Thus, the image sets
are not consistent. This inconsistency can cause complications, for example,
if the body boundaries of the CT data and the SPECT or PET can be
registered, but the internal structures still differ significantly. Combined
CT/SPECT systems with a low-dose CT performed for a period of about
10 min are designed to avoid the breath-holding problem. The CT resolution
for these systems is about 3 mm, allowing for both attenuation correction
and registration of CT/SPECT data for accurate tumour localization. In
spite of the difficulties, many registration methods have been published
[53–68]. For a detailed survey of algorithms developed so far, the reader is
referred to the comprehensive review by Hutton et al. [69]. Below, we will
discuss newer registration methods and/or their application to dosimetry.
One type of registration method depends on mutual information, an
intensity-based similarity measure [70]. A patient result that used a
mutual-information-based registration for both attenuation correction and
tumour localization [14, 37] is shown in figure 3.5.
Sjögreen et al. have developed a registration protocol that has been
applied both to whole-body scintillation camera images, and to CT/SPECT
Figure 3.5. Results for one slice from a three-dimensional registration between
a CT image set and a SPECT image set. The pelvic area of a non-Hodgkin’s
lymphoma patient (# 33) being treated with 131 I tositumomab is shown. The
CT slice (at the top) is from a scan that was taken 13 days before the treatment
infusion. The two tumours, shown by black outlines, had their edges manually
chosen in this image. The SPECT slice (at the bottom) is from a scan that
was taken 42 hours after the treatment infusion. The tumour outlines were
placed on the bottom image by virtue of the registration. The anterior pelvic
tumour (ant pel. tum.) could be evaluated even though its percent injected
dose per gram was only 0.68 times that of the left pelvic tumour (left pel.
tum.). 131 I activity in the urine of the bladder (not labelled) is a strong feature
of the SPECT image. (See plate 2 for colour version.)
image registration [71]. Both mono- (emission-to-emission) and multi-

(emission-to-transmission) modality image registration has been addressed.
The method performs an optimization of the mutual information between
the images. For the whole-body registration method, a spatial transforma-
tion based on a second-order polynomial-warping algorithm is applied,
including local transformation of regions corresponding to the legs, pelvis,

upper abdomen and head. For the CT/SPECT registration, scatter images
have been used as a registration feature. The photopeak and Compton scat-
ter images are acquired simultaneously in dual-acquisition mode, and the
registration can thus be performed via the scatter images, which to a larger
extent reflect the body boundary. A patient positioning procedure is applied,
which mainly consists of using a flat and similarly shaped couch for both
CT and SPECT acquisitions, and applying co-calibrated directional lasers.
Spatial transformations include rigid translations and three-dimensional
rotations, and the option of applying transformations locally. The mutual
information, or the normalized mutual information, is used as the similarity
measure, calculated either globally or based on local regions. The contribu-
tion from the CT/SPECT image registration to the overall dosimetry error
has also been evaluated by introducing deliberate misalignments between
the Monte Carlo simulated SPECT images and corresponding CT images
of an anthropomorphic computer phantom [72]. The activity and absorbed
dose obtained when using re-registered CT images compared with the
corresponding results when using perfectly matched CT images have
shown that that the registration method does not introduce significant errors.
Recently, a new procedure to allow three-dimensional registration of CT
and SPECT has been proposed [73]. The rationale for the procedure is that it
should be easier to find the correct superimposition for two transmission
image sets, because they are quite similar, than for a transmission data set
(CT) and an emission date set (SPECT) which can have many different
features. In this particular procedure, both of the transmission sets involve
planar images. Implementation is perhaps simplest for a dual-head SPECT
camera system. The procedure involves two extra radionuclide transmission
images, obtained before or after the SPECT acquisition. One is from anterior
to posterior and the other is from left lateral to right lateral. These are
compared with re-projections of the volumetric CT data along the same
axes. When the CT volume image set has the correct shift, scaling and rota-
tion, its projections should match those of the transmission images. Then the
same transformation of the CT should bring it into registration with the
SPECT data set acquired with the dual-head camera. The authors state
that acceptable accuracy is attained in initial phantom tests [73].
3.4.8. Quantitative imaging with PET

PET is arguably the most advanced of the non-invasive nuclear medical
imaging techniques. It provides both qualitative and quantitative informa-
tion about the volume distribution of biologically significant radiotracers.
The sensitivity of in vivo tracer studies is highly developed with PET,
which is based on electronic collimation and thereby offers a wide acceptance
angle for detecting emitted annihilation photons. Consequently, the sensitivity
of PET per disintegration with comparable axial fields of view is two orders of
magnitude greater than that of SPECT cameras. Another boost for sensitivity
comes from the fact that PET usually detects molecules labelled with short-
lived radioisotopes for which the decay rate is very high. Within the spectrum
of medical imaging modalities, sensitivity ranges from the detection of milli-
molar concentrations in functional MRI to pico-molar concentrations in
PET which is a difference involving a factor of 109 [74].
It is also generally held that PET offers better spatial resolution than
SPECT. Furthermore, PET has provided the possibility of quantitative
measurements of tracer concentration in vivo for a longer time than
SPECT has. However, in practice, there are several issues that must be
addressed in order to fully realize the potential of PET. That is, the
measured line integrals must be corrected for a number of background
and physical effects before reconstruction. The needed processes include
subtraction of random coincidences, detector normalization, deadtime,
attenuation and scatter corrections. Detector normalization is similar to
correction for camera–collimator-system non-uniformities based on field-
flood measurements in SPECT. Deadtime, attenuation and scatter correc-
tion are all also needed in SPECT. Since attenuation and scatter correction
are especially important in clinical PET, they are briefly discussed in section
3.4.9 below.
Advantages. PET is based on electronic collimation and thereby offers

a wide acceptance angle for detecting emitted annihilation photons. Con-
sequently, the sensitivity of PET per disintegration with comparable axial
fields of view is two orders of magnitude greater than that of SPECT
cameras. A further sensitivity advantage for detecting low molecular concen-
tration arises in PET when detecting molecules labelled with short-lived
radioisotopes. With respect to imaging of molecular pathways and interac-
tions, the specificity of PET arises due to the use of radioisotopes such as
18
F and, to a lesser extent, 11 C to label biochemical/physiological and
pharmaceutical molecules within a radioactive form of one of its natural
constituent elements. SPECT’s ability to image some chemical processes is
hindered by the fact that the isotopes are relatively large atoms and cannot
be used to label some compounds due to steric reasons. Moreover, quantifi-
cation techniques are well established with PET whereas they have only been
more recently developed with SPECT.
Disadvantages. The high cost of PET and the need for a nearby or an
on-site cyclotron and for radiochemical/radiopharmaceutical support has
limited PET to only a few clinical sites, to date. The major drawback is
that a short half-life puts a practical upper limit on the activity of the manu-
factured isotope. Moreover, for treatment planning, a radioisotope from the
same element as the treatment element is required. 124 I, 86 Y and 64 Cu are
available and have undergone some usage, however.
3.4.9. Effect of attenuation and scatter correction on PET imaging

Attenuation correction in PET is now widely accepted by the nuclear medi-
cine community as a vital component for the production of artefact-free,
quantitative data. In PET, correction for attenuation depends on the total
distance travelled by both annihilation photons and is independent of the
emission point along the ray defined by these photons. The most accurate
attenuation correction techniques are based on measured transmission
data acquired before (pre-injection), during (simultaneous) or after (post-
injection) the emission scan. Alternative methods to compensate for
photon attenuation in reconstructed images use assumed distribution and
boundary of attenuation coefficients, segmented transmission images, or
consistency condition criteria [4]. Transmission-based attenuation correction
has been traditionally performed in the case of PET whereas only more
recently has it been applied to SPECT imaging. There are many possible
explanations for that difference, one of them being that PET started
mainly as a research tool where there was greater emphasis on accurate
quantitative measurements.
Techniques based on transmission image segmentation and tissue classi-
fication tools have also been proposed to minimize the acquisition time and
increase the accuracy of attenuation correction, while still preserving or even
reducing the noise level [75]. Representative coronal slices of a clinical study
at the level of the thorax are shown in figure 3.6. The figure illustrates the
improvement in image quality of the emission-data reconstructions when
using fuzzy clustering-based segmented attenuation correction as compared
with measured attenuation correction. The images are less noisy and show
more uniform uptake of the tracer.
3.4.10. Reconstruction algorithms for fully three-dimensional PET

The development of fully three-dimensional reconstruction algorithms has
been necessary in order to take advantage of the acquisition of PET data
without septa. Various methods have been proposed to solve the problem
of recovering the image from the resulting four-dimensional data sets.
There are two major classes of image reconstruction algorithms used in
PET: direct analytical methods and iterative methods. At present, the most
widely used methods of image reconstruction are still direct analytical
methods because they are relatively quick. However, the images tend to be
streaky and display interference between regions of low and high tracer
concentration. Images produced by iterative methods are computationally
much more intense and, at present, the computational time of reconstruction
is prohibitive. However, with the development of parallel architectures and
because of the potential for using these architectures in conjunction with
iterative techniques, the future for iterative reconstruction seems bright.
Auxiliary contribution from PET 77
Figure 3.6. Illustration of the improvement in image quality when using

segmented attenuation correction. A. Reconstructions using measured attenua-
tion correction. B. Reconstructions using fuzzy clustering-based segmented
attenuation correction. Reprinted with permission from the Institute of Physics
[75].
During the past decade, much research and development has concen-
trated on the development of improved fully three-dimensional reconstruction
algorithms and scatter compensation techniques required for quantitative
PET. Increasingly sophisticated scatter correction procedures are under
investigation: particularly those based on accurate scatter models and
iterative reconstruction-based scatter compensation approaches [76]. In
summary, PET quantification remains an open issue that requires further
R&D efforts [77].
3.5. AUXILIARY CONTRIBUTION FROM PET
FDG-PET is widely used in the diagnosis, staging and assessment of tumour

response to therapy, because metabolic changes generally precede the more
conventionally measured parameter of change (tumour size). Although
metabolic imaging is an obvious choice, the way to perform imaging is still
an open issue. Combined PET/CT systems are pushing ahead for acquiring
co-registered anatomical and functional images in a single scanning session

allowing more accurate and effective attenuation correction [78]. The tracers
or combinations of tracers to be used [79], when the imaging should be done
after therapy, what are the optimal acquisition and processing protocols, are
all unanswered questions. Moreover, each tumour–therapy combination
may need to be independently optimized and validated. Results, to date,
for two diseases are discussed immediately below.
The use of PET with FDG, FDOPA and 15 O-water to assess the
metabolic status of melanoma metastases after treatment with dacarbazine
and -interferon has been recently reported [79]. This approach, without
pre-treatment imaging, cannot assess subtle increases or decreases in
tumour metabolism. However, the approach should be able to assess whether
the response has been complete or whether residual viable tumour is present.
This method is appropriate if a reasonable probability exists that a single
course of therapy will result in a complete response.
In imaging of BexxarTM patients with FDG-PET, response has been
shown to occur as early as 6–7 days. That is: (1) FDG-PET imaging found
tumour metabolism to be reduced at 33–70 days after radioimmunotherapy
(RIT), compared with a baseline scan. The standardized uptake value lean
(SUVlean ) for tumours in eight patients was reduced, and reduced
proportionately to the final response (complete response patients more
than partial response patients, and partial response patients more than
non-response patients) [80]. (2) In five of six other patients, the same
SUVlean was already decreased at 6–7 days after the therapy administration.
3.6. TREATMENT PLANNING
When using imaging for dosimetry, one often assumes that the results during
therapy will be the same as those during evaluation, except for scaling by the
ratio of the administered activities. Since the administration for the first
imaging could possibly affect the pharmacokinetics, this assumption may
not hold. If it does not, the choice of administered activity might still be
made in the same way but measurements would be needed during therapy
to quantify the discrepancy. It is well known that in ablation of thyroid
remnants, measurements seem to imply a reduction in the therapy activity
compared with what one calculates by scaling. The explanation for the
observed results is still under discussion, however.
In BexxarTM RIT of lymphoma, the ratio of the therapy percent–
injected dose divided by the evaluation percent–injected dose has been
checked for variations from 1. These checks were carried out at 2 to 3 days
after the administration [81]. There is an extra biological complication
here in that the RIT consists of the administration of an anti-CD20
murine monoclonal IgG2a antibody (tositumomab) pre-dose followed by
References 79
its 131 I-radiolabelled conjugate (iodine 131 I tositumomab). The evaluation was
carried out with a reduced activity of iodine 131 I tositumomab but with the
same pre-dose. For 31 chemotherapy-relapsed patients, the hypothesis that
the ratio equals 1 was rejected using conjugate-view imaging [82]. However,
the average of the ratio was only slightly less than one (0.93). For seven
previously untreated patients, results from imaging using either conjugate
views or SPECT agreed with the hypothesis of the ratio being equal to 1
[81]. So, for both types of patient, there will not be much error on average
with assuming a ratio of 1.
ACKNOWLEDGMENTS
One author (KFK) acknowledges the support of Grant R01 CA87955

awarded by the National Cancer Institute, United States Department of
Health and Human Services. Another author (HZ) acknowledges the
support of grant SNSF 3152-062008 from the Swiss National Science Foun-
dation. Another author (MLJ) acknowledges the support from the Swedish
Cancer Foundation. The contents and views in this chapter are solely the
responsibility of the authors and do not necessarily represent the official
views of the institutions providing the support.
REFERENCES
[1] Leichner P K et al. 1993 An overview of imaging techniques and physical aspects of
treatment planning in radioimmunotherapy Med. Phys. 20 569–77
[2] Siegel J A et al. 1999 MIRD Pamphlet No 16: Techniques for quantitative
radiopharmaceutical biodistribution data acquisition and analysis for use in human
radiation dose estimates J. Nucl. Med. 40 37S–61S
[3] Snyder W S, Ford M R, Warner G G and Watson S B 1975 ‘S’, Absorbed Dose per
Unit Cumulated Activity for Selected Radionuclides and Organs. MIRD Pamphlet
No 11 (Society of Nuclear Medicine)
[4] Zaidi H and Hasegawa B H 2002 Determination of the attenuation map in emission
tomography J. Nucl. Med. 43 in press
[5] Liu A, Williams L E and Raubitschek A A 1996 A CT assisted method for absolute
quantitation of internal radioactivity Med. Phys. 23 1919–28
[6] Zaidi H 1996 Comparative methods for quantifying thyroid volume using planar
imaging and SPECT J. Nucl. Med. 37 1421–6
[7] Alaamer A S, Fleming J S and Perring S 1993 Evaluation of a technique for the
quantification of radioactivity and volume of an object using SPECT Nucl. Med.
Commun. 14 1061–70
[8] Fleming J S and Alaamer A S 1998 A rule based method for context sensitive
threshold segmentation in SPECT using simulation Phys. Med. Biol. 43 2309–23
[9] Zaidi H 1996 Organ volume estimation using SPECT IEEE Trans. Nucl. Sci. 43
2174–82
[10] Erdi Y E et al. 1997 Segmentation of lung lesion volume by adaptive positron
emission tomography image thresholding Cancer 80 2505–9
[11] Boudraa A E et al. 1996 Delineation and quantitation of brain lesions by fuzzy
clustering in positron emission tomography Comput. Med. Imaging Graph. 20
3141
[12] Koral K F et al. 1989 Calculating radiation absorbed dose for pheochromocytoma
tumors in 131I MIBG therapy Int. J. Radiat. Oncol. Biol. Phys. 17 21118
[13] Tang H R et al. 2001 Neuroblastoma imaging using a combined CT scanner–
scintillation camera and 131I-MIBG J. Nucl. Med. 42 23747
[14] Koral K F et al. 2000 Initial results for hybrid SPECTconjugate-view tumor
dosimetry in 131I-anti-B1 antibody therapy of previously untreated patients with
lymphoma J. Nucl. Med. 41 157986
[15] Sisson J C et al. 1994 Predictors of toxicity in treating patients with neuroblastoma by
radiolabeled metaiodobenzylguanidine Eur. J. Nucl. Med. 21 4652
[16] Wahl R L et al. 1998 Iodine-131 anti-B1 antibody for B-cell lymphoma an update on
the Michigan Phase I experience J. Nucl. Med. 39 21S27S
[17] Thomas S R, Maxon H R and Kereiakes J G 1976 In vivo quantitation of lesion
radioactivity using external counting methods Med. Phys. 3 2535
[18] Hammond N D, Moldofsky P J, Beardsley M R and Mulhern C B Jr 1984 External
imaging techniques for quantitation of distribution of I-131 Fðab0 Þ2 fragments of
monoclonal antibody in humans Med. Phys. 11 77883
[19] Eary J F, Appelbaum F L, Durack L and Brown P 1989 Preliminary validation of
the opposing view method for quantitative gamma camera imaging Med. Phys. 16
3827
[20] Ogawa K, Harata H, Ichihara T, Kubo A and Hashimoto S 1991 A practical method
for position dependent Compton-scatter correction in single photon emission CT
IEEE Trans. Med. Imaging 10 40812
[21] Shulkin B L et al. 1988 Conjugate view gamma camera method for estimating tumor
uptake of iodine-131 metaiodobenzylguanidine J. Nucl. Med. 29 5428
[22] Matthay K K et al. 2001 Correlation of tumor and whole-body dosimetry with tumor
response and toxicity in refractory neuroblastoma treated with 131I-MIBG J. Nucl.
Med. 42 171321
[23] Doherty P, Schwinger R, King M and Gionet M 1985 Distribution and dosimetry of
indium-111 labeled Fðab0 Þ2 fragments in humans. Fourth International Radiopharma-
ceutical Dosimetry Symposium, Oak Ridge, TN, CONF-851113 (DE86010102)
[24] Zaidi H 1996 Quantitative SPECT: Recent developments in detector response
attenuation and scatter correction techniques Physica Medica 12 10117
[25] Rosenthal M S et al. 1995 Quantitative SPECT imaging a review and
recommendations by the Focus Committee of the Society of Nuclear Medicine
Computer and Instrumentation Council J. Nucl. Med. 36 1489513
[26] Tsui B M, Zhao X, Frey E C and McCartney W H 1994 Quantitative single-photon
emission computed tomography basics and clinical considerations Semin. Nucl. Med.
24 3865
[27] Dewaraja Y K, Ljungberg M and Koral K F 2000 Accuracy of 131I tumor
quantification in radioimmunotherapy using SPECT imaging with an ultra-high-
energy collimator: Monte Carlo study J. Nucl. Med. 41 176067
[28] Sayeram S et al. 2001 Compensation for collimator detector response in tumor
SPECT imaging using In-111 with dual energy photons J. Nucl. Med. 42 197P
References 81
[29] Gonzales-Trotter D E, Bowsher J E and Jaszczak R J 2001 Absolute quantitation of

a spherical I-131 activity distribution using a high-resolution rotating collimator:
A phantom study IEEE Trans. Nucl. Sci. 48 6573
[30] Links J M, Zubieta J K, Meltzer C C, Stumpf M J and Frost J J 1996 Influence of
spatially heterogeneous background activity on ‘hot object’ quantitation in brain
emission computed tomography J. Comput. Assist. Tomogr. 20 6807
[31] Gonzalez D E, Jaszczak R J, Bowsher J E, Akabani G and Greer K L 2001 High-
resolution absolute SPECT quantitation for I-131 distributions used in the
treatment of lymphoma: A phantom study IEEE Trans. Nucl. Sci. 48 70714
[32] Hudson H M and Larkin R S 1994 Accelerated image-reconstruction using ordered
subsets of projection data IEEE Trans. Med. Imaging 13 6019
[33] Alaamer A S, Fleming J S and Perring S 1994 Evaluation of the factors affecting the
accuracy and precision of a technique for quantification of volume and activity in
SPECT Nucl. Med. Commun. 15 75871
[34] Tang T R 1998 A Combined X-ray CT–Scintillation Camera Imaging System for
Measuring Radionuclide Uptake in Tumors. Thesis. University of California
[35] Fessler J A and Hero A O 1995 Penalized maximum-likelihood image-reconstruction
using space-alternating generalized EM algorithms IEEE Trans. Imag. Proc. 4
141729
[36] Dewaraja Y K, Ljungberg M and Koral K F 2001 Evaluation of object shape effects
in I-131 SPECT tumor activity quantification Eur. J. Nucl. Med. 28 9006
[37] Koral K F et al. 2000 Tumor-absorbed-dose estimates versus response in
tositumomab therapy of previously untreated patients with follicular non-
Hodgkin’s lymphoma: preliminary report Cancer Biother. Radiopharm. 15 34755
[38] Lange K and Carson R 1984 E-M reconstruction algorithms for emission and
transmission tomography J. Comput. Assist. Tomogr. 8 30616
[39] Nickoloff E L, Perman W H, Esser P D, Bashist B and Alderson P O 1984 Left-
ventricular volume—Physical basis for attenuation corrections in radionuclide
determinations Radiology 152 51115
[40] Koral K F et al. 1994 CT–SPECT fusion plus conjugate views for determining
dosimetry in iodine-131-monoclonal antibody therapy of lymphoma patients J.
Nucl. Med. 35 171420
[41] Tang H R et al. 1999 Implementation of a combined X-ray CT–scintillation camera
imaging system for localizing and measuring radionuclide uptake: Experiments in
phantoms and patients IEEE Trans. Nucl. Sci. 46 5517
[42] Sorenson J A 1974 Quantitative measurement of radioactivity in vivo by whole-body
counting. In Instrumentation in Nuclear Medicine G a S ed J Hine (New York:
Academic Press) vol 2 pp 31148
[43] Chang L T 1978 A method for attenuation correction in radionuclide computed
tomography IEEE Trans. Nucl. Sci. 25 63843
[44] Jaszczak R J, Greer K L, Floyd C E, Harris C C and Coleman R E 1984 Improved
SPECT quantification using compensation for scattered photons J. Nucl. Med. 25
893900
[45] Beekman F J, Kamphuis C and Frey E C 1997 Scatter compensation methods in 3D
iterative SPECT reconstruction: a simulation study Phys. Med. Biol. 42 161932
[46] Frey E C and Tsui B M W 1993 A practical method for incorporating scatter in a
projector–backprojector for accurate scatter compensation in SPECT IEEE Trans.
Nucl. Sci. 40 110716
[47] Beekman F J, de Jong H W and Slijpen E T 1999 Efficient SPECT scatter calculation
in non-uniform media using correlated Monte Carlo simulation Phys. Med. Biol. 44
N183N192
[48] de Jong H W and Beekman F J 2001 Rapid SPECT simulation of downscatter in non-
uniform media Phys. Med. Biol. 46 62135
[49] Beekman F J, Slijpen E T, de Jong H W and Viergever M A 1999 Estimation of the
depth-dependent component of the point spread function of SPECT Med. Phys. 26
231122
[50] King M A, Long D T and Brill B A 1991 SPECT volume quantitation: Influence of
spatial resolution source size and shape and voxel size Med. Phys. 18 101624
[51] Hawkins W G, Yang N-C and Leichner P K 1991 Validation of the circular harmonic
transform (CHT) algorithm for quantitative SPECT J. Nucl. Med. 32 14150
[52] King M A, Schwinger R B, Doherty P W and Penney B C 1984 Two-dimensional
filtering of SPECT images using the Metz and Wiener filters J. Nucl. Med. 25 123440
[53] Andersson J L R, Sundin A and Valind S 1995 A method for coregistration of PET
and MR brain images J. Nucl. Med. 36 130715
[54] Borgefors G 1988 Hierachical chamfer matching a parametrical edge matching
algorithm IEEE Trans. Pattern Anal. Mach. Intell. 10 84965
[55] Eberl S et al. 1996 Automated interstudy image registration technique for SPECT and
PET J. Nucl. Med. 37 13745
[56] Flux G 1995 Multimodality Image Registration and its Application to the Dosimetry of
Intralesional Radionuclide Therapy. Thesis. Institute of Cancer Research & Royal
Marsden Hospital, University of London
[57] Hill D L G and Hawkes D J 2000 Across-modality registration using intensity-based
cost functions. In Handbook of Medical Imaging—Processing and Analysis ed I N
Bankman (San Diego: Academic)
[58] Maes F, Collignon A, Vandermeulen D, Marchal G and Suetens P 1997
Multimodality image registration by maximization of mutual information IEEE
Trans. Med. Imaging 16 18798
[59] Maguire G Q, Noz M E, Lee E M and Schimpf J H 1986 Correlation methods for
tomographic images using two and three dimensional techniques. In Information
Processing in Medical Imaging ed S L Bacharach (Dordrecht: Martinus Nijhoff )
pp 26679
[60] Maintz J B A, van den Elsen P A and Viergever M A 2001 3D multimodality medical
image registration using morphological tools Imag. Vision Comput. 19 5362
[61] Pelizzari C A, Chen G T Y, Spelbring D R, Wiechselbaum R R and Chen C 1989
Accurate three-dimensional registration of CT, PET, and/or MR images of the
brain J. Comput. Assist. Tomogr. 13 206
[62] Rangarajan A, Chui H and Duncan J S 1999 Rigid point feature registration using
mutual information Med. Imag. Anal. 3 42540
[63] Studholme C, Hill D L G and Hawkes D J 1996 Automated 3D registration of MR
and CT images of the head Med. Imag. Anal. 1 16375
[64] Studholme C, Hill D L G and Hawkes D J 1999 An overlap invariant entropy
measure of 3D medical image alignment Pattern Recogn. 32 7186
[65] van den Elsen P A, Pol E J D and Viergever M A 1993 Medical image matching—A
review with classification IEEE Eng. Med. Biol. Mag. 12 2639
[66] Wells W M, Viola P, Atsumi H, Nakajima S and Kikinis R 1996 Multi-modal volume
registration by maximization of mutual information Med. Imag. Anal. 1 3551
References 83
[67] Woods R P, Mazziotta J C and Cherry S R 1992 Rapid automated algorithm for
aligning and reslicing PET images J. Comput. Assist. Tomogr. 16 620–33
[68] Woods R P 2000 Within-modality registration using intensity-based cost functions. In
Handbook of Medical Imaging—Processing and Analysis ed I N Bankman (San Diego:
Academic)
[69] Hutton B F, Braun M, Thurfjell L and Lau D Y H 2002 Image registration an
essential tool for nuclear medicine Eur. J. Nucl. Med. 29 559–77
[70] Meyer C R et al. 1997 Demonstration of accuracy and clinical versatility of mutual
information for automatic multimodality image fusion using affine and thin-plate
spline warped geometric deformations Med. Imag. Anal. 1 195–206
[71] Sjögreen K, Ljungberg M, Wingšrdh K, Erlandsson K and Strand S-E 2001
Registration of emission and transmission whole-body scintillation camera images
J. Nucl. Med. 42 1563–1570
[72] Sjögreen K 2001 Image Processing for Quantitative Scintillation-Camera Imaging:
Application to Radionuclide Therapy. Thesis. Lund University
[73] DeVries D J and Moore S C 2001 Registration of CT and SPECT volumes using CT
reprojections and planar transmission images Proc IEEE Nuclear Science Symposium
and Medical Imaging Conference, San Diego, CA, in press
[74] Jones T 1996 The role of positron emission tomography within the spectrum of
medical imaging Eur. J. Nucl. Med. 23 207–11
[75] Zaidi H, Diaz-Gomez M, Boudraa A E and Slosman D O 2002 Fuzzy clustering-
based segmented attenuation correction in whole-body PET imaging Phys. Med.
Biol. 47 1143–60
[76] Zaidi H 2000 Comparative evaluation of scatter correction techniques in 3D positron
emission tomography Eur. J. Nucl. Med. 27 1813–26
[77] Zaidi H 2001 Scatter modelling and correction strategies in fully 3-D PET Nucl. Med.
Commun. 22 1181–4
[78] Beyer T et al. 2000 A combined PET/CT scanner for clinical oncology J. Nucl. Med.
41 1369–79
[79] Dimitrakopoulou-Strauss A, Strauss L G and Burger C 2001 Quantitative PET studies
in pretreated melanoma patients a comparison of 6-[18F]fluoro-L-dopa with 18F-FDG
and 15O-water using compartment and noncompartment analysis J. Nucl. Med. 42 248–
56
[80] Torizuka T et al. 2000 Metabolic response of non-Hodgkin’s lymphoma to 131I-anti-
B1 radioimmunotherapy evaluation with FDG PET J. Nucl. Med. 41 999–1005
[81] Koral K F et al. 1999 I-131 anti-B1 therapy/tracer uptake ratio using a new procedure
for fusion of tracer images to computed tomography images Clin. Cancer Res. 5
3004s–3009s
[82] Zasadny K R et al. 1997 Do tracer dosimetry studies predict normal organ and
tumor uptake of I-131-anti-B1 (anti CD-20) antibody in patients receiving
radioimmunotherapy for non-Hodgkins lymphoma? Radiology 205 509–509
Chapter 4
Computational methods in internal

radiation dosimetry
Pat B Zanzonico and John W Poston, Sr
4.1. INTRODUCTION
Radiation dosimetry deals with the determination of the amount and the
spatial and temporal distribution of energy deposited in matter by ionizing
radiation. Internal radionuclide dosimetry specifically deals with radiation
energy deposition in tissue for radionuclides within the body. Internal
dosimetry has been applied to the determination of tissue doses and related
quantities for occupational exposures in radiation protection, environmental
exposures in radiation epidemiology, and diagnostic and therapeutic
exposures in nuclear medicine. With the increasing therapeutic application
of internal radionuclides in medicine, larger administered activities are
used and higher normal-tissue doses, with associated radiation injury, may
result. Treatment planning for radionuclide therapy thus requires, in
addition to reliable dose–response relationships for target tissues and
dose–toxicity relationships for normal tissues, increasingly more accurate
and precise dose estimates for target tissue and at-risk normal tissues.
Radiation dosimetry in nuclear medicine continues to evolve—from
estimation of population- and organ-averaged doses for risk estimation for
stochastic effects to calculation of individualized patient doses that reflect
the heterogeneity of dose distributions within tumours and organs and the
probability of deterministic effects [1–4]. Reviewed in this chapter, the
standard computational methods developed for internal radiation dosimetry,
such as the ‘MIRD schema’, have generally been based on ‘average’ kinetic
and anatomic (i.e., anthropomorphic) models. These methods yield popula-
tion- and organ-averaged doses for estimation of stochastic risk associated
with diagnostic administered activities. In other chapters of this book,
strategies for individualized patient dosimetry, more appropriate for radio-
nuclide therapy, are developed.
84
Radiation quantities and units 85
4.2. RADIATION QUANTITIES AND UNITS
4.2.1. Stochastic versus deterministic quantities

Dosimetric quantities may be characterized as ‘stochastic’ or ‘deterministic’
(i.e., ‘non-stochastic’) [4–8]. A quantity subject to statistical fluctuations is
termed ‘stochastic’, while the mean of a large number of determinations of
the quantity is termed ‘deterministic’. Each stochastic quantity, then, has a
corresponding deterministic quantity. The field of ‘microdosimetry’ deals
with the number, size and spatial and temporal distributions of individual
energy-deposition events, particularly in microscopic structures of the
order of molecules, macromolecules and supramolecular complexes in size,
characterizing such events in terms of inherently stochastic quantities [5].
4.2.2. Definitions of dosimetric quantities

A physically and biologically meaningful specification of radiation ‘dose’ is a
problematic concept and has been specified in a variety of ways [9, 10]. The
following definitions are extracted from [6] and [8].
‘Exposure’, X, is defined as
dQ
X ð4:1Þ
dm
where dQ is the absolute value of the total charge of ions of one sign
produced in air when all the electrons (negatrons and positrons) liberated
by photons (X- and -rays) in air are completely stopped in air, and dm is
the mass of air in which the electrons (negatrons and positrons) were liber-
ated by photons (X- and -rays).
Exposure is essentially the ionization equivalent of air kerma (defined
below) and, as in the case of kerma, it may often be convenient to refer to
a value of exposure at a point inside a material different from air. In such
a case, the exposure specified will be that which would be determined for a
small quantity of air placed at the point of interest. Exposure is a deter-
ministic quantity and thus corresponds to a sufficiently large irradiated
volume and/or a sufficiently large amount of radiation to yield insignificant
statistical fluctuations in its measured value.
The ‘absorbed dose’, D, is probably the most widely used quantity to
characterize ‘dose’ and is defined as
dE
D ð4:2Þ
dm
where dE is the mean energy imparted by ionizing radiation to matter, and
dm is the mass of matter to which the energy is imparted.
86 Computational methods in internal radiation dosimetry
Absorbed dose is a deterministic quantity and is the deterministic

correlate of the stochastic quantity, ‘specific energy’, z:
"
z ð4:3Þ
m
where " is the energy imparted to matter by a single energy-deposition event
and dm is the mass of the matter.
It is conceptually straightforward to relate absorbed dose to exposure
[11]. The average energy required to produce an ion pair in air is nearly
constant for all electron energies, corresponding to a value of the mean
energy per ion pair (W) of 33.7 eV/ion pair [12]. Therefore, since the conven-
tional unit of exposure, the roentgen (R), corresponds to 1:61 1012 ion pair/
gm of air (i.e., 1 R ¼ 1:61 1012 ion pair/gm), 1 eV ¼ 1:602 1012 erg, and
1 rad ¼ 100 erg/gm, the absorbed dose to air for an exposure of 1 R to air is
ion pair/g air eV erg 1 rad
1:610 1012 33:7 1:602 1012
R air ion pair eV 100 erg=g
rad
¼ 0:869 ð4:4Þ
R air
Therefore,
Dair ðradÞ ¼ 0:869Xair ðRÞ ð4:5Þ
where Dair is the absorbed dose (in rad) to air, and Xair is the exposure (in R)
to air.
Under conditions of charged particle equilibrium for a point in a non-air
medium (such as tissue), equation (5) may be modified to yield the absorbed
dose in a non-air medium for a given exposure in air [11]:
ðen =Þmed
Dmed ðradÞ ¼ 0:869 X ðRÞ ð4:6aÞ
ðen =Þair air
¼ 0:869ðen =Þ med
air Xair ðRÞ ð4:6bÞ
¼ fmed Xair ðRÞ ð4:6cÞ
where Dmed is the absorbed dose (in rad) to a point in a non-air medium,
ðen =Þmed is the mass energy absorption coefficient (e.g., in cm2 /gm) in a
non-air medium, ðen =Þair is the mass-energy absorption coefficient (e.g.,
in cm2 /gm) in air, ðen =Þ med
air ðen =Þmed =ðen =Þair is the mass-energy
absorption coefficient ratio used to convert air kerma to dose in the
medium surrounding the source, Xair is the exposure (in R) to air, and
fmed ¼ 0:869ðen =Þ med
air is the exposure-to-absorbed-dose (i.e., R-to-rad)
conversion factor (in rad/R) for the non-air medium.
Fortuitously, the exposure-to-absorbed-dose (i.e., R-to-rad) conversion
factor (fmed ) is very nearly unity (0.87–0.968) for a wide range of photon
energies (0.01–10 MeV) for air, water and soft tissues. Generally, therefore,
Radiation quantities and units 87
the absorbed dose (in rad) may be considered to be numerically equal to the
exposure (in R).
‘Kerma’, K, is defined as
dEtr
K ð4:7Þ
dm
where dEtr is the sum of the initial kinetic energies of all the charged ionizing
particles liberated by uncharged ionizing particles (including photons) in
matter, and dm is the mass of matter in which the charged ionizing particles
were liberated.
In the case in which the matter is air, kerma is often referred to as ‘air
kerma’ or ‘free air kerma’.
The quality as well as the quantity of radiation are important deter-
minants of the frequency and/or severity of radiogenic effects. The ‘quality’
of a radiation is related to the characteristics of the microscopic spatial
distribution of energy-deposition events, determined by the mass, charge
and energy (i.e., velocity) of the charged particles composing the radiation
or, in the case of X-rays, -rays and neutrons, the charged particles produced
by the radiation. Sparsely ionizing radiations such as X- and -rays and
intermediate-to-high-energy electrons and particles are characterized as
‘low-quality’ radiations, while densely ionizing radiations such as low-
energy electrons (e.g., Auger electrons), protons, neutrons and particles
are typically characterized as ‘high-quality’ radiations. Importantly, for the
same absorbed dose, the frequency and/or severity of radiogenic biological
effects are generally less for sparsely ionizing, low-quality radiations than
for densely ionizing, high-quality radiations.
The quality of radiation is quantitatively characterized by the ‘linear
energy transfer’, L or LET, or the ‘restricted linear energy transfer’, L or
LET of a material for charged particles:

dE
L ð4:8Þ
dl
where dE is the energy lost by a charged particle in traversing a distance in
matter due to those collisions in which the energy loss is less than the
energy cut-off , and dl is the distance traversed in matter.
The restricted linear energy transfer (L ) requires the specification of a
cut-off energy (), necessitated by primary energy-deposition events result-
ing in relatively high-energy, relatively long-range secondary electrons (i.e.,
-rays) whose energy deposition events may be considered as separate from
those along the track of the primary radiation. In radiation protection,
this feature is generally disregarded by specification of the ‘unrestricted
linear energy transfer’, L1 or LET1 (also known simply as the ‘linear
energy transfer’, L or LET), where the energy cut-off () is set equal to
infinity (1).
LET is the deterministic correlate of the stochastic quantity, ‘lineal

energy’, y:
"
y ð4:9Þ
l
where " is the energy imparted to matter by a single energy-deposition event,
and l is the mean chord length of the volume of the matter.
For the same absorbed dose, the frequency and/or severity of radiogenic
biological effects are generally less for low-LET radiations than for high-LET
radiations. The influence of radiation quality on the frequency and/or severity
of biological effects is quantified by the ‘relative biological effectiveness’, or
RBE(A) of radiation A:
Dreference
RBEðAÞ ð4:10Þ
DA
where Dreference is the absorbed dose of reference radiation (typically a widely
available sparsely ionizing radiation such as 60 Co -rays) required to produce
a specific, quantitatively expressed frequency and/or severity of a specific
biological effect, and DA is the absorbed dose of radiation A required to
produce the same specific, quantitatively expressed frequency and/or severity
of the same specific biological effect, with all pertinent parameters, except the
radiation itself, maintained as identical as possible. Because the relative RBE
represents a ratio of absorbed doses, it is a dimensionless quantity.
Because the actual value of RBE depends on many factors, such as
the nature of the biological effect, the absorbed dose, the absorbed dose
rate, etc., a simplified version of the relative biological effectiveness, the
‘quality factor’, Q or QF, was devised for purposes of radiation protection.
However, because of variations in energy, LET and therefore RBE and Q
along the radiation track, the so-called ‘dose equivalent’, H, at a point
must be related to the ‘mean quality factor’, Q , and the mean absorbed
dose at that point:
D
HQ ð4:11Þ
is the mean quality factor,
where Q
ð
Q ¼ ð1=DÞ QðLÞDðLÞ dL; ð4:12Þ
D is the linear-energy-transfer-averaged absorbed dose at the point of

calculation of the effective dose, QðLÞ is the quality factor for radiation of
linear energy transfer L at the point of calculation of the effective dose,
and DðLÞ is the absorbed dose for radiation of linear energy transfer L at
the point of calculation of the effective dose.
However, because of the difficulty of determining the energy, LET, and
Q distributions at a point, the dose-equivalent concept is itself of limited
The MIRD schema 89
practical utility. A different quantity, the ‘equivalent dose’, HT;R , in tissue or

organ T due to radiation R, therefore has been devised:
HT;R wR DT;R ð4:13Þ
where wR is the radiation (R) weighting factor, a dimensionless quantity

designed to account for differences in relative biological effectiveness
among different radiations given by equation (4.12), and DT;R is the mean
absorbed dose to tissue or organ T due to radiation R.
When a tissue or organ is irradiated by a combination of different
radiations (i.e., radiations of different qualities), the ‘equivalent dose’, HT ,
in tissue or organ T is the sum of all the mean tissue or organ doses, DT;R ,
due to each component radiation R multiplied by its respective weighting
factor, wR :
X
HT wR DT;R ð4:14Þ
R
Like the RBE, the quality factors, mean quality factors and radiation weight-
ing factors are dimensionless quantities.
For a tissue or organ, the equivalent dose is conceptually different from
the dose equivalent. The dose equivalent is based on the absorbed dose at a
point in tissue weighted by the LET-dependent distribution of quality factors
(Q(L)) at that point. The equivalent dose, in contrast, is based on the average
absorbed doses (DT;R ) in the tissue or organ weighted by the radiation
weighting factor (wR ) for the radiation(s) actually impinging on the tissue
or organ or, in the case of internal radionuclides, as it is actually emitted
by the source.
4.3. THE MIRD SCHEMA
One of the most widely used and authoritative approaches for internal dose
calculations in medicine was developed by the Medical Internal Radiation
Dose (MIRD) Committee of the Society of Nuclear Medicine (SNM) and
generally is referred to as the ‘MIRD schema’ or ‘MIRD formalism’ [4,
13–16]. The International Commission on Radiological Protection (ICRP)
has developed a similar methodology and similar reference data [17]. The
MIRD schema, including notation, terminology, mathematical method-
ology, and reference data, has been disseminated in the form of the collected
MIRD Pamphlets and associated publications [2, 13, 18, 19]. With the
publication of ORNL/TM-8381/V1-7 [20], age- and gender-specific body
habiti other than the original 70 kg adult anthropomorphic model (‘Refer-
ence Man’ or ‘Standard Man’) [14, 21] are now incorporated into the
MIRD schema. In addition, several computerized versions of the MIRD
schema, including MIRDOSE [22], DOSCAL (which incorporates mean

tumour doses) [23] and MABDOS (with curve fitting and modelling features)
[24–26], have been developed.
Calculation of the absorbed dose to an organ of the body from an
internally distributed radionuclide is based conceptually on answering two
very simple questions. First, what is the total number of nuclear transitions
(decays) that occur in the organ over the time-period of interest? Second,
how much energy is deposited in the organ per nuclear transition per unit
mass of the organ? However, in reality, the calculation of internal absorbed
dose requires knowledge of a number of factors and rests on a number of
assumptions, including the anthropomorphic models of the ‘standard’
human body and its major internal organs. Fundamentally, one needs to
know:
. the amount of radioactivity administered—the administered activity;
. the rate of radioactive decay of the administered radionuclide—the
physical half-life (or decay constant);
. each type of radiation emitted by the decaying radionuclide and its
frequency and average energy of emission—the equilibrium dose
constant;
. the fraction of the administered activity which localizes in each tissue or
organ (or ‘source region’)—the uptake or, more completely, the time–
activity function;
. the length of time the radioactive material resides in each tissue or
organ—the effective half-time (or residence time)—as derived from the
time–activity function;
. the total number of decays (nuclear transitions) which occur in each
tissue or organ (or source region)—the cumulated activity;
. the fraction of radiation energy which is absorbed in the tissue or organ
itself as well as in other tissues and organs (or ‘target regions’)—the
absorbed fractions; and
. the mass of each tissue or organ (target region).
As it is applied to diagnostic radiopharmaceuticals, its traditional applica-
tion, the MIRD schema implicitly assumes that activity and cumulated
activity are uniformly distributed within source regions and that radiation
energy is uniformly deposited within target regions. Moreover, dosimetry
for diagnostic radiopharmaceuticals is generally based on (a) average
time–activity data in animal models and/or in a small cohort of human
subjects and (b) age- and gender-specific ‘average’ models of human
anatomy. The traditional MIRD schema does not incorporate tumours as
either source or target regions.
As illustrated in figures 4.1 to 4.3 and developed below, the basic
calculation in the MIRD schema, yielding the mean absorbed dose to
target region rk from the activity (i.e., cumulated activity) in source region
The MIRD schema 91
Figure 4.1. ‘Step 1’ in the MIRD schema for internal radionuclide dosimetry:
compilation of the pertinent physical data (i.e., radionuclide-specific nuclear
decay data) and calculation of the pertinent physical parameters (i.e., absorbed
fractions and S factors for selected source region–target region pairs in a
mathematically formulated ‘average’ anatomic model). Adapted, in part,
from [14] and [20].
rh , follows [4, 13–16]:

P
ðrk A~h i i i ðrk rh Þ
D rh Þ ¼ ð4:15aÞ
mk
X
¼ A~h i i ðrk rh Þ ð4:15bÞ
i
¼ A~h Sðrk rh Þ ð4:15cÞ

where A~h is the cumulated activity in source region rh , that is, the total
number of decays in source region rh , mk is the mass of target region rk , i
is the equilibrium dose constant for radiation i, that is, the average energy
emitted per decay in the form of radiation i [27], i ðrk rh Þ is the absorbed
fraction in target region rk for radiation i emitted in source region rh , that
is, the fraction of energy of radiation i emitted in source region rh absorbed
in target region rk , i ðrk rh Þ is the specific absorbed fraction in target
acquisition of the pertinent biological data (i.e., time–activity data for the
radiopharmaceutical under consideration in selected organs) and calculation
of the pertinent biological parameters (i.e., cumulated activities, or residence
times, in the selected organs).
region rk for radiation i emitted in source region rh , that is, the fraction of
energy of radiation i emitted in source region rh that is absorbed per unit
mass in target region rk
i ðrk rh Þ
i ðrk rh Þ ð4:16Þ
mk
and Sðrk rh Þ is the radionuclide-specific S factor for target region rk
and source region rh , that is, the absorbed dose to target region rk per unit
cumulated activity in source region rh
P
ðr rh Þ
Sðrk rh Þ i i i k ð4:17Þ
mk
An important simplification of absorbed dose calculations was intro-
duced by combining the radionuclide-specific equilibrium dose constant
i , the source region-to-target region absorbed fraction i ðrk rh Þ, and
the target region mass mk into a single quantity, the S factor, as defined by
equation (4.17) [15]. As a result, given the cumulated activity A~h in a given
source region rh , one can use the tabulated S factors to yield, by a simple
multiplication, the absorbed dose contribution D ðrk rh Þ to a target
region rk . Besides computationally simplifying the determination of
The MIRD schema 93
integration of the pertinent physical and biological parameters to calculate the
mean absorbed doses to selected target regions in the mathematically formu-
lated anatomic model. Adapted, in part, from [13].
absorbed dose, the S factor conceptually clarified this task by isolating all
non-time-dependent and, largely, non-biology-dependent dosimetric factors
into a single parameter.
The total mean absorbed dose Dðrk Þ to target region rk is then calculated
by summing of the absorbed dose contributions from all source regions rh :
X A~h Pi i i ðrk rh Þ

Dðrk Þ ¼ ð4:18aÞ
h
mk
X X
¼ A~h i i ðrk rh Þ ð4:18bÞ
h i
X
¼ A~h Sðrk rh Þ ð4:18cÞ
h
The cumulated activity, or time integral of activity, A~h , in a given target

region rh is calculated by integration from the time of administration of the
radiopharmaceutical (t ¼ 0) to the time of its complete elimination (t ¼ 1)
of the time–activity function, Ah ðtÞ:
ð1
A~h ¼ Ah ðtÞ dt ð4:19Þ
0
The time–activity function includes both the physical decay constant of the
radionuclide and the biological disappearance constant(s) (see the detailed
discussion below). It is usually determined by discrete serial measurements.
Despite the complexity of the multiple underlying biological processes,
the time–activity data of virtually any radiopharmaceutical in any tissue or
organ can be accurately represented by an exponential function:
X
Ah ðtÞ ¼ ðAh Þ eðh Þj t ð4:20Þ
j
where ðAh Þj is the (extrapolated) activity at time t ¼ 0 for the jth exponential
component of the time–activity function in source region rh , and ðh Þj is the
clearance constant of the jth exponential component of the time–activity
function in source region rh , that is, the fraction per unit time of activity
eliminated for the jth exponential component of the time–activity function
in source region rh .
The clearance constant, ðh Þj , is actually the ‘effective’ (e) clearance
constant because it includes the effects of both biological clearance and
physical (i.e., radioactive) decay:
ðh Þj ½ðe Þh j ð4:21aÞ
¼ p þ ½ðb Þh j ð4:21bÞ
where p is the physical decay constant of the radionuclide, and ½ðb Þh j is the
biological clearance constant of the jth exponential component of the time–
activity function in source region rh , that is, the fraction per unit time of
activity biologically eliminated for the jth exponential component of the
time–activity function in source region r.
For each exponential component, the half-life (or half-time) is related to
its corresponding clearance constant:
ln 2
T1=2 ¼ ð4:22aÞ

0:693
¼ ð4:22bÞ

and the effective half-time, Te ¼ ðT1=2 Þe , includes the effects of both the
biological and physical half-lives:
1 1
Te ¼ þ ð4:23aÞ
Tp Tb
Tp Tb
¼ ð4:23bÞ
Tp þ Tb
where Tp is the physical half-life of the radionuclide, and Tb is the biological

half-time of the jth exponential component of the time–activity function.
Patient- and position-specific dosimetry 95
Integration of the time–activity function, Ah ðtÞ ¼ ðA0 Þh eðe Þh t , yields

the cumulated activity in source region rh :
X ðAh Þj X
A~h ¼ ¼ 1:44 ðAh Þj ½ðTe Þh j ð4:24aÞ
j
ðh Þj j
In the dosimetry literature, the quantity of cumulated activity has largely

been replaced by the residence time, h , in source region rh [13], equivalent
to the cumulated activity per unit administered activity:
A~h
h ¼ ð4:25aÞ
AA
X ðA~h Þj
¼ 1:44 ½ðTe Þh j ð4:25bÞ
j
AA
X
¼ 1:44 ðfh Þj ½ðTe Þh j ð4:25cÞ
j
where AA is the administered activity, and ð fh Þj is the (extrapolated) fraction

of the administered activity at time t ¼ 0 for the jth exponential component
of the time–activity function in source region rh .
Therefore,
A~h ¼ AA h ð4:26Þ
4.4. PATIENT- AND POSITION-SPECIFIC DOSIMETRY
Dr Robert Loevinger, one of the originators of the MIRD schema, has stated
that ‘ . . . there is in principle no way of attaching a numerical uncertainty to
the profound mismatch between the patient and the model (the totality of all
assumptions that enter into the dose calculation). The extent to which the
model represents in some meaningful way a patient, or a class of patients, is
always open to question, and it is the responsibility of the clinician to make
that judgment’ [13]. In contrast to diagnostic radiopharmaceuticals, therapeutic
radiopharmaceuticals engender much smaller risk–benefit ratios and therefore
markedly reduced tolerances for such inaccuracies in dose estimation. To the
extent, then, that specific patients deviate kinetically and anatomically from
the respective kinetic and anatomic averages, tissue dose estimates will be
inaccurate and individualized (i.e., patient-specific) dosimetry may be required.
4.4.1. Patient-specific dosimetry: case study—radioiodine therapy of

metastatic thyroid cancer
The basis of radioiodine treatment of metastatic thyroid cancer is the specific
uptake of administered radioiodide by and the resulting radiogenic damage
to differentiated thyroid cancer cells—assuming the cancerous thyroid has

been surgically removed, any residual normal thyroid ablated, and blood
levels of thyroid stimulating hormone (TSH) temporarily increased by with-
drawal of replacement thyroid hormone. Historically, the most commonly
employed dose prescription algorithm has been a fixed administered activity
[16], typically 100 mCi, possibly adjusted depending upon the extent of
disease [28]. While simple and convenient, this approach ignores altogether
patient-specific kinetic and anatomic factors. Alternatively, since one does
not know the actual, generally low radioiodine uptakes in and resulting
absorbed doses to the metastases, one should administer the maximum
activity which will not induce serious side effects in order to maximize the
therapeutic effect. Benua et al. [29] empirically determined the criteria that
define the maximum ‘safe’ administered activity: the mean absorbed dose
to blood (as a surrogate for the bone marrow) must not exceed 200 rad
and, at 48 h post-administration, the projected lung and the total body
retention must not exceed 80 and 120 mCi, respectively. In practice, the
200 rad absorbed dose limit to blood is generally the therapy-limiting
criterion. The implementation of this patient-specific treatment planning
paradigm is described below [16].
Following oral administration of a 2 mCi tracer administration of 131 I
iodide, serial measurements are performed daily (1 to 4 days post-administra-
tion) of the activity concentration in blood and the activity in the total body.
The cumulated activities in the total body and the cumulated activity
concentration in blood are equated with the areas (determined by numerical
integration) under the total body time–activity curve and the area under the
blood time–activity concentration curve, respectively [16, 29]:
area under curve
ðn
exp½0:0861ti fAi ½expð0:0861ðti þ 1 ti ÞAi Þgðti þ 1 ti Þ
¼
i¼1 Ai
ln
exp½0:0861ðti þ 1 ti ÞAi
expð0:861tn An Þ
þ ð4:27Þ
0:0861
where Ai is the decay-corrected activity (%) in total body or the activity
concentration (%/l) in blood) at time ti post-administration (in days).
Thus, the time–activity data in the tumour and total body and the time–
activity concentration data in blood are integrated numerically up to the last
measured datum (An at time tn post-administration), corresponding to the
first term in the sum on the right-hand side of equation (4.27), and analytically
thereafter, conservatively assuming elimination of the radionuclide by physi-
cal decay only, corresponding to the second term in the sum. An important
advantage of numerical integration is its generality: it is adaptable to non-
monotonic time–activity curves (e.g., the blood time–activity concentration
curve for radioiodide) and no simplifying assumptions are introduced

regarding the analytic form of the time–activity curve. On the other hand,
a disadvantage of the assumption of radionuclide elimination by physical
decay only following the last measurement is that it results in a systematic
overestimation, perhaps substantial, of the overall area under the time–
activity curve. Furhang et al. [30] have recently modified the ‘Benua formal-
ism’ [29] by replacing numerical integration of the time–activity data with
analytical integration of exponential fits to these data.
The absorbed doses (in rad/mCi) to blood and to tumour are equated
with the sum of the mean total body absorbed dose from penetrating ()
radiation and the mean self-irradiation absorbed dose from non-penetrating
() radiation [30]. The mean total body absorbed dose from -radiation is
calculated using the height- and mass-dependent mean total-body geometric
factor and specific -ray constant for iodine-131, assuming uniform total
body distribution of cumulated activity [31–34]. The mean self-irradiation
absorbed dose from -radiation is calculated using the equilibrium absorbed
dose constant for -radiation for 131 I [27], assuming complete local absorp-
tion [31–34]. The pertinent equations are:
mean blood absorbed dose

¼ mean absorbed dose from total body activity
þ mean absorbed dose from ‘local’ (blood) activity; ð4:28Þ
where
mean gamma absorbed dose from total body activity (rad/mCi)

¼ 0:000235 g (cm) (R cm2 /mCi h)
area under total body timeactivity curve (% administered
activity day) ð4:29Þ
mean gamma blood absorbed dose from local activity (rad/mCi)

¼ 0:233 np (g rad/Ci h) area under blood timeactivity
concentration curve (% administered activity day) ð4:30Þ
mean beta radiation tumour dose from local activity (rad/mCi)

¼ 0:233 np (g rad/Ci h) area under tumour timeactivity
curve (% administered activity day/L)=[tumour mass (kg)] ð4:31Þ
and g is the height- and body-mass-dependent total body mean geometric

factor [31–34], is the specific -ray constant for 131 I [31–34] (2.23 R cm2 /
mCi h [11]), f is the exposure-to-absorbed dose conversion factor (1 rad/R
for water/soft tissue (see above)), np is the equilibrium dose constant for
non-penetrating radiations for 131 I (0.41 g rad/Ci h [27]).
4.4.2. Adaptation of the MIRD schema to patient-specific dosimetry

Even though the standard anthropomorphic models used in the MIRD
schema represent normal human anatomy and thus do not include tumours,
the schema can be adapted to patient-specific normal-organ dosimetry for
planning radionuclide therapy based on the maximum tolerated activity
[16, 35]. Not surprisingly, the most important quantitative adjustment in
this adaptation involves the tumour-bearing organ(s). The proposed adapta-
tion is rather crude and, importantly, has not in any sense been validated. (In
the following development, all absorbed doses are still mean values even
though, to simplify the notation, the overstrike bar has been eliminated
over the symbol, D).
For organ ‘non-self’ irradiation (source region rh 6¼ target region rk ), S
factors are relatively insensitive to organ (i.e., source and target region) size
and shape. Therefore, unless the source and/or target regions are grossly
abnormal (e.g., due to the presence of tumour), Reference Man (or Reference
Woman) S factors may be applied to specific patients for calculating the
organ non-self absorbed dose contribution:
patient Reference Man
if rh 6¼ rk ð4:32Þ
Sðrk rh Þ Sðrk rh Þ
For organ ‘self’ irradiation (source region rh ¼ target region rk ), S factors

are approximately inversely proportion to organ mass—because most of
the self-dose to any organ is contributed by non-penetrating radiations
which are completely absorbed locally while absorbed dose is inversely
proportional to the organ mass. Therefore, for a normal organ (i.e., an
organ without tumour), S factors adjusted for the difference in masses
between the patient and Reference Man organs may be applied to specific
patients for calculating the self-absorbed dose contribution:
Reference Man
patient Reference Man target region ðrk Þ mass
ð4:33Þ
Sðrk rh Þ Sðrk rh Þ patient target
region ðrk Þ mass
if rh ¼ rk .
In principle, patient organ masses may be estimated by computed
tomography or magnetic resonance imaging. In practice, however, masses
of normal organs may not be available. A more practical, though certainly
less accurate, adaptation of equation (33) based on the patient and Reference
Man total body masses may then be used:
Reference Man
patient Reference Man total body mass
ð4:34Þ
Sðrk rh Þ Sðrk rh Þ patient total
body mass
if rh ¼ rk and the patient target region ðrk Þ mass is not available, assuming,
arbitrarily, that organ mass is directly proportional to total body mass.
For a tumour-bearing organ, adaptation of the MIRD schema is some-
what more complicated. First, the self-irradiation absorbed dose and S factor
for the tumour-bearing organ can be separated into their penetrating and
non-penetrating radiation components:
Dðrh rh Þ Dp ðrh rh Þ þ Dnp ðrh rh Þ ð4:35Þ
Sðrh rh Þ Sp ðrh rh Þ þ Snp ðrh rh Þ ð4:36aÞ
P
i ðnp Þi ðnp Þi ðrh rh Þ
¼ Sp ðrh rh Þ þ ð4:36bÞ
mh
P
i ðnp Þi
¼ Sp ðrh rh Þ þ ð4:36cÞ
mh
np
¼ Sp ðrh rh Þ þ ð4:36dÞ
mh
where Sðrh rh Þ is the total self-irradiation S factor in source and target
region rh , Sp ðrh rh Þ is the self-irradiation S factor for penetrating radiations
in source and target region rh , Snp ðrh rh Þ is the self-irradiation S factor
for non-penetrating radiations in source and target region rh , ðnp Þi is the
equilibrium dose constant for non-penetrating radiation i, ðnp Þi ðrh rh Þ is
the self-irradiation absorbed fraction in non-penetrating radiations for
source and target region rh ,
ðnp Þi ðrh rh Þ ¼ 1 ð4:37Þ
P
mh is the mass of the source and target region rh , and np i ðnp Þi is the
total equilibrium dose constant for non-penetrating radiations.
Therefore,
np
Sp ðrh rh Þ ¼ Sðrh rh Þ : ð4:38Þ
mh
The patient-specific self-irradiation S factor for penetrating radiations can
then be calculated:
Reference Man
patient Reference Man target region ðrk Þ mass
: ð4:39Þ
Sðrh rh Þ Sðrh rh Þ patient target
region ðrk Þ mass
The self-irradiation absorbed dose for non-penetrating radiations in source

and target region rh is
np ~
Dnp ðrh rh Þ ¼ Ah : ð4:40Þ
mh
Finally, the total absorbed dose to the patient target region rk is the sum of
the non-self-irradiation dose contributions (equation (15c)) and, primarily,
the self-irradiation absorbed doses from penetrating radiations with patient
S factors (equation (39)), Dp ðrh rh Þ, and from penetrating radiations,
Dnp ðrh rh Þ (equation (40)). Note that the absorbed dose from penetrating
radiations (X- and -rays) includes contributions from activity in both the
normal liver and from a tumour in the liver because of the relatively long
distances penetrated by such radiations. In contrast, the absorbed dose
from non-penetrating radiations (-particles) is contributed only by activity
in the normal liver itself because such radiations are completely absorbed
within macroscopic tumours, that is, the radiation cannot penetrate into
the surrounding normal liver.
To determine masses of the tumour-bed organ plus the tumour(s),
regions of interest (ROIs) may be drawn around the entire liver, including
the tumour(s) as well as the normal liver, on computed tomograms (CTs)
or magnetic resonance images (MRIs). ROIs can be drawn specifically
around the individual tumours to determine their respective masses. The
normal tumour-bed organ mass can be calculated by subtracting the
combined masses of the individual tumours from the mass of the entire
tumour-bed organ plus tumour(s). An analogous ROI analysis can be
applied to the gamma camera images to determine the activities and cumu-
lated activities specifically in the tumour-bed organ and tumour(s).
4.4.3. Position-specific dosimetry: calculation of non-uniform dose

distributions
In addition to patient-specific kinetics and anatomy, the issue of spatial non-
uniformity of dose has become increasingly important at the macroscopic
[36–60] and microscopic [61–79] levels. There are at least three approaches
to the calculation of macroscopic non-uniform dose distributions [60], dose
point-kernel convolution, Monte Carlo simulation, and voxel S factors,
and these will be elucidated further in other chapters. The dose point-
kernel is currently the most widely used of these approaches [36–38, 41, 42,
53–56], primarily because of the demanding computational requirements of
Monte Carlo simulation and the limited availability of voxel S factors. (A
dose point-kernel is the radial distance-dependent absorbed dose about an
isotropic point source in an infinite homogeneous medium—typically a
soft tissue-equivalent medium such as water.) With the wider availability
of high-speed desktop computers and of compatible simulation codes, the
Summary 101
use of Monte Carlo analysis has increased [39, 40, 80]. Monte Carlo-based
dosimetry can more accurately account for tissue variations in mass density
and atomic number as well as edge effects which may be important at the
periphery of the body and at soft tissue–lung and –bone interfaces [39].
For example, if the relevant distribution data were somehow available
(e.g., by autoradiography of biopsy specimens), Monte Carlo analysis
might be particularly applicable to normal lung dosimetry in radioiodine
treatment of metastatic thyroid cancer, particularly when in the form of
dosimetrically problematic miliary disease. This method remains computa-
tionally time-consuming, however [39]. Tabulations of voxel S factors,
conceptually equivalent to voxel source-kernels (the mean absorbed dose
to a target voxel per radioactive decay in a source voxel, both of which are
contained in an infinite homogeneous soft-tissue medium) [49, 59], are
becoming available [60]. In contrast to dose point-kernel- and Monte
Carlo-based techniques, the voxel S factor method does not require
specialized computer facilities and is relatively fast, and thus may emerge
as the practical method of choice for calculation of macroscopic non-
uniform dose distributions.
Once a dose distribution has been calculated, a corresponding ‘dose–
volume histogram’ can be derived [81]. Basically, a dose–volume histogram
is a graph of the fraction of the tumour or organ volume receiving a specified
dose versus the dose (differential form) or the fraction of the tumour or organ
volume receiving less than a specified dose versus the dose (integral, or
cumulative, form). It graphically presents the minimum, mean and maximum
doses and the dispersion about the mean dose. The greater this dispersion,
the more non-uniform is the dose distribution.
An important practical component of macroscopic non-uniform dosi-
metry is the ability to fuse, or register, tomographic images from multiple
modalities [46, 47, 56, 82–84]. Dose distributions, calculated from three-
dimensional activity distributions measured by scintigraphic imaging (i.e.,
SPECT or PET), may be presented as isodose contours or colour-coded
images. By image fusion, such isodose contours or colour-coded images
can be superimposed on or juxtaposed with the corresponding anatomy to
allow correlation of doses with tumour and normal organs (as imaged by
CT or MRI) [36, 56, 59, 60].
4.5. SUMMARY
Radiation dosimetry deals with the determination of the amount and the
spatial and temporal distribution of energy deposited in matter by ionizing
radiation. Internal radionuclide radiation dosimetry specifically deals with
the deposition of radiation energy in tissue due to a radionuclide within
the body. However, unlike external radiation dose (which can often be
measured), internal radiation dose must be calculated. These procedures

have evolved over more than 60 years from relatively simple approaches to
those with a high level of sophistication. Internal dosimetry has been applied
to the determination of tissue doses and related quantities for occupational
exposures in radiation protection, environmental exposures in radiation
epidemiology, and diagnostic and therapeutic exposures in nuclear medicine.
This chapter has reviewed the concepts associated with the computational
approaches to internal radiation dosimetry.
REFERENCES
[1] Fisher D R 1994 Radiation dosimetry for radioimmunotherapy. An overview of

current capabilities and limitations Cancer 73 905–11
[2] Howell R W, Wessels B W, Loevinger R, Watson E E, Bolch W E, Brill A B, Charkes
N D, Fisher D R, Hays M T, Robertson J S, Siegel J A and Thomas S R 1999 The
MIRD perspective 1999. Medical Internal Radiation Dose Committee J. Nucl.
Med. 40 3S–10S
[3] Stabin M, Tagesson M, Thomas S R, Ljungberg M and Strand S E 1999 Radiation
dosimetry in nuclear medicine Appl. Radiat. Isot. 50 73–87
[4] Zanzonico P B 2000 Internal radionuclide radiation dosimetry: a review of basic
concepts and recent developments J. Nucl. Med. 41 297–308
[5] Rossi H 1968 Microscopic energy distributions in irradiated manner. In Radiation
Dosimetry, vol I: Fundamentals eds F Attix, W Roesch and E Tochlin (New York:
Academic) pp 43–92
[6] ICRU 1980 Radiation Quantities and Units. International Commission on Radiation
Units and Measurements (ICRU) Report 33. International Commission on
Radiation Units and Measurements (ICRU), Bethesda, MD
[7] ICRU 1983 Microdosimetry. International Commission on Radiation Units and
Measurements (ICRU) Report 36. International Commission on Radiation Units
and Measurements (ICRU), Bethesda, MD
[8] NCRP 1985 SI Units in Radiation Protection and Measurements. National Council
on Radiation Protection and Measurements (NCRP) Report 82. National Council on
Radiation Protection and Measurements (NCRP), Bethesda, MD
[9] BEIR 1990 Health Effects of Exposure to Low Levels of Ionizing Radiation (BEIR
V). Biological Effects of Ionizing Radiation (BEIR) Committee, National Research
Council. National Academy Press, Washington, DC
[10] ICRP 1990 1990 Recommendations of the International Commission on Radiological
Protection. International Commission on Radiological Protection (ICRP) Publica-
tion 60, Annals of the ICRP 21 (Elmsford, NY: Pergamon)
[11] Johns H and Cunningham J 1974 The Physics of Radiology (Springfield: Thomas)
[12] Attix F 1986 Introduction to Radiological Physics and Radiation Dosimetry (New
York: Wiley)
[13] Loevinger R, Budinger T, Watson E, Atkins H, Blau M, Lathrop K, Poston J,
Robertson J, Thomas S and Weber D 1991 MIRD Primer for Absorbed Dose
Calculations (revised edition) (New York: Society of Nuclear Medicine)
References 103
[14] Snyder W, Ford M, Warner G and Fisher H Jr 1969 Estimates of absorbed fractions
for monoenergetic photon sources uniformly distributed in various organs of a
heterogeneous phantom. MIRD Pamphlet No 5 J. Nucl. Med. 10 (Suppl. 3) 5–52
[15] Snyder W, Ford M, Warner G and Watson S 1975 ‘S’, Absorbed dose per unit
cumulated activity for selected radionuclides and organs. Medical Internal
Radiation Dose (MIRD) Pamphlet No 11 (New York: Society of Nuclear Medicine)
[16] Zanzonico P, Brill A and Becker D 1995 Radiation dosimetry. In Principles of Nuclear
Medicine eds H Wagner, Z Szabo and J Buchanan (Philadelphia: Saunders) 2nd
edition pp 106–34
[17] ICRP 1988 Radiation Dose to Patients from Radiopharmaceuticals. International
Commission on Radiological Protection (ICRP) Publication 53 (New York:
Pergamon)
[18] Watson E 1999 Foreword J. Nucl. Med. 40 1S–2S
[19] Loevinger R 1989 The MIRD perspective. In Dosimetry of Administered
Radionuclides eds S Adelstein, A Kassis and R Burt (Washington, DC: American
College of Nuclear Physicians/Department of Energy) pp 29–43
[20] Cristy M and Eckerman K 1987 Specific Absorbed Fractions of Energy at Various
Ages from Internal Photon Sources (Parts I–VII). Oak Ridge National Laboratory
Report ORNL/TM-8381/V1-7 (Springfield, VA: National Technical Information
Service, Dept of Commerce)
[21] ICRP 1975 Report of the Task Group on Reference Man ICRP Publication 23 (Oxford:
Pergamon)
[22] Stabin M 1996 MIRDOSE: personal computer software for internal dose assessment
in nuclear medicine J. Nucl. Med. 37 538–46
[23] Sgouros G, Bigler R and Zanzonico P 1988 DOSCAL: a tumor-incorporating mean
absorbed dose calculation program (Abstract) J. Nucl. Med. 29 874
[24] Johnson T 1988 MABDOS: a generalized program for internal dosimetry Comput.
Meth. Programs Biomed. 37 159–67
[25] Johnson T K, McClure D and McCourt S 1999 MABDOSE. II: Validation of a
general purpose dose estimation code [In Process Citation] Med. Phys. 26 1396–403
[26] Johnson T K, McClure D and McCourt S 1999 MABDOSE. I: Characterization of a
general purpose dose estimation code [In Process Citation] Med. Phys. 26 1389–95
[27] Weber D, Eckerman K, Dillman L and Ryman J 1989 MIRD: Radionuclide Data and
Decay Schemes (New York: Society of Nuclear Medicine)
[28] Beierwaltes W H, Nishiyama R H, Thompson N W, Copp J E and Kubo A 1982
Survival time and ‘cure’ in papillary and follicular thyroid carcinoma with distant
metastases: statistics following University of Michigan therapy J. Nucl. Med. 23
561–8
[29] Benua R, Cicale N and Sonenberg M 1962 The relation of radioiodine dosimetry to
results and complications in the treatment of metastatic thyroid cancer Am. J.
Roentgenol. 87 171–82
[30] Furhang E E, Larson S M, Buranapong P and Humm J L 1999 Thyroid cancer
dosimetry using clearance fitting J. Nucl. Med. 40 131–6
[31] Marinelli L, Quimby E and Hine G 1948 Dosage determination with radioactive
isotopes. II. Practical considerations in therapy and protection Am. J. Roentgenol.
Radium Therapy 59 260–80
isotopes. I. Fundamental dosage formulae Nucleonics 2 56–66
isotopes. II. Biological considerations and practical applications Nucleonics 2
449
[34] Loevinger R, Holt J G and Hine G J 1956 Internal administered radioisotopes. In
Radiation Dosimetry eds G J Hine and G L Brownell (New York: Academic)
pp 80173
[35] Zanzonico P B 2001 Radiation dosimetry of rhenium-188 Lipiodol: a workshop. In
Lecture Notes: Regional Training Workshop on Radionuclide Treatment of Liver
Cancer. NMS/11/RTC-2001-01/RCA-1 (Perth) (Perth: International Atomic Energy
Agency (IAEA)) pp 12861
[36] Erdi A K, Wessels B W, DeJager R, Erdi Y E, Atkins F B, Yorke E D, Smith L, Huang
E, Smiddy M, Murray J et al. 1994 Tumor activity confirmation and isodose curve
display for patients receiving iodine-131-labeled 16.88 human monoclonal antibody
Cancer 73 93244
[37] Erdi A K, Yorke E D, Loew M H, Erdi Y E, Sarfaraz M and Wessels B W 1998 Use of
the fast Hartley transform for three-dimensional dose calculation in radionuclide
therapy Med. Phys. 25 222633
[38] Erdi A K, Erdi Y E, Yorke E D and Wessels B W 1996 Treatment planning for radio-
immunotherapy Phys. Med. Biol. 41 200926
[39] Furhang E E, Chui C S and Sgouros G 1996 A Monte Carlo approach to patient-
specific dosimetry Med. Phys. 23 15239
emitter therapy Med. Phys. 24 116372
[41] Giap H B, Macey D J, Bayouth J E and Boyer A L 1995 Validation of a dose-
point kernel convolution technique for internal dosimetry Phys. Med. Biol. 40
36581
[42] Giap H B, Macey D J and Podoloff D A 1995 Development of a SPECT-based three-
dimensional treatment planning system for radioimmunotherapy J. Nucl. Med. 36
188594
[43] Howell R W, Rao D V and Sastry K S 1989 Macroscopic dosimetry for radioimmuno-
therapy: nonuniform activity distributions in solid tumors Med. Phys. 16 6674
[44] Humm J L 1986 Dosimetric aspects of radiolabeled antibodies for tumor therapy
J. Nucl. Med. 27 14907
[45] Humm J L and Cobb L M 1990 Nonuniformity of tumor dose in radioimmuno-
therapy [see comments] J. Nucl. Med. 31 7583
[46] Koral K F, Zasadny K R, Kessler M L, Luo J Q, Buchbinder S F, Kaminski M S,
Francis I and Wahl R L 1994 CT–SPECT fusion plus conjugate views for
determining dosimetry in iodine-131-monoclonal antibody therapy of lymphoma
patients J. Nucl. Med. 35 1714-20
[47] Koral K F, Lin S, Fessler J A, Kaminski M S and Wahl R L 1997 Preliminary results
from intensity-based CT–SPECT fusion in I-131 anti-B1 monoclonal-antibody
therapy of lymphoma Cancer 80 253844
[48] Leichner P K and Kwok C S 1993 Tumor dosimetry in radioimmunotherapy:
methods of calculation for beta particles Med. Phys. 20 52934
[49] Liu A, Williams L, Wong J and Raubitschek A 1997 A voxel source kernel (VSK)
method for rapid, patient-specific dose estimates in radioimmunotherapy (RIT)
(Abstract) J. Nucl. Med. 38(suppl) 106P
References 105
[50] Liu A, Williams L E, Lopatin G, Yamauchi D M, Wong J Y and Raubitschek A A

1999 A radionuclide therapy treatment planning and dose estimation system J. Nucl.
Med. 40 11513
[51] Ljungberg M and Strand S E 1988 Dose planning with SPECT Int. J. Cancer Suppl. 2
6770
[52] Mayer R, Dillehay L E, Shao Y, Song S, Zhang Y G, Bartholomew R M and
Williams J R 1994 A new method for determining dose rate distribution from
radioimmunotherapy using radiochromic media Int. J. Radiat. Oncol. Biol. Phys.
28 50513
[53] Sgouros G, Barest G, Thekkumthala J, Chui C, Mohan R, Bigler R E and Zanzonico
P B 1990 Treatment planning for internal radionuclide therapy: three-dimensional
dosimetry for nonuniformly distributed radionuclides J. Nucl. Med. 31 188491
[54] Sgouros G, Chiu S, Pentlow K S, Brewster L J, Kalaigian H, Baldwin B, Daghighian
F, Graham M C, Larson S M and Mohan R 1993 Three-dimensional dosimetry for
radioimmunotherapy treatment planning J. Nucl. Med. 34 1595601
[55] Sgouros G 1999 Treatment planning for internal emitter therapy: methods,
applications and clinical implications. In Proceedings of the Sixth International
Radiopharmaceutical Dosimetry Symposium (ORISE 99-0164) eds A Schlafke-
Stelson, M Stabin and R Sparks (Oak Ridge, TN: US Dept of Energy and Oak
Ridge Associated Universities) pp 1325
[56] Kolbert K S, Sgouros G, Scott A M, Bronstein J E, Malane R A, Zhang J, Kalaigian
H, McNamara S, Schwartz L and Larson S M 1997 Implementation and evaluation
of patient-specific three-dimensional internal dosimetry J. Nucl. Med. 38 3018
[57] Uchida I, Yamada Y, Oyamada H and Nomura E 1992 Calculation algorithm of
three-dimensional absorbed dose distribution due to in vivo administration of
nuclides for radiotherapy Kaku Igaku 29 1299306
[58] Wessels B W, Yorke E D and Bradley E W 1990 Dosimetry of heterogeneous uptake
of radiolabeled antibody for radioimmunotherapy Front. Radiat. Ther. Oncol. 24
1048
[59] Akabani G, Hawkins W, Eckblade M and Leichner P 1997 Patient-specific dosimetry
using quantitative SPECT imaging and three-dimensional discrete Fourier transform
convolution J. Nucl. Med. 38 30814
[60] Bolch W E, Bouchet L G, Robertson J S, Wessels B W, Siegel J A, Howell R W, Erdi
A K, Aydogan B, Costes S, Watson E E, Brill A B, Charkes N D, Fisher D R, Hays M
T and Thomas SR 1999 MIRD Pamphlet No 17: The dosimetry of nonuniform
activity distributions—radionuclide S values at the voxel level. Medical Internal
Radiation Dose Committee J. Nucl. Med. 40 11S36S
[61] Adelstein S, Kassis A and Sastry K 1986 Cellular vs organ approaches to dose
estimates. In Proceedings of the Fourth International Radiopharmaceutical Dosimetry
Symposium (Conf-85113 (DE86010102) eds A Schlafke-Stelson and E Watson
(Oak Ridge, TN: US Dept of Energy and Oak Ridge Associated Universities) pp 1325
[62] Goddu S M, Rao D V and Howell R W 1994 Multicellular dosimetry for micrometa-
stases: dependence of self-dose versus cross-dose to cell nuclei on type and energy of
radiation and subcellular distribution of radionuclides J. Nucl. Med. 35 52130
[63] Goddu S, Howell R, Bouchet L, Bolch W, Rao D, Budinger T, Fisher D, Hays M,
Lathrop K, Robertson J, Siegel J, Thomas S, Watson E and Wessels B 1997
MIRD Cellular S Factors: Self-Absorbed Dose per Unit Cumulated Activity for
Selected Radionuclides and Monoenergetic Electrons and Alpha Particle Emitters
Incorporated into Different Cell Compartments (Reston, VA: Society of Nuclear

Medicine)
[64] Griffiths G L, Govindan S V, Sgouros G, Ong G L, Goldenberg D M and Mattes M J
1999 Cytotoxicity with Auger electron-emitting radionuclides delivered by antibodies
Int. J. Cancer 81 98592
[65] Humm J L, Howell R W and Rao D V 1994 Dosimetry of Auger-electron-emitting
radionuclides: Report No 3 of AAPM Nuclear Medicine Task Group No 6 Med.
Phys. 21 190115 [published erratum appears in Med. Phys. 1995 22 1837]
[66] Humm J L, Macklis R M, Lu X Q, Yang Y, Bump K, Beresford B and Chin L M 1995
The spatial accuracy of cellular dose estimates obtained from 3D reconstructed serial
tissue autoradiographs Phys. Med. Biol. 40 16380
[67] Humm J 1987 A microdosimetric model of astatine-211 labeled antibodies for
radioimmunotherapy Int. J. Radiat. Oncol. Biol. Phys. 13 176773
[68] Macklis R M, Lin J Y, Beresford B, Atcher R W, Hines J J and Humm J L 1992
Cellular kinetics, dosimetry, and radiobiology of alpha-particle radioimmuno-
therapy: induction of apoptosis Radiat. Res. 130 2206
[69] O’Donoghue J A and Wheldon T E 1996 Targeted radiotherapy using Auger electron
emitters Phys. Med. Biol. 41 197392
[70] Wrenn M E, Howells G P, Hairr L M and Paschoa A S 1973 Auger electron dosimetry
Health Phys. 24 64553
[71] Yorke E D, Williams L E, Demidecki A J, Heidorn D B, Roberson P L and Wessels
B W 1993 Multicellular dosimetry for beta-emitting radionuclides: autoradiography,
thermoluminescent dosimetry and three-dimensional dose calculations Med. Phys. 20
54350
[72] Zalutsky M R, Stabin M, Larsen R H and Bigner D D 1997 Tissue distribution and
radiation dosimetry of astatine-211-labeled chimeric 81C6, an alpha-particle-emitting
immunoconjugate Nucl. Med. Biol. 24 25561
[73] Kassis A I, Adelstein S J, Haydock C and Sastry K S 1980 Radiotoxicity of 75Se and
35
S: theory and application to a cellular model Radiat. Res. 84 40725
[74] Kassis A I, Adelstein S J, Haydock C, Sastry K S, McElvany K D and Welch M J
1982 Lethality of Auger electrons from the decay of bromine-77 in the DNA of
mammalian cells Radiat. Res. 90 36273
[75] Kassis A I, Adelstein S J, Haydock C and Sastry K S 1983 Thallium-201: an
experimental and a theoretical radiobiological approach to dosimetry J. Nucl. Med.
24 116475
[76] Kassis A I, Sastry K S and Adelstein S J 1985 Intracellular distribution and radio-
toxicity of chromium-51 in mammalian cells: Auger-electron dosimetry J. Nucl.
Med. 26 5967
[77] Kassis A I, Fayad F, Kinsey B M, Sastry K S, Taube R A and Adelstein S J 1987
Radiotoxicity of 125I in mammalian cells Radiat. Res. 111 30518
[78] Kassis A I, Sastry K S and Adelstein S J 1987 Kinetics of uptake, retention, and
radiotoxicity of 125IUdR in mammalian cells: implications of localized energy
deposition by Auger processes Radiat. Res. 109 7889
[79] Kassis A I 1992 The MIRD approach: remembering the limitations J. Nucl. Med. 33
7812 [editorial; comment]
planning system Acta Oncol. 35 36772
References 107
[81] O’Donoghue J A 1999 Implications of nonuniform tumor doses for radioimmuno-

therapy [In Process Citation] J. Nucl. Med. 40 133741
[82] Scott A M, Macapinlac H A, Divgi C R, Zhang J J, Kalaigian H, Pentlow K, Hilton
S, Graham M C, Sgouros G, Pelizzari C et al 1994 Clinical validation of SPECT and
CT/MRI image registration in radiolabeled monoclonal antibody studies of
colorectal carcinoma J. Nucl. Med. 35 197684
[83] Scott A M, Macapinlac H, Zhang J J, Kalaigian H, Graham M C, Divgi C R, Sgouros
G, Goldsmith S J and Larson S M 1994 Clinical applications of fusion imaging in
oncology Nucl. Med. Biol. 21 77584
[84] Scott A M, Macapinlac H, Zhang J, Daghighian F, Montemayor N, Kalaigian H,
Sgouros G, Graham M C, Kolbert K, Yeh S D et al 1995 Image registration of
SPECT and CT images using an external fiduciary band and three-dimensional
surface fitting in metastatic thyroid cancer J. Nucl. Med. 36 1003
Chapter 5
Mathematical models of the human

anatomy
John W Poston Sr, Wesley E Bolch
and Lionel G Bouchet
5.1. INTRODUCTION
5.1.1. Early approaches to dose assessment

The first scientists to be involved in estimating the dose due to internally
distributed radionuclides were very familiar with the techniques used for
dosimetry of external radiation sources and radium implants [1]. These
techniques were already relatively well developed. However, unlike external
radiation exposure, the radiation-absorbed dose due to an internally
deposited radionuclide usually had to be calculated rather than measured.
Procedures to be followed in these dose calculations probably began with
the formulations of Marinelli and his colleagues in the 1940s [2, 3]. A
widely used elaboration of the Marinelli formulation was due to Loevinger
and his colleagues was published in the 1950s [4, 5].
Early approaches to such calculations were very simple and took into
account the activity and the ‘effective half-life’ of the radionuclide in the
organ, the energy deposited in the organ, and an approximate organ mass.
Typically, it was assumed that the radionuclide was uniformly distributed
in the homogeneous organ of interest. Usually, the calculations were made
of the absorbed dose rate from -emitting and -emitting radionuclides
separately. In the -radiation calculations, it was assumed that all the -
radiation energy was absorbed in the organ containing the radionuclide.
For -emitters, a ‘geometry factor’ was applied to account for the radiation
energy escaping the organ. Dose calculations were completed for simple
shapes such as right circular cylinders as well as models for specific organs
such as the thyroid gland and the total body [4,5].
108
Introduction 109
5.1.2. Need for mathematical models in dose assessment

While these relatively simple approaches to dosimetry provided reasonable
estimates of the doses due to internally deposited radionuclides, researchers
and practitioners longed for better representations of individual organs as
well as the human body and techniques that could provide more accurate
dose estimates. There was a constant desire to model the organs of the
body, and the entire body itself, in a more realistic fashion. This desire led
to a progression of developments that ultimately produced the very realistic
and complex mathematical models (called phantoms) in use today. However,
this progression took some time, starting with very simple shapes and very
simple calculational approaches in the late 1950s and early 1960s.
The ICRP in its recommendations on internal exposure assumed that
organs of the body could be represented by spheres of varying radii [6]. Each
organ was represented by a sphere and the effective radius of the sphere was
specified. In these simple calculations, the radionuclide source was assumed
to be at the centre of the sphere and the ‘effective absorbed energy’ was
calculated for each organ. Corrections were made for photon energy lost
from the sphere. In addition, the ICRP specified a radius for the total body
and assumed that it could be represented by a sphere with a radius of 30 cm.
It is interesting to note that the same 30 cm radius as used for the organs
designated the ‘muscle,’ the gastrointestinal tract and the small intestine.
Increases in the size and speed of digital computers allowed improvements
in the models and the approaches taken to radiation transport calculations. In
addition, as the number of radionuclides in use in nuclear medicine increased
and more was understood about the distribution of these materials in the
organs, there was a need for more detailed descriptions of the organs in
which specific regions or structures within the organ were delineated.
5.1.3. Conceptual role of mathematical phantoms

Conceptually, the purpose of the phantom is to represent the organ or tissue
of interest, to allow the radionuclide of interest to be contained in a specific
volume, and to provide a medium that absorbs and scatters the radiation
emitted in a manner similar to tissue. A phantom is a mathematical model
designed to represent an organ or tissue of the body, an organ system or
perhaps the entire body. Early designs were simple shapes such as right
circular cylinders, spheres or discs. Usually, the composition of the phantom
was specified and the constituents represented the major elements found in
soft tissue, muscle, bone, lungs, etc. However, in early phantoms, it was
often assumed that the material was simply water. As will be discussed
below, the complexity of the phantoms increased as technology developed.
These developments included improvements of the details of the geometrical
design as well as expansions of the elemental composition of the phantom.
110 Mathematical models of the human anatomy
5.2. HISTORICAL DEVELOPMENTS
5.2.1. Simple models of Brownell, Ellett and Reddy

Brownell et al. published perhaps the most extensive early compilation of
data for photon dosimetry in MIRD Pamphlet No 3 [7]. This pamphlet,
which summarized much of their earlier research, provided estimates
obtained via Monte Carlo calculations of the photon-absorbed fractions of
energy for centrally located point sources in three phantoms with simple
shapes. Phantom shapes included elliptical cylinders, right circular cylinders
and spheres. All phantoms were assumed to consist of a homogeneous tissue
composed of carbon, hydrogen, oxygen and nitrogen, with trace amounts of
chlorine and sodium.
For phantoms shaped as elliptical cylinders, Brownell and his colleagues
considered 13 monoenergetic photon sources as well as radium. Photon
energies ranged from 0.02 to 2.75 MeV. There were 19 elliptical cylinders
ranging in mass from 2 to 200 kg (although one wonders why such a large
mass was of interest). For right circular cylinders, only six energies
were considered (ranging from 0.04 to 1.46 MeV) but the same 19 masses
were included in the calculations. For spherical shapes, photon energies
ranged from 0.02 to 2.75 MeV and phantoms with the same 19 masses
were considered.
In addition, these authors calculated absorbed fractions of energy for
uniform distributions of activity in some simple phantoms. Results were
presented for ellipsoids, small spheres, thick ellipsoids and flat ellipsoids.
Again, monoenergetic photon sources in the range 0.02 to 2.75 MeV and
phantoms of various masses were considered. For ellipsoids, the masses
ranged from 2 to 200 kg, as in previous calculations. For small spheres,
thick ellipsoids and flat ellipsoids only masses in the range 0.3 to 6 kg were
considered.
5.2.2. Early models developed by Snyder

In the late 1950s and 1960s, many research groups were developing simple
phantoms and Monte Carlo techniques to be used in dose assessment. One
major effort was led by W S Snyder at the Oak Ridge National Laboratory
(ORNL). Initially, the ORNL research group focused on the use of a right
circular cylinder with dimensions suitable for simulating the trunk of an
adult human (30 cm diameter with a 60 cm height). This phantom was sub-
divided into a large number of regions and energy deposition was scored
in each of these regions. Both external and internal sources of radiation
were considered and absorbed dose to a specific organ was derived from
consideration of the dose to the sub-regions in which the organ was assumed
to reside.
Historical developments 111
5.2.3. The Snyder–Fisher phantom
In 1966, Fisher and Snyder, at the Oak Ridge National Laboratory (ORNL),
reported on the development of an adult phantom for use in radiation
absorbed dose calculations [8]. The adult phantom was assumed to be
standing erect with the arms at the sides of the body. Three specific regions
were defined: 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 modelled as a larger elliptical
cylinder, 20 cm 40 cm with a height of 70 cm. The legs below the buttocks
were modelled as a truncated elliptical cone with a height 80 cm. Regions of
little dosimetric concern 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 define the locations of specific organs in the phantom. Approxi-
mately 120 sub-regions were defined in the phantom that 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.
Also included in this research was the development of phantoms for
humans of other ages. These phantoms represented humans with ages 0
(newborn), 1, 5, 10 and 15 years. These phantoms will be discussed in
more detail in section 5.2.6 below.
In 1967, Fisher and Snyder reported on the development of an adult
phantom with 22 internal organs and more than 100 sub-regions [9]. This
phantom represented the next step in the development of anthropomorphic
phantoms for use in dose calculations. However, this phantom also was
homogeneous and the skeleton and lung regions were not defined—in
terms of the assignment of different densities and elemental compositions
for these regions. 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 skeletal
volume. 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. [10, 11] but were about 50% lower than those
obtained using the ICRP spherical organ methods [6].
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 -rays [12]. 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 centre of the phantom.
A heterogeneous phantom was the next logical step in the development

of anthropomorphic phantoms. Snyder and his colleagues [13] published an
extensive compilation of calculations of the absorbed fractions of energy for
monoenergetic photon 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 [14]. This phantom
ultimately became known in the nuclear medicine community as the ‘MIRD
Phantom’ (see section 5.2.4).
5.2.4. The MIRD-5 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 23 [14] 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 International Commission on Radiological Protection
(ICRP). These estimates were later published as the ICRP Publication 30
series [15].
The heterogeneous phantom contained three principal idealizations:
1. It had simple geometrical shapes that approximated the organ and body
shapes and dimensions. 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 of the organs (i.e., the source organ).
3. The Monte Carlo method was applied in the calculations to obtain
a set of photon histories upon which estimates of the energy
deposited in the organs and tissues of the phantom could be
based. 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 to 50 000 histories) and, for low energy photons, the absorbed
fractions of energy for many organs were considered 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 [16]. 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 the following.
1. The head section was represented by a right elliptical cylinder topped by
half an 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 modelled as
organs with walls and contents.
6. A source and target region called the ‘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 hetero-
geneous phantom have been made (see section 5.3 below). However, the
fundamental use of the phantom for internal dose calculations has been
essentially unchanged since its inception.
5.2.5. Photon and electron transport

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 the ALGAM code developed at ORNL [17]. 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 photon transport 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 inter-
action [18, 19]. Subsequently investigations using a coupled electron/photon
transport code indicated that for most situations, this apparent deficiency
was not significant [20, 21].
5.2.6. Similitude and paediatric phantoms
Development of the adult human phantom by Snyder and his colleagues (see
section 5.2.3 above) was paralleled by the development of phantoms represent-
ing humans of other ages [8]. These phantoms represented children with ages
of 0 (newborn), 1, 5, 10 and 15 years. 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 ‘paediatric phantoms’ were obtained
by applying a set of simple transformations to the axes of the Cartesian
coordinate system in which the adult phantom was located. These phantoms
became known as the ‘similitude phantoms’ because of their resemblance to
the adult. This approach has its limitations because children generally are
not just ‘little adults’. However, these phantoms were the first developed to
answer a real need in the nuclear medicine community [22].
Improvements in the paediatric models were closely linked with the
development of the heterogeneous adult phantom. Even though these new
phantoms were heterogeneous, the paediatric phantoms were obtained
through the same transformation method. That is, these phantoms also
were similitude phantoms. The outside dimensions were obtained by apply-
ing 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 [23].
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 paediatric 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 modelled after the adult, were designed independently.
Three ‘individual phantoms’ were designed by Hwang et al. [24, 25].
These were the newborn, 1- and 5-year-old models. A separate effort was
undertaken by Jones et al. [26] for the 15-year-old and Deus and Poston
[27] 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 [28, 29].
The current stylized models 115
5.2.7. Development of MIRDOSE codes

Internal dose calculations using a stylized model generally are performed by
combining tabulated dose conversion factors (e.g., absorbed fractions, S
values) with radionuclide activity information in source organs or regions
[30]. Because this process can be tedious to perform and involves repetitive
tasks, it lends itself to treatment with computer programs. The MIRDOSE
computer program was developed to perform internal dose calculations
using specific stylized models [31, 32]. The software was originally presented
in 1984 by Watson and Stabin [31] and has been continuously supported and
updated since then, with version 2 available in 1987, and version 3 in 1994
[32]. Using MIRDOSE, dose calculations are performed by searching
through internal databases of radionuclide decay data and specific absorbed
fractions of energy (specific for a stylized model) to generate S values of
interest. These S values are combined with user-entered source organ
residence times to calculate the dose in the different organs and regions of
the body. Libraries used in the version 2 of the MIRDOSE computer
program were based on 59 radionuclides derived from references [33–35]
and on the paediatric phantom series of Cristy and Eckerman [28, 29]. In
version 3, libraries were updated to use 224 radionuclides [33] for 10 mathe-
matical phantoms (the six ORNL phantom series [28, 29] and the four
pregnant female phantom series at different gestation times [36]) with 28
source and 28 target regions. In addition, special models for the gastro-
intestinal tract [15] and the bladder [37] were implemented in both
MIRDOSE 2 and 3. To replace the ICRP Publication 30 bone model [15]
used in MIRDOSE2 for skeletal dosimetry, the new model of the skeletal
system developed by Eckerman [38] was implemented in MIRDOSE3, thus
allowing for the calculation of doses in 15 different bone sites. To support
tumour dosimetry, the ability to calculate absorbed doses to small, unit
density spheres was added to MIRDOSE3 using absorbed fractions of
energy for both photon and electron sources [7, 39–41].
5.3. THE CURRENT STYLIZED MODELS
5.3.1. The ORNL phantom series

The Oak Ridge National Laboratory (ORNL) in 1980 developed a new series
of stylized models of various ages [28]. The series included an adult male, a
newborn and individuals of ages 1, 5, 10 and 15 years developed from anthro-
pological data (legs, trunk and head) and from age-specific organ masses
published in ICRP Publication 23 [14]. 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
paediatric dosimetry over the similitude paediatric phantoms. These phantoms
also 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 [42, 43], and a new model of the thyroid. While the ORNL paediatric
model series was initially published in 1981 [28], these models were not readily
utilized until 1987 with the publication of ORNL/TM-8381 [29]. In this
report, the only major change in the phantom series was that the 15-year-
old model was assumed also to represent the average adult female. For this
purpose, the breast, uterus and ovaries were modified according to published
reference average values [14]. The phantoms were used with the ETRAN
Monte Carlo photon transport code [44, 45] to calculate specific absorbed
fractions of energy in all five paediatric phantoms, as well as in the adult
male, for 12 photon energies (0.01 to 4.0 MeV). Electron transport was not
considered in these simulations and the electron energy was assumed to be
locally deposited.
5.3.2. The MIRD stylized models

Since the publication of the stylized dosimetric model of Snyder et al. in
MIRD Pamphlet 5 Revised [16], 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 brain [46] representing average reference brain and head for
a newborn, 1-, 5-, 10- and 15-year-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 radio-
pharmaceuticals [47, 48]. 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 sub-regions 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 masses [14, 49] and shapes from analysis of MRI
images. Twenty-one sub-regions were modelled within the head and neck,
including five regions representing bony structures (simulated as a homoge-
nized 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 sub-regions
were delineated: the caudate nuclei, cerebellum, cerebral cortex, lateral
ventricles, lentiform nuclei (a composite region of both the putamen and
the globus pallidus), thalami, third ventricle, and 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 skin.
In both the ORNL phantoms of Cristy and Eckerman [28, 29] and in the
MIRD adult model developed by Snyder [16] 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 [50, 51]. Snyder and
Ford [52] designed a dosimetric model of the human kidney to study the
administration of the mercurial diuretic chlormerodrin (neohydrin) labelled
with 197 Hg and 203 Hg. 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 197 Hg and 203 Hg in the cortex, in the medulla,
in the bladder and in the total body (blood).
McAfee published in 1970 a multi-region kidney model of an adult [53]
representing the renal cortex and medulla as two concentric elliptical shells,
and the renal pelvis as a wedge-shaped hollow structure at the centre 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 197 Hg- and 203 Hg-labelled clormerodrin [54].
Patel described a multi-compartment kidney model in 1988 [55]. This
model, which was similar to the model used by Snyder and Ford [52],
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 code [17, 56] and absorbed
fractions and specific absorbed fractions of energy were calculated for 12
monoenergetic photon sources in the range 0.01 to 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 32 P,
51
Cr, 57 Co, 67 Ga, 99m Tc, 111 In, 123 I, 131 I, 127 Xe, 133 Xe and 201 Tl.
A new kidney model recently has been adopted by the MIRD Committee
and will be published as MIRD Pamphlet No 19 [57]. Following the increased
use of radiopharmaceuticals in therapeutic nuclear medicine and recent
associated kidney toxicity [58], the MIRD Committee developed six advanced
stylized models of the kidney (newborn, 1-, 5-, 10- and 15-year-old 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 defined
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 [53].
Each region was derived from dimensions and volumes given in ICRP Publi-
cation 23 [14] for both the newborn and the adult, and assumed constant
volume ratios between the different kidney sub-regions for the other ages. In
these models each medullary pyramid was modelled 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 Pamphlets No 15 and 19, the EGS4 Monte Carlo trans-
port code [18, 59] 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 [30],
mean absorbed doses to the target regions per unit cumulated activity in
the source region (S values) were tabulated for specific radionuclides.
5.3.3. 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 were misrepresented 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 modelling these organs, leading
either to the organ not being represented, or to its being over-simplified.
Development of new radio-agents 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 [60]. This new organ was
modelled as a single sphere located just below the urinary bladder, with a
volume consistent with Reference Man of ICRP Publication 23 [14]. 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 incorrectly modelled in the
ORNL phantom series. They developed a new model of the lower part of the
colon [61], and separated the rectum from the lower large intestine. This new
model included an explicitly defined rectum, an anatomically corrected
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 [14].
Because the intestinal tract and curvatures cannot be modelled with
simple geometrical shapes that can be easily coupled to a Monte Carlo trans-
port 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 mucosal layer. In 1994,
Poston et al. developed a revision of the GI tract of the MIRD adult phan-
tom to better represent these sensitive targets [20, 21]. The actual wall of the
GI tract was divided in its thickness into four regions of varying radiosensi-
tivities; 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 ten small, concentric layers (100 mm
thick for the adult) and the dose to each layer was recorded separately.
Then, the determination of the mucosal layer thickness for each section
would give directly the dose to the mucosa. However, since it is not possible
to determine directly the mucosal 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
mucosal layer along the GI tract from anatomic slides of cross sections of
a human GI tract. Different mucosal 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 [18].
Stubbs et al. [62] 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 for four radionuclides important to nuclear
medicine. They presented S values (in units of Gy/Bq s) for 90 Y, 99m Tc, 123 I
and 131 I. They concluded, as did Poston et al., that the ICRP ‘one-half’
assumption was overly conservative for non-penetrating 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 [63]. Results were obtained using this
model for monoenergetic electrons in the range 10 to 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
MCNP4A [19] and a new model of the small intestine 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.
5.3.4. Other stylized models of the human anatomy

Other modifications and additions to the ORNL and MIRD stylized models
include a peritoneal cavity [64], a new model of the nasal cavity and major
airway [65], and a new model of the long bones [66]. The peritoneal cavity
model was developed in 1989 by Watson et al. [64]. 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 15 O-
labelled gas PET imaging agents [65]. 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 [66].
This effort was undertaken 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.
5.3.5. Use of stylized models in therapeutic nuclear medicine

Because stylized models are based on an average individual (Reference Man),
associated dose calculations can only represent the average dose to a
population. Although this average dose can sometimes be applied to specific
individuals, large variations in patient weight, size and organ mass are
expected, and can significantly change the dose received by an organ. Con-
sequently, Reference Man S values have to be scaled to match a specific
individual. In MIRD Pamphlet 11, Snyder et al. discussed methods of scaling
radionuclide S values from the 70 kg Reference Man to individual patients
[67]. For target organs, rT , sufficiently distant from the source organ, rS ,
values of specific absorbed fraction ðrT rS Þ for photon irradiation are
reasonably independent of organ mass, and thus no scaling from Reference
Man to the patient is required. In these cases, increases (or decreases) in
target organ mass, mT , are compensated (scaled) by proportional increases
(or decreases) in the absorbed fraction ðrT rS Þ. As the target organ
approaches the source organ, however, this compensation no longer exists
for photons. When rT and rS coincide (the target organ becomes the source
organ), the dose from photon emitters above 100 keV becomes proportional
2=3
to mS . For electron sources, in which full energy absorption is typically
Tomographic models 121
assumed, the dose from electrons scales as the inverse first power of the organ
mass (i.e., m1S ).
To better match a stylized model to a specific patient, models of
different sizes also can be used as proposed by Clairand et al. [68]. Using
the MIRD stylized model, they developed six new models of different
heights: three male models of heights 160, 170 and 180 cm and three
female models of heights 150, 160 and 170 cm. Using a statistical analysis
of anthropometric data gathered from 684 forensic autopsy cases for the
years 1987–1991 (all subjects selected were Caucasian adults who died of
external causes and showed no pathological changes), they derived organ
masses as a function of height [69]. Only organs weighed during autopsy
were taken into account in this study (heart, kidneys, liver, lungs, pancreas,
spleen, testes and thyroid), and other organs were obtained from the ORNL
models.
To support the dosimetry in therapeutic nuclear medicine, simple
tumour models also have been introduced in stylized phantoms [70–74].
These approaches allow for the determination of the photon dose from a
tumour to adjacent organs, but also provide the photon dose from body
organs to the tumour, which for some radionuclides can be non-negligible
[75]. Clairand et al. developed a computer program (DOSE3D) to simulate
spherical tumours inside the ORNL adult phantom [70]. With this program,
the user can select one or more tumours of different sizes, and the EGS4
Monte Carlo transport code is used for the particle transport. Similarly,
Johnson et al. developed the MABDOSE software to simulate spherical
tumours inside the ORNL phantom series [71–74]. In this code, the phantom
series is first voxelized at a user-selectable resolution (10, 5, 2, or 1 mm), and
the ETRAN Monte Carlo photon transport code used for the particle
transport.
5.4. TOMOGRAPHIC MODELS
As discussed above, stylized mathematical models of human anatomy

provide a relatively efficient geometry for use with Monte Carlo radiation
transport codes. Nevertheless, these models are only approximations to the
true anatomical features of individuals for which dose estimates are required.
An alternative class of anatomic models is based upon three-dimensional
imaging techniques such as magnetic resonance (MR) imaging or computed
tomography (CT). These tomographic models of human anatomy represent
large arrays of image voxels that are individually assigned both a tissue type
(e.g., soft tissue, bone, air, etc.) and an organ identity (heart wall, femur
shaft, tracheal airway, etc.). Thus, image segmentation is needed to process
the original image into a format acceptable for radiation transport using
codes such as EGSnrc or MNCP [19, 76].
5.4.1. Methods of construction
Two general anatomic sources exist for the construction of tomographic

models of human anatomy: (1) cadavers and (2) live subjects, typically medical
patients. Each data source has its distinct advantages and disadvantages.
Cadaver imaging generally offers a substantially improved opportunity for
full anatomic coverage including the extremities. With CT image acquisition,
higher-resolution scans can be performed, as radiation dose considerations
and patient motion are not of concern. For these same reasons, multiple
scans on the same cadaver can be performed using different technique factors
(kVp, mA, filtration, etc.). Problems associated with cadaver imaging include
tissue shrinkage, body fluid pooling, air introduction to the gastrointestinal
tract, collapse of the lungs, and general organ settling. Perhaps the greatest
disadvantage, however, is that cadaver imaging will most likely not involve
the use of CT tissue contrast agents needed for soft-tissue image segmentation.
Computed tomography is perhaps the imaging modality of choice for
construction of full-body tomographic computational models. Skeletal
tissues are more readily defined under CT imaging, whereas image distor-
tions of skeletal tissues are problematic under MR imaging. With live patient
imaging, CT contrast agents provide acceptable definitions of soft tissue
borders needed for image segmentation, and thus MR does not offer a
distinct advantage over CT for this data source. In cadaver imaging, the
lack of contrast agents hinders border definitions for soft-tissue organs. To
compensate, MR imaging of the cadaver might be considered. However,
the frozen state of the cadaver tissues can alter both spin–lattice and spin–
spin relaxation times (T1 and T2, respectively), complicating and distorting
MR images of the subject. Thus, CT imaging is recommended for both live
and deceased subjects upon whom tomographic models are to be
constructed. With continuing advances in multi-detector, multi-slice CT
imaging, scan times are minimized, offering improved opportunities for
live-subject imaging as potential data sources. CT–MR image fusion is
another possibility for input data to image segmentation. However, patient
motion artefacts within MR images, and their generally lower image resolu-
tions, currently restrict opportunities for CT–MR co-registration to the head
region where patient motion is more easily minimized.
Following image acquisition, the next step in model construction is image
segmentation. In cases where the CT image provides strong tissue contrast,
automated pixel-growing methods of image segmentation can be applied to
rapidly delineate organ boundaries. In these methods, a central pixel is
tagged and the algorithm groups all neighbouring pixels within a defined
grey-level interval (e.g., skeletal regions, air lumen of nasal passages, certain
soft-tissue organs under CT contrast). For those organs with poor tissue
contrast, organ boundaries must be segmented manually. Final review of
the tomographic model by medical personnel trained in radiographic anatomy
is highly recommended. Standardized software packages, such as IDL

(Research Systems, Boulder, CO), can offer all necessary tools for construction
of tomographic computational models.
5.4.2. Review of representative models

Several research groups have been involved in the development of tomo-
graphic computational models over the past decade. Most of the efforts
have been directed towards adult models, but a few important studies have
resulted in paediatric tomographic models for radiation dosimetry studies.
In 1994, Zubal and colleagues published their work in manually
segmenting 129 CT transverse slices of a living male adult [77]. Their goal
was to create a tomographic computational model for use in simulating
nuclear medicine imaging data [78]. A total of 36 tissue regions were
identified within the three-dimensional voxel array. The original CT
images were obtained in two imaging sessions: one of 78 slices from the
neck to mid-thigh with 10 mm slice thicknesses (pixel size of 1 mm), and
one of 51 slices from the head to the neck region with 5 mm slices (pixel
size of 0.5 mm). Image processing techniques were applied to achieve a
final isotropic cubic voxel resolution of 4 mm on a side. Similar techniques
also have been applied in the creation of a high-resolution, MRI-based
voxel head model [79]. Stabin et al. have coupled the Zubal total-body
model to the MCNP radiation transport code [80]. Yoriyaz et al. have
reported estimates of cross-organ photon dose based on MCNP calculations
within the Zubal model [81, 82]. Tagesson et al. used the segmented Zubal
head model to estimate absorbed fractions and radionuclide S values for a
variety of radiopharmaceuticals used in neuroimaging [83].
In 1997, Jones reported research performed at the NRPB on a voxel-
based computational model of the adult (NORMAN) [84]. This model was
used to estimate organ doses from external photon sources over a range of
energies and irradiation geometries. When comparing the results with
those obtained using a MIRD-type stylized model, differences in organ
doses were found to range from a few per cent to over 100% at photon
energies between 10 and 100 keV; differences were less than 20% at photon
energies exceeding 100 keV. A subsequent study found that estimates of
organ doses from internal photon sources differed widely as assessed using
two different models, and differences in cross-organ doses generally increased
with decreasing photon energy [85].
The National Library of Medicine’s Visible Human Project also
provides an excellent data source from which one may construct tomo-
graphic computational models for use in diagnostic X-ray simulations for
adult patients. This data set consists of MRI, CT and anatomical images.
The CT data consist of axial CT scans of the entire body taken at 1 mm
intervals at a resolution of 512 512 pixels. Xu and his students have
recently reported on their efforts to merge this data set with both the MCNP
and EGS4 radiation transport codes [86]. Computational results have been
reported for internal electron sources [87], external photon sources [88],
neutrons [89, 90] and electrons [91].
At present, three tomographic computational models have been
constructed for specific use in paediatric dosimetry studies [92, 93]. Two of
these are the BABY and CHILD models developed at GSF in Germany
by Zankl and colleagues. BABY is constructed from a whole-body CT
exam of an 8-week-old female 24 h after death. CHILD was constructed
from a CT exam of a live 7-year-old female who was undergoing subsequent
whole-body irradiation prior to bone marrow transplantation. Image
segmentation was performed which assigned voxels to 54 regions in the
baby and 64 regions in the child. The models have been used to determine
paediatric organ doses in a variety of medical exposure scenarios including
nuclear medicine [94], projection radiography [95–98] and computed tomo-
graphy [99, 100]. The third paediatric model, ADELAIDE, is based upon
44 consecutive CT scans of a 14-year-old female [101]. This particular
model is restricted to the torso only, and has been used with the EGS4
transport code to assess organ doses received during CT exams.
Recent efforts at GSF have resulted in a new adult reference male
tomographic model, GOLEM. This model was segmented from whole-
body medical images of a living 38-year-old male with external dimensions
close to those of the ICRP Reference Man [102–105]. Recent German–
Japanese collaborations also have been undertaken to establish a reference
tomographic computational model for the adult Asian male [106].
5.4.3. Comparisons with stylized mathematical models

The obvious difference between tomographic and stylized mathematical
models is the improved geometric descriptions of organ position, organ
shape and body contour offered by the former. In estimating internal
organ absorbed dose, however, only the absorbed fraction of energy and
the target organ mass are of consequence. When the tomographic computa-
tional model results in estimates of the absorbed fractions and organ masses
that are essentially equivalent to those in the stylized model, any improve-
ment in anatomic description is of no consequence. When differences in
dosimetry are noted, they must be considered in regard to the development
of each model. As noted above, stylized models are generally constructed
based upon estimates of average organ size and position as documented in
ICRP Publication 23 for Reference Man or Reference Woman [14]. With a
tomographic model, organ positions and locations can be very different
inasmuch as the individual does not represent the average characteristics
of reference subject populations. The tomographic model is highly patient
specific to the individual that was imaged. When applied to other patients,
however, dosimetry data gathered from a single tomographic model could be

less desirable than that obtained from a reference-stylized model, particularly
when the current patient differs substantially from the tomographic model
subject. Ideally, an entire series of whole-body tomographic computational
models would be needed. Patient-specific estimates of internal organ doses
could be made through selecting one of an array of tomographic models
that best describes the current patient of interest.
Early attempts at the development of tomographic models were severely
hindered by data storage and computer memory restrictions during radiation
transport. At present, advances in computer hardware and software have
all but eliminated these issues, except in extreme cases. Nevertheless, certain
limitations of tomographic models over stylized models continue to exist that
are of particular note. Voxel dimensions within tomographic models are
initially restricted to those used in acquiring the image. Image slice
interpolation algorithms subsequently may be applied to reduce voxel
dimensions, although residual errors may persist as a result of partial-
volume averaging within the original image data set. Small anatomic
structures such as organ epithelial layers and GI mucosa may be missed
altogether within the segmented regions of the tomographic model. For
these structures, the simplicity of geometric representation within the stylized
mathematical models is a distinct advantage. Peter et al. have recently
reviewed problems related to discretization errors within tomographic
computational models used in simulation studies of radiographic imaging
[107].
One organ of particular importance to radionuclide therapy is the
skeletal system. This anatomic region is particularly difficult to construct
using stylized geometrical descriptions, and thus tomographic computational
models offer a distinct advantage in providing macroscopic details of bone
site contours, shapes and locations. Nevertheless, voxel dimensions are
much larger than microscopic tissue sites of interest (marrow cavities, bone
trabeculae, endosteal tissues, etc.). Consequently, microimaging technologies
applied to ex vivo samples are necessary to produce usable images for
charged-particle transport within trabecular regions of the skeleton [108].
Final assessment of marrow and endosteal tissue doses from photon sources
external to the skeleton can be made via coupling of the photon transport
within the whole-body tomographic computational model, while transport
of photon-produced secondary electrons will invoke microscopic voxelized
geometries of trabecular bone regions. Eckerman et al. have used this
dual-transport technique for skeletal dosimetry utilizing (1) chord-based
models of electron transport in trabecular bone, and (2) the ORNL-series
of stylized models for photon transport [38]. Although no studies have
been published to date, this overall dual-transport methodology can be
extended to the use of macroscopic and microscopic tomographic models
of the skeletal system. These concepts are discussed further in chapter 13.
5.5. SUMMARY
To some extent, mathematical formulations of organs and tissues of the body

have been used in the dosimetry of internally distributed radionuclides for
nearly 60 years. The concepts are not new ones. However, over this time,
the sophistication of these ‘phantoms’ has 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. The ability to model the entire body of a ‘Reference Man’ and to
specify the location, shape, volume and mass of organs in the body was
the next logical step. These developments were followed almost immediately
by the development of mathematical models for humans of other ages as well
as attempts to model specific regions in important organs such as the kidneys
and brain. Currently, anatomic models, based upon three-dimensional
imaging techniques such as magnetic resonance imaging or computed
tomography, are being widely developed and used for dosimetry. These
techniques lend themselves to the design and specification of mathematical
models representing the individual patient and, in addition, allow the defini-
tion of structures and/or regions in organs so that the distribution of the
radionuclide can be better represented and the radiation-absorbed dose can
be assessed more accurately.
REFERENCES
[1] NCRP 1985 The experimental basis for absorbed-dose calculations in medical uses of
radionuclides NCRP Report No 83 (Bethesda, MD: National Council on Radiation
Protection and Measurements)
[2] Marinelli L D 1942 Dosage determination with radioactive isotopes Am. J.
Roentgenol. Rad. Therapy 47 210–16
[3] Marinelli L D, Quimby J H and Hine G J 1948 Dosage determination with
radioactive isotopes II. Practical considerations in therapy and protection Am. J.
Roentgenol. Rad. Therapy 59 260–81
[4] Loevinger R, Japha E M and Brownell G L 1965 Discrete radioisotope sources.
In Radiation Dosimetry eds Hine G J and Brownell G L (New York: Academic)
pp 693–799
[5] Loevinger R, Hiolt J G and Hine G J 1965 Internally administered radionuclides.
In Radiation Dosimetry eds Hine G J and Brownell G L (New York: Academic)
pp 801–73
[6] ICRP 1959 Report of committee II on permissible dose for internal radiation
International Commission on Radiological Protection (Oxford: Pergamon)
[7] Brownell G L, Ellett W H and Reddy A R 1968 MIRD Pamphlet No 3: Absorbed
fractions for photon dosimetry J. Nucl. Med. Suppl 29–39
[8] Fisher H L J and Snyder W S 1966 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) pp 221–28
References 127
[9] Fisher H L J and Snyder W S 1967 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) pp 24557
[10] Ellet W H, Callahan A B and Brownell G H 1964 Gamma-ray dosimetry of internal
emitters: Monte Carlo calculations of absorbed dose from point sources Brit. J.
Radiol. 37 4552
[11] Ellet W H, Callahan A B and Brownell G H 1965 Gamma-ray dosimetry of internal
emitters II: Monte Carlo calculations of absorbed dose from uniform sources Brit. J.
Radiol. 38 54144
[12] Snyder WS 1967 The variation of dose in man from exposure to a point source of
gamma rays ORNL-4168 Oak Ridge National Laboratory Oak Ridge, TN 25761
[13] Snyder W S, Ford M R, Warner G G and Fisher Jr H L 1969 MIRD Pamphlet No 5:
Estimates of absorbed fractions for monoenergetic photon sources uniformly
distributed in various organs of a heterogeneous phantom J. Nucl. Med. 10 152
[14] ICRP 1975 Report on the task group on Reference Man ICRP Publication 23
(Oxford: International Commission on Radiological Protection)
[15] ICRP 1979 Limits for intakes of radionuclides by workers Publication 30.
[16] Snyder W S, Ford M R and Warner G G 1978 MIRD Pamphlet No 5, Revised:
Estimates of specific absorbed fractions for photon sources uniformly distributed in
various organs of a heterogeneous phantom (New York: Society of Nuclear Medicine)
[17] Warner G G and Craig A N J 1968 ALGAM: A computer program for estimating
internal dose in a man phantom ORNL-TM-2250 (Oak Ridge, TN: Oak Ridge
National Laboratory)
[18] Nelson W R, Hirayama R H and Roger D W O 1985 The EGS4 code system SLAC
Report 265 (Stanford, CA: Stanford Linear Accelerator Center)
[19] Briesmeister J F 1993 MCNP—a general Monte Carlo N-particle transport code,
version 4A LA-12625-M (Los Alamos, NM: Los Alamos National Laboratory)
[20] Poston J W Jr, Kodimer K A, Bolch W E and Poston J W Sr 1996 Calculation of
absorbed energy in the gastrointestinal tract Health Phys. 71 3006
[21] Poston J W Jr, Kodimer K A, Bolch W E and Poston J W Sr 1996 A revised model
for the calculation of absorbed energy in the gastrointestinal tract Health Phys. 71
30714
[22] Kereiakes J G, Seltzer R A, Blackburn B and Saenger E L 1965 Doses to infants and
children: a plea for a standard child Health Phys. 11 9991004
[23] Poston J W 1990 The development of early pediatric models and their application to
radiation absorbed dose calculations. In Dosimetry of Administered Radionuclides
eds S J Adelstein, A I Kassis and R.W.B) (Washington, DC: American College of
Nuclear Physicians) pp 10521
[24] Hwang J M L, Shoup R L, Warner G G and Poston J W 1976 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)
[25] Hwang J M L, Shoup R L and Poston J W 1976 Mathematical description of a
newborn human for use in dosimetry calculations ORNL/TM-5453 (Oak Ridge,
TN: Oak Ridge National Laboratory)
[26] Jones R M, Poston J W, Hwang J L, Jones T D and Warner G G 1976 The develop-
ment and use of a fifteen-year-old equivalent mathematical phantom for internal dose
calculations ORNL/TM-5278 (Oak Ridge, TN: Oak Ridge National Laboratory)
[27] Deus S F and Poston J W 1976 The development of a mathematical phantom

representing a 10-year-old for use in internal dosimetry calculations. In Proceedings
of the Symposium on Radiopharmaceutical Dosimetry HEW Publication (FDA) 76-
8044 Oak Ridge, TN
[28] Cristy M 1980 Mathematical phantoms representing children of various ages for use in
estimates of internal dose ORNL/NUREG/TM-367 (Oak Ridge, TN: Oak Ridge
[29] Cristy M and Eckerman KF 1987 Specific absorbed fractions of energy at various ages
from internal photon sources ORNL/TM-8381 (Oak Ridge, TN: Oak Ridge National
Laboratory)
[30] Loevinger R, Budinger T F and Watson E E 1991 MIRD Primer for Absorbed Dose
Calculations (New York: The Society of Nuclear Medicine) p 128
[31] Watson E E and Stabin M 1984 Basic alternative software package for internal
radiation dose calculations. In 17th Midyear Topical Symposium of the Health
Physics Society eds R L Kathren, D P Higby and M A McKinney (Richland, WA)
in nuclear medicine J. Nucl. Med. 37 53846
[33] Weber D A, Eckerman K F, Dillman L T and Ryman J C 1989 MIRD: Radionuclide
Data and Decay Schemes (New York: Society of Nuclear Medicine)
[34] ICRP 1983 Radionuclide transformations Publication 38 International Commission
on Radiological Protection (Oxford: Pergamon)
[35] Kocher D 1981 Decay data tables DOE TIC-11026 (Oak Ridge, TN: Oak Ridge
[36] Stabin M, Cristy M and Watson E E 1995 Mathematical models of the adult female at
various stages of pregnancy ORNL/TM-12907 (Oak Ridge, TN: Oak Ridge National
Laboratory)
[37] Cloutier R J, Smith S A, Watson E E, Snyder W S and Warner G G 1973 Dose to the
fetus from radionuclides in the bladder Health Phys. 25 14761
[38] Eckerman K F 1985 Aspects of the dosimetry of radionuclides within the skeleton
with particular emphasis on the active marrow. In Proceedings of the Fourth
International Radiopharmaceutical Dosimetry Symposium eds A T Schlafke-Stelson
and E E Watson (Oak Ridge, TN: ORAU) pp 51434
[39] Ellett W H and Humes R M 1971 MIRD Pamphlet No 8: Absorbed fractions for
small volumes containing photon-emitting radioactivity J. Nucl. Med. 2532
small spheres J. Nucl. Med. 29 803
spheres of various sizes J. Nucl. Med. 35 1526
[42] Coffey J L and Watson E E 1981 S-values for selected radionuclides and organs with
the heart wall and heart contents as source organs. In Third International Radiophar-
maceutical Dosimetry Symposium eds E E Watson, A T Schlafke-Stelson, J L Coffey
and R J Cloutier (Rockville, MD: US Department of Health & Human Services)
pp 56394
[43] Coffey J L 1978 A revised mathematical model of the heart for use in radiation
absorbed dose calculation University of Tennessee, Knoxville, TN
[44] Seltzer S M 1988 An overview of ETRAN Monte Carlo Methods. In Monte Carlo
Transport of Electrons and Photons eds T M Jenkins, W R Nelson, A Rindi, A E
Nahum and D W O Rogers (New York: Plenum) pp 15381
References 129
[45] Seltzer S M 1991 Electronphoton Monte Carlo calculations: the ETRAN code
Appl. Radiat. Isot. 42 91741
[46] Bouchet L G, Bolch W E, Weber D A, Atkins H L and Poston J W Sr 1999 MIRD
head and brain J. Nucl. Med. 40 62S101S
[47] London E D 1993 Imaging Drug Action in the Brain (Boca Raton, FL: CRC)
[48] van Heertum R L and Tikofsky R S 1995 Cerebral SPECT Imaging (New York:
Raven)
[49] Blinkov S M and Glezer I I 1968 The Human Brain in Figures and Tables (New York:
Plenum)
[50] Petegnief Y, Petiet A, Peker M C, Bonnin F, Meulemans A and Le Guludec D 1998
Quantitative autoradiography using a radioimager based on a multiwire propor-
tional chamber Phys. Med. Biol. 43 362938
[51] Willis K W, Martinez D A, Hedley-Whyte E T, Davis M A, Judy P F and Treves S
1977 Renal localization of 99mTc-stannous glucophetonate and 99mTc-stannous
dimercaptosuccinate in the rat by frozen section autoradiography. The efficiency
and resolution of technetium-99m Radiat. Res. 69 47588
[52] Snyder W S and Ford M R 1967 A dosimetric study for the administration of
neohydrin labeled with 203Hg and 197Hg ORNL-4168 (Oak Ridge, TN: Oak Ridge
National Laboratory) pp 26773
[53] McAfee J G 1969 Problems in evaluating the radiation dose for radionuclides
excreted by the kidneys. In Medical Radionuclides: Radiation Dose and Effects eds
R J Cloutier, C L Edwards and W S Snyder (Oak Ridge, TN: US Atomic Energy
Commission) pp 27194
[54] Blau M, McAfee J G, Roher R H, Snyder W S and Smith E M 1975 MIRD dose
estimate No. 6: Hg-197- and Hg-203-labeled chlormerodrin J. Nucl. Med. 16 121417
[55] Patel J S 1988 A revised model of the kidney for medical internal radiation dose
(College Station, TX: Texas A&M University)
[56] Hui T Y E 1985 A preliminary model of the circulating blood for use in radiation
dose calculations (College Station, TX: Texas A&M University)
[57] Bouchet L G, Bolch W E, Blanco P, Rajon D A, Clairand I, Konijnnenberg M W
and Wessels B W 2002 MIRD Pamphlet No 19: Absorbed fractions of energy and
radionuclide S values for six pediatric multiregion models of the kidney (in press)
[58] Boerman O C, Oyen W J and Corstens F H 2001 Between the Scylla and Charybdis
of peptide radionuclide therapy: hitting the tumor and saving the kidney Eur. J.
Nucl. Med. 28 14479
[59] Ford R L and Nelson W R 1978 The EGS code system: computer programs for Monte
Carlo simulation of electromagnetic cascade shower Report 210 (Stanford: Stanford
Linear Accelerator Center)
[60] Stabin M 1994 A model of the prostate gland for use in internal dosimetry J. Nucl.
Med. 35 51620
[61] Mardirossian G, Tagesson M, Blanco P, Bouchet LG, Stabin M, Yoriyaz H, Baza S,
Ljungberg M, Strand S E and Brill A B 1999 A new rectal model for dosimetry
applications J. Nucl. Med. 40 152431
[62] Stubbs J B, Evans J F and Stabin M 1998 Radiation absorbed doses to the walls of
hollow organs J. Nucl. Med. 39 198995
[63] Bhuiyan N U 2000 A revised dosimetric model for calculation of electron dose in the
small intestine (College Station, TX: Texas A&M University)
[64] Watson E E, Stabin M, Davis J L and Eckerman K F 1989 A model of the peritoneal
cavity for use in internal dosimetry J. Nucl. Med. 30 200211
[65] Deloar H M, Watabe H, Nakamura T, Narita Y, Yamadera A, Fujiwara T and Itoh
M 1997 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 160313
[66] Clairand I, Bouchet L G and Bolch W E 2000 A new macroscopic model of the long
bones for skeletal dosimetry J. Nucl. Med. 41 1062
[67] Snyder W S, Ford M R, Warner G G and Watson S B 1975 MIRD Pamphlet No 11:
‘S,’ absorbed dose per unit cumulated activity for selected radionuclides and organs
(New York: Society of Nuclear Medicine)
[68] Clairand I, Bouchet L G, Ricard M, Durigon M, Di Paola M and Aubert B 2000
Improvement of internal dose calculations using mathematical models of different
adult heights Phys. Med. Biol 45 277185
[69] de la Grandmaison G L, Clairand I and Durigon M 2001 Organ weight in 684 adult
autopsies: new tables for a Caucasoid population Forensic Sci. Int. 119 14954
[70] Clairand I, Ricard M, Gouriou J, Di Paola M and Aubert B 1999 DOSE3D: EGS4
Monte Carlo code-based software for internal radionuclide dosimetry J. Nucl. Med.
40 151723
[71] Johnson T K, McClure D L, McCourt S L, Andl G J and Berman B D 1998
MABDOSE: A computer program for the calculation of dose. In Proceedings of the
Sixth International Radiopharmaceutical Dosimetry Symposium ed A T Schlafke-
Stelson (Oak Ridge, TN: ORAU)
[72] Johnson T K 1988 MABDOSE: A generalized program for internal radionuclide
dosimetry Comput. Meth. Progr. Biomed. 27 15967
general purpose dose estimation code Med. Phys. 26 138995
[74] Johnson T K, McClure D and McCourt S 1999 MABDOSE. II: Validation of a
general purpose dose estimation code Med. Phys. 26 1396403
[75] Johnson T K and Colby S B 1993 Photon contribution to tumor dose from
considerations of 131 I radiolabeled antibody uptake in liver, spleen, and whole
body Med. Phys. 20 166774
transport. I. EGSnrc, the new EGS4 version Med. Phys. 27 48598
[77] Zubal I G, Harrell C R, Smith E O, Rattner Z, Gindi G and Hoffer P B 1994
Computerized 3-dimensional segmented human anatomy Med. Phys. 21 299302
[78] Zubal I G and Harell C R 1992 Voxel based Monte Carlo calculations of nuclear
medicine images and applied variance reduction techniques Image Vision Comput.
10 34248
[79] Zubal I G, Harrell C R, Smith E O, Smith A L and Krischlunas P 1996 High
resolution, MRI-based, segmented, computerized head phantom. In Proceedings
of the Sixth International Radiopharmaceutical Dosimetry Symposium eds A T S-
Stelson, M Stabin and R B Sparks (Oak Ridge, TN: ORAU) pp 31924
[80] Stabin M, Yoriyaz H, Brill A B and Dawant M 1999 Monte Carlo calculation of
dose conversion factors for a new generation of dosimetry phantoms J. Nucl.
Med. 40 310P11P
[81] Yoriyaz H, dos Santos A, Stabin M and Cabezas R 2000 Absorbed fractions in a
voxel-based phantom calculated with the MCNP-4B code Med. Phys. 27 155562
References 131
[82] Stabin M and Yoriyaz H 2001 Photon specific absorbed fractions calculated in the
trunk of an adult male voxel-based phantom Health Phys. 82 2144
[83] Tagesson M, Zubal I G, Ljungberg M and Strand S E 1996 S values for subregions in
the brain. In Proceedings of the Sixth International Radiopharmaceutical Dosimetry
Symposium eds A T S-Stelson, M Stabin and R B Sparks (Oak Ridge, TN:
ORAU) pp 32530
[84] Jones D G 1997 A realistic anthropomorphic phantom for calculating organ doses
arising from external photon irradiation Radiat. Prot. Dosim. 72 2129
[85] Jones D G 1998 A realistic anthropomorphic phantom for calculating specific
absorbed fractions of energy deposited from internal gamma emitters Radiat.
Prot. Dosim. 79 41114
[86] Xu X G, Chao T C and Bozkurt A 2000 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 47686
[87] Chao T C and Xu X G 2001 Specific absorbed fractions from the image-based VIP-
Man body model and EGS4-VLSI Monte Carlo code: internal electron emitters
Phys. Med. Biol 46 90127
[88] Chao T C, Bozkurt A and Xu X G 2001 Conversion coefficients based on the VIP-
Man anatomical model and EGS4 Health Phys. 81 16383
[89] Bozkurt A, Chao T C and Xu X G 2000 Fluence-to-dose conversion coefficients
from monoenergetic neutrons below 20 MeV based on the VIP-man anatomical
model Phys. Med. Biol. 45 305979
[90] Bozkurt A, Chao T C and Xu X G 2001 Fluence-to-dose conversion coefficients
based on the VIP-Man anatomical model and MCNPX code for monoenergetic
neutrons above 20 MeV Health Phys. 81 184202
[91] Chao T C, Bozkurt A and Xu X G 2001 Organ dose conversion coefficients for
0.110 MeV electrons calculated for the VIP-Man tomographic model Health
Phys. 81 20314
[92] Zankl M, Veit R, Petoussi N, Mannweiler E, Wittmann A and Drexler G 1994
Realistic computerized human phantoms Life Sci. Space Res. 14 42331
[93] Zankl M, Veit R, Williams G, Schneider K, Fendel H, Petoussi N and Drexler G
1988 The construction of computer tomographic phantoms and their application
in radiology and radiation protection Radiat. Environ. Biophys. 27 153–64
[94] Petoussi-Henß N and Zankl M 1998 Voxel anthropomorphic models as a tool for
internal dosimetry Radiat. Prot. Dosim. 79 41518
[95] Petoussi-Henß N, Panzer W, Zankl M and Drexler G 1995 Dose–area product and
body doses Radiat. Prot. Dosim. 57 36366
[96] Veit R and Zankl M 1992 Influence of patient size on organ doses in diagnostic
radiology Radiat. Prot. Dosim. 43 24143
[97] Veit R and Zankl M 1993 Variation of organ doses in pediatric radiology due to
patient diameter calculated with phantoms of varying voxel size Radiat. Prot.
Dosim. 49 35356
[98] Zankl M 1993 Computational models employed for dose assessment in diagnostic
radiology Radiat. Prot. Dosim. 49 33944
[99] Kalender W A, Schmidt B, Zankl M and Schmidt M 1999 A PC program for estimating
organ dose and effective dose values in computed tomography Eur. Radiol. 9 55562
[100] Zankl M, Panzer W, Petoussi-Henß N and Drexler G 1995 Organ doses for children
from computer tomographic examinations Radiat. Prot. Dosim. 57 39396
[101] Caon M, Bibbo G and Pattison J 1999 An EGS4-ready tomographic computational

model of a 14-year-old female torso for calculating organ doses from CT examina-
tions Phys. Med. Biol. 44 221325
[102] Smith T, Petoussi-Henss N and Zankl M 2000 Comparison of internal radiation
doses estimated by MIRD and voxel techniques for a ‘family’ of phantoms Eur. J.
Nucl. Med. 27 138798
[103] Smith T J, Phipps A W, Petoussi-Henss N and Zankl M 2001 Impact on internal
doses of photon SAFs derived with the GSF adult male voxel phantom Health
Phys. 80 47785
[104] Zankl M, Panzer W and Herrmann C 2000 Calculation of patient doses using a
human voxel phantom of variable diameter Radiat. Prot. Dosim. 90 15558
[105] Zankl M and Wittmann A 2001 The adult male voxel model ‘Golem’ segmented
from whole-body CT patient data Radiat. Environ. Biophys. 40 15362
[106] Saito K, Wittmann A, Koga S, Ida Y, Kamei T, Funabiki J and Zankl M 2001
Construction of a computed tomographic phantom for a Japanese male adult and
dose calculation system Radiat. Environ. Biophys. 40 6975
[107] Peter J, Tornai M P and Jaszczek R J 2000 Analytical versus voxelized phantom
representation for Monte Carlo simulation in radiological imaging IEEE Trans.
Med. Imaging 19 55664
[108] Jokisch D W, Bouchet L G, Patton P W, Rajon D A and Bolch W E 2001 Beta-
particle dosimetry of the trabecular skeleton using Monte Carlo transport within
3D digital images Med. Phys. 28 150518
Chapter 6
Monte Carlo codes for use in therapeutic

nuclear medicine
Michael Stabin and Habib Zaidi
6.1. INTRODUCTION
The Monte Carlo method is widely used for solving problems involving
statistical processes and is very useful in medical physics due to the stochastic
nature of radiation emission, transport and detection processes. The method
is very useful for complex problems that cannot be modelled by computer
codes using deterministic methods or when experimental measurements
may be impractical. However, due to computer limitations, the method has
not yet fully lived up to its potential. With the advent of high-speed super-
computers, the field has received increased attention, particularly with
parallel algorithms and high performance platforms, which have much
higher execution rates. Recent advances in computer hardware and software
have lowered, if not removed, the barriers to wider acceptance of Monte
Carlo calculations by life scientists.
Many Monte Carlo programs have been in use in the field of nuclear
imaging, internal dosimetry and treatment planning with many of them
available in the public domain [1, 2]. Those Monte Carlo software
packages are excellent research tools; however, becoming familiar with
the codes can be time consuming, and most codes are not completely
free of bugs. Moreover, the simulation package must be integrated in
the local software environment. For that purpose, modifications may be
necessary, which require a detailed working knowledge of the code struc-
ture. This chapter summarizes state-of the-art developments of widely used
Monte Carlo codes in therapeutic nuclear medicine in connection with
computing facilities available for routine and research applications.
Current trends and some strategies for future development in the field
will also be discussed.
133
134 Monte Carlo codes for use in therapeutic nuclear medicine
6.2. HISTORICAL DEVELOPMENTS
The ‘Monte Carlo method’ describes a very broad area of science, in which
many processes, physical systems and phenomena are simulated by statistical
methods employing random numbers (see chapter 1). In the specific applica-
tion of interest here, the transport of ionizing radiation particles is simulated
by creation of particles or rays from a defined source region, generally with
random initial orientation in space, with tracking of the particles as they
travel through the system, sampling the probability density functions
(pdfs) for their interactions to evaluate their trajectories and energy deposi-
tion at different points in the system. The interactions determine the penetra-
tion and motion of particles but, more importantly, the energy deposited
during each interaction gives the radiation absorbed dose, when divided by
the appropriate values of mass. With sufficient numbers of interactions,
the mean absorbed dose at points of interest will be given with acceptable
uncertainties. The central issues include how well the real system of interest
can be simulated by a geometrical model, how many histories (i.e., how much
computer time) are needed to obtain acceptable uncertainties (usually
around 5%, no more than 10%), and how measured data can be used to
validate the theoretical calculations.
In the general sense, a problem that can be treated by a Monte Carlo
method is determined by a set of parameters that in aggregate determine
completely the behaviour of an element of the system in all situations
modelled within that system. Specifically, in applications of Monte Carlo
in therapeutic nuclear medicine, we are interested in modelling the creation
and interaction of all particles (speaking somewhat imprecisely, as photons,
which are of high interest in this situation, are best described as waves
rather than as particles) throughout their ‘lifetime’ in the body. Particles
are generally ‘created’ at random positions within a source region (which
might be an entire organ or some smaller structure), with defined initial
orientations, and then travel random distances before interacting with
atoms within the tissues of the body and depositing energy in these inter-
actions, and eventually being absorbed completely or escaping the body.
A complete characterization of the source will include its spatial distribu-
tion, the angular distribution of its emission spectrum (most often
random in 4 space, unless some unique situation exists which restricts
the orientation of the emissions—this is more common in situations
involving external sources, which may be collimated or shielded in certain
ways), and its energy distribution. Once a particle is created and begins its
journey away from the source location, the three-space location of the first
interaction must be chosen. This is done by sampling the known
probability distributions for its interactions with atoms in the body. The
probabilities of interaction for photons are very well known—photo-
electric, Compton, and pair production events occur with well defined
probabilities in materials of different atomic number. Specification of the

atomic number composition of the spatial regions involved in the simula-
tion will thus permit the characterization of the probability distributions
for interaction in any location, which can be sampled using random
number generators, as described elsewhere in this text. The process can
be computationally intensive, as ‘ray tracing’, or calculation of projection
paths to the geometric limits of defined spatial regions, can involve many
calculations. ‘Russian roulette’ routines with fictitious particles [3] were
developed to provide a shortcut method to an equivalent result. With the
development of faster and faster computers, computational speed has
become a less important issue. Groups of individual interactions are
generally treated in ‘steps’ by the computer, and application of known
interaction types and probabilities is more complicated. Good electron
transport algorithms are considerably more difficult to implement than
photon transport algorithms. Many individual codes have been developed
with individualized Monte Carlo routines to model photon transport and
attenuation in shields, the human body, etc. On the other hand, only a
handful of Monte Carlo codes have been developed that truly model electron
transport, and only a few of them have produced widely accepted results. In a
number of applications in the human body, it is acceptable to assume that
electrons do not need to be transported, i.e., that photoelectrons or emitted
electrons and particles are absorbed exactly where they are created. In
many other applications, particularly in therapy, it is important to know
the spatial distribution of electron dose, and so electron transport is
needed.
Application of Monte Carlo processes to model any system of even
moderate complexity is not possible without the use of a computer. Tables
of random digits were printed in the years before the widespread availability
of computing resources to the average researcher or student. Use of such
tables to sample source emissions and particle interactions would be, of
course, laborious and tedious to the extreme, without the possibility of
generating any meaningful results in a reasonable time frame. Therefore,
Monte Carlo methods did not really reach any prominence in serious
scientific inquiry until computing resources became available to the general
scientific community. Computing power is now quite vast; computers that
in the 21st century sit on our desktops hold more power than was made
available by computing systems of just a few decades ago that filled entire
rooms. Large dedicated workstations and distributed computing networks
continue to increase available and inexpensive computing power by orders
of magnitude every few years. Currently, therefore, the ability of the
investigator to model microscopic (e.g., radiation transport) or macroscopic
(e.g., weather patterns) processes is limited only by his imagination, and by
the availability of well defined pdfs which apply to the situation to be
modelled.
6.2.1. The ‘Reference Man’ phantoms and the ALGAMP code
The first real treatment of internal dose assessment was done in the Manhattan
project, developed by K Z Morgan, W Snyder, and others [4]. Their focus was
on the protection of radiation workers and, to a degree, the general public,
from the production and use of radioactive materials, in particular in relation
to the war effort. They did not, however, employ Monte Carlo methods in
these calculations. Their model of the human body and each of its parts was
a simple homogeneous spherical construct, in which electrons were completely
absorbed and for which photon absorption could be modelled by a simple
analytical expression, knowing the absorption coefficient for the organ or
tissue of interest. More specifically, for soft tissue, bone, lung, or any mixture
thereof (such as the whole body), the photon absorbed fraction for an organ
irradiating itself is given as (1 er ), where is the absorption coefficient
for the tissue composition of interest and r is the assigned effective radius of
the object. This information was developed for the very large number of radio-
nuclides of interest to radiation safety at that time—products of the fission
process, activation products from neutron interactions in the structural and
other materials in the reactors, and others. The results were given in the
context of a system of radiation protection guidelines for workers, i.e.,
amounts of permissible intake of each radionuclide per working year (given
permissible dose limits), and permissible concentrations in air and water
(given the intake limits and assumptions about standard breathing and
drinking rates) [4].
The first breakthrough in the use of Monte Carlo methods was the
development of the Fisher–Snyder heterogeneous, hermaphrodite, anthropo-
morphic model of the human body in the 1970s [5]. This model, or
‘phantom’, consisted of spheres, ellipsoids, cones, tori, and subsections of
such objects, combined to approximate the geometry of the body and its
internal structures. For the majority of cases, electron energy was assumed
to be absorbed where created (exceptions included hollow organs with
separate wall and contents sections, and the bone and marrow, in which
‘crossfire’ between adjacent regions can occur due to the small dimensions
of the regions). For photons, Monte Carlo methods were developed using
a computer code called ALGAM [6], which created photons at random posi-
tions within any ‘source’ region (organ or tissue assumed to be contaminated
with radioactivity), gave these photons a random orientation in 4 space, and
then followed them through various Compton and photoelectric interactions
(pair productions events were quite rare, as starting energies did not exceed
4 MeV) until the photon reached a certain critical low (‘cut-off ’) energy and
was assumed to be locally absorbed, or until it escaped the surface of the
body (at which point the probability of scatter from an air molecule and
redirection towards the body was assumed to be negligibly low). With
repeated sampling of the source, which in this time generally involved only
tens of thousands of trials (histories), a statistical average behaviour of

particles originating in this source could be obtained for other regions of
the body of interest to radiation dose assessment (’target’ regions). This
behaviour was reported as the fraction of energy emitted in the source that
was absorbed in a target (absorbed fraction, AF), with an associated uncer-
tainty (reported as the coefficient of variation). These AFs were thus a
considerable improvement over the values given in ICRP Publication 2 [4],
as the organ geometries were more realistic, and, more importantly, the
organs could irradiate each other, whereas in the ICRP 2 model an organ
could irradiate only itself. These AFs were used later by the ICRP in updated
assessments for workers; of more interest in this text is that they found more
immediate application in dose assessments for nuclear medicine patients,
thanks to the monumental efforts of the newly formed Medical Internal
Radiation Dose (MIRD) Committee of the Society of Nuclear Medicine.
In a flurry of publications in its early years, this committee published
decay data, methods for kinetic analyses, the AFs from the ALGAM calcu-
lations, dose conversion factors for over 100 nuclides of interest to nuclear
medicine, dose calculations for various radiopharmaceuticals, methods for
small scale dose calculations with electrons, and other interesting practical
scientific documents.
The original phantom developed was intended mainly to represent a
healthy average adult male, which well characterized the working population
of its time. As noted above, the phantom did have both male and female
organs (naturally with no cross-irradiation terms given!), but most structures
represented the organs of ‘Reference Man’, as defined by the ICRP from an
extensive range of medical and other literature, restricted primarily to
European and North American populations. Both due to the makeup of
the nuclear medicine population and the diversifying worker population,
the need for other phantoms arose. In 1987, Cristy and Eckerman [7] of
Oak Ridge National Laboratory (ORNL) developed a series of phantoms
representing children of different ages, one of which (the 15-year-old) also
served as a model for the adult female. AFs for these phantoms were
developed using the ALGAMP code (‘P’ signifying a ‘parametrized’ version
of the code—allowing substitution of parameters giving the radii and posi-
tions of the various organs at different ages). These values were published
in an ORNL document, but never officially adopted in the MIRD or other
peer-reviewed literature. Nonetheless, they were widely accepted and used
for dose calculations in different age individuals. In 1995, Stabin et al. [8]
released a set of phantoms representing the adult female at different stages
of gestation, to satisfy the need to calculate radiation doses to the foetus
from nuclear medicine and other applications. These calculations were also
done with the ALGAMP code.
Other models were developed by independent investigators to represent
certain organs or organ systems not included in these phantoms, including
subregions of the brain [9, 10], the eye [11], the peritoneal cavity [12], the
prostate gland [13], bone and marrow [14, 15], rectum [16], spheres of varying
size [17, 18], and others.
6.2.2. Development of MIRDOSE codes

The MIRDOSE computer program [19] was originally developed by Dr.
Stabin to eliminate the tedium of repetitive internal dose calculations (look-
ing up dose conversion factors from tables and adding contributions from
every source to every target, even if of minor importance) to automate the
calculation of organ doses in nuclear medicine. The original code was
developed in 1983 on a Tektronix stand-alone workstation, which had 32k
of total memory and storage limited to 8 inch soft disks. This version
(MIRDOSE 1) was never distributed. In its second incarnation in about
1985, MIRDOSE 2, the code was rewritten to be usable on the IBM Personal
Computer, which was rapidly growing in popularity. The code employed 59
radionuclides, and had available the AFs from the Cristy/Eckerman phan-
tom series. Users entered cumulated activities for the source organs involved,
and obtained dose estimates for chosen target organs, with contributions
from all source organs included, and appropriate ‘remainder of the body’
corrections [20] applied.
The MIRDOSE3 code represented the migration of the MIRDOSE
code from the DOS to the Windows environment. Written in the Microsoft
Visual Basic1 environment, users could select from a menu of over 200
radionuclides, whose decay data can be used with absorbed fractions from
any of 10 phantoms (the six Cristy/Eckerman paediatric phantoms and the
four phantoms representing the nonpregnant or pregnant adult female).
The user entered residence times [21] for all source organs involved, and
the program calculated and applied the appropriate S values to obtain
organ dose estimates (for all available, not selected, target regions). The
program also provided calculation of the ICRP-defined effective dose
equivalent [22] and effective dose [23]. The software thus facilitated dose
calculations under different conditions, in different phantoms, etc. The
program has been used to calculate dose estimates for a variety of nuclear
medicine compounds, and its use is widely cited in the literature.
The evolution of the MIRDOSE software continues today. In order to
provide more computer platform independence, Dr. Stabin recently rewrote
the MIRDOSE code entirely in the Java language and incorporated a curve-
fitting algorithm for kinetic data developed by Dr. Sparks [24]. Java is
compiled on each machine before execution, thus allowing the code to be
used on Unix, Macintosh, Windows NT and other platforms. The code
was renamed ‘OLINDA’ (Organ Level INternal Dose Assessment), partly
to distinguish it from the activities of the MIRD Committee (which had
expressed concern that the name MIRDOSE might imply that it was a
Public domain Monte Carlo codes 139
product of that committee), and partly to integrate the name into a new
unified system of internal and external dose assessment. This unified
system is deployed on an Internet website for rapid electronic access [25].
This site, called the RAdiation Dose Assessment Resource (RADAR),
provides decay data for over 800 radionuclides, absorbed fractions for all
available stylized phantoms and some voxel phantoms, kinetic data, dose
factors (for all phantoms and nuclides), risk information, and other data
via electronic transfer to users worldwide.
6.3. PUBLIC DOMAIN MONTE CARLO CODES
Table 6.1 lists widely used public domain Monte Carlo codes together with a
short description of their key features. Most of the packages mentioned
below run virtually on different platforms and operating systems and are
available free of charge from the authors or through the official channels
(RSIC or NEA).
6.3.1. The EGS code

The electron gamma shower (EGS) computer code system is a general-
purpose package for Monte Carlo simulation of the coupled transport of
electrons and photons in an arbitrary geometry for particles with energies
from a few keV up to several TeV [26]. The code represents the state-of-
the-art of radiation transport simulation because it is very flexible, well docu-
mented and extensively tested. Some have referred to the EGS code as the
de facto gold standard for clinical radiation dosimetry. EGS is written in
MORTRAN, a FORTRAN pre-processor with powerful macro capabilities.
EGS is a ‘class II’ code that treats knock-on electrons and bremsstrahlung
photons individually. Such events require pre-defined energy thresholds
and pre-calculated data for each threshold, determined with the cross-section
generator PEGS.
6.3.2. The MCNP code

MCNP is a general-purpose Monte Carlo code that can be used for neutron,
photon, electron, or coupled neutron/photon/electron transport [28]. The
code treats an arbitrary three-dimensional configuration of materials in
geometric cells bounded by first- and second-degree surfaces and fourth-
degree elliptical tori. For photons, the code takes account of incoherent
and coherent scattering, the possibility of fluorescent emission after photo-
electric absorption, absorption in pair production with local emission of
annihilation radiation, and bremsstrahlung. Electron transport in MCNP
is based on the ETRAN system as in ITS where a condensed random walk
Table 6.1. Key features of public domain Monte Carlo codes used in therapeutic
nuclear medicine applications.
MC code General description
ALGAM [6] Monte Carlo calculations using the OGRE system are employed for
photon transport. The user can specify a source arbitrarily through a
subroutine. Simulation for internal dosimetry calculations is
specifically included.
MIRDOSE [19] Dose calculations based on Monte Carlo pre-calculated specific
absorbed fractions for 10 phantom models using the ALGAMP code.
EGS4 [26] Coupled photon/electron transport in any material through
user-specified geometries. Simulation for internal dosimetry
calculations is not specifically included and requires an extensive
amount of user programming in Mortran.
ITS including Coupled photon/electron transport in any material through slabs,
TIGER cylinders or combinatorial. Simulation for internal dosimetry
CYLTRAN and calculations is not specifically included and requires an extensive
ACCEPT [27] amount of user programming in Fortran.
MCNP [28] Coupled neutron/photon/electron transport in any material through
user-generalized geometry. Simulation for internal dosimetry
calculations is not specifically included and requires an extensive
amount of user manipulation of input date files to model complex
geometries.
GEANT [29] Coupled photon/electron transport in any material through
combinatorial geometry. Simulation for internal dosimetry
calculations not specifically included and requires an extensive amount
of user programming in Cþþ (GEANT 4).
ETRAN [30] Coupled photon/electron transport in plane-parallel slab targets that
have a finite thickness in one dimension and are unbound in the other
two dimensions. Simulation for internal dosimetry calculations is not
specifically included and requires an extensive amount of user
manipulation of input date files to model complex geometries.
based on accepted multiple-scattering theories is implemented. Important

features that make MCNP 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.
6.3.3. The ETRAN code

ETRAN computes the transport of electrons and photons through plane-
parallel slab targets that have a finite thickness in one dimension and are
unbound in the other two dimensions [30]. The incident radiation can
consist of a beam of either electrons or photons with specified spectral and
Public domain Monte Carlo codes 141
directional distribution. Options are available by which all orders of the

electron–photon cascade can be included in the calculation. Thus electrons
are allowed to give rise to secondary knock-on electrons, continuous
bremsstrahlung and characteristic X-rays; and photons are allowed to
produce photo-electrons, Compton electrons, and electron–positron pairs.
Annihilation quanta, fluorescence radiation, and Auger electrons are also
taken into account. According to user specifications, the Monte Carlo
histories of all generations of secondary radiations are followed.
6.3.4. ITS
The Integrated TIGER Series (ITS) of coupled electron/photon Monte Carlo
transport codes is a powerful tool for determining state-of-the-art descriptions
of the production and transport of the electron/photon cascade in time-
independent, multi-material, multi-dimensional environments [27]. ITS is a
collection of programs sharing a common source code library that can solve
sophisticated radiation transport problems. A total of eight codes are in the
collection which can be split into six groups: the TIGER codes (for one-
dimensional slab geometries), the CYLTRAN codes (for two-dimensional
cylindrical geometries), the ACCEPT codes (for arbitrary three-dimensional
geometries), the standard codes (for normal applications), the P codes (for
applications where enhanced ionization/relaxation procedures are needed),
and the M codes (for applications which involve two- or three-dimensional
macroscopic electromagnetic fields). The user selects the appropriate code
from the library and supplies it with any special requirements and the physical
description of the problem to be solved in an input file.
6.3.5. The GEANT code

The GEANT package was originally designed for high-energy physics experi-
ments, but has found applications also outside this domain in the areas of
medical and biological sciences, radiation protection and astronautics [29].
The main applications of GEANT are the transport of particles through an
experimental setup for the simulation of detector response and the graphical
representation of the setup and of the particle trajectories. The two functions
are combined in the interactive version of GEANT. The code has been
rewritten in Cþþ and exploits advanced software-engineering techniques
and object-oriented technology to achieve transparency (GEANT 4).
6.3.6. Other Monte Carlo codes

Many other codes have been developed or adapted by users to match their
specific needs. Interesting initiatives include the so-called ‘all-particle
method’ [31] and EPIC (electron–photon three-dimensional Monte Carlo
transport) codes planned to be developed at the Lawrence Livermore National

Laboratory (LLNL), which houses the world’s most extensive nuclear and
atomic cross section database that parametrizes the interactions of photons,
electrons/positrons, neutrons, protons, and other heavy charged particles
[32]. These planned codes never received any support, and as such were shelved
and were never ready for public release. Among the codes that have received
considerable attention during the past few years is PENELOPE, which
performs Monte Carlo simulation of coupled electron–photon transport in
arbitrary materials and complex quadric geometries [33]. A mixed procedure
is used for the simulation of electron and positron interactions, in which
‘hard’ events (i.e., those with deflection angle and/or energy loss larger than
pre-selected cut-offs) are simulated in a detailed way, while ‘soft’ interactions
are calculated from multiple scattering approaches. The mixed simulation
algorithm for electrons and positrons implemented in PENELOPE reproduces
the actual transport process to a high degree of accuracy and is very stable
even at low energies. This is partly due to the use of a sophisticated transport
mechanics model based on the so-called random hinge method, with
energy-loss corrections for soft events. Other differentiating features of the
simulation are a consistent description of angular deflections in inelastic
collisions and of energy-loss straggling in soft stopping events. Binding effects
and Doppler broadening in Compton scattering are also taken into account.
FLUKA is another ambitious fully integrated all-particle Monte Carlo
simulation package [34]. At the beginning, electron–photon transport was
the same as EGS, but the code has evolved in a very different direction
during the past few years. The code looks promising and is now available
from the authors upon request.
6.4. LIMITATIONS OF CURRENT NUCLEAR MEDICINE DOSE

CALCULATIONS
6.4.1. Introduction
Currently, and for the past 20–30 years, internal dose calculations for nuclear
medicine have used the standardized hermaphroditic phantoms with stylized
geometries (spheres, ellipsoids, cylinders, etc.) described above to represent
the major organs, skeletal system, whole body, and bone marrow of ‘Refer-
ence Man’ (70 kg), ‘Reference Woman’ (57 kg), and ‘Reference’ children and
pregnant women. Doses calculated using these models do not include the
detail needed for accurate assessment of risks in therapeutic administrations
of radiopharmaceuticals and their results have been shown to be poor predic-
tors of marrow radiotoxicity. These models give only average dose to whole
organs (not dose distributions within organs or tissues with possibly nonuni-
form activity distributions) and the reported dose is applicable only to a
Limitations of current nuclear medicine dose calculations 143
99m
Table 6.2. Radiation dose estimates for the reference adult for Tc
glucoheptonate.
Estimated radiation dose
Organ mGy/MBq rad/mCi
Kidneys 4:4 102 1:6 101

Liver 3:2 103 1:2 102
Lungs 1:7 103 6:2 103
Ovaries 5:5 103 2:0 102
Red marrow 2:5 103 9:1 103
Testes 3:7 103 1:4 102
Urinary bladder wall 7:4 102 2:7 101
Effective dose equivalent 1:0 102 mSv/MBq 3:8 102 rem/mCi
Effective dose 7:4 103 mSv/MBq 2:8 102 rem/mCi
person whose size and weight are close to that of the reference individual
after which the model was derived. Thus the doses reported with such
models really represent the dose to a phantom, not to a patient. If the bioki-
netic data to be applied were taken from the actual patient, then these data
are patient specific. In diagnostic applications in nuclear medicine, usually a
standardized kinetic model is also applied. Typically results are shown in
table 6.2.
These dose estimates are based on a standard kinetic model for gluco-
heptonate, and AFs for the adult male phantom in the Cristy/Eckerman
phantom series. If one were to use patient-specific biokinetic data for a
nuclear medicine therapy agent with this same phantom, the result would
be the same as in table 6.2 (except that effective dose quantities may not be
used in therapy applications). One can make the dose estimates more patient
specific through mass-based adjustments to the organ doses:
. Absorbed fractions for electrons and alphas scale linearly with mass.
. Absorbed fractions for photons scale with mass to a power of 1/3.
One generally cannot:
. Account for patient-specific differences in organ geometry.
. Account for patient-specific marrow characteristics.
. Calculate dose distributions within organs.
To perform real patient-specific dose calculations, one needs a patient-
specific physical model to be used with patient-specific biokinetic data. A
‘one-dose-fits-all’ approach to radiation therapy with these internal emitter
treatments is not likely to be effective (due to the narrow range between
tumour ablation and bone marrow toxicity). Individual patients not only
have significantly different uptake and retention half-times of activity of
the radioactive agent, but also have significantly different physical

characteristics and radiosensitivities. Many cancer patients have failed
other treatments, and may enter radiotherapy with compromised marrow
due to their previous treatments. Thus, their therapies should be optimized,
taking into account individual parameters as much as is possible.
If one were to approach the radiation oncologist or medical physicist in
an external beam therapy program and suggest that all patients with a certain
type of cancer should receive exactly the same treatment schedule (beam
type, energy, beam exposure time, geometry, etc.), the idea would certainly
be rejected as not being in the best interests of the patient. Instead, a
patient-specific treatment plan would be implemented in which treatment
times are varied to deliver the same radiation dose to all patients. Patient-
specific calculations of doses delivered to tumours and normal tissues have
been routine in external beam radiotherapy and brachytherapy for decades.
The routine use of a fixed GBq/kg, GBq/m2 , or simply GBq, administration
of radionuclides for therapy is equivalent to treating all patients in external
beam radiotherapy with the same treatment schedule. Varying the treatment
time to result in equal absorbed dose for external beam radiotherapy is
equivalent to accounting for the known variation in patients’ physical
characteristics, and radionuclide biokinetics to achieve similar tumour
doses in internal-emitter radiotherapy, while watching the doses received
by healthy tissues.
6.5. IMPROVEMENTS IN MODELS FOR NUCLEAR MEDICINE

THERAPY
6.5.1. Current approaches to patient-specific dose calculations

In radionuclide therapy of cancer, the practice of managing patient therapy
varies considerably, with some methods being based on patients’ radiation
doses and others focusing only on administered activity. In the dose-based
methods, both biokinetic data and dose conversion factors may be
individualized, in lieu of using published, standardized data. The subject’s
actual biokinetic data are usually used, at least when performing clinical
trials for the purpose of establishing the dosimetry in a new agent; in routine
clinical practice, such data are not often gathered. This places limitations on
the efficacy of the therapy, because knowledge of individual-specific
biokinetics, both for the tumour(s) and critical normal tissue(s) is essential
for proper optimization of therapy. Dose conversion factors from standard
phantoms, as have established the state of the art for dosimetry for the
past 30 years, are generally employed, sometimes adjusting for patient-
specific parameters. Recent advances in imaging technology and radiation
transport methods permit the possibility of far more patient-specific dose
Improvements in models for nuclear medicine therapy 145
calculations than are possible with such standardized phantoms. A variety of

approaches to patient-specific dose calculations are observed in practice, as
follows.
6.5.1.1. No patient-specific parameters (an activity-based approach)

When one has a very wide therapeutic ‘window’ (i.e., the difference between
the radiation dose delivered to the tumour and the highest dose that is
delivered to a radiosensitive normal tissue, such as marrow), one may consider
giving the same amount of activity to all patients without worry about opti-
mizing the therapy. When the window is smaller, optimization becomes
more critical. For over 50 years, 131 I treatment of thyroid cancer has been
based empirically on administered activity rather than on actual radiation
doses delivered [35]. The conventional approach for the post-surgical treat-
ment of thyroid cancer when there is functioning thyroid tissue is to administer
activities less than 7500 MBq, since it has been shown that the red marrow dose
per MBq administered is of the order of 0.3 mGy/MBq (it is of the order of
0.15 mGy/MBq for thyroidectomized thyroid cancer patients). This approach
has been shown to result in minimal haematologic toxicity, as the red marrow
can receive a dose of 200 cGy without significant myelotoxicity [36]. Giving all
patients the same administered activity of 131 I is thought by many physicians to
be as good as patient-specific activity administration in functioning thyroid
cancer patients. Some believe, however, that successful treatment of these
patients involves radiation dose thresholds and that the success rates using a
patient-specific administered activity approach are equal to or better than
those reported with the empiric fixed-dose methods of 131 I administration
[37–40].
6.5.1.2. Minimal patient-specific parameters (also an activity-based

approach)
Often in trials with internal emitters, activity is administered to all patients
based on a fixed amount of activity per kg of body weight or per m2 of
body surface area [41]. This is an improvement over giving every patient
the same amount of activity, but does not account for patient-specific
differences in biokinetics. Two patients of the same weight who had effective
clearance half-times (e.g., from an organ or the total body) that were a factor
of two different presumably would receive doses to those regions that were a
factor of two different.
6.5.1.3. Hybrid patient-specific/standard-phantom-based dosimetry

(a dose-based approach)
A more patient-specific treatment regimen involves determination of
absorbed dose to tumour or to the tissue/organ expected to have toxicity
prior to administration of the therapy radioactivity. Radiation absorbed

dose is defined as the energy absorbed per unit mass and is determined
from the product of cumulated activity (or residence time) and a dose conver-
sion factor (e.g., the S value, defined as the mean absorbed dose to a target
organ per unit cumulated activity in the source organ, in the MIRD schema
[21]). Each of these two factors can be patient dependent and can directly
influence the calculated patient radiation dose to regions of the body (e.g.,
red marrow, tumours, etc.). Cumulated activity is directly proportional to
the product of the uptake and retention half-time of activity of the radio-
labelled agent in the source regions of interest. Dose conversion factors,
like biokinetic data, should reflect individual patient characteristics as
much as possible. For example, radiation absorbed dose estimates using
‘off the shelf’ S values (e.g., MIRD 11 [42], MIRDOSE3 [19], etc.) for
most organs should be adjusted based on direct patient image data [43]
using standard methods which correct for variations in patient organ
masses. More complex, but similar, adjustments can be made to existing
marrow S values, as will be discussed shortly. All dose estimates utilize
models; it is the patient specificity of these models which must be adjusted.
The administration of a patient-specific amount of radionuclide is based
on the individual patient’s biokinetics determined from a dosimetric study
after administration of a tracer amount of the radiolabelled therapeutic
drug product. The measured effective half-time, or residence time, of the
radioactivity in the red marrow, total body, or other normal organ(s) is
used to determine the necessary patient-specific therapy activity of the
radiolabelled drug to deliver the desired (or maximum possible) radiation
absorbed dose to the tumour while limiting the severity of toxicity to the
bone marrow (for a non-myeloablative therapy) or other dose-limiting
normal organ (for a myeloablative treatment). This method is potentially
equivalent to the administration of activity based on MBq/kg, but only if
there is no patient-to-patient variation in uptake or retention of the material.
Some investigators apply patient-specific kinetics to the standard phantoms,
and in addition may make some adjustments to the final results based on
patient-specific physical parameters. Some examples (this is not an
exhaustive review of the literature, just an overview of a few cases which
demonstrate certain common techniques) are listed below.
6.5.1.3.1. Juweid et al. [44], employing a 131 I FðabÞ2 anti-carcinoembryonic

antigen monoclonal antibody against medullary thyroid cancer, used patient
image data to obtain kinetics in normal organs, blood and tumours. Dose
escalation was based on prescribed radiation dose estimates to critical
organs, as determined by a pretherapy dosimetric study. Blood biokinetic
data were used with the Sgouros [45] method to predict marrow activity;
organ, marrow and tumour doses were calculated using the MIRDOSE
software [19], using the standard adult phantom.
6.5.1.3.2. Wiseman et al. [46] in a dosimetry study involving non-Hodgkin’s

lymphoma (NHL) patients treated with a 90 Y IgG1 monoclonal antibody,
determined administered activity using a MBq/kg approach, and then
retrospectively studied radiation doses. They combined patient kinetic data
for whole organs (taken from nuclear medicine camera image data) with
kinetic data on blood (which was extrapolated using the Sgouros method
to estimate marrow residence time) to calculate organ and marrow dose
estimates in over 50 patients. Marrow mass was calculated (in the Sgouros
method) for individual patients, adjusting the model suggested standard
mass by considering the patient’s height and weight. Doses to spleen were
also adjusted in cases in which the spleen mass was known (from computed
tomography or other measurements) to be markedly different from that in
the standard phantom. Otherwise, MIRDOSE estimates for the reference
adult were reported for all patients (the reference adult male was used for
all patients). Similarly, Divgi et al. [47], studying a 131 I antibody against
renal cell carcinoma, used data from planar nuclear medicine camera
images mixed with survey meter measurements to establish whole body
kinetics and used MIRD Pamphlet No 11 [42] total body-to-total body S
values to calculate total body dose. A liver time–activity curve was
constructed, also from nuclear medicine image data, and liver dose was
calculated with standard MIRD S values but modified to represent the
patient’s liver mass, as determined from CT images.
6.5.1.3.3. Wahl et al. [48], studying a 131 I IgG2a monoclonal antibody

against NHL, used nuclear medicine camera images to obtain biokinetic
data for normal organs and tumours. The administered activity was based
on a prescribed radiation dose to the total body, as determined by a pre-
therapy dosimetric study. They used dose to the total body to evaluate
marrow response. Their total body doses were made patient specific through
the use of absorbed fractions for different sized ellipsoids whose mass corre-
sponded approximately to the total mass of the patient. Other organ and
tumour doses were calculated using MIRDOSE.
6.5.1.3.4. Juweid et al. [49] treated patients with 131 I-labelled anti-carcino-
embryonic antigen monoclonal antibodies with administered activities based
either on the patient’s body surface area or on a prescribed red marrow
dose as determined by a pre-therapy dosimetric study.
6.5.1.4. Extension of mathematical phantom results

The MABDose [50] and DOSE3D [51] computer codes adapt the standard
geometrical phantoms, allowing placement of a tumour or tumours in
various locations to estimate dose contributions from these tumours to
normal organs, but do not at present use patient images. These codes
work with stylized representations of average individuals, and give average
dose to whole organs. The RTDS code [52] employs either the standard
MIRDOSE phantom set (pre-calculated dose conversion factors for stan-
dard phantoms) or its own transport algorithms in a limited body space,
based on voxel source kernels (which are useful in a single tissue type) to
produce average organ doses or dose distributions within specified organs
or tissues of the body.
6.5.1.5. Voxel source kernel approaches

As a step in the direction of providing pre-calculated dose conversion factors
for use with three-dimensional voxel data, voxel source kernels were
developed by Williams et al. [53] and by the MIRD Committee [54]. These
dose conversion factors in principle allow calculation of three-dimensional
dose information from patient-specific data, but only in limited spatial
areas where there are no changes in material composition or density (these
kernels were developed only for a soft tissue medium, and thus do not
work in lung or bone, or where soft tissue/bone or lung interfaces occur).
6.5.1.6. Patient-specific dosimetry based on three-dimensional image data

Many specially designed computer codes have been developed for patient-
specific dosimetry and treatment planning. A few groups have managed to
fuse three-dimensional anatomical data, from CT or MRI images, with
three-dimensional data on radionuclide distribution, from SPECT or PET
images, to provide a three-dimensional representation of the radionuclide
distribution in the tissues of the patient. Efforts in this area include the
3D-ID code at Memorial Sloan-Kettering Cancer Center [55, 56], the
SIMDOS code from the University of Lund [57], the RMDP code from
the Royal Marsden Hospital in the UK [58], the VOXELDOSE code from
Rouen, France [59], and the SCMS code [60]. A detailed description of
some of these tools is provided in the subsequent chapters of the book.
Most of them are still research tools and have not entered the clinical area
yet. The code with the most clinical experience to date is the 3D-ID code
[56, 61]. This code produces three-dimensional dose distributions as well as
dose–volume histograms (functions which show what fraction of an organ
received what dose) for normal organs and tumours. The RMDP and
VOXELDOSE codes combine the MIRD voxel source kernels for soft
tissue with point kernels to give dose distributions in unit density soft
tissue regions of the body. The SCMS code of Yoriyaz et al. [60] uses the
MCNP Monte Carlo software to transport electrons and photons in hetero-
geneous voxel-based phantoms, using fused CT and SPECT image data.
They show an example of how different the geometries of the standardized
phantom may be from that of real human subjects. Figure 6.1 shows the
Figure 6.1. Images showing the GI tract (top) and kidney–liver region
(bottom) from the Yale voxel phantom [62] (left) and the Cristy/Eckerman
geometrical phantom [7] (right).
difference between the GI tract and liver–kidney regions of the Cristy/Ecker-

man adult male phantom and that of the voxel-based 75 kg individual image
provided by the group at Yale [62]. Both the realism of the voxel-based phan-
toms and their representation of organ overlap are clearly superior to that of
the standardized geometrical phantoms.
6.5.2. Innovation in software development tools

It is generally accepted that the biomedical research community will benefit
from recent innovations in programming languages, techniques and para-
digms. Several programming and software development approaches have
been suggested in the literature. The choice of software development tools
is further dictated by the user requirements. Unlike procedural programming
languages, which separate data from operations on data defined by
procedures and functions, object-oriented programming languages consist
of a collection of interacting high-level units, the objects that combine
both data and operations on data. This renders objects not much different
from ordinary physical objects. This resemblance to real things gives objects
much of their power and appeal. Not only can they model components
of real systems, but equally well fulfil assigned roles as components in soft-
ware systems. The object-oriented methodology enables an evolutionary
approach to software development that still maintains a high degree of
modularity [63].
When developing software systems it is possible to make use of object

orientation through all the phases, in contrast to traditional methods
based on structured analysis and structured design. It has been shown that
object-oriented programming helps to improve design features and efficacy
assessment of software development, which is not easily obtained with
other programming paradigms [64]. Our experience in undertaking object-
oriented software development allows us to claim that it is worthwhile to
invest in the initial painful period to gain expertise in a new software
development technique [65]. The potential advantages in the later stages
of the software life cycle make investments in the early stages worthwhile.
The powerful constructs promoted by object technology can yield an
elegant, quality code. However, potential users should keep in mind that
while success is possible, the object-oriented paradigm itself does not
guarantee this. Several iterations may be necessary before the final goal is
achieved. Since computers are continually increasing in speed and
memory, it might seem at first that it is only a matter of time before
Monte Carlo calculations become used routinely. However, the same
advances in technology that led to faster computers also lead to bigger
and more difficult problems. Thus there is a continuing need for new
ideas in computer programming and algorithm development. Convincingly
demonstrating that the new methods are truly more effective than the
previous methods requires careful matching of the advantages of the
approaches compared.
6.5.3. Parallel computing aspects

Although variance reduction techniques have been developed to reduce
computation time, the main drawback of the Monte Carlo method is that
it is extremely time-consuming. To obtain the good statistics required for
image reconstruction or radiation dosimetry studies requires tracking
hundreds of millions of particles. Consequently, a large amount of CPU
time (weeks or even months) may be required to obtain useful simulated
data sets. The development of advanced computers with special capabilities
for vectorized or parallel calculations opened a new way for Monte Carlo
researchers. Parallel computers are becoming increasingly accessible to
medical physicists. This allows research into problems that may otherwise
be computationally prohibitive to be performed in a fraction of the real
time that would be taken by a serial machine. Historically, however, most
programs and software libraries have been developed to run on serial,
single-processor computers. A modification or adaptation of the code is
therefore a prerequisite to run it on a parallel computer. However, it is
worth pointing out that among all simulation techniques of physical
processes, the Monte Carlo method is probably the most suitable one for
parallel computing since the results of photon histories are completely
independent from each other. Moreover, computer aided parallelization

tools designed to automate as much as possible the process of parallelizing
scalar codes are becoming available. Although parallel processing seems to
be the ideal solution for Monte Carlo simulation, very few investigations
have been reported and only a few papers have been published on the subject
[2, 66, 67].
Sequential programs make the most effective use of the available
processing power: they alone guarantee maximum use of the CPU. In parallel
programs, communication management introduces an unavoidable over-
head, resulting in less efficient use of the overall CPU power. Scalar or
serial Monte Carlo codes track the history of one particle at a time, and
the total calculation time is the sum of the time consumed in each particle
history. Many Monte Carlo applications have characteristics that make
them easy to map onto computers having multiple processors. Some of
these parallel implementations require little or no inter-processor communi-
cation and are typically easy to code on a parallel computer. Others require
frequent communication and synchronization among processors and in
general are more difficult to write and debug. A common way to parallelize
Monte Carlo is to put identical ‘clones’ on the various processors; only the
random numbers are different. It is therefore important for the sequences
on the different processors to be uncorrelated so each processor does not
end up simulating the same data [68]. That is, given an initial segment of
the sequence on one process, and the random number sequences on other
processes, we should not be able to predict the next element of the sequence
on the first process. For example, it should not happen that if we obtain
random numbers of large magnitude on one process, then we are more
likely to obtain large numbers on another. In developing any parallel
Monte Carlo code, it is important to be able to reproduce runs exactly in
order to trace program execution.
Since a Monte Carlo particle history is a Markov chain, the next inter-
action or movement of a particle is always determined by the current state of
the particle. The histories of two particles become identical only when the
same random number sequence is used to sample the next state. To ensure
that the seed tables on each processor are random and uncorrelated,
Mascagni et al. described a canonical form for initializing separate cycles
of the Fibonacci generators [69]. There are, however, many approaches to
vectorized and parallel random number generation in the literature. We
can distinguish three general approaches to the generation of random
numbers on parallel computers: centralized, replicated and distributed. In
the centralized approach, a sequential generator is encapsulated in a task
from which other tasks request random numbers. This avoids the problem
of generating multiple independent random sequences, but is unlikely to
provide good performance. Furthermore, it makes reproducibility hard to
achieve: the response to a request depends on when it arrives at the generator,
and hence the result computed by a program can vary from one run to the
next. In the replicated approach, multiple instances of the same generator
are created (for example, one per task). Each generator uses either the
same seed or a unique seed, derived, for example, from a task identifier.
Clearly, sequences generated in this fashion are not guaranteed to be
independent and, indeed, can suffer from serious correlation problems.
However, the approach has the advantages of efficiency and ease of
implementation and should be used when appropriate. In the distributed
approach, responsibility for generating a single sequence is partitioned
among many generators, which can then be parcelled out to different
tasks. The generators are all derived from a single generator; hence, the
analysis of the statistical properties of the distributed generator is simplified.
In a review of vectorized Monte Carlo, Martin and Brown [70] described
variations of event-based algorithms together with speed-up results
published by different groups. During the past two decades, investigations
were carried out to run different Monte Carlo codes on multiple-transputer
systems [67], vector parallel supercomputers [71], parallel computers [66] and
a cluster of workstations in a local area network using PVM [72]. There are
large discrepancies in the performance ratio reported by different authors. In
particular, Miura reported a speed-up of about 8 with the vectorized ESG4
code (EGS4V) [73]. A linear decrease in computing time with the number of
processors used was also demonstrated for eight [66] and up to 32 processors
[74]. This speed-up is especially significant in SPECT simulations involving
higher energy photon emitters, where explicit modelling of the phantom
and collimator is required.
6.5.4. Towards clinical applications of ‘on-the-fly’ Monte Carlo-based

dosimetry calculations
It is clear that a new generation of dose modelling tools must be developed to
be used with internal emitter therapy in nuclear medicine. It is unacceptable
to use standardized, geometrical phantoms to do dose calculations for
individual patients if we are to give meaningful information to the physician
to be used in planning patient therapy. ‘One dose fits all’ is not a scientific
approach to the problem. The evolution of methodology that was followed
for external beam radiotherapy treatment planning must be followed for
internal emitter therapy. The technology now exists to develop patient-
specific three-dimensional dose maps, based on fusion of CT (or MRI) and
SPECT (or PET) data, with individualized Monte Carlo calculations done
in a reasonable amount of time using high powered computing workstations
or distributed computing networks. The main difficulty to be overcome is the
resistance of patients and the medical staff to the acquisition of multiple
planar and SPECT images, using a tracer amount of activity, needed to
adequately characterize the dosimetry.
References 153
6.6. SUMMARY
It is gratifying to see in overview the progress that internal dose modelling has
made, from simple spherical geometries, through stylized models of reference
individuals, and most recently towards truly patient-specific models based on
patient images. Challenges remain, particularly in the areas of bone and
marrow dosimetry, in image quantification, and in correlating radiation dose
with effect. As these challenges are met, and experience is gained, patients can
expect to receive therapy with internal emitters, based on accurate dose calcula-
tions and sound risk/benefit evaluations made by their attending physician.
ACKNOWLEDGMENTS
One of the authors (HZ) acknowledges the support of grant SNSF 3152-
062008 from the Swiss National Science Foundation. The contents and
views in this chapter are solely the responsibility of the author and do not
necessarily represent the official views of the institution providing support.
REFERENCES
[1] Andreo A 1991 Monte Carlo techniques in medical radiation physics Phys. Med. Biol.
36 861–920
[3] Bielajew A F and Rogers D W O 1989 Monte Carlo Transport of Electrons and
Photons eds N W Jenkins, A Rindi, A Nahum and D W O Rogers (New York:
Plenum Press) pp 407–19
[4] ICRP 1959 Report of Committee II on Permissible Dose for Internal Radiation. ICRP
Publication 2 I.C.o.R. Protection (New York: Pergamon)
[5] Snyder W, Ford M R and Warner G 1978 Estimates of specific absorbed fractions for
photon sources uniformly distributed in various organs of a heterogeneous phantom.
MIRD Pamphlet No 5, revised NM/MIRD (New York: Society of Nuclear Medicine)
[6] Warner G G and Craig A M J 1968 ALGAM, a computer program for estimating
internal dose for gamma-ray sources in a man phantom ORNL-TM-2250 (Oak
Ridge National Laboratory)
[7] Cristy M and Eckerman K F 1987 Specific absorbed fractions of energy at various ages
from internal photon sources. I Methods, II one year old, III five year old, IV ten year
old, V fifteen year old male and adult female, VI new-born and VII adult male ORNL/
TM 8381/V1-V7 (Oak Ridge National Laboratory)
[8] Stabin M et al. 1995 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)
[9] Eckerman K, Cristy M and Warner G G 1981 Dosimetric Evaluation of Brain Scanning
Agents eds E E Watson, A T Schlafke-Stelson, J L Coffey and R J Cloutier, Third
International Radiopharmaceutical Dosimetry Symposium; HHS Publication FDA
81-8166, US Dept of Health and Human Services, Oak Ridge, TN (Rockville, MD:
Food and Drug Administration) pp 527–40
[10] Bouchet L G, Bolch W E, Weber D A, Atkins H L and Poston J W 1999 MIRD
head and brain. Medical Internal Radiation Dose J. Nucl. Med. 40 62S–101S
[11] Holman B L et al. 1983 Biodistribution and dosimetry of N-isopropyl-p-[123I]iodo-
amphetamine in the primate J. Nucl. Med. 24 922–31
[12] Watson E E, Stabin M, Davis J L and Eckerman K F 1989 A model of the peritoneal
cavity for use in internal dosimetry J. Nucl. Med. 30 2002–11
[13] Stabin M 1994 A model of the prostate gland for use in internal dosimetry J. Nucl.
Med. 35 516–20
[14] Eckerman K F and Stabin M 2000 Electron absorbed fractions and dose conversion
factors for marrow and bone by skeletal regions Health Phys. 78 199–214
[16] Mardirossian G et al. 1999 A new rectal model for dosimetry applications J. Nucl.
Med. 40 1524–31
[19] Stabin M 1996 MIRDOSE personal computer software for internal dose assessment
[20] Cloutier R J, Watson E E, Rohrer R H and Smith E M 1973 Calculating the radiation
dosage to an organ J. Nucl. Med. 14 53–5
[21] Loevinger R, Budinger T and Watson E 1988 MIRD Primer for Absorbed Dose
Calculations S.o.N. Medicine, NM/MIRD (New York)
[22] ICRP 1979 Limits for Intakes of Radionuclides by Workers. ICRP Publication 30
I.C.o.R. Protection (New York: Pergamon)
[23] ICRP 1991 Recommendations of the International Commission on Radiological
Protection ICRP. Publication 60 I.C.o.R. Protection (New York: Pergamon)
[24] Stabin M and Sparks R B 1999 MIRDOSE4 does not exist J. Nucl. Med. Suppl. 40
309P
[25] Stabin M et al. 2001 RADAR—the radiation dose assessment resource. An online
source of dose information for nuclear medicine and occupational radiation safety
J. Nucl. Med. Suppl. 42 243P
[26] Nelson W R, Hirayama H and Rogers D W O 1985 The EGS4 Code System (Stanford
Linear Accelerator Center)
[27] Halbleib J A, Kensek R P, Valdez G D, Seltzer S M and Berger M J 1992 ITS: the
Integrated TIGER Series of electron/photon transport codes—version 3.0 IEEE
[28] Hendricks J S and Briesmeister J F 1992 Recent MCNP developments IEEE Trans.
Nucl. Sci. 39 1035–40
[29] Brun R, Bruyant F, Maire M, McPherson A C and Zanarini P 1994 GEANT detector
description and simulation tool W5013 (CERN)
[30] Berger M J and Seltzer S M 1968 Electron and Photon Transport Programs. I. Intro-
duction and Notes on Program DATAPAC 4, II. Notes on Program ETRAN-15 Rep
NBS 9836 and 9837 (NBS)
References 155
[31] Cullen D E, Balinger C T and Perkins S T 1991 The all particle method: 1991 status
report UCRL-JC-108040 (Lawrence Livermore National Laboratory)
[32] Zaidi H 2000 Comparative evaluation of photon cross section libraries for materials
of interest in PET Monte Carlo simulations IEEE Trans. Nucl. Sci. 47 2722–35
[33] Sempau J, Acosta E, Baro J, Fernandez-Varea J M and Salvat F 1997 An algorithm
for Monte Carlo simulation of the coupled electron–photon transport Nucl. Instr.
Meth. B 132 377–90
[34] Fasso A, Ferrari A and Sala P R 2001 Proceedings of the MonteCarlo 2000 Conference
eds A Kling, F Barao, M Nakagawa, L Tavora and P Vaz (Berlin: Springer) pp 159–64
[35] Zanzonico P B, Brill A B and Becker D V 1995 Principles of Nuclear Medicine eds
H N Wagner, Z Szabo and J W Buchanan (Philadelphia, PA: Saunders) 2nd
edition, pp 106–34
[36] Benua R S, Cicale N R, Sonenberg M and Rawson R W 1962 The relation of
radioiodine dosimetry to results and complications in the treatment of metastatic
thyroid cancer Am. J. Roentgenol. 87 171–82
[37] Maxon H R et al. 1992 Radioiodine-131 therapy for well-differentiated thyroid
cancer—a quantitative radiation dosimetric approach outcome and validation in 85
patients J. Nucl. Med. 33 1132–36
[38] Samuel A M and Rajashekharrao B 1994 Radioiodine therapy for well-differentiated
thyroid cancer: a quantitative dosimetric evaluation for remnant thyroid ablation
after surgery J. Nucl. Med. 35 1944–50
[39] Schlesinger T, Flower M A and McCready V R 1989 Radiation dose assessments in
radioiodine (131I) therapy. 1. The necessity for in vivo quantitation and dosimetry in
the treatment of carcinoma of the thyroid Radiother. Oncol. 14 35–41
[40] Maxon H R 1999 Quantitative radioiodine therapy in the treatment of differentiated
thyroid cancer Q. J. Nucl. Med. 43 313–23
[41] DeNardo S J 2000 Tumor-targeted radionuclide therapy trial design driven by patient
dosimetry J. Nucl. Med. 41 104–6
[42] Snyder W, Ford M R, Warner G and Watson S 1975 ‘S’, absorbed dose per unit
cumulated activity for selected radionuclides and organs. MIRD Pamphlet No 11,
NM/MIRD (New York: Society of Nuclear Medicine)
[43] J A Siegel et al. 1999 MIRD Pamphlet No 16: Techniques for quantitative
radiopharmaceutical biodistribution data acquisition and analysis for use in human
radiation dose estimates J. Nucl. Med. 40 37S–61S
[44] M E Juweid et al. 2000 Initial experience with high-dose radioimmunotherapy of
metastatic medullary thyroid cancer using 131I-MN-14 FðabÞ2 anti-carcinoembryonic
antigen MAb and AHSCR J. Nucl. Med. 41 93–103
[45] Sgouros G 1993 Bone marrow dosimetry for radioimmunotherapy theoretical
considerations J. Nucl. Med. 34 689–94
[46] Wiseman G A, White C A, Witzig T A et al. 1999 Final dosimetry results of IDEC-
Y2B8 phase I/II 90yttrium radioimmunotherapy trial in non-Hodgkin’s lymphoma
J. Nucl. Med., Suppl. 40 64P
[47] Divgi C R et al. 1998 Phase I/II radioimmunotherapy trial with iodine-131-labeled
monoclonal antibody G250 in metastatic renal cell carcinoma Clin. Cancer Res. 4
2729–39
[48] Wahl R L, Zasadny K R, MacFarlane D et al. 1998 Iodine-131 anti-B1 antibody for
B-cell lymphoma an update on the Michigan Phase I experience J. Nucl. Med., Suppl.
39 21S–27S
[49] Juweid M E et al. 1999 Prediction of hematologic toxicity after radioimmunotherapy

with (131)I-labeled anticarcinoembryonic antigen monoclonal antibodies J. Nucl.
Med. 40 1609–16
general purpose dose estimation code Med. Phys. 26 1389–95
Monte Carlo code-based software for internal radionuclide dosimetry J. Nucl.
Med. 40 1517–23
[52] Liu A et al. 1999 A radionuclide therapy treatment planning and dose estimation
system J. Nucl. Med. 40 1151–3
[53] Williams L E, Liu A, Raubitschek A A and Wong J Y 1999 A method for patient-
specific absorbed dose estimation for internal beta emitters Clin. Cancer Res. 5
3015s–3019s
[54] Bolch W E et al. 1999 MIRD Pamphlet No 17: The dosimetry of nonuniform activity
distributions—radionuclide S values at the voxel level. Medical Internal Radiation
Dose Committee J. Nucl. Med. 40 11S–36S
[55] Sgouros G et al. 1990 Treatment planning for internal radionuclide therapy three-
dimensional dosimetry for nonuniformly distributed radionuclides J. Nucl. Med. 31
1884–91
[56] Sgouros G et al. 1993 Three-dimensional dosimetry for radioimmunotherapy
treatment planning J. Nucl. Med. 34 1595–601
[58] Guy M J, Flux G D, Papavasileiou P, Flower M A and Ott R J 2002 RMDP-MC: a
dedicated package for I-131 SPECT quantification, registration, patient-specific
dosimetry and Monte-Carlo ed M Stabin, 7th International Radiopharmaceutical
Dosimetry Symposium, Oak Ridge, TN (Oak Ridge: Associated Universities) in press
[59] Gardin I et al. 2002 VoxelDose: a computer program for 3D dose calculation in
therapeutic nuclear medicine ed M Stabin, 7th International Radiopharmaceutical
Dosimetry Symposium, Oak Ridge, TN (Oak Ridge: Associated Universities) in press
absorbed dose distribution estimates for patient-specific dosimetry J. Nucl. Med. 42
662–9
[61] Kolbert K S et al. 1997 Implementation and evaluation of patient-specific three-
dimensional internal dosimetry J. Nucl. Med. 38 301–8
[62] Zubal I G et al. 1994 Computerized 3-dimensional segmented human anatomy Med.
Phys. 21 299–302
[63] Zaidi H 2000 Object-oriented programming environments: a promising tool for
biomedical imaging research Pol. J. Med. Phys. Eng. 6 71–89
[64] Valdes I H 1994 Advantages of object-oriented programming MD Comput. 11 282–3
[65] Zaidi H, Herrmann Scheurer A and Morel C 1999 An object-oriented Monte Carlo
simulator for 3D positron tomographs Comput. Meth. Programs Biomed. 58 133–45
[66] Zaidi H, Labbé C and Morel C 1998 Implementation of an environment for Monte
Carlo simulation of fully 3D positron tomography on a high-performance parallel
platform Parallel Comput. 24 1523–36
[67] Ma C 1994 Implementation of a Monte Carlo code on a parallel computer system
Parallel Comput. 20 991–1005
References 157
[68] De Matteis A and Pagnutti S 1995 Controlling correlations in parallel Monte Carlo
Parallel Comput. 21 73–84
[69] Mascagni M, Cuccaro S A, Pryor D V and Robinson M L 1995 A fast, high-quality,
and reproducible lagged-Fibonacci pseudorandom number generator J. Comput.
Phys. 15 211–19
[70] Martin W R and Brown F B 1987 Status of vectorized Monte Carlo code for particle
transport analysis Int. J. Supercomputer Appl. 1 11–32
[71] Smith M F, Floyd C E and Jaszczak R J 1993 A vectorized Monte Carlo code for
modeling photon transport in SPECT Med. Phys. 20 1121–7
[72] Kirkby D R and Delpy D T 1997 Parallel operation of Monte Carlo simulations on a
diverse network of computers Phys. Med. Biol. 42 1203–8
[73] Miura K 1987 EGS4V: vectorization of the Monte Carlo cascade shower simulation
code EGS4 Comput. Phys. Commun. 45 127–36
[74] Dewaraja Y, Ljungberg M, Majumdar A, Bose A and Koral K F 2002 A parallel
Monte Carlo code for planar and SPECT imaging: implementation, verification
and applications in I-131 SPECT Comput. Meth. Prog. Biomed. in press
Chapter 7
Dose point-kernels for radionuclide

dosimetry
Manuel Bardie`s, Cheuk Kwok and George Sgouros
7.1. INTRODUCTION
Dose point-kernels (DPKs) are widely used for radionuclide dosimetry.

More than three decades ago Loevinger et al. [1] justified the use of ‘Point
Source Dose Functions’ (ancestors of DPKs) for emissions: ‘The problem
of determining the absorbed dose around a localized particle source in
tissue can be separated in two steps. First, the distribution of dose around
a (hypothetical) point source of particles in tissue is determined. Second,
the dose distribution in and around a localized particle source is calculated
by a suitable summation of the elementary point source distribution.’
The superposition concept, which cannot be dissociated from DPK, is
implicit in this approach. Any source volume can be broken down into a
set of point sources, and the determination of dose distribution around
point sources allows non-point sources to be treated by ‘suitable summation’.
For Berger [2], once again in the case of emissions, the specific
absorbed fraction ðxÞ used to calculate the dose rate at some distance x
from an isotopic point source can also be regarded as a ‘point kernel’ because
of its use in applications of the superposition principle. Thus,
RðxÞ ¼ AnkEðxÞ ð7:1Þ
where RðxÞ is the mean dose rate (in cGy/sec), x is the distance to the source
(in cm), A is the activity (in Bq), n is the number of particles emitted per
decay, k ¼ 1:6 108 g cGy/MeV, E is the mean energy from particles
emitted by the source (in MeV), and ðxÞ is the point kernel (in g1 ).
The relation is apparent here between the DPK and the absorbed fraction,
as defined by the MIRD Committee [3].
MIRD Pamphlet No 7 [4] expounded the formalism of the calculation of
doses delivered at some distance from point sources of electrons or particles
158
Introduction 159
in water and other media. The definition of the variables used differentiated
the specific absorbed fraction (for sources of electrons or particles) from
‘energy-dissipation distribution’ as defined by Spencer [5] and, in particular,
allowed the introduction of ‘scaled absorbed-dose distribution’.
Thus, for a point source of electrons of energy E0 , the dose rate Rðx; E0 Þ
is given by
Rðx; E0 Þ ¼ AnkE0 ðx; E0 Þ: ð7:2Þ
This formulation (very close to the preceding one) defines ðx; E0 Þ as the ‘point
isotropic specific absorbed fraction’ expressed in g1 . Thus, the quantity
4x2 ðx; E0 Þ dx is the fraction of emitted energy absorbed in a spherical
layer of radius x and thickness dx in a medium of density (g/cm3 ). If it is
assumed that all the energy emitted is finally absorbed by the medium, then
ð1
4 x2 ðx; E0 Þ dx ¼ 1; ð7:3Þ
0
which constitutes the normalization requirement for ðx; E0 Þ.

Dimensionless ‘energy-dissipation distribution’ Jðu; E0 Þ, as defined
according to Spencer [6], gives the energy dissipated by the primary electrons
(which is not strictly equivalent to the energy absorbed in the medium) in the
spherical layers of radius u around a point source:
E0
Jðu; E0 Þ ¼ 4x2 ðx; E0 Þ ð7:4Þ
LðE0 Þ
where LðE0 Þ is the mean energy loss of an electron per unit path length (in
MeV/cm) at source energy E0 , and u ¼ x=rðE0 Þ is the distance to the source
expressed as a fraction of the initial electron range rðE0 Þ, in cm, according
to the continuous-slowing-down-approximation (csda). The distance rðE0 Þ is
sometimes noted as r0 in the literature. Normalization of the distance ðuÞ
and introduction of the factor E0 =LðE0 Þ allow the dependence of Jðu; E0 Þ
on E0 to be reduced. Scaled absorbed dose distribution, Fð; E0 Þ, is defined by
dx
Fð; E0 Þ ¼ 4x2 ðx; E0 Þ ¼ 4x2 x90 ðx; E0 Þ ð7:5Þ
d
where ¼ x=x90 is the ratio of the distance from the source x to the 90-
percentile distance.
Fð; E0 Þ is dimensionless and relatively independent of the medium
crossed (even if x90 itself depends on the medium crossed). No explicit
definition of the DPK is provided in the article, but the concept of scaled
absorbed dose distribution was frequently applied in subsequent works
and presented as the DPK. A minor modification often encountered relates
to normalization, which can be performed relative to x90 or r0 , depending on
the author. In this chapter, Fð; E0 Þ will be used for DPK, although the
terminology and the normalization factor actually used in the articles cited
may differ slightly.
160 Dose point-kernels for radionuclide dosimetry
These equations can be generalized to particle sources, which

introduces
R ðxÞ ¼ An kEAv ðxÞ ð7:6Þ
and
dx
F ðÞ ¼ 4x2 ðxÞ ¼ 4x2 x90 ðxÞ: ð7:7Þ
d
The modification of the variables is self-explanatory.
In principle, the resolution of a dosimetric problem consists in considering
the contribution of all emissions from a source volume to an entire target
volume. This implies the integration in six dimensions of a dose deposition
function at some distance from a point source. This function can be defined
as the DPK. It is readily apparent that the value of this breakdown relates to
the possibility of using a single function applicable to the entire area of
interest. This is feasible for an isotropic homogeneous medium, in which case
the determination of dose deposition at some distance from a point source
allows any dosimetric problem to be solved by ‘suitable summation’. Thus, a
DPK approach is relevant mainly in the case of a homogeneous medium.
Another limitation relates to the calculation mode for DPK. In general,
the mean dose (a non-stochastic value) deposited at some distance from a
point source is considered, which places us implicitly in the domain of macro-
dosimetry.
This chapter is divided into three parts: the first concerns how DPKs are
obtained (the methods used and the limitations of these methods); the second
describes the various data sets available in the literature (photons or
electrons); and the third considers the domains for the use of DPKs in radio-
nuclide dosimetry.
7.2. METHODS USED TO GENERATE DOSE POINT-KERNELS
The methods used to generate DPKs have kept pace with changes in the
experimental data available and benefited from spectacular progress in the
field of numerical calculation. This section describes the different methods
used to generate point-kernels and their changes over time.
Until 1955, the only experimental data available concerned a dozen
radionuclides in air. It has been shown that these data could be represented
by the formula [1, 7]

k0 x ½1ðx=cÞ ð1xÞ
ðxÞ ¼ c 1 e þ x e x < c
ðxÞ2 c
ð7:8Þ
k0 ð1xÞ
ðxÞ ¼ e x c
x
Methods used to generate dose point-kernels 161
where k0 is a normalization constant, the apparent coefficient of absorption

for the medium, and c a dimensionless parameter. The coefficients and c
were given for soft tissues and emitters

1:37 E
¼ 18:6ðE0 0:036Þ 2 cm2 =g
E
8
<2
> 0:17 E0 0:5 MeV ð7:9Þ
c ¼ 1:5 0:5 E0 1:5 MeV
>
:
1 1:5 E0 3 MeV
where E is the mean energy of the spectrum, E0 the maximal energy, and E

the mean energy of a spectrum representing an allowed shape of maximum
energy E0 .
At the end of the 1950s, Spencer [5] resolved electron transport equations
numerically (according to the continuous slowing down approximation), while
taking multiple diffusion into account according to the moments method.
These calculations, as validated from experimental measurements in different
media and for different energies [8], considered energy distribution as a
function of the distance for isotropic point sources of monoenergetic electrons
(20 keV to 10 MeV) in different homogeneous media. As a result of adoption of
the csda, statistical variations in energy loss straggling were not taken into
account. Consequently, there was a tendency to underestimate a dose
deposited at some distance from the emission point.
Cross et al. [9, 10] conducted important studies for measurement and
calculation of DPK (planar source or point source) based on the moments
method of Spencer. Important progress was made by Berger [11], who used
the Monte Carlo method to simulate electron transport. Berger was able to
take both multiple scattering and energy loss straggling into account. Energy
loss from -rays and bremsstrahlung production were also incorporated.
Although the point-kernels presented in MIRD Pamphlet No 7 [4] are
derived from the work of Spencer, those discussed later [12] were calculated
using the ETRAN Monte Carlo code [13]. More recently, other point-kernels
have been calculated for 147 radionuclides, using a code (ACCEPT) derived
from ETRAN [14].
The EGS4 Monte Carlo code has also been used to generate point-
kernels for monoenergetic electrons (50 keV to 3 MeV) and for eight radio-
nuclides (32 P, 67 Cu, 90 Y, 105 Rh, 131 I, 153 Sm, 186 Re and 188 Re) [15]. Although
the results obtained with ETRAN versions prior to 1986 may differ from
those obtained with EGS4 [16], due to incorrect sampling of energy loss
straggling in ETRAN, there are no important differences now in the results
obtained with these two codes [13].
Regardless of the method employed, current results are concordant
enough to allow DPKs to be considered as reliable data usable for more
complex geometries in the input of calculation programs.
7.3. REVIEW OF DOSE POINT-KERNELS AVAILABLE FOR

RADIONUCLIDE DOSIMETRY
7.3.1. Photons
Monoenergetic. According to MIRD formalism [17], photon DPK in units
of (cGy/decay) at a distance r from a monoenergetic source of energy E
(MeV) in an unbounded homogeneous medium can be expressed as
en r
KðrÞ ¼ 1:6 108 E e Ben ðrÞ ð7:10Þ
4r2
where and en are respectively the linear attenuation and energy absorption
coefficients of the photon in the medium. The quantity Ben is the energy
absorption buildup factor, defined as the ratio between the absorbed dose
rate due to both primary and scattered photons and that due to primary
photons alone.
Using the moments method of Spencer and Fano, Berger [17] has
published Ben tables for photon energies from 15 keV to 3 MeV for distances
r from 0.05 to 20 mean free paths in water (1 mean free path ¼ 1/). At
individual photon energies within the energy range, Berger also expressed
the buildup factor as a ten-term polynomial of distance. Coefficients of the
polynomial have no obvious simple energy dependence. It has been shown
by Brownell et al. [18] that Berger’s method gives results comparable with
those obtained using a Monte Carlo code for photon transport in a large
soft-tissue volume.
Berger’s tabulated Ben values were fitted to analytic functions other
than polynomials by Kwok et al. [19] and Leichner [20]. Such fittings of
data minimize computer storage space. The parameters of the fitting function
have a simple relationship with photon energy for Kwok et al., while a
common expression applies to the absorbed fractions for photons and
particles for Leichner.
More recently, several groups of researchers have used Monte Carlo
codes to derive photon DPKs in water, namely ETRAN [13], EGS4 [21–
23] and MCNP [24, 25]. The agreement between the codes themselves and
with Berger’s calculation is within 3–5% [25–27].
Radionuclides. The DPK for any specific -emitting nuclide in a homoge-

neous medium can be obtained by summing the monoenergetic photon
DPK Ki over the photon energy spectrum ðni ; Ei Þ, where ni is the probability
of emitting energy Ei per decay and i and en;i are, respectively, the linear
attenuation and energy absorption coefficients of the photon with energy Ei
in the medium:
en;i
Ki ðrÞ ¼ 1:6 108 ni Ei expði rÞBen ði rÞ: ð7:11Þ
4r2
Review of dose point-kernels available for radionuclide dosimetry 163
Monte Carlo codes have been used directly to derive the photon DPKs of
many radionuclides including 123 I, 124 I, 125 I, 131 I, 111 In, 64 Cu, 67 Ga, 68 Ga,
99m
Tc, 131 Cs, 103 Pd and 71 Ge [21, 25, 28, 29]. Some of the DPKs have also
been parametrized to analytic functions for easy computation on personal
computers [21].
7.3.2. Electrons
For dosimetry of electrons, two major types of approach are possible,
depending on the problem involved. First, the problem to be resolved may
consist in comparing doses (or absorbed fractions) delivered to the same
target by different radionuclides. In this case, it seems preferable to calculate,
once and for all, a table of absorbed fractions for monoenergetic electrons
and then integrate these results into the emission spectra of the radionuclides
considered. Second, when various geometric configurations need to be
studied for the same radionuclide, it is often preferable to use the DPK of
the radionuclide considered.
In both cases, the application of data for monoenergetic electrons to all
electron emissions of a radionuclide should be performed with due care. First
of all, it is essential to consider the -emission spectra and not mean energy
[30]. Some spectra can be neglected if their emission percentage is too low to
contribute significantly to the delivered dose, provided that this choice is
indicated and clearly documented. Monoenergetic emissions (Auger or
conversion electrons) should be taken into account if they contribute signifi-
cantly to the delivered dose.
For example, the electron emission values given in table 7.1 for 131 I have
been published by the French Primary Laboratory of Ionizing Radiations
[31]. A comparison of these values with those cited by the MIRD Committee
[32] (in boldface in the table) indicates that the difference actually concerns
less than 1% of the energy emitted by transition. This section cites various
DPK sets used in the dosimetry of radionuclides.
Monoenergetic DPK. The tables in MIRD Pamphlet No 7 [4] provide

Fð; E0 Þ values for nine energies in water ranging from 25 keV to 4 MeV.
These DPKs were generated from Spencer’s energy-dissipation distributions.
Insofar as the data show little variation as a function of energy, it is possible
to interpolate these values, if necessary.
Berger [12] published an article initially intended as a ‘preliminary report’,
but which became one of the most frequently cited references in the DPK field.
These ‘improved point-kernels’ were obtained by Monte Carlo simulation
using the ETRAN code [11]. Angular deflections and energy loss straggling
due to multiple Coulomb scattering (by atoms and orbital electrons) were
taken into account, and energy transport by secondary bremsstrahlung was
included. The Fð; E0 Þ tables were given for 36 energies ranging from
131
Table 7.1. Electron emission data for I.
Energy/particle, Particles/transition, Energy/transition, ðiÞ

EðiÞ 100nðiÞ
(keV) (%) g cGy/mCi h fJ/Bq s
eAL 3:9 1:5 5:7 0:5 0.0004 0.031

eAK 29:02 5:51 0:7 0:1 0.0004 0.027
ec1K 45:622 0:016 3:55 0:08 0.0035 0.259
ec1L 75:065 0:335 0:47 0:01 0.0007 0.056
ec1MN 79:65 0:45 0:105 0:002 0.0002 0.013
ec3K 129:27 0:03 0:48 0:09 0.0013 0.099
ec4K 142:649 0:016 0:050 0:002 0.0002 0.011
ec3L 158.48–159.15 0:23 0:04 0.0008 0.058
ec3MN 162.9–163.8 0:065 0:010 0.0002 0.017
ec4L 171.16–172.43 0:0110 0:0004 0.0000 0.003
ec7K 249:74 0:03 0:254 0:009 0.0013 0.102
ec7L 278.85–279.52 0:045 0:002 0.0003 0.020
ec7MN 283.3–284.2 0:0112 0:0005 0.0000 0.005
ec14K 329:92 0:02 1:55 0:05 0.0109 0.819
ec14L 359.03–359.70 0:25 0:01 0.0019 0.144
ec17K 602:417 0:016 0:0285 0:0003 0.0004 0.028
b
1ðmaxÞ 247:9 0:6 2:11 0:03 0.0031 0.235
b
2ðmaxÞ 309:9 0:6 0:63 0:01 0.0012 0.088
b
3ðmaxÞ 333:8 0:6 7:21 0:07 0.0148 1.116
b
4ðmaxÞ 606:3 0:6 89:9 0:6 0.368 27.59

5ðmaxÞ 629:7 0:6 0:072 0:007 0.0003 0.026
b
6ðmaxÞ 806:9 0:6 0:1 0.0006 0.045
0.5 keV to 10 MeV, which in practice allowed scanning of all of the necessary
energies in the domain of -emitting radionuclides. These data have been
widely used in dosimetric models or in DPK calculations for radionuclides.
Subsequently, it appeared that an incorrect sampling of energy-loss straggling
in ETRAN affected the validity of these data [13, 16]. The corrected values
now available extend from 20 keV to 20 MeV [33].
The tables for -ray dose distributions in water published by Cross et al.
[14, 34] were calculated using the ACCEPT code (derived from ETRAN).
For monoenergetic emissions, comparison with values obtained with the
EGS4 code [15] for six energies ranging from 50 keV to 3.0 MeV indicated
that differences between the two data sets were slight (a few per cent).
Fð; E0 Þ values were given for nine energies from 25 keV to 4 MeV.
Radionuclides. The tables in MIRD Pamphlet No 7 [4] give F ðxÞ values

for 75 radionuclides in water. The values used for spectra were taken
essentially from a compilation by Martin et al. [35]. These data can be
extrapolated to media other than water.
The monoenergetic DPK values published by Berger [12] were used by

Prestwich et al. [36] to calculate DPK for six radionuclides of interest in
nuclear medicine (32 P, 67 Cu, 90 Y, 131 I, 186 Re and 188 Re). Only the emissions
of each radionuclide were considered. The results are presented in table form
as well as in analytic form (via a fit of the data by a sum of log-normal and
exponential functions).
The values published by Simpkin and Mackie [15], obtained using the
EGS4 Monte Carlo code [37], concern eight radionuclides (32 P, 67 Cu, 90 Y,
105
Rh, 131 I, 153 Sm, 186 Re and 188 Re). The calculation was performed for
spectra and monoenergetic emissions.
The tables for -ray dose distributions in water published by Cross et al.
[14, 34] were calculated using the ACCEPT code (derived from ETRAN). The
integration of monoenergetic DPKs for 147 radionuclides was performed for
emissions with a relative intensity greater than 1% and conversion or Auger
electrons with energy above 30 keV and intensity greater than 1%. The results
are in agreement with those obtained experimentally or by simulation using
other codes (ETRAN or EGS4). As experimental values measured in water
were not available, validation was performed by extrapolation of values
measured in other media (air, gas, plastic or metal scintillators). The results
are shown as the product of the dose rate at distance r from a point source
multiplied by the squared distance, i.e., r2 J 0 ðrÞ, in nGy h1 Bq1 cm2 ,
which in fact amounts to the product obtained by multiplying r2 by a dose
per decay (similar to an S-factor in MIRD terms) [38].
Cross [39] has more recently proposed empirical expressions for calcula-
tion of doses delivered at some distance from point sources. The principle
consists in adding an ‘end-of-range’ term to an expression similar to that
proposed by Loevinger. The adjustment was made for 60 radionuclides
using data derived from the ACCEPT code. The error involved is of the
order of a few per cent in most cases.
The DPKs proposed for electrons rarely descend below 10–20 keV, which
for most simulation codes corresponds to the limit for validity of the multiple
diffusion theory. It is possible to simulate interactions on an ad hoc basis, and
various codes have been proposed for this purpose [40]. However, their use is
generally limited to energy depositions very near the emission point (e.g., for
Auger emitters). These applications most often relate to microdosimetric
studies, which are outside the field of DPK utilization.
7.3.3. Combined approaches

Leicher [20] proposed a unified approach to photon and particle dosimetry
based on a fit of Berger tables for photons [17] and electrons [4]. The empirical
function proposed is also valid for photons and particles.
Other point-kernels generated by the EGS4 Monte Carlo code have
more recently been proposed by Furhang et al. [21] for 14 radionuclides in
an aqueous medium (123 I, 124 I, 125 I, 131 I, 111 In, 64 Cu, 67 Cu, 67 Ga, 68 Ga, 186 Re,
188
Re, 153 Sm, 117m Sn and 99m Tc). The results are expressed by providing
for each radionuclide the coefficients of a series of polynomials and
exponentials approximating the calculated values. Once again, the equations
proposed are easy to include in more complex calculation programs.
Bolch et al. [41] extended the DPK concept to cubic geometry by
presenting S value tables at the voxel level. The objective was to use these
S values in cases in which cumulated activity is determined from tomoscinti-
graphic images or quantitative autoradiographies, i.e., from numerical
images sampled in three dimensions (voxels). For each target voxel, the
mean dose is obtained by summing the doses delivered by each surrounding
voxel:
X
N
ðvoxelk Þ ¼
D Avoxelh Sðvoxelk voxelh Þ ð7:12Þ
h¼0
The determination of Sðvoxelk voxelh Þ is performed by Monte Carlo

calculation using the EGS4 code in a homogeneous medium (soft tissue).
In fact, this calculation could have been performed perfectly well directly
from DPKs.
The calculation was performed for different radionuclides (32 P, 89 Sr,
90
Y, 99m Tc and 131 I) for cubic voxels with 3 and 6 mm sides (i.e., for PET
and SPECT respectively) and for 131 I for cubic voxels with 0.1 mm sides
(for autoradiography).
It may be concluded that numerous DPK sources exist, involving either
raw data directly or relations obtained by adjustment of these values. The use
of approximated relations for data (experimental or calculated), which were
employed in the past to reduce calculation time, seems less justified today.
Insofar as any data fit introduces an element of uncertainty into the initial
scheme, it seems more logical to use raw data directly.
The calculation of raw data is generally performed from Monte Carlo
simulation codes. A review of current studies indicates that the various
codes proposed give similar results and are sufficiently concordant with
experimental values for DPKs, so that they can be considered as reliable
and used as input data in more elaborate calculation models.
7.3.4. Impact of the medium

Point-kernel dose calculations implicitly assume that the absorbed dose
is deposited within a uniform, isotropic medium (usually water). The
effect of accounting for differences in density and atomic number
variations and the potential errors associated with neglecting such variations
have been studied for photon dosimetry using Monte Carlo calculations
[21]. The results are abstracted in table 7.2. The first column of values ()
Table 7.2. Variation of photon absorbed fractions for selected radionuclides

and organs when making different assumptions regarding organs’ density and
atomic number.
Radionuclide Source Target ;Z

125
I Pancreas Spleen 1:8 104 1:7 104 1:7 104
Left kidney Liver 5:9 107 4:9 107 3:3 107
Left kidney Right kidney 7:0 106 4:0 106 4:4 107
131
I Pancreas Spleen 7:2 105 6:7 105 6:7 105
Left kidney Liver 5:0 106 4:8 106 4:8 106
Left kidney Right kidney 1:4 105 1:2 105 1:2 105
presents calculations performed assuming a homogeneous water phantom

with air beyond the outer contour of the reference man phantom. The
second column ( ) presents calculations accounting for density ()
variations, and the third (;Z ) for both density and atomic number (Z)
variations.
Table 7.2 shows that density and atomic number variations may be
neglected for geometries in which the source, target and separating
medium are similar. The pancreas, spleen and separating medium, for
example, have similar densities and atomic numbers. Correspondingly, the
magnitude of the error in neglecting and Z variations is small. Conversely,
vertebral bone separates the left and right kidneys, so that failure to correct
for and Z could lead to dose overestimations of more than one order of
magnitude (depending on the photon energy spectrum).
The effect of accounting for differences in density and atomic number
variations on photon DPK was also addressed by Monroe et al. [29] and
Janicki et al. [25]. Both of these groups introduced a correction factor F to
allow for a change in the effective path length between a photon point
source and the point where the dose is evaluated. If the path passes through
different materials j of varying thickness tj and a linear attenuation coefficient
j , and w is the linear attenuation coefficient of water. F is multiplied by the
DPK for water to become a modified DPK. This approach also assumes that
an equivalent thickness of water, producing the same primary photon
attenuation as the actual path through a heterogeneous geometry, predicts
scatter build-up correctly:
X
F ¼ exp ðj w Þtj : ð7:13Þ
j
The effect of tissue inhomogeneity is much more complicated on electron/

DPK than on photon DPK, because of rapid variation of the electron
scattering cross-section with the atomic number of the scatterer and electron
energy. Kwok et al. [42, 43] and Nunes et al. [44, 45] used both Monte Carlo
calculations and experiments to assess the effect of bone and air on dose
distribution of point sources of low-energy electrons and emitters near
tissue boundaries. When a point source of 147 Pm (maximum energy
225 keV) or 204 Tl (maximum energy 763 keV) was positioned at a planar
interface of cortical bone (CB) and red marrow (RM), the radiation dose
in RM within 6 mm from the interface was increased by 9 1% (SE) for
both sources, as compared with the dose in the same region inside a homo-
geneous volume of RM. The dose in a spherical shell of RM 20 mm thick
next to a spherical interface of CB/RM received from a monoenergetic
point source of electrons at the centre of a sphere of RM with a radius of
500 mm was increased by up to 22 1% (SE) compared with that for the
same scoring region in the homogeneous case. The maximum dose increase
occurred at 400 keV electron energy. The dependence of dose enhancement
on the curvature of the interface was also demonstrated. Conversely, when
a 32 P point source was positioned at a planar interface of air and soft
tissue, the radiation dose in soft tissue 5.5 mm from the boundary was reduced
by 26:44 0:02% (SE) from that in the homogeneous situation. There was
no simple modification of electron/ DPK due to tissue inhomogeneity.
An ingenious way to calculate the DPK in any homogeneous medium
from that in air (or water) was suggested by Cross and co-workers [14], and is
known as the ‘scaling factor’ method. Marcu and Prestwich [46] recently
extended the applicability of the scaling factor method to dissimilar media
with a planar interface. The investigation was done for a planar source of
32
P in water and the source was located at various distances from a planar
interface between water and any other homogeneous medium with atomic
number Z in the range 8 < Z < 50. Dose deposition discrepancies of less
than 5% were detected for the depth within which at least 95% of the emitted
energy was deposited. It will be of interest to develop procedures to apply the
scaling factor method to non-planar geometries and to more than one closely
spaced boundary.
7.4. A GENERAL APPROACH FOR USE OF DOSE

POINT-KERNELS IN CALCULATIONS
It is rather difficult to classify the different approaches for the use of DPKs in
radionuclide dosimetry. The range of applications is rather broad (calcula-
tion of absorbed doses, absorbed fractions or S factors), the scales considered
extend from the cell to the organ, and the methods used are numerous
(numerical integration or convolution; hybrid methods associating Monte
Carlo simulation and the convolution of radioactive sources by DPKs).
The classification adopted here is based on the scale of the problem, from
cellular to organ dimensions.
A general approach for use of dose point-kernels in calculations 169
7.4.1. Absorbed fractions on a cellular scale

DPKs are frequently used for calculations on a cellular scale, for which the
medium can often be considered as homogeneous (generally equivalent to
water) and isotropic.
Models have been developed to evaluate the dosimetric properties of
different emitters in the context of targeted radiotherapy. For example,
in the case of homogeneous uptake of electron emitters (energy E0 ) through-
out the cell (radius RS ), Berger [2] has shown that the self-absorbed fraction
of energy can be calculated by
ð
1 2RS x x 3
ðRS RS Þ ¼ 1 1:5 þ 0:5 Fð; E0 Þ dx: ð7:14Þ
r0 0 2RS 2RS
Likewise, for uptake on the cell surface (e.g., in radioimmunotherapy), the
self-absorbed fraction [47] is given by
ð
1 1 x
ðRS RS Þ ¼ 1 Fð; E0 Þ dx ð7:15Þ
2r0 0 2RS
where Fð; E0 Þ is the scaled point kernel (considered above) and r0 the
maximal electron range according to the csda.
Various publications have proposed absorbed fractions on a cellular
scale. All describe the geometry of the problem analytically (even though
the integration or convolution used to approach the result is generally
performed numerically). They often apply a geometrical reduction factor
(ranging from 0 to 1) describing the efficacy of irradiation at some distance
from the source point in terms of the fraction of the sphere centred on the
emission point intersecting target volume. Berger [2] noted that ‘One can
interpret ðxÞ very simply as the fraction of the spherical shell of radius x
around the point of observation that lies within the source region’.
The frequent use of DPK on a cellular scale can be explained at least in
part by the simplification introduced through spherical symmetries, both for
the geometrical model and energy deposition at some distance from a point
source.
7.4.2. Absorbed fractions on a millimetre scale

When a dose is delivered on a millimetre scale, e.g., in the case of tumour
mass during targeted therapy trials in animals, the use of DPK can be
envisaged (generally for -emitters), provided that the medium is always
homogeneous. In this case, various possibilities can be considered, depending
on how activity distribution is determined.
If the activity is described analytically, the preceding case applies, i.e.,
appropriate integration of DPK [48]. Thus, it is possible to convolute activity
distribution by DPK [19]. If the activity is described discretely, e.g., in the
case of serial autoradiographs of tumor sections, it is possible to convolute

each volume element (voxel) by the DPK to obtain the dose throughout
the volume [49, 50].
A mouse model for calculating cross-organ doses for 90 Y-labelled
immunoconjugates was proposed by Hui et al. [51]. This approach used
the DPK convolution proposed in MIRD Pamphlet No 7 [4] to calculate
self-absorbed fractions for organs with a density close to that of water,
and Monte Carlo simulation [37] for organs with a different density (lungs,
bone and marrow).
It is possible to simulate the geometry of the problem (size, relative position
of source regions and targets) by a Monte Carlo approach and then use DPK to
calculate the doses delivered at some distance from the emission points. This
approach has been used to design a ‘mouse bone marrow dosimetry model’
[52] or in the case of dose distribution for follicular lymphomas [53].
7.4.3. Absorbed fractions on an organ scale

Doses delivered by emissions are generally considered on an organ scale, in
which case emissions are regarded as non-penetrating. The calculation of S
factors for the best-known anthropomorphic phantoms is based in part on
DPKs [54–56]. In this case, activity is considered to be uniformly distributed
in organs modelled by simplified geometric forms.
DPKs have also been proposed for customized dosimetry in the case of
therapeutic applications of radionuclides [28, 57, 58]. Activity is determined
from quantitative tomoscintigraphy. Thus, three-dimensional discrete
distribution is considered, and activity is regarded as uniform in each
volume element or voxel. The calculation is performed by DPK convolution
of activity distribution.
In the case of discrete convolutions, the use of fast Fourier transform
(FFT) [57] or fast Hartley transform (FHT) [59] allows a significant reduc-
tion of calculation time, which could facilitate the use of these methods in
a clinical context. However, it should be recalled that these methods apply
to a homogeneous medium.
7.5. CONCLUSIONS
Dose point-kernels constitute one of the tools available for performing

dosimetric calculations in nuclear medicine [60]. In general, doses delivered
by emitters are considered on a microscopic scale and those by emitters
on an organ scale. Thus, DPK dosimetric methods can be applied on a micro-
scopic or macroscopic scale, provided that the medium in which radiations
propagate is considered to be homogeneous. Is it possible to go beyond
this limitation of homogeneity?
References 171
DPKs constitute the most suitable approach when the medium is

actually homogeneous and isotropic, but provide only a stop-gap solution
when it is heterogeneous and an estimation of delivered doses is required
within a reasonable time. In fact, the relevant alternative to using DPK in
a heterogeneous medium consists in simulating particle transport and
energy depositions integrally by Monte Carlo methods. This alternative is
being used increasingly in medical physics, e.g., in nuclear medicine.
However, the methods applied still require abilities in physics and computer
science and involve a computing power that limits their extension. In the near
future, it is likely that simulation methods will satisfy current needs for custo-
mized dosimetry on a per patient basis in targeted radiotherapy and other
applications. In the meantime, uses of DPKs in a heterogeneous medium
can probably be explored in dosimetry for brachytherapy and superposition
methods [23].
In principle, the field of DPK applications concerns cases in which the
mean absorbed dose at some distance from a point source is valid, represent-
ing a stochastic and macrodosimetric magnitude. However, this is not always
true, e.g., when dose calculations are made on a cellular scale and it is
necessary to verify that the situation is outside the field of application of
microdosimetric methods [61].
DPKs have been used for more than 40 years in the field of radionuclide
dosimetry. They represent the energy transfer function at some distance from
a point source and can be used in many dosimetric approaches that apply the
superposition principle. Values available in the literature both for and
electron/ emissions can be considered as reliable. The large number of
approaches proposed, whether for simple or complex geometries described
analytically or discretely, are indicative of the value and versatility of this
type of data in medical physics.
REFERENCES
[1] Loevinger R, Japha E M and Brownell G L 1956 Discrete Radioisotope Sources in Radia-
tion Dosimetry ed G J Hine and G L Brownell (New York: Academic) pp 693–755
[2] Berger M J 1970 Beta-ray dosimetry calculations with the use of point kernels. In
Medical Radionuclides: Radiation Dose and Effects eds R J Cloutier, C L Edwards
and W S Snyder (Washington, DC: US Atomic Energy Commission) pp 63–86
[3] Loevinger R and Berman M 1968 MIRD Pamphlet No 1: A schema for absorbed-
dose calculation for biologically distributed radionuclides J. Nucl. Med. 9 7–14
[4] Berger M J 1971 MIRD Pamphlet No 7: Distribution of absorbed doses around point
sources of electrons and beta particles in water and other media J. Nucl. Med. 12 5–23
[5] Spencer L V 1955 Theory of electron penetration Phys. Rev. 98 1597–615
[6] Spencer L V 1959 Energy Dissipation by Fast Electrons. Monograph No 1 (National
Bureau of Standards)
[7] Loevinger R and Holt J G 1956 Internally administered radioisotopes. In Radiation
dosimetry ed G J Hine and G L Brownell (New York: Academic) pp 801–73
[8] Spencer L V 1966 Energy Dissipation by Fast Electrons. Monograph No 1, US

Government Printing Office (National Bureau of Standards)
[9] Cross W G and Williams G 1967 The distribution of absorbed energy from a point
beta source Canadian J. Phys. 45 2021–40
[10] Cross W G, Williams G, Ing H, Freedman N O and Mainville J 1982 Tables of beta-
ray dose distributions in water, air and other media AECL-7617 (Atomic Energy of
Canada Limited)
[11] Berger M J 1963 Monte Carlo calculation of the penetration of diffusion of fast
charged particles. In Methods in Computational Physics F.S.a.R.M. ed B Alder
(New York: Academic) pp 135–215
[12] Berger M J 1973 Improved point kernels for electrons and beta-ray dosimetry NBSIR-
73-107 (National Bureau of Standards)
[13] Seltzer S M 1988 An overview of ETRAN Monte Carlo methods. In Monte Carlo
Transport of Electrons and Photons ed T M Jenkins, W R Nelson and A Rindi
(New York: Plenum) pp 153–81
[14] Cross W G, Freedman N O and Wong P Y 1992 Tables of beta-ray dose distributions
in water AECL-10521 (Atomic Energy of Canada Limited)
[15] Simpkin D J and Mackie T R 1990 EGS4 Monte Carlo determination of the beta dose
kernel in water Med. Phys. 17 179–86
[16] Rogers D W O and Bielajew A F 1988 A comparison of EGS4 and ETRAN. In Monte
Carlo Transport of Electrons and Photons ed T M Jenkins, W R Nelson and A Rindi
(New York: Plenum) pp 323–44
[17] Berger M J 1968 Energy deposition in water by photons from point isotropic sources.
MIRD Pamphlet No 2 J. Nucl. Med. 9 15–25
[18] Brownell G L, Ellet W H and Reddy A R 1968 MIRD Pamphlet No 3: Absorbed
fractions for photon dosimetry J. Nucl. Med. 27 27–39
[19] Kwok C S, Prestwich W V and Wilson B C 1985 Calculation of radiation doses for
nonuniformly distributed beta and gamma radionuclides in soft tissue Med. Phys.
12 405–12
[20] Leichner P K 1994 A unified approach to photon and beta particle dosimetry J. Nucl.
Med. 35 1721–9
[22] Luxton G 1994 Comparison of radiation dosimetry in water and in solid phantom
materials for I-125 and Pd-103 brachytherapy sources: EGS4 Monte Carlo study
Med. Phys. 21 631–41
[23] Carlsson A K and Ahnesjo A 2000 Point kernels and superposition methods for
scatter dose calculations in brachytherapy Phys. Med. Biol. 45 357–82
[24] Tsoulfanidis N and Shrestha B 1997 Photon dose equivalent rate from a cylindrical
source using a point kernel technique Health Phys. 72 931–5
[25] Janicki C, Duggan D M and Rahdert D A 2001 A dose-point-kernel model for a
low energy gamma-emitting stent in a heterogeneous medium Med. Phys. 28
1397405
[26] Lubberink M et al. 1999 Positron emission tomography and radioimmunotargeting—
aspects of quantification and dosimetry Acta Oncol. 38 3439
[27] Luxton G and Jozsef G 1999 Radial dose distribution, dose to water and dose rate
constant for monoenergetic photon point sources from 10 keV to 2 MeV: EGS4
Monte Carlo model calculation Med. Phys. 26 25318
References 173
[28] Giap H B, Macey D J, Bayouth J E and Boyer A L 1995 Validation of a dose-point

kernel convolution technique for internal dosimetry Phys. Med. Biol. 40 36581
[29] Monroe J I et al. 2001 Experimental validation of dose calculation algorithms for the
GliaSite RTS, a novel 125I liquid-filled balloon brachytherapy applicator Med. Phys.
28 7385
[30] Akabani G, Poston J W Sr and Bolch W E 1991 Estimates of beta absorbed fractions
in small tissue volumes for selected radionuclides J. Nucl. Med. 32 8359
[31] Lagoutine F, Coursol N and Legrand J 1982 Table de Radionucleídes (Gif sur Yvette,
France: Laboratoire National Henri Becquerel, Commissariat à l’Energie Atomique)
[32] Weber D A, Eckerman K F, Dillman L T and Ryman J C 1989 MIRD: Radionuclide
Data and Decay Schemes (New York: Society of Nuclear Medicine)
[33] Seltzer S M 1991 Electronphoton Monte Carlo calculations: the ETRAN code Appl.
Radiat. Isot. 42 91741
[34] Cross W G, Freedman N O and Wong P Y 1992 Beta-ray dose distributions from
point sources in an infinite water medium Health Phys. 63 16071
[35] Martin M J and Blichert-Toft P H 1970 Radioactive atoms—Auger electrons, alpha-,
beta-, gamma-, and X-ray data Nucl. Data Tables 8 1198
[36] Prestwich W V, Nunes J and Kwok C S 1989 Beta dose point kernels for radionuclides
of potential use in radioimmunotherapy J. Nucl. Med. 30 103646
[37] Nelson W R, Hirayama H and Rogers D W 1985 The EGS4 Code System Report No
265 (Stanford Linear Accelerator Center)
Calculations, Revised (New York: Society of Nuclear Medicine)
[39] Cross W G 1997 Empirical expressions for beta ray point source dose distributions
Rad. Prot. Dosim. 69 8596
[40] Paretzke H G, Goodhead D T, Kaplan I G and Terrissol M 1995 Track structure
quantities in atomic and molecular data for radiotherapy and radiation research
IAEA-TECDOC-799 (International Atomic Energy Agency)
[41] Bolch W E et al. 1999 MIRD Pamphlet No 17: The dosimetry of nonuniform activity
distributions. Radionuclide S values at the voxel level J. Nucl. Med. 40 11S36S
[42] Kwok C S, Bialobzyski P J, Yu S K and Prestwich W V 1990 Effect of tissue
inhomogeneity on dose distribution of point sources of low-energy electrons Med.
Phys. 17 78693
[43] Kwok C S, Bialobzyski P J and Yu S K 1991 Effect of tissue inhomogeneity on dose
distribution of continuous activity of low-energy electrons in bone marrow cavities
with different topologies Med. Phys. 18 53341
[44] Nunes J, Prestwich W V and Kwok C S 1993 Experimental determination of 32P dose
backscatter factors at and near soft-tissue boundaries Med. Phys. 20 223–1
[45] Nunes J, Prestwich W V and Kwok C S 1993 An evaluation of the EGS4 and
CYLTRAN Monte Carlo codes with regard to boundary beta-ray dosimetry by
comparison with experimental beta-ray dose backscatter factors Med. Phys. 20 124350
[46] Marcu S M and Prestwich W V 1998 Application of the scaling factor method to
estimation of beta dose distributions for dissimilar media separated by a planar
interface Med. Phys. 25 147886
[47] Sastry K S R, Haydock C, Basha A M and Rao D V 1985 Electron dosimetry for
radioimmunotherapy: optimal electron energy Rad. Prot. Dosim. 13 24952
[48] Langmuir V K and Sutherland R M 1988 Dosimetry models for radioimmunotherapy
Med. Phys. 15 86773
[49] Roberson P L, Buchsbaum D J, Heidorn D B and Ten Haken R K 1992 Three-

dimensional tumor dosimetry for radioimmunotherapy using serial autoradiography
Int. J. Rad. Oncol., Biol., Phys. 24 32934
[50] Roberson P L, Heidorn D B, Kessler M L, Ten Haken R K and Buchsbaum D J 1994
Three-dimensional reconstruction of monoclonal antibody uptake in tumor and
calculation of beta dose-rate nonuniformity Cancer 73 9128
[51] Hui T E et al. 1994 A mouse model for calculating cross-organ beta doses from
yttrium-90-labeled immunoconjugates Cancer 73 9517
[52] Muthuswamy M S, Roberson P L and Buchsbaum D J 1998 A mouse bone marrow
dosimetry model J. Nucl. Med. 39 12437
[53] Hui T E et al. 1992 Localised beta dosimetry of 131I-labeled antibodies in follicular
lymphoma Med. Phys. 19 97104
‘S’, absorbed dose per unit cumulated activity for selected radionuclides and organs
[55] Ford M R, Snyder W S, Dillman L T and Watson S B 1977 Maximum permissible
concentration (MPC) values for spontaneously fissioning radionuclides Health
Phys. 33 3543
from internal photon sources. I Methods, II One year old, III Five year old, IV Ten year
old, V Fifteen year old male and adult female, VI New-born and VII Adult male ORNL/
TM-8381/V1 to ORNL/TM-8381/V7 (Oak Ridge National Laboratory)
Fourier transform convolution J. Nucl. Med. 38 30814
[58] Lampinen J S, Pohjonen H K and Savolainen S E 1998 Calculating internal dose by
convolution from SPECT/MR fusion images Ann. Nucl. Med. 12 15
[59] Erdi A K et al. 1998 Use of the fast Hartley transform for three-dimensional dose
calculation in radionuclide therapy Med. Phys. 25 222633
[60] Bardiès M and Pihet P 2000 Dosimetry and microdosimetry of targeted radiotherapy
Curr. Pharm. Design 6 1469502
[61] ICRU 1983 Microdosimetry ICRU Report 36 (International Commission on Radia-
tion Units and Measurements)
Chapter 8
Radiobiology aspects and radionuclide

selection criteria in cancer therapy
Amin I Kassis
8.1. INTRODUCTION
The deposition of energy by ionizing radiation is a random process. Follow-

ing energy absorption by cells, certain molecular modifications will be
induced that may lead to cell death. Despite the fact that this process is
stochastic in nature, the death of a few cells, in general, within a tissue or
an organ will not have a significant effect on its function. As the dose
increases, more cells will die and eventually organ/tissue function will be
impaired.
8.2. RADIOBIOLOGICAL EFFECTS
8.2.1. Molecular lesions

Damage to the genome of cells is the basis for most radiation effects which
display themselves generally as single- and double-strand breaks within
DNA (SSB and DSB, respectively), base damage, and cross-links between
DNA strands and nuclear proteins [1]. Most cells types, however, are
equipped with a host of enzymes that are efficient at repairing damage to
DNA. To aid in their function, such cells are also outfitted with machinery
that holds up the mitotic cycle until repairs can be made. Repair of DNA
damage from low-linear-energy-transfer (low-LET) radiations (photons
and energetic electrons) is very efficient and sensitive to dose rate [2]. On
the other hand, the repair of damage from high-LET densely ionizing radia-
tion ( particles and Auger effects in DNA) is more complex.
Another manifestation of damage to DNA is the appearance of
chromosomal and chromatidal aberrations. Chromosome-type aberrations
175
176 Radiobiology aspects and radionuclide selection criteria
are induced in G1 -phase chromosomes or in unreplicated regions of S-phase

chromosomes. Chromatid-type aberrations are induced in replicated regions
of S-phase chromosomes and/or in G2 -phase chromosomes. The shape of the
dose–effect relationship observed for chromosomal aberrations from low-
LET radiation follows the general formula D þ D2 . With increasing
LET, the relationship goes from linear–quadratic to linear. Lowering the
dose rate also changes the low-LET aberration curve from the quadratic
type to one increasingly more linear.
8.2.2. Cellular responses
8.2.2.1. Survival
Several in vitro assays have been described to measure the ability of cells to
proliferate. In practice, these assays measure the capacity of cells to success-
fully reproduce and, thus, to form a colony. Using a colony-forming assay, it
is possible to determine the decrease in survival, expressed as a surviving
fraction, as a function of a graded radiation dose.
Radiation survival curves are log–linear plots of surviving fraction (log)
versus dose (linear). The shape of the survival curve constructed through
such a set of survival points varies and will depend on certain biological,
physical, and chemical factors. In general, two types of dose–survival
curve have been described (figure 8.1). For the exponential survival curve,
the slope is always constant and can be expressed by
S=S0 ¼ eD ð8:1Þ
where S=S0 is the surviving fraction of the irradiated cells, D is the dose
delivered, and ¼ 1=D0 , the dose needed to reduce survival to 0.37 (for this
type of curve, D0 is also known as D37 ). Such exponential survival curves
are observed when mammalian cells are exposed to ‘densely ionizing’ radiation
(e.g., particle emissions, DNA-incorporated Auger electron emitters).
In the second type of dose–response relationship, expressed by the sig-
moidal survival curve (figure 8.1), the efficiency of cell kill is not constant: at
low doses, a slow decrease in survival is observed and the curve has a shoulder;
at higher doses, an exponential decrease in survival is seen. This type of survival
curve occurs routinely when mammalian cells are exposed to sparsely ionizing
radiation (e.g., X-rays, particles, extranuclear Auger electrons). Such curves
can be fitted using a second-order polynomial equation:
S=S0 ¼ aD þ bD2 ð8:2Þ
where a and b are the fit parameters and solving for D when S=S0 ¼ 0:37
yields the D37 value. Alternatively, and preferably, the curve can be fitted
Radiobiological effects 177
Figure 8.1. Dose–response survival curves for mammalian cells in vitro.
using a linear–quadratic equation:

2
S=S0 ¼ eðD þ D Þ
ð8:3Þ
where equals the rate of cell kill by a single-hit mechanism, D is the dose
delivered, equals the rate of cell kill by a double-hit mechanism, and solving
for D when S=S0 ¼ 0:37 yields the D37 value. Rewriting equation (8.3)
provides a method for calculating the rate constants and by graphic
linearization [3]:
lnðS=S0 Þ
¼ D: ð8:4Þ
D
When lnðS=S0 Þ=D is plotted against dose D, a linear relationship is obtained
in which the y intercept with zero dose is the linear inactivation constant
and the slope of the line equals the
pffiffiffi quadratic inactivation constant . Note
that the reciprocals of and equal, respectively, D0 and D0 , the
latter representing the D0 value of the exponentially decreasing portion of
the curve. Such plots of cell inactivation can assist in making independent
components of such inactivations readily apparent, and they have signifi-
cantly improved the fit of data for mammalian cell survival curves and the
resolution of problems associated with the conceptual difficulty of zero
slope with zero dose. Note that if radiation is protracted or the dose rate is
low, as often occurs with radionuclides, the term predominates.
8.2.2.2. Division delay

Dividing cells pass through four phases that are based on two observable
events, mitosis (M) and DNA synthesis (S). Two gap phases that occur
before and after the S phase are known as G1 and G2 , respectively.
The delay in the progression of dividing cells through their cell cycle
following irradiation is a well known phenomenon. Usually it is reversible
and dose dependent, occurs only at specific points in the cell cycle, and is
similar for both surviving and nonsurviving cells. Cells undergoing mitosis
(M phase) continue through division basically undisturbed; those in the G1
phase of the cell cycle have very little delay, those in the S phase a moderate
delay, and those in the G2 phase maximum delay. The net result is that many
cells accumulate at the G2 =M boundary, and the ratio of mitotic cells to
nonmitotic cells (i.e., the mitotic index) is altered. The length of the delay
and the decrease in mitotic index are both functions of dose.
8.2.2.3. Redistribution
The radiosensitivity of cells is a function of their position within the cell cycle.
Cells in late S phase are most resistant (the survival curve following
irradiation has a wide shoulder), while those in the G2 and M phases are
most radiosensitive (the survival curves have a steep slope and no shoulder).
Consequently, following irradiation, the cells in the most sensitive phase
will be killed preferentially. This redistribution of cells will lead to partial
cell-cycle synchrony and a change in the overall radiosensitivity of the
cell cohort. One would expect a more radioresistant cell population, but in
reality the population rapidly becomes desynchronized and the net effect is
sensitization of the surviving population.
8.2.2.4. Repair
Mammalian cells are generally capable of repairing some of the damage
induced by radiation. This phenomenon is dose-rate dependent. As the
dose rate decreases, the capacity of cells to repair radiation damage increases
and this is manifested by a widening of the shoulder of the survival curve
accompanied by an increase in D0 (e.g., X-ray irradiation). In essence, two
types of repair have been described. The first, sublethal damage (SLD)
repair, occurs when a dose of radiation is split into two fractions and suffi-
cient time is allowed (0.5–2 h) for any/all of the radiation-induced damage
to be repaired. Naturally, if no repair is allowed to occur, for example, by
the immediate application of a second dose of radiation, the cells will die.
Sublethal damage and its repair have been shown to be important factors in
the sparing of normal tissues during fractionated radiotherapy. The second,
potentially lethal damage (PLD) repair, is observed only when mammalian
cells are grown under suboptimal conditions following a single dose of radia-
tion. Under such circumstances, an increase (rather than a decrease) in survival
is observed. This phenomenon is believed to be a result of the delayed entry of
the irradiated cells into mitosis, thereby allowing the cells to repair the PLD.
Both SLD and PLD as well as their repair have been reported only for X-ray
and irradiation (i.e., low-LET-type survival curves that have considerable
shoulders) and are practically nonexistent for neutron and particle irradiation
(i.e., high-LET radiation).
8.2.2.5. Oxygen effect

Oxygen radiosensitizes mammalian cells to the damaging effects of radiation.
Consequently, hypoxic cells can be as much as threefold more radioresistant
than well oxygenated cells. It is thought that following irradiation, oxygen
enhances free radical formation and/or blocks the reversal of certain reversible
chemical alterations that have occurred. Here again, it is important to note
that the oxygen effect is greatest for photons and high-energy particles
(low-LET-type survival curves) and is practically absent for particles and
neutrons (high-LET-type survival curves).
8.2.3. Tissue responses
8.2.3.1. Heterogeneity of response

Following irradiation, the response of an organ/tissue/tumour will depend to
a large degree on its inherent radiosensitivity. In the clinical situation, the
term maximum tolerated dose (MTD) is used to indicate the highest dose
that a normal organ can withstand. Once the MTD has been exceeded, the
patient will exhibit the particular signs and symptoms associated with
dysfunction of the organ being irradiated.
The MTD to irradiation of assorted tissues and organs is variable. The
tolerance doses for the gonads and bone marrow are low (100–200 cGy);
those for the intestine, kidneys and heart are moderate (2000–4500 cGy);
and those for mature bone and cartilage, the bladder and the central
nervous system are high (5000–7000 cGy). These doses are for external
beam therapy. With radionuclides, where the dose rate is much lower, the
MTD may be somewhat higher. Consequently, knowledge of the MTD for
the various organs and/or tissues being irradiated is important and can
help to predict with some certainty the risks associated with a radiothera-
peutic dose.
8.2.3.2. Protraction and fractionation effects

Radiotherapists have long realized that dividing the radiation dose into daily
or weekly fractions is more effective in eradicating tumours than giving a
single dose, while reducing the undesirable effects in normal tissues. The
size and number of the fractions, the treatment time and the total dose
given depend mainly on the radiosensitivity of the tumour being eliminated
and the tolerance of the surrounding normal tissues and organs. Since the
response of a tumour to irradiation is determined in part by the depopulation
of its dividing cells, it is expected that tumour cells react to dose fractionation
in a fashion similar to that of acutely responding rather than late responding

normal tissues, i.e., have high = ratios (higher ratios indicate that a lower
proportion of the damage will be repaired). In fact, Thames and Hendry [4]
and Williams et al. [5] have reported significant differences between the =
ratios for tumours and normal tissues. These authors have observed that while
growing tumours exhibit a predominant = ratio of 10 to 20, normal
tissues show a central tendency around 3 to 5. To the extent that such
animal data are relevant to human tumours, it is not surprising that dose
fractionation favours tumour control and minimizes normal tissue damage.
8.2.3.3. Bystander effects

Recent studies have demonstrated that a radiobiological phenomenon
termed ‘bystander effect’ can be observed in mammalian cells grown and
irradiated in vitro. Bystander damage describes biological effects, originating
from irradiated cells, in cells not directly affected by radiation-induced
ionizations. Investigators have reported that cells display lower survival
and a higher rate of genetic changes than would be predicted from direct-
ionization-only models [6–12] when a small fraction of a cell population
growing in vitro has been traversed by particles. Evidence from these
reports challenges the past half-century’s central tenet that radiation conveys
damage to DNA either through direct ionization or indirectly via, for
example, hydroxyl radicals produced in water molecules in the immediate
vicinity of DNA.
To determine whether a bystander effect could be demonstrated with an
in vivo system, we recently investigated the ability of tumour cells that are
labelled with lethal doses of DNA-incorporated 125 I to exert a radiobiological
effect on surrounding unlabelled tumour cells growing subcutaneously in nude
mice (Kassis et al., unpublished results). To this end, human colon adeno-
carcinoma LS174T cells, pre-incubated with lethal doses of the thymidine
analogue 5-[125 I]iodo-20 -deoxyuridine (125 IUdR), were co-injected with
unlabelled cells subcutaneously into nude mice, and tumour growth was
measured at various times to determine the tumour’s responsiveness to the
co-injected ‘dying’ 125 I-labelled cells. Our data demonstrate that these 125 I-
labelled cells have a substantial inhibitory effect on the growth of unlabelled
tumour cells (figure 8.2). Since the radiation dose deposited by the radio-
labelled cells into the growing tumour is less than 1/100th that needed to
inhibit/delay tumour growth, we conclude that the results are a consequence
of a bystander effect that is initiated by the decay of the DNA-incorporated
125
I atoms. These in vivo findings, together with those recently published by
Howell and co-investigators [8, 13] in which a bystander effect was observed
following the decay of DNA-incorporated tritium in vitro, should significantly
impact the current dogma for assessing the therapeutic potential of internally
administered radionuclides.
Figure 8.2. Tumour growth inhibition by the bystander effect. During the 15
day period, the presence of 0:2 106 125 I-labelled cells leads to the deposition
of 2 cGy only in the growing tumours (unpublished results).
8.2.4. Radiation quality

particle, particle and Auger electron emitters are three types of radio-
nuclide that can be used for tumour therapy (table 8.1). Particles from
these radioactive emissions produce tracks along which energy is transferred
to and deposited in biological matter. The intensity of the energy transfer
varies and depends on the energy, charge and mass of the traversing particle.
The term linear energy transfer (LET) describes the transfer of energy (e.g.,
keV) along the track traversed (e.g., micrometre) by the particle. The LET of
particle emitters, such as 131 I and 90 Y, is low and, depending on their
energy, the particles traverse several millimetres. On the other hand, the
LET of particles emitted from 211 At and 213 Bi and of Auger electrons
emitted by 125 I is high. Actually, for Auger emitters, the electrons produce
ionizations that are clustered around the point of decay (i.e., they are not
along a linear track). While the tracks of particles are several cell
diameters in length, those of Auger cascade electrons are localized at the
site of decay in a sphere with radius of several nanometres. The tracks at
low LET are sparsely ionizing and many (thousands) are needed to produce
a detectable biological effect (e.g., cell death). At high LET the distance
between ionizations becomes shorter and the tracks are dense so that
these particles are much more efficient at producing lethal effects. For
example, several particle traversals (fewer than five) through a cell nucleus
or approximately 50 to 100 DNA-incorporated 125 I decays are sufficient to
sterilize a cell.
Table 8.1. Physical characteristics of therapeutic radionuclides.
Decay mode Particles Energy Range
Electrons Medium to high 1–12 mm

(0.5–2.3 MeV)
Helium nuclei High 50–100 mm
(several MeV)
EC/IC Auger electrons Very low Several nm
(eV–keV)

EC ¼ electron capture; IC ¼ internal conversion.
8.3. TARGETING PRINCIPLES IN RADIONUCLIDE THERAPY
8.3.1. Choice of radionuclide
8.3.1.1. particles
Current radionuclide therapy in humans is based almost exclusively on -
particle-emitting isotopes. Typically, the electrons that are emitted from
the nuclei of the decaying atoms (1 electron/decay) have various energies
up to a maximum and thus have a distribution of ranges (table 8.2). As
each particle emitted traverses matter, it follows a contorted path, loses its
Table 8.2. Physical characteristics of particle emitters.
Radionuclide Half-life E R † E ðmaxÞ R ðmaxÞ †

(keV) (mm) (keV) (mm)
33
P 25.4 d 77 0.09 249 0.63
169
Er 9.4 d 99 0.14 350 1.1
177
Lu 6.7 d 133 0.23 497 1.8
67
Cu 61.9 h 141 0.26 575 2.1
131
I 8.0 d 182 0.39 610 2.3
153
Sm 46.8 h 224 0.54 805 3.3
198
Au 64.8 h 312 0.89 961 4.2
109
Pd 13.5 h 361 1.1 1028 4.5
186
Re 3.8 d 349 1.1 1077 4.8
165
Dy 2.3 h 440 1.5 1285 5.9
89
Sr 50.5 d 583 2.2 1491 7.0
32
P 14.3 d 695 2.8 1710 8.2
188
Re 17.0 h 764 3.1 2120 10.4
90
Y 64.1 h 935 4.0 2284 11.3

Mean (E ) and maximum (E ðmaxÞ ) energy of particles emitted per disintegration [14].
†
Mean (R ) and maximum (R ðmaxÞ ) range of particle in water [15].
Targeting principles in radionuclide therapy 183
kinetic energy and eventually comes to a stop. Since the LET of these light,
charged particles is very low (0.2 keV/mm), except for the few nanometres at
the end of their range just before they stop, they are sparsely ionizing and
quite inefficient at damaging DNA and killing cells. Consequently, their
use as therapeutic agents predicates the presence of high radionuclide
concentrations within the targeted tissue and the traversal of several thou-
sand electrons per mammalian cell nucleus.
An important ramification of the long range (mm) of each emitted
electron is the production of cross-fire, a circumstance that negates the need
to target every cell within the tumour. For microscopic disease, however,
long-range and some medium-range emitters may deposit a significant fraction
of the energy of their particles outside the tumour [16]. This is particularly of
concern in the selection of a radiopharmaceutical for the palliation of bone
pain from metastatic osseous lesions [17]. Long-range emitters (e.g., 32 P) run
the risk of significantly irradiating bone marrow as well as bony lesions,
whereas short-range emitters (e.g., 33 P) are calculated to result in a significantly
lower dose to bone marrow relative to bone/bony lesions.
The matter of inhomogeneity in the distribution of a radionuclide and
its consequent dose has also been addressed. O’Donoghue [18] has derived
a method for calculating an equivalent uniform biologically effective dose
based on the absorbed-dose distribution represented by the biologically
effective dose–volume histogram. For larger tumours he postulates that a
combination of radionuclide and external beam therapy may result in
optimal dose distribution.
Many of the -particle-emitting radionuclides used for therapy also
release photons that generally do not add significantly to the dose delivered
to the target tissue. However, these photons may contribute considerably to
the whole-body dose. For example, 3.7 GBq of 131 I distributed throughout
the whole body would deposit approximately 60 cGy per day. Since the
bone marrow is usually the dose-limiting organ (200–300 cGy), the success
or failure of therapy will depend on not exceeding its MTD.
8.3.1.2. particles
An particle that is emitted by any of the radionuclides suitable/available
for therapy (table 8.3) is identical to a helium nucleus with energies ranging
from 4 to 9 MeV. The corresponding tissue ranges are 40 to 100 mm.
particles travel in straight lines, initially depositing their energy, in soft
tissue, at approximately 80 keV/mm and increasing their rate of energy
deposition towards the end of their tracks. The typical energy deposition
from an particle traversing the diameter of a cell nucleus is about
600 keV; this translates to an absorbed dose of about 0.14 Gy to the cell
nucleus per hit. These numbers depend, among other parameters, on the
size of the cell nucleus.
Table 8.3. Physical characteristics of particle emitters.
Radionuclide Half-life E R †

(MeV) (mm)
211
At 7.2 h 6.79 60
212
Bi 60.6 min 7.80 75
213
Bi 45.7 min 8.32 84

Mean energy (E ) of particles emitted per disintegration [14].
†
Mean range (R ) of particles calculated using second order polynomial
regression fit [19]: R ¼ 3:87E þ 0:75E 2 0:45, where R is the range (mm) in
unit density matter and E is the particle energy (MeV).
Whereas the stochastics of energy distribution are unimportant for

absorbed doses greater than 1 Gy for particles because each cell experiences
thousands of individual particle tracks, the average specific energy deposited
per unit mass (i.e., the absorbed dose) is not a suitable parameter for cells
traversed by a few particles. Several authors have thus described micro-
dosimetric approaches for calculating the specific energy deposited in
individual cells and predicting the response of cells to -particle irradiation
[20, 21]; others (e.g., [22]) have related the specific energy distributions to
the fraction of cell survivors.
The magnitude of cross-dose (from radioactive sources associated with
one cell to an adjacent cell) is an important factor when evaluating particles
for therapy. This will vary considerably depending on the size of the labelled
cell cluster and the fraction of cells labelled [23].
Finally, when the particle emitter is covalently bound to the nuclear
DNA of tumour cells (i.e., in the case of cell self-irradiation), the contribu-
tion of heavy ion recoil of the daughter atom must also be considered in
assessing the radiobiological effects [24].
8.3.1.3. Auger electrons

During the decay of certain radioactive atoms, a vacancy is formed (most
commonly in the K shell) as a consequence of electron capture (EC), with
the prompt emission of a neutrino from the atomic nucleus and/or internal
conversion (IC). Such vacancies are rapidly filled by electrons dropping in
from higher shells. As a result of nuclear rearrangements, some Auger
electron emitters also emit a photon that may itself be converted to a
nuclear vacancy resulting in a second shower of Auger electrons (e.g., 125 I).
Similarly, other metastable nuclei (e.g., 99m Tc, 80m Br) emit a photon,
which may or may not be converted into an electron vacancy (only 3% of
the 140 keV photons of 99m Tc are converted). This process, creation of
an electron vacancy within a shell and its filling up, leads to a cascade of
atomic electron transitions that move the vacancy towards the outermost
Targeting principles in radionuclide therapy 185
Table 8.4. Physical characteristics of Auger electron emitters.
Radionuclide Half-life Electrons Average energy (eV)

per decay deposited in 5 nm
sphere
51
Cr 27.7 d 6 210
67
Ga 3.3 d 5 260
75
Se 120 d 7 270
99m
Tc 6h 4 280
77
Br 57 h 7 300
123
I 13.1 h 11 420
111
In 2.8 d 8 450
125
I 60.1 d 20 1000
201
Tl 73.1 h 20 1400
193m
Pt 4.3 d 26 1800
195m
Pt 4.0 d 33 2000

Average yield of Auger and Coster–Kronig electrons.
shell. Each inner-shell electron transition results in the emission of a charac-

teristic X-ray photon or an Auger, Coster–Kronig or super-Coster–Kronig
monoenergetic electron (collectively called Auger electrons). Typically an
atom undergoing EC and/or IC emits several Auger electrons with energies
ranging from a few eV to approximately 80 keV. Thus, the range of Auger
electrons in water is from a fraction of a nanometre to several tens of
micrometres. In some cases (e.g., 125 I, 195m Pt) as many as 20 to 33 Auger
electrons are emitted per decay on average (table 8.4). In addition to the
shower of low-energy electrons, this form of decay leaves the daughter
atom with a high positive charge resulting in subsequent charge transfer
processes.
The short range of Auger electrons requires their close proximity to
radiosensitive targets for radiotherapeutic effectiveness. This is essentially a
consequence of the precipitous drop in energy density as a function of
distance in nanometres (e.g., figure 1 in Kassis et al. [25] and figure 4 in
Kassis et al. [26]). These Auger electron emitters need to be located in the
cell nucleus, close to or incorporated into DNA.
8.3.2. Half-life
For a radiopharmaceutical with an infinite residence time in a tumour or
tissue, a radionuclide with a long physical half-life will deliver more decays
than one with a short half-life if both radiopharmaceuticals have the same
initial radioactivity. Moreover, there can be a striking difference in the
time-dependent dose rate delivered by the two. If the number of radionuclide
atoms per unit of tissue mass is n and the energy emitted (and absorbed) per
decay is E, then the absorbed-dose rate is proportional to nE=T where T is

the half-life. The ratio E=T is an important indicator of the intrinsic radio-
therapeutic potency of the radionuclide [27]. In general, for biological
reasons (e.g., repair of SLD), higher dose rates delivered over shorter treat-
ment times are more effective than lower dose rates delivered over longer
periods. Thus, a radionuclide with a shorter half-life will tend to be more
biologically effective than one with similar emission energy but a longer
half-life. In addressing this complicated and multifaceted phenomenon,
O’Donoghue and Wheldon [28] have noted that the relative effectiveness of
125
I and 123 I, when the same number of atoms of each is bound to a
tumour, depends on the tumour doubling time and the rate at which the
radiopharmaceutical dissociates from the target. When both are very long,
the longer-lived 125 I is theoretically more effective; otherwise the shorter-
lived 123 I is preferred.
8.3.3. Choice of vector or ligand

The selection of a suitable carrier molecule rests on many factors. These
include (i) the biological specificity and in vivo stability, (ii) the biological
mechanism(s) that bind it to target cells and the affinity of the carrier for
these sites, (iii) the stability of the complex thus formed, (iv) the chemical
properties of the carrier molecule which must permit the conjugation of a
therapeutic radionuclide without degradation of the intrinsic characteristics
of the molecule. Finally, the physical half-life of the radionuclide must be at
least equal to, and preferably much longer than, the biological half-life of the
carrier molecule.
One of the simplest targeting agents is radioiodine, administered as
iodide for treating functional cancer of the thyroid and, potentially, breast
cancers that display the iodine transporter. In the palliation of metastatic
bone pain, a simple cation, radiostrontium, has been used. Its targeting
relies on the accretion of alkaline-earth cation congeners of calcium on
bony surfaces. The oxides of radiophosphorus as orthophosphates and the
somewhat more complex phosphonates labelled with radiorhenium and
radiosamarium have also been proposed.
Short peptides labelled with 90 Y have been used to target neuroendo-
crine tumours [29]. Meta-[131 I]iodobenzylguanidine (M131 IBG), an iodinated
neurotransmitter, is taken up by neural crest tumours [30]. Radiocolloid
labelled with the particle emitter 211 At has been successfully employed in
treating ovarian ascites tumours in mice [31, 32].
Radioiodinated pyrimidine deoxyribose nucleosides (e.g., radio-
iododeoxyuridine) have been used for the experimental treatment of ovarian,
brain and spinal cord tumours in animals [33–36]. Because the carrier is
incorporated into the DNA of dividing cells, it is most effective when labelled
with Auger-electron-emitting radionuclides.
Experimental therapeutics 187
Successful therapy of lymphomas and leukaemias has been achieved

with intact antibodies and antibody fragments labelled with - and -
particle-emitting radionuclides [37]. Specificity in this case relies on
tumour-associated antigens located on the malignant cell surface.
8.3.4. Properties of targets

Critical tumour characteristics include (i) their accessibility, (ii) the number
of carrier-molecule binding sites and (iii) the distribution of binding sites
among the targeted and nontargeted cells and their relationship to the cell
cycle. The microscopic environment of the target, including tumour
vascularity, vascular permeability, oxygenation, microscopic organization
and architecture, is also important [38, 39].
The route of administration (e.g., intravenous, intralymphatic, intra-
peritoneal, intrathecal) must also be considered. Some pathways may
provide a mechanical means of maximizing tumour-to-nontumour ratios.
Finally, specific activity of the radiopharmaceutical should be taken into
account, especially when receptors can be easily saturated and weaker non-
specific binding competes with the targeting molecule. On the other hand,
certain treatments, including the use of radiolabelled antibodies, may be
assisted by a mass effect and optimized by the addition of unlabelled
immunoglobulin.
8.4. EXPERIMENTAL THERAPEUTICS
8.4.1. particle emitters

Investigators have assessed the in vitro toxicity of -particle-emitting
radionuclides for many years. Early work, carried out by radiobiologists,
concentrated on tritium and 14 C, whereas later studies examined the
therapeutic potential of more energetic particle emitters (table 8.2). In
these studies, the survival curves either have a distinct shoulder [40–42] or
are of the high-LET type [43–45]. Invariably, the D0 calculated is several
thousand decays. The decay of such radionuclides (e.g., tritium, 14 C and
131
I) has been shown to lead to various molecular alterations (e.g., SSB,
DSB, chromosomal aberrations), but only when the cells are exposed to
very large numbers of decays (10 000–150 000) [40, 46–48].
Despite the rather low in vitro toxicity of -particle-emitting radio-
nuclides, they continue to be pursued for targeted therapy. This is in part
due to their availability and also to the long range of the emitted electrons,
which can lead to the irradiation of all the cells that are within the maximum
range and path of the particle (i.e., cross-fire). As mentioned above, the main
advantage of cross-fire is that it negates the necessity of the radiotherapeutic
agent’s being present within each of the targeted cells, i.e., it permits a certain
degree of heterogeneity. However, three factors will mainly determine
whether the dose delivered to the targeted tissue will be therapeutically
effective: (i) it is essential that the radiotherapeutic agent concentrate
within foci throughout the targeted tissue, (ii) the distances between these
hot foci must be equal to or less than twice the maximum range of the emitted
energetic particles and (iii) the concentration of the radiotherapeutic agent
within each hot focus must be sufficiently high to produce a cumulative cross-
fire dose to the surrounding targeted cells of approximately 10 000 cGy. Since
the dose is inversely proportional to the square of the distance, it is important
to note that the concentration of the therapeutic agent needed to deposit such
cytocidal doses will decrease precipitously when the distance between the hot
foci decreases.
Experimentally, these predictions have been substantiated in various
animal-tumour therapy studies. In many of these studies, radionuclides
have been targeted after conjugation to monoclonal antibodies. For example,
investigators have assessed the therapeutic efficacy of 131 I-labelled mono-
clonal antibodies in rodents bearing subcutaneous tumours. Although a
substantial proportion of cells within a tumour mass show reduced/no
expression of the targeted antigen and, therefore, are not targeted by the
radioiodinated antibody, 131 I-labelled antibodies that localize in high
concentrations in tumours are therapeutically efficacious and can lead to
total tumour eradication in some instances [49–52]. Thus, even when 131 I is
not-so-uniformly distributed within a tumour, the decay of this radionuclide
can lead to tumour sterilization as long as it is present in sufficiently high
concentrations (figure 8.3). Similar results have also been reported with
other -particle-emitting isotopes, in particular 90 Y [42, 53, 54] and 67 Cu
[55, 56].
Several ‘two- and three-step’ approaches have also been proposed to
target radioactivity to a tumour [57–61]. In general, an antibody that is
not internalized by the targeted cell is injected prior to the administration
of a small radiolabelled molecule that has a strong affinity to the antibody.
Figure 8.3. Tumour growth post-treatment with 131 I-labelled monoclonal anti-
body [51].
Figure 8.4. Therapeutic activity of 90 Y-DOTA-biotin post monoclonal anti-

body–streptavidin conjugate administration in tumour-bearing mice [61]. l
and g: controls; T: 800 mCi 90 Y-DOTA-B following pretargeting.
The most extensively studied approach utilizes the high avidity of avidin (Av)
or streptavidin (SAv) for biotin, a vitamin found in low concentration
in blood and tissues. Because the noncovalently bound Av/SAv–biotin
complex has an extremely low dissociation constant (kd ) of about 1015 M,
investigators have used these two molecules as binding pairs to bridge
molecules that have no affinities for each other and to target various radio-
nuclides. A recent report [61] that examined the efficacy of this approach
in tumour-bearing mice has demonstrated that 100% of the mice treated
with the 90 Y-labelled biotin derivative have been cured of their disease
(figure 8.4).
8.4.2. particle emitters

The potential application of -particle-emitting radionuclides as targeted
therapeutic agents has been of interest for more than 25 years. If selectively
accumulated in the targeted tissues (e.g., tumours), the decay of such radio-
nuclides with their high LET and short range of a few (5–10) cell diameters
should result in highly localized energy deposition in the targeted tumour
cells and minimal irradiation of surrounding normal host tissues. As a
consequence of the LET of these particles (80 keV/mm), one to four traversals
of particles through a mammalian cell nucleus will kill the cell [21, 24, 62,
63]. In comparison, the LET of negatrons emitted by the decay of emitters is
very much smaller (0.2 keV/mm) and, thus, thousands of particles must
traverse a cell nucleus for sterilization of the cell.
The investigation of the therapeutic potential of particle emitters has
focused on three radionuclides: 211 At, 212 Bi and 213 Bi (table 8.3). In vitro
studies [21, 62] have demonstrated that the decrease in survival of mam-
malian cells after exposure to uniformly distributed particles from these
radionuclides is monoexponential (figure 8.5) but that, as predicted theoreti-
cally [64] and shown experimentally [24], these curves develop a tail when the
dose is nonuniform (figure 8.5). Such studies have shown that (i) these
Figure 8.5. Survival of mammalian cells exposed in suspension to 211 At-

astatide [62], 212 Bi-DTPA [21] or 211 AtUdR (only 50% of cells labelled)
[24]. Inset: theoretical expectations [64].
particle emitters are highly toxic [21, 62, 65–70], (ii) cells in monolayers
require more particle traversals than spherical cells [21, 62], (iii) only a
few (1–4) particle traversals through a mammalian cell nucleus are neces-
sary to sterilize a cell [21, 62, 70].
At the molecular level, the traversal of particles through a mammalian
cell nucleus leads to the efficient production of chromosomal aberrations and
DSB. For example, the incubation of mammalian cells with 211 At-astatide
causes a significant increase in chomosomal aberrations [62], but these
decline with the passage of time. More recently, Walicka et al. [24] have
established that more than ten DSBs are produced per decay of DNA-
incorporated 211 At, a value much higher than that obtained following the
decay of the DNA-incorporated Auger electron emitter 125 I [71, 72].
Investigators have also assessed the therapeutic potential of particle
emitters in tumour-bearing animals [31, 32, 68, 73–80]. Bloomer et al. [31,
32] have reported a dose-related prolongation in median survival when
mice bearing an intraperitoneal murine ovarian tumour are treated with
211
At-tellurium colloid administered directly into the peritoneal cavity.
While this -particle-emitting radiocolloid is curative without serious
morbidity, colloids of -particle-emitting radionuclides (32 P, 165 Dy, 90 Y)
are not (figure 8.6). In another set of in vivo studies examining the therapeutic
efficacy of a 212 Bi-labelled monoclonal antibody, the radionuclide is most
effective when used with a carrier having target specificity [75]. Finally, a
recent report by McDevitt and co-workers [80] has demonstrated that
225
Ac-labelled internalizing antibodies are therapeutically effective in mice
bearing solid prostate carcinoma or disseminated lymphoma. The targeting
ligand is radiolabelled with 225 Ac (T1=2 ¼ 10 days), a radionuclide that
decays to stable 209 Bi via a cascade of six daughters (221 Fr, 217 At, 213 Bi,
Figure 8.6. Percentage change in median survival of tumour-bearing mice

treated with - and -emitting radiocolloids [32].
213
Po, 209 Tl, 209 Pb) with the emission of five and three particles. While the
proposed approach seems promising, the use of 225 Ac-labelled radio-
pharmaceuticals in humans is very likely to generate high normal-tissue
toxicity (as a consequence of the particle recoil-energy-induced dissociation
of the five chemically different radioactive atoms from the targeting carrier)
post decay of the 225 Ac-labelled molecules that have not been internalized by
the targeted tumour cells.
8.4.3. Auger electron emitters

The toxicity to mammalian cells of radionuclides that decay by EC and/or IC
has, for the most part, been established with the Auger-electron-emitting
radionuclide 125 I. Because of its predominant (93%) IC decay following EC,
125
I is a prolific emitter of Auger electrons (mean of 20 per decaying atom
[81–84]). Dosimetric calculations have predicted that the electrons most
frequently produced by 125 I dissipate their energy in the immediate vicinity
of the decaying atom and deposit 106 to 109 rad/decay within a 2 nm sphere
around the decay site [26, 82, 85]. Thus, the biological effects are expected to
depend critically on the proximity of the radionuclide to DNA.
The radiotoxicity of 125 I has been studied in vitro under a number of condi-
tions including (i) when the radioelement is incorporated into the DNA duplex
as iododeoxyuridine (IUdR) [26, 40, 44, 86–89], (ii) when it is intercalated
between the stacked bases of DNA as 3-acetamido-5-[125 I]iodoproflavine
(A125 IP) [90, 91], (iii) when it is bound to the minor and major grooves of
DNA as an iodinated Hoechst dye [92], (iv) as a radiolabelled triplex-forming
oligonucleotide [93], (v) when it is bound to transcription elements as iodinated
tamoxifen and oestrogen [94, 95] and (vi) when it is located in the cytoplasm of
cultured cells as iodorhodamine (125 I-DR) [96].
Figure 8.7. Survival of cells following exposure to 125 I decays: DNA-incorpo-

rated [26] or within cytoplasm [96]. The survival curve for 131 IUdR is included
for comparison [41].
The decay of 125 I incorporated into cellular DNA as IUdR leads to an

exponential decrease in clonal survival (figure 8.7). Similar responses are
seen with 77 Br [25] and 123 I [97]. When plotted as a function of the total
number of decays, the three curves have different slopes. If the dose to the
cell nucleus is calculated, a single cell-survival curve is obtained (figure
8.8), suggesting that the radiation dose to the cell nucleus is a parameter
that describes these observed biological effects adequately. When the inter-
calating agent A125 IP is incubated with cells, the survival curve is still
exponential, but each decay is 1.5 times less effective than that of 125 IUdR,
i.e., its relative biological effectiveness (RBE) is lower (figure 8.8). On the
other hand, a recent study of survival on exposure to an 125 I-labelled
oestrogen ligand reports a D0 similar to 125 IUdR [98].
Figure 8.8. Survival fraction of V79 cells plotted as function of radiation dose
to cell nucleus. RBE of each agent is calculated at the D37 [90].
Contrary to these survival curves, incubation of mammalian cells with

an 125 I-labelled DNA groove binder (Hoechst 33342) results in a linear–
quadratic survival curve [92]. A similar low-LET-like survival curve is also
seen when 125 I decays within the cytoplasm (figure 8.8). Furthermore, the
RBE of these agents is even lower, i.e., many more decays are required for
equivalent toxicity [96].
From these experiments, it has been concluded that the position of the
Auger electron emitter with regard to DNA is a major determinant of radio-
toxicity and that under certain circumstances decays result in a high-LET-
like response. The amount of energy deposited in the nucleus is also
important, and decays taking place in the cytoplasm are much less effective.
Molecular studies have shown that the decay of 125 I in iodouracil leads
to the creation of carbon fragments from the pyrimidine ring [99, 100]. The
incorporation of the radionuclide into duplex oligonucleotides leads to
shattering of the nucleic acid strands [101]. Similarly, its binding to the
minor groove of plasmids results in a high yield of DSB [102, 103]. Whereas
such studies have predicted correctly that the decay of DNA-incorporated
125
I would be highly toxic to mammalian cells, they have assumed that the
radiotoxicity of 125 IUdR would be due solely to the deposition of energy
directly into the DNA molecule. Experiments with radical scavenging
agents have established that this is not the case and that clusters of watery
radicals, formed following the decay of 125 I within chromatin, damage not
only DNA at the site of decay but also a number of distant bases due to
the packaging and compaction of DNA [71, 72]. As a result, the cyotoxicity
of 125 IUdR in mammalian cells can be modified strikingly by radical
scavengers [104–106]. Indeed, it now seems likely that indirect effects
caused by clusters of aqueous radicals are the principal mechanism of cell
damage.
The extreme degree of cytotoxicity observed with DNA-incorporated
125
I has been exploited in experimental radionuclide therapy using
125
IUdR. Since this radiopharmaceutical breaks down rapidly after systemic
administration, studies have been carried out in animal models and in
patients where locoregional administration is feasible. For example, the
injection of 125 IUdR into mice bearing an intraperitoneal ascites ovarian
cancer has led to a 5-log reduction in tumour cell survival [33]. Similar results
are obtained with 123 IUdR [34]. When 125 IUdR is administered intra-
cerebrally to rats bearing an intraparenchymal gliosarcoma, the survival of
treated animals is significantly prolonged [35]. Therapeutic doses of
125
IUdR injected intrathecally into rats with intrathecal 9L-rat-gliosarcoma
tumours significantly delay the onset of paralysis in these animals [107].
Most recently, Kassis et al. [36, 108] have shown that the therapeutic
efficacy of 125 IUdR in rats bearing intrathecal human TE-671 rhabdomyo-
sarcoma is substantially enhanced by the co-administration of methotrexate,
an antimetabolite that enhances IUdR uptake by DNA-synthesizing cells
Figure 8.9. Therapy of rats bearing human intrathecal rhabdomyosarcoma.

125
IUdR methotrexate was injected intrathecally [108].
(figure 8.9). This chemo–radio combination in fact leads to 5–6-log kill and
cures approximately 30% of the animals.
REFERENCES
[1] Ward J F 1986 Mechanisms of DNA repair and their potential modification for
radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 12 1027–32
[2] Goodhead D T 1985 Saturable repair models of radiation action in mammalian cells
Radiat. Res. 104 S-58–67
[3] Chapman J D 1980 Biophysical models of mammalian cell inactivation by radiation.
In Radiation Biology in Cancer Research eds R E Meyn and H R Withers (New York:
Raven) pp 21–32
[4] Thames H D and Hendry J H 1987 Fractionation in Radiotherapy (London: Taylor
and Francis)
[5] Williams M V, Denekamp J and Fowler J F 1985 A review of = ratios for
experimental tumors: implications for clinical studies of altered fractionation Int.
J. Radiat. Oncol. Biol. Phys. 11 87–96
[6] Nagasawa H and Little J B 1992 Induction of sister chromatid exchanges by extremely
low doses of -particles Cancer Res. 52 6394–6
[7] Azzam E I, de Toledo S M, Gooding T and Little J B 1998 Intercellular
communication is involved in the bystander regulation of gene expression in
human cells exposed to very low fluences of alpha particles Radiat. Res. 150 497–504
[8] Bishayee A, Rao D V and Howell R W 1999 Evidence for pronounced bystander
effects caused by nonuniform distributions of radioactivity using a novel three-
dimensional tissue culture model Radiat. Res. 152 88–97
[9] Nagasawa H and Little J B 1999 Unexpected sensitivity to the induction of
mutations by very low doses of alpha-particle radiation: evidence for a bystander
effect Radiat. Res. 152 552–7
References 195
[10] Zhou H, Randers-Pehrson G, Waldren C A, Vannais D, Hall E J and Hei T K 2000

Induction of a bystander mutagenic effect of alpha particles in mammalian cells
Proc. Natl Acad. Sci. USA 97 2099–104
[11] Azzam E I, de Toledo S M and Little J B 2001 Direct evidence for the participation
of gap junction-mediated intercellular communication in the transmission of damage
signals from -particle irradiated to nonirradiated cells Proc. Natl Acad. Sci. USA
98 473–8
[12] Brenner D J, Little J B and Sachs R K 2001 The bystander effect in radiation
oncogenesis: II. A quantitative model Radiat. Res. 155 402–8
[13] Bishayee A, Hall H Z, Stein D, Rao D V and Howell R W 2001 Free radical-initiated
and gap junction-mediated bystander effect due to nonuniform distribution of
incorporated radioactivity in a three-dimensional tissue culture model Radiat. Res.
155 335–44
[14] Kocher D C 1981 Radioactive Decay Data Tables: A Handbook of Decay Data for
Application to Radiation Dosimetry and Radiological Assessments DOE/TIC-11026
(Springfield, VA: National Technical Information Center, US Department of Energy)
[15] ICRU 1984 Stopping Powers for Electrons and Positrons Report 37 (Bethesda, MD:
International Commission on Radiation Units and Measurements) p 206
[16] Wheldon T E 1994 Targeting radiation to tumours Int. J. Radiat. Biol. 65 109–16
[17] Bouchet L G, Bolch W E, Goddu S M, Howell R W and Rao D V 2000 Considera-
tions in the selection of radiopharmaceuticals for palliation of bone pain from
metastatic osseous lesions J. Nucl. Med. 41 682–7
[18] O’Donoghue J A 1999 Implications of nonuniform tumor doses for radioimmuno-
therapy J. Nucl. Med. 40 1337–41
[19] ICRU 1993 Stopping Powers and Ranges for Protons and Alpha Particles Report 49
(Bethesda, MD: International Commission on Radiation Units and Measurements)
p 256
[20] Charlton D E and Sephton R 1991 A relationship between microdosimetric spectra
and cell survival for high-LET irradiation Int. J. Radiat. Biol. 59 447–57
[21] Charlton D E, Kassis A I and Adelstein S J 1994 A comparison of experimental and
calculated survival curves for V79 cells grown as monolayers or in suspension
exposed to alpha irradiation from 212Bi distributed in the growth medium Radiat.
Prot. Dosim. 52 311–15
[22] Roeske J C and Stinchcomb T G 1999 The use of microdosimetric moments
in evaluating cell survival for therapeutic alpha-particle emitters Radiat. Res. 151
31–8
[23] Goddu S M, Howell R W and Rao D V 1994 Cellular dosimetry: absorbed fractions
for monoenergetic electron and alpha particle sources and S-values for radionuclides
uniformly distributed in different cell compartments J. Nucl. Med. 35 303–16
[24] Walicka M A, Vaidyanathan G, Zalutsky M R, Adelstein S J and Kassis A I 1998
Survival and DNA damage in Chinese hamster V79 cells exposed to alpha
particles emitted by DNA-incorporated astatine-211 Radiat. Res. 150 263–8
[25] Kassis A I, Adelstein S J, Haydock C, Sastry K S R, McElvany K D and Welch M J
1982 Lethality of Auger electrons from the decay of bromine-77 in the DNA of
mammalian cells Radiat. Res. 90 362–73
[26] Kassis A I, Sastry K S R and Adelstein S J 1987 Kinetics of uptake, retention, and
radiotoxicity of 125IUdR in mammalian cells: implications of localized energy
deposition by Auger processes Radiat. Res. 109 78–89
[27] O’Donoghue J A 1994 The impact of tumor cell proliferation in radioimmuno-

therapy Cancer 73 974–80
[28] O’Donoghue J A and Wheldon T E 1996 Targeted radiotherapy using Auger
electron emitters Phys. Med. Biol. 41 1973–92
[29] Smith M C, Liu J, Chen T, Schran H, Yeh C-M, Jamar F, Valkema R, Bakker W,
Kvols L, Krenning E and Pauwels S 2000 OctreoTher: ongoing early clinical
development of a somatostatin-receptor-targeted radionuclide antineoplastic
therapy Digestion 62 (suppl 1) 69–72
[30] Loh K-C, Fitzgerald P A, Matthay K K, Yeo P P B and Price D C 1997 The treat-
ment of malignant pheochromocytoma with iodine-131 metaiodobenzylguanidine
(131I-MIBG): a comprehensive review of 116 reported patients J. Endocrinol.
Invest. 20 648–58
[31] Bloomer W D, McLaughlin W H, Neirinckx R D, Adelstein S J, Gordon P R, Ruth
T J and Wolf A P 1981 Astatine-211–tellurium radiocolloid cures experimental
malignant ascites Science 212 340–1
[32] Bloomer W D, McLaughlin W H, Lambrecht R M, Atcher R W, Mirzadeh S,
Madara J L, Milius R A, Zalutsky M R, Adelstein S J and Wolf A P 1984 211At
radiocolloid therapy: further observations and comparison with radiocolloids of
32
P, 165Dy, and 90Y Int. J. Radiat. Oncol. Biol. Phys. 10 341–8
[33] Bloomer W D and Adelstein S J 1977 5-125I-iododeoxyuridine as prototype for
radionuclide therapy with Auger emitters Nature 265 620–1
[34] Baranowska-Kortylewicz J, Makrigiorgos G M, Van den Abbeele A D, Berman R M,
Adelstein S J and Kassis A I 1991 5-[123I]iodo-20 -deoxyuridine in the radiotherapy of
an early ascites tumor model Int. J. Radiat. Oncol. Biol. Phys. 21 1541–51
[35] Kassis A I, Wen P Y, Van den Abbeele A D, Baranowska-Kortylewicz J, Makrigiorgos
G M, Metz K R, Matalka K Z, Cook C U, Sahu S K, Black P M and Adelstein S J 1998
5-[125I]iodo-20 -deoxyuridine in the radiotherapy of brain tumors in rats J. Nucl. Med.
39 1148–54
[36] Kassis A I, Dahman B A and Adelstein S J 2000 In vivo therapy of neoplastic
meningitis with methotrexate and 5-[125I]iodo-20 -deoxyuridine Acta Oncol. 39
731–7
[37] Foon K A 2000 Monoclonal antibody therapies for lymphomas Cancer J. 6 273–8
[38] Jain R K 1990 Tumor physiology and antibody delivery Front. Radiat. Ther. Oncol.
24 32–46
[39] Netti P A, Hamberg L M, Babich J W, Kierstead D, Graham W, Hunter G J, Wolf
G L, Fischman A, Boucher Y and Jain R K 1999 Enhancement of fluid filtration
across tumor vessels: implication for delivery of macromolecules Proc. Natl Acad.
Sci. USA 96 3137–42
[40] Chan P C, Lisco E, Lisco H and Adelstein S J 1976 The radiotoxicity of iodine-125 in
mammalian cells. II. A comparative study on cell survival and cytogenetic responses
to 125IUdR, 131IUdR, and 3HTdR Radiat. Res. 67 332–43
[41] Burki H J, Koch C and Wolff S 1977 Molecular suicide studies of 125I and 3H
disintegration in the DNA of Chinese hamster cells Curr. Top. Radiat. Res. Q. 12
408–25
[42] Govindan S V, Goldenberg D M, Elsamra S E, Griffiths G L, Ong G L, Brechbiel M W,
Burton J, Sgouros G and Mattes M J 2000 Radionuclides linked to a CD74 antibody
as therapeutic agents for B-cell lymphoma: comparison of Auger electron emitters
with -particle emitters J. Nucl. Med. 41 2089–97
References 197
[43] Ragni G and Szybalski W 1962 Molecular radiobiology of human cell lines II.
Effects of thymidine replacement by halogenated analogues on cell inactivation by
decay of incorporated radiophosphorus J. Mol. Biol. 4 338–46
[44] Burki H J, Roots R, Feinendegen L E and Bond V P 1973 Inactivation of
mammalian cells after disintegrations of 3H or 125I in cell DNA at 1968C Int. J.
Radiat. Biol. 24 363–75
[45] Liber H L, LeMotte P K and Little J B 1983 Toxicity and mutagenicity of X-rays and
[125I]dUrd or [3H]TdR incorporated in the DNA of human lymphoblast cells Mutat.
Res. 111 387–404
[46] Cleaver J E and Burki H J 1974 Biological damage from intranuclear carbon-14
decays: DNA single-strand breaks and repair in mammalian cells Int. J. Radiat.
Biol. 26 399–403
[47] Gutiérrez S, Carbonell E, Galofré P, Creus A and Marcos R 1998 The alkaline single-
cell gel electrophoresis (SCGE) assay applied to the analysis of radiation-induced
DNA damage in thyroid cancer patients treated with 131I Mutat. Res. 413 111–19
[48] Hengstler J G, Bockisch A, Fuchs J, Grimm W, Görges R, Oesch-Bartlomowicz B,
Zapf A-O, Lade K, Tanner B, Teichmann E, Thelen M, Gebhard S and Oesch F
2000 Induction of DNA single-strand breaks by 131I and 99mTc in human
mononuclear blood cells in vitro and extrapolation to the in vivo situation Radiat.
Res. 153 512–20
[49] Esteban J M, Schlom J, Mornex F and Colcher D 1987 Radioimmunotherapy of
athymic mice bearing human colon carcinomas with monoclonal antibody B72.3:
histological and autoradiographic study of effects on tumors and normal organs
Eur. J. Cancer Clin. Oncol. 23 643–55
[50] Riklund K E, Makiya R A, Sundström B E, Thornell L-E and Stigbrand T I 1990
Experimental radioimmunotherapy of HeLa tumours in nude mice with 131I-
labeled monoclonal antibodies Anticancer Res. 10 379–84
[51] Buchsbaum D, Khazaeli M B, Liu T, Bright S, Richardson K, Jones M and Meredith
R 1995 Fractionated radioimmunotherapy of human colon carcinoma xenografts
with 131I-labeled monoclonal antibody CC49 Cancer Res. 55 (suppl) 5881s–7s
[52] Sato N, Saga T, Sakahara H, Yao Z, Nakamoto Y, Zhang M, Kuroki M, Matsuoka
Y, Iida Y and Konishi J 1999 Intratumoral distribution of radiolabeled antibody
and radioimmunotherapy in experimental liver metastases model of nude mouse
J. Nucl. Med. 40 685–92
[53] Otte A, Mueller-Brand J, Dellas S, Nitzsche E U, Herrmann R and Maecke H R
1998 Yttrium-90-labelled somatostatin-analogue for cancer treatment Lancet 351
417–18
[54] Chinn P C, Leonard J E, Rosenberg J, Hanna N and Anderson D R 1999 Preclinical
evaluation of 90Y-labeled anti-CD20 monoclonal antibody for treatment of non-
Hodgkin’s lymphoma Int. J. Oncol. 15 1017–25
[55] DeNardo G L, Kukis D L, DeNardo S J, Shen S, Mausner L F, O’Donnell R T,
Lamborn K R, Meyers F J, Srivastava S C and Miers L A 1997 Enhancement of
67
Cu-2IT-BAT-Lym-1 therapy in mice with human Burkitt’s lymphoma (Raji)
using interleukin-2 Cancer 80 2576–82
[56] DeNardo G L, Kukis D L, Shen S, Mausner L F, Meares C F, Srivastava S C, Miers
L A and DeNardo S J 1997 Efficacy and toxicity of 67Cu-2IT-BAT-Lym-1
radioimmunoconjugate in mice implanted with human Burkitt’s lymphoma (Raji)
Clin. Cancer Res. 3 71–9
[57] Hnatowich D J, Virzi F and Rusckowski M 1987 Investigations of avidin and biotin
for imaging applications J. Nucl. Med. 28 1294–302
[58] Goodwin D A, Meares C F, McCall M J, McTigue M and Chaovapong W 1988 Pre-
targeted immunoscintigraphy of murine tumors with indium-111-labeled bifunc-
tional haptens J. Nucl. Med. 29 226–34
[59] Le Doussal J-M, Gruaz-Guyon A, Martin M, Gautherot E, Delaage M and Barbet J
1990 Targeting of indium 111-labeled bivalent hapten to human melanoma mediated
by bispecific monoclonal antibody conjugates: imaging of tumors hosted in nude
mice Cancer Res. 50 3445–52
[60] Kassis A I, Jones P L, Matalka K Z and Adelstein S J 1996 Antibody-dependent
signal amplification in tumor xenografts following pretreatment with biotinylated
monoclonal antibody and avidin or streptavidin J. Nucl. Med. 37 343–52
[61] Axworthy D B, Reno J M, Hylarides M D, Mallett R W, Theodore L J, Gustavson L
M, Su F-M, Hobson L J, Beaumier P L and Fritzberg A R 2000 Cure of human
carcinoma xenografts by a single dose of pretargeted yttrium-90 with negligible
toxicity Proc. Natl Acad. Sci. USA 97 1802–7
[62] Kassis A I, Harris C R, Adelstein S J, Ruth T J, Lambrecht R and Wolf A P 1986
The in vitro radiobiology of astatine-211 decay Radiat. Res. 105 27–36
[63] Raju M R, Eisen Y, Carpenter S and Inkret W C 1991 Radiobiology of particles III.
Cell inactivation by -particle traversals of the cell nucleus Radiat. Res. 128 204–9
[64] Humm J L, Chin L M, Cobb L and Begent R 1990 Microdosimetry in
radioimmunotherapy Radiat. Prot. Dosim. 31 433–6
[65] Kurtzman S H, Russo A, Mitchell J B, DeGraff W, Sindelar W F, Brechbiel M W,
Gansow O A, Friedman A M, Hines J J, Gamson J and Atcher R W 1988 212Bismuth
linked to an antipancreatic carcinoma antibody: model for alpha-particle-emitter
radioimmunotherapy J. Natl Cancer Inst. 80 449–52
[66] Larsen R H, Bruland Ø S, Hoff P, Alstad J, Lindmo T and Rofstad E K 1994
Inactivation of human osteosarcoma cells in vitro by 211At-TP-3 monoclonal
antibody: comparison with astatine-211-labeled bovine serum albumin, free
astatine-211 and external-beam X rays Radiat. Res. 139 178–84
[67] Strickland D K, Vaidyanathan G and Zalutsky M R 1994 Cytotoxicity of -particle-
emitting m-[211At]astatobenzylguanidine on human neuroblastoma cells Cancer Res.
54 5414–9
[68] Larsen R H and Bruland Ø S 1998 Intratumour injection of immunoglobulins
labelled with the -particle emitter 211At: analyses of tumour retention,
microdistribution and growth delay Br. J. Cancer 77 1115–22
[69] Palm S, Bäck T, Claesson I, Delle U, Hultborn R, Jacobsson L, Köpf I and Lindegren S
1998 Effects of the alpha-particle emitter At-211 and low-dose-rate gamma-radiation
on the human cell line Colo-205 as studied with a growth assay Anticancer Res. 18
1671–6
[70] Nikula T K, McDevitt M R, Finn R D, Wu C, Kozak R W, Garmestani K, Brechbiel M
W, Curcio M J, Pippin C G, Tiffany-Jones L, Geerlings M W Sr, Apostolidis C, Molinet
R, Geerlings M W Jr, Gansow O A and Scheinberg D A 1999 Alpha-emitting bismuth
cyclohexylbenzyl DTPA constructs of recombinant humanized anti-CD33 antibodies:
pharmacokinetics, bioactivity, toxicity and chemistry J. Nucl. Med. 40 166–76
[71] Walicka M A, Adelstein S J and Kassis A I 1998 Indirect mechanisms contribute to
biological effects produced by decay of DNA-incorporated iodine-125 in
mammalian cells in vitro: double-strand breaks Radiat. Res. 149 134–41
References 199
[72] Kassis A I, Walicka M A and Adelstein S J 2000 Double-strand break yield

following 125I decay: effects of DNA conformation Acta Oncol. 39 721–6
[73] Kozak R W, Atcher R W, Gansow O A, Friedman A M, Hines J J and Waldmann T
A 1986 Bismuth-212-labeled anti-Tac monoclonal antibody: -particle-emitting
radionuclides as modalities for radioimmunotherapy Proc. Natl Acad. Sci. USA
83 474–8
[74] Harrison A and Royle L 1987 Efficacy of astatine-211-labeled monoclonal antibody
in treatment of murine T-cell lymphoma Natl Cancer Inst. Monogr. 3 157–8
[75] Macklis R M, Kinsey B M, Kassis A I, Ferrara J L M, Atcher R W, Hines J J,
Coleman C N, Adelstein S J and Burakoff S J 1988 Radioimmunotherapy with
alpha-particle-emitting immunoconjugates Science 240 1024–6
[76] Link E M and Carpenter R N 1990 211At-methylene blue for targeted radiotherapy
of human melanoma xenografts: treatment of micrometastases Cancer Res. 50
2963–7
[77] Zalutsky M R, McLendon R E, Garg P K, Archer G E, Schuster J M and Bigner D
D 1994 Radioimmunotherapy of neoplastic meningitis in rats using an -particle-
emitting immunoconjugate Cancer Res. 54 4719–25
[78] Kennel S J and Mirzadeh S 1998 Vascular targeted radioimmunotherapy with
213
Bi—an -particle emitter Nucl. Med. Biol. 25 241–6
[79] Adams G P, Shaller C C, Chappell L L, Wu C, Horak E M, Simmons H H, Litwin S,
Marks J D, Weiner L M and Brechbiel M W 2000 Delivery of the -emitting
radioisotope bismuth-213 to solid tumors via a single-chain Fv and diabody
molecules Nucl. Med. Biol. 27 339–46
[80] McDevitt M R, Ma D, Lai L T, Simon J, Borchardt P, Frank R K, Wu K, Pellegrini
V, Curcio M J, Miederer M, Bander N H and Scheinberg D A 2001 Tumor therapy
with targeted atomic nanogenerators Science 294 1537–40
[81] Charlton D E and Booz J 1981 A Monte Carlo treatment of the decay of 125I Radiat.
Res. 87 10–23
[82] Sastry K S R and Rao D V 1984 Dosimetry of low energy electrons. In Physics of
Nuclear Medicine: Recent Advances eds D V Rao, R Chandra and M C Graham
(Woodbury, NY: American Institute of Physics) pp 169–208
[83] Pomplun E, Booz J and Charlton D E 1987 A Monte Carlo simulation of Auger
cascades Radiat. Res. 111 533–52
[84] Charlton D E, Pomplun E and Booz J 1987 Some consequences of the Auger
effect: fluorescence yield, charge potential, and energy imparted Radiat. Res. 111
553–64
[85] Kassis A I, Sastry K S R and Adelstein S J 1985 Intracellular localisation of Auger
electron emitters: biophysical dosimetry Radiat. Prot. Dosim. 13 233–6
[86] Hofer K G and Hughes W L 1971 Radiotoxicity of intranuclear tritium, 125iodine
and 131iodine Radiat. Res. 47 94–109
[87] Porteous D D 1971 The toxicity of 125IUdR in cultured mouse BP8 tumour cells Br.
J. Cancer 25 594–7
[88] Bradley E W, Chan P C and Adelstein S J 1975 The radiotoxicity of iodine-125 in
mammalian cells. I. Effects on the survival curve of radioiodine incorporated into
DNA Radiat. Res. 64 555–63
[89] Hofer K G, Harris C R and Smith J M 1975 Radiotoxicity of intracellular 67Ga, 125I
and 3H: nuclear versus cytoplasmic radiation effects in murine L1210 leukaemia Int.
J. Radiat. Biol. 28 225–41
[90] Kassis A I, Howell R W, Sastry K S R and Adelstein S J 1988 Positional effects of

Auger decays in mammalian cells in culture. In DNA Damage by Auger Emitters eds
K F Baverstock and D E Charlton (London: Taylor and Francis) pp 1–13
[91] Kassis A I, Fayad F, Kinsey B M, Sastry K S R and Adelstein S J 1989 Radio-
toxicity of an 125I-labeled DNA intercalator in mammalian cells Radiat. Res. 118
283–94
[92] Walicka M A, Ding Y, Roy A M, Harapanhalli R S, Adelstein S J and Kassis A I
1999 Cytotoxicity of [125I]iodoHoechst 33342: contribution of scavengeable effects
Int. J. Radiat. Biol. 75 1579–87
[93] Sedelnikova O A, Panyutin I G, Thierry A R and Neumann R D 1998 Radiotoxicity
of iodine-125-labeled oligodeoxyribonucleotides in mammalian cells J. Nucl. Med.
39 1412–18
[94] Bloomer W D, McLaughlin W H, Milius R A, Weichselbaum R R and Adelstein S J
1983 Estrogen receptor-mediated cytotoxicity using iodine-125 J. Cell. Biochem. 21
39–45
[95] Yasui L S, Hughes A and DeSombre E R 2001 Cytotoxicity of 125I-oestrogen decay
in non-oestrogen receptor-expressing human breast cancer cells, MDA-231 and
oestrogen receptor-expressing MCF-7 cells Int. J. Radiat. Biol. 77 955–62
[96] Kassis A I, Fayad F, Kinsey B M, Sastry K S R, Taube R A and Adelstein S J 1987
Radiotoxicity of 125I in mammalian cells Radiat. Res. 111 305–18
[97] Makrigiorgos G M, Kassis A I, Baranowska-Kortylewicz J, McElvany K D, Welch
M J, Sastry K S R and Adelstein S J 1989 Radiotoxicity of 5-[123I]iodo-20 -
deoxyuridine in V79 cells: a comparison with 5-[125I]iodo-20 -deoxyuridine Radiat.
Res. 118 532–44
[98] Yasui L S, Hughes A and DeSombre E R 2001 Relative biological effectiveness of
accumulated 125IdU and 125I-estrogen decays in estrogen receptor-expressing
MCF-7 human breast cancer cells Radiat. Res. 155 328–34
[99] Halpern A and Stöcklin G 1977 Chemical and biological consequences of -decay.
Part 1 Radiat. Environ. Biophys. 14 167–83
[100] Halpern A and Stöcklin G 1977 Chemical and biological consequences of -decay.
Part 2 Radiat. Environ. Biophys. 14 257–74
[101] Martin R F and Haseltine W A 1981 Range of radiochemical damage to DNA with
decay of iodine-125 Science 213 896–8
[102] Kassis A I, Harapanhalli R S and Adelstein S J 1999 Comparison of strand breaks in
plasmid DNA after positional changes of Auger electron-emitting iodine-125 Radiat.
Res. 151 167–76
[103] Kassis A I, Harapanhalli R S and Adelstein S J 1999 Strand breaks in plasmid DNA
after positional changes of Auger electron-emitting iodine-125: direct compared to
indirect effects Radiat. Res. 152 530–8
[104] Walicka M A, Adelstein S J and Kassis A I 1998 Indirect mechanisms contribute to
biological effects produced by decay of DNA-incorporated iodine-125 in
mammalian cells in vitro: clonogenic survival Radiat. Res. 149 142–6
[105] Bishayee A, Rao D V, Bouchet L G, Bolch W E and Howell R W 2000 Protection by
DMSO against cell death caused by intracellularly localized iodine-125, iodine-131
and polonium-210 Radiat. Res. 153 416–27
[106] Walicka M A, Ding Y, Adelstein S J and Kassis A I 2000 Toxicity of DNA-
incorporated iodine-125: quantifying the direct and indirect effects Radiat. Res.
154 326–30
References 201
[107] Sahu S K, Wen P Y C, Foulon C F, Nagel J S, Black P M, Adelstein S J and Kassis

A I 1997 Intrathecal 5-[125I]iodo-20 -deoxyuridine in a rat model of leptomeningeal
metastases J. Nucl. Med. 38 386–90
[108] Kassis A I, Dahman B A and Adelstein S J 2000 In vivo therapy of neoplastic
meningitis with methotrexate and 5-[125I]iodo-20 -deoxyuridine Acta Oncol. 39 731–7
Chapter 9
Microdosimetry of targeted radionuclides

John C Roeske and John L Humm
9.1. INTRODUCTION
Microdosimetry is the study of the stochastic process of energy deposition

from radiation within small volumes. Small in the context of microdosimetry
typically refers to tissue equivalent volumes commensurate with the size of
the biological targets of relevance. This has commonly meant spherical
volumes of between 0.3 and 20 mm—the realm in which energy deposition
can be experimentally measured using a proportional counter. Such
chambers are constructed of tissue equivalent plastic and operated with a
tissue equivalent gas. Since these devices, used for the purpose of measuring
microdosimetric spectra, were developed by Rossi [1] they are frequently
referred to as ‘Rossi chambers’. One of the greatest applications of the
Rossi chambers is to characterize unknown mixed radiation beams found
at reactor facilities. Since Rossi counters measure radiation deposits
within the chamber event by event, accumulation of the magnitude of
events using a multi-channel analyser results in a spectrum of the energy
deposition. Because the magnitude and spectrum of energy deposition
differs for radiations of different LET (linear energy transfer), micro-
dosimetry has developed methods of curve stripping to separate the
contribution to the absorbed dose from photons, electrons, neutrons
(recoil protons) and heavy ions. This has been an important contribution
to radiation protection, where the radiobiological effectiveness (RBE) of
different radiations depends upon the LET of the radiation. In fact, micro-
dosimetry goes further than measuring the LET of the radiation field. By
measuring the stochastics of individual radiation interactions within small,
cell-equivalent volumes, it serves as a methodology of determining the
entire spectrum of radiation deposits and LETs experienced by a cell or
population of cells.
202
Introduction 203
The energy deposition spectra measured by Rossi counters are presented

in two formats: specific energy z and lineal energy y. Specific energy is defined
as z ¼ "=m, where " is the magnitude of the energy deposit within a mass m: It
is the stochastic analogue of absorbed dose D. Lineal energy is defined as
y ¼ "=l, where l is the mean chord length of the simulated cavity. Lineal
energy is the stochastic analogue of LET.
To study energy deposition within volumes < 0:3 mm in diameter, track
structure simulation codes are required. These codes use Monte Carlo
methods to model individual particle histories as they traverse matter.
Coupled with models of the biological target, these track structure codes
have been used to make predictions of DNA strand break and base
damage [2–6]. With the development of increasingly sophisticated radio-
nuclide targeting techniques, an understanding of the microdosimetry for
these systems will assist in the interpretation of biological effects.
Radiolabelled antibodies and other nuclear medicine radionuclide
targeting compounds are used either at diagnostic or therapeutic activities.
Diagnostic activities administered to patients result typically in whole body
doses on the order of 1 cGy. While at first glance one would expect that
microdosimetric methods are required to characterize the stochastic nature
of energy deposition, a closer evaluation reveals that this may not be
necessary. For low LET radiations such as -, X- and -rays emitted by
radionuclides, an absorbed dose of 1 cGy corresponds to an average of
approximately 50 individual track traversals of a cell (depending upon the
assumed cell diameter). The relative deviationpffiffiffiffiffi in the number of hits
(assuming Poisson statistics) is 14% (1= 50). Applying the criteria of
Kellerer and Chmelevsky [7], who state that microdosimetry should be
applied when the relative deviation of the specific energy spectrum is greater
than 20%, the need for microdosimetry at these absorbed doses is borderline
for diagnostic tracers. However, the application of Poisson statistics is
warranted only when the radionuclide distribution is uniform. This is
rarely the case for any radionuclide administered to man. As a consequence,
an extended application of microdosimetry, which deals not only with the
stochastics of the energy deposition process (classical microdosimetry) but
also with the unique biodistribution of the radionuclide, may be warranted
in the analysis of energy deposition at all administered tracer activities.
This concept is exemplified by the study of Makrigiorgos et al. [8] who
performed measurements of the microdistribution of a routine diagnostic
tracer 99m Tc-Microlite from liver tissue samples and showed that, relative
to other cells, the colloidal agent was concentrated 200- to 1000-fold
higher in liver macrophages relative to the average activity per gram of liver.
There are two types of radionuclide for which classical microdosimetry
has the greatest relevance, even at therapeutically administered activities:
and Auger electron emitters. Since and Auger electron therapies using
antibody and other carrier molecules have become a clinical reality over
204 Microdosimetry of targeted radionuclides
the past few years, the remainder of this chapter will focus on a discussion of
each.
9.2. ALPHA EMITTERS
Over the past several years, researchers have taken an interest in the use of
particle emitters as a potential form of therapy for micrometastatic disease
[9–19]. The radiobiological advantages of particles are their high linear
energy transfer (LET) and independence from dose rate and oxygen effects
[20]. When combined with a suitable carrier (e.g., antibody), the short
range of particles in tissue (40–90 mm) results in a highly localized deposi-
tion of energy near the point of emission (presumably in the tumour) and
limits the volume of normal tissues irradiated.
The dosimetry of particle emitters is challenging due to the stochastic
nature of energy deposited in small, subcellular targets as shown schematically
in figure 9.1. In this diagram, some cell nuclei receive multiple particle hits,
while others receive no hits whatsoever. (In this context, we define a hit as any
particle that intersects a cell nucleus, regardless of the energy deposited.)
Additionally, an particle may traverse a variety of path lengths in passing
through a cell nucleus. For example, an particle that intersects a spherical
nucleus through the exact centre will have a path length equal to twice the
radius, whereas an particle that intersects nearly tangent to the cell nucleus
Figure 9.1. Schematic diagram illustrating the stochastic nature of energy

deposited in individual cell nuclei (shaded circles) by particles (lines with
arrows). Some cells will receive no hits, while others will receive a number of
hits based on the Poisson distribution.
Alpha emitters 205
will have little or no path length. Since the energy deposited by an particle is
proportional to its path length, each particle in the above scenario produces
one hit; however, the amount of energy deposited is significantly different.
Since the distribution of hits to the cell nucleus is often characterized by
using Poisson statistics, this problem is often referred to as a compound
Poisson process—a Poisson distribution in the number of events where the
magnitude of each event varies. Other effects that influence the dosimetry of
particles include the variation in energy loss (LET) along an individual
particle’s path, the size and shape of the target, and the source configuration
(within the cell versus outside the cell, etc.). Thus, it is generally accepted
that the analysis of the cellular response to particle irradiation requires
microdosimetric methods. The importance of microdosimetry in therapeutic
applications (e.g., radioimmunotherapy) is that while two different micro-
dosimetric distributions may have the same average specific energy (absorbed
dose), they can result in significantly different levels of cell kill [21, 22].
9.2.1 Monte Carlo simulation of energy deposition by particle emitters

Microdosimetric spectra may be calculated using either analytical or Monte
Carlo methods [23]. Analytical methods use convolutions (via Fourier
transforms) of the single-event spectrum to calculate multi-event distribu-
tions. The single-event spectrum represents the pattern of specific energy
depositions for exactly one particle hit. While analytical codes are
computationally efficient, they are often limited to simple source–target
geometries as the single-event spectrum must be known for each source–
target configuration. Monte Carlo codes offer greater flexibility than
analytical methods, and can simulate a wide variety of geometries and
source configurations. While Monte Carlo codes may be computationally
intensive, the recent availability of fast desktop computers has made the
calculation time less of a concern.
-Particle Monte Carlo codes require the developer to specify both the
source and target geometry, as well as providing a characterization of the
particle energy loss. Idealizations are often made to simplify the coding and
reduce calculation time. A typical algorithm initially selects an particle
emitter and determines whether the source decays based on its effective
half-life. Next, the energy of the particle emission is determined by
random sampling of the branching ratios. Based on isotropic emission, the
directional cosines of the particle track are randomly assigned. In nearly
all Monte Carlo codes, particles are assumed to travel in straight lines.
This approximation is valid for particles having energies less than
10 MeV [24]. In addition, the range of -rays and the width of the particle
track (100 nm) are often ignored since the targets that are often studied (i.e.,
cell nucleus) are much larger than these dimensions [25]. Using this straight-
line approximation, the parametric equations [26] can be solved to determine
whether the particle intersects the cell nucleus:

x ¼ x0 þ t cosð1 Þ ð9:1Þ
0
y ¼ y þ t cosð2 Þ ð9:2Þ
z ¼ z0 þ t cosð3 Þ: ð9:3Þ
In equations (9.1)–(9.3), x0 , y0 and z0 are the coordinates of the particle

emission, and x, y and z are the intercepts of the particle with the nucleus.
The angles 1 , 2 and 3 are the angles of the particle path relative to the x, y
and z axes, respectively. If the nucleus is spherical (or some other simple
geometry), equations (9.1)–(9.3) can be substituted into the target’s
analytical form to determine the distance traversed by the particle upon
entering the leaving the nucleus. For example, a spherical nucleus is given by
ðx x0 Þ2 þ ðy y0 Þ2 þ ðz z0 Þ2 ¼ r2n ; ð9:4Þ
where x0 , y0 and z0 are the centre coordinates of a nucleus with radius rn .
Substituting equations (9.1)–(9.3) into equation (9.4) and solving for t
yields two solutions—t1 corresponding to the distance from the source to
where the particle enters the nucleus, and t2 corresponding to the distance
between the source and where the particle exits the nucleus. Note that when
the particle emission occurs within the nucleus (starter), t1 is chosen to be
zero, while t2 is the intercept with the nuclear surface. If the particle
terminates its track within the nucleus (stopper) the value of t2 is set to the
range of the particle. A hit occurs when the solutions for t1 and t2 are both
real, positive numbers.
Next, the energy deposited within the cell nucleus is calculated. This
energy is given by
ð R t1
dE
"1 ¼ dx: ð9:5Þ
R t2 dx
The energy deposited by a single particle transit is transformed into specific

energy using the relation
z1 ¼ "1 =m: ð9:6Þ
In equation (9.5), ðdE=dxÞ is the stopping power, and the limits of
integration are over the residual range of the particle upon entering
ðR t1 Þ and leaving the target ðR t2 Þ, respectively. The stopping power,
which characterizes the rate of energy loss per unit path length, folds in
the individual processes involved in particle energy loss [27]. Stopping
power data for a variety of media can obtained from the literature [28–32].
Inherent in the stopping power formulation is the continuous slowing
down approximation (CSDA). As the name implies, this approximation
assumes that particles lose energy continuously as they traverse matter.
Alpha emitters 207
Thus, the determination of the specific energy deposited relies heavily on the
choice of stopping powers used. In a comparison of the impact of stopping
power formulation, Stinchcomb and Roeske [23] calculated specific energy
spectra for the same source–target geometry as Humm [22]. The former
case used Janni’s stopping power [33] for protons (appropriately scaled for
particles), while in the latter case Walsh’s stopping power [34] was utilized.
In general, a disparity of 1–7% was observed in the average specific energies
using these two different characterizations of the stopping power. However,
in practical situations, it is expected that the uncertainty in the activity
distribution will be much larger than that of the stopping powers used in
the Monte Carlo calculation.
The process of determining the specific energy deposited for an
individual interaction can be repeated to determine the multi-event specific
energy spectrum. In this case, all sources within the maximum range of the
particle emission are evaluated to determine whether they intersect the
nucleus and the subsequent specific energy deposited is calculated and
summed. This scenario represents one possible outcome. Next, the process
is repeated and, most likely, a different total specific energy will be calculated.
By repeating this process, the entire outcome of specific energies deposited
can be determined. A histogram of these individual results is the multi-
event distribution. An alternative is to determine the single-event specific
energy spectrum. This distribution is calculated by examining only those
particle emissions that result in exactly one hit to the cell nucleus. Each
time an individual particle intersects the nucleus, the specific energy is
determined and recorded. All of these single-event specific energies are
then histogrammed to determine the single-event spectrum. Single-event
spectra are often used in conjunction with analytical codes whereby the
multi-event spectra are calculated by multiple convolutions of the single-
event spectrum. Stinchcomb and Roeske [23] demonstrated that both
methods produce the same multi-event spectra within the uncertainty of
the calculation itself. An example of a single-event spectrum for 212 Bi
distributed on the cell surface is shown in figure 9.2. A multi-event spectrum
for 15 212 Bi sources on the cell surface is shown in figure 9.3.
Microdosimetric spectra are often combined with a model of cell
survival to estimate the survival following a particular type of irradiation.
The surviving fraction following particle irradiation is given by [22, 23, 35]
ð1
SðzÞ ¼ f ðzÞ ez=z0 dz ð9:7Þ
0
where f ðzÞ is the multi-event specific energy spectrum representing the

fraction of cells receiving specific energies between z and z þ dz, z0 is the
specific energy deposited within an individual cell that reduces the average
cell survival to 1=e, and expðz=z0 Þ is the fraction of cells that survive.
The use of the exponential function assumes that there are no bystander
Figure 9.2. Single-event specific energy distribution for 212 Bi sources distribu-
ted on a cell surface. The cell radius is 10 mm and the nuclear radius is 5 mm. The
two peaks correspond to the two emissions from 212 Bi: 6.05 MeV (36%) and
8.78 MeV (64%).
effects and that intratrack interactions are dominant relative to intertrack

interactions. When using the single-event spectrum, the relationship is
given by [23]
zÞ ¼ exp½hnif1 T1 ðz0 Þg
Sð ð9:8Þ
Figure 9.3. Multi-event specific energy spectrum for 15 particle sources

(212 Bi) distributed randomly on a cell surface with radius 10 mm. The nuclear
radius is 5 mm. Note that a significant fraction of the cells receive zero hits to
the cell nucleus.
Alpha emitters 209
where hni is the average number of hits to the cell nucleus, and T1 ðz0 Þ is the
Laplace transform of the single-event spectrum. Note that the same value of
z0 satisfies both equations (9.7) and (9.8).
It is important to point out that z0 is not equal to D0 , which is
determined from the slope of the cell survival curve. Rather it is a more
fundamental quantity as D0 has folded into it not only the effects of the
radiation, but also the effects of the source–target geometry. In the
applications to be discussed a recurrent theme is to use the observed
surviving fraction to determine the inherent cell sensitivity, z0 .
9.2.2. Applications of microdosimetry to particle emitters

Microdosimetry has been used in a number of particle applications,
particularly RIT. These applications can be broadly characterized as
theoretical studies of simple cellular geometries, experimental analysis of
cell survival following particle irradiation, and the microdosimetry of
realistic geometries such as multicellular spheroids and bone marrow. The
work in each of these categories will be discussed separately.
Fisher [36] was among the first to demonstrate the utility of micro-
dosimetry for therapeutic applications. This particular report used an
analytical code based on the previous work of Roesch [37], and suggested
a number of geometries for RIT including sources distributed on and
within individual cells, sources distributed within spherical clusters of cells,
and sources located in cylinders (i.e., blood vessels) that deposited energy
within spherical cell nuclei a short distance away. These geometries have
served as the basis of those used in subsequent theoretical studies. In one
such study, Humm [21] combined the physical calculation of specific
energy (using Monte Carlo methods) with a model for cell survival to analyse
the impact of stochastic energy deposition on the expected surviving fraction
of a group of cells. In particular, two geometries were considered—cells
located outside a capillary and cells located within a tumour consisting of
a uniform distribution of 211 At. The results of this analysis demonstrated
that although the mean dose was similar for these two types of geometry,
there was a significant variation in the expected cell survival due to the
differences in the specific energy spectra. In particular, the fraction of cells
without particle hits increased with distance from the capillary (due to
the short range of the particles) resulting in a bi-exponential cell survival
curve. Initially, the slope of this curve was similar to that of a uniformly
irradiated tumour. However, with increasing doses, the curve was less
steep and asymptotically approached a value corresponding to the fraction
of non-hit cells. Building upon the previous analysis, Humm and Chin [22]
analysed the effects of cell nucleus size, binding fraction, cell volume fraction
and nonuniform binding on calculated specific energy spectra. Their results
demonstrated that the expected cell survival curves, produced by nonuniform
distributions of -particle-emitting radiolabelled antibodies can depart

significantly from the classical mono-exponential curves that were produced
by an uniform, external source of particles. In particular, although the
inherent cell sensitivity (z0 ) was held constant, the slope of the cell survival
curve as a function of absorbed dose to the medium was highly dependent
upon the source configuration. Those cases in which cells were more
uniformly irradiated had steeper cell survival curves than simulations in
which the distribution of emitters was highly heterogeneous. Stinchcomb
and Roeske [38] further studied the effects of cell size and shape on expected
cell survival. In this study, the cell and nucleus were allowed to take on
shapes ranging from spheres to ellipsoids where the ratio of the major-to-
minor axis was varied from 1 to 5, while the volume of the nucleus was
held constant. Separately, the size of the nucleus was varied while the nuclear
shape was held constant. Using calculated specific energy spectra and a
model of cell survival, they determined that the expected cell survival was
not a strong function of the target shape, provided the volume was fixed.
However, they noted significant variations in cell survival as the volume of
the sensitive target was varied. In a recent study by Kvinnsland et al. [39],
microdosimetric spectra were calculated for a heterogeneous expression of
cell surface antigens, as well as various diameters of cells and cell nuclei.
Their results indicated that cell survival may be underestimated by two
orders of magnitude if the mean values of any of these parameters are
used. The largest contributor to this effect was use of the average antigen
concentration. All of these studies point out a number of features related
to particle microdosimetry. In particular, they demonstrate the importance
of accurately simulating the source/target geometry. Inaccuracies (such as
using mean values) may impact the specific energy spectrum and subsequent
estimation of cell survival.
Stinchcomb and Roeske [40] applied the previously developed methods
of predicting cell survival to the analysis of experimentally produced cell
survival curves. Two types of source configuration were considered. The
first case consisted of cells irradiated in suspension by an ionic solution of
212
Bi. In this case, a small fraction of the 212 Bi accumulates on or within
the cell. The second case consisted of 212 Bi chelated to DTPA—a geometry
where the source remains in solution outside the cell. By calculating the
specific energy distribution and knowing the resultant cell survival for the
irradiation condition, the inherent cell survival (z0 ) was determined using
iterative methods. In this particular case, the authors observed a small differ-
ence in z0 for the two source geometries. These differences were attributed to
using the average cell size (as opposed to the cell size distribution), and thus
demonstrated the importance of including the cell size distribution in future
microdosimetric calculations. A microdosimetric approach was used by
Larsen et al. [17] to determine the inherent cell sensitivity of human glioma
and melanoma cell lines. Both cells lines were grown in microcolonies and
Alpha emitters 211
irradiated by 211 At-labelled antibodies. A Monte Carlo method was used to

calculate the single-event spectra of microcolonies modelled as an infinite
plane. This analysis indicated that one or two hits to the cell nucleus reduced
the surviving fraction to 37%. Furthermore, the microdosimetric analysis
revealed that the sensitivity of individual cells was much greater than
would be expected using a nonstochastic dose analysis. Charlton [41] made
refinements to the Larsen [17] model by incorporating the effects of finite
cluster sizes. The results of this work indicated that the dose was a maximum
in the centre of the individual clusters and decreased by nearly a factor of 2
near the edges, dependent upon the size of the clusters. In particular, it was
observed that for smaller clusters of cells (i.e., <25) the average number of
passages for 37% survival was 1.26, which was smaller than the 1.62
estimated using the infinite plane assumption. This analysis indicates the
importance in accurately modelling cell geometry and its effect in interpreting
cell survival studies. Aurlien et al. [42] compared the biological effects of
211
At radioimmunoconjugates to external -rays for osteosarcoma cells.
Experiments were performed with cells in suspension containing a uniform
solution of the -particle-emitting radioactivity. The distribution of cell
sizes was measured and a Monte Carlo code was used to calculate the specific
energy spectra. Their results indicated that when survival was plotted as a
function of the average dose there was a significantly higher level of cell
kill in osteosarcoma cells relative to bone marrow cells. Thus, this form of
therapy may prove efficacious. In a departure from the previous studies,
Charlton et al. [43] applied a model of cell survival to predict cell survival
following particle irradiation. Initially, the microdosimetric spectra were
calculated for cells grown both as a monolayer and in suspension. Using
these spectra, and a model of cell survival that took into account the specific
energy deposited and the incident particle’s LET, a theoretical cell survival
curve was generated. A comparison of theoretical with experimental survival
data indicated agreement within experimental uncertainties. This approach
demonstrated that if knowledge of the cell response as a function of LET
is available, it is possible to use microdosimetric spectra with a model of
individual cell survival to accurately predict cell survival following particle
irradiation.
The previous studies focused on combining microdosimetric spectra
with biological models to analyse or predict survival of cells irradiated in
suspension or as a monolayer. A logical extension of this work is the
modelling of more complex geometries. One such geometry, multicellular
spheroids, mimics small micrometastatic clusters and is therefore
therapeutically relevant. However, because of the additional complexity,
the computation is significantly more difficult and therefore few studies
have been performed in this area. Kennel et al. [44] considered cell survival
following the irradiation of multicellular spheroids with an -particle-
emitting radiolabelled antibody. Following irradiation, the spheroids were
divided into three regions corresponding to an outer, middle and inner layer
of cells. Cell survival was assessed individually for each layer. Additionally, a
Monte Carlo code was developed to calculate specific energy spectra as a
function of depth within the spheroid assuming the activity was concentrated
on the outer 10 mm. Their results demonstrated a wide variation in the
specific energy distribution in the three different levels of the spheroid. The
average dose was a maximum near the outer spheroid surface and decreased
rapidly with depth. Conversely, the fraction of cells receiving zero hits was a
minimum near the surface and increased rapidly with depth. Additionally,
they noted a qualitative correlation that cells near the outer edges of
the spheroids had a lower surviving fraction compared with those inside.
Charlton [45] produced a calculational model for multicellular spheroids
and simulated their subsequent survival. The distribution of particle
tracks throughout cells within the spheroid was calculated and combined
with a cell survival model that takes into account the survival as a function
of LET [46]. The results of this analysis indicated that cell survival decreased
from 57 to 37% as the spheroid diameter increased from 75 to 225 mm for a
uniform source with one decay per cell, and 50% packing. Longer ranged
particle emitters increased the number of hits per cell for the larger spheroids.
Increasing the packing fraction from 40 to 70% decreased the survival from
46 to 26% in 200 mm diameter spheroids, again with one decay per cell. This
significant decrease in survival was due to the increased crossfire dose com-
ponent as cells were brought closer together. In a different simulation, a
small fraction of cells (20%) was assumed not to take up any activity due
to lack of expression of a surface antigen. In this simulation, the unlabelled
fraction did not significantly change the expected cell survival. In both of
these studies, there is evidence of a highly nonuniform specific energy
distribution that varies with distance from the spheroid surface. Thus, a
single specific energy distribution is not representative of that through the
entire tumour. Combining specific energy distributions with a model of cell
survival may provide an overall measure of the therapeutic effectiveness.
However, these cell survival models do not take into account second-order
processes such as bystander effects. These processes are more difficult to
simulate and may play a significant role for these types of geometry. The
ongoing refinement of these cell survival models is currently an area of
active research [47].
Bone marrow is often the dose-limiting organ in RIT. The dosimetry of
bone marrow is difficult due to the complex geometry as well as the presence
of tissue inhomogeneities. Thus, idealized models, as have been utilized in
previous studies, must be replaced by more realistic geometries. The work
to date on estimating specific energy spectra for bone marrow has focused
largely on using histological samples obtained from humans or animal
models. While there has been considerable interest in bone marrow micro-
dosimetry for protection purposes [48–51] this review will focus only on
Auger electron emitters 213
studies related to RIT. Akabani and Zaltusky [52] obtained histological

samples of beagle bone marrow, and manually measured chord length
distributions. Using a Monte Carlo program, they calculated the single-
event spectra for both sources in the extracellular fluid and sources located
on the surface of red marrow cells. These single-event distributions were
combined with a model of cell survival. This analysis demonstrated that
activity concentrated on the cell surface resulted in significantly higher
levels of cell kill than activity in the extracellular fluid. The effect of LET
on the survival of human haemopoietic stem cells in various geometries
was studied by Charlton et al. [53]. These geometries were determined from
human marrow samples obtained from cadavers. The authors measured the
distribution of distances between stem cells and fat cells. Microdosimetric
spectra and cell survival were calculated for three different source/target
geometries: (i) isolated cells labelled on the surface, (ii) non-targeted distribu-
tion of decays in an extended volume and (iii) non-targeted decays in marrow
with 36% of the marrow volume occupied by fat. Two different radionuclides
were considered, 149 Tb and 211 At. These simulations indicated that for
targeted decays, 149 Tb was five times more effective than 211 At when
compared on a hit-by-hit basis. This enhancement was due to the lower
energy of 149 Tb resulting in a higher LET of the incident particles. They
also concluded that cell survival was a function of the position of decay
relative to the cell nucleus. Using a similar model to Charlton [53], Utteridge
et al. [54] considered the risk of developing secondary malignancies (i.e.,
leukaemia) from particles. This risk may be important in evaluating the
future therapeutic application of particles in patients who have an excellent
prognosis. Three -emitting radionuclides were considered based on the
particle’s relative range (short, medium and long). In this analysis, the
authors calculated the fraction of cells that are hit and would survive (as
these would potentially cause secondary malignancies). They determined
that lowest fraction occurred for low energies and the highest fraction
occurred for the highest energy particle emitter.
While particle Monte Carlo codes have been used in a variety of
theoretical and in vitro analyses, direct clinical application of microdosimetry
has been limited due to a lack of knowledge of the source distribution as
a function of time and location on the microscopic level. However, the
previously described studies have provided valuable insight into the stochastics
of particle dosimetry that may be useful in interpreting the clinical results of
patients treated with particle emitters.
9.3. AUGER ELECTRON EMITTERS
Auger electrons are emitted by any radionuclide that decays by either electron
capture or internal conversion. Both these decay processes result in an inner
atomic electron vacancy, usually in the innermost K-shell. This vacancy

initiates a cascade of inner atomic shell transitions in which a number of
low energy Auger, Coster–Kronig and super-Coster–Kronig electrons are
liberated with emission energies corresponding to the differences in the binding
shell energies involved in the transition. The total energy emitted following
electron capture corresponds to the binding energy of a K-shell (or L-shell)
electron for the appropriate atom. This energy is shared between all of the
electrons and any fluorescent X-ray photons emitted. For this reason, the
total energy emitted is small, and has often been disregarded in organ and
total body dosimetry estimates. After internal conversion, the same Auger
cascade ensues, with the difference that a potentially much higher ‘conversion
electron’ from the K-shell (or L-shell) can be ejected with an energy
corresponding to the difference between the nuclear excitation energy and
the appropriate inner shell electron binding energy.
Radionuclides which undergo electron capture and/or internal con-
version are extremely common. They constitute about half of the isotopes
in the table of radionuclides, including many of the isotopes employed in
routine diagnostic nuclear medicine studies, e.g., 99m Tc, 67 Ga, 123 I,
201
Tl, etc. The relevance of the soft electrons emitted as a consequence of
electron capture or internal conversion went unnoticed until the studies by
Ertl et al. [55] and Hofer and Hughes [56]. These independent investigators
discovered that when 125 I, which decays by electron capture followed by
internal conversion, is directly incorporated into the cellular genome, via
the thymidine analogue 125 I-iodiodeoxyuridine, that the cell survival curves
resemble the appearance of particles. Estimates of the radiation absorbed
dose revealed that intranuclear decays of 131 I deposited more dose to the cell
nucleus than 125 I, yet investigators found 125 I was >12 times more lethal than
131
I on a per decay basis [57, 58]. This led radiobiologists to wonder how low
energy and low LET electrons could exhibit high LET properties. The history
of 125 I radiobiology has been reviewed by Sastry [59]. In brief, Charlton and
Booz [60] simulated the atomic relaxation of 125 I following electron capture
and internal conversion and revealed that on average 21 electrons are emitted
per decay. The large number of electrons emitted, and the positive charge
build-up in their wake, results in a massive local energy deposit within a
few Ångströms of the decay site. It was microscopic track structure
simulations that revealed the physical explanation for the high radiotoxicity
of 125 I, when Charlton [61] showed that the density of ionizations around the
decay site of 125 I is greater than in the core of an -particle track. Since this
time, improved calculations of Auger spectra for a number of electron
capture and internal conversion decaying radionuclides have been generated
[62], more elaborate models of the DNA have been created [4, 63], and better
experimental data on strand break damage versus distance from the decay
site has been measured on high resolution denaturing polyacrylamide gels
[64, 65].
Out of this work and that of many other contributors have emerged the
following inferences:
1. The radiobiological effectiveness of Auger electron sources is no

different from any other low LET source, provided the decay does not
occur adjacent to a critical cellular target [58].
2. The range of effectiveness of Auger-electron-emitting isotopes is extremely
short and decay at or within 1 nm of a sensitive structure, such as the
DNA, is subject to high LET effects [66].
3. Since it is the pharmaceutical carrier which determines the sub-cellular
location, the radiobiological effectiveness for any Auger emitting
radionuclide depends not only on the isotope but also on the carrier
molecule [67].
9.3.1. Monte Carlo simulation of Auger decays and their energy deposition
There is considerable stochastic variability in the way that energy is released
from electron capture or internal conversion of decaying radionuclides. This
variability was first modelled by Monte Carlo simulation of the decay process
of 125 I [60]. Since then, Auger electron spectra have been calculated for
several other radionuclides [62]. The number and distribution of energy
between the electrons (and photons) emitted in an Auger cascade varies
between individual disintegrations as governed by the inner shell transition
probabilities. The number of permutations of possible de-excitation path-
ways through the atom can be hundreds, depending on the atomic number
of the isotope. The way in which electron capture (or internal conversion)
and the ensuing Auger cascade are computer simulated is described in
detail by Charlton and Booz [60], Humm [68] and Howell [62]. In brief, a
random number generator is used to select an initial inner shell vacancy
according to the probabilities of capture within each of the respective orbitals
K, L1, L2, L3 etc. The probabilities for all transitions to this vacancy are
normalized to one, where each allowed transition is weighted by its respective
probability according to published values for radiative [69] and non-radiative
[70] transitions, determined from quantum mechanical calculations of
oscillator strengths. A second random number is then used to select one of
these transitions, according to their respective weights. The new electron
occupancy of the orbitals is recorded and the next innermost shell vacancy
is selected. This process then repeats until all of the vacancies have moved
to the outermost shells (the lowest energy state), when further transitions
are no longer possible.
For high Z elements (e.g., tellurium) following K-capture in iodine, the
fluorescent yield is high (0.877). This means that 87.7% of the transitions to
fill a K-shell vacancy emit characteristic X-rays and only 12.3% result in a K-
shell Auger electron. This results in a variable total amount of electron
Figure 9.4. Schematic diagram of an 123 I atom undergoing electron capture.

The crosses denote electron occupancy of each shell. A vacancy is represented
by a empty circle, and a filled vacancy (as a consequence of a transition) by
a circle containing a cross. Each transition moves a vacancy to a higher orbital
until all vacancies occupy the outermost shells.
energy released, contingent upon the photon transitions which occur. One
example of a possible atomic de-excitation pathway following electron
capture for 123 I spectra is schematically depicted in figure 9.4.
Simulation of 10 000 individual atomic de-excitations results in an
output of individual Auger electron and X-ray spectra, containing the
stochastics of the de-excitation process. These data can be averaged to
yield an average Auger electron emission spectrum, an example of which is
shown in figure 9.5. Note that this does not denote an emission spectrum
from an individual decay, but rather a frequency weighted average of
10 000 simulated decays. Dosimetric estimates to different compartments
of the cell have been performed using the average Auger electron spectrum,
and the results are the same as when multiple individual Auger electron
spectra are used. The stochastics of individual spectra is only pertinent
when calculating the energy deposition along the DNA.
Charlton and Humm [71] performed Monte Carlo electron track
simulations for 10 000 individual Auger electron spectra resulting from 125 I
and other radionuclides [66]. They scored the coordinates of each energy
deposition generated by the electron track structure code of Paretzke [72],
relative to the origin of the site of decay placed at the position of a base in
a computer-simulated geometric model of the DNA double helix. This
model consisted of a central cylinder representing the bases surrounded by
123
Figure 9.5. An example of an average electron energy spectrum from I
following 10 000 simulated decays.
a concentric helical cylinder of pitch 0.36 nm, representing the sugar

phosphate strands. The resulting energy depositions within opposing
strand segments of this DNA model were then compared with experimental
data on the radius of the strand break damage measured with DNA
sequencing gels relative to the 125 I decay site by Martin and Haseltine [73].
Matching the theoretical profile of stochastic energy deposition versus
distance from the site of 125 I decay with the measured single-strand break
frequency profile allowed a threshold energy for strand break damage to
be estimated. It was a threshold energy of 17.5 eV for single-strand break
production which provided the best agreement with the experimental data.
To improve statistics, the energy deposition was calculated for the same
individual Auger spectrum 10 times, and it was found that the variability
in local energy deposition was substantial and added yet another layer of
stochastic variability in the data. In summary, variation in energy deposition
results from the inherent variability on atomic de-excitation and from the track
structure of these emissions relative to the modelled target sites. Whereas these
stochastics are not important for cellular level dosimetry estimates, except at
very low numbers of decays per cell, they are important for predicting
strand break damage. This is because any threshold model of strand break
damage does not bear a linear relationship to the mean energy deposition.
The stochastic effects of Auger electron emitters and the high LET
effects are relevant within a few nanometres of the decay site, since most
Auger and Coster–Kronig electrons have ranges of this order. Therefore,
for Auger sources that decay at the extracellular locations (as is the case
for most diagnostic tracers used in nuclear medicine), concerns of high
LET and the consequent radiotoxicity effects are unwarranted. However, the
extreme short range effects of Auger electron emitters do pose a new problem
in assessing the dosimetry of novel radiopharmaceuticals. Humm et al. [74]
and Goddu et al. [75] have recommended some guidelines for their evalua-
tion, which requires the initial determination of the partition of the agent
into extracellular, cell surface bound, intracellular, intranuclear and
appended to the DNA. Cellular S-factors for the former four compartments
have been calculated by Goddu et al. [76, 77] for a number of electron capture
and internal conversion decaying radionuclides. The average radiation dose
to the nucleus (or other compartment) can be determined by the piecewise
addition of each contribution, upon experimental determination of the rela-
tive compartmental partitions.
9.3.2. Targeted therapy with Auger electron emitters

Auger electron emitters have been used in pre-clinical and clinical studies in a
number of radiopharmaceutical forms designed to selectively deliver radia-
tion to tumour cells. These include thymidine precursors [78–82], thymidine
precursor prodrugs [83], minor DNA groove binding bisbenzimidazole
ligands [84], triplet-forming oligonucleotides [85], hormones [86] and anti-
bodies [87–90]. A review of the emerging role of Auger targeting therapies
can be found in Mariani et al. [91]. The most effective therapies target the
Auger source directly to the DNA, such as iododeoxyuridine, Hoechst
33258 or triplex-forming oligonucleotides. One drawback of these agents is
the lack of tumour specificity, although attempts to attach these agents to
tumour-specific carriers, such as antibodies, as a means of selective tumour
targeting and intracellular access are currently under investigation. The
potential therapeutic advantages of using Auger emitters for therapy with
internalizing antibodies are described by Daghighian et al. [92] and Behr
et al. [93]. In brief, the short range of Auger electrons renders them ineffectual
in sterilizing the usual dose limiting tissues such as bone marrow, since they
neither bind nor are internalized by these cells. This was clinically demon-
strated by Welt et al. [89], who administered to patients activities in excess
of 29 GBq (800 mCi) of 125 I-labelled A33 antibody and still did not reach
dose-limiting toxicity. The cell cytoplasm acts to shield the nucleus and
cellular genome from the low energy electron emissions. Antibody binding
to cell surface antigen and eventual internalization increases the spatial
proximity of the Auger source to the targets of cell inactivation, thus
giving rise to a significant geometric and hence dosimetric differential
between tumour and normal cells. However, the deposition of energy from
Auger electron emitters with antibody (or hormone) approaches to the
tumour targeting do not result in the high LET effects observed with 125 I-
iododeoxyuridine. With these approaches the Auger source is treated like a
very soft source, such as tritium. As a consequence, for targeted Auger
Future directions 219
therapy to be successful, the radiolabel carrier must reach every clonogeni-

cally potent tumour cell. This is a severe limitation. However, the theoretical
rewards are enticing and it is this reason which drives researchers to further
pursue the enormous potential of high radiotoxicity associated with Auger
emitters with limited toxicity to normal tissues.
9.4. FUTURE DIRECTIONS
While Monte Carlo codes have produced an increased understanding of

stochastic patterns of energy deposition by particle and Auger electron
emitters in both simple and complex geometries, and for different sub-
cellular distributions, application to clinical practice has been limited. The
principal reason is that these codes are not readily available and that
often, the activity distribution as a function of time at the sub-cellular level
is not well known (especially in vivo). One potential solution to the first
problem is to combine the detailed information provided by microdosimetry
with the ease of use and availability of MIRD. Roeske and Stinchcomb [94]
have provided a formalism for determining three parameters that are
considered important in particle dosimetry: average dose, standard devia-
tion in specific energy, and the fraction of cells receiving zero hits. These
parameters are determined using tables of the S value, and the first and
second moments of the single-event spectra. For example, the average dose
is given by the product of the S value and the cumulated activity. Dividing
the average dose by the first moment of the single-event spectrum yields
the average number of hits, which can be used to determine the fraction
receiving zero hits (assuming Poisson statistics). The standard deviation is
given by the product of the average number of hits and the second
moment of the single-event spectrum. These individual moments may be
determined using either analytical methods or Monte Carlo calculations.
Stinchcomb and Roeske [95] have produced tables of the S value, and the
individual moments for a number of geometries and source configurations
appropriate for RIT. Roeske and Stinchcomb [96] applied these tables to
the analysis of cell survival following particle irradiation. Similar to
previous work, they used microdosimetry to determine the inherent cell
survival parameter, z0 . However, the approach they took was to approximate
the single-event spectrum by its series expansion. Terms containing moments
greater than the second were dropped. In general, this approximation (using
only the first and second moments) resulted in values of z0 that agreed to
within 1–2% of the value calculated by direct determination using the full
spectrum. Thus, the moments approach is validated for the analysis of cell
survival studies.
Monte Carlo codes for the calculation of Auger electron spectra are also
not readily available. However, spectra for several electron capture and
internal conversion decaying radionuclides have been published for radio-

nuclides relevant for diagnostic and therapeutic applications [62], allowing
users to perform their own energy deposition calculations. Most of the
published spectra, whereas adequate for microdosimetric calculations, have
assumed static energy levels, using the Z þ 1 or Z þ N approximation in
the determination of the individual Auger electron emission energies. Such
simplified approaches result in ever-increasing inaccuracies in the energy
estimates as the number of inner shell electron vacancies increases, especially
for high Z atoms. More advanced methods, based on relativistic Dirac–Fock
calculations, provide a perfect energy balance of the considered atomic
system when applied to Monte Carlo simulations of Auger cascades [97].
These and other developments, such as the consideration of shake-up and
shake-off, will improve the accuracy of the Auger energy spectra [98, 99].
Of more immediate practical utility, a dedicated MIRD publication on cellu-
lar S factors has been published, which contains data on several Auger elec-
tron as well as sources [100]. A method of implementation of cellular S
values for Auger sources has been proposed by Humm et al. [74].
Another problem of particle and Auger electron microdosimetry
involves how to undergo the transition from modelled uniform or non-
uniform radionuclide distributions to actual tissue (and sub-cellular)
distributions. There has been significant work over the past few years that
has focused on multicellular spheroids and bone marrow geometries. It is
expected that researchers will obtain tissue samples and perform micro-
autoradiography for other critical organs (e.g., liver, kidneys, etc.) to
perform analyses similar to Akabani et al. [52] and Humm et al. [101].
However, as we move towards more complex geometries, radiobiological
models of cell survival and normal tissue damage will require modification.
This will include the modelling of the effects of nonuniform distributions
of activity, the dynamics of biokinetic processes, variation of cell radio-
sensitivity during the different phases of the cell cycle, and bystander effects.
A final direction involves using microdosimetry to predict cell survival.
Charlton et al. [43] showed the utility of this method by accurately predicting
cell survival using the cell sensitivity as a function of LET. This approach may
be useful in the design of new drugs. For example, computer simulations can
be performed to predict cell survival for a new type of antibody. Various possi-
ble radionuclides can be tested with this antibody to demonstrate which
combination would yield the most effective therapy. The best combination
can then be experimentally verified. Such an approach would potentially
speed the design and testing of new radiolabelled drugs. Another potential,
along the same line, involves using microdosimetry to predict the therapeutic
effectiveness within individual patients. For example, tumour cells may be
obtained from a patient. By analysis of etch pits on a plastic sensitive to
heavy ion tracks, as proposed by Soyland et al. [102–104], the cell response
as a function of LET may be determined. Next, using Monte Carlo codes,
References 221
the activity distribution, cell geometry and LET response may be combined to
predict cell survival. Since cell response in vitro may not correlate with tumour
response in vivo, such an approach will require feedback to refine such models.
It is the hope that such an approach would help avoid useless therapies in
patients who may not benefit (such as those with highly resistant cells), or
aid in determining the optimal dosing within an individual patient. Such appli-
cations are potentially years away from implementation, but may provide a
method of optimizing radionuclide therapy to improve the therapeutic gain.
REFERENCES
[1] Rossi H H 1968 Microdosimetric energy distribution in irradiated matter. In

Radiation Dosimetry vol I Fundamentals eds F H Attix and W C Roesch (New
York: Academic)
[2] Charlton D E, Nikjoo H and Humm J L 1989 Calculation of initial yields of single-
and double-strand breaks in cell nuclei from electrons, protons and alpha particles
Int. J. Radiat. Biol. 56(1) 1–19
[3] Brenner D J 1990 Track structure, lesion development and cell survival Radiat. Res.
124(1) (suppl) S29–37
[4] Pomplun E 1991 A new DNA target model for track structure calculations and its
first application to I-125 Auger electrons Int. J. Radiat. Biol. 59(3) 625–42
[5] Friedland W, Jacob P, Paretzke H G, Merzagora M and Ottolenghi A 1999
Simulation of DNA fragment distributions after irradiation with photons Radiat.
Environ. Biophys. 38(1) 39–47
[6] Nikjoo H, O’Neill P, Terrissol M and Goodhead D T 1999 Quantitative modelling of
DNA damage using Monte Carlo track structure method Radiat. Environ. Biophys.
38(1) 31–8
[7] Kellerer A M and Chmelevsky D 1975 Criteria for the applicability of LET Radiat.
Res. 63(2) 226–34
[8] Makrigiorgos G M, Ito S, Baranowska-Kortylewicz J, Vinter D W, Iqbal A, Van
den Abbeele A D, Adelstein S J and Kassis A I 1990 Inhomogeneous deposition
of radiopharmaceuticals at the cellular level: experimental evidence and dosimetric
implications J. Nucl. Med. 31(8) 1358–63
[9] Kozak R W, Atcher R W, Gansow O A, Friedman A M, Hines J J and Waldmann T
1986 Bismuth-212 labeled anti-Tac monoclonal antibody: alpha-particle-emitting
radionuclides as modalities for radioimmunotherapy Proc. Natl Acad. Sci. 83 474–8
[10] Kurtzman S H, Russo A, Mitchell J B, DeGraff W, Sindelar W F, Brechbiel M W,
Gansow O A, Friedman A M, Hines J J, Gamson J and Atcher R W 1988 212Bismuth
linked to an antipancreatic carcinoma antibody: Model for alpha-particle-emitter
radioimmunotherapy J. Natl Cancer Inst. 80 449–52
[11] Macklis R M, Kinsey B I, Kassis A I, Ferrara J L M, Atcher R W, Hines J J,
Coleman C N, Adelstein S J and Burakoff S J 1989 Radioimmunotherapy with
alpha-particle-emitting immunoconjugates Science 240 1024–6
[12] Zalustky M R, McLendon R E, Garg P K, Archer G E, Schuster J M and Bigner D
D 1994 Radioimmunotherapy of neoplastic meningitis in rats using an alpha-
particle-emitting immunoconjugate Cancer Res. 54 4719–25
[13] Zalutsky M R, Stabin M, Larsen R H and Bigner D D 1997 Tissue distribution and
radiation dosimetry of astatine-211-labeled chimeric 81C6, an alpha-particle-
emitting immunoconjugate Nucl. Med. Biol. 24 255–1
[14] Rotmensch J, Whitlock J L, Schwartz J L, Hines J J, Reba R C and Harper P V 1997
In vitro and in vivo studies on the development of the alpha-emitting radionuclide
bismuth 212 for intraperitoneal use against microscopic ovarian carcinoma Am. J.
Obstet. Gyne. 176 833–41
[15] Kennel S J and Mirzadeh S 1997 Vascular targeting for radioimmunotherapy with
213
Bi Radiochimica Acta 79 87–91
[16] Larsen R H, Vaidyanathan G and Zalutsky M R 1997 Cytotoxicity of alpha-particle-
emitting 5-[211At]astato-20 -deoxyuridine in human cancer cell Int. J. Radiat. Biol. 72(1)
79–90
[17] Larsen R H, Akabani G, Welsh P and Zalutsky M R 1998 The cytotoxicity and
microdosimetry of astatine-211-labeled chimeric monoclonal antibodies in human
glioma and melanoma cells in vitro Rad. Res. 149 155–62
[18] Couturier O, Faivre-Chauvet A, Filippovich I V, Thedrez P, Sai-Maurel C, Barides
M, Mishra A K, Gauvrit M, Blain G, Apostolodis C, Molinet R, Abbe J-C, Bataille
R, Wijdenes J, Chatal J-F and Cherel M 1999 Validation of 213Bi-alpha
radioimmunotherapy for multiple myeloma Clin. Cancer Res. 5 3165s–70s
[19] Sgouros G, Ballangrud A M, Jurcic J G, McDevitt M R, Humm J L, Erdi Y E,
Mehta B M, Finn R D, Larson S M and Scheinberg D A 1999 Pharmacokinetics
and dosimetry of an alpha-particle emitter labeled antibody: 213Bi-HuM195 (Anti-
CD33) in patients with leukemia J. Nucl. Med. 40(11) 1935–46
[20] Hall E J 2000 Radiobiology for the Radiologist (Philadelphia: Lippincott, Williams
and Wilkins)
[21] Humm J L 1987 A microdosimetric model of astatine-211 labeled antibodies for
radioimmunotherapy Int. J. Radiat. Oncol. Biol. Phys. 13 1767–73
[22] Humm J L and Chin L M 1993 A model of cell inactivation by alpha-particle internal
emitters Rad. Res. 134 143–50
[23] Stinchcomb T G and Roeske J C 1992 Analytic microdosimetry for radioimmuno-
therapeutic alpha emitters Med. Phys. 19(6) 1385–93
[24] Polig E 1978 The localized dosimetry of internally deposited alpha-emitters Curr.
Top. Rad. Res. 13 189–327
[25] Olko P and Booz J 1990 Energy deposition by protons and alpha particles in spherical
sites of nanometer to micrometer diameter Radiat. Environ. Biophys. 28 1–17
[26] Eckerman K F, Ryman J C, Taner A C and Kerr G D Traversal of cells by radiation
and absorbed dose fraction estimates for electrons and alpha particles. In
Proceedings of the Fourth International Radiopharmaceutical Dosimetry Symposium
eds A T Schlafke-Stelson and E E Watson (CONF-851113) (Oak Ridge, TN: Oak
Ridge Associated Universities) pp 67–86
[27] Attix F H 1986 Introduction to Radiological Physics and Radiation Dosimetry (New
York: Wiley)
[28] Ziegler J F, Biersack J P and Littmark U 1985 The Stopping and Range of Ions in
Matter (New York: Pergamon)
[29] Ziegler J F 1988 The stopping and range of ions in solids. Ion Implantation: Science
and Technology (San Diego: Academic) pp 3–61
[30] Ziegler J F 1999 The stopping of energetic light ions in matter Appl. Phys. Rev., J.
Appl. Phys. 85 1249–72
References 223
[31] Bichsel H 1992 Stopping powers of fast charged particles in heavy elements NIST
Report IR-4550 (Springfield, VA: US National Technical Information Service)
[32] ICRU 1993 Stopping powers and ranges for protons and alpha particles Report No 49
(Bethesda, MD: International Commission of Radiation Units)
[33] Janni J F 1966 Calculation of energy loss, range, path length, straggling, multiple
scattering and the probability of inelastic nuclear collisions for 0.1 to 1000 MeV
protons Technical Report No AFWL-TR-65-150
[34] Walsh P J 1970 Stopping power and range of alpha particles Health Phys. 19 312–16
[35] Charlton D E and Sephton R 1991 A relationship between microdosimetric spectra
and cell survival for high-LET irradiation Int. J. Radiat. Biol. 59(2) 447–57
[36] Fisher D R 1985 The microdosimetry of monoclonal antibodies labeled with alpha
emitters. In Proceedings of the Fourth International Radiopharmaceutical Dosimetry
Symposium eds A T Schlafke-Stelson and E E Watson (CONF-851113) (Oak Ridge,
TN: Oak Ridge Associated Universities) pp 26–36
[37] Roesch W C 1977 Microdosimetry of internal sources Radiat. Res. 70 494–510
[38] Stinchcomb T G and Roeske J C 1995 Survival of alpha particle irradiated cells as a
function of the shape and size of the sensitive target (nucleus) Radiat. Prot. Dosim.
62(3) 157–64
[39] Kvinnsland Y, Stokke T and Aurilien E 2001 Radioimmunotherapy with alpha-
particle emitters: Microdosimetry of cells with heterogeneous antigen expression
and with various diameters of cells and nuclei Radiat. Res. 155 288–96
[40] Stinchcomb T G and Roeske J C 1994 Analysis of survival of C-18 cells after irradia-
tion in suspension with chelated and ionic bismuth-212 using microdosimetry
Radiat. Res. 140 48–54
[41] Charlton D E 1999 The survival of monolayers of cells growing in clusters irradiated
by 211At appended to the cell surfaces Radiat. Res. 151 750–3
[42] Aurlien E, Larsen R H, Akabani G, Olsen D R, Zalutsky M R and Bruland O S 2000
Exposure of human osteosarcoma and bone marrow cells to tumor-targeted alpha-
particles and gamma-irradiation: Analysis of cell survival and microdosimetry Int. J.
Radiat. Biol. 76(8) 1129–41
[43] Charlton D E, Kassis A I and Adelstein S J 1994 A comparison of experimental and
calculated survival curves for V79 cells grown as monolayers or in suspension
exposed to alpha irradiation from 212Bi distributed in growth medium Radiat.
Prot. Dosim. 52 311–15
[44] Kennel S J, Stabin M, Roeske J C, Foote L J, Lankford P K, Terzaghi-Howe M,
Patterson H, Barkenbus J, Popp D M, Boll R and Mirzadeh S 1999 Radiotoxicity
of bismuth-231 bound to membranes of monolayer and spheroid cultures of
tumor cells Radiat. Res. 151 244–56
[45] Charlton D E 2000 Radiation effects in spheroids of cells exposed to alpha emitters
Int. J. Radiat. Biol. 76(11) 1555–64
[46] Charlton D E and Turner M S 1996 Use of chord lengths through the nucleus to
simulate the survival of mammalian cells exposed to high-LET alpha-radiation
Int. J. Radiat. Biol. 69(2) 213–17
[47] Iyer M and Lehnert B E 2002 Alpha-particle-induced increases in the radioresistance
of normal human bystander cells Radiat. Res. 157 3–7
[48] Humphreys E R and Humm J L 1988 A Monte-Carlo approach to the micro-
dosimetry of 224Ra in murine compact and cancellous bone Health Phys. 54(6)
607–15
[49] Charlton D E, Utteridge T D and Beddoe A H 1996 Microdosimetry of

haemopoietic stem cells irradiated by alpha particles from short-lived products of
222
Rn decays in fat cells and haemopoietic tissue Int. J. Radiat. Biol. 69(5) 585–92
[50] Charlton D E, Salmon P L and Utteridge T D 1998 Monte Carlo/numerical
treatment of alpha-particle spectra from sources buried in bone and resultant
doses to surface cells Int. J. Radiat. Biol. 73(1) 89–92
[51] Austin A L, Ellender M, Haines J W, Harrison J D and Lord B I 2000
Microdistribution and localized dosimetry of the alpha-emitting radionuclides
239
Pu, 241Am and 233U in mouse femoral shaft Int. J. Radiat. Biol. 76(1) 101–11
[52] Akabani G and Zalutsky M R 1997 Microdosimetry of astatine-211 using histological
images: Application to bone marrow Radiat. Res. 148 599–607
[53] Charlton D E, Utteridge T D and Allen B J 1998 Theoretical treatment of human
haemopoietic stem cell survival following irradiation by alpha particles Int. J.
Radiat. Biol. Phys. 74(1) 111–18
[54] Utteridge T D, Charlton D E and Allen B J 2001 Monte Carlo modeling of the effects
of injected short-, medium- and longer-range alpha-particle emitters on human
marrow at various ages Radiat. Res. 156 413–18
[55] Ertl H H, Feinendegen L E and Heiniger H J 1970 Iodine-125, a tracer in
cell biology: physical properties and biological aspects Phys. Med. Biol. 15(3) 447–
56
[56] Hofer K G and Hughes W L 1971 Radiotoxicity of intranuclear tritium, 125 iodine
and 131 iodine Radiat. Res. 47(1) 94–101
[57] Chan P C, Lisco E, Lisco S and Adelstein S J 1978 Cell survival and cytogenetic
responses to 125I-UdR in cultured mammalian cells Curr. Top. Radiat. Res. 12(1–
4) 426–35
[58] Hofer K G 1980 Radiation biology and potential therapeutic applications of
radionuclides Bull. Cancer 67(3) 343–53
[59] Sastry K S 1992 Biological effects of the Auger emitter iodine-125: A review. Report
No 1 of AAPM Nuclear Medicine Task Group No 6 Med. Phys. 19(6) 1361–70
[60] Charlton D E and Booz J 1981 A Monte Carlo treatment of the decay of 125I Radiat.
Res. 87(1) 10–23
[61] Charlton D E 1986 The range of high LET effects from 125I decays Radiat. Res.
107(2) 163–71
[62] Howell R W 1992 Radiation spectra for Auger-electron emitting radionuclides:
Report No 2 of AAPM Nuclear Medicine Task Group No 6 Med. Phys. 19(6)
1371–83
[63] Nikjoo H, Martin R F, Charlton D E, Terrissol M, Kandaiy S and Lobachevsky P
1996 Modelling of Auger-induced DNA damage by incorporated 125I Acta Oncol.
35(7) 849–56
[64] Lobachevsky P N and Martin R F 1996 DNA strand breakage by 125I-decay in a
synthetic oligodeoxynucleotide. 2. Quantitative analysis of fragment distribution
Acta Oncol. 35(7) 809–15
[65] Lobachevsky P N and Martin R F 2000 Iodine-125 decay in a synthetic
oligodeoxynucleotide. I. Fragment size distribution and evaluation of breakage
probability Radiat. Res. 153(3) 263–70
[66] Humm J L and Charlton D E 1989 A new calculational method to assess the
therapeutic potential of Auger electron emission Int. J. Radiat. Oncol. Biol. Phys.
17(2) 351–60
References 225
[67] Narra V R, Howell R W, Harrapanhalli R S, Sastry K S and Rao D V 1992 Radio-

toxicity of some iodine-123, iodine-125 and iodine-131-labeled compounds in mouse
testes: implications for radiopharmaceutical design J. Nucl. Med. 33(12) 2196–201
[68] Humm J L 1984 The analysis of Auger electrons released following the decay of radio-
isotopes and photoelectric interactions and their contribution to energy deposition
(KFA Report Jülich No 1932, PhD thesis, Germany, June)
[69] Salem S I, Panossian S L and Krause R A 1974 Experimental K and L relative X-ray
emission rates At. Data Nucl. Data Tables 14 91–109
[70] Chen M H, Crasemann B and Mark H 1979 Relativistic radiationless transition
probabilities for atomic K and L shells At. Nucl. Data Tables 24 13–37
[71] Charlton D E and Humm J L 1988 A method of calculating initial strand breakage
following the decay of 125-I Int. J. Radiat. Biol. 53 353–65
[72] Paretke H 1987 Radiation track structure. In Kinetics of Non-Homogeneous
Processes ed C R Freeman (New York: Wiley)
[73] Martin R F and Haseltine W A 1981 Range of radiochemical damage to DNA with
decay of iodine-125 Science 21 213(4510) 896–8
[74] Humm J L, Howell R W and Rao D V 1994 Dosimetry for Auger electron emitters:
Report No 3 of AAPM Nuclear Medicine Task Group No 6 Med. Phys. 21 1901–15
[75] Goddu S M, Howell R W and Rao D V 1996 Calculation of equivalent dose for
Auger electron emitting radionuclides distributed in human organs Acta Oncol.
35(7) 909–16
[76] Goddu S M, Rao D V and Howell R W 1994 Cellular dosimetry: absorbed fractions
for monoenergetic electron and alpha particle sources and S-values for radionuclides
uniformly distributed in different cell compartments J. Nucl. Med. 35(2) 303–16
[77] Goddu S M, Rao D V and Howell R W 1994 Multicellular dosimetry for
micrometastases: dependence of self-dose versus cross-dose to cell nuclei on type
and energy of radiation and subcellular distribution of radionuclides J. Nucl.
Med. 35(3) 521–30
[78] Bagshaw K D, Sharma S K, Southall P J, Boden J A, Boxer G M, Partridge T A,
Antoniw P and Pedley R B 1991 Selective uptake of toxic nucleoside (125IUdR) by
resistant cancer Br. J. Radiol. 64(757) 37–44
[79] Mariani G, Di Sacco S, Volterrani D, Di Luca L, Buralli S, Di Stefano R,
Baranowska-Kortylewicz J, Bonora D, Matteucci F, Ricci S, Bellina C R, Falcone
A, Salvadori P A, Mosca F, Adelstein S J and Kassis A I 1996 Tumor targeting
by intra-arterial infusion of 5-[123I]iodo-20 -deoxyuridine in patients with liver
metastases from colorectal cancer J. Nucl. Med. 37(4) (Suppl) 22s–25s
[80] Chiou R K, Dalrymple G V, Baranowska-Kortylewicz J, Holdeman K P, Schneider-
man M H, Harrison K A and Taylor R J 1999 Tumor localization and systemic
absorption of intravesical instillation of radio-iodinated iododeoxyuridine in
patients with bladder cancer J. Urol. 162(1) 58–62
[81] Macapinlac H A, Larson S M, Daghighian F, Finn R, Zhang J, Humm J L, Squire O
and Kemeny N 1996 I-125-50 -Iododeoxyuridine (IUdR) in the treatment of colorectal
cancer metastatic to the liver: 1. Pilot clinical trial J. Nucl. Med. 37 25s–29s
[82] Kassis A I, Adelstein S J and Mariani G 1996 Radiolabeled nucleoside analogs in
cancer diagnosis and therapy J. Nucl. Med. 40(3) 301–19
[83] Baranowska-Kortylewicz J, Kortylewicz Z P, Hoffman D, Winoto A, Lai J and
Dalrymple G V 1996 Colon-specific prodrugs of 5-radioiodo-20 -deoxyuridine Acta
Oncol. 35(7) 959–64
[84] Karagiannis T C, Lobachevsky P N and Martin R F 2000 Cytotoxicity of an 125I-

labelled DNA ligand Acta Oncol. 39(6) 681–5
[85] Panyutin I G, Winters T A, Feinendegen L E and Neumann R D 2000 Development
of DNA-based radiopharmaceuticals carrying Auger-electron emitters for anti-gene
radiotherapy Q. J. Nucl. Med. 44(3) 256–67
[86] DeSombre E R, Hughes A, Hanson R N and Kearney T 2000 Therapy of estrogen
receptor-positive micrometastases in the peritoneal cavity with Auger electron-
emitting estrogens—theoretical and practical considerations Acta Oncol. 39(6)
659–66
[87] Brady L W, Miyamoto C, Woo D V, Rackover M, Emrich J, Bender H, Dadparvar
S, Steplewski Z, Koprowski H and Black P 1992 Malignant astrocytomas treated
with iodine-125 labeled monoclonal antibody 425 against epidermal growth factor
receptor: a phase II trial Int. J. Radiat. Oncol. Biol. Phys. 22(1) 225–30
[88] Derui L, Woo D V, Emrich J, Steplewski Z, Rodeck U, Herlyn D, Koprowski H,
Miyamoto C and Brady L W 1992 Radiotoxicity of 125I-labeled monoclonal
antibody 425 against cancer cells containing epidermal growth factor receptor Am.
J. Clin. Oncol. 15(4) 288–94
[89] Welt S, Scott A M, Divgi C R, Kemeny N E, Finn R D, Daghighian F, St Germain J,
Richards E C, Larson S M and Old L J 1996 Phase I/II study of iodine 125-labeled
monoclonal antibody A33 in patients with advanced colon cancer J. Clin. Oncol.
14(6) 1787–97
[90] Adelstein S J, Kassis A I, Baranowska-Kortylewicz J, van den Abbeele A D, Mariani
G and Ito S 1991 Potential for tumor therapy with I-125 labelled immunoglobulin
Int. J. Radiat. Appl. Inst. B 18 43–4
[91] Mariani G, Bodei L, Adelstein S J and Kassis A I 2000 Emerging roles for
radiometabolic therapy of tumors based on auger electron emission J. Nucl. Med.
41(9) 1519–21
[92] Daghighian F, Barendswaard E, Welt S, Humm J L, Scott A, Willingham M C,
McGuffie E, Old L J and Larson S M 1996 Enhancement of radiation dose to the
nucleus by vesicular internalization of iodine-125-labeled A33 monoclonal
antibody J. Nucl. Med. 37 1052–7
[93] Behr T M, Behe M, Lohr M, Sgouros G, Angerstein C, Wehrmann E, Nebendahl K
and Becker W 2000 Therapeutic advantages of Auger electron- over beta-emitting
radiometals or radioiodine when conjugated to internalizing antibodies Eur. J.
Nucl. Med. 27(7) 753–65
[94] Roeske J C and Stinchcomb T G 1997 Dosimetric framework for therapeutic alpha-
particle emitters J. Nucl. Med. 38 1923–9
[95] Stinchcomb T G and Roeske J C 1999 Values of ‘S’, hz1i and hz12i for dosimetry
using alpha-particle emitters Med. Phys. 26(9) 1960–71
[96] Roeske J C and Stinchcomb T G 1999 The use of microdosimetric moments in
evaluating cell survival for therapeutic alpha-particle emitters Radiat. Res. 151 31–8
[97] Pomplun E, Booz J and Charlton D E 1987 A Monte Carlo simulation of Auger
cascades Radiat. Res. 111(3) 533–52
[98] Pomplun E 2000 Auger electron spectra—the basic data for understanding the
Auger effect Acta Oncol. 39(6) 673–9
[99] Stepanek J 2000 Methods to determine the fluorescence and Auger spectra due to
decay of radionuclides or due to a single atomic-subshell ionization and comparisons
with experiments Med. Phys. 27(7) 1544–54
References 227
[100] Goddu S M, Howell R W, Bouchet L G, Bolch W E and Rao D V 1997 MIRD

Cellular S Values (Reston, VA: Society of Nuclear Medicine)
[101] Humm J L, Macklis R M, Bump K, Cobb L M and Chin L M 1993 Internal
dosimetry using data derived from tissue autoradiographs J. Nucl. Med. 34 1811–17
[102] Soyland C, Hassfjell S P and Steen H B 2000 A new alpha-particle irradiator with
absolute dosimetric determination Radiat. Res. 153 9–15
[103] Soyland C and Hassfjell S P 2000 A novel 210Po-based alpha-particle irradiator for
radiobiological experiments with retrospective alpha-particle hit per cell
determination Radiat. Environ. Biophys. 39 125–30
[104] Soyland C and Hassfjell S P 2000 Survival of human lung epithelial cells following in
vitro alpha-particle irradiation with absolute determination of the number of alpha-
particle traversals of individual cells Int. J. Radiat. Biol. 76 1315–22
Chapter 10
The MABDOSE program for internal

radionuclide dosimetry
Timothy K Johnson
10.1. INTRODUCTION—THE NEED FOR BETTER DOSIMETRY
Internal radionuclide dosimetry from unsealed sources of radioactivity has

always been a challenge. A formalism for estimating organ doses from
internally deposited radionuclides was published in 1968 by the Medical
Internal Radiation Dose (MIRD) Committee of the Society of Nuclear
Medicine [1, 2]. In its most general form, the MIRD formalism reduced
the dosimetry problem to one of adding all of the emitted radiation from a
radionuclide located within a given region, and depositing fractions of that
emitted radiation in individual organs. The practical implementations of
this formalism—the MIRD Pamphlet 11 [3] and the comparable MIRDOSE
program [4]—were based around the construction of a geometric phantom
that approximated the average dimensions and locations of human organs.
The decay of a variety of radionuclides was simulated, with the radionuclides
sequentially located within regions of the geometric phantom corresponding
to regular organ and anatomic spaces. The results of this simulation were
tabulated for each radionuclide.
Unfortunately, the manner in which these tables were published—as static
objects, existing on either hardcopy paper or as software lookup tables—made
them unable to adapt to the changing requirements imposed by rapid advances
in biotechnology. Specifically, the series of tables were not designed to account
for the effects of an arbitrarily shaped source of activity at an arbitrary
location. Additionally, deviation of a person’s anatomy from that of the
geometric phantom could give rise to substantial errors in dose calculations,
since the assigned dose is inversely proportional to the organ’s actual mass.
In general, these tables were used in the evaluation of diagnostic amounts of
radioactivity. The errors associated with this use were not significant in
terms of determining whether a diagnostic amount of activity was safe or
228
Dosimetry software design considerations 229
not. Dose estimates that differed from the ‘true’ value by factors of 2 to 10 still
resulted in numbers that fell within the diagnostic range. This allowed for the
creation of simple dose-per-unit administered activity tables such as are found
on all package inserts for modern day commercial radiopharmaceuticals.
The advent of hybridoma technology raised the possibility of attempting
to treat cancer with antibodies targeted to specific tumour antigens.
Antibodies offer a carrier medium that is potentially selective at the
molecular level. Hybridoma technology created a vehicle whereby the
antibodies could be manufactured in abundance. The possibility of tagging
these ‘magic bullets’ with radioactivity to treat cancers became suddenly a
reality. The desire to use radiolabelled antibodies in a therapeutic modality,
however, placed an entirely new demand on the dosimetry community.
Factors of 2 and 10 were no longer acceptable. Delivering an estimated dose
of 200 mCi to the bone marrow would allow a patient to survive. If this esti-
mate were off by a factor of 2, the difference could result in death.
The initial clinical trials that infused radiolabelled antibodies into
patients demonstrated that the concept of a ‘magic bullet’ was a mistaken
one. Radioactivity did not localize in the tumour exclusively: it concentrated
in other organs that were clearly visible on whole body scans. If the carrier
antibody was classified as a ‘good’ antibody, it localized preferentially in
the tumour, but NOT exclusively. Consequently, investigators were forced
to consider the following question: Given that radioactivity localizes in tissues
other than tumour, how much radioactivity can the patient tolerate? This
question had to be weighed in conjunction with its corollary: How much
radioactivity must be administered to sterilize the tumour? Having to consider
both questions simultaneously, investigators wondered if it was possible to
obtain the latter before the former. Answers required accurate knowledge
of the kinetics of a particular antibody, the physical decay parameters
associated with the radiolabel, and a thorough knowledge of the geometric
composition and anatomic makeup of an individual patient. This knowledge
did not exist in a form that was usable on a practical scale.
10.2. DOSIMETRY SOFTWARE DESIGN CONSIDERATIONS
To begin addressing the problem of tumour dose versus normal organ dose in
the context of radiolabelled antibodies, the sources of error during the
process of forming an internal radionuclide dose estimate were analysed.
Additionally, the pace of computer technology development was evaluated.
These areas of inquiry resulted in a list of design features that an optimal
dosimetry solution should consider. The solution that was arrived at—a
practical software tool for estimating monoclonal antibody doses—was
developed in direct response to the design considerations. These design
considerations are summarized below.
230 The MABDOSE program for internal radionuclide dosimetry
10.2.1. Dosimetry considerations
First, it is impossible to quantify activity accurately at the voxel level using

NaI as a detector. The goal of most imaging studies in nuclear medicine is
to maximize contrast between areas of low and high uptake. Nuclear medi-
cine images are quantum limited, in the sense that photon statistics drive
the overall impression of image quality. In order to obtain adequate count
statistics in a reasonable period of time, cameras routinely acquire images
using a 10% energy window. The consequence of using a window this
wide is that scattered radiation is accepted into the image acquisition
matrix, distorting the impression of underlying activity. Typical resolution
within a patient scatter medium is 1.0–1.5 cm. Although camera manufac-
turers will specify 3–4 mm resolution at 99m Tc energies, this is for a line
source at the face of the collimator, and is not realizable in a patient under
normal acquisition constraints of limited time and limited activity.
Second, other measurements having higher accuracy than NaI(Tl)
imaging are available for quantifying some organ spaces. On a regular
basis, the spaces that can be directly sampled include the urine and blood.
Less frequently, faecal samples can be collected and counted, as can serial
biopsies of superficial tumours and some normal organs. The latter may
include bone marrow, lung and liver. Obviously, it makes no sense to
implement a dosimetry approach that relies solely on imaging data at the
expense of throwing away higher-accuracy data. Unless a methodology
can be adapted to a variety of input data forms, it risks incorporating
unreliable and inaccurate data at the expense of higher-accuracy data.
Third, activity at a given physical location changes as a function of time,
the product of both physical decay and biological accumulation and
excretion mechanisms. It is not enough to calculate the activity at a single
point in time. One must align the location for each activity measurement,
fit those data points to some sort of mathematical function, and integrate
the function to achieve the total amount of radiation released from that
location. This time integral is referred to as the cumulated (or cumulative)
activity. In the limit, this implies that each voxel must be precisely aligned
with its temporally displaced counterparts derived from images obtained at
different times. Alignment of this precision is currently only practised in
radiation oncology teletherapy treatment plans. Because of the poor
resolution in nuclear medicine (1–1.5 cm in patients), sets of marker sources
would have to be placed on each and every patient at each and every time
point for alignment. Attention to marker source placement, collimator
angulation, gantry rotation, bed position, and collimator-to-patient distance
would have to be rigorously observed. This level of rigour is currently not
practised in most nuclear medicine departments. The reality of internal
organs shifting from day to day as a patient moves implies that this is a
practical impossibility.
Dosimetry software design considerations 231
Fourth, the total body remains as the single most important space in
which intravenously administered activity concentrates. This implies that
quantification of the total body should be performed. If one chooses to use
a SPECT acquisition, this becomes time prohibitive. Typically, when
SPECT has been used, investigators have decided to select for quantification
only organs that fall within the field of view of a single acquisition rotation. If
one is serious about quantifying activity at the voxel level, one should
quantify each and every voxel where activity occurs. Investigators, however,
evidently find this too great an expenditure of time since no one has ever
performed it in a clinical trial. The primary reason for not quantifying
total body activity from a series of total body SPECT acquisitions (aside
from the time commitment) is the gross errors introduced into the quantifi-
cation process, upwards of 400%!
Fifth, the most radiosensitive organ in the body is the bone marrow.
The current experience with radiolabelled antibodies suggests that binding
affinities are not so great that one gets wholesale uptake in the tumour
and none anywhere else. With intravenous infusions of radiolabelled anti-
body and the assumption that higher amounts of activity will result in
improved response rates, the bone marrow continues to be the critical
organ. This organ (more properly an organ space) is beneath the resolution
limits of modern scintillation cameras. For the foreseeable future, it must be
dealt with by means of a mathematical model that is predictive of biological
outcome. This model should be capable of incorporating measures of bone
marrow size (reserve) as well as measures of bone marrow viability
(patency) given the reality that radioimmunotherapy patients have usually
undergone several rounds of chemotherapy. This treatment is only
amenable to a mathematical modelling analysis. This statement derives
from the fact that the anatomy of interest is not visualizable on a practical
basis.
Sixth, the fact that cancers are the target for infused radiation means
that one does not have the luxury of being able to use an ‘average’ organ
located at an average position within the mathematical representation of
an average person. Each patient that presents will have a unique number
of cancer foci. Each focus will have a unique position and shape within the
patient’s body. A completely general dosimetry program should have the
ability to allow for the creation and construction of these arbitrary volumes
on a patient-by-patient basis, and incorporate them into the dose calculation
scheme.
Seventh, in 1985 the speed of microprocessors was increasing regularly,
while the surface area occupied by integrated circuits was shrinking. The
likelihood for CRAY supercomputer speeds on a desktop computer system
in the near future was considered high. Calculation algorithms that
previously had been considered time prohibitive might be judged feasible if
given a second look.
10.2.2. Dosimetry approaches
The dosimetry solutions divided into three general approaches: convolution,

adaptation of the existing MIRD paradigm, and reinventing the wheel under
a more general guise. The preceding observations immediately ruled out the
convolution approach: inability to quantify activity at the voxel level,
inability to align each and every voxel with its temporally separated
counterparts, and the use of convolution kernels derived from transport in
a homogeneous media when the human body has a minimum of three distinct
media, weighed heavily against adopting this solution. Although it was
possible to fabricate situations and target geometries where this solution
would yield accurate results, it was equally possible to fabricate situations
where the solution was totally wrong. The quest for generality left this
solution unacceptable.
The second approach—that of trying to adapt the current MIRD
tables—was equally unacceptable. The primary problem was that there was
no simple way of scaling dose estimates to account for differences in patient
geometry from the Reference Man phantoms. Similarly, there was no accurate
method for including a tumour dose that was completely general. With regard
to patient geometry, depending on how large and how close a real target organ
was to a source of activity would determine how good an estimate the
MIRD11 absorbed fraction was. As with the convolution solutions, although
it was possible to invent situations and source/target organ pairs where this
approximation was a good one, it was equally possible to invent situations
where this approximation was poor. With respect to absorbed fractions for
a tumour, the photon contribution would suffer similar shortcomings. It
would deteriorate into an infinite number of perturbations, with each
perturbation appropriate to an individual set of conditions.
The implemented solution involved a return to the fundamentals upon
which the MIRD formalism was based. This solution was envisioned as an
integrated program, and was christened MABDOSE [5–14]. MABDOSE
incorporates solutions that address each of the first six points enumerated
above, and takes advantage of the increases in computer CPU speeds to
address point seven. The foundation of the program is the assumption that
homogeneous activity concentrations exist at a statistical level that is deter-
mined by the user. As a fairly intuitive first approximation, the qualities of
cells that define an organ’s function are also the qualities that should result
in homogeneity of radioactivity concentration. Even in an organ system
having more than one distinct cell type (e.g., the liver, composed of Kupffer
cells and hepatocytes), the relative distribution of cell types throughout the
organ is approximately the same in each unit mass of organ tissue, resulting
in uniform uptake. This makes the idea of activity quantification on an organ
basis a reasonable starting point for a statistical criterion of homogeneous
activity concentration.
Historic development of MABDOSE 233
If a single organ concentrates activity inhomogeneously, that organ can

be subdivided into distinct subregions until the criterion of activity
homogeneity is met. Each subregion is then treated as an independent
activity source. At an acceptance level determined by the user’s segmentation
resolution requirements, any remaining activity variation among a single
organ’s voxels is assumed due to measurement error. The fact that one
typically has available a large ensemble of voxels with which to average
out statistical variations implies that this approach should be favoured
over inaccurate unit voxel approaches. Calculating a total organ activity at
a single time point, then apportioning that activity equally among all
voxels that compose that organ, is the approach adopted by MABDOSE.
10.3. HISTORIC DEVELOPMENT OF MABDOSE
10.3.1. Version 1
The development of the MABDOSE program has evolved in a series of
stages. The first version of the program was presented at the Annual Meeting
of the American Association of Physicists in Medicine in 1986 [5]. To set the
historic record straight, the original version was called MABDOS because
file names were restricted to being six letters under the DOS operating
system (The MIRDOS program that Oak Ridge produced one year later
was subjected to similar constraints). With the relaxation in subsequent
years of filename size limits to allow longer and more descriptive filenames,
the current appellation is MABDOSE. The name is an acronym that
stands for Monoclonal Antibody Dosimetry. The acronym is a bit
misleading, because it is indicative of the program’s origins rather than its
functionality. MABDOSE is a completely general internal radionuclide
dosimetry program. It was initially conceived to address the shortcomings
inherent in the standard approach when dealing with dosimetry problems
posed by monoclonal antibodies as carrier molecules for radionuclides. It
is not restricted, however, to only performing dose estimates for monoclonal
antibodies.
The first version of the MABDOSE program had defining characteristics
that included (i) selection of an arbitrary radionuclide from the spectral library
assembled by Kocher [16], (ii) creation of a graphical interface for defining
mathematical models that predates the appearance of Apple’s Macintosh
GUI, (iii) curve fitting routines as alternatives to mathematical modelling,
(iv) incorporation of numerical integration routines to obtain cumulated
activities, (v) implementation of the ALGAM67 Monte Carlo code for
photon transport, (vi) incorporation of local deposition of non-penetrating
radiation for modelling electron and particulate radiation transport and (vii)
display of mean organ dose in a list format [7]. The program used the
equations for the Adult Reference Man phantom [17] to create a ‘logical’
target geometry, as opposed to a ‘virtual’ target geometry. The difference is
important: the former merely exists as a set of equations that are checked
for determining whether an x=y=z coordinate triplet falls within the volume’s
boundary. Thus, if the left kidney were declared a source organ, and a starting
coordinate desired, the equations for the left kidney would be consulted. A
random number generator would be used to obtain an arbitrary x=y=z co-
ordinate triplet from within the rectangular volume that just superscribed
the organ’s surface boundary. The triplet would then be checked against the
equations governing the left kidney to determine if the triplet satisfied the
defining equations.
The Monte Carlo simulation engine of the MABDOSE code was
derived from ALGAM67, a software code originally obtained from the
Oak Ridge Radiation Shielding Information Center in 1984. Historically,
ALGAM67 was the code that was used to create the MIRD11 tables. It
was deemed prudent at the time to use the same simulation code, because
ultimately the MABDOSE results would be compared back to the
MIRD11 tables as the nominal standard for internal radionuclide dosimetry
calculations.1 The simulation was structured so that each photon greater
than 10 keV was simulated for the same, user-determined, number of
histories. A random starting coordinate was selected from within a declared
source organ, and the photon tracked until it had either deposited all of its
energy or had escaped from the target phantom. Average dose to organs
identified in the Reference Man phantom were reported in a list.
The MABDOSE program was written in Fortran. The initial phase of a
dose estimate (source organ declaration, time–activity data entry, curve
fitting, curve integration, radionuclide selection) and the final phase
(combination of absorbed fractions with cumulated activities, display of
dose by organ) were completed on a Texas Instruments Professional
microcomputer. The intermediate portion of the program (Monte Carlo
simulation of radioactive decay using the declared source organs and
declared radionuclide spectrum) was performed on a Cray supercomputer.
This intermediate step necessitated the uploading of the solution’s inter-
mediate ‘state’, then downloading the results for final dose output.
10.3.2. Version 2
The second version of the program implemented a number of refinements to
Version 1. The speed of microcomputers was increasing exponentially. It was
1
In the interim, three other codes have gained prominence in simulation circles: EGS4, BCMP
and ETRAN. Their popularity does not detract from the applicability of the ALGAM67
algorithm, nor the practical extensions to it that includes non-penetrating radiation from a
nuclide’s spectrum.
Historic development of MABDOSE 235
readily seen that it was only a matter of time before microcomputers would
have calculation speeds approaching those of a CRAY. The simulation
portion of the program was ported to a microcomputer platform, resulting
in a tightly coupled program intended to execute completely from within a
single desktop hardware platform. At the same time, the entire MABDOSE
program was rewritten in the Cþþ language. This rewrite was occasioned by
the large size the original program had grown to. Changes to the code for
refinements, enhancements, and added functionality were impeded by the
use of a non-object-oriented language.
10.3.3. Version 3
The third version of the program opted for the creation of a full-fledged
virtual target geometry dataset. A three-dimensional lattice was physically
created in computer memory, with each voxel in the lattice representing
a single tissue type of the target. Each voxel in the lattice contained a
single index representative of its tissue type. The collection of lattice voxels
created a target geometry that served as the space in which the Monte
Carlo simulation was carried out. The same space served as the means for
declaring sources of activity. The source data structure encapsulated a
column vector whose elements corresponded to all voxels containing the
index representing that source. Selecting a starting coordinate was reduced
to calling a pseudo-random number generator to provide a value distributed
between 0 and 1, then multiplying by the length of the source’s column
vector.
Once created, a three-dimensional target lattice could be stored in a
database library for future recall. The rules governing the six different
Reference Man phantoms specified by Cristy and Eckerman [17] were imple-
mented in a hidden layer. If a target of the selected age and resolution existed
in the library, it was ‘checked out’ for use in the current simulation. If the
target did not exist, it was created by first allocating the lattice in memory.
The equations governing the appropriate age phantom were then consulted
to determine which organ index should be assigned each voxel.
10.3.4. Version 4
The fourth version of the program decoupled the target lattice from the
Reference Man phantom. This was accomplished by prefacing the lattice
data set with a header. The header contains a simple list of each voxel type
that will be encountered when reading in the lattice. Each voxel type is
represented by three fields: the first field is the integer index assigned to
represent that tissue type; the second is an ASCII descriptor of the tissue
type the index represents; and the third is an integer index that represents
the medium type, governing the voxel’s interaction characteristics with
radiation, and currently restricted to selection from among bone, soft tissue
and lung tissue. In a manner analogous to that of decoupling the target lattice
from the Reference Man, it is possible to generalize the assignment of media
types so that they can be incorporated after dynamic declaration (see section
10.5).
10.4. MABDOSE: FROM START TO FINISH
10.4.1. Overview
The majority of an internal radionuclide dose estimate deals with manipulating
data into a form that is usable by other program components. Too much
emphasis has been placed on the fact that MABDOSE uses a Monte Carlo
algorithm for photon transport based on the ALGAM67 code, and that it
does not use one of the more modern codes such as EGS4 or ECMP. The
reality is that MABDOSE accounts for all radiation: the algorithm used for
transporting penetrating (i.e., photon) radiation is based on ALGAM67; the
algorithm used for transporting non-penetrating radiation (i.e., electrons,
particles, photons <10 keV, and particles) is local absorption. It is logically
inconsistent to simulate electron transport when the resolution from nuclear
medicine scans routinely supersedes the range of particles. Instead, R&D
effort on MABDOSE has been devoted to generalizing the user interface for
ease of use, and to expanding the range of problems for which the interface
is applicable. This means that doses at boundaries between source and
target will necessarily have an error associated with them, since particles
will in reality ‘spill over’ the boundary edge and lose some of their energy in
the adjacent tissue, rather than stopping abruptly at the interface boundary.
This error is mitigated somewhat by adjacent whole body activity.
MABDOSE requires four objects be declared during a dosimetry session.
These are (i) a target lattice, (ii) a set of source volumes, (iii) cumulated
activities for each source volume in the set and (iv) a radionuclide from
which the dose is desired. Each of these four objects is discussed below in
detail. The order for making these objects known to the program is general,
with two exceptions. First, a source cannot be declared until a target lattice
has been created. Second, cumulated activities for a source cannot be
generated until the source has been defined. A declared source is not
considered defined until it has at least one time–activity point assigned to it.
10.4.2. The target lattice

The foundation of the MABDOSE program is the concept of the target
lattice. This exists as a three-dimensional array that is dynamically allocated
at run time. Dynamic memory allocation generalizes the memory requirements
MABDOSE: from start to finish 237
necessary for program execution: computers having smaller amounts of RAM

can still execute MABDOSE at the expense of being limited to a coarser
resolution lattice.
As mentioned previously, Version 3 of the program was the first to
implement the target lattice as a virtual three-dimensional array. This lattice
was restricted to the family of Reference Man phantoms, with resolution of
10, 5, 2 and 1 mm3 voxels. The coordinates of each voxel were tested against a
chain of if–then loops to determine which organ system the voxel belonged
to. The voxel was then assigned an integer index representing the organ
system. The defined lattice was stored to disk in a library of target lattices.
If the lattice representing the selected Reference Man phantom at the selected
resolution did not exist in memory, it was created and stored to disk.
Subsequent selection of that particular resolution phantom resulted in the
lattice being read in from disk.
In the original version of MABDOSE, the Reference Man geometry could
be redefined to include spaces that functioned as tumour foci. A graphical
interface for tumour definition allowed the operator to place spherical pertur-
bations within the Reference Man geometry. These perturbations functioned as
additional source/target organs. Although they could in theory represent any
separate and distinct source of—or target for—radiation, they were typically
associated with tumours.
The current version of MABDOSE retains this capability with respect to
lattices composed of the Reference Man phantoms. However, the generaliza-
tion of a source data structure to be potentially whatever is segmented (see
section 10.3.4) leaves the decision of what needs to be segmented up to the
individual. If there are tumour spaces that are deemed significant, and
worthy of segmentation, they can be contoured and assigned an organelle
index. They will then be made available to the MABDOSE system as a
distinct target and potential source for radioactivity within the three-
dimensional target lattice. This obviates the need for all of the graphics
routines that redefine the target lattice. As an example, the Zubal segmented
three-dimensional computed tomography data set ‘YaleMan’ is bundled with
the current version of MABDOSE as a target lattice. A user can employ this
actual human anatomic data set as a target lattice instead of a Reference Man
phantom. Alternatively, any three-dimensional data set can be read in
provided the structures of interest have all been segmented.
10.4.3. Declaring source volumes

The MABDOSE program creates three rectangular panes on the lower half
of the home screen. Once a target lattice is defined (see above), the list of
organelles that compose that target lattice is displayed on the leftmost
pane (figure 10.1). Each organelle is represented by a unique index. A
unique index is assigned to each voxel of the target lattice.
Figure 10.1. The MABDOSE main window, showing descriptors for the
possible voxel ‘types’ in the bottom left most list. The source ‘Kidneys’ has
been constructed in the bottom middle list from the voxel types, ‘L Kidney’
and ‘R Kidney’. When this list is closed, the source ‘Kidneys’ will be added to
the patient source list displayed in the bottom right. Each source must have a
separate time–activity data set associated with it. Cumulated activity is
apportioned equally among voxels that comprise the source.
The target list in the left-most pane corresponds to the possibilities for
declaring and defining a source. A source is a logical construct that has
a physical space in the target lattice associated with it. The underlying
theme of a source is homogeneity of activity concentration: no matter
what sub-volume of a given source is sampled, when compared with another
sub-volume from the same source, the activity concentration will be the
same. This is realizable in a statistical sense. Thus, one could expect both
the left and right kidneys to handle a given radiopharmaceutical in exactly
the same way. One could then construct a single source, called ‘Kidneys’
from the two volumes that comprise the left and right kidney spaces.
The centre pane is essentially a temporary workspace that allows one to
construct a source from the left-most list of organelle possibilities. When a
list of all organelles that are to be treated as a single source has been
constructed, the list in the centre pane is closed, and the source added to
the patient’s current source list. The right-most pane on the lower half of
the program’s main window represents the current source list.
10.4.4. Time-activity data and the generation of cumulated activities
For each declared source, an estimate must be made of the amount of activity
in that source as a function of time. This is frequently, but not exclusively,
derived from nuclear medicine scintigrams. The ability to incorporate higher
accuracy measurements from blood samples, urine collections and tissue biop-
sies motivated the development of the current MABDOSE interface. Each
source has associated with it a list of time–activity data point pairs. The inter-
face object for identifying time–activity point pairs to the MABDOSE system
is a two-column spreadsheet. Right clicking on a declared source displays the
spreadsheet associated with that source. Individual time–activity data are
input according to the abundance obtained for each source.
The time–activity data are fitted to a mathematical function, and
integrated. The integral of the time–activity curve gives the total number
of nuclear disintegrations, a quantity known as the cumulated activity.
Combined with knowledge of the radionuclide’s decay spectrum, the type
and total number of each energy in the spectrum is known. The cumulated
activity is divided by the number of voxels that compose the source being
simulated. Again, the rationale for doing this is the concept of source homo-
geneity: what defines a source is uniformity of function over the set of
discrete voxels that compose the source.
The mathematical functions that are integrated to achieve the cumu-
lated activity are arrived at using two main approaches. The first is curve
fitting. Curve fitting presupposes that the operator knows how the functional
form of the data should look, although not the function’s magnitude. There-
fore, the operator must have an appropriate function beforehand. The
functions that MABDOSE currently implements are a mono-exponential
function, a bi-exponential function, and a linear function. Additionally,
MABDOSE can perform a piecewise integration using discrete trapezoidal
or Riemann steps between data points.
The second technique for obtaining cumulated activities is mathematical
modelling. MABDOSE implements a graphical interface for composing a
mathematical model (figure 10.2). Each source has associated with it a
colour-coded rectangle. These are positioned on the screen in a format that
is convenient to the user. Directed pathways are then established by starting
a line segment within an activity source, and terminating the line segment in
an activity sink. The interrelationships specified by the created model define a
series of differential equations whose solutions are sums of exponentials.
An initial estimate of the exchange rate between linked compartments is
made. Starting from these estimates, a derivative-free version of the
Levenberg–Marquardt algorithm minimizes the sum of squares difference
between the raw data points and the differential equations by altering the
rate constants in a stepwise fashion [18]. The amplitude of each exponential
term corresponds to an initial value problem, and can be calculated given the
Figure 10.2. Sample model definition describing a simple two-compartment

model.
boundary condition of the activity distribution at time t ¼ 0. For antibody

problems, this boundary condition is typically taken to be 100% of the
activity distributed homogeneously in the total body at time t ¼ 0.
This analysis results in a least squares estimate of the individual decay
constants, and a functional representation of the time–activity curve for
each organ system that localizes activity. These differential equations are
integrated using a fourth-order Runge–Kutta method to arrive at cumulated
activities.
10.4.5. Selection of a radionuclide

The list of radionuclides available in MABDOSE is derived from the DLC-80
database of Kocher [16]. The radionuclide list is displayed in a scrollable
window. The display is enhanced by an adjacent window exhibiting an
individual radionuclide’s decay data. Highlighting a radionuclide in the
scrollable list results in the adjacent window showing the nuclide’s spectrum,
with all radiations divided by type (, positrons, electrons, photons), their
percentage abundance, maximum energy and mean energy (if the energy
has an associated distribution).
10.4.6. Simulation of radioactive decay

Prior to simulating radioactive decay, the radiation spectrum is divided into
penetrating and non-penetrating radiation types. For the penetrating radia-
tion, the percentage abundance of each radiation above 10 keV is summed to
form a cumulative probability distribution. This probability distribution
is sampled to determine the energy of the photon to be simulated. This
probability distribution preferentially samples those photons that are most
abundant.
10.4.7. Simulation—cross section lookup tables

Cross section data for the Monte Carlo transport of photons is obtained
from a lookup table maintained in system RAM. The lookup table brackets
an energy range from 2 keV to 16 MeV. This energy range is divided into 13
intervals, with 64 energy bins allocated per interval. Instead of linearly
dividing the energy range into 13 equal intervals, each interval represents a
doubling of the interval preceding it. Thus the first interval ranges from 2
to 4 keV; the second interval from 4 to 8 keV; the third interval from 8 to
16 keV; and so on. This scheme allows for increased resolution at lower
energies where dramatic changes in the photoelectric interaction coefficient
are realized.
The lookup table is constructed for composite media using the DLC-7c
photon cross section library assembled by Roussin et al. [19]. This library is
an elemental cross section database; cross sections for composite materials
(i.e., soft tissue, lung and bone) are constructed using a mass weighted aver-
age of the interaction coefficients for the media’s constituent atoms [20, 21].
The compositions for soft tissue, lung and bone used by MABDOSE are
those specified by Cristy and Eckerman [17].
In accord with the ALGAM transport code used to generate the data for
MIRD Pamphlet No 5 [22] (and subsequently used to generate MIRD
Pamphlet No 11), the lookup table is constructed to reflect incoherent
(Compton) scatter, pair production, and total attenuation coefficients for a
composite medium. With respect to incoherent scatter, the interaction
coefficient is calculated according to the Klein–Nishina equation for a free
electron, and scaled up by each element’s electron density. After weighting
each elemental coefficient by the element’s mass abundance in the composite
media, the coefficients are summed. The interaction coefficient is stored as a
probability after dividing it by the total attenuation coefficient.
With respect to pair production, probabilities are constructed using a log–
log interpolation of elemental pair production cross sections read in directly
from the DLC-7c library. Each elemental coefficient is weighted by the
element’s mass abundance in the composite media and summed. Analogous
to the incoherent scatter coefficient, the composite pair production interaction
coefficient is divided by the total attenuation coefficient and stored as a pure

probability.
With respect to total attenuation, interaction coefficients are con-
structed by summing the incoherent scatter and pair production cross
sections created as above, and augmenting the sum with a log–log
interpolation of elemental photoelectric cross sections read in from DLC-
7c. Appropriate conversion factors are applied in order to store the total
attenuation coefficient in units of cm1 . Coherent scatter interactions are
deemed negligible, and are ignored.
It should be noted that photoelectric probabilities are not stored in the
lookup table. Instead, they are calculated at run time by differencing the
incoherent scatter and pair production probabilities from unity. This
implementation characteristic is a holdover from an era when computer
memory was limited and expensive.
10.4.8. Simulation—radiation transport

The Monte Carlo simulation uses the same transport algorithm used by
Snyder et al. [2] in the ALGAM67 code to generate the MIRD11 tables.
The algorithm uses the maximum attenuation coefficient from among all
materials that make up the target geometry as the basis for decreasing a
photon’s ‘probability of existence’. For the family of Reference Man
phantoms, this number is equal to three, representing bone, soft tissue and
lungs. Use of this data reduction technique obviates the need to sample the
photoelectric probability distribution directly.
A starting coordinate is selected from within a source volume, and an
initial direction chosen from the isotropic distribution. Interaction co-
efficients are read in for the energy photon being transported. The well
known expression for the distance to collision site—log(rand_#( )) (1/)—
is used to transport the photon. At this juncture, a game of chance is
played with probability of acceptance equal to media =max . If a random
number is greater than this value, the photon is allowed to continue along
its current trajectory. If the random number is less than or equal to this
value, an interaction is deemed to have taken place, and energy weighted
by the respective interaction coefficients deposited within the voxel contain-
ing the interaction point. The photon’s new direction is determined from
sampling the Klein–Nishina distribution. The photon continues until either
it escapes from the phantom, or its ‘probability of existence’ becomes less
than 0.00005.
If a photon’s energy is determined to be greater than 1.02 MeV after an
interaction has taken place, pair-production is simulated. The state of the
current photon is stored temporarily while a daughter photon is started.
The daughter is assigned an initial energy of 0.511 keV, and a direction
selected from the isotropic distribution. The daughter is assigned a weight
twice that of the parent, so as to represent a pair of photons. The daughter is

followed until it escapes from the target lattice or its weight falls to less than
0.00005. Radiation transport simulation of the parent photon is then
resumed from where it left off [23].
At the end of the simulation, each voxel’s contents reflect the amount of
penetrating energy absorbed for the photon histories simulated. Since
photon energies have been simulated based on their percentage abundance,
the voxel contents are a statistical measure of the voxel’s absorbed fraction
for penetrating radiation. Each voxel in the lattice is divided by the total
energy simulated, multiplied by the penetrating radiation-per-decay, and
multiplied by the source’s cumulated activity. This yields a dose lattice for
penetrating radiation originating from the given source volume.
Simulation of non-penetrating radiation is then carried out for the same
source volume in a second three-dimensional lattice. The current algorithm
for simulating the transport of non-penetrating radiation—local absorp-
tion—follows the same number of histories as for the simulation of penetrat-
ing radiation. If N non-penetrating radiation histories are followed, and
the source volume is composed of n voxels, on average N=n histories will
originate in each voxel. If the value N=n is multiplied by the radionuclide’s
total non-penetrating energy, an estimate of the dose contribution from
non-penetrating radiation is obtained. This value is assigned to each voxel
of the source only.
At the end of the non-penetrating radiation simulation, each voxel’s
contents reflect the amount of non-penetrating energy absorbed for the
non-penetrating histories simulated. As with the penetrating radiation
simulation, each voxel in the lattice is divided by the total energy simulated
and multiplied by the source’s cumulated activity. This yields a dose lattice
for non-penetrating radiation originating from the given source volume.
The non-penetrating radiation dose lattice is added to the penetrating
radiation dose lattice, with the sum stored in an accumulator dose lattice.
The process is repeated for additional sources if present in the source list.
Although the simulation of non-penetrating radiation in this manner may
seem like a great deal of effort that could be bypassed with a few algebraic
simplifications, it prepares MABDOSE for the incorporation of other
transport algorithms that follow individual particles.
10.4.9. Dose display

Average dose is displayed on a per-organelle basis (figure 10.3). Regular
composite organs (e.g., skeleton, total body) are also displayed. The
beauty of having a lattice-based target geometry is that the mapping of
total dose deposited on a per-voxel basis is maintained. It is the raw result
of the simulation. MABDOSE has implemented the display of dose–
volume histograms (DVHs) for highlighted targets. Alternatively, a series
Figure 10.3. Dose display, accompanied by a DVH for the selected liver
volume.
of organs can be highlighted and the DVH for this composite organ list
displayed.
10.5. FUTURE DEVELOPMENTS
There are a number of enhancements that are envisioned for further develop-
ment. Each of these enhancements provides flexibility to the overall program
that increases the generality of the problem specification.
10.5.1. Automated or semi-automated image segmentation

This remains as one of the truly difficult problems left in the field of medical
physics dosimetry. Use of the blood pool as the means for delivery of
radiolabelled compounds has the consequence of exposing all organs in the
body to the administered radiation. Only rarely can this be ignored (e.g.,
instances where extraction from the blood pool is virtually instantaneous).
It follows that all organs would need to be contoured in a general purpose
dosimetry program, since the user could not know beforehand which
organs would prove to be dose-limiting critical structures. However,
Future developments 245
manual contouring is the only method that is currently error free, and this is
time prohibitive for implementation on a routine patient basis. Robust
segmentation algorithms that require minimal operator interaction are
necessary to surmount this hurdle. The algorithm currently being
investigated is a thin-plate spline routine [24–26]. The routine would operate
on a spiral CT data set that brackets a patient volume from head to foot.
Output from the routine would be a three-dimensional data set with all
organs surfaces segmented according to the RTOG format [27].
10.5.2. Generalization of the atomic composition of target voxels

Currently, a voxel is classified as bone, soft tissue or lung. [Note: a voxel can
be marked as ‘Outside the target-of-interest’, so there are actually four
indices for characterizing a voxel.] This precludes other atomic ‘types’,
such as polystyrene, lucite, nylon, acrylic, aluminium, copper and water,
from which test phantoms are frequently made. The proposed addition
would allow the dynamic creation of an atomic type. It would decouple
the meaning of an index from one of only three types, allowing the user to
associate meaning to a voxel in a completely general way. The user would
define the atomic abundance of individual atoms that compose a substance.
This information would construct effective cross section tables from atomic
cross section tables using mass weighted abundance. These tables would be
read in from information provided in the file header regarding atomic
make-up. As with the target lattice geometries, these cross section tables
could be stored for future reference, or generated dynamically each time a
dosimetry simulation was affected.
10.5.3. Incorporation of electron simulation

Although currently not warranted by image resolution in nuclear medicine,
the ability to simulate electron transport is desirable based on arguments
of physics completeness. It is also possible that higher resolution imaging
modalities may become the devices of choice for measuring and gauging
dose effects. PET imaging has the potential to provide the order of magnitude
increase in resolution that is necessary for assigning activity on a per-voxel
basis. It makes logical sense to expand the transport code to include a
Monte Carlo implementation of electron transport. The incorporation of
the ETRAN code of Berger and Seltzer is especially attractive.
10.5.4 Incorporation of nonlinear models

MABDOSE divides the dosimetry calculation into two parts that correspond
to the terms A~ and S in the MIRD formalism. To calculate A~, mathematical
modelling software was written to implement linear models as previously
mentioned. However, the rationale for using radiolabelled antibodies is that

the carrier medium targets unique binding sites at the cellular level. This
paradigm is somewhat limited if the user is restricted to defining and using
linear models, since targeting cell surface-binding sites implies that the
number of sites can be saturated. If this occurs, a twofold increase in anti-
body will not result in a twofold increase in activity, and the mathematics
necessary to describe the time-course of activity concentration requires
nonlinear mathematics.
The application of nonlinear mathematics is closely aligned with the
idea of a completely segmented anatomic data set. From the physiology
literature can be obtained estimates of the number of binding sites per unit
mass for a given organ. This specification will have built into it an inherent
link with the size or mass of an individual patient’s organ. Thus, a nonlinear
model whose rates of incorporation/extraction/excretion are dependent on
the ensemble of mass-dependent terms would allow the specification of
homoeomorphic models that would be true over a much larger range of
antibody amounts and radioactivity doses. Models that are more predictive
over a wider range of conditions result in higher accuracy dose estimates.
10.6. SUMMARY
Prior to MABDOSE, there was no general method to perform dosimetry

resulting from the introduction of an arbitrary source of activity (as
exemplified by a tumour focus). MABDOSE was developed as both an
extension of the MIRD philosophy and as a software program to implement
that philosophy. MABDOSE accounts for all radiations in all organ systems
that localize activity including tumour. It represents a synthesis of interactive
modelling and dosimetry software. MABDOSE, when finally implemented
as envisioned, will be a completely general, all-inclusive internal radionuclide
dosimetry program.
REFERENCES
[1] Loevinger R and Berman M 1968 A schema for absorbed-dose calculations for
biologically distributed radionuclides J. Nucl. Med. 1 (suppl) 7
[2] Loevinger R and Berman M 1976 A revised schema for calculating the absorbed dose
from biologically distributed radionuclides, MIRD Pamphlet No 1, Revised. With a
foreword by R J Cloutier and E M Smith (New York: Society of Nuclear Medicine)
[3] Snyder W S, Ford M R, Warner G G and Watson S B 1975 ‘S’: absorbed dose per unit
cumulated activity for selected radionuclides and organs, MIRD Pamphlet No 11 (New
York: Society of Nuclear Medicine)
References 247
[5] Johnson T K 1986 A tumor preferential dosimetry schema for generalized use in
radionuclide monoclonal antibody therapy Med. Phys. 13 586 (abst)
[6] Johnson T K and Vessella R L 1987 A generalized dosimetry schema for tumor
preferential uptake of monoclonal antibodies in radionuclide immunotherapy J.
Nucl. Med. 28(4) (suppl) 680 (abst)
[7] Johnson T K 1988 MABDOS: a generalized program for internal radionuclide
dosimetry Comput. Meth. Programs Biomed. 27 159–67
[8] Johnson T K and Vessella R L 1989 On the possibility of ‘real-time’ Monte Carlo
calculations for the estimation of absorbed dose in radioimmunotherapy Comput.
[9] Johnson T K and Vessella R L 1989 On the application of parallel processing to the
computation of dose arising from the internal deposition of radionuclides Comput.
Phys. 3 69–72
[10] Johnson T K, McClure D L, McCourt S L, Andl G J, Berman B D and Newman F D
MABDOSE: a computer program for the calculation of dose. In Proceedings of the
6th International Radiopharmaceutical Dosimetry Sysmposium, Gatlinburg, TN, 7–
10 May 1996 (ORISE 99-0164, January 1999) pp 425–39
[11] Johnson T K, McClure D, McCourt S and Newman F 1997 MABDOSE:
Characterization and validation of a general purpose dosimetry code Med. Phys.
24 1025 (abst).
[12] Johnson T K 1997 Implementation of a 3-D internal dosimetry system on a PC Med.
Phys. 24 990–1 (abst)
[13] Johnson T K, McClure D L and McCourt S L 1999 MABDOSE. I. Characterization
of a general purpose dosimetry code Med. Phys. 26(7) 1389–95
[14] Johnson T K, McClure D L and McCourt S L 1999 MABDOSE II: Validation of a
general purpose dosimetry code Med. Phys. 26(7) 1396–403
[15] Zubal I G, Harrell C R, Smith E O, Rattner Z, Gindi G and Hoffer P B 1994
Computerized 3-dimensional segmented human anatomy Med. Phys. 21 299–302
[16] Kocher D C DLC-80/DRALIST data package (RSIC Data Library DLC-80, Oak
Ridge National Laboratory)
from internal photon sources. I. Methods (ORNL/TM-8381/V1 April)
[18] Press W H, Teukolsky S A, Vetterling W T and Flannery B P 1992 Numerical Recipes
in C: the Art of Scientific Computing 2nd edition (Cambridge: Cambridge University
Press) pp 656–706
[19] Roussin R W, Knight J R, Hubbell J H and Howerton R J 1983 Description of the
DLC-99/HUGO package of photon interaction data in ENDF/B-V format ORNL/
RSIC-46 (Oak Ridge National Laboratory)
[20] Hubbell J H 1969 Photon cross sections, attenuation coefficients and energy absorption
coefficients from 10 keV to 100 GeV NSRDS-NBS 29
[21] Seltzer S M 1993 Calculation of photon mass energy-transfer and mass energy-
absorption coefficients Radiat. Res. 136 147–70
Estimates of specific absorbed fractions for photon sources uniformly distributed in vari-
ous organs of a heterogeneous phantom (New York: Society of Nuclear Medicine)
[23] Warner G G and Craig A M Jr 1986 ALGAM: a computer program for estimating
internal dose from gamma-ray sources in a man phantom ORNL/RSIC CCC-152
(Oak Ridge National Laboratory) reviewed November 1986
[24] Wust P, Gellermann J, Beier J, Wegner S, Tilly W, Troger J, Stalling D, Oswald H,

Hege H C, Deuflhard P and Felix R 1998 Evaluation of segmentation algorithms
for generation of patient models in radiofrequency hyperthermia Phys. Med. Biol.
43 3295–307
[25] Boes J L, Weymouth T E, Meyer C R, Quint L E, Bland P H and Bookstein F L 1990
Generating a normalized geometric liver model with warping Radiology 177(P) 134
(abst)
[26] Baldock R A and Hill B 2000 Image warping and spatial data mapping. In Image
Processing and Analysis: a Practical Approach eds R Baldock and J Graham (New
York: Oxford University Press)
[27] RTOG Data Exchange Specification v3.30, available from http://rtog3dqa.wustl.edu/
exchange_files/tapeexch330full.htm
Chapter 11
The three-dimensional internal dosimetry

software package, 3D-ID
George Sgouros and Katherine S Kolbert
11.1. INTRODUCTION
The standard formalism for patient dosimetry was developed by the Medical
Internal Radiation Dose (MIRD) Committee and is described in reference
[1], and also reviewed in chapter 4 of this book. As indicated above, to
simplify absorbed dose calculations, the MIRD Committee developed ‘S
factor’ tables and an associated procedure for their use [2]. This procedure
has been implemented in a software package, MIRDOSE3 [3]. To generate
tables of S factors for different radionuclides and source–target organ combi-
nations, a standard model of human anatomy was adopted in which organ
position, dimensions, and composition were mathematically defined. The
radioactivity was assumed to be uniformly distributed throughout each
source organ and the S factors were defined as the mean absorbed dose to
a target organ per unit cumulated activity in a source organ. Since the posi-
tion and size of tumours may vary within the body and since a standard
model of human anatomy was adopted for generating the S factor tables,
tumours are not included in the published tables. A number of approaches
have been developed for estimating the absorbed dose to tumours and the
dose contribution from tumours to normal organs. The simplest approxima-
tion is made by assuming that all electrons are deposited locally and that the
relative contribution to the tumour-absorbed dose from photons is negligi-
ble. Alternatively, the fraction of electron energy absorbed may be consid-
ered assuming the tumour can be modelled as a sphere [4, 5]. Using tables
of photon-absorbed fraction to spheres or ellipsoids, the photon self-dose
may be added by assuming that the tumour is a sphere or ellipsoid [6]. If
this assumption is made, the photon dose to and from normal organs may
also be calculated by placing the idealized tumour geometry in a defined posi-
tion relative to the standard geometry used for the S factor calculations [7, 8].
249
250 The three-dimensional internal dosimetry software package, 3D-ID
If a point-kernel convolution technique is used in estimating absorbed dose,

the true tumour and normal tissue geometry as well as the activity distribu-
tion may be taken into account to yield a spatial absorbed dose or dose-rate
distribution [9–20]. Tissue composition and density variations are not easily
accounted for using point-kernel techniques. To account for these, Monte
Carlo techniques are needed to estimate absorbed dose [21–32].
The 3D-ID software package takes the distribution of radiolabelled
antibody for a given patient (e.g., from SPECT or PET) and combines it
with anatomical information (e.g., CT or MRI) to yield absorbed dose
estimates that are specific to a particular patient’s biodistribution and
anatomy [9, 11, 21, 32, 33]. This work introduced the concept of dose–
volume histograms for internally administered radionuclides [9, 21, 34].
The software package, 3D-ID, may be used to carry out both Monte Carlo
and point-kernel-based calculations. It has been used to examine the
impact of different radionuclides on the dose distribution, given a fixed
cumulated activity distribution [21]. More recently, it has been used in a
detailed analysis of tumour dose versus response in the treatment of non-
Hodgkin’s lymphoma using 131 I-anti-B1 antibody [35] and also in thyroid
cancer patients using 124 I PET imaging data with CT [36].
11.2. BACKGROUND TO 3D-ID DEVELOPMENT
3D-ID was developed in response to the increasing use of targeted radio-

nuclide therapy. Although point-kernel convolution as well as patient-
specific Monte Carlo calculations had been implemented in external radio-
therapy [37], no package was available to provide the platform and support
structure required to carry out such calculations for patients treated with
targeted radiopharmaceuticals. Likewise, although the fundamental
schema of the MIRD methodology does not preclude patient-specific
dosimetry calculations, translation of this schema, however, had focused
primarily on diagnostic uses of radionuclides in which standardized
anatomy, uniform activity distribution and estimation of mean absorbed
dose were deemed adequate in assessing the hazards involved in the use of
radionuclides [2]. In targeted therapy, administered activities greatly exceed-
ing those used for diagnosis are combined with therapeutic radionuclides and
tumour to non-tumour ratios that could easily lead to normal organ toxicity.
11.3. PROGRAMMING ASPECTS/DEVELOPMENT PHILOSOPHY
Most of the effort in patient-specific three-dimensional dosimetry has been in

the development of algorithms and computational techniques. Application
of such techniques, however, requires a software system that: translates
Monte Carlo-based implementation 251
studies from multiple imaging modalities (CT or MR and PET or SPECT)

into a single data format, provides the tools needed to draw contours
around regions of interest for identifying source and target regions and
outputs the absorbed dose calculations as images, isodose contours or
dose–volume histograms [9]. Each of these steps requires considerable user
interaction and is input/output intensive. 3D-ID was developed in order to
provide a comprehensive and user-friendly framework in which to carry
out the various and necessary steps to calculate absorbed dose rate or total
dose to any user-defined target region.
In order to relate patient anatomy to radionuclide distribution, it is
necessary to register the appropriate functional and anatomic image. The
non-trivial issue of image registration has been kept separate from the 3D-
ID software system in order to keep the focus on the logistics of image
handling, region-of-interest definition and dose calculation.
11.4. MONTE CARLO-BASED IMPLEMENTATION
Monte Carlo-based dosimetry techniques can account for density () and
atomic number (Z) variations that are not easily accounted for using a
point-kernel approach. A Monte Carlo dosimetry algorithm was developed
for 3D-ID implementation that accounts for individual patient details by
using a SPECT or PET radionuclide activity distribution to initialize a
simulation of photons across a CT geometry. For a given radionuclide, initial
photons are generated according to the voxelized activity distribution
provided by SPECT or PET. Photon energies for the given radionuclide
are sampled from a published decay scheme and transport is performed
using the EGS4 transport code [38]. Efficient particle transport within the
discrete geometry was achieved by treating contiguous voxels having similar
Z as a single medium, and correcting for density variations by adjusting the
path length of the particle in proportion to the density, as determined from
CT. Discrimination between bone and water (which was used to represent
tissue) was carried out using a water–bone density threshold (WBT) with
WBT < 1:4 g cmÿ3 considered water or air and WBT 1:4 g cmÿ3 to be
bone. An additional gain in efficiency was obtained by scoring the absorbed
dose along photon, rather than electron, tracks. The method was verified by
creating discrete density and activity arrays for the Standard Phantom
geometry and comparing specific absorbed fraction for various organ and
energy combinations.
It is important to note that Monte Carlo-based dosimetry in which
activity, density and medium distribution are obtained directly from imaging
can lead to artefacts due to problems with the input information. For
example, images of Monte Carlo-derived dose distributions will depict
higher absorbed doses in the lungs than in tumour lesions in the lungs, even
if the tumour uptake is substantially greater than in normal lung parenchyma.

This is because partial volume effects and imperfect scatter correction cause a
‘smearing’ of radioactivity outside the activity-containing tissue. The lower
density of lung relative to the tumour tissue leads to a much greater dose to
lung than to the tumour (i.e., greater energy deposition per unit mass) [21].
11.5. POINT-KERNEL-BASED IMPLEMENTATION
Point-kernels and point-kernel-based implementations of patient-specific

three-dimensional dosimetry are described in detail in chapter 7 of this
book. In this section the implementation of the point-kernel methodology
as used in 3D-ID will be reviewed.
Although 3D-ID requires a specific format for point-kernel input, the
origin of the point-kernels is left to the software user. The point-kernels
that have been previously used for 3D-ID were generated by Monte Carlo
simulation of photon spectrum transport through water [10]. The point-
kernel calculation provides the user with the option of invoking the
assumption, a priori, that all electron energy will be deposited within each
activity-containing voxel. Since the voxel dimensions used for SPECT or
PET imaging (of humans) are greater than the path length of the electron
or emissions currently used for radionuclide therapy, this assumption is
usually invoked for human dosimetry studies. When 3D-ID is used for
animal or autoradiography studies, the point-kernel scale must be matched
accordingly, and the kernel values at smaller dimensions are generally
dominated by electron emissions.
11.6. INDIVIDUAL MODULES
The 3D-ID software package consists of a set of three modules sharing a

common interface style [9] (figure 11.1). The first step provides an entry
point into the software package for the diverse modality- and vendor-
dependent image formats. Images in standard DICOM or Interfile3.3 can
be read in directly while those in non-standard format can also be read
providing the user is aware of details regarding file size, two-dimensional
image matrix size and voxel bit size, image offset, etc. It is important that
voxel dimensions are correct so that accurate calculations based on the
image elements can be made. All three-dimensional images sets that are to
be subsequently used by 3D-ID are saved in a platform- and system-
independent format with minimal header information defining modality,
imaging isotope, voxel dimensions and patient name and id.
The second step provides tools to define regions of interest and to
calculate source to target dose. Aligned multi-modality image sets are
Individual modules 253
Figure 11.1. 3D-ID system flowchart.
arranged in two rows corresponding to an anatomical modality (CT/MRI)

and to a radionuclide modality (SPECT/PET) (figure 11.2). Images are
displayed separately rather than fused or overlaid so that all available data
from both modalities are preserved.
Contour drawing for region-of-interest (ROI) definition is carried out
on a side-by-side, dual modality display of individual two-dimensional
image slices, thus providing access to information from both imaging
modalities (figure 11.3). Prior to drawing, the cursor appears over both
images so that the user can easily identify corresponding points within the
two images. Contours are drawn using either image and it is possible to
switch back and forth between images during drawing. A number of drawing
tools are provided: region growing which automatically generates a contour
plot by using a percentage of the ratio between a user selected image value
interior to the region and a selected background value outside the region
[39]; edge following which is effective in images of high contrast; and
standard copy and paste of individual ROI contours. ROIs can be quantified
Figure 11.2. Multi-image display from 3D-ID showing four transverse slices
from a set of CT-SPECT aligned images. The top row shows the CT with
ROIs drawn over three lesions identified as: right external iliac in blue, left
external iliac in green, and right common iliac in light blue. The bottom row
shows the corresponding 131 I SPECT slices with the lesions indicated. (See
plate 3 for colour version.)
in terms of total volume, pixel count, and magnitude and location of

maximum value within the ROI.
Once ROI volumes have been defined, they may be identified as source
or target for a particular dose calculation. A point-kernel file must also be
specified to complete input for the dose calculation.
The dose distribution for selected source and target combinations is
calculated by convolving the point-kernel with the activity image. Convolu-
tion is implemented either by fast Fourier transform or by a table lookup
procedure. In either case, the electron dose is deposited within the activity
containing voxel for most human dosimetry calculations. If the electron
range is significant relative to patient organ dimensions, an electron point-
kernel may be added to the early portion of the photon point-kernel to
account for the spatial distribution of the electron dose deposition.
Depending upon the size of the image files, the dose calculation can be a
computationally intensive and time-consuming calculation. In such circum-
stances, batch mode execution of the dose calculation is possible. If multiple
processors or multiple workstations are available, the dose calculation can be
divided into a number of individual parallel batch jobs; the user may select
Individual modules 255
Figure 11.3. Regions of interest are defined by drawing individual contours

using aligned image sets. As contours are drawn, they appear in the multi-
slice, two-row, main display window (seen here behind the drawing window).
the operating system for the dose calculation (currently UNIX or PC) and
the resulting batch file will be appropriately defined. An estimated number
of calculations as well as sample timings are provided as a guideline for
the user.
The dose calculation yields a three-dimensional array of dose values,
which are essentially images of dose distribution and can be viewed as
‘maps’ of the spatially varying dose in a target volume from a particular
source volume or volumes. The modular characteristic of 3D-ID provides
the flexibility to replace the current dose calculation technique with other
techniques that may be either point-kernel or Monte Carlo based.
Once the calculation has been performed, the final step in 3D-ID is the
display and analysis of the three-dimensional dose data set. This includes:
calculation of the mean absorbed dose to a target per unit cumulated activity
from a source (or sources); generation of dose–volume histograms of source
to target regions in order to summarize the dose information in a clinically
interpretable manner; and presentation of all image data in a summary
display of anatomic, functional and dose distribution images.
If absorbed dose instead of dose-rate is desired and time-sequential
SPECT or PET studies are unavailable, the absorbed dose in each target
Figure 11.4. Sample integral dose–volume histogram showing distribution of

dose over four lesions. Histograms were derived from 124 I PET studies in
patients with thyroid carcinoma [36]. Each curve represents an individual
lesion seen in a single patient.
Figure 11.5. The spatial absorbed dose distribution is depicted so that image
intensity corresponds to absorbed dose values. Isodose contour levels for 10%
(green), 25% (blue), 50% (red) and 75% (yellow) of the maximum dose
to the tumour mass are shown. The image was obtained from 3D-ID analysis
of 124 I PET studies in patients with thyroid carcinoma [36]. (See plate 4 for
colour version.)
Considerations in clinical implementations 257
voxel may be scaled by a total source organ cumulated activity that is

typically obtained from planar imaging kinetics. In this way, the spatial
distribution of activity is assumed to be preserved throughout the study
period. This is a first-order approximation, which may be improved upon
if additional (three-dimensional) pharmacokinetic data are obtained. Cumu-
lated activity estimation is external to the system, as this will depend upon the
details of patient imaging and pharmacokinetic data acquisition.
The mean dose over a target volume is determined by taking the mean of
all values in the dose array. Either integral or differential dose–volume
histograms can be generated. In the former, the percentage of volume receiving
less than or equal to a particular dose value is plotted as a function of the dose
value (figure 11.4). In the latter, the percentage of volume receiving a particular
absorbed dose range is plotted versus the dose range.
Multiple iso-dose levels can be displayed in the multi-modality image
display by selecting either a specific percentage of the maximum dose or
the actual dose value (figure 11.5).
11.7. CONSIDERATIONS IN CLINICAL IMPLEMENTATIONS
The logistics associated with performing fully patient-specific three-

dimensional dosimetry are considerable. To define the anatomy and also
to provide tissue density information a CT scan over the region of interest
is required. To provide the activity distribution a PET or SPECT scan is
required. If absorbed dose rather than absorbed dose-rate is desired, then
kinetic information is required. In conventional dosimetry, kinetic informa-
tion that is averaged over the whole source organ volume suffices; in three-
dimensional dosimetry variations in pharmacokinetics within an organ
must be taken into account. Ideally kinetic information on a voxel-by-
voxel basis should be used. This would require multiple SPECT or PET
studies taken over time, each registered to each other, and also registered
to the CT study. Assuming such data are available, a voxel-by-voxel-based
integration of the activity would have to be performed to yield a three-
dimensional representation of residence time or cumulated activity. To
achieve this type of comprehensive three-dimensional dosimetry calculation
a number of logistical, technical and fundamental hurdles must be overcome.
The logistical difficulties arise because it may not be feasible to perform PET
or SPECT imaging of a particular patient multiple times. The time required
for such imaging is considerably greater than planar imaging. This may
lead to difficulties with patient compliance and camera availability. The
technical difficulties include registration of PET or SPECT images to each
other and to a CT study. Although this step is independent of the dosimetry
algorithm itself, the fidelity of the registration impacts on the quality of the
input which, in turn, will determine the reliability of the dose estimates.
Registration accuracy will depend upon the anatomical ROI, the agent being
imaged and the technique that is used for registration. Since it is likely
that the total number of counts associated with any given voxel will be low,
yielding a high standard deviation, the voxel-by-voxel-based integration
outlined above may be fundamentally limited by the error associated with
each voxel value. Integration on a voxel-by-voxel basis may yield unreliable
values for residence time or cumulated activity due to these errors. A
number of techniques and approximations may be adopted to overcome
some of these difficulties. It is possible to derive a residence time or cumulated
activity image, for example, by using planar imaging to obtain kinetics and
SPECT or PET imaging to obtain the spatial distribution. A residence time
or cumulated activity image may be obtained by assuming that all points
making up the spatial distribution of activity follow the same kinetics [33].
11.8. FUTURE POTENTIAL USES
3D-ID has proved to be a particularly versatile software package with features

that extend beyond patient-specific dosimetry. Given appropriate information
regarding anatomy and spatial distribution, it is possible to carry out absorbed
dose calculations for a number of different circumstances. For example, by
using radionuclide kernels with a greater spatial resolution such that dose
contributions are dominated by electron deposition, it is possible to carry
out dosimetry calculations for experimental animals. The advent and increas-
ing interest in small animal imaging makes this application particularly
compelling. Using 3D-ID with high-resolution MR images and assuming a
uniform cumulated activity distribution in individual organs of a mouse, we
have recently generated murine S factors [40]. A similar effort could be under-
taken using autoradiography and histology slides. In short, because 3D-ID
allows input of images and point-kernels from a wide variety of sources,
almost any spatial distribution of radionuclide can be used to generate the
corresponding absorbed dose distribution.
ACKNOWLEDGMENTS
This work was supported by NIH grants R01 CA62444, U01 CA58260 and
P01 CA33049 and DOE grant DE-FG02-86ER-60407.
REFERENCES
Calculations revised edition (New York: Society of Nuclear Medicine)
References 259
‘S’ absorbed dose per unit cumulated activity for selected radionuclides and organs
[6] Ellett W H and Humes R M 1971 MIRD Pamphlet No 8: Absorbed fractions for
small volumes containing photon-emitting radioactivity J. Nucl. Med. 12(5)
(suppl) 25–32
[7] Johnson T K, McClure D and McCourt S 1999 MABDOSE I: Characterization of a
general purpose dose estimation code Med. Phys. 26 1389–95
[8] Johnson T K and Colby S B 1993 Photon contribution to tumor dose from
considerations of 131 I radiolabeled antibody uptake in liver, spleen and whole
body Med. Phys. 20 1667–74
[9] Kolbert K S, Sgouros G, Scott A M et al. 1997 Implementation and evaluation of
patient-specific three-dimensional internal dosimetry J. Nucl. Med. 38 301–8
[11] Sgouros G, Barest G, Thekkumthala J et al. 1990 Treatment planning for internal
radionuclide therapy: three-dimensional dosimetry for nonuniformly distributed
radionuclides J. Nucl. Med. 31 1884–91
[12] Liu A, Williams L E, Wong J Y and Raubitschek A A 1998 Monte Carlo-assisted
voxel source kernel method (MAVSK) for internal beta dosimetry Nucl. Med.
Biol. 25 423–33
[13] Flux G D, Webb S, Ott R J, Chittenden S J and Thomas R 1997 Three-dimensional
dosimetry for intralesional radionuclide therapy using mathematical modeling and
multimodality imaging J. Nucl. Med. 38 1059–66
[14] Giap H B, Macey D J and Podoloff D A 1995 Development of a SPECT-based three-
dimensional treatment planning system for radioimmunotherapy J. Nucl. Med. 36
1885–94
kernel convolution technique for internal dosimetry Phys. Med. Biol. 40 365–81
[17] van Dieren E B, Plaizier M A, van Lingen A, Roos J C, Barendsen G W and Teule G J
1996 Absorbed dose distribution of the Auger emitters 67Ga and 125I and the beta-
emitters 67Cu, 90Y, 131I and 186Re as a function of tumor size, uptake and intracellular
distribution Int. J. Radiat. Oncol. Biol. Phys. 36 197–204
[18] van Dieren E B, van Lingen A, Roos J C, Huijgens P C, Barendsen G W and Teule G J
1994 A dosimetric model for intrathecal treatment with 131I and 67Ga Int. J. Radiat.
Oncol. Biol. Phys. 30 447–54
[19] Erdi A K, Yorke E D, Loew M H, Erdi Y E, Sarfaraz M and Wessels B W 1998 Use
of the fast Hartley transform for three-dimensional dose calculation in radionuclide
therapy Med. Phys. 25 2226–33
[20] Liu A, Williams L E, Lopatin G, Yamauchi D M, Wong J Y and Raubitschek A A

1999 A radionuclide therapy treatment planning and dose estimation system J. Nucl.
Med. 40 1151–3
[21] Furhang E E, Chui C S, Kolbert K S, Larson S M and Sgouros G 1997 Implementa-
tion of a Monte Carlo dosimetry method for patient-specific internal emitter therapy
Med. Phys. 24 1163–72
[22] Bouchet L G, Jokisch D W and Bolch W E 1999 A three-dimensional transport
model for determining absorbed fractions of energy for electrons within trabecular
bone J. Nucl. Med. 40 1947–66
Monte Carlo code-based software for internal radionuclide dosimetry J. Nucl. Med.
40 1517–23
[24] Bolch W E, Bouchet L G, Robertson J S et al. 1999 MIRD Pamphlet No 17:
The dosimetry of nonuniform activity distributions—radionuclide S values at
the voxel level. Medical Internal Radiation Dose Committee J. Nucl. Med. 40
11S–36S
[25] Bardies M and Myers M J 1996 Computational methods in radionuclide dosimetry
[27] Leichner P K 1994 A unified approach to photon and beta particle dosimetry J.
Nucl. Med. 35 1721–9
[28] Kwok C S, Bialobzyski P J, Yu S K and Prestwich W V 1990 Effect of tissue
inhomogeneity on dose distribution of point sources of low-energy electrons Med.
Phys. 17 786–93
[29] Johnson T K and Vessella R L 1989 On the possibility of ‘real-time’ Monte Carlo
calculations for the estimation of absorbed dose in radioimmunotherapy Comput.
[30] Johnson T K 1988 MABDOS: a generalized program for internal radionuclide
dosimetry Comput Meth. Programs Biomed. 27 159–67
[31] Bardies M and Myers M J 1990 A simplified approach to alpha dosimetry for small
spheres labelled on the surface Phys. Med. Biol. 35 1551–61
specific dosimetry Med. Phys. 23 1523–9
[33] Sgouros G, Chiu S, Pentlow K S et al. 1993 Three-dimensional dosimetry for
radioimmunotherapy treatment planning J. Nucl. Med. 34 1595–601
[34] Kolbert K S, Sgouros G, Scott A M et al. 1994 Dose–volume histogram representa-
tion of patient dose distribution in 3-dimensional internal dosimetry J. Nucl. Med. 35
P123–P124
[35] Sgouros G, Squeri S, Kolbert K S et al. 1998 Tumor response following 131I-anti-B1
antibody therapy of non-Hodgkin’s lymphoma: patient-specific tumor dosimetry
and the time-course of tumor volume changes Cancer Biotherm. Radiopharm. 13
303
[36] Mun E F, Kolbert K S, Sheikh A et al. 2002 Patient-specific I-124 PET-based 3-D
dosimetry for I-131 thyroid cancer therapy J. Nucl. Med. 43 86P
[37] Mohan R, Barest G, Brewster L J et al. 1988 A comprehensive three-dimensional
radiation treatment planning system Int. J. Radiat. Oncol. Biol. Phys. 15 481–95
References 261
[38] Nelson W R, Hirayama H and Rogers D W O 1985 The EGS4 code-system SLAC
Report No 265 (Stanford, CA: SLAC)
[39] Erdi Y E, Mawlawi O, Larson S M et al. 1997 Segmentation of lung lesion volume by
adaptive positron emission tomography image thresholding Cancer 80 2505–9
[40] Kolbert K S, Watson T, Matei C, Xu S, Koutcher J A and Sgouros G 2002 Murine-
specific S-factors J. Nucl. Med. 43 88P
Chapter 12
Evaluation and validation of dose

calculation procedures in patient-specific
radionuclide therapy
Michael Ljungberg, Yuni Dewaraja, George Sgouros
and Sven-Erik Strand
12.1. INTRODUCTION
The traditional method of absorbed dose calculation has been based on a

simplified MIRD formalism with S values, which have been calculated from
a generalized geometry. Until recently, the most common approach was to
calculate the absorbed dose averaged over a tumour or organ volume.
Estimates of the radioactivity in these volumes could be obtained from two-
dimensional whole-body scintillation camera imaging or from SPECT or
PET imaging. The two latter technologies allow for three-dimensional
information if proper corrections for various physical and camera-specific
effects, such as photon attenuation, contribution from scatter in the object
and crystal, septum penetration and its degradation in spatial resolution, are
applied. Three-dimensional techniques also have the potential to account for
heterogeneous uptakes and thus enable absorbed dose determination down
to the voxel level. If a three-dimensional attenuation distribution of the patient
that is aligned to the SPECT data is available, these two information sets
(physiology and morphology) can be combined in a calculation scheme to
produce three-dimensional absorbed dose distributions. This is an advantage
compared with mean–average dose calculation and should therefore be the
main future direction for dosimetry development in radionuclide therapy.
12.2. DOSE–VOLUME INFORMATION VERSUS MEAN–

AVERAGE ORGAN/TUMOUR DOSE
The ability of three-dimensional dosimetry to provide absorbed dose
262
Monte Carlo simulation of imaging systems 263
Figure 12.1. An example of an integral dose–volume histogram. The curves

depict the percentage of the bone marrow (BM) and the spleen (S) target
volumes that receives an absorbed dose rate less than or equal to the dose on
the x axis. The dotted lines denote the mean absorbed dose rates to each
organ. Redrawn with permission from [1].
information on a voxel-by-voxel basis means that the absorbed dose is

averaged over a much smaller volume compared with the earlier standard
geometry S factor absorbed-based dose calculations. Regional variation in
absorbed doses can, therefore, be monitored. Three-dimensional informa-
tion, combined with accurate morphological information, provides better
estimates of the absorbed doses since patient-specific geometry and attenua-
tion can be included in the calculation. This puts, however, more demands
on the registration procedure between anatomical and physiological
images. The absorbed dose distribution within a region can also be presented
as dose–volume histograms, i.e., the probability of voxels having a par-
ticular absorbed dose. These histograms provide useful information of
how the absorbed dose is locally distributed. Further processing of this
information will be linked to histograms of proper dose–response functions
for regional treatment but also be able to combine with, for example, external
beam therapy. Figure 12.1 shows an example of an integral dose–volume
histogram.
12.3. MONTE CARLO SIMULATION OF IMAGING SYSTEMS
Since SPECT and planar imaging provide the underlying information for
activity quantification and biokinetics, evaluation and validation of the dosi-
metry should also consider reconstruction methods, registration methods of
CT and SPECT/PET data, scatter and attenuation compensation methods
264 Dose calculation procedures in patient-specific radionuclide therapy
and distance-dependent collimator response corrections. In addition, partial

volume effects and camera limitations, such as count rate and pile-up effects,
need to be evaluated. The latter is important for intra-therapy imaging of
patients undergoing treatments with high activities. The validation can be
performed by physical measurements using phantoms, from computer-
generated data or a combination of both.
12.3.1. Monte Carlo programs for radionuclide imaging

Several Monte Carlo programs, designed to simulate scintillation camera
imaging for both SPECT and PET, have been described in the literature
[2]. The main use of these programs has been to provide a comparison
between ideal images and images improved by correction for, e.g., photon
attenuation and scatter, on a pixel-by-pixel basis. The design of the programs
varies by its complexity and aimed application area.
The SIMIND program, developed at the Lund University, is a Fortran-
90 code that runs on the major platforms. Included in the program is simula-
tion of SPECT and planar imaging using arbitrary voxel-based phantoms. In
its native form, the program can also simulate transmission SPECT imaging.
The user can modify the program by linking user-written scoring; source or
isotope routines where the two latter routines provide spatial coordinates for
the decay and photon energies. These routines access a global common block
and the user-written code that can be linked to the main code without
interference. The code can be coupled with the collimator routine of deVries
et al. [3] where photon transport is modelled realistically including scatter
and penetration events. Recently, the code has also been parallelized using
the MPI system for use on large-scale parallel computers.
The SimSET program is a package, developed at the University of
Washington, which simulates both SPECT and PET imaging. The software
is written in a modular format in C. The main module is the PHG module,
which models photon transport through heterogeneous media for both
SPECT and PET. A collimator module tracks the photons further through
a collimator. Finally, a detector module receives the photons, tracks the
photons through the detector and scores the interactions within the detector.
For accurate modelling of scintillation camera imaging with higher
energy photon emitters it is essential to include complete photon transport
in the collimator. These photons have a high probability of penetrating the
collimator septa or undergoing scatter in the collimator. Only a few Monte
Carlo algorithms have been used in nuclear medicine to model the collimator
in its physical entirety because of the demanding computational require-
ments [3–6]. The MCNP program [6] has been shown to be useful for
investigation of collimator characteristics due to its combinatorial geometry
package that makes it easy to build regular patterns such as hex-hole shaped
collimators.
Monte Carlo simulation of imaging systems 265
12.3.2. Software phantoms

Anthropomorphic software phantoms with internally segmented structures
make clinically realistic Monte Carlo simulations possible. Both voxel-
based and analytic phantoms are used. The voxel-based phantoms are
mostly derived from segmented tomographic images of the human anatomy
obtained by either X-ray CT or MRI. The analytic phantoms consist of
regularly shaped continuous objects defined by combinations of simple
mathematical geometries. An advantage of analytical phantoms over
voxelized phantoms is that changes of the geometry can more easily be
carried out to allow for anatomical variability. Below are examples of
three software torso phantoms useful for nuclear medicine imaging simula-
tions and RNT dosimetry evaluation.
12.3.2.1. The Zubal phantom

This is a computerized anatomically correct three-dimensional volume array,
which represents an adult male. The phantom was created using manual
segmentation of X-ray CT slices of a living male human [7]. All major
internal structures have been segmented and pixels inside each structure
have been given a unique code. These codes are then used to assign activity
or density values to each structure. The phantom data are available as a
128 128 246 matrix with a cubic resolution of 4 mm. A further develop-
ment of the phantom has been made by Dr Stuchly, University of Victoria,
who copied the arms and legs from the Visible Human (VH) and attached
them to the original torso phantom [8]. Because the arms of the VH cadaver
were positioned over the abdominal part, Dr Sjögreen at Lund University
mathematically straightened the arms out along the phantom’s side to
make the phantom useful for whole-body imaging simulations [9]. Figure
12.2 shows rendered images of all three versions. Zubal and colleagues
have also developed a high-resolution brain phantom, based on an MRI
scan, which can be used for detailed investigations in the head.
12.3.2.2. The MCAT phantom

The MCAT phantom has been developed mainly for cardiac studies [10]. The
program creates voxel images of the heart and some of the major organs
(liver, kidney, lungs, ribs, stomach and spleen) from mathematical descrip-
tions. The main advantage is the ability to modify the shape, size and
location of the phantom and organs including motion of the heart.
12.3.2.3. The NCAT phantom

This phantom is an extension on the MCAT but utilizes a technology called
non-uniform rational B-splines (NURBS) when creating the shapes. NURBS
define continuous surfaces and allow the phantom to be defined at any
Figure 12.2. The left image shows the rendered image of the original Zubal
phantom. The middle image shows the extensions of arms and legs that Stuchly
et al made using the VH data set. The right image shows the straightening of the
arms that makes the phantom useful for patient-like whole-body simulations in
RNT.
spatial resolution [11, 12]. By fitting NURBS to patient data, the phantom
generated is more realistic than methods based on solid geometry definitions.
The NCAT phantom also includes temporal functions, such as modelling of
the heart motion and the respiratory movements. All organ and skeletal
models (except for the heart model) are based on CT information of the
VH male data set. The heart originates from a gated MRI scan of a human.
12.4. EVALUATION OF IMAGE QUALITY DEGRADATION BY

MONTE CARLO METHODS
The accuracy of the activity quantification is an essential part of the absorbed

dose calculation. SPECT image quality and quantification accuracy are
degraded by scatter, penetration, attenuation and finite spatial resolution.
Methods that compensate for these effects have been described in the litera-
ture, but most have been developed for 99m Tc (photon energy ¼ 140 keV)
since this radionuclide is the most widely used in diagnostic nuclear methods.
Radionuclides that have been widely used in therapy and diagnosis prior to
therapy are, for example, 131 I and 111 In. Some of these radionuclides emit
several photons in the decay. These additional photons may degrade the
accuracy in the quantification of the principal energy depending on their
energy and abundance. Consider 131 I as an example of a useful radionuclide
Evaluation of image quality degradation by Monte Carlo methods 267
that has multiple photon emission. The four highest-intensity emissions of

131
I are: 284 keV (6.1%), 364 keV (82%), 637 keV (7.2%) and 723 keV
(1.8%) [11]. Although the 637 and 723 keV photons are low in intensity,
they contribute significantly to the image since higher-energy photons have
a relatively low attenuation in the patient and a higher probability of
penetrating the collimator (which is optimized for 364 keV). The 637 and
723 keV photons that have lost part of their energy due to Compton scatter
can be included in the 364 keV photopeak window events. A comparison
between energy pulse-height distributions from a ‘simple’ decay, such as
99m
Tc, and the more complicated 131 I that involves multiple photon emission
is shown in Figure 12.3. Here the un-scattered primary component has been
separated.
Figure 12.4 shows image quality degradation by comparing an ideal 131 I
WB study without any physical or camera-based limitation (A) and a realistic
WB study with all degradation effects (B). The differences in image quality
between A and B mainly consist of degraded spatial resolution and contrast
resolution, due to limited system resolution, scatter and septum penetration.
The difference is striking and shows how complicated it will be to properly
obtain a correct activity measurement from scintillation camera imaging.
131
12.4.1. Evaluation of planar I imaging
A fundamental imaging parameter to study is the spatial resolution where the
point-spread function (PSF) describes the imaging system’s response to an
infinitely small point-source (a function). Figure 12.5 shows a comparison
between measured and simulated energy spectra and PSFs for a 131 I point
source in air for a high-energy collimator (HEGAP) and a ultra-high-
energy collimator (UHEGAP). The reason for using two collimators was
to investigate how well the UHEGAP collimator, with its thicker lead
septa, reduces septum penetration. The good agreement between measured
and simulated energy spectra was achieved only after inclusion of backscatter
photons in the simulation model [13, 14]. Validation of the energy spectrum
was also carried out for a case where there was significant object scatter using
a phantom consisting of a hot sphere centred in a cylindrical water tank [14].
The wide tails of the 131 I PSF (see figure 12.5) are due to septum penetration,
which degrades focal quantification accuracy because of the spread of
regional counts to surrounding areas. Correcting for this is especially
important in 131 I RNT where antibody uptake by surrounding organs is
significant compared with tumour uptake.
Monte Carlo is ideal for assessing penetration and scatter since, unlike
in experimental measurement, these events can be tracked separately in a
simulation. For 131 I, a quantitative comparison of scatter and penetration
has been performed for the HEGAP and the UHEGAP collimators [13].
The geometric, scatter and penetration components of 364 keV photo-peak
Figure 12.3. Energy spectral distributions detected in the scintillation camera,

for (A) 99m Tc and (B) 131 I. Solid lines show the total detected spectra and
dashed lines show the primary spectra due to photons that have not been
scattered in the object, the camera head, or penetrated through the collimator
septa. The spectra were obtained by Monte Carlo simulation of a point
source, located centrally in an elliptical water-phantom, at a depth of 10 cm.
events were separated for a planar acquisition of a point source in air.

‘Geometric’ implies traversing the collimator hole without interaction;
‘scatter’ implies Compton scattering in the collimator lead while ‘penetra-
tion’ implies penetration of one or more septa without scatter. The results
are shown in table 12.1. The UHEGAP collimator significantly reduces the
scatter and penetration components compared with the HEGAP collimator.
This was achieved at the expense of some loss in resolution and sensitivity.
Figure 12.4. Simulated whole-body images of 131 I with the activity distribution
corresponding to a 131 I labelled monoclonal antibody distribution and four
simulated tumours. Image A represents the true distribution, obtained by analy-
tical integration. Image B represents a measured image, obtained by Monte
Carlo simulation and detector parameters representing a typical scintillation
camera system. The intensity distribution is altered, i.e., the contrast resolution
is heterogeneous, due to the variable attenuation and scattering properties of the
body.
The suppression of collimator scatter and penetration with the use of the
UHEGAP collimator is also evident in the comparison of 131 I energy spectra
and PSFs for the two collimators (figure 12.5).
131
12.4.2. Evaluation of SPECT I imaging
Validation of 131 I SPECT simulations has been carried out for a clinically
realistic situation using a computerized version of the experimental Radio-
logical Support Devices (RSD) thorax phantom. The physical phantom
was altered by inserting a water-filled plastic sphere, representing a
tumour, close to the liver. The organ and tumour relative activity concentra-
tion ratios were realistic values based on 131 I RIT patient images. Because of
the demanding computational requirements, the simulation was carried out
using the parallel implementation of SIMIND on an IBM-SP2 distributed
memory architecture [15]. There was good agreement between the measured
and simulated images, as demonstrated in figure 12.6.
Tumour activity quantification, such as the one simulated in figure 12.6,
is affected by the spread or blurring of regional counts to surrounding areas
due to the finite spatial resolution of the system. The ‘spill-out’ of counts
270
Figure 12.5. Comparison of the measured and simulated 131 I energy spectra and PSFs for Picker’s collimators HEGAP and UHEGAP.
(Reprinted by permission of the Society of Nuclear Medicine from [13].)
Dose calculation procedures in patient-specific radionuclide therapy
Table 12.1. Monte Carlo comparison of geometric, penetration and scatter

component of events within the photo-peak window for the point source in air.
Collimator Geometric (%) Penetration (%) Scatter (%)
HEGAP 27.3 43.3 29.4

UHEGAP 72.4 17.3 10.3
from the target to the background decreases the tumour activity quantifica-
tion, as the source structure gets smaller while the ‘spill-in’ contribution from
the background to the target increases the tumour activity quantification.
The spill-out depends on the shape and size of the source structure and the
system resolution while spill-in depends on these parameters as well as the
activity level and distribution in the background. An experimental evaluation
of the effect of different tumour sizes and shapes on activity quantification
can be technically difficult to achieve, while Monte Carlo simulation is a
powerful tool to carry out such a study. The effects of object shape, size
and background on SPECT 131 I activity quantification without detector
response compensation has been reported in a recent Monte Carlo study
[16]. The results of the study are summarized in table 12.2. Here b is the
background to tumour activity concentration ratio. Note that the activity
error is 0% for one of the simulations because the calibration factor was
also determined from this simulation. The activity quantification was carried
out using a constant calibration factor and no partial volume correction. The
physical object boundary was used to define the volume of interest (VOI),
which is consistent with patient quantification procedures where CT-defined
tumour boundaries are superimposed on to the registered SPECT image.
Results of table 12.2 show that the activity error increases significantly
with decreasing tumour size because of the increased spill-out of counts.
The activity error for cylindrical tumours is consistently higher than the
error for the spherical tumours because spill-out is more significant for
non-spherical objects. Also, activity error depends significantly on b because
of the effect of spill-in.
Figure 12.6. Images and profiles corresponding to a slice of the heart/thorax

phantom are shown. The true activity distribution as well as the measured
and simulated reconstructed images are shown. (Reprinted by permission of
Elsevier from [15].)
Table 12.2. The activity error presented as the percentage difference between
the true 131 I activity and the SPECT estimated activity for different sizes,
shapes and b values.
Tumour volume (cm3 ) Tumour shape b Activity error (%)
50 Sphere 1/5 12.7

50 Sphere 1/3 6.7
50 Sphere 0 23.0
100 Sphere 1/5 9.8
100 Sphere 1/3 6.6
100 Sphere 0 15.2
200 Sphere 1/5 0
200 Sphere 1/3 þ0.5
200 Sphere 0 1.3
50 Cylinder 1/5 44.0
50 Cylinder 0 79.0
100 Cylinder 1/5 32.0
100 Cylinder 1/3 22.0
100 Cylinder 0 50.0
200 Cylinder 0 18.0
12.4.3. Other radionuclides

In RNT with 90 Y-labelled monoclonal antibodies, 111 In is often used as an
analogue tracer because of the absence of -ray emissions associated with
90
Y. Some of the same problems associated with 131 I imaging apply to
111
In, which has -ray emissions at 172 keV (91%) and 245 keV (94%). In
a Monte Carlo study, the fraction of 111 In photons, which penetrate or
scatter in the collimator, was determined for different values of collimator
lead content [17]. For the 172 keV window the penetration component was
up to 15% while the scatter component was up to 7%. For the 245 keV
window the penetration component was up to 58% while the scatter
component was up to 23%.
Monte Carlo simulation using a modified MCNP code has been used to
characterize scatter and penetration of 123 I, which has several high-energy
emissions above the main photo-peak at 159 keV [4]. This work showed
that for sources in air and for the 159 kV window, the fraction of counts
due to high-energy photons was 10–20% for a medium-energy collimator
and 30–50% for a low-energy collimator. For the medium-energy collimator
the reported scatter :primary and penetration :primary ratios were 8–12%
and 12–18%, respectively while these ratios for the low-energy collimator
were 20–50% and 65–120%, respectively.
Evaluation of absorbed dose calculation procedures 273
The multiple -rays associated with 67 Ga range in energy from 91 to

1856 keV. The challenging problems of scatter, penetration and lead X-
rays in 67 Ga SPECT has been investigated in a Monte Carlo study [18].
The study shows that as photo-peak energy increases the fraction of detected
photons classified as geometric decreases rapidly while the collimator scatter
and penetration fractions increase. The reported spill-down of higher
energies to lower-energy windows ranged from 7% to 20%.
12.5. EVALUATION OF ABSORBED DOSE CALCULATION

PROCEDURES BY MONTE CARLO METHODS
The Monte Carlo method has been shown to be very efficient in both
calculating the absorbed dose and also as a tool for evaluating various
parts of the calculation schemes [19]. The key point is that if a simulation
of SPECT images (including all inherent limitations such as photon attenua-
tion, scatter contribution, collimator and septum penetration effects) can
accurately be made, then the output from a proposed calculation procedure
can be compared directly with the corresponding dose calculation using the
known activity and attenuation distribution on which the imaging simula-
tions were based.
Figure 12.7 shows the basic steps in such an evaluation. An advantage
of Monte Carlo methodology is that simulations can be made of realistic
distributions of, in principle, any radiopharmaceutical distribution and
that complex density distributions can be included.
Figure 12.7. A flow-chart describing different steps for which a quantitative

SPECT dose planning software could be validated.
Figure 12.8. Example of how the Monte Carlo method can help evaluate the
accuracy in a three-dimensional absorbed dose calculation program. The
figure shows a transversal section through the lungs (top row) and a section
through the kidneys (bottom row). From the definition of the true activity
distribution (A) ideal images without scatter and attenuation effects can be
simulated (B) and used as reference with which the particular scatter and
attenuation method should be compared (C). Note (a) the higher absorbed
dose shown as a border around the lungs which originated from in-spill counts
from the nearby located tissue and (b) the hot-spot that is a result of an air-
cavity that result in a very local high absorbed dose due to the low density of
the air.
The Monte Carlo method offers a key advantage in SPECT quantifica-

tion of, e.g., 131 I compared with experimental measurements since results of
correction methods can be compared with the ideal. These different results
can then be input to a three-dimensional absorbed dose calculation program
(point-dose kernel program on full photon/electron program based on MC
methodology). The final absorbed dose images as a function of different
correction methods can then be compared with ‘gold standards’ (figure
12.7). Figure 12.8 shows such a comparison of simulated 131 I distributions
in the Zubal phantom [19].
Verification of the dosimetry program has been described extensively
elsewhere in this book. We here only draw attenuation to a few references
that have been based on Monte Carlo simulations. Furhang et al. [20] have
described a streamlined approach for carrying out Monte Carlo calculations
in a voxelized geometry. A detailed validation and comparison of the Monte
Carlo method versus MIRD standard geometry S factors was provided. The
study also examined the impact of non-homogeneous density distribution
and variation in the atomic number. The group has also described a full
implementation of patient-specific MC dosimetry, showing spillover
artefacts [1]. The algorithm was validated by comparing with dosimetric
quantities using the MIRD Standard Man phantom for representative case
studies. Validation by comparing with other published data has also been
made by Tagesson et al. [21]. An implementation of a patient-specific
point-dose kernel dosimetry program for SPECT/PET images with informa-
tion from anatomical modalities has been described by Kolbert et al. [22].
Evaluations based on direct measurements 275
These are further described in chapter 11. An MCNP4B-based program for

patient-specific dosimetry has been described by Yoriyaz et al. [23] where the
Zubal phantom was used as a patient and absorbed photon fractions verified
against reference values from a standard geometry phantom [24].
12.6. EVALUATIONS BASED ON DIRECT MEASUREMENTS
Due to logistical constraints, absorbed dose distributions within patients are

necessarily calculated using limited data sets. Quantitative imaging informa-
tion, as discussed above, is obtained to provide the absorbed dose-program
with relevant information on the activity kinetics and anatomy of the patient.
Dosimetry calculations are performed either using predefined standard
anatomy with associated conversion factors or in three dimensions taking
into account the individual anatomy and attenuation properties of each
patient. Although sophisticated computational techniques have been applied
to these problems, there is a great need for direct absorbed dose measure-
ments to verify the complete calculation/treatment chain.
Measurements of absorbed dose for verification of calculations can be
divided into verification in phantoms or in vivo. The latter part could be
either in experimental animals or in humans. A previous overview of
absorbed dose estimation methods in radionuclide therapy has been
summarized by Strand et al. [25]. Radionuclide dosimetry measurements,
especially when applying the dosimeter in vivo or in contact with the skin,
means that besides energy imparted by photons also the particle
contribution has to be considered.
12.6.1. Thermoluminescence dosimeters

The most-used dosimeter for verifications of absorbed dose calculations has
been the thermoluminescence dosimeter (TLD). The properties of TLDs
have extensively been investigated, because of their wide usage in external
beam therapy [26]. Wessels and Griffith [27] in the mid-1980s suggested
applying mini-TLDs in vivo. Using sufficiently small TLDs, absorbed dose
non-uniformities can be measured within tissues and tumours. The
CaSO4 (Dy) dosimeter has been used as a mini-TLD due to its high light
output at low absorbed doses. The method has been extensively explored
by several groups for its dosimetric characteristics [28–30] and parameters
such as the sensitivity, signal loss, pH-sensitivity and supra-linearity have
been studied. Signal loss in tissue was found to be as high as 70% after 9
days, and varied with the temperature in the tissue. Strong pH dependence
was also found. The increase in sensitivity, i.e., the supra-linearity factor,
was 2–3 at 20 Gy. The mini-TLDs can be used in vivo if care is taken for
their signal loss due to environmental impact on the dosimetry material.
There is a recent development of encapsulated mini-TLDs that should

be able to resist the mentioned environmental factors and should be more
rigorous for in vivo use. These show no pH dependence and are not affected
by the aquatic environment [31].
Volumetric dosimetry using TLDs has been suggested by Aissi and
Poston [32] using a mixture of paraffin, tetrachlorobenzene and LiF TLD-
100. The dosimeter is tissue equivalent and can be shaped arbitrarily.
12.6.2. Biological dosimeters

Ionizing radiation induces a number of specific and stable alterations in
biological materials. Such changes have been examined as a possible
approach for estimating radiation exposure and absorbed dose. Historically,
this approach has been driven by the need to provide a retrospective evalua-
tion of absorbed dose following accidental or uncontrolled exposures to
radiation. In the search for a reliable biological dosimeter, investigators
have considered procedures based on biochemical indicators, peripheral
blood cell count (lymphocytes, reticulocytes and other peripheral blood
mononuclear cells), chromosomal aberrations, and formation of micronuclei
[33–37]. Chromosomal aberrations include transient changes such as
acentrics (lacking a centromere), dicentrics (having two centromeres) and
ring chromosomes as well as stable changes such as chromosomal trans-
locations (transposition of chromosomal segments between two different
chromosomes). The former are typically observed by Giemsa staining; the
latter require fluorescence in situ hybridization (FISH) with chromosome-
specific DNA libraries. Micronuclei are small secondary nuclei that arise
from acentric chromosomal fragments or whole chromosomes that fail to
incorporate into the daughter nuclei during mitosis [38]. A detailed review
of biological dosimetry is outside the scope of this section; rather, this section
will focus on biological dosimetry as used to validate or evaluate computa-
tional methods for radionuclide dosimetry.
One of the earliest applications of biological dosimetry in radionuclide
therapy was in the treatment of thyroid cancer with radioiodine [39].
Recent studies have compared the calculated with biologically derived
absorbed dose to bone marrow after single and repeated treatment with
131
I [40, 41]. Chromosomal aberrations in peripheral lymphocytes were
assessed by conventional means (Giemsa staining) and also by FISH.
Calibration of the biological dosimeters was obtained, in vitro, using 60 Co
(high-dose-rate) and 131 I (low-dose-rate) irradiation. The lowest dose
detected, in vitro, was 0.2 Gy. Consistent with previous studies, frequency
of aberrations was not significantly dependent on dose-rate. Induced trans-
locations were about three- to eightfold more frequent than dicentrics for
each dose of irradiation, again consistent with earlier studies [42, 43].
Based upon the frequency of chromosomal aberrations in peripheral
lymphocytes 4 days after administration of 3.7 GBq 131 I, an absorbed dose of

0.5 Gy was estimated for a patient population in which thyroid uptake
ranged from 0 to 9.1% of administered activity. This is close to the 0.3 to
0.4 Gy range expected for a 3.7 GBq 131 I administration for patients with
similar thyroid uptake values [44, 45]. Interestingly, the biologically
determined absorbed dose correlated with total body retention but not
with thyroid uptake.
Patients that were retreated over a 5 year time-period showed an increase
in chromosomal aberrations before and after each administration of 3.7 GBq
131
I consistent with the 0.5 Gy estimate obtained for patients treated only once.
The cumulated increase in chromosomal aberrations, however, did not reflect
the expected cumulative absorbed dose. Based on the product of 0.5 Gy and
the number of re-treatments, a cumulative dose of 1 to 3.5 Gy should have
been obtained; instead the cumulated chromosomal aberrations (including
translocations) yielded a dose of only 0.5 to 1.23 Gy. This finding was
attributed to the possible loss of cells with chromosomal abnormalities via
apoptosis.
The micronucleus assay has also been used in examining radiotoxicity
after 131 I treatment of thyroid cancer or hyperthyroidism [46–53]. These
studies all observed an increase in micronucleus frequency after 131 I treat-
ment. In patients with thyroid cancer receiving 3.7 GBq 131 I, the biologically
determined absorbed dose was typically 0.3 Gy. In one report, the same
technique was used to investigate biological dose for thyroid cancer and
hyperthyroid patients, the latter receiving three- to tenfold less radioactivity
then the former. Surprisingly, the results showed the same increase in
micronuclei, corresponding to a dose of 0.3 Gy [49], although this is not
supported by conventional dosimetry calculations which yield approximately
a twofold increase in the marrow absorbed dose per unit administered
activity [54].
Biological dosimetry has also recently been applied to investigate
marrow absorbed dose in radioimmunotherapy and in radionuclide therapy
for bone pain palliation [55–59]. In a series of patients participating in a
Phase I radioimmunotherapy trial of 90 Y-cT84.66, a 90 Y-labelled human/
mouse chimeric antibody against the carcinoembryonic (CEA) antigen
[59]. The frequency of stable chromosomal translocations (SCTs) in 18
patients receiving one or more cycles of the radiolabelled antibody was
measured after each cycle. Red marrow doses were estimated assuming a
marrow activity concentration equal to 0.3 times that of blood [58]. A
linear correlation was observed (r ¼ 0:79 and 0.89) between increases in
SCT for chromosomes 3 (r ¼ 0:79) and 4 (r ¼ 0:89) and calculated marrow
absorbed dose. Despite this observation, a weak correlation was observed
between haematological toxicity (fractional decrease in platelet and white
blood cell count) and red marrow dose (r ¼ 0:61 and 0.53, respectively) or
translocation frequency (r < 0:4).
In a detailed series of murine studies examining marrow dosimetry for

bone-targeting radionuclides, survival of granulocyte-macrophage colony-
forming cells (GM-CFCs) was calibrated and used as a biological dosimeter
[55, 56, 58]. Calibration for dose-rate was obtained using an external 137 Cs
source that could be differentially attenuated to achieve a given exponentially
decreasing dose-rate profile. An initial dose rate of 0.37 cGy/h per kBq of
femur activity was measured for 90 Y-citrate with the dose-rate decreasing
exponentially with a 62 h half-time. The same approach was used to
demonstrate that shorter-ranged Auger electron emitters such as 117m Sn
and 33 P are promising candidates for bone pain palliation because of lower
marrow absorbed dose and lower expected toxicity. Since GM-CFC cells
are progenitor cells that reside in the marrow, these studies required highly
invasive procedures to collect adequate marrow for colony formation
assays. To overcome the highly invasive nature of these studies, the induction
of micronuclei in peripheral reticulocytes was also examined [57]. These
studies gave initial dose-rates of 0:0020 0:0004 and 0:0026 0:0002 cGy/h
per kBq administered for 32 P-orthophosphate and 90 Y-citrate, respectively.
Corresponding values obtained using the GM-CFC survival assay were
0:0031 0:0004 and 0:0030 0:0003 cGy/h per kBq administered.
To summarize, biological dosimeters have been used extensively in the
treatment of thyroid disease with 131 I. These studies have shown that
biologically estimated blood or marrow absorbed doses are in generally
good agreement with calculated values. Use of biological dosimeters in radio-
immunotherapy and bone palliation therapy has only recently been examined
and although a correlation between biologically estimated and calculated
absorbed dose has been established in patients, absolute comparison of the
two approaches has not yet been carried out.
12.6.3. Other dosimetry methods

MOSFET dosimeters [60] can be used as in vivo probes since the absorbed
dose and absorbed dose rate can be measured simultaneously. One drawback
is the electrical coupling; another is its non-tissue equivalence.
A novel idea is to use laser induced fluorescence (LIF) in crystal
materials [61]. Here the crystal is mounted on an optical fibre. The dosimeter
can be very small and no electrical circuitry is connected to the patient. The
dosimeter signal is linear with absorbed dose over several decades and
simultaneously the absorbed dose and dose rate can be measured [62, 63].
Electron paramagnetic resonance (EPR) signal of irradiated mineralized
tissues has been suggested as a dosimetry system to be used in the dosimetry
of bone-seeking radiopharmaceuticals [64, 65]. Chemical dosimeters could be
an alternative with the benefit of the possibility of having any shape.
Methods suggested for in vivo dosimetry are electron spin resonance (ESR)
using L--alanine, almost tissue equivalent regarding mass stopping power
(Scol =) and mass energy absorption coefficient (en =) [66, 67]. The system,
however, has low sensitivity and can only be used for higher absorbed doses,
as expected in therapeutic nuclear medicine. Another system that could be
suggested is the gel dosimetry system read with ordinary MR cameras.
Here one uses gels containing ferrous ions. These dosimeters are soft tissue
equivalent, however, also requiring relative high absorbed doses [68]. They
can be made in arbitrary shapes.
12.6.4. Phantoms for experimental measurements

In an early attempt to verify the MIRD formalism Grönberg et al. [69] built a
phantom of the MIRD mathematical model and filled it with water. The
urinary bladder, liver or kidney was filled with 99m Tc, 123 I or 131 I and
sealed LiF dosimeters were placed at the liver, kidneys, lungs and ovaries.
Measurements showed that the calculated absorbed dose compared with
the measured only differed within 20%. Aissi et al. [32, 81] compared the
MIRD calculations with experimental absorbed doses in the MIRD
mathematical phantom obtained with a volumetric dosimeter. Results
showed an agreement between the calculated and measured data with a
slight overestimation of calculated values.
Kwok et al. [70] calculated the absorbed dose for 32 P in interfaces and
compared those with LiF measurements. They simulated a soft-tissue–
bone planar interface by a polystyrene (PST)–aluminium junction and
measured the change in dose from the dose value in homogeneous PST
due to a point source of 32 P using LiF thermoluminescent dosimeters.
With the point source at the interface, the dose rates increased by up to
12%. With the point source at a PST–air planar interface to simulate a
soft-tissue–air junction, the dose rates decreased as much as 25%. Giap
et al. [71] compared the absorbed dose calculations based on a three-
dimensional activity map from SPECT and point dose kernels with
measurements by TLDs in a large water phantom. Comparisons were also
made with MIRD S values in an Alderson abdominal anthropomorphic
phantom. The results showed that their three-dimensional method was in
good agreement with both the TLD measurements and the MIRD
formalism. Their measurements were within 8% of calculated values and
in good agreement (14%) with MIRD formalism. TLDs have also been
used in a water phantom for verification of calculated absorbed doses for
131
I by Aschan et al. [72].
12.6.5. In vivo verifications—experimental animals

Mini-TLDs have been used in animals in several experiments [73]. In another
study correction for signal loss, spectral shift of photon energy spectra and
calibration in equivalent media was done for rats injected with 131 I-labelled
antibodies. Absorbed dose calculations were based on animal-specific S

values [74]. The corrected mini-TLD measurements overestimated the
absorbed dose, which might be explained due to no correction for void
volume [75].
Desrosiers et al. [76] have used EPR for determination of absorbed
doses to bone tissue from different radiopharmaceuticals aimed at therapy
and palliation. Results for beagle bone exposed to radiopharmaceuticals
under clinical conditions indicated that the EPR gave approximately the
calculated absorbed dose based on MIRD formalism, but suggest that the
absorbed dose distribution may be non-uniform. In another study, Ignatiev
et al. [65] used EPR dosimetry in dogs injected with 90 Y.
12.6.6. In vivo verifications—humans

Using TLDs on the skin of patients the absorbed dose in radionuclide
therapy has been measured. In clinical settings, also, mini-TLDs have been
used [77, 78]. TLDs were used by Aschan et al. [72] in patients with pseudo-
myxoma undergoing radioimmunotherapy with 131 I B72.3 monoclonal
antibody. They determined the absorbed dose to the kidneys from TLD
measurements on the skin and compared with MIRD formalism and
point-dose kernel technique. The overall accuracy in phantom measurements
was 15% and in patients around 50%.
12.6.7. Stack phantom for realistic RNT dosimetry verifications

Verification of arbitrary absorbed dose distributions requires access to
phantoms that easily can configure arbitrary activity distributions. A novel
idea is to use the stack phantom for such purposes [79]. The idea is based
on the stack phantom for diagnostic imaging [80]. For clinical nuclear
medicine with this technique, it is possible to obtain ‘ideal’ experimental
images without disturbance from photon interaction with the phantom
material.
In principle, the method is based on discrete sampling of a radioactivity
distribution in three-dimensional objects by means of equidistant two-
dimensional planes. These two-dimensional planes or templates can be
obtained by processing X-ray CT images. The automation of the CT
images is obtained by image processing programs to make it possible to
introduce different pathological patterns, simply by simulating some regions
inside the image with different sphere sizes and greyscale levels. These digital
activity distribution maps, visualized on the computer screen, can then be
printed out on paper sheets by using an ordinary ink-jet printer, where the
ink has been mixed with an appropriate radionuclide, The radioactive
sheets can then be stacked or mounted in an equidistantly spaced pattern
to form a three-dimensional structure with some arbitrary dense material
References 281
stacked in between. This stacking material is optional. If attenuation and

scatter free projections are the goal some low-density material may be
used. When the demand is to obtain patient-like projections, some tissue-
equivalent material should be used.
Utilizing the stack phantom any software phantom geometry and
activity distribution can easily be transferred into a real phantom to perform
direct absorbed dose measurements for verification of the dosimetry/treat-
ment calculations.
ACKNOWLEDGMENTS
This work was partially funded by US Public Health Service grant RO1
CA80927, the Swedish Cancer Foundation, The Gunnar Nilsson Founda-
tions and the Bertha Kamprads Foundations.
REFERENCES

emitter therapy Med. Phys. 24(7) 1163–72
[2] Ljungberg M, Strand S-E and King M A 1998 Monte Carlo Calculation in Nuclear
Medicine: Applications in Diagnostic Imaging (Bristol: IOP Publishing)
[3] DeVries D J, Moore S C, Zimmerman R E, Friedland B and Lanza R C 1990
Development and validation of a Monte Carlo simulation of photon transport in
an Auger camera IEEE Trans. Med. Imag. 4 430–8
[4] Baird W H, Frey E C, Wang W T and Tsui B M W 2000 Characterization of collima-
tor scatter and penetration response for I-123 imaging J. Nucl. Med. 41(5) 133P
[5] Pollard K P, Bice A N, Eary J F, Durack L D and Lewellen T K 1992 A method for
imaging therapeutic doses of iodine-131 with a clinical gamma camera J. Nucl. Med.
33 771–6
[6] Yanch J C and Dobrzeniecki A B 1993 Monte Carlo simulation in SPECT: complete
3D modeling of source, collimator and tomographic data acquisition IEEE Trans.
Nucl. Sci. 40 198–203
[7] Zubal I G and Harrell C R 1994 Computerized three-dimensional segmented human
anatomy Med. Phys. 21 299–302
[8] Dawson T W, Caputa K and Stuchly M A 1997 A comparison of 60 Hz uniform
magnetic and electric induction in the human body Phys. Med. Biol. 42(12) 2319–29
[9] Sjögreen K, Ljungberg M, Wingårdh K, Erlandsson K and Strand S-E 2001
Registration of emission and transmission whole-body scintillation camera images
J. Nucl. Med. 42(10) 1563–70
[10] Pretorius P H, Xia W, King M A, Tsui B M, Pan T S and Villegas B J 1997 Evaluation
of right and left ventricular volume and ejection fraction using a mathematical
cardiac torso phantom J. Nucl. Med. 38(10) 1528–35
[11] Segars W P, Lalush D S and Tsui B M W 2002 Modeling respiratory mechanics in the
MCAT and spline-based MCAT phantoms IEEE Trans. Nucl. Sci. 48 89–97
[12] Segars W P 2001 Development of a new dynamic NURBS-based cardiac-torso (NCAT)

phantom (University of North Carolina)
[13] Dewaraja Y K, Ljungberg M and Koral K F 2000 Accuracy of 131I tumor quanti-
fication in radioimmunotherapy using SPECT imaging with an ultra-high-energy
collimator: Monte Carlo study J. Nucl. Med. 41(10) 1760–7
[14] Dewaraja Y K, Ljungberg M and Koral K F 2000 Characterization of scatter
and penetration using Monte Carlo simulation in 131I imaging J. Nucl. Med. 41(1)
123–30
[15] Dewaraja Y K, Ljungberg M, Bose A, Majumdar A and Koral K 2002 A parallel
Monte Carlo code for planar and SPECT imaging: implementation verification and
applications in I-131 SPECT Comp. Meth. Progr. Biomed. 67 115–24
[16] Dewaraja Y K, Ljungberg M and Koral K 2001 Monte Carlo evaluation of object
shape effects in I-131 SPECT tumor activity quantification Eur. J. Nucl. Med. 28
900–6
[17] Moore S C 1998 Design of a collimator for imaging 111-In. In Monte Carlo
Calculation in Nuclear Medicine: Applications in Diagnostic Imaging ed M Ljungberg,
S-E Strand and M A King (Bristol: IOP Publishing) pp 183–93
[18] DeVries D J, King M A and Moore S C 1999 Characterization of spectral and spatial
distributions of penetration, scatter and lead X-rays in Ga-67 SPECT. Conference
Records of the IEEE Medical Imaging Conference (M9-6) pp 1701–10
[19] Ljungberg M, Sjögreen K, Liu X, Frey E C, Dewaraja Y K and Strand S-E 2002
A three-dimensional absorbed dose calculation method based on quantitative
SPECT for radionuclide therapy evaluation for 131I using Monte Carlo simulation
J. Nucl. Med. 43 1101–9
[20] Furhang E E, Chui C S and Sgouros G A 1996 Monte Carlo approach to patient-
specific dosimetry Med. Phys. 23(9) 1523–9
planning system Acta Oncol. 35(3) 367–72
[22] Kolbert K S, Sgouros G, Scott A M, Bronstein J E, Malane R A, Zhang J et al. 1997
Implementation and evaluation of patient-specific three-dimensional internal
dosimetry J. Nucl. Med. 38(2) 301–8
[23] Yoriyaz H, Stabin M and dos Santos A 2001 Monte Carlo MCNP4B-based
absorbed dose distribution estimates for patient-specific dosimetry J. Nucl. Med.
42(4) 662–9
[24] Yoriyaz H, dos Santos A, Stabin M and Cabezas R 2000 Absorbed fractions in a
voxel-based phantom calculated with the MCNP-4B code Med. Phys. 27(7) 1555–62
[25] Strand S-E, Jönsson B-A, Ljungberg M and Tennvall J 1993 Radioimmunotherapy
dosimetry—a review Acta Oncol. 33 807–17
[26] Horowitz Y S 1984 Thermoluminescence dosimetry. In Thermoluminiscence and
Thermoluminescent Dosimetry ed Y S Horowitz (Boca Raton, FL: Chemical
Rubber Company) pp 43–130
[27] Wessels B W and Griffith M H 1986 Miniature thermoluminescent dosimeter
absorbed dose measurements in tumor phantom models J. Nucl. Med. 27 1308–14
[28] Demidecki A J, Williams L E, Wong Y C, Wessels B W, Yorke E D, Strandh M et al.
1993 Considerations on the calibration of small thermoluminescent dosimeters used
for measurement of beta particle absorbed doses in liquid environments Med. Phys.
20(4) 1079–87
References 283
[29] Strand S-E and Strandh M 1994 Parameters to consider for measurements of
absorbed doses in vivo with mini-thermoluminescent dosimeters Cancer 73(3) 985–8
[30] Strandh M and Strand S-E 1996 In vivo absorbed dose measurements with mini-
TLDs. Parameters affecting the reliability Acta Oncol. 35(6)
[31] Martin S, Lisbona A, Richard J, Morteau S, Denizot B and Bardies M 2000
Production of new thermoluminescent mini-dosimeters Phys. Med. Biol. 45(2) 479–94
[32] Aissi A and Poston J W 1984 An improved volumetric dosimeter for internal dose
verification Health Phys. 46(2) 371–6
[33] Skalka M 1971 Biochemical Indicators of Radiation Injury in Man (Vienna:
International Atomic Energy Agency)
[34] Bender M A and Gooch PC 1962 Types and rates of X-ray-induced chromosome
aberrations in human blood irradiated, in vitro Proc. Natl Acad. Sci. USA 48
52232
[35] Chambers K A, Harrington N P, Ross W M and Filion L G 1998 Relative alterations
in blood mononuclear cell populations reflect radiation injury in mice Cytometry
31(1) 45–52
[36] Pantelias G E and Maillie H D 1984 The use of peripheral blood mononuclear cell
prematurely condensed chromosomes for biological dosimetry Radiat. Res. 99(1)
140–50
[37] Lenarczyk M and Slowikowska M G 1995 The micronucleus assay using peripheral
blood reticulocytes from X-ray-exposed mice Mutat. Res. 335(3) 229–34
[38] Fenech M and Morley A A 1985 Measurement of micronuclei in lymphocytes Mutat.
Res. 147(1/2) 29–36
[39] Boyd E, Buchanan W W and Lennox B 1961 Damage to chromosomes by therapeutic
doses of radioiodine Lancet i 977–8
[40] M’Kacher R, Schlumberger M, Legal J D, Violot D, Beron-Gaillard N, Gaussen A
et al. 1998 Biologic dosimetry in thyroid cancer patients after repeated treatments
with iodine-131 J. Nucl. Med. 39(5) 825–9
[41] M’Kacher R, Legal J D, Schlumberger M, Voisin P, Aubert B, Gaillard N et al. 1996
Biological dosimetry in patients treated with iodine-131 for differentiated thyroid
carcinoma J. Nucl. Med. 37(11) 1860–4
[42] Tucker J D, Ramsey M J, Lee D A and Minkler J L 1993 Validation of chromosome
painting as a biodosimeter in human peripheral lymphocytes following acute
exposure to ionizing radiation in vitro Int. J. Radiat. Biol. 64(1) 27–37
[43] Schmid E, Zitzelsberger H, Braselmann H, Gray J W and Bauchinger M 1992
Radiation-induced chromosome aberrations analysed by fluorescence in situ
hybridization with a triple combination of composite whole chromosome-specific
DNA probes Int. J. Radiat. Biol. 62(6) 673–8
[44] Zanzonico P B 1997 Radiation dose to patients and relatives incident to 131I therapy
Thyroid 7(2) 199–204
[45] Smith T and Edmonds C J 1984 Radiation dosimetry in the treatment of thyroid
carcinoma by 131I Radiat. Prot. Dosim. 5 141–9
[46] Wuttke K, Streffer C, Muller W U, Reiners C, Biko J and Demidchik E 1996
Micronuclei in lymphocytes of children from the vicinity of Chernobyl before and
after 131I therapy for thyroid cancer Int. J. Radiat. Biol. 69(2) 259–68
[47] Watanabe N, Yokoyama K, Kinuya S, Shuke N, Shimizu M, Futatsuya R et al. 1998
Radiotoxicity after iodine-131 therapy for thyroid cancer using the micronucleus
assay J. Nucl. Med. 39(3) 436–40
[48] Ramirez M J, Surralles J, Galofre P, Creus A and Marcos R 1997 Radioactive iodine
induces clastogenic and age-dependent aneugenic effects in lymphocytes of thyroid
cancer patients as revealed by interphase FISH Mutagenesis 12(6) 44955
[49] Monsieurs M A, Thierens H M, van de Wiele C V, Vral A M, Meirlaen I A, de Winter
H A et al. 1999 Estimation of risk based on biological dosimetry for patients treated
with radioiodine Nucl. Med. Commun. 20(10) 911–17
[50] Livingston G K, Foster A E and Elson H R 1993 Effect of in vivo exposure to iodine-131
on the frequency and persistence of micronuclei in human lymphocytes J. Toxicol.
Environ. Health 40(2/3) 367–75
[51] Lambert V, Thierens H, Monsieurs M, Roncancio C and Laurent C 2001 Trans-
location frequencies measured in patients one year after radioactive iodine therapy
for thyrotoxicosis Int. J. Radiat. Biol. 77(6) 679–85
[52] Gutierrez S, Carbonell E, Galofre P, Creus A and Marcos R 1997 Micronuclei induc-
tion by 131I exposure: study in hyperthyroidism patients Mutat. Res. 373(1) 39–45
[53] Catena C, Villani P, Nastasi R, Conti D, Righi E, Salerno G et al. 1994 Micronuclei
and 3AB-index in patients receiving iodine-131 therapy J. Nucl. Biol. Med. 38(4)
58693
[54] McEwan A C 1977 Absorbed doses in the marrow during 131I therapy Br. J. Radiol.
50 329–31
[55] Goddu S M, Howell RW, Giuliani D C and Rao D V 1998 Biological dosimetry of
bone marrow for incorporated yttrium-90 J. Nucl. Med. 39(3) 547–52
[56] Goddu S M, Bishayee A, Bouchet L G, Bolch W E, Rao D V and Howell R W 2000
Marrow toxicity of 33P- versus 32P-orthophosphate: implications for therapy of bone
pain and bone metastases J. Nucl. Med. 41(5) 941–51
[57] Lenarczyk M, Goddu S M, Rao D V and Howell R W 2001 Biologic dosimetry of
bone marrow: induction of micronuclei in reticulocytes after exposure to 32P and
90
Y J. Nucl. Med. 42(1) 162–9
[58] Siegel J A, Wessels B, Watson E E, Stabin M, Vriesendorp H M, Bradley E W et al.
1990 Bone marrow dosimetry and toxicity for radioimmunotherapy Antibody
Immunocon. Radiopharm. 3 213–33
[59] Wong J Y, Wang J, Liu A, Odom-Maryon T, Shively J E, Raubitschek A A et al. 2000
Evaluating changes in stable chromosomal translocation frequency in patients
receiving radioimmunotherapy Int. J. Radiat. Oncol. Biol. Phys. 46(3) 599–607
[60] Gladstone D J, Lu X Q, Humm J L, Bowman H F and Chin L M 1994 A miniature
MOSFET radiation dosimeter probe Med. Phys. 21 1721–8
[61] Miller S D and Endres G W 1990 Laser-induced optically stimulated M centre
luminescence in LiF Radiat. Prot. Dosim. 33(1) 59–62
[62] Strandh M, Johansson J and Strand S E A 2001 BaF(Br,I)Eu dosimeter based on laser
induced fluorescence for in in vivo measurements (http://www.doseinfo-radar.com/1-
pos.htm)
[63] Strandh M, Strand S-E, Johansson J and Malmgren L 2002 Dosimetry based on
laser induced fluorescence in LiF for in vivo measurements (http://www.doseinfo-
radar.com/1-pos.htm)
[64] Schauer D A, Desrosiers M F, Le F G, Seltzer S M and Links J M 1994 EPR
dosimetry of cortical bone and tooth enamel irradiated with X and gamma rays:
Study of energy dependence Radiat. Res. 138 1–8
[65] Ignatiev E A, Lyubashevskii N M, Shishikina E A and Romanyukha A A 1999 EPR
dose reconstruction for bone-seeking 90Sr Appl. Radiat. Isot. 51(2) 151–9
References 285
[66] Wielopolski L, Maryanski M, Ciesielski B, Forman A, Reinstein L E and Meck A G

1987 Continuous three dimensional radiation dosimetry in tissue equivalent phantoms
using electron paramagnetic resonance in L--alanine Med. Phys. 14 646–52
[67] Olsson S, Lund E and Erickson R 1996 Dose response and fading characteristics of an
alanine-agarose gel Appl. Radiat. Isot. 47(11) 1211–17
[68] Olsson L E, Bäck S Å J, Magnusson P and Haraldsson P 1998 3D Dosimetry Using
Gels and MRI eds J D Hazle and A L Boyer (College Park, American Association
of Physicists in Medicine. Imaging in Radiation Therapy) pp 475–504
[69] Grönberg T, Persson B and Strand S-E 1977 Experimental evaluation of organ doses
from exposure to Tc-99m, I-123 and I-131 Proc. IVth Int. Cong. Int. Radiat. Prot.
Assoc. (Paris 1977) pp 493–6
[70] Kwok C S, Irfan M, Woo M K and Prestwich W V 1987 Effect of tissue
inhomogeneity on beta dose distribution of 32P Med. Phys. 14 98–104
kernel convolution technique for internal dosimetry Phys. Med. Biol. 40(3) 365–81
[72] Aschan A C, Toivonen M J, Lampinen J S, Tenhunen M, Kairemo K J A, Korppi-
Tommola E T et al. 1999 The use of TL detectors in dosimetry of systemic
radiation therapy Acta Oncol. 38(2) 189–96
[73] Klein J L, Nguyen T H, Laroque P, Kopher K A, Williams J R, Wessels B W et al.
1989 90Yttrium and 131iodine radioimmunoglobulin therapy of an experimental
hepatoma Cancer Res. 49 6383–9
[74] Hui T E, Fisher D R, Kuhn J A, Williams L E, Nourigat C, Badger C C et al. 1994 A
mouse model for calculating cross-organ beta doses from yttrium-90-labeled
immunoconjugates Cancer 73(3) 951–7
[75] Strand S E, Strandh M, Ljungberg M and Tagesson M 1994 Mini-Tl-dosimeters for in
vivo measurements J. Nucl. Med. 35 78P
[76] Desrosiers M F, Avila M J, Schauer D A, Coursey B M and Parks N J 1993
Experimental validation of radiopharmaceutical absorbed dose to mineralized bone
tissue Appl. Radiat. Isot. 44 459–63
[77] Chiou R K, Wessels B W, Woodson M and Limas C 1991 Study of clinical thermo-
luminescent dosimeters (Cl.TLD) in direct measurement of absorbed dose for
radioimmunotherapy Int. J. Radiat. Appl. Instrum. A 42 181–6
[78] Leichner P K, Yang N-C, Wessels B W, Hawkins W G, Order S E and Klein J L 1990
Dosimetry and treatment planning in radioimmunotherapy Front. Radiat. Ther.
Oncol. 24 109–20
[79] El-Ali H H, Strand S-E, Larsson S A, Johnsson C, Palmer J, Ljungberg M et al. 2002
A new phantom based on ink jet technique for imaging and dosimetry. In Proceedings
of the Seventh International Radiopharmaceutical Dosimetry Symposium, Nashville,
TN
[80] Larsson S A, Jonsson C, Pagani M, Johansson L and Jacobsson H 2000 A novel
phantom design for emission tomography enabling scatter- and attenuation-‘free’
single-photon emission tomography imaging Eur. J. Nucl. Med. 27 131–9
[81] Aissi A and Poston J W 1987 Comparison of measured and calculated internal
absorbed doses in a heterogenous phantom Phys. Med. Biol. 32 1245–56
Chapter 13
Monte Carlo methods and mathematical

models for the dosimetry of skeleton and
bone marrow
Lionel G Bouchet, Wesley E Bolch, Michael Stabin,
Keith F Eckerman, John W Poston, Sr
and A Bertrand Brill
13.1. INTRODUCTION
13.1.1. Anatomy of the skeletal system

The entire skeleton in an adult consists of 200 distinct bones divisible into
four classes: long, short, flat and irregular bones [1]. These 200 bones form
a rigid framework to which the softer tissues and organs of the body are
attached. Microscopically, the human skeletal system is a complex structure
that can be classified into two distinct types: trabecular bone and cortical
bone (figure 13.1). The trabecular bone (also called cancellous bone) is a
porous, spongy structure found in the interior of the flat bones and at the
ends of the long bones. It is formed of a complex network of bone trabeculae
and tissue cavities. Each cavity is lined by a layer of osteogenic cells called the
endosteum and is filled with marrow that is either haematopoietically active
(red marrow) or inactive (yellow marrow). The cortical bone is a hard, dense
structure also called compact bone. It is found in the shafts of the long bones
and in the outer cortex of all bones. The dominant microstructure of the
cortical bone is the osteon, which is formed by a central Haversian canal
lined with a layer of endosteum and surrounded by bone lamellae. Typical
dimensions of the microstructures of an adult trabecular and cortical bone
are given in table 13.1.
Radiosensitive cells in the human skeletal system have been identified as
(a) haematopoietic cells present in bone marrow, (b) endosteal cells lying
close to bone surfaces and (c) epithelial cells close to bone surfaces in the
286
Introduction 287
Figure 13.1. Diagram of the upper half of a femur showing the trabecular bone
regions in the femur head (NMR image) and neck, and the cortical bone regions
in the diaphysis.
air sinuses of the skull [2]. Dosimetrically important regions in therapeutic

nuclear medicine are considered to be the haematopoietic stem and precursor
cells in the marrow cavities and the osteogenic cells on the surface of the bone
trabeculae. Doses to epithelial cells lining ethmoid, sphenoid, mandibular
and maxillary sinuses from emitters in bone or in gaseous form [3–5] are
generally not applicable for nuclear medicine dosimetry.
Table 13.1. Typical geometrical dimensions of trabecular and cortical micro-

structures of an adult [6].
Feature Typical dimensions (mm)
Trabecular bone
Bone trabeculae in parietal bone 511
Bone trabeculae in cervical vertebra 279
Bone trabeculae in femur head 232
Marrow cavity in parietal bone 389
Marrow cavity in cervical vertebra 910
Marrow cavity in femur head 1157
Cortical bone
Osteon length (between Volkmann canals) 2500
Concentric lamellae thickness 7
Osteon diameter 200–400
Haversian canal diameter 20–200
288 Dosimetry of skeleton and bone marrow
13.1.2. Dosimetry of the skeletal system

In radioimmunotherapy (RIT) and radionuclide bone pain palliation, the
skeletal system, and more precisely the bone marrow, has been one of the
main dose-limiting organs [7–11]. Dose calculations in nuclear medicine
are principally based on the MIRD methodology [12]. In a practical
approach, the MIRD methodology simply decomposes the average dose to
a target region from a specific source region into a ‘biological’ term,
A~source , and a ‘physical’ term, S(target source):
ðtarget
D sourceÞ ¼ A~source Sðtarget sourceÞ; ð13:1Þ
where A~source is the cumulated activity or the total number of nuclear

transitions of the radionuclide in the source region, and S(target source)
is the S value or the average absorbed dose in the target per unit cumulated
activity in the source region. The cumulated activity is generally determined
on a per-patient basis from SPECT, planar imaging (scintigraphy) [13] or
blood activity count [9–11]. S values are tabulated for specific radionuclide,
source and target combinations. The calculation of an S value requires
knowledge of the radiations emitted by the radionuclide and information
on the distribution of absorbed energy within target regions of interest in
the anatomical model for monoenergetic sources distributed within source
regions of the model; such data are referred to as absorbed fractions of
energy. The calculation of absorbed fractions of energy requires the use of
a model of the anatomy of interest (source, target and surrounding media)
and a particle transport code. For most source–target regions in the body,
both models are well differentiated and anatomical models are based on
the combination of simple mathematical shapes approximating the anatomy
[14], and the particle transport is based on Monte Carlo techniques [15]
and=or point kernel algorithms [16].
Macroscopically, the skeletal system is well defined throughout the
body, and its geometry can be modelled using simple mathematical shapes
[15–17]. However, it is an inhomogeneous region with radiosensitive tissues
embedded within a dense bone matrix defined at the microscopic level.
Therefore, a different approach must be employed to incorporate both the
macroscopic and microscopic anatomy and, because directly ionizing and
indirectly ionizing radiations have widely different interaction ranges, they
are separately considered in the dosimetry modelling.
There are three approaches that have been developed and used to
calculate the energy deposited by indirectly-ionizing radiations in the
sensitive regions of the skeletal system. These three methods are all
based on the use of a homogeneous macroscopic model of the skeletal
system (uniform mixture of bone and soft tissue with density 1.4 g cm3 )
within mathematical anthropomorphic phantoms [15, 16] that is coupled
to a photon transport Monte Carlo code. The first method was developed
Introduction 289
by Snyder et al. for the tabulation of dose conversion factors for various
radionuclides [18], and was subsequently used in MIRD Pamphlets Nos.
5 and 11 [14, 15]. Snyder et al. calculated the energy deposited by
photon interactions in the macroscopic skeletal model, and partitioned it
by target mass to generate the energy deposited in the sensitive tissues of
the skeletal system. This approach assumed that the range of secondary
electrons was smaller than the marrow cavity sizes and equivalence in
the mass absorption coefficients for marrow and bone. In an effort to
improve upon these assumptions, Eckerman presented a new approach in
which photon fluence calculated in a homogenous skeletal region of the
anthropomorphic phantom is coupled with dose per unit fluence response
factors derived considering the creation and transport of the secondary
electrons in a microscopic model of the skeletal region [16, 19, 20]. This
second approach was used in the tabulation of specific absorbed fractions
of energy (absorbed fraction of energy per unit mass of target region)
published in ORNL TM 8381 [16], which were subsequently used in
MIRDOSE 3 [21]. A similar approach proposed by Bouchet et al. consisted
of calculating the energy of the secondary electrons created by photon
interactions in the macroscopic model of the skeletal system and coupling
them to electron absorbed fractions of energy calculated using a
microscopic model of the skeletal system [22]. The last two methods
differed principally in the use of two different microscopic models of the
skeletal system that will be presented and compared in subsequent
sections.
For directly ionizing radiations, there are essentially two approaches to
calculate the energy deposited in the sensitive regions of the skeletal system.
The first is to focus the modelling on closely representing the anatomical
microstructures of the skeletal system and to couple the model to a Monte
Carlo particle transport code. However, because of its intricate anatomy
and its complex microstructure, it is difficult to exactly model the skeletal
system without using three-dimensional imaging, and only a few investiga-
tors have used three-dimensional data. The second approach is to model
the particle transport in such a way that the traversed distance across a
given microscopic region (bone trabeculae or marrow cavity for example)
by a particle matches exactly the real path of the particle in the skeletal
microstructure. This approach initially was proposed by the research
group of Spiers [6, 23–29] and subsequently was used in several skeletal
dosimetry models [19, 20, 30].
In the next sections of this chapter, the different microscopic models of
the cortical and trabecular bone regions that have been used with Monte
Carlo techniques to generate dosimetric data for nuclear medicine are
presented. Because a model is based upon specific assumptions, and is
derived for a reference individual, correction methods for patient-specific
applications are given as well.
13.2. TRABECULAR BONE MODELS
13.2.1. Research at the University of Leeds

The foundation of all bone dosimetry modelling was established through the
research of Spiers and his students at the University of Leeds between 1949
and 1981 [6, 23–29]. From this research group, three major dissertations on
bone dosimetry are of particular note: one by Darley in 1972 [25], one by
Whitwell in 1973 [26], and one by Beddoe in 1976 [6]. Spiers determined
through this research that the microstructure of trabecular bone could not
be well described using simple geometrical shapes. Instead, he used frequency
distributions of linear path lengths through trabeculae and marrow cavities
as a quantitative description of the three-dimensional structure of this
region of the skeleton. For this purpose, Darley and Beddoe designed an
optical bone scanner capable of automatically measuring bone and cavity
space chord length distributions of thin sections of human trabecular bone
[26, 27]. Assuming a direction of alignment of the cavities, they derived
corresponding omnidirectional chord length distributions. A total of seven
bone sites were measured including the cervical vertebra, lumbar vertebra,
femur head, femur neck, iliac crest, parietal bone, and rib. Whitwell used
the experimentally measured omnidirectional chord length distributions to
calculate dose conversion factors (ratio of absorbed dose in target to the
absorbed dose in hypothetical trabecular bone assuming complete absorp-
tion of the emissions) for seven radionuclides (14 C, 18 F, 22 Na, 32 P, 45 Ca,
90
Sr, 90 Y) distributed within the volume of trabeculae. A Monte Carlo
process was used to sample the chord length distributions and select electron
paths through marrow cavities and bone trabeculae. The energy deposited in
those different regions was determined using range–energy relationships in
both the bone trabeculae and marrow cavities (continuous slowing down
approximation) assuming a straight electron path through these regions.
Although these calculations were performed assuming the electron deposits
all its energy in the trabecular bone, Whitwell applied correction factors to
simulate the escape of electrons from the trabecular bone for high-energy
electron sources. In this study, only bone- and surface-seeking radionuclides
were considered, and only the marrow cavity and endosteum were chosen as
target regions.
13.2.2. Model of trabecular bone in MIRD Pamphlet 11

The research of Whitwell focused on radionuclides of interest for radiation
protection and, therefore, nuclear medicine dosimetry was not of primary
importance. To apply Whitwell’s data to additional radionuclides, Snyder
et al. approximated the specific absorbed fractions for monoenergetic sources
by expressing the skeletal-averaged dose conversion factors for the seven
Trabecular bone models 291
radionuclides as a function of their average energy [26, 27]. Using these

specific absorbed fractions, trabecular bone S values were tabulated as
part of MIRD Pamphlet No. 11 [14]. In these reports, four source regions
were considered (trabecular bone, cortical bone, red marrow, and yellow
marrow) and four target regions (skeletal bone, red marrow, yellow
marrow, and skeletal endosteum). For the skeletal source–target combina-
tions not considered by Whitwell, Snyder et al. applied one of the following
five assumptions to estimate values of absorbed fraction: (a) no energy
deposition in the target region, (b) full energy deposition in the target
region, (c) conservation of energy within neighbouring regions, (d) uniform
energy deposited per unit mass in both the source and target regions, and/or
(e) the reciprocity theorem [12, 31].
13.2.3. Skeletal dosimetry in ICRP Publication 30

In 1979, the ICRP issued its Publication 30 [32], which recommended
absorbed fractions of energy for particles for use in radiation protection.
For particles originating in the bone volume, a single value of absorbed
fraction was recommended. For particles originating on the bone surface,
one absorbed fraction for low-energy particles (average energy less than
0.2 MeV), and one for high-energy particles (average energy greater than
or equal to 0.2 MeV) were recommended. These absorbed fraction values
were based on the dose conversion factors from Whitwell [26, 27], and
were intended to be conservative, as their objective was to provide a
system of radiation protection for workers and not for clinical RIT.
Subsequently, these relatively energy-independent absorbed fractions of
energy were implemented in the MIRDOSE2 program [21] for use in nuclear
medicine dosimetry. In this same computer program, the self-absorbed
fraction to the marrow was assumed to be unity as suggested in Part 3 of
the ICRP Publication 30.
13.2.4. Spherical model of trabecular bone

Chen and Poston [33] proposed a model in 1982 that assumed the marrow
spaces in trabecular bone were spherical in shape. The marrow space
dimensions and the trabecular thicknesses were chosen randomly from the
chord-length distributions and other data of Spiers and his colleagues
[6, 23–29]. A computer code, DAB-BE, was written which could be used to
calculate the absorbed energies deposited by particles or electrons inside
human trabecular bone. The radiation source geometry could be either
uniformly distributed in the whole bone or located at a fixed point inside
the bone. Bremsstrahlung production was not considered in these calcula-
tions. Results were presented for six monoenergetic electron sources (0.1,
0.5, 0.8, 1.0, 1.5 and 2.0 MeV) distributed uniformly in a mathematical
representation of the arm [18]. In addition, a fixed point source of 1.0 MeV
monoenergetic electrons was studied.
This spherical model was improved by Zuzarte de Mendoca by
consideration of the endosteal layer as well as the marrow space and
trabecular dimensions [34]. She used the Monte Carlo code EGS4 [35] and
considered both surface and volume sources. These calculations produced
absorbed fractions and specific absorbed fractions of energy for six mono-
energetic electron sources (0.1, 0.2, 0.5, 1.0, 2.0 and 4.0 MeV) for source–
target combinations in the spine, the cranium and the pelvis. She found
that the absorbed fractions in bone marrow were strongly dependent on
the ratio of trabecular bone to bone marrow volume. The higher the ratio,
the lower the absorbed fraction in bone marrow.
Parry [36] continued this research and devised a modified model of the
trabecular bone for use in the EGS4 code [35]. He calculated electron
absorbed fractions for red marrow and the endosteal tissue on trabecular
bone. The model was based on measured omnidirectional chord length
distributions through trabeculae and marrow cavities in nine skeletal
locations. Absorbed fractions were calculated for only two target regions
and all other skeletal regions were estimated based on these results. In
addition, the absorbed fractions were used to calculate S values for a variety
of -emitting radionuclides at each of 15 skeletal regions containing red bone
marrow. Absorbed fractions and S values were calculated for sources in the
marrow, on the bone surface and in the bone volume. Comparisons were
made between these results and those published in ICRP Publication 30
[32] as well as those determined using MIRDOSE2 software [21].
13.2.5. The Eckerman trabecular bone model

Eckerman [19, 20] presented new electron absorbed fractions of energy in the
trabecular bone region based on the chord length distributions measured by
Darley [6, 23, 25]. Using Whitwell’s approach [26, 27] of combining Monte
Carlo sampling of chord length distributions in bone cavities and trabeculae
with electron range–energy relationships, Eckerman derived absorbed
fractions of energy for monoenergetic electrons for the seven trabecular
bone sites measured by Darley. Source regions explicitly considered in this
model were marrow cavity, bone trabeculae and bone surface, with target
regions being the marrow cavity, endosteum and bone trabeculae. To
simulate a bone surface source, the Eckerman model assumed that electrons
were emitted at the surface between the trabeculae and marrow cavities. For
the endosteal target, a 10 mm layer of soft tissue was assumed on the bone
surface as recommended in ICRP-30 [32]. Because the radiosensitive active
marrow target was part of the simulated marrow cavity, Eckerman derived
the absorbed fraction to this sensitive structure by partitioning the energy
deposited in the full marrow cavity by the cellularity factor (percentage of
total marrow mass that is active). This model simulated no escape of the
electron from the trabecular bone, and assumed that all energy emitted
was deposited in the trabecular bone. In 1994, results from this model
were implemented in the MIRDOSE3 program [21]. Reference masses for
the endosteum and active marrow in trabecular bone of the 15 skeletal
regions defined in the anthropomorphic phantoms [16] were derived as
described in [20] and used to convert Eckerman’s absorbed fractions of
energy to radionuclide S values. This implementation allowed for explicit
calculation of regional absorbed dose in the 15 different bone sites as well
as for the whole skeleton, for two trabecular bone target regions (red
marrow and the endosteum) and three trabecular bone sources (trabecular
bone volume, red marrow and trabecular bone surface).
13.2.6. The Bouchet et al. trabecular bone model

In 1999, Bouchet et al. extended the one-dimensional transport model for
electrons in trabecular bone to a three-dimensional model [30]. This model
was based on the chord length distributions of the group of Spiers. A
series of three-dimensional spherical half-spaces of bone or marrow, with
radii consistent with the measured chord distributions, were presented to
the particle during Monte Carlo radiation transport. In this manner,
allowances were made for potential differences between the particle path-
length and its linear displacement through the media, for backscattering
electrons, as well as for the production of both -rays and bremsstrahlung.
Electron transport in the endosteum was performed by creating a three-
dimensional spherical half-space with radii selected randomly from a
random entry angle in a 10 mm thick layer. The EGS4-PRESTA Monte
Carlo transport code [35, 37] was used to simulate the electron interaction
within this three-dimensional model, assuming no escape of energy from
the trabecular bone. Similar to Eckerman’s simulations, Bouchet et al.
calculated absorbed fractions of energy for monoenergetic electrons for the
seven trabecular bone sites for which the Spiers group tabulated the chord
length distributions. Source and target regions considered were the
trabecular marrow space, the trabecular bone volume and the trabecular
bone endosteum. Results from these simulations were combined with
reference masses derived from ICRP Publication 70 for 22 skeletal sites
and nine source–target tissue combinations within the trabecular bone [38].
13.2.7. Differences between the three-dimensional model of Bouchet et al.

and the one-dimensonal model of Eckerman
The benefits of a three-dimensional transport model over a one-dimensional
model were investigated by Bouchet et al. [30]. They constructed a one-
dimensional model for electron transport in trabecular bone following the
chord length sampling methodology employed in their three-dimensional

model. The energy deposited in a given region was scored assuming a straight
electron trajectory through each tissue region and using the range–energy
relationships adopted by Whitwell and Eckerman. They reported improve-
ments over one-dimensional transport at low electron energies (less than
200 keV). Large differences were observed when the source and the target
were not adjacent regions (30–50% and upward for energies less than
100 keV) due to the possible variations in the range of the electrons modelled
by the EGS4 Monte Carlo transport code (i.e., range straggling).
Three additional differences in the methodologies used by Eckerman
and by Bouchet et al. are associated with the selection of source and target
for the active marrow and the bone surfaces. The first difference is associated
with the derivation of absorbed fractions for self-irradiation of red marrow.
In the Eckerman model, absorbed fractions of energy in the red marrow were
calculated by multiplying the absorbed fractions in the marrow cavity by
reference cellularity values for each bone region. In the Bouchet et al.
model, it was assumed that the absorbed fractions of energy to the red
marrow and marrow cavity were numerically equal. As a result of this
difference in methodology, skeletal average S values for an active marrow
source and target calculated by Eckerman were determined to be 50%
lower than those calculated by Bouchet et al.. Recent research at the
University of Florida has explored these differences using three-dimensional
radiation transport techniques in trabecular bone based upon images
acquired through NMR microscopy [39], and based upon calculations for
the femur and humerus, has concluded that (i) the Eckerman model provides
acceptable estimates of AF(red marrow red marrow) at energies above
200 keV, (ii) the Bouchet et al. model provides acceptable estimates of
these same values at energies less than 20 keV and (iii) neither model
accurately predicts the absorbed fraction for self-irradiation of the active
marrow in the energy range 20–200 keV. Results from these two models
were recently combined using the University of Florida findings to generate
more accurate absorbed fractions of energy for the active marrow [40].
The second difference between these involves bone surface sources. For
a bone surface source, the Eckerman model uses a two-dimensional planar
surface source, selected at the interface between the trabeculae and
marrow cavities as recommended in ICRP 30 [32]. The Bouchet et al.
model assumes that the source is distributed throughout the 10 mm layer of
soft tissue (endosteum) on the bone interface. Consequently, the Eckerman
skeletal average S values for a ‘bone surface’ source and a marrow target
are 15% lower than those of Bouchet et al. for low-energy sources.
The third difference is associated with an endosteal target. Both models
assume the same 10 mm layer of soft tissue on the bone surface. However,
because the Eckerman model is a one-dimensional model, electrons are
assumed to enter this target with a uniform distribution of angles as
compared with the three-dimensional transport of Bouchet et al. that

assumes a uniform distribution of the cosine of the angle. The one-dimen-
sional assumption leads to a decrease in the pathlength through the endo-
steum. As a result of the differences associated with the modelling of the
surface source and endosteal target, calculated S values for a surface
source and an endosteal target are 60% lower in the MIRDOSE3/Eckerman
model than in the Bouchet et al. model for low-energy sources. Resolution of
these last two issues awaits acquisition and analysis of further data.
13.2.8. Model of a trabecular bone lesion

A different type of trabecular bone model was developed in 1995 by
Samaratunga et al. [41]. This group developed an anatomical model of
trabecular bone designed specifically to calculate the dose to skeletal
metastases from the surface seeker 186 Re. This model was based on
anatomical measurements of chord length distributions of cavities and
bone trabeculae on 25 samples from skeletal metastases. Bone trabeculae
were represented by ellipsoids located in an infinite marrow-tissue medium.
Finally, the transport of electrons was simulated in a three-dimensional
geometry using the Monte Carlo electron transport code EGS4. Even
though the procedure was very sophisticated, the model developed by
Samaratunga was intended for transport of electrons in trabecular bone
metastases, and therefore was not applicable for Reference Man tabulations.
13.2.9. High-resolution imaging for trabecular bone dosimetry

An alternative to chord-based transport models for trabecular bone dosimetry
is the use of three-dimensional digital images of these skeletal sites coupled
directly to Monte Carlo radiation transport codes. This technique offers
several distinct advantages over chord-based models. First, absorbed fractions
of energy are acquired within a skeletal site for which mass estimates also are
known. In chord-based models, only the absorbed fraction is obtained from
the transport calculations; independent assignments of target tissue mass are
required to report absorbed dose. Second, the full three-dimensional structure
of the anatomical site is available for particle tracking, enabling one to include
energy loss to the cortical bone cortex of the skeletal site for high-energy
sources. Most current chord-based transport models continue to utilize an
infinite trabecular-region transport geometry. Third, explicit partitioning
of the marrow source and target tissues into active (red marrow) and inactive
(yellow marrow) is accommodated within three-dimensional models of
trabecular bone, avoiding scaling approaches necessary within chord-based
models to account for marrow cellularity.
Three imaging techniques have been applied to the general study of
trabecular bone microarchitecture: nuclear magnetic resonance (NMR)
microscopy (resolution: 50–100 mm), micro-computed tomography (resolu-

tion: 20–50 mm) and synchrotron microtomography (resolution: <20 mm).
Jokisch et al. investigated the use of NMR microscopy to establish three-
dimensional electron transport models of trabecular bone [42, 43]. In these
studies, cadaveric sources of thoracic vertebral bodies were subjected to
marrow digestion, immersion in Gd-doped water, and imaging at 14.1 T
using a three-dimensional spin–echo pulse sequence. Electron transport with
the EGS4 code was conducted under a dual transport geometry: one utilizing
a segmented, voxel model of the bone trabeculae and marrow cavities, and one
utilizing a stylized representation of the trabecular spongiosa and cortex of
cortical bone. These studies were extended by Bolch et al. to include explicit
consideration of active marrow as both a source and target region [44]. This
study showed that, for nonspecific uptake of radioactivity in the active
marrow, marrow cellularity is an important parameter to consider in
patient-specific assessments of active marrow self-irradiation for high-energy
therapy radiopharmaceuticals. For low-energy emitters, reference values of
cellularity are sufficient for dose estimation.
In studies by Patton et al., energy loss to cortical bone was explored and
compared with results from infinite trabecular region transport [45]. This study
indicated that S value corrections on the order of 5–8% for 32 P and 8–11% for
90
Y were required to properly account for the finite size of skeletal site. In a
subsequent study, Patton et al. established that the overall methodology
employed in NMR microscopy and radiation transport was reproducible to
1–2% at energies below 1 MeV, and no more than 4% at 4 MeV [46].
Furthermore, the study established that nearly identical absorbed fraction
energy profiles could be obtained via either marrow-intact or marrow-free
NMR imaging, thus suggesting the potential for in vivo patient-specific
dosimetric models of the skeleton. This conclusion was supported by the
theoretical study by Rajon et al. [47] noting that image resolutions no finer
than 300 mm were sufficient for the development of three-dimensional
dosimetry models of trabecular bone for high-energy electron emitters
employed in radionuclide therapies.
13.3. CORTICAL BONE MODELS
13.3.1. Research at the University of Leeds

Similar to trabecular bone dosimetry, the foundation of cortical bone
dosimetry was established by F W Spiers at the University of Leeds. In
1976, Beddoe used the bone scanning microscope to measure transverse
chord lengths through Haversian cavities and the intervening cortical bone
matrix [6]. Initially developed for path-length measurements in trabecular
bone, this bone-scanning microscope was refined to measure the small
cavities in cortical bone. Stained sections of transverse slices of cortical
Cortical bone models 297
bone (20–30 mm thick) were scanned to obtain transverse chord length

distributions of the Haversian cavity sizes and cortical bone matrix, the
latter defined as the distances between Haversian cavities in the plane
perpendicular to the bone axis. With this optical scanning system, an effective
resolution of about 8 mm was achieved. Beddoe made measurements of the
cortex in three cortical bones: the humerus, tibia and femur. In his measure-
ments, Beddoe assumed that the Haversian cavities ran parallel to the long
axis of the bone, which is generally the case for the long bones. He also
assumed that the number of transverse canals (Volkmann canals) appearing
in the transverse planes was negligibly small. This last assumption was
verified visually on his stained sections of cortical bones. Beddoe included
measurements of the iliac crest, but his results showed irregular fluctuations
due to scanning only one small section.
For dosimetry calculations, Beddoe followed the technique employed by
Whitwell [26, 27] in the dosimetry of trabecular bone. Assuming a straight elec-
tron path through Haversian cavities and cortical bone matrix, a Monte Carlo
sampling technique was used to randomly select electron paths through the
Haversian canal and cortical bone matrix. The energy deposited in these
regions was determined using range–energy relationships in bone and soft
tissues, and was used to calculate dose conversion factors for several radionu-
clides. These dose conversion factors were only published in the dissertation of
Beddoe, and only for a limited number of radionuclides (14 C, 18 F, 22 Na, 32 P,
45
Ca, 90 Sr, 90 Y). Moreover, the only source considered in his study was the
bone volume, where both the endosteum and Haversian canals are of interest
for nuclear medicine and health physics dosimetric applications.
13.3.2. The ICRP-30 cortical bone model

In ICRP Publication 30, recommended absorbed fractions for cortical bone
are only given for the endosteum as target region and for the cortical surface
and bone volume as source regions [32]. For emitters originating on the
cortical surface, the Commission recommended an absorbed fraction to
the endosteum of 0.25 for mean energies below 200 keV, and an absorbed
fraction of 0.015 for mean energies greater than or equal to 200 keV. For
emitters originating in the cortical bone volume, an energy-independent
absorbed fraction of energy to the endosteum of 0.015 was recommended.
These absorbed fractions values were based on the dose conversion factors
of Beddoe, and were intended to be conservative, as their objective was to
provide a system of radiation protection for workers.
13.3.3. The Eckerman cortical bone model

Eckerman and Stabin [20] presented absorbed fraction for monoenergetic
electron sources distributed in the volume and on the surface of cortical
bone matrix. Details regarding these calculations were, however, not

provided in reference [20]. The absorbed fractions were based on a cylindrical
model of the Haversian canal located within an infinite cortical bone matrix.
The diameter of the Haversian canal was taken to be 40 mm and the energy
deposition within the 10 mm layer lining the canal was evaluated based on
the range–energy relationship (CSDA approximation).
13.3.4. The Bouchet et al. cortical bone model

In 1999, Bouchet et al. developed a three-dimensional transport model for
electrons in cortical bone [48] using the same methodology they used for
trabecular bone dosimetry [30]. They modelled the transport geometry
experienced by an electron travelling in the cortical bone using the transverse
chord length distributions through Haversian cavities and cortical bone
matrix measured by Beddoe. Because the chord length distributions were
measured within the plane perpendicular to the bone axis, each region was
represented by parallel portions of cylinders with radii sampled from the
chord length distributions. By sampling many different chords, for many
electrons, the average transport of an electron in cortical bone was simulated.
However, the escape of the particle from the cortical bone was not simulated
and, therefore, these cylinders were not limited by any planes along the
direction of their axes. Electron transport in the endosteum was performed
by creating a three-dimensional cylindrical half-space with radii selected
randomly from an entry angle in a 10 mm thick layer. The EGS4-PRESTA
Monte Carlo transport code was used to simulate the electron interaction
within this three-dimensional model, thus allowing for modelling of (a) the
non-linear electron trajectories, (b) full transport of the electrons with
-rays and bremsstrahlung, (c) the distribution of sizes of Haversian canals
and (d) cross-osteon irradiation. Absorbed fraction results for source and
target regions—the cortical endosteum, the Haversian canal or the cortical
bone matrix—were tabulated for the three cortical bone sites for which
Beddoe tabulated the chord length distributions. Results from these simula-
tions were combined with reference masses derived from ICRP Publication
70 for 22 skeletal sites and nine source–target tissue combinations within
the trabecular bone [38].
13.3.5. The research of Akabani

In 1993, Akabani developed an anatomical model [49] for the dosimetry of a
single Haversian canal, and derived dose factors for several -emitting radio-
nuclides (32 P, 45 Ca, 89 Sr, 90 Sr, 90 Y, 131 I, 153 Sm). This single Haversian canal
was modelled as a cylinder surrounded by an infinite cortical bone matrix.
Six Haversian canal sizes were individually considered with radii of 5, 10,
20, 30, 40 and 50 mm. The Monte Carlo transport code EGS4-PRESTA
Improving the models: patient-specific dosimetry 299
was used to derive electron absorbed fractions of energy for two sources
(Haversian canal and bone surface) and three targets (Haversian canal,
endosteum and the bone matrix). From these calculations, dose conversion
factors were tabulated for these seven radionuclides as a function of the
Haversian canal radius. Although elegant in its design, this anatomical
model only considered a single size for the Haversian canal, and it did not
consider the cross-osteon irradiation.
13.4. IMPROVING THE MODELS: PATIENT-SPECIFIC

DOSIMETRY
13.4.1. The clinical experience—ability to predict toxicity with current

dosimetric models
Nuclear medicine therapy is used increasingly in the treatment of cancer
(thyroid, leukemia/lymphoma with RIT, primary and secondary bone
malignancies, and neuroblastomas) and marrow toxicity generally limits
the amount of treatment that can be administered safely. The most difficult
problem in calculating bone marrow dose arises from the complex distribu-
tion of bone marrow in bone. The microanatomy of the marrow and its
diffuse intra-skeletal distribution make it difficult to calculate red marrow
dose accurately using only classical physical approaches, as discussed
above. Although stem cells and their primordial derivatives are presumed
to be the primary targets, the stroma that supports the red and yellow
marrow also plays an important role in facilitating marrow function, and
is an important secondary target [50, 51]. Dose to the bone marrow comes
mainly from activity within the marrow itself (specific binding to a cellular
component of either the marrow or blood), with the next highest contribu-
tion coming from tracer accumulation, when it occurs, in the surrounding
bone, with lesser contributions from more remote sources. Due to limited
spatial resolution, nuclear medicine imaging devices are unable to resolve
the source distribution in bone, and other methods are needed to calculate
the contributions from these sources.
Bone marrow toxicity is the major dose-limiting factor in RIT, but there
is no consensus on how to calculate that dose accurately, or of individual
patients’ ability to tolerate the planned therapy [8, 13, 52–59]. Experience
with external beam therapy has shown that there is a narrow margin between
the delivered dose that kills a tumour and that which causes serious injury to
the patient. Given that with modern imaging methods the dose from external
beam treatments can be accurately calculated, the remaining major issues are
tumour sensitivity and patient tolerance. In nuclear medicine, in contrast to
external beam radiotherapy, marrow absorbed dose may not be accurately
calculated, even when a treatment-planning dosimetry study is performed
[60]. Nonetheless, both external beam and internal emitter therapy share the
same uncertainty regarding predicted patient tolerance. Nuclear medicine
physicians in general have less experience in the therapy regime than those
who deal with external treatment approaches, and typically undertreat
patients in order to minimize the risk of even low-grade toxicity. If treatment
of cancer with radionuclides and RIT are to become a primary means of
treating cancer patients, the activity to be administered to individual patients
should be based on accurate evaluations of the dose to be delivered to the
tumour and of patient tolerance [61].
Radionuclide therapy has been most effective in the treatment of thyroid
cancer because of the high specific uptake of radioactive iodine by thyroid
tissue, but there is still no agreement on the means for choosing an optimal
therapeutic radiation dose to be delivered. The hope for RIT derives from the
notion that it can achieve selective high tumour uptake based on improved
targeting of tumour specific/associated antigens. Intravascular tumours,
including leukaemia and lymphoma, have the highest likelihood of successful
treatment with RIT, partly because tumour targeting efficiency is high, and
partly because of the high radiation sensitivity of lymphocytes, and to a
lesser extent of granulocytes. A major problem with RIT treatment of
most solid tumours has been low uptake per unit tumour mass, so that the
delivered dose is often too low for successful therapy. If one gives very
large doses intending to ablate marrow and tumour followed by marrow
transplantation, then lungs, liver or kidneys will become the dose limiting
organs. Bone marrow transplantation (BMT) is being employed routinely
in the treatment of breast cancer with high doses of chemotherapy and/or
of some solid tumours with RIT. A reasonable goal would be to develop
an improved means of defining an appropriate intermediate dose that
provides high ‘cure’/complete remission rates, without the need for BMT.
Progress is being made in the development of improved calculation
methods and tools for patient-specific treatment-planning focusing primarily
on tumour dose optimization [62–64]. Still, problems remain in the accurate
assessment of bone marrow dose and in the estimation of marrow reserve.
Better patient-specific knowledge of both dose and marrow reserve is
needed if one is to develop a good predictor of tolerance dose. Two analytical
approaches have been taken. The first is the physical method, in which one
measures the biokinetics and computes dose based on knowledge of the
amount injected, the radionuclide emission characteristics, and the measured
residence times in the different organs using accepted dose calculation
methods [12]. A complementary approach attempts to use patient-specific
characteristics to estimate residual damage from prior therapeutic treatments
to estimate the amount of added dose that would be tolerated by normal
organs, including bone marrow.
The importance of including pre-therapy patient status in toxicity
predictions has been studied by several groups [59, 65], and it is clear that
prior patient therapy and disease status must be factored into treatment
decisions. Nonetheless, to date poor correlations have been observed
between administered dose and marrow toxicity, which are somewhat
improved by more carefully calculated marrow doses [54]. This suggests
the need for more patient-specific information, including better dose
estimates and better information on other patient characteristics, including
perhaps patient-specific indicators of response (i.e., biomarkers).
Marrow toxicity from RIT is manifest by haematological changes in
circulating platelets, lymphocytes, granulocytes and reticulocytes, and later
changes are seen in red blood cells. Due to the high sensitivity of the
marrow, second organ toxicity is not a problem unless doses which are
known to be lethal to the marrow are intentionally given. Attempts to
correlate haematological toxicity with marrow dose, when marrow cells are
specifically targeted, have not been particularly successful in the past, in part
due to uncertainties in the actual absorbed dose, but also due to the
difficulty in assessing marrow functional status prior to therapy [8, 13, 53–
58]. The best correlations obtained to date have had a value of r around 0.8
[53, 57, 66, 67]. Whole-body absorbed dose and red marrow absorbed dose
are usually the best indicators of haematological toxicity, as measured by
platelet toxicity grade, with red marrow dose being slightly better. However,
the poor correlation coefficients obtained from these analyses emphasize the
need to better account for prior patient therapy and bone marrow reserve.
That red marrow absorbed dose is the best (albeit not appropriate on a
patient-by-patient basis) predictor of haematological toxicity has been
confirmed using data from other antibody trials [68]. Most recently, a
survey of the literature using data from different antibodies, different patient
populations and different disease sites has confirmed this observation over a
much larger scale [69]. Blumenthal et al. showed that plasma FLT3-L (a
stromal-cell produced positive stimulatory cytokine) levels predicted excess
platelet toxicity in patients receiving RIT after chemotherapy [70].
Many other studies have found little or no correlation between calculated
marrow dose for large numbers of patients and observed toxicity, as deter-
mined by various haematological parameters (platelet nadir, percentage
platelet drop, percentage lymphocyte drop, percentage granulocyte drop,
etc.). The primary reason for this is probably that all available marrow dose
models are patterned after a normal healthy young adult with good marrow
reserve, while the treated population is generally older (with inherently
different trabecular bone structures and less marrow reserve), often with
compromised marrow due to disease, prior chemotherapy or radiotherapy.
13.4.2. Correcting the models using patient-specific data

It is unreasonable to expect that doses calculated from a single model
representing an average healthy adult will yield good correlations of
radiation dose with observed effects in a heterogeneous patient population.

Many patients are older, with presumably less active marrow mass than
younger people, have more osteoporotic trabecular bones, and many have
disease or have had prior therapies which may affect marrow status and
trabecular bone micro-anatomy. Simple adjustments to standard model S
values may be made, assuming that patient marrow mass changes in some
predictable way with lean body mass, body surface area, or other easily
measured anatomic features. Such approximations will of course be
imperfect, but better than simply applying the standard model to all patients
equally. Some improvements in correlations between marrow dose and
observed marrow toxicity can be realized by calculating patient-specific
contributions to marrow dose and by making patient-specific adjustments
to relevant parameters such as marrow mass and cellularity (as implied by
patient age, disease, and prior therapeutic treatment) [54, 67].
More patient-specific evaluations can be envisioned, which could be
realized with some additional investigational effort. Functioning marrow
could be evaluated through the use of marrow scanning agents, such as
99m
Tc-sulfur colloid, 18 F-fluorothymidine [71], or other marrow avid
agents. If some quantitative measure of uptake in different regions can be
obtained, model-specific trabecular bone dose conversion factors [19, 20,
38] may be modified to be more specific to the patient. For example, skeletal
averaged S values in MIRDOSE3 [19, 20] were calculated from skeletal
averaged absorbed fractions:
X
AFðRM RMÞ ¼ CFbone fRM;bone AFðMS RMÞbone ð13:2Þ
bone
where CFbone is the cellularity factor for a given bone, fRM;bone is the
fraction of the red marrow (RM) associated with a given bone, and
AFðMS RMÞbone is the absorbed fraction for the energy originating in
the marrow of a given bone and irradiating the marrow space of that
bone. The values of fRM;bone and CFbone may be assigned based on infor-
mation gained in individual radionuclide marrow scans; that is, the fraction
of marrow calculated to be in each of the model regions and the cellularities
may be calculated on a patient-specific basis. In the absence of patient-
specific values of cellularity derived from marrow biopsies, values from
age-dependent models (e.g., [72, 73]) may be applied or, as will be discussed
in the next section, magnetic resonance imaging methods could also be used
to determine patient-specific active marrow mass and location [74–77]. Bone
marrow uptake of RIT agents in expected marrow sites which have
diminished or absent marrow function on radionuclide scans may be
assumed to be tumour, and the computed dose calculated using a more
specific model (e.g., [41]). Time–activity integrals may be calculated and
doses estimated for the different bone marrow regions in which marrow
localization has been scored, assigning the total marrow residence time to
each region using the values of fRM;bone . Whole marrow weighted average
dose and differential and integral dose–volume histograms then may be
calculated.
Determining marrow time–activity integrals for internal emitters is an
area of difficulty, as marrow activity is heterogeneous even in normal
subjects, and more so in patients with cancer, with or without a history of
previous treatment. Different imaging methods rely on either (i) the quanti-
fication of activity in a certain segment of the marrow (e.g., the lumbar spine
or sacrum) and extrapolation of total marrow activity assuming the total
amount of red marrow in the body and the fraction of the total that resides
in individual segments [13], or (ii) the assumption that the dose is derived
only from blood contributions, with the assumption of normal marrow
size [9] and/or (iii) combined blood and marrow-fixed activity methods.
The correlation between dose calculated in the different manners, biomarker
results, and clinical predictors with observed toxicity and patient cure rates
should be used to assess the potential utility of the different dose calculation
and marrow indices for improving treatment outcome. This is not currently
in standard practice, but should be considered if we are to seek more
meaningful dose and risk assessment in individual patients, as is routine in
external beam therapy.
13.4.3. In vivo high-resolution imaging

Accurate dose calculations in external beam radiotherapy treatments have
been made possible with routine use of three-dimensional anatomical
images (CT/MR). For skeletal dosimetry, patient-specific high-resolution
three-dimensional images need to provide information on the microstructure
of the trabecular bone, on the distribution of the red marrow target and
associated sensitive tissues, and on the distribution of the radionuclide.
The use of three-dimensional anatomical images in the determination of
trabecular bone microstructure is limited by the image resolution. In order to
limit partial volume effect, the voxel size has to be smaller than the structure
to be imaged. In 2000, Rajon et al. presented a theoretical study of the varia-
tions of the absorbed dose (S values) as a function of image resolution, and
determined that voxel sizes of 300–400 mm were sufficient for accurate
marrow dose calculations (calculated relative error less than 5% for voxel
size < 400 mm) [47]. As previously described, resolution of 20–50 mm has
been achieved with NMR microscopy to determine the microstructures of
trabecular bone from cadaveric bone samples. Modern high-field clinical
MR scanners (1.5–3 T) also have been used to provide in vivo high-resolution
images of the distal radius and calcaneus of osteoporotic patients with in-
plane resolution of 150 mm and slice thickness 300 mm [78]. Although
sufficient for dosimetry of trabecular bone, this high resolution is achieved
using special coils with a very small field of view, therefore limiting their
applications to body extremities (hands and feet) which are of no direct

interest for dosimetry. With the development of fast helical CT scanning
using multislices, CT imaging may become ideal for in vivo determination
of trabecular bone microstructures due to its high spatial integrity and its
higher contrast between bone and soft tissue. Current state-of-the-art
clinical CT scanners allow in-plane image resolution of 200 mm with
500 mm slice thickness to be acquired over a 20 cm length in less than 2
minutes (Rajon et al. [47] estimated that a 500 mm voxel size would lead to
a dose calculation error < 7% for a marrow source and target). Although
the number of detectors, their size and their efficiency currently limit the
acquisition time and the image resolution, Saito and Artawara recently
reported the development of a real-time three-dimensional CT scanner
using 256 detectors [79].
As previously discussed, patient-specific determination of marrow
cellularity and marrow reserve is important to improve bone dosimetry
calculations. In most dosimetric models, a reference marrow cellularity has
been used to convert absorbed fractions of energy calculated from a
Monte Carlo transport code to S values [19, 20, 38]. Several investigators
have developed techniques to measure in vivo distribution of marrow
cellularity in trabecular bone using clinical MR scanners and have correlated
in vivo measurements to bone biopsies [74–77, 80, 81]. Although these
techniques have been used to assess the composition of bone marrow prior
to and after bone marrow transplant [82], they have not yet been applied
in dosimetry. It is expected that patient-specific knowledge of bone
marrow cellularity throughout trabecular bone would increase prediction
of bone marrow toxicity, and allow optimal selection of the amount of
therapeutic radiation dose to be delivered.
13.5. SUMMARY
Skeletal dosimetry is complicated by the intricate nature of its microstructure

and by the irregular distribution of the radiosensitive red marrow tissue, both
of which can vary greatly between patients and throughout various bone
regions. Accurate patient-specific skeletal dose calculations are a primary
concern in radioimmunotherapy and radionuclide bone pain palliation as
the red marrow is usually the dose-limiting tissue. Significant improvements
have been made in the use of skeletal dosimetric models for bone marrow
dose calculations, mainly from the development of better models (more
appropriate for therapeutic nuclear medicine dosimetry) and a better under-
standing of the models and how to use them. Correlations of radiation
toxicity to marrow dose, however, remain weak, and will only improve as
more patient-specific dosimetric models are developed and used. The increase
in the speed of digital computers has recently permitted the coupling of
References 305
high-resolution three-dimensional digital images of the skeletal micro-

structures acquired in vitro to a Monte Carlo transport code. New imaging
techniques also allow the in vivo distribution of marrow cellularity in
trabecular bone to be determined. Application of these and other techniques
in vivo will allow improvement in the quality of patient-specific marrow dose
calculations for a better prediction of bone marrow toxicity and an optimal
selection of the amount of therapeutic radiation dose to be delivered.
REFERENCES
[1] Gray H 1974 Gray’s Anatomy (Philadelphia, PA: Running)

[2] ICRP 1995 Basic anatomical and physiological data for use in radiological protection:
the skeleton ICRP Publication 70, International Commission on Radiological
Protection (Oxford: Pergamon)
[3] Rowland R E, Stehney A F and Lucas H F 1978 Dose–response relationships for
female radium dial workers Radiat. Res. 76 368–83
[4] Littman M S, Kirsh I E and Keane A T 1978 Radium-induced malignant-tumors of
mastoid and para-nasal sinuses Am. J. Roentgenol. 131 773–85
[5] Rowland R E, Stehney A F and Lucas H F 1983 Dose–response relationships for
radium-induced bone sarcomas Health Phys. 44 15–31
[6] Beddoe A H 1976 The Microstructure of Mammalian Bone in Relation to the
Dosimetry of Bone-Seeking Radionuclides (Leeds: University of Leeds)
[7] Lewington V J 1996 Cancer therapy using bone-seeking isotopes Phys. Med. Biol. 41
2027–42
[8] Siegel J A, Wessels B W, Watson E E, Stabin M, Vriesendorp H M, Bradley E W,
Badger C C, Brill A B, Kwok C, Stickney D R, Eckerman K F, Fisher D R,
Buchsbaum D J and Order S E 1990 Bone marrow dosimetry and toxicity for
radioimmunotherapy Ab. Immunconj. Radiopharm. 3 213–33
[9] Sgouros G 1993 Bone marrow dosimetry for radioimmunotherapy: theoretical
considerations J. Nucl. Med. 34 689–94
[10] Sgouros G, Stabin M, Erdi Y, Akabani G, Kwok C, Brill A B and Wessels B 2000 Red
marrow dosimetry for radiolabeled antibodies that bind to marrow, bone, or blood
components Med. Phys. 27 2150–64
[11] Shen S, DeNardo G L, Sgouros G, O’Donnell R T and DeNardo S J 1999 Practical
determination of patient-specific marrow dose using radioactivity concentration in
blood and body J. Nucl. Med. 40 2102–6
Calculations (New York: Society of Nuclear Medicine)
[13] Siegel J A, Lee R E, Pawlyk D A, Horowitz J A, Sharkey R M and Goldenberg D M
1989 Sacral scintigraphy for bone marrow dosimetry in radioimmunotherapy Int. J.
Radiat. Appl. Instrum. B 16 553–9
‘S’, absorbed dose per unit cumulated activity for selected radionuclides and organs
Estimates of specific absorbed fractions for photon sources uniformly distributed
in various organs of a heterogeneous phantom (New York: Society of Nuclear

Medicine)
from internal photon sources ORNL/TM-8381 (Oak Ridge, TN: Oak Ridge National
Laboratory)
[17] Clairand I, Bouchet L G and Bolch W E 2000 A new macroscopic model of the long
bones for skeletal dosimetry J. Nucl. Med. 41 1062
[18] Snyder W S, Ford M R, Warner G G and Watson S B 1974 A tabulation of dose
equivalent per microcurie-day for source and target organs of an adult for various
radionuclides ORNL-5000 (Oak Ridge, TN: Oak Ridge National Laboratory)
[19] Eckerman K F 1985 Aspects of the dosimetry of radionuclides within the skeleton
with particular emphasis on the active marrow. In Proceedings of the Fourth
International Radiopharmaceutical Dosimetry Symposium eds A T Schlafke-Stelson
and E E Watson (Oak Ridge, TN: ORAU) pp 514–34
[20] Eckerman K F and Stabin M 2000 Electron absorbed fractions and dose conversion
factors for marrow and bone by skeletal regions Health Phys. 78 199–214
[22] Bouchet L G, Clairand I and Bolch W E 2000 Improvement of skeletal dosimetry
associated with photon sources J. Nucl. Med. 41 1049
[23] Beddoe A H, Darley P J and Spiers F W 1976 Measurements of trabecular bone
structure in man Phys. Med. Biol. 21 589–607
[24] Darley P J 1968 Measurement of linear path length distributions in bone and bone
marrow using a scanning technique. In Symposium on Microdosimetry EAEC
Report EUR 3747 Ispra, Italy pp 509–26
[25] Darley P J 1972 An Investigation of the Structure of Trabecular Bone in Relation
to the Radiation Dosimetry of Bone-Seeking Radionuclides (Leeds: University of
Leeds)
[26] Whitwell J R 1973 Theoretical Investigations of Energy Loss by Ionizing Particles in
Bone (Leeds: University of Leeds)
[27] Whitwell J R and Spiers F W 1976 Calculated beta-ray dose factors for trabecular
bone Phys. Med. Biol. 21 16–38
[28] Spiers F W 1949 The influence of energy absorption and electron range on dosage in
irradiated bone Br. J. Radiol. 22 521–33
[29] Spiers F W 1966 A review of the theoretical and experimental methods of determining
radiation dose in bone Br. J. Radiol. 39 216–21
[30] Bouchet L G, Jokisch D W and Bolch W E 1999 A three-dimensional transport model
for determining absorbed fractions of energy for electrons within trabecular bone J.
Nucl. Med. 40 1947–66
[31] Howell R W, Wessels B W, Loevinger R, Watson E E, Bolch W E, Brill A B, Charkes
N D, Fisher D R, Hays M T, Robertson J S, Siegel J A and Thomas S R 1999 The
MIRD perspective 1999 J. Nucl. Med. 40 3S–10S
[32] ICRP 1979 Limits for intakes of radionuclides by workers Publication 30, Part 1.
[33] Chen H T and Poston J W 1982 Estimates of absorbed energy in trabecular bone due
to beta-particles or electrons Health Phys. 43 647–53
[34] Zuzarte de Mendonca A 1993 Trabecular Bone Dosimetry Using a Monte Carlo Code
(College Station, TX: Texas A&M University)
References 307
[35] Nelson W R, Hirayama R H and Roger D W O 1985 The EGS4 code system SLAC
Report 265 (Stanford, CA: Stanford Linear Accelerator Center)
[36] Parry R A 1995 Revised Estimates of Electron Absorbed Fractions and Radionuclide S-
Values in Trabecular Bone (College Station, TX: Texas A&M University)
[37] Bielajew A F and Rogers D W O 1987 PRESTA, the ‘parameter reduced electron
step transport algorithm’ for electron Monte Carlo transport Nucl. Instrum. Meth. B
18 165–81
[39] Patton P 2000 NMR Microscopy for Skeletal Dosimetry: an Investigation of Marrow
Cellularity on Dose Estimates (Gainesville, FL: University of Florida)
[40] Stabin M, Eckerman K F, Bolch W E, Bouchet L G and Patton P W 2002 Evolution
and status of bone and marrow dose models Cancer Biother. Radiopharmaceut. In press
[41] Samaratunga R C, Thomas S R, Hinnefeld J D, Von Kuster L C, Hyams D M,
Moulton J S, Sperling M I and Maxon H R 1995 A Monte Carlo simulation
model for radiation dose to metastatic skeletal tumor from rhenium-186(Sn)-
HEDP J. Nucl. Med. 36 336–50
[42] Jokisch D W, Patton P W, Inglis B A, Bouchet L G, Rajon D A, Rifkin J and Bolch
W E 1998 NMR microscopy of trabecular bone and its role in skeletal dosimetry
[43] Jokisch D W, Bouchet L G, Patton P W, Rajon D A and Bolch W E 2001 Beta-
particle dosimetry of the trabecular skeleton using Monte Carlo transport within
3D digital images Med. Phys. 28 1505–18
[44] Bolch W E, Patton P W, Rajon D A, Shah A P, Jokisch D W and Inglis B A 2002
Considerations of marrow cellularity in 3-dimensional dosimetric models of the
trabecular skeleton J. Nucl. Med. 43 97–108
[45] Patton P W, Rajon D A, Shah A P, Jokisch D W, Inglis B A and Bolch W E 2002 Site-
specific variability in trabecular bone dosimetry: considerations of energy loss to
cortical bone Med. Phys. 29 6–14
[46] Patton P W, Jokisch D W, Rajon D A, Shah A P, Myers S L, Inglis B A and Bolch
W E 2002 Skeletal dosimetry via NMR microscopy: investigations of sample
reproducibility and signal source Health Phys. 82 316–26
[47] Rajon D A, Jokisch D W, Patton P W, Shah A P and Bolch W E 2000 Voxel size
effects in three-dimensional nuclear magnetic resonance microscopy performed for
trabecular bone dosimetry Med. Phys. 27 2624–35
[48] Bouchet L G and Bolch W E 1999 A three-dimensional transport model for deter-
mining absorbed fractions of energy for electrons within cortical bone J. Nucl. Med.
40 2115–24
[49] Akabani G 1993 Absorbed dose calculations in Haversian canals for several beta-
emitting radionuclides J. Nucl. Med. 34 1361–6
[50] Breems D A, Blokland E A, Neben S and Ploemacher R E 1994 Frequency analysis of
human primitive haematopoietic stem cell subsets using a cobblestone area forming
cell assay Leukemia 8 1095–104
[51] Neben S, Anklesaria P, Greenberger J and Mauch P 1993 Quantitation of murine
hematopoietic stem cells in vitro by limiting dilution analysis of cobblestone area
formation on a clonal stromal cell line Exp. Hematol. 21 438–43
[52] Bolch W E, Bouchet L G, Robertson J S, Wessels B W, Siegel J A, Howell R W,
Erdi A K, Aydogan B, Costes S, Watson E E, Brill A B, Charkes N D, Fisher D R,
Hays M T and Thomas S R 1999 MIRD pamphlet No 17: the dosimetry of

nonuniform activity distributions—radionuclide S values at the voxel level.
Medical Internal Radiation Dose Committee J. Nucl. Med. 40 11S–36S
[53] DeNardo D A, DeNardo G L, O’Donnell R T, Lim S M, Shen S, Yuan A and
DeNardo S J 1997 Imaging for improved prediction of myelotoxicity after radio-
immunotherapy Cancer 80 2558–66
[54] Lim S M, DeNardo G L, DeNardo D A, Shen S, Yuan A, O’Donnell R T and
DeNardo S J 1997 Prediction of myelotoxicity using radiation doses to marrow
from body, blood and marrow sources J. Nucl. Med. 38 1374–8
[55] Eary J F, Krohn K A, Press O W, Durack L and Bernstein I D 1997 Importance of
pre-treatment radiation absorbed dose estimation for radioimmunotherapy of non-
Hodgkin’s lymphoma Nucl. Med. Biol. 24 635–8
[56] Breitz H B, Fisher D R and Wessels B W 1998 Marrow toxicity and radiation
absorbed dose estimates from rhenium-186-labeled monoclonal antibody J. Nucl.
Med. 39 1746–51
[57] Behr T M, Sharkey R M, Juweid M E, Dunn R M, Vgg R C, Siegel J A
and Goldenberg D M 1996 Hematological toxicity in the radioimmunotherapy of
solid cancers with 131I-labeled anti-CEA NP-4 IgG1: dependence on red marrow
dosimetry and pretreatment. In Proceedings of the Sixth International Radiopharma-
ceutical Dosimetry Symposium Gatlinburg, TN, pp 113–25
[58] Juweid M E, Zhang C H, Blumenthal R D, Hajjar G, Sharkey R M and Goldenberg
D M 1999 Prediction of hematologic toxicity after radioimmunotherapy with (131)I-
labeled anticarcinoembryonic antigen monoclonal antibodies J. Nucl. Med. 40 1609–16
[59] Emami B, Lyman J, Brown A, Coia L, Goitein M, Munzenrider J E, Shank B, Solin
L J and Wesson M 1991 Tolerance of normal tissue to therapeutic irradiation Int. J.
Radiat. Oncol. Biol. Phys. 21 109–22
[60] Erdi A K, Erdi Y E, Yorke E D and Wessels B W 1996 Treatment planning for
radioimmunotherapy Phys. Med. Biol. 41 2009–26
[61] DeNardo G L, DeNardo S J, Macey D J, Shen S and Kroger L A 1994 Overview of
radiation myelotoxicity secondary to radioimmunotherapy using 131I-Lym-1 as a
model Cancer 73 1038–48
[62] Sgouros G, Barest G, Thekkumthala J, Chui C, Mohan R, Bigler R E and Zanzonico
P B 1990 Treatment planning for internal radionuclide therapy: three-dimensional
dosimetry for nonuniformly distributed radionuclides J. Nucl. Med. 31 1884–91
specific dosimetry Med. Phys. 23 1523–9
[65] DeNardo G L, Lamborn K R, DeNardo S J, Goldstein D S, Dolber-Smith E G,
Kroger L A, Larkin E C and Shen S 1995 Prognostic factors for radioimmunotherapy
in patients with B-lymphocytic malignancies Cancer Res. 55 5893s–98s
[66] Sgouros G, Jureidini I M, Scott A M, Graham M C, Larson S M and Scheinberg D A
1996 Bone marrow dosimetry: regional variability of marrow-localizing antibody
J. Nucl. Med. 37 695–8
[67] Wessels B, Breitz H B, Meredith R, Juweid M E, Pastor M, Colburn M and DeNardo
G L 2000 Bone marrow dosimetry model adjustments based on age, gender and prior
therapy J. Nucl. Med. 41 83P
References 309
[68] Sgouros G, Deland D, Loh A C, Divgi C R and Larson S M 1997 Marrow and whole-
body absorbed dose vs marrow toxicity following 131I-G250 antibody therapy in
patients with renal-cell carcinoma J. Nucl. Med. 38 252P
[69] O’Donoghue J A, Baidoo N and Sgouros G 1998 Calculated red marrow absorbed
dose predicts hematological toxicity in radioimmunotherapy J. Nucl. Med. 39 186P
[70] Blumenthal R D, Lew W, Juweid M, Alisauskas R, Ying Z and Goldenberg D M 2000
Plasma FLT3-L levels predict bone marrow recovery from myelosuppressive therapy
Cancer 88 333–43
[71] Shields A F, Grierson J R, Dohmen B M, Machulla H J, Stayanoff J C, Lawhorn-
Crews J M, Obradovich J E, Muzik O and Mangner T J 1998 Imaging
proliferation in vivo with [F-18]FLT and positron emission tomography Nat. Med.
4 1334–6
[72] Mauch P, Constine L, Greenberger J, Knospe W, Sullivan J, Liesveld J L and Deeg
H J 1995 Hematopoietic stem cell compartment: acute and late effects of radiation
therapy and chemotherapy Int. J. Radiat. Oncol. Biol. Phys. 31 1319–39
[73] Koo K H, Dussault R, Kaplan P, Kim R, Ahn I O, Christopher J, Song H R and
Wang G J 1998 Age-related marrow conversion in the proximal metaphysis of the
femur: evaluation with T1-weighted MR imaging Radiology 206 745–8
[74] Ballon D, Dyke J, Schwartz L H, Lis E, Schneider E, Lauto A and Jakubowski A
2000 Bone marrow segmentation in leukemia using diffusion and T2 weighted echo
planar magnetic resonance imaging NMR Biomed. 13 321–28
[75] Ballon D, Jakubowski A A, Tulipano P K, Graham M C, Schneider E, Aghazadeh B,
Chen Q S and Koutcher J A 1998 Quantitative assessment of bone marrow hemato-
poiesis using parametric magnetic resonance imaging Magn. Reson. Med. 39 789–800
[76] Ballon D, Jakubowski A A, Graham M C, Schneider E and Koutcher J A 1996
Spatial mapping of the percentage cellularity in human bone marrow using
magnetic resonance imaging Med. Phys. 23 243–50
[77] Ballon D, Jakubowski A, Gabrilove J, Graham M C, Zakowski M, Sheridan C and
Koutcher J A 1991 In vivo measurements of bone marrow cellularity using volume-
localized proton NMR spectroscopy Magn. Reson. Med. 19 85–95
[78] Link T M, Majumdar S, Grampp S, Guglielmi G, van Kuijk C, Imhof H, Glueer C
and Adams J E 1999 Imaging of trabecular bone structure in osteoporosis Eur.
Radiol. 9 1781–8
[79] Saito Y and Ortawara B S 2001 Development and evaluation of a real-time 3-
dimensional CT (4D-CT) scanner. In 87th Annual Meeting of the Radiological Society
of North America Chicago, IL
[80] Nonomura Y, Yasumoto M, Yoshimura R, Haraguchi K, Ito S, Akashi T and Ohashi
I 2001 Relationship between bone marrow cellularity and apparent diffusion coeffi-
cient J. Magn. Reson. Imaging 13 757–60
[81] Schick F, Einsele H, Kost R, Duda S, Jung W I, Lutz O and Claussen C D 1994
Hematopoietic reconstitution after bone marrow transplantation: assessment with
MR imaging and H-1 localized spectroscopy J. Magn. Reson. Imaging 4 71–8
[82] Schick F, Einsele H, Weiss B, Forster J, Lutz O, Kanz L and Claussen C D 1996
Assessment of the composition of bone marrow prior to and following autologous
BMT and PBSCT by magnetic resonance Ann. Hematol. 72 361–70
Chapter 14
Monte Carlo modelling of dose distributions

in intravascular radiation therapy
Michael Stabin and Dennis M Duggan
14.1. INTRODUCTION
The treatment of coronary artery disease, specifically to prevent restenosis

after percutaneous transluminal coronary angioplasty (PTCA) with high
doses of ionizing radiation, is a relatively new field of investigation. Monte
Carlo approaches are particularly well suited to the calculation of dose
distributions in this area. In this section, we will briefly review the technical
basis for this area of dose calculation and summarize current efforts to
calculate dose, including Monte Carlo applications.
Coronary arteries have diameters generally between 3 and 5 mm at their
origin, and then taper to smaller sizes [1]. Atherosclerotic plaque is formed
as fatty debris from blood accumulates in an artery, followed by the
proliferation of smooth muscle cells from the vessel’s internal elastic
membrane and media. Accumulation of collagen, elastin, and lipids
ensues, with the formation of the plaque, which consists of lipids, necrotic
cells and collagen. Over time, the core of the plaque necroses, calcifies and
haemorrhages. This causes a reduction in the blood flow (stenosis) within
the artery, with the threat of myocardial ischaemia or infarction. PTCA
can be performed with a number of techniques, to re-establish a stable
lumen with an acceptable inner diameter for appropriate blood flow to the
coronary muscle. Techniques include the use of balloons (to mechanically
expand the lumen), cutting and ablating devices (to remove the plaque)
and stent implantation (to stabilize the artery and provide structure).
Unfortunately, in about 35–40% of patients undergoing PTCA, the
coronary arteries restenose within a few months or years, and the benefits
of the original therapy are lost. The reasons for this are complex, but the
process seems to involve the following steps: fracture of the plaque from
balloon expansion occurs, followed by lining of the injury by platelets, and
310
Introduction 311
formation of a fibrous mesh of platelets and red blood cells. Monocytes then
phagocytize the mesh, and smooth muscle cells migrate into the intimal
subendothelial space, leading to neointimal formation. Over weeks, the
neointima becomes less cellular and the healing site resembles a plaque. In
about 40% of patients, the neointimal hyperplasia is excessive and results
in more narrowing of the lumen than was resolved by angioplasty [1].
Various preclinical and clinical studies have shown that the delivery of
between perhaps 15 and 50 Gy of ionizing radiation to the coronary artery
wall after balloon angioplasty can markedly reduce the rate of restenosis,
with few or no complications [1]. Due to the geometry, internal sources are
thought to be more efficacious than external beams of radiation, thus a
science of ‘intravascular brachytherapy’ (IVB) or ‘intravascular radionuclide
therapy’ is under development. This form of therapy principally takes two
approaches:
1. The use of temporary implants—seeds, wires and balloons with radio-
active walls or filled with radioactive solutions to deliver between 15
and 20 Gy to a 2–3 cm length of vessels between 2 and 5 cm in diameter.
The desired dose rate is about 5 Gy/min, to deliver the dose in a short
time and reduce complications from reduced blood flow. The technique
of afterloading with high dose rate sources is common.
2. The use of permanent implants—i.e., radioactive stents, that deliver a
similar dose, but with a lower dose rate.
The effectiveness of using high dose-rate intravascular radioisotope sources
for the inhibition of arterial restenosis has been demonstrated in a variety
of animal models and clinical trials [2]. This technology has been applied
primarily to the coronary arteries for the inhibition of intimal hyperplasia
following balloon angioplasty and/or stent placement. This neointimal
hyperplasia is considered to be pathologic, as the narrowing or restenosis
of a coronary artery can be life threatening [3, 4]. The current concept of
the mechanism of action is based on the inhibition of myofibroblast
stimulation in the adventitia by radiation [3]. Vessel wall irradiation with
particles has been demonstrated to prevent the onset of restenosis after
stent implantation or angioplasty in a porcine model [5-8].
Calculation of the radiation dose delivered to the coronary artery wall, as
well as to other organs and tissues in the body, is important to fully evaluate
the possible risks and benefits of this procedure. Computational methods for
these calculations are well developed and simply need to be implemented
appropriately. Both analytical and Monte Carlo methods are available and
have been used to calculate dose distributions around these sources. Monte
Carlo methods have the advantage of explicit treatment of the radiation
transport of electrons and photons at interfaces of materials of different
density and composition (e.g., at the interface of an atherosclerotic plaque
and soft tissue). The different approaches to the delivery of this radiation
312 Monte Carlo modelling of dose distributions
dose include the use of radioactive wires, balloons with radioactive walls or
filled with radioactive solutions, and stents. There are also numerous candidate
radionuclides, including and emitters, of low and high energy, which are
currently under investigation.
14.2. CANDIDATE RADIONUCLIDES AND TECHNOLOGIES
Treated lesions are generally 1–5 cm in length [1]. In conventional

brachytherapy, in interstitial applications, photons of 20–30 keV are
known to be efficacious in some sites such as the prostate, while in other
sites and in intracavitary applications, photons of >50 keV are needed.
Electron sources of high energy (e.g., 90 Sr/90 Y) are sometimes used to treat
superficial tumours (e.g., of the eye). For intravascular applications, it is
thought that photon energies need to exceed 20 keV and electron energies
need to be at least 1 MeV [1]. Much experience has been gained with the
use of 192 Ir in conventional brachytherapy, and its use is prevalent in studies
for IVB. 192 Ir has a mixture of photons between 300 and 500 keV and two s
of maximum energy 500–650 keV. Much of the photon energy will be
deposited away from the source, in the heart and other tissues of the patient,
as well as outside the patient (thus delivering a dose to the attending
physician and other staff). sources, such as 32 P and 90 Y (maximum ener-
gies around 1.7 and 2.2 MeV, respectively) offer the advantage of delivering
all of their energy locally, to the vascular wall and immediately surrounding
tissue. Other sources, notably 188 Re (two important s with maximum
energies between 1.9 and 2.2 MeV, with a photon of 155 keV at 15%) have
been suggested as well.
For temporary implants, calculations have shown that, for emitters,
between 10 and 30 GBq of activity (250–500 mCi) will be needed to attain
a dose rate of 4–5 Gy/min and thus maintain an acceptable dwell time.
Beta emitters can probably achieve the same goal with activities more like
0.5–1.5 GBq (14–38 mCi). With the higher activities of the emitters,
there are more radiation safety problems created for source handling,
shielding and storage, management of staff procedures, etc., while with the
sources the principal radiation safety concerns involve doses to the
extremities during source manipulation. If one is using a wire or seed type
source, it has been shown that problems of source centring are more
important for the emitters than for emitters [8] (see analysis below).
The issue of source centring is an important one; if the source is not well
centred, one side of the artery will receive a substantially higher dose than
the other, resulting in underdosing to one side (and perhaps suboptimal
therapy) and overdosing to the other (raising the risk of radiation-related
complications, e.g., direct tissue damage, induction of aneurisms). Centring
of these sources can be aided with specifically designed devices. Alternatively,
Radiation dosimetry studies 313
the use of liquid-filled balloons has been proposed, as this inherently solves
the problem of source uniformity. In this application, however, there is the
risk of balloon rupture and release of the radioactive liquid into the blood-
stream. This would presumably act like an intravenous injection of the
material, with its subsequent distribution, retention and radiation dosimetry
being determined by the radionuclide and chemical compound (if any) to
which it is attached. The use of a balloon with radioactive walls can give
the same uniformity with a much lower risk of releasing the radioactive
materials.
The use of radioactive stents has been proposed, to provide structure
and stability to the affected artery as well as delivering the radiation dose
to the vessel wall. In these applications, to date, only emitters (principally
32
P) have been proposed. The activity may be coated on the surface of the
stent (which raises the possibility of activity leaching over time into the
bloodstream) or impregnated into the metal of the stent (by a manufacturing
process or neutron activation of extant metal components). These stents have
the advantages that low activities (kBq to MBq range) can be employed (as
the dose is delivered over many hours or days as the radionuclide decays in
situ), and that the stent conforms in shape to the lesion site. Non-radioactive
stents, however, do not inhibit the process of restenosis; in fact, there is
evidence that they may even stimulate it [1].
These are not the only candidate technologies and nuclides available or
suggested. Many strategies and nuclides have been proposed, including low
energy X- or -ray emitters, the use of inserted X-ray generation devices,
and others. As many ideas are being generated and tested, one cannot
generalize about the future. Many factors, scientific, economic and logistic,
will influence the final suitability of different technologies and will determine
those that find the greatest acceptance.
14.3. RADIATION DOSIMETRY STUDIES
14.3.1. Vessel wall dose—calculations

For both electron and photon sources, calculation of radiation dose distribu-
tions may be performed by either analytical (point-kernel) techniques or
Monte Carlo transport simulations. Point-kernel techniques employ
lookup tables of dose–response values, usually based originally on some
Monte Carlo studies carried out for point or other simple sources in a
uniform infinite medium, and extend the results to more complex geometries
through numerical methods. In Monte Carlo studies, the problem geometry
is described, and the history of many elementary particles is simulated using
random sampling of known probability distributions (for emissions from an
isotropic source, cross sections for interactions, production of secondary
Figure 14.1. Isodose distributions, based on point-kernel calculations, for a

linear array of 192 Ir seeds (adapted from a study by Wiedermann et al [8],
as given in AAPM Report No 66 [1]. Reprinted with permission from AAPM.
particles, etc.). Point kernel methods generally apply to only a single medium
(generally water), but can be adapted to approximate particle behaviour in
other media [9,10]. Monte Carlo methods have the advantage of explicitly
treating particle behaviour in various media. For example, in this applica-
tion, a fluid in a filled balloon, the wall of the balloon, and the vessel wall
will all have similar compositions (water or soft tissue equivalent), but a
central guide wire or source wire will be different (metallic) and the plaque
will also be different (calcified tissue, with a composition and density signifi-
cantly different from that of soft tissue). Monte Carlo methods can have the
disadvantage of only being able to treat certain predefined geometries in
combinations (e.g., cylinders, spheres).
Most early attempts to characterize the dosimetry of various intra-
vascular sources employed point-kernel approaches, while later ones
employ more Monte Carlo applications. Figure 14.1 shows isodose distribu-
tions, based on point-kernel calculations, for a linear array of 192 Ir seeds
(adapted from a study by Wiedermann et al. [8], as given in AAPM
Report No 66 [1]). Prestwich et al. [11] used analytical kernels, which were
based ultimately on measurements, to calculate dose from a 32 P coated
stent. They showed one-dimensional radial dose distributions; two-dimen-
sional isodose contours and three-dimensional dose distributions for a
3 mm 20 mm cylindrical source. Li et al. [12] used MCNP to calculate
doses from a single strut of a stent with radioisotope (48 V, 32 P) distributed
throughout, then used superposition to calculate dose distributions for an
Figure 14.2. Dose rate distribution in a plane transverse to the axis of a 48 V

impregnated stent obtained by Li et al. Reprinted from [15] with permission
from AAPM.
entire stent containing 48 V and 32 P (figure 14.2). Janicki et al. [13] used a more
realistic geometry but still used an analytic kernel, and calculated three-
dimensional dose distributions in tissue around a 3:5 mm 20 mm wire
mesh stent. Later, Janicki et al. [9] extended this model to cases in which
the stent was surrounded by layers of materials with different compositions
and densities (figure 14.3). Duggan et al. [14] performed a calculation similar
to that of Prestwich et al. but with a Monte-Carlo generated kernel from
Simpkin et al. [15]. McLemore [16] also used MCNP to calculate dose
from a single strut with a 103 Pd coating on the outside and then used super-
position to calculate the dose from the entire stent.
Many complete sources, with realistic geometries and materials, have
now been simulated. These include entire radioactive stents [17]. Seed or
wire sources are usually encapsulated, often in high atomic number
materials. This can have a dramatic effect on the dose distribution, especially
for or low-energy photon sources. Monte Carlo simulation is the most
accurate technique for predicting the dose distribution for such sources.
Examples of simulations of encapsulated sources that are now in clinical
use include that of the Novoste BetaCath 90 Sr seed by Soares et al. [18]
and the Guidant encapsulated 32 P wire source by Mourtada et al. [19]
(figure 14.4). In addition, many other proposed sources have been simulated,
including a coil-shaped wire source [20] and a soft X-ray device [21].
As mentioned above, an important issue in the use of radioactive
wires or seeds is the centring of the source within the vessel. A study done
by Amols et al. [22] showed that centring is substantially more important
for sources than for photon sources (although it is clearly important for
both). Figure 14.5 shows that a centring error of 0.5 mm in a 5 mm diameter
Figure 14.3. Comparison of calculations with all water DPK model (W) and
multilayer DPK model (ML) with radiochromic film measurements (blue)
for a 32 P stent in a block of solid water with a layer of Teflon in between the
stent and the radiochromic film. Reprinted with permission from AAPM
(Janicki et al [9]).
vessel will result in an error of 1.6 for 192 Ir (i.e., the actual dose received on
the overdosed side will be 1.6 times that calculated for a centred source) and
2.1 for 90 Y (figure 14.5).
Extensive calculations, using both the MCNP 4B [23] and EGS4 [24]
Monte Carlo codes, were made for a number of fixed artery sizes and for
Figure 14.4. Comparison of Monte Carlo and measured values for the Guidant
encapsulated 32 P wire source [19]. Reprinted with permission from AAPM.
Figure 14.5. Dose asymmetry (ratio of maximum to minimum dose in vessel

wall) resulting from inaccurate centring of 5 mm long 192 Ir and 90 Sr sources
within 5 mm diameter artery obtained by Amols et al. Adapted from [22]
and reprinted with permission from ASTRO.
wire and liquid-filled balloon sources by Stabin et al. [25]. They performed
calculations for monoenergetic photons and electrons over a broad range
of energies, then applied decay spectra for over 800 radionuclides to the
monoenergetic results, and incorporated the data into an interactive PC
computer program that provided rapid comparisons of candidate radio-
nuclides and geometries. Their blood vessel models were simple cylinders
3 cm in length, half of whose inner circumference had an atherosclerotic
plaque and half which did not. Results were given for radial dose as a
function from the vessel wall (diameters of 1.5, 3 and 4.5 mm), and compar-
isons for some candidate nuclides were given (figure 14.6).
14.3.2. Vessel wall dose—measurements

Measurements are made using thin strips of film (GAF Chromic) or thin
TLD chips, to evaluate the doses over the short distances (from perhaps
0.02 to 10 mm) of interest in this application. Janicki et al. [9] made measure-
ments and performed point-kernel calculations for a Palmaz–Schatz stent
containing 32 P (figures 14.7–14.10).
14.3.3. Dose to body organs in the case of a balloon rupture

In the case of a balloon rupture, the radioactive material in the balloon will
enter the bloodstream, and may distribute to other organs, delivering an
Figure 14.6. Dose comparison for 186 Re and 188 Re balloons and an 192 Ir wire in
a 3 mm diameter vessel. In all cases, the activities were calculated so that the
vessel would receive a dose of 30 Gy in 5 min at 0.5 mm from the vessel surface
(comparisons performed on the non-plaque side of the vessel). The calculated
activities were 2.48 GBq of 186 Re, 1.02 GBq of 188 Re and 64.1 GBq of 192 Ir.
Figure 14.7. Diagram of solid water phantom used to expose radiochromic film
to the -emitting stent. Four stents were exposed. For two stents, a layer of
Teflon was placed on the stent before the radiochromic film was stacked on
top. Low and high refer to low- and high-sensitivity radiochromic film, respec-
tively. The measured thicknesses of the different layers were Teflon, 386 mm;
low-sensitivity radiochromic film, 108 mm; and high-sensitivity radiochromic
film, 257 mm. Reprinted with permission from Janicki et al [9].
Figure 14.8. Photograph of a stent mounted on a polymethyl-methacrylate

(PPMA) rod and inserted into a solid water block like the one shown in
figure 14.7. On the right is the radiochromic film exposed while placed directly,
emulsion side toward the stent, on top of the block. Reprinted with permission
from Janicki et al. [9].
unwanted radiation dose to these organs. A comparison was made of the

possible radiation dose distributions for four candidate radionuclides, 32 P,
90
Y, 186 Re and 188 Re (table 14.1). The pharmaceutical or carrier to which
the nuclide is attached will determine its behaviour in vivo; here the forms
assumed were 32 P phosphate, 90 Y in ionic form and the rhenium isotopes
as perrhenate. The kinetic model chosen for 32 P and 90 Y came from ICRP
Figure 14.9. Radiochromic film results. Reprinted with permission from Janicki
et al. [9].
Figure 14.10. Comparison of the radiochromic film results and the DPK multi-
layer (ML) model. Reprinted with permission from Janicki et al. [9].
Publication 30 [26]; the model for perrhenate was the MIRD Dose Estimate
Report No 8 model (Lathrop et al. [27]) for sodium pertechnetate in nonrest-
ing subjects, which was simply adapted for use with the rhenium isotopes, as
suggested by the data of Kotzerke et al. [28]. Organ doses for the four radio-
pharmaceuticals assuming release of unit quantities into the bloodstream are
shown in table 14.1. For 32 P and 90 Y, the organ receiving the highest dose is
red marrow with a dose of 8.12 and 3.26 Gy per GBq respectively. With 186 Re
and 188 Re, the organs receiving the highest dose are the large intestine,
thyroid and urinary bladder.
Calculations show that one would require 1.04 GBq of 32 P or 0.92 GBq
90
of Y in a balloon source to give 30 Gy in 5 min at 0.5 mm in a 3 mm
diameter artery. It is clear from these data and the dose estimates in the
above table that use of 32 P or 90 Y in the forms assumed here (liquid sources
of 32 P as phosphate or ionic 90 Y) will represent a significant risk to the
patient, as the predicted absorbed doses to red marrow from the quantities
of activity assumed are near or above median lethal doses. For the rhenium
isotopes, there are significant doses predicted to the intestines, but not at
levels that would be considered life threatening. Other strategies may
improve the radiation dose picture, such as the use of perchlorate to block
thyroid uptake, or attaching the rhenium isotopes to other chemical
compounds, such as MAG3, and to obtain different biological behaviour.
In these cases, obviously different biokinetic behaviour will result in different
dose estimates, which need to be calculated in each case to evaluate the
possible radiation doses.
Radiation effects—artery walls 321
Table 14.1. Organ doses from injection of radionuclides into the bloodstream
due to balloon rupture (Gy per GBq injected).
32 90 186 188
P Y Re Re
Adrenals 7:60 10ÿ1 9:07 10ÿ2 3:17 10ÿ2 4:45 10ÿ2

Brain 7:60 10ÿ1 9:07 10ÿ2 3:10 10ÿ2 4:35 10ÿ2
Breasts 7:60 10ÿ1 9:07 10ÿ2 3:09 10ÿ2 4:34 10ÿ2
Gallbladder wall 7:60 10ÿ1 9:07 10ÿ2 3:38 10ÿ2 4:66 10ÿ2
LLI wall 7:60 10ÿ1 9:07 10ÿ2 1:95 100 1:97 100
Small intestine 7:60 10ÿ1 9:07 10ÿ2 3:66 10ÿ2 4:97 10ÿ2
Stomach 7:60 10ÿ1 9:07 10ÿ2 2:64 10ÿ1 4:33 10ÿ1
ULI wall 7:60 10ÿ1 9:07 10ÿ2 1:95 100 1:98 100
Heart wall 7:60 10ÿ1 9:07 10ÿ2 3:14 10ÿ2 4:42 10ÿ2
Kidneys 7:60 10ÿ1 9:07 10ÿ2 3:20 10ÿ2 4:48 10ÿ2
Liver 7:60 10ÿ1 3:89 100 3:18 10ÿ2 4:46 10ÿ2
Lungs 7:60 10ÿ1 9:07 10ÿ2 3:13 10ÿ2 4:38 10ÿ2
Muscle 7:60 10ÿ1 9:07 10ÿ2 3:18 10ÿ2 4:45 10ÿ2
Ovaries 7:60 10ÿ1 9:07 10ÿ2 3:67 10ÿ2 4:99 10ÿ2
Pancreas 7:60 10ÿ1 9:07 10ÿ2 3:24 10ÿ2 4:55 10ÿ2
Red marrow 8:12 100 3:26 100 3:19 10ÿ2 4:48 10ÿ2
Bone surfaces 1:04 101 3:91 100 3:32 10ÿ2 4:52 10ÿ2
Skin 7:60 10ÿ1 9:07 10ÿ2 3:10 10ÿ2 4:35 10ÿ2
Spleen 7:60 10ÿ1 9:07 10ÿ2 3:19 10ÿ2 4:48 10ÿ2
Testes 7:60 10ÿ1 9:07 10ÿ2 3:18 10ÿ2 4:47 10ÿ2
Thymus 7:60 10ÿ1 9:07 10ÿ2 3:12 10ÿ2 4:38 10ÿ2
Thyroid 7:60 10ÿ1 9:07 10ÿ2 6:06 10ÿ1 1:10 100
Urinary bladder wall 1:91 100 8:02 10ÿ1 1:04 100 1:51 100
Uterus 7:60 10ÿ1 9:07 10ÿ2 3:52 10ÿ2 4:87 10ÿ2
Total body 1:52 100 5:21 10ÿ1 4:37 10ÿ2 5:74 10ÿ2
32 90 186 188
Chemical forms assumed: P as sodium phosphate, Y in ionic form, Re and Re as
perrhenate.
14.4. RADIATION EFFECTS—ARTERY WALLS
As this is a new technology and some of the effects we might expect are
expressed over longer times, there will be a delay before we know what
complications to expect. However, there is some clinical experience with
arterial damage and complications from external beam radiation. Acute
and chronic morphological changes have been seen in femoral arteries in
dogs given 40 Gy in 10 days [29]. In one radiation therapy centre, 20 patients
had blood vessel-associated injuries associated with radiation therapy over a
20 year period, including arterial rupture and occlusion, over a 7–24 year
period after therapy [30]. To effectively employ this promising technology,
we must make use of the best information that we can obtain from
radiation dose calculations and measurements, and proceed with care in

patient studies.
REFERENCES
[1] Task Group 60 of the American Association of Physicists in Medicine 1999 AAPM
Report No 66, Intravascular brachytherapy physics Med. Phys. 26 119–52
[2] Waksman R 1999 Vascular Brachytherapy 2nd edition (Armonk, NY: Futura)
[3] Wilcox J, Waksman R, King S and Scott N 1996 The role of the adventitia in the
arterial response to angioplasty: the effect of intravascular radiation Int. J. Radiat.
[4] Shi Y, Pieniek M, Fard A, O’Brien J, Mannion J D and Zalewski A 1996 Adventitial
remodeling after coronary arterial injury Circulation 93 340–8
[5] Waksman R, Robinson K, Croker I, Gravanis M B, Palmer S J, Wang C, Cipolla G D
and King S B III 1995 Intracoronary radiation before stent implantation inhibits
neointima formation in stented porcine coronary arteries Circulation 92 1383–6
[6] Waksman R, Robinson K, Croker I, Wang C, Gravanis M B, Cipolla G D, Hillstead
R A and King S B III 1995 Intracoronary low-dose beta-irradiation inhibits
neointima formation after coronary artery balloon injury in the swine restenosis
model Circulation 92 3025–31
[7] Waksman R, Robinson K, Croker I, Gravanis M B, Cipolla G D and King S B III
1995 Endovascular low-dose irradiation inhibits neointima formation after coronary
artery balloon injury in swine Circulation 91 1533–9
[8] Wiedermann J, Marboe C, Amols H, Schwartz A and Weinberger J 1994 Intra-
coronary irradiation markedly reduces restenosis after balloon angioplasty in a
porcine model J. Am. Coll. Cardiol. 23 1491–8
[9] Janicki C, Duggan D M, Gonzalez A, Coffey C W II and Rahdert D A 1999 Dose
model for a beta-emitting stent in a realistic artery consisting of soft tissue and
plaque Med. Phys. 26 2451–60
[10] Janicki C, Duggan D M and Rahdert D A 2001 A dose-point-kernel (DPK) model
for a low energy gamma-emitting stent in an heterogeneous medium Med. Phys. 28
1397–405
[11] Prestwich W V, Kennett T J and Kus F W 1995 The dose distribution produced by a
32P-coated stent Med. Phys. 22 313–20
[12] Li A N, Eigler N L, Litvack F and Whiting J S 1998 Characterization of a positron
emitting V48 nitinol stent for intracoronary brachytherapy Med. Phys. 25 20–8
[13] Janicki C, Duggan D M, Coffey C W, Fischell D R and Fishchell T A 1997
Radiation dose from a phosphorus-32 impregnated wire mesh vascular stent Med.
Phys. 24 437–45
[14] Duggan D M, Coffey C W and Levit S 1998 Dose distribution for a 32P-impregnated
coronary stent: comparison of theoretical calculations and measurements with radio-
chromic film Int. J. Radiat. Oncol. Biol. Phys. 40 713–20
[15] Simpkin D J and Mackie T R 1990 EGS4 Monte Carlo determination of the betadose
kernel in water Med. Phys. 17 179–86
[16] McLemore L 2000 Dosimetry characterization of a palladium 103 implanted stent for
intravascular brachytherapy, Master’s Thesis, University of Texas
References 323
[17] Reynaert N F, Verhaegen F, Taeymans Y, Van Eijkeren M and Thierens H 1999

Monte Carlo calculations of dose distributions around 32 P and 198 Au stents for
intravascular brachytherapy Med. Phys. 26 1484–91
[18] Soares C G, Halpern D G and Wang C 1998 Calibration and characterization of
betaparticle sources for intravascular brachytherapy Med. Phys. 25 339–46
[19] Mourtada F A, Soares C G, Seltzer S M and Lott S H 2000 Dosimetry characteriza-
tion of 32 P catheter based vascular brachytherapy source wire Med. Phys. 27 1770–6
[20] Hafeli U O, Roberts W K, Meier D S, Ciezki J P, Pauer G J, Lee E J and Weinhous M S
2000 Dosimetry of a W-188/Re-188 beta line source for endovascular brachytherapy
Med. Phys. 27 668–75
[21] Gierga D P and Shefer R E 2001 Characterization of a soft X-ray source for intra-
vascular radiation therapy Int. J. Radiat. Oncol. Biol. Phys. 49 847–56
[22] Amols H, Zaider M, Weinberger J, Ennis R, Schiff P and Reinstein L 1996 Dosimetric
considerations for catheter-based beta and gamma emitters in the therapy of neointimal
hyperplasia in human coronary arteries Int. J. Radiat. Oncol. Biol. Phys. 36 913–21
[23] Briesmeister J 1993 MCNP—A general Monte Carlo n-particle transport code. MCNP
User’s Manual Los Alamos National Laboratory
[24] Bielajew A and Rogers D 1987 PRESTA: the parameter reduced electron-step trans-
port algorithm for electron Monte Carlo transport Nucl. Instrum. Meth. B18 165–81
[25] Stabin M, Konijnenberg M, Knapp Jr F F and Spencer R H 2000 Monte Carlo
modeling of radiation dose distributions in intravascular radiation therapy Med.
Phys. 27 1086–92
[26] International Commission on Radiological Protection. Limits for Intakes of Radio-
nuclides by Workers 1979 ICRP Publication 30 (New York: Pergamon)
[27] Lathrop K, Atkins H, Berman M, Hays M and Smith E 1976 MIRD Report No 8:
Summary of current radiation dose estimates to normal humans from 99mTc as
sodium pertechnetate J. Nucl. Med. 17 74–77
[28] Kotzerke J, Fenchel S, Guhlmann A, Stabin M, Rentschler M, Knapp F Jr and
Roeske S 1998 Pharmacokinetics of 99Tcm-pertechnetate and 188Re-perrhenate after
oral administration of perchlorate: Option for subsequent care after the use of
liquid 188Re in a balloon catheter Nucl. Med. Comm. 19 795–801
[29] Fonkalsrud E W, Sanchez M, Zerubavel R and Mahoney A 1977 Serial changes in
arterial structure following radiation therapy Surg. Gynecol. Obstet. 145 395–400
[30] McCready R A, Hyde G L, Bivins B A, Mattingly S S and Griffen W O Jr 1983
Radiation-induced arterial injuries Surgery 93 306–12
Chapter 15
The Monte Carlo method as a design tool in

boron neutron capture synovectomy
David Gierga, Jacquelyn Yanch and Ruth Shefer
15.1. INTRODUCTION
Boron neutron capture synovectomy (BNCS) is a potential application of the

10
B(n,)7 Li reaction for the treatment of rheumatoid arthritis. Rheumatoid
arthritis is a chronic disease characterized by painful inflammation of the
membrane (the synovium) lining the inner joint capsule of articulating
joints. If left unchecked, synovial inflammation can lead to cartilage destruc-
tion, bone erosion and joint disability [1]. BNCS is envisioned as a two-part
procedure involving, first, the injection of a 10 B-labelled compound directly
into the joint, and second, irradiation of the joint with a beam of epithermal
neutrons. The resulting 10 B(n,)7 Li reaction imparts highly localized energy
to regions selectively loaded with 10 B. The goal of BNCS is thus to deliver
enough energy to ablate the synovium, thereby relieving the symptoms of
rheumatoid arthritis.
The boron neutron capture reaction is also under investigation as a
means of treating intracranial tumours and metastatic melanoma of the
periphery by many different groups worldwide [2–5]. A number of Phase I
and Phase I/Phase II clinical trials of boron neutron capture therapy
(BNCT) are currently underway or have recently been completed. However,
while both BNCT and BNCS rely on the same nuclear reaction to impart
local dose to diseased cells, there are significant differences in the two clinical
approaches that necessitate the design of neutron beams specific to each
application. To date, published investigations of the development of
BNCS have been limited to our group; hence this chapter will focus on the
Monte Carlo-based neutron beam design that we have performed. Specifi-
cally, we will present both our strategy for the design of clinically relevant
neutron beams for BNCS and the results of this design process. The
optimized neutron delivery system has been constructed; this assembly will
324
Introduction 325
be described and results of experimentally verifying the simulation predic-

tions will be presented. The chapter will conclude with a discussion of
Monte Carlo and experimental determination of patient dose and optimal
patient shielding configurations.
15.1.1. Rheumatoid arthritis and current methods of treatment

Rheumatoid arthritis (RA) is a chronic, systemic disease of unknown origin
characterized by painful inflammation of articular joints. The incidence in
women is about 2–3 times higher than in men. RA occurs in all racial and
ethnic groups, and affects approximately 1% of the worldwide adult
population. RA can occur at all ages, but 80% of RA patients experience
the onset of the disease between the ages of 35 and 50 [1]. Patients with
RA can experience considerable distress as a result of joint pain and
significantly restricted mobility. In addition to these debilitating conditions,
those afflicted with RA have a shorter median life expectancy compared with
the general population [6].
The first clinical signs of the disease are usually fatigue, fever and
morning joint stiffness. Affected joints eventually become enlarged, tender
and painful to move. RA initially involves small joints such as hands,
wrists, knees and feet. As the disease progresses, elbows, shoulders, hips
and ankles may be affected. The first manifestation of RA is a proliferation
of synovial cells (synovitis), although the cause of the initial inflammatory
response is unknown. During the onset of RA, synovial cells start to
proliferate, increasing in number and in size. The amount of synovial fluid
generated by the synovial cells also increases. As a result of synovial
proliferation, a vascularized tissue (the pannus) forms and attacks the
articular cartilage and bone, which if left unchecked can lead to cartilage
destruction, bone erosion and joint disability.
The primary treatment for RA consists of using various drugs
administered to reduce synovial inflammation. In approximately 10% of
patients, however, one or more joints remain unresponsive to drug
treatment and surgical management may be necessary. For patients who
have suffered severe cartilage and bone damage and suffer from significant
joint dysfunction, the only alternative is total joint replacement. The long-
term durability of the total knee replacement is a concern. Subsequent to
surgery, a 6 to 8 week period of physical therapy is required to allow the
joint to become weight-bearing. Recovery from the surgery could also be
complicated if the patient has multiple joints with RA. The cost of the
procedure is also significant, ranging from $25 000 to $30 000 as of 1990
[7]. For these reasons, despite the high success rate of total knee replace-
ment surgery, the procedure is only advised when there is severe joint pain
from cartilage destruction and when the joint has not responded to drug
therapy [7].
326 The Monte Carlo method as a design tool
For patients who experience continued pain and inflammation from

synovitis, an alternative to total knee replacement is available to preclude
or slow the progression of cartilage or bone damage. This option is termed
synovectomy, which is the physical removal of the inflamed synovium
from the joint. The removal of this inflamed tissue leads to a reduction in
joint swelling and pain, with the aim of providing better joint function.
Synovectomy was initially performed as an open surgical procedure,
although this has been replaced by arthroscopic synovectomy in recent
years. Open surgical synovectomy has a number of disadvantages, including
the potential for haemorrhage, the risks of anaesthesia, a long recovery
period (3–6 weeks), unpredictable efficacy, and the ability to only remove
about 80% of the synovium. Arthroscopic synovectomy is a much simpler
procedure than open synovectomy, with post-operative hospital stays
averaging only 3 days.
In many countries, radiation synovectomy has been used as an
alternative to surgical synovectomy. Radiation synovectomy involves the
intra-articular injection of a radionuclide (usually a emitter), and relies
on the energy deposition of the particles to ablate the inflamed synovium.
Radiation synovectomy is a much simpler procedure than both open and
arthroscopic surgery, with only a local anaethesia necessary, and the
elimination of any rehabilitation time that would follow surgery. The most
frequently used isotope is 90 Y, which has a 2.7 day half life, and a maximum
energy of 2.2 MeV. Radiation synovectomy using 90 Y (in the form of a
silicate colloid) has been compared with open surgical synovectomy in
knee joints, with both procedures showing similar recurrence rates [8].
Although radiation synovectomy using 90 Y or other emitters has been
shown to be effective, the US Food and Drug Administration has not
approved its use because of the substantial risk of leakage of the radionuclide
from the treated joint. Healthy tissue doses can be quite large, with doses
between 250 and 500 cGy for the liver and between 5000 and 10 000 cGy
for the lymph nodes [9].
15.2. BORON NEUTRON CAPTURE SYNOVECTOMY
Boron neutron capture synovectomy (BNCS) has been proposed as a method

of performing radiation synovectomy that does not involve the use of any
radioactive material [10]. BNCS is a two-part procedure (illustrated schema-
tically in figure 15.1) in which a non-radioactive compound containing 10 B is
injected into the joint fluid; next, after allowing time for synovial uptake of
the compound, a beam of neutrons causes the 10 B to fission releasing two
high-LET, high-RBE particles that travel distances less than the diameter
of a cell. The 10 B(n,)7 Li reaction delivers intense radiation damage to
those cells that have previously been loaded with 10 B or their nearest
Boron neutron capture synovectomy 327
Figure 15.1. Schematic illustration of boron neutron capture synovectomy

showing a coronal view of a rheumatoid human knee joint, the largest articulat-
ing joint in the body. In a rheumatoid knee the joint space is enlarged resulting
from increased synovial fluid and thickened inflamed synovial tissue which, in
this example, has begun to erode adjoining articular cartilage. For BNCS a
boron-labelled compound will be injected directly into the joint space using a
standard lateral approach. The joint will then be irradiated by a low-energy
neutron beam.
neighbours. Like radiation synovectomy, BNCS results in synovial ablation

by the delivery of radiation energy to the diseased membrane. The procedure
could be performed on an outpatient basis and would require no rehabilita-
tion. Unlike radiation synovectomy, there is no radiation hazard associated
with leakage of the injected compound out of the joint. The 10 B remains
stable both before and after the neutron bombardment. The radiation dose
is only delivered while the tissue is undergoing neutron irradiation.
The use of the 10 B(n,)7 Li nuclear reaction to treat diseases is not a
novel concept. Boron neutron capture therapy (BNCT) was proposed in
1936 by Locher [11] as a potential treatment for cancer. Early clinical
assessment of BNCT for malignant gliomas using thermal neutrons was
initiated in the 1950s at Brookhaven National Laboratory (BNL) and at
MIT [12, 13]. These clinical trials were not successful because of the
insufficient penetration of thermal neutrons and the lack of tumour
specificity of the boron compounds used. These factors contributed to
excessive doses to normal brain tissue, and the trials were ended in 1961.
BNCT trials were resumed, however, in Japan in 1968, and continue there
to this day [2]. Recent advances in epithermal neutron beams and in
boron-delivery agents led to a resumption of BNCT clinical trials: at MIT
for intracerebral or sub-cutaneous melanoma and glioblastoma multiforme
[3], and at BNL [4] and the High Flux Reactor, Petten, The Netherlands [5]
for glioblastoma. These clinical trials all use a nuclear reactor as an epither-
mal neutron source. Accelerator-based BNCT for glioblastoma is also under
development [14–17].
There are several important differences between BNCS and BNCT.
First, the target tissue in BNCS is the synovium, which is only 0.4 to
1.3 cm below the skin surface (depending on the affected joint), while brain
tumours treated by BNCT can occur at depths up to 7 cm. Second, in vivo
10
B uptake studies have indicated that boron levels in the synovium on the
order of thousands or tens of thousands of ppm are readily achievable [7].
In BNCS, the boron-labelled compound is injected directly into the fluid
adjacent to the target tissue. In BNCT, conversely, the boronated compound
is delivered systemically, so 10 B levels in the target tissue are significantly
lower, usually on the order of 30–50 ppm. Finally, the joints to which
BNCS would be most frequently applied (knee and finger joints) are not
located near any radiosensitive organs of the body, whereas this is not the
case in BNCT clinical trials in the brain. All of these factors result in signifi-
cantly different neutron beam requirements for BNCS compared with
BNCT.
15.3. NEUTRON BEAM DESIGN CALCULATIONS FOR BNCS
BNCS could, in principle, be performed using any prolific neutron source,

including nuclear reactors, isotopic sources or charged-particle accelerators.
Isotope sources, however, may have limited practical application as a result
of their relatively low neutron output; storage of neutron-emitting sources
can also be problematic. Accelerators are compact and inexpensive
compared with nuclear reactors, and could be more easily installed in a
hospital environment. Research in BNCS at the Laboratory for Accelerator
Beam Applications (LABA) at MIT has therefore focused on accelerator
sources for BNCS. A tandem electrostatic accelerator [18] capable of
generating the proton and deuteron currents needed for the production of
high-intensity neutron beams is in operation at MIT LABA; this accelerator
was used extensively in the experimental evaluation of neutron beams for
BNCS and in the investigation of the efficacy of BNCS in an animal model
(see section 15.5 below).
The accelerator-based charged-particle reactions considered for BNCS
included 7 Li(p,n), 9 Be(p,n), 9 Be(d,n), 13 C(d,n), 2 H(d,n) and 3 H(d,n) [19, 20].
However, these and all other charged particle reactions generate energetic
neutrons. Since the cross-section for the boron neutron capture reaction
increases as the neutron energy decreases (i.e., very low energy neutrons
are desired for BNCS) then regardless of the neutron-producing reaction
used, the emitted neutrons must be moderated or filtered down to an
Neutron beam design calculations for BNCS 329
energy suitable for clinical application. How to generate a clinically useful

spectrum was then investigated using Monte Carlo simulation of neutron
transport, as described below.
The Monte Carlo simulation code MCNP4B [21] was used for all beam
design calculations. MCNP is public-domain simulation code developed at
the Los Alamos National Laboratory and capable of handling coupled
neutron–photon–electron transport problems. The code includes very
robust geometry specifications, physics and cross section packages, and
variance reduction capabilities. The code has been widely benchmarked
[22–24] and examples of MCNP used to model neutron capture therapy
problems abound in the literature [14, 15, 25–29].
15.3.1. Optimal beam energy

The first step in beam development was to determine the range of neutron
energies that would be most useful for joint irradiation. Given the shallow
depth of the synovium the optimal beam energy was expected to be lower
than the energies determined to be optimal for BNCT. Using Monte Carlo
simulation, an extensive ‘ideal beam’ study was performed in which the
dosimetric effects of 20 monoenergetic neutron beams, ranging from
0.025 eV to 10 keV, were evaluated in a tissue-equivalent model of the
human knee [7]. Both planar and isotropic beams were investigated. Each
beam was evaluated in terms of its effect on two therapeutic ratios, namely
the synovium dose to skin dose ratio and the synovium dose to bone surface
dose ratio. A 10 B concentration of 1000 ppm was assumed in the synovium.
This concentration is quite conservative relative to values measured in both
ex vivo and in vivo experiments [7, 30].
Plotting therapeutic ratios as a function of incident beam energy showed
similar trends for both therapeutic ratios and for both planar and isotropic
beams. Dose ratios are high (approximately 80–120) for all energies below
roughly 500 eV or 1 keV after which energy they begin to fall rapidly [7].
These results indicated that neutron beams with energies from thermal to
approximately 0.5 keV to 1 keV are optimal for BNCS. This energy range
is significantly lower than the range considered useful for BNCT and was
used as a goal in subsequent studies to design a practical and therapeutically
useful neutron beam.
15.3.2. Optimal beam design

Monte Carlo calculations using MCNP were performed to examine the
ability of various moderator/reflector configurations to generate high-
intensity neutron beams with low fast-neutron contamination. Early in the
investigation the 2 H(d,n) and 3 H(d,n) reactions were dropped from consid-
eration due to insufficient intensity in the moderated beam. However,
Table 15.1. Summary of charged-particle nuclear reactions considered for

BNCS.
Reaction Ion Average neutron Maximum Neutron yield

energy energy at 08 neutron energy (n/min/mA)
(MeV) (MeV) (MeV)
7
Li(p,n) 2.5 0.6 0.79 5:34 1013
9
Be(p,n) 4.0 1.06 2.12 6:0 1013
9
Be(d,n) 2.6 2.02 6.95 8:4 1013
1.5 1.66 5.81 1:67 1013
13
C(d,n) 1.5 1.08 6.77 1:09 1013

This yield was determined by matching simulated and measured phantom dose rates at MIT
LABA. Yields in the literature range from 1:98 1013 to 9:64 1013 n/min/mA [32, 33].
encouraging results were obtained with the 9 Be(p,n) reaction at proton

energies of 4.0, 3.7 and 3.4 MeV [31], the 9 Be(d,n) reaction at deuteron
energies of 2.6 and 1.5 MeV [32, 33], the 13 C(d,n) reaction at a deuteron
energy of 1.5 MeV [34], and the 7 Li(p,n) reaction at a proton energy of
2.5 MeV [35]. Table 15.1 provides initial yield and spectral information for
these reactions at various particle bombarding energies.
A number of moderator/reflector configurations were evaluated by
examining the dosimetric effect of each moderated beam in tissue-equivalent
phantoms of the human knee and finger [20]. Each joint phantom (see figure
15.2) consisted of cylindrically concentric layers of tissue: bone, articular
cartilage, joint fluid space, synovial lining, subsynovium, and fat and skin.
The outer diameters of the knee and finger joint are 8.7 and 2.23 cm,
respectively. Modelling of the human finger served a dual purpose since
the dimensions of the arthritic finger are similar to the dimensions of the
arthritic knee joint in rabbits. The arthritic rabbit model has been extensively
used by various investigators in the development and evaluation of
treatments for RA; we have also used this model to evaluate synovial
Figure 15.2. A cross-section through the cylindrically symmetric phantom

representing the knee or the finger joint. Realistic tissue compositions were
used and dimensions of each phantom were obtained from MR images.
uptake of boronated compounds in vivo [10] as well as to investigate the

efficacy of BNCS (see below). Initial beam design simulations were carried
out assuming a synovial concentration of 1000 ppm. Because of the
magnitude of this concentration the boron in the synovium was explicitly
modelled in all calculations [20]. All healthy tissues were assumed to contain
1 ppm 10 B. No data exist regarding RBE values of neutrons and 10 B(n,)
reactions in joint tissues. Therefore the RBE values commonly used for
BNCT of brain tumors were adopted for this work: 4.0, 3.8 and 1.0 for the
10
B reaction products, neutrons and photons, respectively.
For all moderator/reflector configurations evaluated the moderator
material was assumed to be D2 O and the length was varied to examine the
effect on therapeutic ratios and dose rate. Extremely large therapeutic dose
ratios were obtained with relatively small thicknesses of D2 O (i.e., high
fast/thermal neutron ratios) due to the high 10 B concentration assumed in
the target tissue. That is, neutron beams with significant fast neutron
contamination may still be therapeutically useful since the joint will be
exposed to the neutron beam for only a short period of time. High boron
levels thus lead to considerable flexibility in beam design for BNCS as
compared with the requirements for BNCT.
A number of reflector materials were examined including lead, graphite,
7
Li2 CO3 and D2 O. Highest dose rates at the phantom position were obtained
with a graphite reflector. Use of D2 O as a reflector material (i.e., simply
extending the moderator diameter) led to the highest therapeutic ratios as
a result of its low atomic number, with the ratios using graphite only slightly
lower. Since graphite is significantly less expensive than D2 O and provided
the highest dose rate, it became the reflector of choice. Thus, a moderator/
reflector assembly composed of D2 O and graphite was selected for subse-
quent optimization [20].
Results of these studies indicated that a superior beam would be
produced by the 7 Li(p,n) reaction due to its high yield of low energy
neutrons. However, given the practical difficulties in constructing and
cooling a lithium target [36], our initial efforts to design and construct a
useful BNCS beam have focused on the beryllium reactions listed above.
Beryllium is well suited as an accelerator target material because of its
mechanical strength, its high melting point and its high heat conductivity.
Two methods of improving both therapy time and therapeutic ratio
were investigated [20]. The first involved parallel beam irradiation. That is,
the joint (or the beam) would be moved such that the beam would be incident
on the back of the joint for the second half of the irradiation. It was
found that significant improvements in both therapy time and therapeutic
ratios were possible using two parallel-opposed beams. However, little
further improvement was seen as the number of irradiation directions
increased beyond two since the beam is larger than the diameter of the
phantom [19].
Figure 15.3. Two-dimensional plot through the MCNP model of the BNCS
moderator/reflector assembly including target, cooling apparatus, phantom
and graphite side and back reflectors around the phantom.
The second method of improvement involved the use of reflecting

material placed immediately behind and to the sides of the joint during
irradiation. This material provides the opportunity for neutrons to scatter
back into the joint, increasing the synovium dose per incident neutron.
Each scattering also serves to reduce the neutron energy. Thus, the effect
of the back and side reflectors is to significantly reduce therapy time as
well as to increase therapeutic ratios. While a number of possible low-Z
materials could be used for this purpose, graphite was again chosen due to
its low cost, availability and solid form.
Figure 15.3 shows an MCNP-generated plot of the optimized beamline
configuration including graphite reflector, D2 O moderator, knee phantom
surrounded by graphite side and back reflectors, and target with complete
target cooling apparatus. Using the MCNP model, therapy parameters for
joint treatments using various neutron-producing reactions were calculated
as a function of 10 B uptake over a range of 1 to 20 000 ppm. This range of
10
B concentration was based on in vivo uptake studies using arthritic rabbits
which showed that average boron levels of 19 000 for 30 min post-injection
are readily achievable in the synovium. Boron tissue concentrations were
explicitly included in the Monte Carlo simulations. Results are shown in
table 15.2 for parallel irradiations of human knee and finger joints, using
optimized graphite side and back reflector geometry. The therapy times
listed in table 15.2 are based on an accelerator beam current of 1 mA,
synovial boron concentrations of 1000 ppm 10 B, and the delivery of
10 000 RBE-cGy to the synovium, a value based on empirical estimates of
the dose required to produce a clinical effect in particle synovectomy [37].
Table 15.2. BNCS therapy parameters for human knee and human finger
joints. These results are for the optimized configuration shown in figure 15.3,
including side and back reflectors, assuming 1000 ppm 10 B uptake in the syno-
vium, and parallel-opposed irradiations.
Neutron source Moderator Therapy time Skin dose Bone

length (cm) (min mA) (RBE-cGy) ratio
Human knee
2.6 MeV 9 Be(d,n) 23 13 335 38
1.5 MeV 9 Be(d,n) 20 48 586 22
4.0 MeV 9 Be(p,n) 15 7.3 203 73
3.7 MeV 9 Be(p,n) 15 10 190 79
2.5 MeV 7 Li(p,n) 15 7 129 108
Human finger
2.6 MeV 9 Be(d,n) 23 6.1 227 42
1.5 MeV 9 Be(d,n) 20 24 406 24
4.0 MeV 9 Be(p,n) 15 3.6 161 62
3.7 MeV 9 Be(p,n) 15 4.8 136 69
2.5 MeV 7 Li(p,n) 15 3.3 91 108
Simulating a low boron concentration of 1000 ppm leads to therapy times of

generally less than 15 min for a knee joint and less than 7 min for a finger
joint. These times decrease significantly when higher synovial boron concen-
trations are assumed.
Based on the design predictions and results of the simulation studies, a
fully operational BNCS beam line, including accelerator target, cooling
system, and neutron moderator and reflector assembly has been constructed
and installed on the high-current tandem accelerator at MIT LABA. A
photograph of the installed assembly is shown in figure 15.4. The beamline
consists of a 9 cm diameter, 23 cm long D2 O moderator surrounded by an
18 cm thick graphite reflector. The graphite reflector has been resin coated
on its inner surface to make it impervious to the D2 O. The diameter of the
D2 O moderator was chosen to be approximately the size of a human knee
joint. The accelerator target assembly can be placed at any position along
the central axis of the reflector/moderator assembly, which allows flexibility
in the amount of D2 O between the target and the moderator exit window
[20]. The neutron target itself consists of a beryllium tube 4 cm long and
3.2 cm in diameter which seals to an aluminium tube. The target length
was chosen to minimize the number of secondary electrons escaping from
the target during irradiation, thereby enabling the accelerator beam current
to be measured more accurately. Beryllium was chosen as the target material
because it is chemically stable and has good heat transfer characteristics
including high melting point and thermal conductivity of 1283 8C and
1.84 W/cm 8C, respectively. The beryllium target is electrically insulated
Figure 15.4. Photograph of the reflector (graphite) and moderator (D2 O)

assembly installed on a dedicated beamline at MIT LABA.
from the aluminium tube, and target temperature is measured by a thermo-

couple attached to the centre of the target, on the coolant side of the
beryllium. The target is cooled using the submerged jet-impingement
technique, described by Blackburn et al. [36]. This technique can remove
peak heat densities of over 5 kW/cm2 and average heat densities of 1–2 kW/
cm2 over an 8 cm2 target area. The target cooling system is schematically
illustrated as part of the MCNP model shown in figure 15.3. The aluminium
target tube is surrounded by two concentric coolant tubes. The water coolant
enters the outer tube and is forced through a Teflon plug, which leads to a jet of
water impinging on the beryllium target. Water then flows between the inner
coolant tube and the aluminium target tube before exiting the assembly.
15.4. EXPERIMENTAL CHARACTERIZATION OF THE BNCS

BEAM
The neutron beamline assembly shown in figure 15.4 was used in a dose
escalation study to characterize the effects of BNCS in an animal model
(see section 15.5). Prior to the initiation of these animal experiments,
however, the MCNP predictions of the neutron and photon output of the
BNCS moderator/reflector assembly were experimentally verified. Accurate
knowledge of the thermal neutron flux in the phantom is important to
quantify the boron dose and thus treatment time. Photon and fast neutron
doses are also important in terms of dose to healthy structures such as the
Experimental characterization of the BNCS beam 335
skin and bone. Experimental validation of the MCNP-predicted in-phantom

dose rates was performed using the dual ionization chamber technique in
conjunction with foil activation analysis. This method of mixed field
dosimetry is discussed in detail by Rogus et al. [38], and has been used
extensively to characterize accelerator and reactor-produced epithermal
beams for BNCT [38–40]. Given the difficulty of measuring the various
dose components in a realistic (i.e., heterogeneous and small) model of the
rabbit knee, beam characterization was performed in a water-filled phantom;
experimental results were compared with MCNP predictions using a model
of the same phantom.
Measurements were performed using the 1.5 MeV 9 Be(d,n) neutron
source for moderator lengths of 8 and 23 cm, which result in very different
neutron spectra at the output of the moderator/reflector assembly [20].
Figure 15.5 compares the thermal neutron and fast neutron dose rates
determined by both experiment and simulation for the moderator length of
23 cm. The simulation results have been scaled to experiment at a depth of
4 cm. The experimental thermal neutron dose rates agree very well with
simulation over the range of depths in the phantom. The data agree within
one standard deviation from 2 to 10 cm, and agree within two standard
deviations at the 1 cm dose point. The experimental fast neutron dose rates
do not show such good agreement with simulation, but the data are still
within two standard deviations.
Figure 15.6 compares the measured neutron dose rates with simulation
for the 8 cm moderator length. In figure 15.6(a), simulation results are
compared with the results of several experiments. Initial experiments only
measured the thermal neutron dose at depths of 1 and 3 cm, while later
experiments measured the full range of depths. It is difficult to compare
with simulation at only one or two positions, especially if the neutron yield
is uncertain, since the full shape of the dose profile is not available. The
results from experiment 3 agree best with simulation when scaled at the
4 cm depth, and the effective neutron yield determined as a result of this
scaling agrees well with the neutron yield determined from the 8 cm data,
as well as the neutron yield determined from the fast neutron data at
23 cm. The results of all experiments are consistent in that they do agree
within one to two standard deviations (depending on the depth in phantom)
with each other and with simulation. Figure 15.6(b) compares simulation and
experiment for the fast neutron dose rates in the phantom, with very good
agreement between the measured and simulated values [20].
The neutron yield for the 1.5 MeV 9 Be(d,n) reaction varies widely in the
literature. An effective neutron yield was determined by scaling simulation to
experiment. Thermal and fast neutron dose rates were scaled at a depth of
4 cm in the phantom, as shown in figures 15.5 and 15.6. This allows the
shape of the measured and simulated dose profiles to be compared even
though there is uncertainty in the neutron yield. The consistency of the
Figure 15.5. Comparison of experimental and simulation results for (a) the
thermal neutron dose rate and (b) the fast neutron dose rate for a BNCS
target position of 23 cm. Simulation and experiment have been normalized at
a depth of 4 cm.
Experimental characterization of the BNCS beam 337
Figure 15.6. Comparison of experimental and simulation results for (a) the
thermal neutron dose rate and (b) the fast neutron dose rate for a BNCS
target position of 8 cm. Simulation and experiment have been normalized at a
depth of 4 cm.
Table 15.3. Experimentally determined neutron yields for the 1.5 MeV
9
Be(d,n) reaction, determined by scaling thermal and fast neutron doses at
two target positions.
Dose component Scaling factor (n/min mA)
8 cm moderator length 23 cm moderator length
Thermal neutron 1:82 1013 1:62 1013

Fast neutron 1:64 1013 1:58 1013
Averaged 1:67 10 1:1 1012
13
measured and calculated results was also assessed in terms of the effective
neutron yield, as determined by scaling either the fast or thermal neutron
dose rates at both target positions. Since the neutron yield of the 1.5 MeV
9
Be(d,n) reaction is a constant, the effective neutron yield determined by
scaling either the thermal or fast neutron dose, under any initial conditions
(i.e., for any target position), should also be constant.
Table 15.3 shows the effective neutron yield, determined by scaling
thermal and fast neutron doses at two target positions. The neutron yields,
for both dose components and target position, are very consistent and differ
by no more than 13%, with an average value of 1:67 1013 n/min mA. The
uncertainty in the effective neutron yield is estimated as 1:1 1012 n/min
mA, based on the standard deviation of the four values given in table 15.3.
Determining the neutron yield in this manner should be reasonable (assuming
the neutron energy and angular distributions used in the simulation are fairly
accurate), since the level of agreement in both the thermal and fast neutron
doses is good for two target positions that lead to very different neutron spec-
tra in the phantom.
15.5. EXPERIMENTAL INVESTIGATION OF BNCS IN AN

ANIMAL MODEL
The efficacy of BNCS in an animal model was investigated through a series of

rabbit irradiations using the BNCS neutron moderator/reflector assembly
shown in figure 15.4. The purpose of these irradiations was to determine,
in an animal model, if the combination of boron and neutrons could kill
arthritic tissue without adversely affecting healthy tissues. In addition, the
10
B(n,)7 Li dose necessary to achieve synovial necrosis was investigated by
examining histological sections taken from rabbits irradiated over a wide
range of doses. The empirical estimate of the required dose, based on
particle radiation synovectomy studies, is 10 000 cGy [37]. However, a
detailed dose study for particle radiation synovectomy has not been
Experimental investigation of BNCS in an animal model 339
reported in the literature. The dose-delivering particles in BNCS are very

different from those encountered in particle radiation synovectomy and,
in addition, the RBE values that should be used for BNCS are not known.
The dose escalation study was therefore performed to characterize the
therapeutic dose necessary in BNCS, and to study the effects of BNCS as a
function of dose. This section will present a description of the rabbit
irradiations and a determination of the BNCS dose response in rabbits.
The animals used in this dose escalation study were New Zealand white
rabbits, using the antigen-induced arthritis (AIA) model. The complete
protocol for the arthritis induction is described in Binello [7]. The rabbits
were irradiated three days after the arthritis induction. After the rabbit
was anaesthetized, the knee was injected intra-articularly with a K2 10 B12 H12
solution containing 150 000 ppm 10 B. An injection of this concentration has
been shown to lead to an average 10 B uptake of 19 000 ppm in the synovium
[7, 30] for 30 min post-injection. Rabbits were irradiated with synovial doses
ranging from 800 to 81 000 RBE-cGy. The 1.5 MeV 9 Be(d,n) charged particle
reaction produced using the MIT LABA accelerator was used for these
experiments. The beryllium target was placed 8 cm from the exit of the
moderator/reflector system. MCNP was used to model the rabbit irradiation
geometry and to calculate accelerator currents and irradiation times for
various synovial dose levels.
Three types of control rabbit were also used. One set of controls
consisted of arthritic rabbits not subjected to either the neutron irradiation
or the injection of the boron compound (arthritic only). The second set
received the boron compound, but were not irradiated (compound only).
The third set consisted of rabbits that were irradiated for the same amount
of time as the 81 000 RBE-cGy rabbits, but without boron in the synovium
(neutron only). The synovium will receive some dose from the incident
neutron beam and induced photons, but not from boron neutron capture
events.
Twenty-eight rabbits were irradiated as part of the BNCS dose escala-
tion study. Rabbits were sacrificed three days post-irradiation. At each
dose point two animal knees were kept whole and decalcified for whole-
knee histological examination, and the third knee was dissected, synovium
removed, sectioned and stained with hematoxylin and eosin and examined
for signs of tissue necrosis, cellular debris and inflammatory cell infiltrates
[7]. A brief summary of the results is presented here.
The arthritic only control rabbits showed the expected features of the
rabbit AIA model. The compound only and neutron only controls did not
exhibit any effects different from the arthritic only controls. This implies
that any effects seen in the other rabbits were a result of the combination
of boron injection and subsequent neutron irradiations. At the 800 RBE-
cGy dose, a decrease in the density of inflammatory cells and a decrease in
the thickness of the synovial lining was observed. These effects were also
observed, to a greater extent, at the 2030 RBE-cGy dose, suggesting a dose-

dependent response. At 3660 RBE-cGy, a further reduction in the number of
inflammatory cells was observed, but still without examples of severe radia-
tion-induced necrosis.
Clear evidence of synovial necrosis was observed at the 6900 and
7300 RBE-cGy doses. There was a decrease in the number and size of the
synovial cells. There was evidence of cartilage damage, but it was unclear
if this resulted from the arthritic condition or from the irradiation. At the
very high doses (38 900 and 81 000 RBE-cGy), synovial necrosis was
accompanied by what appeared to be radiation-induced damage to the
femoral cartilage. The empirical estimate of the therapeutic dose from
radiation synovectomy is 10 000 RBE-cGy, and there is a significant
uncertainty associated with this estimate. Necrosis in this study was observed
at 6900 and 7300 RBE-cGy (with an estimated uncertainty of 23%), about
30% lower than expected. Given the uncertainties in the dose calculation,
this agreement is entirely within reason. Also, no synovial doses between
7300 and 38 900 RBE-cGy were examined; this dose range should be
examined in further study. Since control rabbits showed no adverse effects,
it can be concluded that the combination of boron and neutrons is effective
in killing arthritic tissue in a rabbit model.
15.6. WHOLE-BODY DOSIMETRY FOR BNCS
The previous sections have focused on simulation and experiment to

characterize the dose delivery to the targeted joint. The dose to the rest of
the patient is also important, and should be minimized. Monte Carlo
calculations were performed to characterize the effective dose for a number
of neutron sources and shielding configurations that could be utilized in
BNCS. To validate that Monte Carlo methods can be used to predict
effective doses, experiments were performed using a whole-body phantom
and the BNCS moderator/reflector assembly installed at LABA [20]. Experi-
ments were done for both shielded and unshielded configurations. These
results were compared with Monte Carlo simulations of the same configura-
tions to assess the accuracy of MCNP predictions of whole-body dose.
The MCNP model of the BNCS moderator/reflector assembly was
combined with an anthropomorphic phantom [41] of the human body in
order to study the whole-body effective dose received during a knee
treatment. The anthropomorphic phantom, shown in figure 15.7(a), is
based on data from a number of sources, including MIRD [42], Cristy and
Eckerman [43], Tsui et al. [44] and the Visible Man project [45]. Dosimetry
calculations were performed for several of the candidate charged-particle
reactions for BNCS. Several shielding configurations and materials were
tested in an effort to reduce the effective dose. Shielding options included
Whole-body dosimetry for BNCS 341
Figure 15.7. (a) MCNP simulation model including the anthropomorphic

phantom and the BNCS moderator/reflector assembly. (b) Photograph of the
phantom positioned for experiments without whole-body shielding. (See plate
5 for colour version.)
embedding the moderator/reflector assembly in a wall of boronated poly-

ethylene, adding a neutron beam delimiter to the cylindrical graphite
reflector assembly, and inserting additional shielding around the patient’s
legs. Results were generated for 19 000 ppm of 10 B uptake in the synovium.
The most effective shielding configuration was a combination of wall,
delimiter and leg shields. This shielding configuration reduced the unshielded
doses by a factor of roughly 2. For synovial boron uptake levels of
19 000 ppm, the effective doses for a shielded patient ranged from 0.13 rem
to 0.72 rem for the 4 MeV 9 Be(p,n) and 2.6 MeV 9 Be(d,n) reactions, the
softest and hardest spectra examined, respectively. Simulations indicated
that a major component of whole-body dose resulted from neutrons entering
the knee joint, and then scattering through the rest of the body; it is difficult
to reduce this dose component without degrading neutron delivery to the
targeted area.
Since the main clearance pathway of many boron compounds from the
joint is through the urine, the dose to the bladder was specifically examined as
a function of boron uptake. Depending on the source reaction, the bladder
dose increases significantly when the boron uptake is in the range of 10–
1000 ppm, and the whole-body effective dose increases significantly when
the bladder boron uptake reaches several hundred to a thousand ppm. The
estimated boron uptake in the bladder corresponding to an uptake of
19 000 ppm K2 B12 H12 in the synovium was 2630 ppm. For the 4 MeV
9
Be(p,n) neutron source, this would increase the bladder dose from
0.12 rem to 0.45 rem.
Experiments were performed to verify that Monte Carlo methods could
be used to predict effective doses, using a whole-body phantom and the
BNCS moderator/reflector assembly installed at LABA. The whole-body
phantom, composed of two water-filled tanks to represent the torso and
one leg, is shown in figure 15.7(b). Experiments were done for both shielded
and unshielded configurations. These results were compared with Monte
Carlo simulations of the experimental configurations.
A combination of two dosimeters was chosen to characterize the mixed-
field radiation dose for the whole-body dosimetry experiments: thermo-
luminescent dosimeters (TLDs) and bubble detectors. TLDs were chosen
to characterize the photon dose, while the bubble detectors were chosen to
measure the thermal and fast neutron dose. TLDs are attractive because
they are sensitive to low doses, the dosimeters are small and thus will not
significantly perturb the radiation field, and tissue equivalence is possible.
Bubble detectors are also sensitive to low doses, are tissue equivalent, and
insensitive to photons. Bubble detectors are available with and without
thermal neutron sensitivity, so the neutron dose over a wide range of energies
can be characterized.
In general, the total equivalent doses for both the shielded and
unshielded experiments agreed well with simulation. The measured values
Summary and conclusions 343
were less than the simulated values, which suggests that from the perspective
of radiation protection, the MCNP calculations are conservative. In most
cases, the measured individual dose components agreed within a factor of
2 with the simulated values. This level of agreement is considered reasonable
for shielding problems and, in particular, for shielding measurements in a
mixed radiation field. The exception to this was the thermal neutron doses
for the shielded phantom, which agreed, on average, within a factor of 4
with simulation.
There are several reasons why the simulated and measured dose values
were only expected to agree within a factor of a few. For the simulation, the
results are only as accurate as the initial source specification. As described
earlier, the neutron and photon sources may not be robust enough to
accurately predict doses far from the target. In addition, the simulation
results have been shown to be sensitive to the surrounding materials in the
radiation vault. While an attempt has been made to accurately describe the
vault in the Monte Carlo model, it is possible that further refinements in
any approximations in the model of the vault may lead to better agreement
between simulation and measurement. Furthermore, the simulation model
neglects all radiation streaming effects, i.e., it assumes that the boronated
polyethylene wall contains no cracks. Streaming effects could have a large
effect on the shielded simulations, since the dose rates have been greatly
reduced by the presence of the shield.
The determination of the photon dose rate is dependent on a correction
for the induced TLD response from thermal neutrons, which can be quite
significant. The thermal neutron response for the TLDs has been experimen-
tally characterized in a thermal neutron beam at the MIT Research Reactor.
The correction is also dependent, however, on the measured thermal neutron
dose determined by the bubble detectors. The measured thermal neutron
doses have been shown to be about a factor of 4 higher than calculated for
the shielded results, and the experimentally determined photon response is
about a factor of 3 lower than calculated. This may result from an over-
correction of the induced thermal neutron response in the TLDs.
15.7. SUMMARY AND CONCLUSIONS
Boron neutron capture synovectomy is a promising modality for the treat-

ment of rheumatoid arthritis and has been shown to be effective in causing
synovial ablation in an animal model. Based on a single injection of a
boron-labelled compound into the affected joint followed by neutron beam
irradiation, BNCS is expected to be a simple procedure performed on an
outpatient basis. The goal is to destroy the inflamed synovial tissue, thereby
alleviating the symptoms of pain and disability. Similar to all existing
methods of treating RA, BNCS addresses the symptoms of this disease
and not the cause, which remains unknown. Hence, recurrence is expected
2ÿ5 years later. However, a repeat BNCS procedure, unlike a repeat surgical
synovectomy, is expected to be a simple and straightforward procedure.
Monte Carlo methods have been used extensively in the design and
evaluation of several aspects of BNCS. Monte Carlo calculations indicate,
for a variety of neutron-producing charged particle reactions, that BNCS
treatment times are very reasonable. Times based on a low 10 B concentration
of only 1000 ppm are generally less than 15 min for a knee joint and less than
7 min for a finger joint, assuming an accelerator current of 1 mA. These times
decrease even further when higher synovial boron concentrations are
assumed. Associated whole-body effective doses can be kept low with
appropriate shielding. Whenever possible, Monte Carlo predictions have
been verified experimentally using a combination of detector types. Given
the wide range of neutron energies encountered and the simultaneous
presence of photons, detection and dosimetry is challenging. Combinations
of ion chambers, foil activation, bubble detectors and thermoluminescent
dosimeters have been used in the experimental evaluation of Monte Carlo
predictions for BNCS. Agreement between simulation and experimental
measurement of in-phantom therapy dose-rates was within two standard
deviations.
REFERENCES
[1] Lipsky P E 1991 Rheumatoid Arthritis. Harrison’s Principles of Internal Medicine ed

J D Wilson et al. (New York: McGraw-Hill) 12th edition pp 1437–43
[2] Hatanaka H and Nakagawa Y 1994 Clinical results of long surviving brain tumor
patients who underwent boron neutron capture therapy Int. J. Radiat. Oncol. Biol.
Phys. 28 1061–6
[3] Busse P M, Zamanhof R, Harling O K, Kaplan J, Kaplan C, Chuang Goorley J,
Kiger W S, Riley K, Tang L, Solares G and Palmer M 2001 The Harvard–MIT
BNCT Program: overview of the clinical trials and translational research. In Frontiers
in Neutron Capture Therapy eds M F Hawthorne, K Shelley and R J Wiersema (New
York: Kluwer–Plenum) vol 1 pp 37–60
[4] Diaz A Z, Chanana A D, Capala J, Chadha M, Coderre J A, Elowitz E H, Iwai J, Joel
D D, Liu H B, Ma R, Shady M D, Slatkin N, Tyson G W and Wielopolski L 2001
Boron neutron capture therapy for glioblastoma multiforme: results from the
initial phase I/II dose escalation studies. In Frontiers in Neutron Capture Therapy
eds M F Hawthorne, K Shelley and R J Wiersema (New York: Kluwer–Plenum)
vol 1 pp 61–72
[5] Sauerwein W, Moss R, Hideghety K, Stecher-Rasmussen F, deVries M, Reulen H-J,
Gvtz C, Paquis P, Wiestler F and Gabel D 2001 Status report on the European
clinical trial of BNCT at Petten. In Frontiers in Neutron Capture Therapy eds
M F Hawthorne, K Shelley and R J Wiersema (New York: Kluwer–Plenum) vol 1
pp 81–86
References 345
[6] Schiff M 1997 Emerging treatments for rheumatoid arthritis Am. J. Med. 102 (suppl
1A) 11S–15S
[7] Binello E 1999 Efficacy of boron neutron capture synovectomy in an animal model.
PhD Thesis. Massachusetts Institute of Technology
[8] Gumpel J M and Roles N C 1975 A controlled trial of intra-articular radiocolloids
versus surgical synovectomy in persistent synovitis Lancet 1 488–9
[9] Davis M A and Chinol M 1989 Radiopharmaceuticals for radiation synovectomy:
evaluation of two yttrium-90 particulate agents J. Nucl. Med. 30 1047–55
[10] Yanch J C, Shortkroff S, Shefer R E, Johnson S, Binello E, Gierga D, Jones A G,
Young G, Vivieros C, Davison A and Sledge C 1999 Boron neutron capture
synovectomy: treatment of rheumatoid arthritis based on the 10B(n,)7Li nuclear
reaction Med. Phys. 26 364–75
[11] Locher G L 1936 Biological effects and therapeutic possibilities of neutrons Am. J.
Roentgenol. 36 1–13
[12] Slatkin D N 1991 A history of boron neutron capture therapy of brain tumors:
postulation of a brain radiation dose tolerance limit Brain 114 1609–29
[13] Farr L, Sweet W H, Robertson J S, Foste C G, Locksley H B, Sutherland D L,
Mendelsohn M L and Stickley E E 1954 Neutron capture therapy with boron in
the treatment of glioblastoma multiforme Am. J. Roentgenol. 71 279–91
[14] Yanch J C, Zhou X-L, Shefer R E and Klinkowstein R E 1992 Accelerator-based
epithermal neutron beam design for neutron capture therapy Med. Phys. 19 709–21
[15] Allen D A, Beynon T D, Green S and James N D 1999 Toward a final design for the
Birmingham boron neutron capture therapy neutron beam Med. Phys. 26 77–86
[16] Chris Wang C-K, Blue T E and Gahbauer R 1988 A neutronic study of an accelerator-
based neutron irradiation facility for boron neutron capture therapy Nucl. Technol. 84
93–107
[17] Bleuel D L, Donahue R J, Ludewigt B A and Vujic J 1998 Designing accelerator-
based epithermal neutron beams for boron neutron capture therapy Med. Phys. 25
1725–34
[18] Klinkowstein R E, Shefer R E, Yanch J C, Howard W B, Song H, Binello E, Black-
burn B W, Daigle J L, Sears S M, Goldie C H and Ledoux R J 1997 Operation of a
high current tandem electrostatic accelerator for boron neutron capture therapy. In
Advances in Neutron Capture Therapy eds B Larson, J Crawford and R Weinreich
(Amsterdam: Elsevier) vol I pp 522–7
[19] Binello E, Shefer R E and Yanch J C 1997 Neutron beam design for boron neutron
capture synovectomy. In Advances in Neutron Capture Therapy (Amsterdam: Elsevier)
vol II pp 459–63
[20] Gierga D P, Yanch J C and Shefer R E 2000 Development and construction of a
neutron beamline for accelerator-based neutron capture synovectomy Med. Phys.
27 203–14
[21] Briesmeister J F 1997 MCNP—A General Monte Carlo N-Particle Transport Code
Version 4B 1997. Los Alamos National Laboratory Report LA-12625-M
[22] Whalen D J, Cardon D A, Uhle J L and Hendricks J S 1991 MCNP: Neutron Bench-
mark Problems Los Alamos National Laboratory Report LA-12212
[23] Whalen D J, Hollowell D E and Hendricks J S 1991 MCNP: Photon Benchmark
Problems Los Alamos National Laboratory Report LA-12196
[24] Gierga D P and Adams K J 1999 Electron/Photon Verification Calculations using
MCNP4B Los Alamos National Laboratory Report LA-13440
[25] Zamenhof R, Redmond E, Solares G, Katz D, Riley K, Kiger S and Harling O K

1996 Monte-Carlo based treatment planning for boron neutron capture therapy
using custom design models automatically generated from CT data Int. J. Radiat.
[26] Allen D A, Beynon T D and Green S 1999 Design for an accelerator-based orthogonal
epithermal neutron beam for boron neutron capture therapy Med. Phys. 26 71–6
[27] Lee C L, Zhou X L, Kudchadker R J, Harmon F and Harker Y D 2000 A Monte
Carlo dosimetry-based evaluation of the 7Li(p,n)7Be reaction near threshold for
accelerator-based boron neutron capture therapy Med. Phys. 27 192–202
[28] Kiger W S, Sakamoto S and Harling O K 1999 Neutronic design of a fission con-
verter-based epithermal neutron beam for neutron capture therapy Nucl. Sci. Eng.
131 1–22
[29] Reed M K, Dobelbower M C, Woollard J E and Blue T E 1998 Mixed-field dosimetry
measurement of a target assembly for an accelerator-based neutron source for boron
neutron capture therapy Nucl. Instrum. Meth. A 419 160–6
[30] Yanch J C, Shortkroff S, Shefer R E, Binello E, Gierga D, Jones A G, Young G,
Vivieros C and Blackburn B W 2001 Progress in the development of boron neutron
capture synovectomy for the treatment of rheumatoid arthritis. In Frontiers in
Neutron Capture Therapy eds M F Hawthorne, K Shelley and R J Wiersema (New
York: Kluwer–Plenum) vol 2 pp 1389–98
[31] Howard W B, Yanch J C, Grimes S M, Massey T N, Al-Quaraishi S I, Jacobs D K
and Brient C E 2001 Measurement of the 9Be(p,n) thick target spectrum for use in
accelerator-based boron neutron capture therapy Nucl. Sci. Eng. 138 145–60
[32] Meadows J W 1991 The thick-target 9Be(d,n) neutron spectra for neutron energies
between 2.6 and 7.0 MeV Argonne National Laboratory Nuclear Data and Measure-
ments Series ANL/NDM-124
[33] Guzek J 1998 PhD Thesis, University of Witwatersrand, South Africa
[34] Colonna N, Beaulieu L, Phair L, Wozniak G J, Moretto L G, Chu W T and Ludewigt
B A 1999 Measurements of low-energy (d,n) reactions for BNCT Med. Phys. 26 793ÿ79
[35] Liskien H and Paulsen A 1975 Neutron production cross sections and energies for the
reaction 7Li(p,n)7Be and 7Li(p,n)7Be At. Data Nucl. Data Tables 15 57–84
[36] Blackburn B W, Yanch J C and Klinkowstein R E 1998 Development of a high-power
water cooled beryllium target for use in accelerator-based boron neutron capture
therapy Med. Phys. 25 1967–74
[37] Deutsch V, Brodack J W and Deutsch K R 1993 Radiation synovectomy revisited
Eur. J. Nucl. Med. 20 1113–27
[38] Rogus R D, Harling O K and Yanch J C 1994 Mixed field dosimetry of epithermal
neutron beams for boron neutron capture therapy at the MITR-II research reactor
Med. Phys. 21 1611–25
[39] Raaijmakers C P J, Konijnenberg M W, Verhagen H W and Miknheer B J 1995
Determination of dose components in phantoms irradiated with an epithermal
neutron beam for boron neutron capture therapy Med. Phys. 22 321–9
[40] White S M 1998 Beam characterization for accelerator-based boron neutron capture
therapy using the 9Be(d,n) nuclear reaction. MS Thesis. Massachusetts Institute of
Technology
[41] Lambeth M 1997 Development of a computerized anthropomorphic phantom for
determination of organ doses for diagnostic radiology. BS Thesis. Massachusetts
Institute of Technology
References 347
[42] Snyder W S, Ford M R, Warner G G and Fisher H L Jr 1969 MIRD Pamphlet No 5

Supplement 3: Estimates of absorbed fractions for monoenergetic phantom sources
uniformly distributed in various organs of a heterogeneous phantom J. Nucl. Med.
10 46–52
[43] Cristy M and Eckerman K F 1987 Specific absorbed fractions of various ages from
internal photon sources. I. Methods ORNL/TM-8381/VI
[44] Tsui B M W, Zhao X D, Gregoriou G K, Lalushl D S, Frey E C, Johnston R E and
McCartney W H 1994 Quantitative cardiac SPECT reconstruction with reduced
image degradation due to patient anatomy IEEE Trans. Nucl. Sci. 41 2838–44
[45] Lorensen W 1995 Marching Through the Visible Man Proceedings of Visualization ’95
(IEEE Press)
Chapter 16
Summary
Habib Zaidi and George Sgouros
Monte Carlo analysis in nuclear medicine has been used for several decades.
Over this time, its use has evolved from a tool available to only a few expert
physicists to one that may reside on a desktop computer and that is now
potentially available for routine use. This evolution has been driven by
four fundamental and critical developments:
1. The ever-increasing sophistication of imaging instrumentation and
associated image analysis and processing methodologies, an area in
which simulation studies have become essential.
2. The emergence of targeted radionuclides as a therapeutic modality,
requiring improved and patient-specific dosimetry at the macro and
micro level.
3. The availability of three- and four-dimensional computer phantoms
providing flexibility and accurate modelling of populations of patient
anatomies, and attenuation and scatter properties as well as biodistribu-
tions of radiopharmaceuticals in the patients.
4. The near exponential increase in the speed and proportional reduction
in size and cost of computing power.
These four interrelated developments have fostered the work reviewed in this
book.
The recent development of combined modality systems (e.g., SPECT/
PET/CT/MRI) has created new opportunities for Monte Carlo simulation
in imaging. These new imaging modalities in turn make it possible to
meet the new requirements on dosimetry that have been driven by thera-
peutic nuclear medicine. The increasing availability of end-user image
formation/analysis and dosimetry software packages that incorporate
Monte Carlo calculations reflects all four of the developments listed
above.
348
Summary 349
It is important to note that although Monte Carlo techniques have existed

for decades, their use in therapeutic nuclear medicine and, in particular, in
imaging-based patient-specific dosimetry is still in its infancy. This book
may, therefore, be viewed as an early album containing several of the many
and varied snapshots of this rapidly growing area.
Biosketches
The authors in this book are drawn from the leading research groups around
the world. Habib Zaidi is senior physicist and head of the PET Instrumenta-
tion and Neuroscience Laboratory at Geneva University Hospital and his
research activities centre on modelling nuclear medical imaging systems
using the Monte Carlo method, dosimetry, image correction, reconstruction
and quantification techniques in emission tomography. Dr George Sgouros
is Associate Member, Department of Medical Physics at Memorial Sloan-
Ketting Cancer Center and his research activities are in the field of dosimetry
and systemic radiation therapy physics.
Habib Zaidi, PhD

Dr Habib Zaidi is senior physicist and head of the PET Instrumentation and
Neuroscience Laboratory at Geneva University Hospital. He received a PhD
in medical physics from Geneva University for a dissertation on Monte Carlo
modelling and scatter correction in positron emission tomography. His
research activities centre on modelling nuclear medical imaging systems
using the Monte Carlo method, dosimetry, image correction, reconstruction
and quantification techniques in emission tomography as well as functional
brain imaging, and more recently on novel design of dedicated high-resolution
PET scanners in collaboration with CERN. He is an associate editor for Medi-
cal Physics journal, member of the editorial board of the International Journal
of Nuclear Medicine, regional editor for Electronic Medical Physics News, a
publication of the International Organization for Medical Physics (IOMP),
and scientific reviewer for several medical physics, nuclear medicine and
computing journals. He is also affiliated to several international medical
physics and nuclear medicine organizations and a member of the professional
relations committee of the IOMP. He is involved in the evaluation of research
350
Biosketches 351
proposals for European and International granting organizations and partici-

pates in the organization of international symposia and conferences.
email: [email protected]
Web: http://dmnu-pet5.hcuge.ch/
George Sgouros, PhD

Dr George Sgouros is Associate Member, Department of Medical Physics at
Memorial Sloan-Kettering Cancer Center. His research activities are in the
field of dosimetry and systemic radiation therapy physics. He received his
bachelor’s degree from Columbia University School of Engineering and
Applied Science and his PhD from Cornell University, Graduate School of
Medical Sciences. He is the author of numerous chapters, peer-reviewed
publications and conference proceedings in targeted radionuclide therapy
and radionuclide dosimetry. He is a member of the Steering Committee of
the American Association of Physicists in Medicine (AAPM), Task Group
on internal emitter dosimetry and also a member of the Society of Nuclear
Medicine’s Medical Internal Radiation Dose (MIRD) Committee.
Pedro Andreo, PhD, DSc

Professor Pedro Andreo graduated (MSc) in Theoretical Physics in 1974, and
got his PhD degree in 1982 at the University of Zaragoza (Spain). He moved
to the Department of Medical Radiation Physics, Karolinska Institute–
University of Stockholm (Sweden) as a Research Fellow in 1987, becoming
Associate Professor and Doctor in Sciences (DSc) in Radiation Physics at
the University of Stockholm in 1989. He was appointed Full Professor in
Radiotherapy Physics at the University of Lund (Sweden) in 1993, and
head of the Radiotherapy Physics group of this University Hospital. Between
1995 and 2000 he was the Head of the Dosimetry and Medical Radiation
Physics Section of the International Atomic Energy Agency (IAEA),
during which period he was the Secretary of the IAEA/WHO network of
Secondary Standards Dosimetry Laboratories. Since 2000 he has been Full
Professor of Medical Radiation Physics of the University of Stockholm.
His scientific activities have emphasized the use of the Monte Carlo
method in radiotherapy physics, mainly for absolute dosimetry and for
treatment planning with electron and photon beams, where he was one of
the pioneers in the field. His most recent activities have been focused on
fundamental radiotherapy dosimetry of therapeutic proton and heavy ion
beams using Monte Carlo methods.
e-mail: [email protected]
Web: http://www.ki.se/onkpat/radfys/
352
Manuel Bardies, PhD

Dr Manuel Bardies was born in 1963. He obtained his MSc in Biomedical
Engineering in 1988, then a PhD in Radiological Physics in 1991. During
his PhD, he stayed at the Hammersmith Hospital in London for a year,
with Dr M J Myers. He then joined INSERM (National Institute of
Health and Medical Research) in 1992, and is working there as a research
scientist in Professor J F Chatal’s team. His fields of interest are all related
to dosimetry for targeted radiotherapy, including quantitative imaging,
dose calculations and small scale dosimetry.
Wesley Bolch, PhD

Wesley Bolch, PhD, PE, CHP, is Professor of Radiological and Biomedical
Engineering at the University of Florida in the Department of Nuclear and
Radiological Engineering. His current research interests include: (1) NMR
microscopy of trabecular bone for improved dosimetric models of the
skeleton, (2) development of tomographic computational models of newborn
anatomy in support of paediatric radiology dosimetry, (3) incorporation of
uncertainty analysis in internal dosimetry models, and (4) molecular
models of radiation damage to DNA. Dr Bolch is a member of the Society
of Nuclear Medicine’s Medical Internal Radiation Dose (MIRD) Commit-
tee. He presently serves on the Board of Directors for the Health Physics
Society.
email: wbolch@ufl.edu
Web: http://www.nuceng.ufl.edu/
Lionel Bouchet, PhD

Lionel G Bouchet was born in Grenoble, France, in 1971. He attended Lycée
Pierre du Terrail high school in the French Alps, graduating in 1989. From
1989 until 1991, he prepared the entry exam to the French Engineering
School at the Lycée Champollion in Grenoble, France. He entered the
Engineering Physics School of Grenoble (ENSPG), France where he
graduated in 1994. The same year, he entered graduate school at Texas
A&M University in the Department of Nuclear Engineering, where he
developed the new MIRD mathematical head and brain dosimetric model,
which earned him an MSc degree in December 1994. In 1995 he commenced
his doctoral research in Medical Health Physics in the Department of
Nuclear and Radiological Engineering at the University of Florida, where
he developed in collaboration with the MIRD Committee several dosimetric
models for use in nuclear medicine. In 1999, he joined the Department of
Neurological Surgery at the University of Florida as a Postdoctoral Research
Biosketches 353
Assistant working on image-guided radiotherapy and radiosurgery. In July

2000, he began an appointment as an Assistant Professor in the Department
of Neurological Surgery at the University of Florida. His current research
interests include three-dimensional image guidance in surgical procedures
and in radiosurgery/radiotherapy, patient-specific virtual reality as applied
to neurosurgical procedures, radiation transport and detection, and
patient-specific internal dosimetry for nuclear medicine.
email: [email protected]fl.edu
A Bertrand Brill, MD, PhD

Randy Brill received his MD at the University of Utah in 1956, and his PhD at
the University of California, Berkeley (Biophysics) in 1961. His dissertation
advisor was Hardin Jones, and the topic was radiation leukemogenesis, a
study based on the Hiroshima/Nagasaki A-bomb survivors. Randy started
his career as a Public Health Service Officer assigned to the Atomic Bomb
Casualty Committee following which he returned to the University of
California to complete his PhD. He was assigned to Johns Hopkins
(Assistant Professor, Radiology), where he set up a radiation epidemiology
study to investigate the possible relation between 131 I administered to people
and leukaemia, as a counterpart to the then known effects in the A-bomb
survivors. Randy returned to academic medicine (Vanderbilt University)
after 7 years in the PHS. He was an Associate Professor, then Professor, in
Medicine/Physics/Radiology, co-Director, then Director of the Division of
Nuclear Medicine and Biophysics. After 15 years, he left for Brookhaven
National Laboratory where he served as Nuclear Medicine Program Coordi-
nator for 7 years. Thereafter, he served as Professor and Research Director of
the Nuclear Medicine Department at the University of Massachusetts Medical
School for 10 years, following which he returned to Vanderbilt as Research
Professor of Radiology and Physics. His current work involves the develop-
ment of new nuclear medicine imaging receptors, and radiation dosimetry,
particularly oriented to radioimmunotherapy and bone marrow effects.
Yuni Dewaraja, PhD

Yuni Dewaraja received her BS in Electrical Engineering from the University
of Western Australia in 1986 after which she worked for 2 years as an engineer
for the Atomic Energy Authority in Colombo, Sri Lanka. She received her MS
in Nuclear Engineering from Kansas State University in 1990. She received her
PhD in Nuclear Engineering from the University of Michigan in 1994 with a
dissertation entitled ‘Imaging neutron activation analysis and multiplexed
gamma-ray spectrometry’. In 1996, she joined the Division of Nuclear
354
Medicine at the University of Michigan Medical Center where she is now an

Assistant Research Scientist. Dr Dewaraja’s current research activities include
quantitative SPECT imaging, Monte Carlo methods in nuclear medicine
imaging and three-dimensional absorbed dose calculation. Dr Dewaraja is a
member of the Society of Nuclear Medicine and the IEEE.
e-mail: [email protected]
Dennis M Duggan, PhD

Dennis M Duggan received a PhD in Physics from the University of
Southern California in 1986. Over 18 years, he worked as a computer
programmer, microwave and electro-optical engineer, and solid-state
physicist before beginning a fellowship in medical physics under Charles W
Coffey II, PhD at the University of Kentucky in 1992. He followed Charles
Coffey to Vanderbilt University and finished his fellowship there in 1994, and
has been there ever since. His present research interests include radiation
dosimetry, both theoretical and experimental, and quality assurance for all
types of conformal radiotherapy including brachytherapy, both interstitial
and intravascular, stereotactic radiosurgery, and intensity modulated radio-
therapy, and the application of functional imaging, such as positron emission
tomography and magnetic resonance spectroscopy imaging, to radiation
therapy planning.
John Humm, PhD

John Humm was awarded a PhD in 1983 for a dissertation entitled ‘The
analysis of Auger electrons released following the decay of radioisotopes
and photoelectric interactions and their contribution to energy deposition’,
performed at the Institute for Nuclear Medicine, Nuclear Research Center
in Jülich, Germany. He then worked at the MRC Radiobiology Unit in
Harwell, Oxfordshire from 1983 to 1987, Charing Cross Hospital, London,
in 1988 and Harvard Medical School, Boston, from 1989 to 1993. John
Humm is currently the chief of the Nuclear Medicine Physics section at
Memorial Sloan-Kettering Cancer Center, New York, where he has been
since 1993. His interests are wide, spanning radiation biology to imaging
and radiotherapy physics, with a special focus on targeted therapy, micro-
dosimetry, and more recently positron emission tomography and molecular
imaging.
Web: http://www.ski.edu/
Biosketches 355
Timothy K Johnson, PhD

Dr Timothy K Johnson is currently Associate Professor in the Department of
Radiation Oncology at the University of Colorado Health Sciences Center,
Denver, Colorado. Dr Johnson received his MS degree in Medical Physics
from the University of Colorado in 1981 and his PhD from the University
of Minnesota in Biophysical Sciences, Medical Physics in 1990. From 1991
to 2001, he was the director of the Graduate Program in Medical Physics at
the University of Colorado. He is a diplomate of the American Board of
Radiology in Diagnostic Radiological Physics and Nuclear Medical Physics.
Dr Johnson has served on AAPM Task Group No 7, Radionuclide Therapy
and Data Acquisition Methods, since its inception. His research interests
include activity quantification, mathematical modelling, image registration,
image fusion, image segmentation, and Monte Carlo radiation transport.
Amin I Kassis, PhD

Dr Amin I Kassis, Associate Professor of Radiology at the Harvard Medical
School and Director of Radiation Biology, is a leader in the field of radiation
biology. A major objective of his research has been an understanding of the
implications of densely ionizing radiations (Auger electrons and particles)
emitted by diagnostic and therapeutic radionuclides. He has defined and
established the biophysical relationship between the intracellular localization
of low-energy electron-emitting radionuclides and the biological conse-
quences of the resulting microdistribution of energy and has demonstrated
the limitations of conventional MIRD dosimetry in estimating radiation
risks to the patient. Another focus of his research has been the development
of radionuclide carrier systems suitable for the precise delivery of diagnostic
and therapeutic radioactive moieties to cancerous cells.
Web: http://www.jpnm.org/
Katherine S Kolbert, MS
Katherine Kolbert has a master’s degree in computer science from the Courant
Institute at New York University, New York, and is currently with the Depart-
ment of Medical Physics at Memorial Sloan-Kettering Cancer Center in New
York City. She has developed a three-dimensional patient-specific dosimetry
software package as well as other image analysis tools useful to the nuclear
medicine–physics community. Copies of her software packages have been
distributed to researchers around the world. She is also an author and contri-
butor to numerous peer-reviewed papers and abstracts.
Web: http://www.ski.edu/
356
Kenneth F Koral, PhD

Ken Koral received his BS degree in physics from Case Institute of
Technology and his PhD in nuclear physics from its successor, Case Western
Reserve University, Cleveland, Ohio. His dissertation advisor was Philip R
Bevington, author of Data Reduction and Error Analysis for the Physical
Sciences. Ken started his research career in medical physics during a year’s
postdoc in the United Kingdom. William H Beierwaltes recruited him to
the University of Michigan Medical Center where his current title is Senior
Research Scientist. Long-time collaborators and friends are Les Rogers
and Neal Clinthorne. Ken’s main research goal has been activity quantifica-
tion in nuclear-medicine single-photon emission computed tomography.
Assisted by Jeff Fessler and Yuni Dewaraja, he is the principal investigator
of a US National Cancer Institute grant entitled ‘Techniques for calculating
tumor dosimetry from imaging’. Ken belongs to a family consisting of
himself, his wife Mary and their three adopted children, one each from
Viet Nam, India and South Korea. Those children have children: Maekong,
Dante and Meagen Ann. They all enjoy movies, reading and hiking. Ken is a
member of the Society of Nuclear Medicine and of the American Association
of Physicists in Medicine.
Web: http://www.rad.med.umich.edu/
Cheuk S Kwok, PhD

Cheuk S Kwok received his PhD in Radiation Biophysics, Council for
National Academic Awards, England, in 1979. He worked initially as a
Medical Physicist with the Department of Clinical Physics and Bioengineer-
ing, West of Scotland Health Boards, Scotland, for about 2 years and then
worked until the late 1990s as a Medical Physicist with the Ontario Cancer
Treatment and Research Foundation, Ontario, Canada. Academic affiliation
was with the Departments of Radiology and Physics, McMaster University,
Ontario, Canada. He currently works as an Associate Professor with the
Department of Optometry and Radiography, Hong Kong Polytechnic
University and will work as a Senior Research Physicist, Department of
Radiation Oncology, City of Hope National Medical Center, California,
from September 2002. His current research is related to antibody-based
targeted therapy of cancer and combination of specific traditional Chinese
medicine with biological agents in cancer prevention and treatment.
Dr Kwok has been awarded 26 peer-reviewed research grants at local and
international levels and has published more than 50 papers in international
refereed journals, conference proceedings and monographs.
Web: http://www.cityofhope.org/
Biosketches 357
Michael Ljungberg, PhD

Michael Ljungberg began his studies in physics, mathematics and radiation
physics in 1979 at the Lund University, Sweden, and received a BSc degree
in physics in 1983. He started his research project in the Monte Carlo field
with a project aimed at developing a Monte Carlo code for calibration of
whole-body counters. This track developed into a more general Monte
Carlo program, SIMIND, for simulation of nuclear medicine imaging and
SPECT. This is a program that today is internationally recognized and used
by several groups. Parallel with his continuing development of SIMIND, he
started working in 1985 with quantitative SPECT and the problem of attenua-
tion and scatter. He graduated for a PhD degree in 1990 and received a
research assistant position at the Department of Radiation Physics, Lund
University, where he continued work on the SIMIND code and on quantita-
tive SPECT and in particular on 131 I imaging. In 1994, he became an associate
professor at Lund University. His current research includes an extensive
ongoing project in oncological nuclear medicine, developing methods based
on quantitative SPECT and Monte Carlo absorbed dose calculations for
patient-specific dosimetry and the development of co-registration methods
for an accurate three-dimensional dose planning scheme for internal radio-
nuclide therapy. In Lund, he is also involved in undergraduate education of
medical physicists, course developments and supervising of PhD students.
For several years, Dr Ljungberg has also worked as a certified medical
physicist in diagnostic nuclear medicine at Helsingborg Hospital.
Web: http://www.radfys.lu.se/
John W Poston Sr, PhD, Professor

Since 1985, Dr John W Poston Sr has been a Professor in the Department of
Nuclear Engineering at Texas A&M University in College Station, Texas. He
received his PhD in Nuclear Engineering from the Georgia Institute of
Technology in 1971. Professor Poston has served as President of the Health
Physics Society (HPS) and was elected as a Fellow of this Society in 1987.
He was elected Fellow of the American Nuclear Society (ANS) in 1996. In
the autumn of 2001, he was elected Fellow of the American Association for
the Advancement of Science (AAAS), 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
for 12 years and in 2002 was elected to Honorary Membership (lifetime) on the
Council. His areas of expertise include internal dose assessment as well as
external dosimetry of mixed radiation fields.
358
John Roeske, PhD

John C Roeske obtained his undergraduate degree in Physics and Mathe-
matics from DePaul University in Chicago. He subsequently attended the
University of Chicago and received his doctorate in Medical Physics in
1992. His dissertation research focused on the dosimetry of radiolabelled
antibodies for the treatment of ovarian cancer. Following graduation, Dr
Roeske joined the faculty of the Department of Radiation and Cellular
Oncology at the University of Chicago, and is currently an Associate
Professor. His research has concentrated on the microdosimetry of
particles as well as the dosimetry of targeted therapy. His current interests
include intensity-modulated radiotherapy and functional imaging.
Web: http://www.radonc.uchicago.edu/
Ruth Shefer, PhD

Dr Ruth Shefer is president and co-founder of Newton Scientific, Inc. in
Cambridge, Massachusetts, a company specializing in the development of
particle accelerators and radiation sources for medicine and industry. Her
research interests include the development of specialized neutron and X-
ray sources for radiation therapy and diagnostic applications, and the
development of accelerators and targets for radioisotope production. She
received a BA in Physics from the University of Pennsylvania and a PhD
in Physics from the Massachusetts Institute of Technology.
Web: http://www.newtonscientificinc.com/
Michael Stabin, PhD

Dr Stabin’s areas of specialty are the dosimetry of internal emitters and
radiation protection. He worked for 15 years at the Radiation Internal
Dose Information Center in Oak Ridge, Tennessee, and currently works at
Vanderbilt University in Nashville, Tennessee. The emphasis of his current
research is on radiation dosimetry for nuclear medicine patients. He works
as an associate editor for several journals and is active in the Health Physics
Society and the Society of Nuclear Medicine.
Web: http://www.doseinfo-radar.com/
Sven-Erik Strand, PhD

Professor Sven-Erik Strand was born 1946, received his MSc in 1972, his
PhD in 1979 and his Medical Bachelor in 1981. He became Associate
Biosketches 359
Professor in 1981 and full Professor in 1995. Since 1981 he has been a lecturer
at the University of Lund and teaching director at the Department of
Radiation Physics. Since 1997 he has been director of the Jubileum Institute.
From September 1991 until December 1992 he spent a sabbatical year at the
Nuclear Medicine Department, Brookhaven National Laboratory. He has
published about 160 regular articles, four book chapters and holds two
patents. He has served as member of various international committees, such
as the ICRU task group on internal dosimetry, the AAPM Nuclear Medicine
Task Groups on dosimetry, and the EANM task group on radionuclide
therapy dosimetry. He has served as reviewer for peer-written manuscripts
in several journals and organized international symposia and conferences.
Sven-Erik Strand’s research activities are in the field of systemic radiation
therapy physics including: registration of time-sequence three-dimensional
pre-therapy radionuclide images from SPECT/PET with CT/MRI anatomical
information for improved image quantification and to provide the basis for a
three-dimensional radiotherapy planning method, the application of radio-
activity quantification methods to determine tumour and non-target tissue
radionuclide uptake; the use of regional radionuclide pharmacokinetics from
time course quantitative SPECT/PET images, Monte Carlo methods to
produce three-dimensional radiation dose distributions. He also participated
in the EMERALD project and received the Bergman-Loevinger award in
2002 for his research in the field of internal dosimetry.
Web: http://www.radfys.lu.se/
Jacquelyn C Yanch, PhD

Jacquelyn Ciel Yanch earned her PhD in Physics from the University of
London in 1988. In 1989 she joined the faculty of the Nuclear Engineering
Department at the Massachusetts Institute of Technology with a secondary
appointment in MIT’s Whitaker College of Health Sciences and Technology.
Research in her laboratory, the MIT Laboratory for Accelerator Beam
Applications, centres around medical and biological applications of low-
energy accelerators including the development of novel therapies for
treatment of disease and the development of a charged particle microbeam
for single cell, single particle irradiations.
Web: http://web.mit.edu/laba/www/
Pat Zanzonico, PhD

Dr Pat Zanzonico received his PhD from the Cornell University Graduate
School of Medical Sciences in 1972. He is currently Associate Attending
360
Physicist and Associate Laboratory Member at Memorial Sloan-Kettering

Cancer Center and Manager of the Center’s Nuclear Medicine Research
Laboratory. Dr Zanzonico is also Associate Professor of Physics in
Radiology at Weill-Cornell Medical College. He is a member of the Editorial
Board and past Associate Editor of the Journal of Nuclear Medicine, a
Member of the National Council on Radiation Protection and Measurement
(NCRP), and Secretary of the Nuclear Medicine Section of the New York
Academy of Medicine. He is actively involved in biomedical research on
immune effector-cell trafficking, patient-specific dosimetry for radionuclide
therapies, radionuclide-based methods for detecting and localizing tumour
hypoxia, and small-animal imaging.
Index
Numbers in bold refer to figures and those in italics refer to tables.
3D-ID, 148, 249, 250 class I, 14, 139

class II, 14
Absorbed dose, 28, 39, 41, 46, 61, 66, 68, 69, 74, Collimators, 34, 61, 71
85, 86, 87, 88, 89, 90, 92, 96, 97, 98, 99, cone beam, 42
100, 108, 111, 112, 114, 115, 116, 120, 126 design, 31, 37, 62
Absorbed fractions, 38, 110 efficiency, 18, 63
Activity distribution, 58, 63, 66, 101 hole, 37, 268
particles, 15, 30, 175, 176, 179, 181, 184, lead, 267, 26
189, 190, 204, 205, 206, 209, 219 high-energy, 37, 62, 267
Analytic reconstruction, 71 medium energy, 272
Annihilation photons, 74, 75, 76 parallel-hole, 36, 37
Attenuation, 31 penetration, 37, 61, 62, 72, 262
Chang, 70 pinhole, 37
coefficients, 9, 68, 76, 241 scatter, 269, 273
non-uniform, 31 septa, 62, 264, 268
Auger electrons, 87, 141, 164, 176, 181, 184, Computational methods, 84, 276, 311
185, 191, 213 Coster–Kronig, 185, 213
Correction factors, 290
Backscattering, 393 Count rate, 72, 264
Bias, 18, 20, 42, 66, 70 Cumulated activities, 91, 92, 96, 100, 234
Bone, 44, 56, 69, 110, 118 Cross section library, 12, 242
cortical, 39, 120 XCOM, 10, 11
trabecular, 120, 125 PHOTX, 9, 10
Bone-marrow, 59 PETSIM, 9, 10, 11
Boron neutron capture therapy, 29, 324, 327 GEANT, 9, 10, 140, 141
137
Boron neutron capture synovectomy (BNCS), Cs, 35, 278
324, 326 CT images, 72, 73, 123, 147
Brachytherapy, 144, 171 Hounsfield units, 69
Brain, 30, 186
Blood flow, 12, 46 Dead time, 56, 66, 75
Detection, 13, 19, 31, 75, 133
Centroids, 12, 115 forced, 71
Charged particles, 15, 87, 142, 182 Detectors, 304, 342, 343, 344
bremsstrahlung, 13, 14, 40, 41, 113 BaF2, 34
361
362 Index
Detectors (contd) Klein–Nishina, 12, 241, 242

BGO, 7, 8, 33, 34 Kernels, 36, 39, 46, 101, 148, 158
block detectors, 34
CsF, 34 Light output, 275
energy resolution, 13, 34 Luminescence, 275
NaI, 230
NaI(Tl), 34, 70, 230 MABDOS, 90, 121, 147, 232, 233
Diagnostic imaging, 22, 280 MCNP, 3, 9, 43, 123, 139
DICOM, 252 MCNP-4B, 30, 39
Directional cosines, 205 MIRD, 28, 38, 39, 56, 84, 89, 91, 92, 93, 95,
DOSE3D, 121, 147 110, 112, 123
Dosimetry, 14, 28 MIRDOSE, 90, 115, 116, 148
external dosimetry, 139 Mechanical, 187, 215, 310, 331
internal dosimetry, 33, 39, 45, 84 Microdosimetry, 85, 202, 203, 205
Dual-isotope, 71 Monoclonal antibodies, 28, 147, 188
Dynamic allocation, 236 Monte Carlo codes, 1, 7, 15, 36, 38, 133, 135,
139, 140
EGS4, 9, 12, 14, 21, 30, 40 event-based, 152
PEGS, 139 photon history, 13
PRESTA, 293, 298 Monte Carlo methods, 3, 4, 22, 34, 40, 43, 44,
EGSnrc, 21, 40 66, 135, 203
Eidolon, 36 analogue, 16, 19
Electron, 12 modelling, 9, 22, 31, 34, 38, 44
energy, 14, 116, 119, 136, 167, 216, 217 simulations, 7, 29, 34, 36, 38, 41
interaction, 14, 293, 298 Monte Carlo techniques, 1, 2, 22, 28, 29, 31,
transport, 1, 7, 13, 14, 18, 22, 39, 115 41, 110, 249, 288
Emission, 10, 18, 30, 31 MORTRAN, 139
Energy distribution, 34, 40, 135, 161 Magnetic resonance imaging, 126, 302
pulse-height distribution, 267 Maximum-Likelihood-Expectation-
Energy resolution, 13, 34 Maximization (MLEM), 67, 68
ETRAN, 14, 40 MRI images, 57, 116, 148
18
F-Fluorodeoxyglucose (FDG), 37 Neutrons, 2, 30, 87, 124
FORTRAN, 5, 139, 234 beam, 30, 43, 324, 327, 328, 329
Fortran-90, 264 Noise, 42, 65, 67, 68, 70, 72, 74
FWHM, 34, 36 Poisson, 18, 203, 204, 205, 219
Non-homogeneous, 70, 71, 274
Gamma rays, 56, 61, 62, 69, 111 Nonlinear, 67, 246, 298
Gaussian, 13, 63 Nonuniform, 142, 189, 209, 212, 220
GEANT, 9, 10, 141
15
Geometrical efficiency, 18 O-water, 78
Optical coupling, 34
Heart, 42, 116, 121, 179 Ordered Subsets Expectation Maximization
Human anatomy, 90, 98, 121, 249 (OSEM), 63, 64, 68
ORNL phantoms, 117, 118, 120, 121
131
I, 36, 38, 43, 45, 68, 96, 97, 98, 117, 119,
161, 163, 164, 165, 166, 186 Parallel computers, 6, 150, 151, 152, 264
111
In, 117, 161, 163, 166 Parallel processing, 31, 150
Interfile, 252 Path length, 21, 159, 167, 205, 206, 251, 252,
Intravascular radiation therapy, 29 290
Iterative reconstruction, 67, 70, 76, 77 PET scanners, 13, 35, 36
ITS, 9, 40, 43, 139, 141 design, 34
Index 363
detector module, 264 Probability density functions, 2, 3, 33, 134

scintillation crystal, 61 Probability distributions, 13, 134, 241, 242,
sensitivity, 34, 74, 75 313
time-of-flight, 34 Pulse sequence, 296
PETSIM, 9, 10, 11
Phantoms, 39, 65, 109, 110, 111, 114, 115, 265 Quality factor, 88, 89
3D Hoffman, 12, 36, 37
MCAT, 265 Radiation protection, 29, 42, 43, 84, 87, 88,
NCAT, 265 102, 136, 141, 202, 290, 291, 297, 343
anthropomorphic, 39, 111 Radiation therapy, 22, 29, 143, 321
female organs, 137 Radiation transport, 2, 6, 7, 12, 22, 28, 31, 39,
mathematical, 28, 114, 115 109, 121, 123, 124, 139, 141, 144, 233,
shape-based, 110 243, 293, 294, 295, 296, 311
voxel-based, 123, 149, 264, 265 Radioactive decay, 90, 101, 234, 241
VIP-Man, 39, 40 Radiobiology, 214
Zubal phantom, 265 Radioimmunotherapy, 28, 78, 277, 278, 280,
Photomultipliers, 13 288, 304
Photomultiplier tubes (PMTs), 34 Radionuclide dosimetry, 84, 91, 92, 93, 158,
Photon fluence, 289 160, 168, 171, 228, 233, 234, 246, 275, 276
Photon histories, 112, 150, 243 Radionuclide therapy, 29, 71, 84, 98, 125, 144,
history weight, 18 182, 193, 221, 250, 252, 262, 275, 276,
Photon history generator (PHG), 264 277, 280, 300, 311
Photon interactions, 7, 14, 112, 113, 289 Random coincidences, 75
bound electrons, 9 Random numbers, 1, 3, 4, 5, 6, 16, 29, 33, 134,
coherent scattering, 7, 13, 139 151
Compton scattering, 14, 70, 142, 268 Reference man, 56, 89, 98, 112, 118, 120, 124,
cross section tables, 245 126, 136, 142, 167, 232, 234, 236, 237,
incoherent scattering, 7, 12, 17 242, 295
lookup tables, 241, 242, 254, 313 Reflector, 30, 329, 330, 331, 332, 333, 334,
pair production, 9, 134, 136, 139, 242 338, 339
photoelectric absorption, 12, 13, 139
photoelectric effect, 12 Sampling, 1, 3, 13, 16, 17, 18, 20, 21, 28, 33,
Photon transport, 7, 12, 18, 112, 113, 116, 40, 137, 162, 164, 205, 280, 292, 293, 297,
121, 125, 135, 142, 162, 233, 236 298, 313
heterogeneous media, 264 direction, 242
Photopeak, 34, 74, 265 selection of the type of photon interaction,
Pile-up, 264 134
Planar imaging, 31, 57, 257, 258, 263, 288 Sampling techniques, 33
Point-spread function (PSFs), 267 collimator, 268, 69, 273
exponential, 38 Compton, 2, 13, 34, 70, 142, 267
Gaussian, 63 Compton window, 74
geometric component, 267 cumulative distribution function, 16, 17
penetration component, 267, 272 direct method, 17
scatter component, 267, 272 effective scatter (ESSE), 71
radially symmetric, 38 energy threshold, 34, 139
simulated, 38 mixed methods, 17
Position blurring, 13 multiple scattering, 14, 20, 21, 142, 161
Positrons, 47, 85, 141, 142, 241 object, 267
emitters, 39, 46, 47 patient, 61, 230
range, 13 rejection method, 17
Positron emission tomography (PET), 7, 31, scatter, 34, 35, 36
57 scattered photons, 12, 16, 34, 61, 70, 162
364 Index
Sampling techniques (contd) reconstruction, 66

single scattering, 14 resolution, 64, 71, 75
subtraction, 70 simulation, 67, 152, 269
transmission images, 34, 60, 61, 68, 75, 76, transmission, 34, 68, 76, 264
77, 264 Statistical uncertainty, 3
transport equation, 31, 161 Statistical variations, 161, 233
Scatter correction methods, 36, 61, 70 Stratification, 19
convolution, 70
99m
dual-energy window, 70 Tc, 34, 67, 117, 119, 163, 166, 203, 214,
energy-based, 36, 70 266, 267, 269, 302
inverse Monte Carlo, 36 Therapy, 30, 36, 47, 55, 56, 59, 61, 62, 66, 77,
iterative reconstruction, 77 96, 118, 135, 143, 144, 145, 146, 152, 153,
model, 77 181, 182, 184, 187, 204, 218, 220, 273,
Monte Carlo method, 71 278, 279, 296, 299, 300, 301, 310, 311,
response functions, 71 313, 322, 325, 331, 332, 333, 344
triple-energy window, 70 Treatment planning, 29, 43, 45, 46, 47, 59, 75,
Wiener filtering, 71 84, 96, 133, 148, 153, 299
201
Scatter fractions, 35, 36 Tl, 117, 214
Scatter response functions, 70 Tomographic imaging, 46, 57, 58, 71, 72, 101,
Scintillation camera, 34, 36, 59, 72, 231, 262, 265
264, 267, 268 Transmission scanning, 34
sensitivity, 37, 46, 74, 268 single-photon sources, 34
Segmentation, 55, 57, 58, 64, 71, 76, 121, 122, Transmission simulation, 34
124, 233, 237, 245, 265 Tumour, 30, 36, 38, 55, 56, 57, 59, 60, 61, 64,
Shielding, 13, 20, 43, 44, 234, 312, 325, 340, 65, 67, 72, 73, 77, 78, 84, 90, 96, 97, 98,
341, 342, 344 100, 101, 144, 145, 147, 148, 179, 186,
SIMIND, 264, 269 186, 187, 189, 208, 218, 220, 228, 229,
SimSET, 264 231, 232, 237, 246, 249, 251, 262, 267,
Single-photon emission computed 269, 271, 299, 300, 312
tomography (see SPECT), 31
Software, 36, 47, 115, 121, 123, 125, 133, 138, User interface, 236
141, 146, 148, 149, 150, 228, 229, 234,
245, 246, 249, 250, 251, 252, 258, 264, Validation, 29, 34, 165, 263, 264, 267, 269,
265, 273, 281, 292, 348 274, 335
phantoms, 265, 281 Variance, 3, 16, 20, 22, 57, 64
Source geometry, 43, 211, 291 Variance reduction, 1, 3, 18, 20, 29, 36, 71,
Spatial resolution, 13, 34, 37, 62, 66, 68, 71, 140, 150, 329
74, 258, 259, 266, 271 forced detection, 71
collimator, 36, 38, 62, 65, 264 Russian roulette, 20, 135
FWHM, 36 splitting, 20
system resolution, 42, 267, 271 stratification, 19
Specific absorbed fraction, 28, 38, 39, 91, 113, Vector processors, 151
115, 116, 117, 120, 158, 159, 251, 289, vectorizable, 4
290, 292 vectorization, 3
SPECT, 31, 32, 34, 36, 38, 42, 45, 57, 58, 59, Vessel wall, 311, 313, 314, 317
60, 62, 66, 67, 68, 69, 70, 71, 72, 73, 74,
75, 76, 77, 79, 148, 152, 231, 251, 254, Weight factor, 20, 89
255, 257, 262 Whole-body imaging, 59, 265
contrast, 66, 71, 72, 89, 267, 269
133
myocardial perfusion studies, 42 Xe, 117
projection data, 42, 45, 56, 66, 67 X-rays, 87, 141, 273
quantification, 64, 274 XCOM, 10, 11

Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine Series in Medical Physics and Biomedical Engineering (IoP, 2003)

Uploaded by

Copyright:

Available Formats

Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine Series in Medical Physics and Biomedical Engineering (IoP, 2003)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Therapeutic Applications of Monte Carlo Calculations in Nuclear Medicine Series in Medical Physics and Biomedical Engineering (IoP, 2003)

Uploaded by

Copyright:

Available Formats

THERAPEUTIC APPLICATIONS OF MONTE CARLO

CALCULATIONS IN NUCLEAR MEDICINE

Other books in the series

Forthcoming titles in the series

George Sgouros, PhD

Institute of Physics Publishing

All rights reserved. No part of this publication may be reproduced, stored in

British Library Cataloguing-in-Publication Data

Commissioning Editor: John Navas

Published by Institute of Physics Publishing, wholly owned by The Institute

Typeset by Academic+Technical Typesetting, Bristol

1 THE MONTE CARLO METHOD: THEORY AND

2 MONTE CARLO TECHNIQUES IN NUCLEAR

2.2. Monte Carlo techniques in medical radiation physics 29

3 MEDICAL IMAGING TECHNIQUES FOR RADIATION

3.6. Treatment planning 78

4 COMPUTATIONAL METHODS IN INTERNAL RADIATION

5 MATHEMATICAL MODELS OF THE HUMAN ANATOMY 108

5.4.3. Comparisons with stylized mathematical models 124

6 MONTE CARLO CODES FOR USE IN THERAPEUTIC

7 DOSE POINT-KERNELS FOR RADIONUCLIDE

7.4.2. Absorbed fractions on a millimetre scale 169

8 RADIOBIOLOGY ASPECTS AND RADIONUCLIDE

9 MICRODOSIMETRY OF TARGETED RADIONUCLIDES 202

10 THE MABDOSE PROGRAM FOR INTERNAL

10.3. Historic development of MABDOSE 233

11 THE THREE-DIMENSIONAL INTERNAL DOSIMETRY

12 EVALUATION AND VALIDATION OF DOSE

12.3. Monte Carlo simulation of imaging systems 263

13 MONTE CARLO METHODS AND MATHEMATICAL

13.3.3. The Eckerman cortical bone model 297

14 MONTE CARLO MODELLING OF DOSE DISTRIBUTIONS

15 THE MONTE CARLO METHOD AS A DESIGN TOOL IN

Habib Zaidi, PhD

George Sgouros, PhD

Pedro Andreo, PhD, DSc, Professor

Manuel Bardies, PhD

Wesley Bolch, PhD

Lionel Bouchet, PhD

A Bertrand Brill, PhD

Yuni Dewaraja, PhD

Dennis Duggan, PhD

Keith Eckerman, PhD

David Gierga, PhD

John Humm, PhD

Timothy K Johnson, PhD

Amin I Kassis, PhD

Kenneth F Koral, PhD

Cheuk S Kwok, PhD

Michael Ljungberg, PhD

John W. Poston Sr, PhD, Professor

John Roeske, PhD

Ruth Shefer, PhD

Michael Stabin, PhD

Sven-Erik Strand, PhD, Professor

Jacquelyn Yanch, PhD

Pat Zanzonico, PhD

implementations of the Monte Carlo approach in the special cases of micro-

The Monte Carlo method: theory and