UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Monte Carlo dosimetry of total body irradiation Burns, Levi 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2018_may_burns_levi.pdf [ 5.8MB ]
Metadata
JSON: 24-1.0365701.json
JSON-LD: 24-1.0365701-ld.json
RDF/XML (Pretty): 24-1.0365701-rdf.xml
RDF/JSON: 24-1.0365701-rdf.json
Turtle: 24-1.0365701-turtle.txt
N-Triples: 24-1.0365701-rdf-ntriples.txt
Original Record: 24-1.0365701-source.json
Full Text
24-1.0365701-fulltext.txt
Citation
24-1.0365701.ris

Full Text

Monte Carlo Dosimetry of Total Body IrradiationbyLevi BurnsB.Sc (Hons.), Queen’s University, 2016A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of ScienceinTHE FACULTY OF GRADUATE AND POSTDOCTORALSTUDIES(Physics)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)April 2018c© Levi Burns, 2018AbstractThe total body irradiation (TBI) technique at the Vancouver Cancer Centre uses asweeping Cobalt-60 beam with patient-specific lung compensators and a stationaryflattening filter, with the patient lying in supine and prone positions during eachfraction. The dose calculations for this technique are limited to the dose deliveredto a point at patient mid-separation at the level of the umbilicus, and the mean doseto the body must fall within a 10% tolerance of prescription. While the currenttechnique is effective, it has been in place for over 23 years, and detailed dosedata to critical organs is necessary before the technique can be upgraded to a moreconformal technique that would offer an improved patient experience.For this purpose, a Monte Carlo simulation technique has been developed andapplied to collect organ dose data from TBI treatments based on retrospective dataof patients recently treated with 12 Gy in 6 fractions. 20 patients, including adultsand pediatric patients, are simulated by constructing Monte Carlo phantoms in eachof the supine and prone positions based on planning CT images. The supine andprone dose distributions are summed with a deformable registration tool and thedoses to the lungs, kidneys, thyroid, and liver are analyzed as well as the dosedelivered to the body. It is determined that while all doses fall within prescription,there is a trend where smaller patients receive lower mean body doses and vice-versa (mean body dose range: 10.93-12.01 Gy). For most patients, the lungs andliver consistently receive doses below the mean body dose, and the thyroid andkidneys consistently receive higher doses than the mean body dose.This thesis presents an overview of the background physics and biology ofTBI as well as a comprehensive survey of different techniques described in theliterature. The Monte Carlo simulation technique used for the retrospective studyis described with calibration, validation, and optimization details, and the organdose results of the retrospective study are shown.iiLay SummaryTotal body irradiation (TBI) is a treatment used primarily to prepare a patient for abone marrow transplant. While the TBI technique at the Vancouver Cancer Centreis effective, an upgrade is being considered based on recent advances in radia-tion therapy technology that would improve the patient experience. Before such amodern technique can be installed, organ dose data from the current technique arerequired at a more detailed level than what has been needed clinically to date. Thisthesis presents a method of collecting this organ dose data by performing com-puter simulations of the TBI treatments of recently treated patients. It is shownthat there are consistent organ dose patterns across the patient population, and thiswill inform the design of the next generation TBI technique.iiiPrefaceThis thesis was completed by the author, Levi Burns, at the Vancouver CancerCentre, and represents a continuation of work done by a previous UBC doctoralstudent, Dr. Tony Teke. Use of patient data was cleared by the institutional researchethics board (REB #H17-02276).A report based on preliminary versions of Chapters 4 and 5 has been acceptedfor publication:L Burns, T Teke, I A Popescu, C Duzenli. Monte Carlo dosimetry of organdoses from a sweeping-beam total body irradiation technique: feasibility and firstresults. Proceedings of the World Congress on Medical Physics and BiomedicalEngineering. Accepted March 2018, in press.Further publications are in preparation which will build on the results presentedin Chapters 4 and 5.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . xAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Background biology and physics . . . . . . . . . . . . . . . . . . . . 32.1 Bones and blood . . . . . . . . . . . . . . . . . . . . . . . . . . 32.1.1 Stem cells . . . . . . . . . . . . . . . . . . . . . . . . . . 32.1.2 Hematopoiesis and bone marrow . . . . . . . . . . . . . . 42.1.3 Blood cells . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2.1 The hallmarks . . . . . . . . . . . . . . . . . . . . . . . . 72.2.2 Hematological malignancies . . . . . . . . . . . . . . . . 92.2.3 Hematological malignancies in Canada . . . . . . . . . . 102.2.4 Bone marrow transplants . . . . . . . . . . . . . . . . . . 12v2.3 Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.3.1 Light-matter interactions . . . . . . . . . . . . . . . . . . 132.3.2 Radiobiology and dose . . . . . . . . . . . . . . . . . . . 162.3.3 Radioactivity and Cobalt-60 radiotherapy units . . . . . . 183 Total body irradiation (TBI) . . . . . . . . . . . . . . . . . . . . . . 213.1 Historical development . . . . . . . . . . . . . . . . . . . . . . . 213.2 Modern TBI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.2.1 Fractionation, dose, and dose rate . . . . . . . . . . . . . 233.2.2 Worldwide heterogeneity in techniques . . . . . . . . . . 243.2.3 Advances in conventional TBI . . . . . . . . . . . . . . . 263.2.4 Toxicities . . . . . . . . . . . . . . . . . . . . . . . . . . 293.2.5 Helical tomotherapy and VMAT . . . . . . . . . . . . . . 333.2.6 Total marrow irradiation and total lymphatic irradiation . . 373.2.7 Treatment planning systems and in-vivo dosimetry . . . . 393.2.8 Monte Carlo dosimetry . . . . . . . . . . . . . . . . . . . 413.3 TBI at the Vancouver Cancer Centre . . . . . . . . . . . . . . . . 434 Monte Carlo dosimetry of the TBI technique in Vancouver . . . . . 494.1 Introduction to Monte Carlo . . . . . . . . . . . . . . . . . . . . 494.1.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.1.2 Monte Carlo simulations in radiation oncology physics . . 504.1.3 EGSnrc . . . . . . . . . . . . . . . . . . . . . . . . . . . 534.2 Phantom production for TBI simulations . . . . . . . . . . . . . . 544.3 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584.3.1 Cobalt source model . . . . . . . . . . . . . . . . . . . . 584.3.2 Simulation procedure and de-noising . . . . . . . . . . . 584.3.3 Optimization . . . . . . . . . . . . . . . . . . . . . . . . 604.3.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . 614.3.5 Validation measurements . . . . . . . . . . . . . . . . . . 644.4 Deformable registration of dose distributions . . . . . . . . . . . 68vi5 Retrospective dose analysis of a TBI patient cohort . . . . . . . . . . 725.1 Purpose and aim . . . . . . . . . . . . . . . . . . . . . . . . . . . 725.2 Methods: patient selection and contouring . . . . . . . . . . . . . 735.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . 766 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 836.1 Ongoing investigations . . . . . . . . . . . . . . . . . . . . . . . 836.2 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88viiList of TablesTable 2.1 Incidence and mortality statistics of hematological cancers inCanada . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10Table 4.1 Abridged results of an optimization study . . . . . . . . . . . . 61Table 4.2 Measurement and simulation results from the validation mea-surements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69Table 5.1 Aggregate organ dose results across the 20 simulated patients . 77Table 5.2 Standard deviation data of the doses delivered to each organ . . 77viiiList of FiguresFigure 2.1 Diagram of a typical long bone in the human body . . . . . . 5Figure 2.2 The hierarchies of different types of blood cells . . . . . . . . 6Figure 2.3 Relative probabilities of photon interactions in medical physics. 16Figure 2.4 The Cobalt-60 unit at the Vancouver Cancer Centre . . . . . . 19Figure 3.1 Schematic representation of a TBI treatment at the VCC . . . 45Figure 3.2 CT images of a TBI patient with radio-opaque markers . . . . 46Figure 4.1 Flowchart describing the basic elements of a Monte Carlo sim-ulation of charged particle transport . . . . . . . . . . . . . . 52Figure 4.2 A section of an egsphant file modelling a lung compensator . . 54Figure 4.3 Slices of a Monte Carlo phantom of a patient . . . . . . . . . 57Figure 4.4 Isodoses from three dose distributions produced in an opti-mization study . . . . . . . . . . . . . . . . . . . . . . . . . 62Figure 4.5 Setup of the calibration measurement . . . . . . . . . . . . . 64Figure 4.6 Simulated dose distribution from the calibration measurement. 65Figure 4.7 Setup used for the validation measurements . . . . . . . . . . 67Figure 4.8 Simulated dose distribution for the validation measurements. . 68Figure 4.9 Final results of a Monte Carlo simulated TBI patient treatment. 70Figure 5.1 Organ contours shown for a simulated patient. . . . . . . . . . 75Figure 5.2 Variation of TBI mean body dose with patient size for the 20patients. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78Figure 5.3 PDD curves for parallel-opposed beams at extended SSD . . . 79Figure 5.4 The homogeneity indices of the dose delivered to the body con-tour against body size . . . . . . . . . . . . . . . . . . . . . . 80Figure 5.5 Organ dose data plotted against mean body dose for lungs, kid-neys, liver and thyroid. . . . . . . . . . . . . . . . . . . . . . 81Figure 5.6 Dose-volume histogram for a simulated patient. . . . . . . . . 82ixList of AbbreviationsALL acute lymphoblastic leukemiaAML acute myeloblastic leukemiaBMT bone marrow transplantCLL chronic lymphoblastic leukemiaCML chronic myeloblastic leukemiaCT computed tomographyDVH dose-volume histogramGVHD graft-versus-host diseaseHI homogeneity indexHL Hodgkin’s lymphomaHSC hematopoietic stem cellHT helical tomotherapyIMRT intensity-modulated radiation therapyMC Monte CarloMOSFET metal oxide semiconductor field effect transistorMU monitor unitMVCT megavoltage computed tomographyNHL non-Hodgkin’s lymphomaOAR organ at riskxOSLD optically stimulated luminescent dosimeterPDD percent depth dosePOP parallel-opposed pairPTV planning target volumeRIC reduced intensity conditioningQA quality assuranceSSD source-to-surface distanceTBI total body irradiationTLD thermoluminescent dosimeterTLI total lymphatic irradiationTMI total marrow irradiationTMLI total marrow and lymphatic irradiationTPS treatment planning systemVCC Vancouver Cancer CentreVMAT volumetric modulated arc therapyxiAcknowledgementsI have been enormously fortunate throughout my graduate studies to be able tobrag to people about the amazing company I have had along the way. Betweenall of the physicists, physics assistants, students, and staff at BC Cancer, I wasnever deprived of mentorship, technical help, or entertainment as my degree movedalong, for which I am boundlessly grateful.There are several contributions to recognize specifically. I would like to thankConrad Yuen, Susan Zhang, and Tony Popescu for their support in bringing myproject to the finish line and beyond, as well as Claudia Mendez and Ron Horwoodfor their assistance with measurements, Nevin McVicar for his MIM wizardry,and Steven Thomas for his revisions and helpful comments on this thesis. To TonyPopescu and Tony Teke, thank you for your patience and enthusiasm in teaching methe ways of Monte Carlo and for your encouragement throughout the project. Mostimportantly, to Cheryl Duzenli for taking me on as a graduate student, for beingendlessly supportive of me in my pursuits, for creating opportunities for me, and foryour guidance, advice, physics wisdom, and open door; it has been a pleasure anda privilege to work with you. Lastly, to Byron Wilson and Youssef Ben Bouchta,for the brainstorming, life chats, and camaraderie in our grotto downstairs, and themany laughs that I look forward to continuing beyond our time as grad students.xiiChapter 1Introduction1.1 PurposeIn treating cancer with radiotherapy, the goal is to treat a target at a prescriptiondose while minimizing the dose delivered to organs at risk. Radiation oncology isa collaboration between skilled professionals including, but certainly not limitedto, radiation oncologists who select the planning target volume (PTV) and the pre-scription dose, radiation therapists who work directly with patients to prepare anddeliver their treatments, and medical physicists who ensure that planned treatmentswill correctly deliver the prescription dose to the target volumes, commission newtechniques, perform quality assurance, and carry out research to improve existingtreatments, among other responsibilities.Total body irradiation (TBI) is a radiotherapy treatment typically employed toprepare a patient for a bone marrow transplant. TBI is a unique treatment in that theentire body is the PTV, rather than a localized tumour or specific region of interest.Treating the entire body at a single prescription dose is a challenging task for manyreasons including the large size of a patient compared to the size of conventionalradiation field sizes, the heterogeneous tissue composition of the body, and the factthat the dose delivered by a radiotherapy beam decreases with depth as it travelsthrough the patient.The TBI technique at the Vancouver Cancer Centre (VCC) is effective andhas been in place for over 23 years. However, given the advances in radiationoncology physics in these years, it has become feasible to raise the quality of ourtechnique to a new standard with improved homogeneity of dose throughout thepatient and more flexibility to reduce the dose delivered to critical organs that donot require the full body prescription. The team at the VCC is considering such an1upgrade which will move from a sweeping Cobalt-60 beam to a technique usinga linear accelerator (linac). Before this transition can be made, a more detailedunderstanding of the doses delivered to each organ at risk (OAR) in the body isrequired from the current method.In this thesis, a Monte Carlo dosimetry technique is developed and appliedto simulate TBI treatments of a retrospective patient cohort. The resulting three-dimensional dose distributions will both serve as a basis for our upgrade and pro-vide a new quality assurance (QA) tool for the existing treatment.1.2 OverviewChapter 2 provides a brief introduction to topics in physics, anatomy, physiology,and medicine that are relevant to the research. It is written to answer the followingquestions to a non-specialist audience: what is cancer, what does it mean for ourblood to have cancer, and how does physics play a role in treating it?Chapter 3 concerns total body irradiation (TBI). Most of the chapter is a litera-ture review of modern TBI techniques, and section 3.3 details the current techniqueat the VCC. This chapter is written with the dual intention of situating the researchof this thesis in the literature and serving as a reference for the VCC clinical teamas the upgraded techniques continues to be developed.Chapter 4 describes the Monte Carlo method as it pertains to radiation oncol-ogy physics. This chapter explains the methodology of the research project in pro-ducing simulated organ dose data, including a brief optimization study and detailsof how our simulations are validated against dosimeter measurements.Chapter 5 presents a retrospective dose-volume analysis of organs at risk in TBIpatients treated recently at our centre, and is an illustration of how the designedMonte Carlo dosimetry technique can be applied. 20 patients were simulated withbody and organ doses presented. The ongoing and future work based on this projectare described in Chapter 6.2Chapter 2Background biology and physics2.1 Bones and blood2.1.1 Stem cellsIn 1665, the term “cell” was coined by Robert Hooke in his ground-breaking work,Micrographia, used to describe the microscopic units that make up a slice of cork.The following two centuries brought scientists closer and closer to the conclusionthat this subdivision into small units applies to living organisms as well. The of-ficial formulation of biological cell theory is considered to have occurred in 1838and 1839 by Matthias Jakob Schleiden and Theodor Schwann. Several years later,Rudolf Virchow built on these ideas and provided history with a famous line, omniscellula e cellula: All cells come from cells [1].More recent work allows 21st century scientists to be more specific. Organismsbegin as one cell that will undergo many divisions to become a full lifeform, andmost of these cells become highly specialized for a certain purpose; for example,muscle cells are very different from brain cells. A given specialized cell cannotgive rise to a different specialized cell by ordinary cell division. Stem cells, onthe other hand, can divide to give rise to many different kinds of specialized cellsthrough a process called differentiation. The division of stem cells is asymmetricsuch that the products include not only a new cell that will differentiate, but also anew stem cell to replace the mother stem cell [2].The first cell in a human is the zygote, a fertilized egg produced by a femaleegg cell and a male sperm cell. The zygote is the ancestor of all other cells inthe organism and thus sits at the highest level of the stem cell hierarchy, classified3as a totipotent stem cell [2, 3]. The next cells in the development line are calledpluripotent stem cells, which can divide into any type of human tissue but areunable to produce a full organism independently. Following these are multipotentstem cells, which can differentiate into different cell types of a given lineage. Onepopulation of multipotent stem cells in the human body, and one of the best-studiedtypes of stem cell, is the hematopoietic stem cell (HSC)1. Like all stem cells, HSCshave the capacity for self-renewal and give rise to a line of more specialized cells,these being the cells of the blood.We can now refine Virchow’s famous statement for the purposes of this the-sis: all blood cells come from hematopoietic stem cells. This process is calledhematopoiesis.2.1.2 Hematopoiesis and bone marrowHematopoiesis occurs throughout the yolk sac in the first few weeks of fetal life.From six weeks until six to seven months of gestation, the liver and the spleen arethe major blood-producing organs. Towards the end of pregnancy, the bone marrowbecomes the most important site of hematopoiesis, and becomes the only source ofnew blood cells after a few weeks past birth [5]. Thus, further discussion of bloodformation first requires a discussion of bones.An adult human skeleton has 206 bones [3]. Bone tissue can be classified aseither cortical bone, which is dense, solid, and forms the exterior portion of mostbones, or as trabecular bone (also known as spongy bone) which threads throughthe interior of the bones. About 80% of adult human bone is cortical, althoughthe ratio of cortical to trabecular bone varies widely within individual bones in thebody [6, 7].Internal cavities exist in most bones and these are usually filled with bone mar-row, which is a softer tissue [3, 6]. There are two types of bone marrow: yellowbone marrow, which contain fatty tissue that can serve as an energy source, andred bone marrow, which is where hematopoiesis takes place. The bone marrowprovides a microenvironment that is suitable for HSCs to divide, self-renew, anddifferentiate. Accordingly, the bones play a key role in regulating blood formation1There is some debate in the literature as to whether HSCs should be considered pluripotent inlight of recent research into their plasticity [4], but this is outside the scope of this discussion.4Figure 2.1: Diagram of a typical long bone in the human body. Blood cellsare formed in the red bone marrow. Image source: Anatomy and Physi-ology, OpenStax [3] (edited).[8]. A diagram of a typical long bone is given in Figure 2.1.In infants, almost all bones in the body contain blood-producing red marrowregions, but these regions are gradually replaced by yellow marrow regions withage. In adults, hematopoietic marrow exists only in the central skeleton and theproximal ends of the femurs and humeri (that is, the ends that are closer to thecenter of the body), and even these regions only consist of approximately 50% redmarrow. Under extraordinary circumstances, such as the destruction of bone mar-row by bone cancers, the spleen and liver can recommence their natal productionof blood cells through extramedullary hematopoiesis [3, 5].There are relatively few HSCs in the bone marrow cell population, roughly oneper 20 million nucleated cells, but each HSC is capable of producing one million5Figure 2.2: Different types of blood cells, shown in hierarchical order begin-ning with the asymmetric division of a hematopoeitic stem cell (HSC).The blood cells on the left hand side are of myeloid lineage, and on theright hand side, lymphoid lineage. Image source: Anatomy and Physi-ology, OpenStax [3].mature blood cells after twenty divisions. The types of blood cells that will arisefrom each division depend on a complex interplay of growth factors and extracel-lular signals received by each HSC from the cellular environment [5]. Each HSCdivision gives rise to either a myeloid stem cell or a lymphoid stem cell whichdifferentiate further as shown in Figure 2.2.62.1.3 Blood cellsIn an average adult, blood constitutes roughly 8% of the body mass [3]. Blood iscomprised of various cells and cellular fragments collectively known as the formedelements, as well as the fluid that suspends these elements, which is called plasma.The formed elements consist of platelets, erythrocytes (red blood cells), andleukocytes (white blood cells), as shown in Figure 2.2. Platelets, each with a life-time of about 10 days, make up less than 1% of the formed elements and are pri-marily involved in healing damaged tissue and preventing blood loss followinginjury to a blood vessel. Erythrocytes make up over 99% of the formed elementsand each cell has a lifetime of about 120 days. They are responsible for trans-porting inhaled oxygen from the lungs to nourish the other cells in the body, andfor transporting carbon dioxide produced as byproducts of bodily functions to thelungs for exhalation.Many types of leukocytes exist, and these are primarily involved in immunity.The eosinophils, neutrophils, and basophils are collectively referred to as granu-locytes. The granulocytes and monocytes follow a myeloid lineage. On the otherhand, lymphocytes (of lymphoid lineage) are labelled as either B lymphocytes orT lymphocytes depending on whether the precursor lymphoid stem cell matured inthe bone marrow (B cells) or in the thymus gland (T cells). While red blood cellsand platelets are confined within the vascular system, leukocytes are able to leavethe circulation to reach other tissues. Lymphocytes are the only leukocytes that canreturn to the circulation afterwards [3, 6].2.2 Cancer2.2.1 The hallmarksOne day, we imagine that cancer biology and treatment [...] willbecome a science with a conceptual structure and logical coherencethat rivals that of chemistry or physics.— Weinberg and Hanahan, The Hallmarks of Cancer, 2000 [9]Cancer has been recognized as a disease for millennia, with the earliest knowndescription of a breast tumour recorded in hieroglyphs by Imhotep, an ancient7Egyptian physician, dating to 2500 BC. In brief, cancer can be considered a diseaseof uncontrolled cell growth. But, given the thousands of different forms of cancerthat can affect virtually any part of the body, the symptoms of cancer vary vastly,and physicians for much of history were mystified by the complex illness. Littleprogress in cancer treatment was made for several thousand years.Fortunately, scientific advances in the 19th and 20th centuries began to producetreatments that would go on to meaningfully prolong the lives of patients of manycancer types, whether by cure or palliation. The advent of molecular biology inthe late 20th century in particular has given researchers an unprecedented wealth ofinformation about the disease at a microscopic level [10, 11].In 2000, a review article by Douglas Hanahan and Robert Weinberg was pub-lished that set the tone for discussions of cancer biology into the 21st century. Theyoutlined six “Hallmarks of Cancer” that govern the production and the growth oftumours, synthesizing an enormous amount of information collected about the dis-ease up to that point in time. Since then, further research, debate, and discussionhave identified two additional hallmarks, leading to a total of eight features that aremore descriptive of cancer than the simple “uncontrolled cell growth” [9, 10, 12]:1. Self-sustaining growth signalling; that is, cancer cells provide their own in-structions to grow, rather than being directed to do so externally2. Insensitivity to anti-growth signals that ordinarily suppress tumour growth3. Evasion of programmed cell death (apoptosis)4. Limitless replicative potential; normally, cells can only divide a certain num-ber of times5. Ability to build their own supply of blood (angiogenesis)6. Ability to metastasize; that is, to travel through the body and initiate a newcancer site elsewhere7. A rewiring of cellular metabolism processes, redefining how cancer cellsproduce their energy8. Evasion of immune system responses82.2.2 Hematological malignanciesRather than taking the form of a solid tumour like many other types of cancer,cancers of the blood begin as an affliction of blood-producing HSCs. From there,production of blood cells is altered. Hematological malignancies can be broadlysplit into three categories: leukemia, lymphoma, and myeloma. These can be splitfurther into many subcategories. Here, we go into detail for a subset of classifica-tions of blood disease encompassing the majority of cases.Leukemia can be either myelogenous or lymphoblastic, corresponding to a dis-ease of either too many myeloblasts or too many lymphoblasts, respectively. Theseare also divided into either acute or chronic disease, as the symptoms can eitherarise and progress quickly on the span of days or weeks, or slowly on the spanof months or years. Acute lymphoblastic leukemia (ALL) is most common inyoung children. Acute myeloblastic leukemia (AML) is the most common type ofleukemia in adults, with the median age of diagnosis being 65 years. Chronic lym-phoblastic leukemia (CLL) most often appears in patients in their seventies, wherethere are too many abnormal B lymphocytes. Chronic myeloblastic leukemia(CML) is of particular interest in the development of cancer research as a whole, asit represents the first time that a specific type of cancer was linked to a consistentchromosomal abberation: the Philadelphia chromosome, a translocation of chro-mosomes 9 and 22 [13, 14]. The chromosomal basis of cancer is described furtherin section 2.3.2. Recent research has identified leukemic stem cells which are re-sistant to traditional therapeutic approaches and may explain why some leukemiasrelapse [15].Lymphoma involves the lymphatic system, which includes the spleen, tonsils,lymph vessels that transport waste products and pathogens away from our cells, andlymph nodes that filter lymph fluid. It is traditionally classified as either Hodgkinor non-Hodgkin lymphoma (HL or NHL): while NHL has many forms and candevelop almost anywhere in the body, in either indolent or aggressive forms andin B or T lymphocytes, HL is well-defined by the presence of Reed-Sternbergcells, which are mature B cells that have become large, malignant, and have severalnuclei. HL typically develops in lymph nodes of the upper body [13, 14]. Oneprecursor for lymphoma is human immunodeficiency virus (HIV) [5].9New cases DeathsMen (103,100) Women (103,200) Men (42,600) Women (38,200)Non-Hodgkin lymphoma 4.5% 3.6% 3.5% 3.1 %Hodgkin lymphoma 0.6% 0.4% 0.2% 0.2 %Leukemia (all types) 3.5% 2.5% 3.9% 3.3 %Multiple myeloma 1.6% 1.2% 1.9% 1.7 %Table 2.1: Projected numbers of new cases and deaths from hematologicalmalignancies in Canada for the year 2017. Numbers are given as per-centages of the total number of new cases and deaths in Canada acrossall cancer types, given in parentheses. Data from reference [20].Finally, multiple myeloma - the most common form of myeloma - is a B-cellmalignancy, where plasma cells are overproduced, which in turn overproduce im-munoglobins that build up in the bone marrow. The median age at diagnosis is70 years with bone pain being a common symptom. While multiple myeloma ishighly treatable, it is a complex disease that remains incurable [16, 17].Beyond the scope of this thesis, each of these cancer types can be subdividedfurther. In fact, there are over 70 types of leukemia, and over 90 types of lymphoma[18, 19].2.2.3 Hematological malignancies in CanadaEvery year, the Canadian Cancer Society publishes a comprehensive report de-scribing Canadian cancer incidence and mortality by disease type. This sectionis a brief overview of the information from the 2017 report concerning leukemia,lymphoma, and multiple myeloma [20].In 2017, it is estimated that 206,200 new cancer cases were diagnosed inCanada with 80,800 cancer deaths. Hematological malignancies are relativelyrare compared to other cancers, with information about new cases and deaths forleukemia, multiple myeloma, HL, and NHL given in Table 2.1. For comparison,20.7% of new cases in men were expected to be prostate cancer, and 25.5% of newcases in women were expected to be breast cancer.The largest risk factor for developing cancer is age, and 89% of new cases are10expected to be in people over 50. This means that the rate of cancer incidence isbound to increase in a country with an aging population such as Canada. However,age-standardized incidence rates and age-standardized mortality rates can be pro-duced that compare cancer cases across many years as if the population structurewas not aging. These reveal that for many cancers, the age-adjusted mortality andincidence rates are actually decreasing, despite an increasing number of cases dueto aging demographics, and the total age-adjusted mortality rate from all forms ofcancer is decreasing (although this varies substantially based on the type).Lymphoma, both HL and NHL, are two cancers with notable declines in mor-tality rates. For HL, mortality has decreased by 2.6% per year since 1992, owing toimproved treatments. For NHL, mortality decreased between 2.3% and 2.5% peryear between 2000 and 2012. A notable contributor to this decrease is the use ofhighly active antiretroviral therapy (HAART) beginning in the late 1990s to com-bat HIV, corresponding to a decrease in the number of aggressive lymphoma casesowing to the virus.The incidence statistics for leukemia vary drastically for pediatric patients com-pared to the general public. While only 0.7% of new cancer cases in Canada occurin people 19 years or younger, leukemia and lymphoma are two of the most com-mon cancers in the patient cohort aged 0-14 years (32% of childhood cancers areleukemias, and 11% are lymphomas). Overall, survival for childhood cancers ishigher than for adult cancers.In a study comparing Canadian cancer cases in two time periods, between1992-1994 and between 2006-2008, five-year age-standardized net survival in-creased from 53% to 60% for all cancers combined. The largest increases in sur-vival across all cancers between these two time periods were seen in NHL (15%increase), leukemia (15% increase, varying by subtype from 9% for AML to 25%for CML), and multiple myeloma (14% increase). In 2006-2008, HL had the third-highest rate among cancers of five-year survival (87% in women and 83% in men).A final note in this section is that survival statistics describe patients treated inthe past. In reality, if advances in medicine are constantly improving healthcare, thesurvivability of diseases is always higher than previously reported due to advancesin detection and treatment compared to of past patients.112.2.4 Bone marrow transplantsBone marrow transplants (BMTs) are a treatment option for some patients withhematological malignancies. The procedure involves collecting stem cells to de-liver to a patient by infusion. BMTs can be allogeneic, where they receive stemcells from a matched donor, or autologous, where the stem cells are collected fromthe patient and then re-infused after a treatment. In the earlier days of chemother-apy, the biological dose limit that could be delivered to the bone marrow was alimiting factor in the amount of chemotherapy that could be given to a patient [11].Bone marrow transplants effectively raise the dose limit of chemotherapy or ra-diation that can be delivered without mortality, and are sometimes referred to asrescue treatments [13, 14].Cancer patients who receive bone marrow transplants must first receive a condi-tioning regimen to destroy unhealthy stem cells. This is done with either chemother-apy, radiation, or a combination of both. Regimens including TBI are an option forpatients with relapsed disease who may have become chemo-resistant from ear-lier treatments, or for patients where sanctuary sparing from chemotherapy maybe a particular concern (e.g. in the testes or central nervous system). A pro-phylactic treatment is normally given to suppress the immune system to preventthe onset graft-versus-host disease (GvHD). There is also a graft-versus-tumour orgraft-versus-leukemia effect that some recent treatments take advantage of by us-ing reduced intensity conditioning [13, 21]. With an autologous transplant, thereis no risk of GvHD, although the relapse rates may be higher. The results of someclinical trials are discussed in section 3.2.4.While many centres have historically been reluctant to treat patients of an ad-vanced age with BMTs (with upper age limits around 55-60 years), a recent clin-ical trial demonstrated that while the number of comorbidities does reduce thechance of patient survival following such a BMT, age alone does not predict sur-vival. Based on this, referrals for a BMT should be based on patient conditionrather than biological age [21, 22].The terms “bone marrow transplant” and “stem cell transplant” are used in-terchangeably by the general public. BMTs originally involved collecting donorbone marrow from the bone itself, while more recent technology allows for stem12cell collection from the blood, called a “peripheral blood stem cell transplanta-tion”. The only difference between this and a BMT are the source of the collectedstem cells [14]. A recent meta-analysis spanning 1,224 patients of either peripheralblood transplants or bone marrow transplants have shown similar overall survivaland mortality rates, with peripheral blood transplants having improved disease-freesurvival, a decrease in relapse, and increased GvHD for hematologic malignancytreatment [23].2.3 Physics2.3.1 Light-matter interactionsThe role that physics plays in the treatment of cancer begins with an understandingof light-matter interactions [24–26].Electromagnetic radiation, which encompasses all types of electromagneticwaves including radio waves, visible light, gamma rays, and the therapeutic X-raysused in radiotherapy, is conceptualized as small packets of energy called photons.As radiation travels from an isotropic source, it spreads over a larger area as it prop-agates and spreads out. As a result, the number of photons per unit area decreases,and the radiation becomes less intense in proportion with the square of the distancetravelled. This effect is called the inverse square law,Φ1Φ2=r22r21(2.1)where Φ1 and Φ2 are the photon fluences at two points corresponding to distancesr1 and r2 from the source of the radiation.When photons travel through matter, the distance travelled before interactingwith the medium can be estimated from the linear attenuation coefficient, µ , whichis the fraction of photons that attenuate per unit thickness of attenuating material.The value of µ depends on the energy of the photon and the electron density of thematerial. The attenuation of a photon beam over a distance x is exponential,N = Noe−µx (2.2)13where No and N correspond to the number of photons in the beam at the origin andat a distance x, respectively.There are four light-matter interactions that form the basis of radiation oncol-ogy physics:Rayleigh scattering: An incoming photon interacts with the electrons of anatom in the medium, and is deflected away in a new direction without loss ofenergy to the medium.Photoelectric effect: An incoming photon is fully absorbed by an atom in themedium. An amount of energy, Etr, is then transferred away from the atomby an ejected electron carrying the energy hν of the incident photon less thebinding energy Eb that originally held this electron in place in the atom,Etr = hν−Eb. (2.3)The resulting inner-shell electron vacancy can then be filled by a higher-shell electron in the atom. This transition results in a new photon beingemitted from the atom with an energy equal to the difference between thebinding energies of the shells in the transition. For large transitions, thesephotons can eject more electrons from the atom which are then called Augerelectrons.Compton scattering: An incoming photon of energy hν interacts with anelectron of an atom in the medium. The electron is knocked out of the atomand carries away with it an energy given by Etr,Etr =hνmec2(1− cos(θ))1+ hνmec2 (1− cos(θ))(2.4)where c is the speed of light, me is the mass of an electron, and θ is the scat-tering angle of the photon. The incident photon continues in a new directionafter deflecting with the atom, with less energy.Pair production: An incoming photon interacts with the nuclear field of anatom and is converted to an electron and a positron. The photon must have14energy of at least 2mec2, or 1.022 MeV, corresponding to the rest mass of anelectron and positron combined. The total energy transferred to the chargedparticles is given byEtr = hν−1.022 MeV. (2.5)Another interaction that is possible is triplet production, where a photon in-teracts with the field of an electron in an atom to produce three chargedparticles. The threshold energy of this interaction is 2.044 MeV which isabove the energy scales used in this thesis with a Cobalt-60 unit (describedin section 2.3.3).The attenuation coefficient µ can be written in terms of the interaction cross-sections for each of these processes,µ = σR + τ+σC +κ (2.6)where the four variables on the right-hand side correspond to Rayleigh scattering,photoelectric absorption, Compton scattering, and pair production. The relativeprobabilities of these interactions depend on the energy of the photons and theatomic number of the medium. These probabilities for water are shown in Figure2.3.Energy is transferred from photon beams to a medium in two steps. First,through these processes, energy is transferred from a photon beam to charged par-ticles. Next, these charged particles deposit their energy elsewhere in the medium.Charged particles interact by collisional effects throughout their tracks, or theyexperience radiative energy losses where photons are produced that may go on toproduce more charged particles. This radiative interaction is called bremsstrahlung,German for “braking radiation”.Radiation comprising of photons is called indirectly ionizing radiation, andradiation comprising of charged particles is called directly ionizing radiation. Thephotons incident on a patient from radiotherapy sources, in and of themselves,are unable to cause any sort of biological or chemical changes. It is the chargedparticles they set in motion that are are able to cause damage to cells, explained inthe next section.15Figure 2.3: Relative probabilities of the four photon interactions describedin section 2.3.1 as a function of photon energy for radiation travellingthrough water. The energies of concern in this thesis are 1.17 MeV and1.33 MeV, described further in section 2.3.3.2.3.2 Radiobiology and doseDNA is the double-stranded helix that contains our genetic code in cell nuclei.DNA is tightly wound into a set of 46 structures called chromosomes, divided into23 homologous pairs. When cells divide through mitosis, the genetic informationis duplicated with a copy given to each daughter cell.The strands of DNA are composed of many atoms with charged and unchargedportions. When charged particles are incident on DNA, there are many featuresin the orderly pattern of molecules and atoms in DNA that can be disrupted, andmolecules may be broken apart due to these interactions. When a single DNAstrand is broken (a single strand break, or SSB), there are several biological re-pair mechanisms available to fix this broken strand and repair the double helix16structure. However, when both DNA strands are broken in locations that are veryclose to each other (a double strand break, or DSB), it is much more difficult torepair the DNA correctly [27]. Chromosomal aberrations result from the incom-plete or incorrect repair of chromosomes, such as the Philadelphia chromosometranslocation mentioned in section 2.2.2. If there are chromosomal aberrations thatgo unrepaired, the cell will become dysfunctional in mitosis; either one or bothdaughter cells will be abnormal, or the process of mitosis may not be able to takeplace at all. As a result, ionizing radiation acts not to kill a cell immediately, butrather to induce reproductive death in the cell.The induction of reproductive cell death can occur in both normal and can-cerous tissue. Unfortunately, it is impossible to avoid completely the radiation ofhealthy tissue in radiotherapy. It is known that radiation can induce cancer as wellas treat it. Notably, from following survivors of the atomic bombings in Hiroshimaand Nagasaki in 1945, it is known that whole-body exposures to radiation can in-duce leukemia [5, 27]. Typically, these cancers have a long latency period beforemanifesting symptomatically, which can occur decades after a radiation therapytreatment [28]. Therefore, the aim of radiotherapy becomes to maximize the cellkilling in cancerous tissue while sparing normal tissue to reduce the number orseverity of treatment-related toxicities (side-effects).The amount of cell killing increases in proportion with the radiation dose, thatis, the amount of energy absorbed per unit mass in the medium. Dose is givena unit called the gray (Gy), where 1 Gy = 1 J/kg. Also commonly used is thecentigray (cGy), with 100 cGy = 1 Gy. Because photons must first transfer theirenergy to damaging charged particles before energy is deposited in the medium,the maximum dose does not occur immediately at the surface of a patient whenradiotherapy is delivered as a photon beam. Rather, there is a build-up regionbetween the surface and the depth of maximum dose dmax. This depth is 0.5 cmfor Cobalt-60 photon beams and increases with photon energy in other treatmentmodalities. After reaching dmax, the dose delivered decreases steadily with depththrough the patient, measured relative to the maximum dose at dmax by a quantitycalled percent depth dose (PDD).Dose can be measured by several means experimentally with dosimeters. Thereare many types of dosimeters and each has strengths and weaknesses that make17them suitable for measurement of dose in different settings. The dosimeter shouldprovide an accurate measurement of the dose without changing the radiation beam.Another dosimetry technique, discussed at length in Chapter 4, is Monte Carlodosimetry, where computer simulations are used rather than a physical device inthe path of a beam. While Monte Carlo can be time consuming, requires heavycomputational resources and careful validation measurements, it produces a fulldose distribution of a given target without having to use clinical machine resources.2.3.3 Radioactivity and Cobalt-60 radiotherapy unitsIt’s these few eternally unstable elements that possibly hold the key tothe fate of the world. They were first developed and employed formass destruction. Now Canada has taken the lead in showing howthey can be used for the mass benefit of mankind.— Eric Hutton, The Atom Bomb That Saves Lives, 1952 [29]The previous sections describe the goal of radiotherapy and the mechanismby which therapeutic X-rays interact with matter. Finally, radioactivity enables adiscussion of the origin of this cell-damaging radiation.Matter is comprised at a fundamental level by atoms consisting of protons andneutrons in their nuclei. Some atomic nuclei are unstable and eventually undergodecay, leading to a new daughter nucleus along with byproducts depending on thedecay mode. The rate of a radioactive decay is proportional only to the number ofradioactive nuclei N in a given sample,dNdt=−λN (2.7)where λ is the decay constant for the given decay mode.In external beam radiation therapy, the most important radioactive isotope isCobalt-60, with a half-life of 1921.5 days, or roughly 5 years and 3 months. 99.8%of Cobalt-60 decays occur first by decay to an excited state of Nickel-60, whichthen emits two photons of energies 1.173 MeV and 1.332 MeV [24]. Other subse-quent decays from Cobalt-60 give rise to other photons with energies below and inbetween these two values [30]. It is these X-rays that provide useful radiation for18Figure 2.4: The Cobalt-60 unit at the Vancouver Cancer Centre. The radioac-tive Cobalt source is housed in a lead shielding unit in the treatmenthead.radiotherapy2. The implementation of Cobalt-60 for radiation therapy was a Cana-dian innovation, championed by Harold Johns with the first treatments on humancancer patients taking place in Saskatoon, Saskatchewan and London, Ontario in1951 [29, 31]. In these units, a Cobalt-60 source is sealed in a lead shielding unit,until a treatment occurs when the source moves to the treatment head for its decayX-rays to pass through the beam collimator. A picture of the Cobalt unit in use atthe VCC is shown in Figure 2.4.At this energy range, in tissue, Compton scattering dominates heavily over anyother light-matter interaction, until the photons scatter to lower energies where2Often, the energy of Cobalt-60 is simply written as 1.25 MeV, the average of the two mostcommon photons.19the photoelectric effect and Rayleigh scattering become noticeable. While Cobalt-60 photon energies are marginally above the threshold for pair production, thesmall amount by which they are over the threshold (and the low atomic numberof tissue material) make this interaction infinitesimally rare. On the other hand,in lead, Cobalt-60 photons are still mostly attenuated by the Compton effect, butphotoelectric absorption and pair production are occurring as well, each at roughly5% of the frequency of the Compton interactions [26].20Chapter 3Total body irradiation (TBI)3.1 Historical developmentThe discoveries of the X-ray by Wilhelm Roentgen in 1895 and of radioactivity byMarie Curie in 1902, which lead to their awarding of Nobel Prizes in 1901 and 1903respectively [32], quickly gave rise to many studies investigating the applicationsof these discoveries in medicine. Very shortly afterwards, TBI was proposed in1905 by Friedrich Dessauer, a German biophysicist, in the same year that Einsteinpublished his famous papers that included a description of the photoelectric effect.TBI was being advocated at the time as a treatment for leukemias. Dessauer hadsuggested TBI as a simultaneous irradiation of the patient by three different sourcesfrom the superior, inferior, and anterior directions to the patient [33].The first unit constructed specifically for TBI was constructed at MemorialHospital in New York [34], championed in May 1931 by Arthur Heublein whopassed away less than a year later. This unit allowed for two patients to be irradi-ated simultaneously in adjacent rooms by the same low-voltage X-ray tube, for ap-proximately 20 hours per day over one to two weeks [35]. The earliest publicationdescribing “a new method of X-ray therapy consisting of continuous irradiation ofthe entire body at long distances from the tube” was published posthumously byHeublein in June 1932 [36].Early interest in TBI was held not only by physicians, but also in secret bythe officials of the Manhattan Project in World War II, and later on by the UnitedStates Department of Defense. Information about whole-body radiation exposureswas considered valuable in the contexts of occupational exposures of workers de-veloping nuclear weapons, determining whether the pilots of nuclear-powered air-planes would be at risk of developing radiation sickness, and in preparing a defense21and response strategy in the event of a nuclear strike. The involvement of the De-partment of Defense in hospital clinical trials was a source of controversy when itcame to light, along with other ethical issues involved with these studies, and thefinal TBI trial that was funded by the Department of Defense at the University ofCincinnati was closed in 1971. In the 1990s, the administration of President Clin-ton struck the Advisory Committee on Human Radiation Experiments to study andreport on such issues in detail. The final report of the committee included a chapteron TBI, among discussions of what research conducted with patients is consideredtherapeutic or non-therapeutic [37]. Fortunately, only this small subset of researchin TBI is shrouded in ethical concerns.Historically, TBI was used to treat malignancies that spanned the whole body,or as last resort treatment for patients with poor outcomes [37]. While the myeloab-lative, or cancer-fighting, properties of TBI were studied early on, its progress wasdismal in fighting cancer compared to developments that were being made withchemotherapeutic approaches. However, TBI was recognized as an effective im-munosuppressant and plays an important role in the history of organ transplantationto condition patients; in 1959, a successful solid organ transplant was performedfor the first time between non-identical twins, a kidney, breaching a genetic barrierto organ transplantation for the first time, in a patient who had first received TBI[38, 39].Pioneering research in the use of bone marrow transplants occurred in Seattle,championed by Dr. E Donnell Thomas in the early 1970s [34], who received the1990 Nobel Prize in Physiology or Medicine for his work [32, 40]. The Seattleteam conducted many groundbreaking experiments with TBI and bone marrowtransplants, both in animal trials and in human clinical trials [39], some of whichare outlined in subsequent sections of this report. In current practice, when TBI isgiven before a transplant, it is prescribed as only one part of the patient care planalongside chemotherapy.223.2 Modern TBI3.2.1 Fractionation, dose, and dose rateEarly TBI prescriptions were typically given in one large fraction of 8-10 Gy,which involved a high risk of treatment-related mortality [40]. Many TBI stud-ies throughout the 1980s and 1990s established a clear benefit of fractionation: byspreading the treatment over multiple days, patients were able to tolerate largertotal doses. These studies included not only experiments with dogs receiving mar-row infusions, many of which were from the Seattle group [41, 42], but also otherstudies of a theoretical nature based on radiobiological considerations [43, 44]. Arandomized clinical trial in 1990 in Seattle found that overall survival from patientsin two treatment arms, one receiving 12 Gy in 6 days and one receiving 15.75 Gy in7 days, had similar overall survival, but the higher-dose arm saw more treatment-related mortality while the lower-dose arm had more relapse-related mortality [45].12 Gy in 6 fractions has widely become the most common TBI prescription.The dose rate used in TBI treatments is typically low. A clinical trial in 1982found that a dose rate above 6 cGy per minute was significantly correlated to mor-tality in acute leukemia patients receiving TBI, although it was confounded withother factors in the study, and a definitive conclusion could not be drawn [46]. Atrial with dogs found that a lower dose rate lead to improved survival when receiv-ing single-fraction doses of TBI, but in fractionated schemes, the results were lessclear [42]. A more recent meta-analysis did not find a dose rate effect on the in-cidence of interstitial pneumonitis (a severe scarring of the lung tissue that can befatal) [47], while a study produced around the same time indicated that there is aneffect above 15 cGy per minute [48]. It has been suggested elsewhere that low TBIdose rates are important for optimal outcomes and reducing the risk of interstitialpneumonitis [49], but this has not been adopted universally and is not used as alimit in helical tomotherapy or volumetric modulated arc therapy techniques.There are several distinct conditioning regimens that exist, reflected in differenttypes of TBI prescriptions. The prescription of 12 Gy in 6 fractions falls under themyeloablative conditioning category where stem cell support is later required torescue the patient. Nonmyeloablative regimens are those that produce minimal23loss of mature blood cells and stem cell support is not required afterwards for thepatient. In between these two categories is reduced intensity conditioning (RIC),where it is possible that the patient could recover without further support, but therewould be significant morbidity and mortality. Some lower-dose TBI prescriptionsare given as parts of RIC techniques and suit patients who are not eligible for fullmyeloablative conditioning [50]. RIC, where a trade-off is made between reducedincidence of toxicities and an increased risk of relapse, has gained popularity sincethe late 1990s [51]. A graft-versus-tumour or graft-versus-leukemia effect can alsobe harnessed with RIC to combat the disease [52]. A recent study comparing 45patients who received RIC to 135 patients receiving myeloablative conditioningfound similar rates of overall survival and relapse, but with lower rates of toxicitiesin the RIC group [53].Some recent early-stage clinical trials have revisited the concepts of dose es-calation and TBI-only conditioning for patients with refractory disease or resistantrelapses. One study for lymphoma patients had three patients in arms of 16 Gy, 18Gy, and 20 Gy treatments, using lung and kidney shielding [54]. The feasibility ofTBI-only conditioning was demonstrated, with only one patient in the 18 Gy armdeveloping severe lung toxicity. Another study of four leukemia patients receivingautologous transplants following TBI doses of 16 Gy again showed that this pro-cedure was tolerable [55], likely due to improved treatment delivery systems andsupportive care strategies in the time elapsed since the 12 Gy in 6 fraction regimenbecame the standard of care.3.2.2 Worldwide heterogeneity in techniquesClinics with TBI programs treat a very small volume of TBI patients compared tothe total volume of patients receiving any kind of radiation treatment. A recent sur-vey across Canadian cancer centres identified 12 centres that deliver TBI and foundthat around 400 patients per year receive TBI across the country [56]. For manyreasons, there is enormous diversity in how TBI is delivered across the world; arecent European survey of 56 centres from 23 countries found no pair of centersdelivering TBI the same way [57], which is also true of the Canadian survey. Thesereasons include the small patient volume, the number of degrees of freedom in the24planning of a TBI treatment - including the dose rate, fractionation schedule, beamenergy, immobilization techniques, and organ dose constraints - and the inherentphysics challenge of delivering a uniform dose to a large and non-uniform humanbody. The very small number of TBI patients makes it difficult to justify the con-struction of a dedicated TBI unit in a clinic, which leads to existing units beingrepurposed based on the equipment and expertise available at each centre. Whilethese conditions allow for each treatment centre to develop a technique that bestsuits their personnel and resources, it renders inter-centre collaboration difficult,with little standardization in place, and the design of multicentre clinical trials tostudy TBI becomes difficult.Information on different TBI techniques in the world is sampled from threerecent surveys: the mentioned Canadian survey of 12 centres likely representingall of Canadian TBI [56]; the mentioned European survey of 56 centres locatedin 19 European countries as well as Australia, Israel, Saudi Arabia, and Tunisia[57]; and an informal survey of 62 centres with about half of the responses fromthe USA, with the rest throughout the rest of the world (reflecting some double-responses between this survey and the first two) [58].The most common TBI prescription around the world is 12 Gy in 6 fractions.The European survey focused on myeloablative transplantation only and found atotal dose range from 8 Gy (in 1 fraction) to 14.4 Gy (in 8 fractions). The Cana-dian survey included TBI for all purposes, including doses of 2 Gy (in 1 fraction)up to 13.5 Gy (in 8 fractions), as well as unspecified fractionation schedules tabu-lated as “other”. The dose rate range for most centres is below 20 cGy per minute,although some European countries report up to 37.5 cGy per minute and someCanadian centres up to 50 cGy per minute. Lung shielding is common practice,being reported by 47 out of 56 centres in the European survey, 11 out of 12 centresin the Canadian survey, and 41 out of 59 respondents in the informal survey, al-though the degree of shielding varies substantially. A small number of centres usechest wall compensation with either electrons, photons, or both, to avoid underdos-ing the tissue anterior to the lungs. Some other organs are infrequently shielded aswell, including the kidneys, liver, heart, eye lenses, and thyroid gland.The apparatus involved in each technique varies widely as well. Most cen-tres use linear accelerators (51 out of 56 centres in the European survey, 10 out25of 12 responses in the Canadian survey, and 55 out of 61 responses in the infor-mal survey) with 6 MV beam energies being used in more than half of centres,and other energies ranging up to 25 MV. The other centres use Cobalt-60 tech-niques. More than half of centres use two fields per fraction in parallel-opposedpair (POP) techniques, with either anterior-posterior fields or bilateral fields, withother infrequent techniques including moving treatment beds, multiple SSDs in agiven fraction, field junctioning, helical tomotherapy (HT), volumetric modulatedarc therapy (VMAT), and other individualized approaches. Bolus use varies fromlarge amounts of bolus placed around the body to no bolus at all, or used only insmall amounts in areas with small amounts of scatter such as the neck and legs.For in-vivo dosimetry, semiconductors are most popular in Europe (37 out of 56centers), while the remaining European sites and the Canadian survey show a mixof semiconductors, MOSFETs, OSLDs, TLDs, and ion chambers.3.2.3 Advances in conventional TBIThis section will present some recent research and development in traditional TBItechniques, and samples the innovation involved in designing them.Various techniques are used in TBI to ensure that the entire patient fits within aradiation beam. The VCC technique, using a sweeping-beam Cobalt-60 source, isdescribed as it was first implemented over 23 years ago by Hussein and El-Khatib[59]. Minor changes have been made to our technique since then, described indetail in its current form in section 3.3. Another Canadian Cobalt-60 TBI tech-nique has been described at McGill University [60], where the larger SSD (220cm) is used to irradiate the full patient without sweeping. This Cobalt unit wasinstalled in the same room as a 10 MV linac that was used for other specialized,less frequent procedures to optimize the use of clinical space. This technique wasdecommissioned in 2015.Many of the studies described in this chapter use parallel-opposed pair (POP)treatments where the patient lies supine and then prone, with the gantry at 0 de-grees. Other centres use gantry angles of 90 or 270 degrees with the patient posi-tioned either lying down or sitting. These treatment designs allow SSDs to reachup to 400 cm [61–63] to ensure complete coverage of the patient and a sufficiently26low dose rate, with one group reporting SSDs of 480 cm [64] or higher. Severalother centres position the patient in the beam by having them stand rather than sitor lie down [65, 66].Rather than using a moving source or a very large SSD to cover the entire pa-tient, the patient can also be moved through a fixed beam on a translating couch[67–70]. One group has described attaching a tray table above a translating couchthat can also move, to ensure appropriate positioning of lead lung shield penumbraswhile the patient is moving underneath [67]. Using a variable rather than constantcouch velocity offers improvements in dose homogeneity [68]. A recent reportfrom the Tom Baker Cancer Centre in Calgary describes the use of continuouslymoving MLC leaves in the direction of the patient motion with a fixed aperturewidth in the lateral direction of the patient, offering further improvements in dosehomogeneity with a 15% decrease in lung dose and a reduction of the lateral bodyedge dose, from 20% over prescription to a 3% variation from prescription [69].The dynamic MLCs render it unnecessary to produce patient-specific organ com-pensators, removing a step from the clinical workflow.Moving the patient during a treatment requires the use of specialized equip-ment. There are other reports presenting TBI techniques that use equipment al-ready available in most clinics, without needing to reinstall or recommission exist-ing units or requiring very large treatment vaults, offering ways to implement newTBI programs with minimal disruption to the clinic. For example, one paper de-scribes a POP technique with the gantry angle at 270 degrees and a seated patient.Cerrobend lung compensators are used during only one of the treatment fractions,and the only beam modulation required are two discrete adjustments to the colli-mator jaw setting during each fraction. The technique still gives acceptable dosehomogeneity with a simple setup [63]. Another reported technique is arc-basedwhere the gantry sweeps out a 100 degree angle while the patient lies supine orprone, feasible at any clinic using linear accelerators with flattening filters and arcrotation features. This strategy has been shown to be effective for patients up to 30cm in thickness, and includes in-vivo dosimetry with nine semiconductors and anion chamber [71].A field-in-field IMRT forward-planning approach has been presented, using an18 MV linac at 380 cm SSD. In this technique, reduced-weight fields are used in27planning to reduce dose in what would otherwise be higher-dose regions in thebody. This technique does not require compensators or bolus, as homogeneityis achieved using the IMRT subfields [72]. Inverse planning has been proposedfor TBI as well, using a modulated-arc TBI delivery system and cerrobend lungcompensators. Using PinnacleTM (Philips Medical Systems, Andover, MA, USA),body and lung contours are produced for each patient, and the shapes of the lungcompensators are imported to Pinnacle and placed above the patient for the dosecalculations [73]. Most modern linear accelerators would be capable of using thisinverse-planned technique in a standard sized vault, and it shows improved homo-geneity over comparable forward-planning techniques. An interesting point is thatthe dose rate is restricted to 50 MU per minute in any beams that pass through thelungs (corresponding to a 10 cGy per minute dose rate), with respect to the earlierdiscussion of dose rate in TBI, while all other beams are delivered at 300 MU perminute.Two further technical details are worth describing. First, beam spoilers areoften used in linac-based TBI methods [61, 62, 64, 68, 72]. These are used tointentionally introduce electron contamination to reduce the depth of maximumdose in TBI treatments, to eliminate the amount of skin sparing that is typical ofhigh-energy X-rays. Typically, these spoilers are on the order of one centimeterthick, made out of acrylic material, and positioned relatively close to the beamhead, within around 30 cm. Second, some TBI techniques collect full body CTscans in their planning pipelines [71–73]. In these cases, it is often necessary dueto the scan length capacity of CT scanners to collect two separate image volumesand then concatenate them manually in the treatment planning system. This canintroduce registration error in the area of overlap, although the overlap region istypically selected in the region of the thigh, away from critical organs at risk.Two more studies are included in this section to further demonstrate the rangeof creativity in TBI planning. A South Korean group, in a 2017 proof-of-conceptstudy, reported using custom-designed compensators for a bilateral beam TBI treat-ment where the patient-specific compensators would be designed by 3D printingmoulds and pouring Cerrobend into them, based on 3D optical scanning of thepatients [74]. These compensators of variable thickness can accommodate forvariations in patient thickness and would also allow for planning to be completed28without requiring CT scans. The technique does not presently account for tissueheterogeneity in the body, but could be adapted for this purpose in future studies.Finally, an American group in 2016 considered using protons for TBI in a re-cent theoretical study [75]. Conventional proton therapy takes advantage of theBragg peak, an effect where massive charged particles deposit most of their en-ergy in tissue at a very narrow and well-defined depth compared to photons andelectrons, offering better sparing of normal tissues surrounding the target volume.Leading up to the Bragg peak is a broad-slow-rising dose deposition curve, whichcurrently serves no clinical purpose. The report suggests using an SSD of sevenmetres to contain a patient of thickness up to 29 g/cm2 entirely within this broad-slow-rising region. Feasibility was shown with both phantom measurements andMonte Carlo measurements of a patient CT dataset at 250 MeV, 275 MeV and 300MeV. This technique would not require beam spoilers or patient-specific compen-sation, as long as a double scattering system was used to produce the necessarylarge field size. To cite the paper directly, “to design a proton system just to makeTBI treatment easier would not be economically wise”, although this approach il-lustrates the extent of TBI innovation in the literature.3.2.4 ToxicitiesMany studies have been published with toxicity data for treatments that includeTBI. For several reasons, it is difficult to speak directly about the toxicities associ-ated with TBI. First, TBI delivery does not occur in isolation as its own treatmenttechnique in modern practice; rather, it precedes a bone marrow transplant, andis typically accompanied by chemotherapy. As a result, toxicities thought to becaused by TBI could be attributed, either in part or in full, to these other concur-rent treatments. Second, TBI patients have typically received complex care plansin years leading up to a TBI treatment, either from their current or previous dis-ease, which may contribute to toxicities as well. Third, with the broad diversityof TBI techniques used across the world, it is difficult to compare similar patientsreceiving similar treatments from different centres to a high degree of precision.That being said, it remains useful to examine toxicity findings on a clinic-by-clinicbasis for consistencies, and some studies are able to produce aggregate data for29several centres.The toxicities of most concern from TBI are the late effects, rather than theacute effects. A study of acute effects in 162 patients tracked and followed theacute toxicities of TBI patients, evaluating them prior to each radiotherapy fractionof a 12 Gy in 6 fraction protocol where the lung dose was limited to 10 Gy [76].Nausea and vomiting were the most common conditions that patients complainedof, in 42.6% and 22.8% of patients, respectively. Typically, a prophlactic treatmentfor gastrointestinal reactions is given to patients to mitigate these effects. Skinreactions including erythema were also reported in 41.4% of patients, as well asfatigue in 49.2% of female patients and 28.3% of male patients. Overall, in termsof acute toxicities, TBI is well-tolerated.Of the late effects, lung toxicity is the most concerning toxicity from TBI. In-terstitial pneumonitis is a known risk from administering TBI, and as describedearlier, most centres implement lung shielding to prevent this. While many centresmaintain a low dose rate to the lungs to prevent lung toxicity, there is conflictingevidence as to whether or not this is a relevant precaution with fractionated TBIprotocols [47, 48]. Incidence rates of interstitial pneumonitis have been reportedas 11% without lung shielding or 2.3% with lung shielding in a study of 12 Gyin 6 fractions [47], 14.3% for patients receiving over 9.4 Gy to the lung and 3.8%receiving less than 9.4 Gy to the lung in a 10 Gy in with three fractions deliv-ered once per day [77], and 33% in a study of 101 patients receiving 13.5 Gy in9 fractions [78]. In the latter study, survival at one year post-transplant was 28%for patients who had severe lung toxicities, and 81% for those who did not. Thestudies also suggested that allogeneic transplants have a higher risk of lethal pul-monary complications [77] and that prior chemotherapy is a risk factor [78]. In astudy of pediatric patients, there was a reported rate of 23.3% for interstitial pneu-monitis in a cohort of 129 patients receiving no more than 10 Gy to the lungs inTBI prescriptions of 10.5-14 Gy [79]. Another study of both pediatric and adultpatients receiving either 11 Gy or 12 Gy to the lungs in a 12 Gy TBI dose sawthat decreasing the lung dose reduced the rate of pneumonitis from 22.2% to 8.5%across the whole cohort of 257 patients; in the subset of 40 pediatric patients, therate lowered from 25% at 12 Gy to 4.2% at 11 Gy [80]. Because it is not possibleto determine by a biopsy if the pneumonitis is radiation-induced or infectious, it30would not be prudent to state that TBI is the only cause of these toxicities, althoughthe statistical association warrants caution nonetheless.The kidneys are another OAR during TBI, with some centres administeringkidney shields in addition to more common lung shields. One study uses both CTand ultrasound to design the fabrication of these shields in an anterior-posteriorPOP technique [81]. A single-centre study of a pediatric TBI cohort of 92 pa-tients found only one patient with unresolved renal dysfunction at one year afterthe transplant, with 28% of patients experiencing renal dysfunction that resolvedwithin months of onset [82]. In this study, they limited the kidney dose to 10 Gyfor patients with a history of renal dysfunction, but the patients otherwise had treat-ments of 11.1 Gy or 12 Gy. One meta-analysis found that there was only a kidneydose-response function for adult populations, but no similar response for pediatricpopulations [83]. Another meta-analysis found a threshold biologically effectivedose of 16 Gy for renal toxicities, which would be lower than the 21.2 Gy biolog-ically effective dose of a standard 12 Gy in 6 fraction scheme, although the age ofthe patients in the reviewed studies was not specified [84].Cataractogenesis is also seen following TBI treatments, and like the kidneys, asmall number of centres also use eye shields for patients to prevent this risk. Thesame meta-analysis that previously found a threshold dose for renal toxicity alsosuggests a biologically effective dose threshold of 40 Gy to the eyes, which is notexceeded in fractionated dose schemes [84]. An earlier meta-analysis of 1,063 Eu-ropean TBI patients with acute leukemia saw a 60% risk of cataract developmentat 10 years in patients who received a single dose and 43% in patients receiving afractionated scheme [85, 86]. The use of eye shields is considered controversial asthe eyes are sanctuary sites from chemotherapy, and their use can result in cylindri-cal areas of underdosage in the brain behind the eyes, leading to higher chances ofrelapse. One study aimed to determine if this risk could be observed, and studieda pediatric cohort of 188 children receiving TBI where 139 had eye shields and 49did not [87]. The incidence rate of cataracts was 90% without shielding and 31%with shielding, and when they did develop, patients with eye shielding developedthem at a longer time after treatment and with less severity. Two patients witheye shielding relapsed, but this small proportion did not suggest that eye shield-ing alone increases the risk of relapse. A recent multi-centre trial found that for31patients under three years of age receiving allogeneic transplants, whether or notTBI was included in the conditioning regimen was an independent risk factor fordeveloping cataracts [88].Pediatric patients make up a significant portion of TBI treatments and severalstudies have focused on this patient population. A 2005 study of 42 children re-ceiving 9.9 Gy in 3 fractions with a median 5.7 years of follow-up found that 78%of patients developed cataracts, 12% developed hypothyroidism and 14% devel-oped thyroid carcinoma [89]. For patients treated under three years of age, 29% ofpatients developed an osteochondroma (a benign bone tumour) at a median follow-up of 9.2 years, some of which required surgical removal. 74% of patients hadsome restrictive pulmonary toxicity but only 45% of patients continue to experi-ence mild problems. Overall, while some late effects from the treatment couldpresent as late as 10 years after treatment, these symptoms rarely affect qualityof life significantly. Follow-up is necessary for these patients for family planningpurposes as testicular and ovarian dysfunction had been noted as well. Anotherstudy of 62 pediatric patients receiving autologous bone marrow transplants at 12years of follow-up, 30 without TBI and 32 with TBI, found an overall higher rateof toxicity in conditioning with TBI compared to conditioning without TBI [90]. Amain toxicity noticed was growth delay, and hormone deficiencies were a concernas well. Growth hormone treatment may be beneficial in the post-transplant careplan for these patients. Secondary cancers were also more common in the TBI armof the trial with six malignant tumours as well as eight benign tumours, appearingwhile the non-TBI arm saw only one patient develop leukemia. The previouslydiscussed multi-centre trial for patients under three years of age found, across 717patients, that the most common late effects were growth disturbance, cataracts, andhypothyroidism, and also suggests annual thyroid palpitation and earlier mammo-gram screening for females in the long-term care plans [88].Another OAR that is not as frequently discussed in the context of TBI arethe breasts for female patients. Based on long-term follow-up data from Seattle,3,337 female survivors of bone marrow transplant procedures were studied, and itwas found that the cumulative incidence of breast cancer at 25 year follow-up was11.0%, although this figure is 3% for patients who did not receive TBI while it is17% for patients who did receive it as part of their transplant [91].323.2.5 Helical tomotherapy and VMATAs more sophisticated beam modulation techniques have been developed, moreconformal treatment options have been devised for TBI. The two main alternativesto the previously described POP techniques are helical tomotherapy (HT) and vol-umetric modulated arc therapy (VMAT). Full descriptions of these techniques canbe found elsewhere; in short, HT involves a couch translation while the gantry fullyrotates around the patient [92], while VMAT uses an optimization technique to pro-duce multiple treatment arcs with continuously changing gantry rotation speeds,MLC patterns, and other settings [93].An early feasibility study of using HT for TBI investigated the effects of sev-eral treatment parameters on treatment planning [94]. Smaller field sizes resulted inimproved dose homogeneity, but also increased treatment times. Greater amountsof MLC modulation also resulted in longer treatments, but lead to lower doses tocritical organs. Treatment planning with HT resulted in a dose reduction to criticalorgans by 35-70% of the prescription dose, including lungs, eyes, heart, liver, andkidneys, without the need for special blocks or other equipment during treatment.Additionally, megavoltage CT (MVCT) images were collected for positioning. Itwas found that using a limited MVCT approach would reduce the target localiza-tion time by 60%, instead of using an MVCT of the whole body. Depending on thetreatment parameters selected, treatment times were in the range of 16-31 minutesper fraction for an anthropomorphic phantom.Another feasibility study of TBI with HT directly compared HT plans to ex-tended SSD treatment plans on four sets of patient CT data [95]. The HT planstreated the upper body with an HT technique and used a standard, nominal SSDtreatment for the lower limbs. The extended SSD plans included an electron boosttreatment. In this study, each treatment took about 4 hours to plan, including anhour to produce lung shields and compensators for the extended SSD treatment.The lung median doses were 5.4 Gy for HT plans and 8.34 Gy and 8.95 Gy forextended SSD plans (left and right lung, respectively). The mean dose deliveredto the PTV was 12.3 Gy using HT plans and 10.3 Gy for the conventional TBIplans. The total beam-on time was longer for the planned HT deliveries, around15.4 minutes per fraction, compared to 11.1 minutes per fraction for extended SSD33treatments, but offers superior dose homogeneity and lung sparing.Another group has described an HT TBI technique using MVCT for both pa-tient positioning and dose reconstruction [96], and reported treating four patientswith this technique [97]. For each fraction, MVCT images were collected anddeformed to the planning CT image with a deformable registration. Planned anddelivered doses, as measured by the TPS calculations and the MVCT reconstruc-tions respectively, differed by 2.7% at most. The beam-on time per fraction wasbetween 19 and 23 minutes, not including time for MVCT data collection, intra-fraction contouring, or analysis between fractions. Another group studied how tooptimize the amount of time required to collect MVCT images for patient posi-tioning while maintaining an acceptable quality of the data, using three cadaversas imaging subjects [98]. Scan times ranged from 4 to 16 minutes per MVCT de-pending on the couch speed used (1 mm/s to 4 mm/s). An iterative reconstructionalgorithm allowed for faster couch speeds, and hence shorter scan times, that pro-duced images of acceptable quality for patient setup. This was at a cost of muchhigher computational times over a filtered back projection algorithm.Another study focused on defining appropriate target volumes in HT TBI andexamined the impact of positioning errors on treatment as measured in a TPS [99].They examined three PTVs: one equal to the external body contour, one with a 5mm air margin outside the patient to account for patient motion during treatments(similar to [95]), and one with a 5 mm margin trimmed into the patient to excludethe skin and restrict the skin dose (similar to [100]). They determined that setup er-rors up to 5 mm provided clinically acceptable dosimetric results, and that design-ing plans with PTVs including air margins were more vulnerable to larger setuperrors. Given this, the authors reason that it is better to emphasize strict patient im-mobilization rather than including an air contour in HT planning to accommodatefor patient movement.An initial study of HT TBI on a cohort of 10 pediatric patients was carriedout, with these patients being suitable for an HT trial due to their small size [101].For larger patients, several reports describe needing two CT scans for planningtreatments that use HT or VMAT for the whole body, due to the limited length of asingle CT scan [101, 102]. In these cases, the patient is rotated 180 degrees duringtreatment between the fields planned on the upper or lower body. For this pediatric34study, patients smaller than 145 cm in height could be treated in 17 minutes ofbeam-on time per fraction, while patients who needed to be rotated and treated intwo positions required up to 34 minutes of beam-on time per fraction. MVCT wasalso used for patient setup. The pediatric patients required strict immobilization,including sedation or anesthesia despite the twice-daily treatments, due to longtreatment times in a single position as well as claustrophobia and loud noises in theunit.Some clinical trials have been undertaken using HT techniques for TBI. Onegroup treated four AML patients, finding improved critical organ sparing with lungdoses of approximately 7 Gy and kidney doses of approximately 8 Gy from a 12Gy in 6 fraction prescription [97]. The team also noted the benefit of not hav-ing to produce compensators or other cumbersome procedures. The main toxicityreported was low-grade dermatitis. Two patients live disease free at most recentfollow-up and two patients succumbed to GvHD, but had no signs of disease atdeath. Another clinic using HT for TBI treatments reported a unique skin reactionin a band at the level of the femur in four out of twelve patients, arising one totwo months after the stem cell transplant [103]. The region coincides with the fieldjunction between upper body and lower body treatments where the fields overlap to24 Gy over 6 fractions, which had been designed to ensure that no part of the targetwas missed. The lesions were improved within a week by applying moisturizingagents.TBI techniques using VMAT have also been explored. One issue with HT isthe requirement for long beam-on times. A 2015 paper planned a 12 Gy in 10fraction treatment using RapidArc (Varian Medical Systems, Palo Alto, CA, USA)which would only take 7.2 minutes of beam-on time per fraction, compared to HTwhich can reach over half an hour [100]. This issue of long beam-on times wasof particular concern due to a lack of clinical resources in the country of origin ofthe paper where beam-on time is valuable. By using six arc fields in VMAT planswith 177 control points each, taking five hours of planning time, the study teamreached mean organ doses of 8.6 Gy to lung and 9.9 Gy to kidneys while ensuring95% of the PTV received the prescription dose. This also eliminated the need foran electron chest wall boost and demonstrated the feasibility of using VMAT forTBI in this setting.35A clinical trial using VMAT TBI treated seven leukemia patients between July2013 and July 2014 with 13.2 Gy in 8 fractions at 6 MV [102]. They used 8 VMATsegments, requiring a 180 degree patient rotation halfway through the treatment.The fractions each took 1.5-2 hours to execute due to extensive positioning require-ments, and each treatment required 40-45 hours of planning time with the currentlyavailable computational resources. The optimizer was steered towards lung spar-ing by placing helping structures inside the lung as necessary. A full body elasticgel bolus as well as a thermoplastic mask were used. While doses were reduced tocritical organs effectively, the increasing planning time could be a burden, but forthis centre, the alternative for the patients would have been to travel to other moredistant centres. There was a very low rate of early toxicities and five of the sevenpatients are disease-free at short-term follow-up.An issue with HT or VMAT is the possibility that circulating malignant cellswill not be completely irradiated, as the whole body is never being irradiated si-multaneously as for POP techniques with large field sizes. A 2010 study of TBIused stochastic and deterministic modelling to investigate this issue, treating bloodperfusion in the body as a sinusoid with a peak-to-peak displacement equal to theheight of a standard adult patient [104]. It was determined that TBI is not likely tobe prone to this issue.Two further studies have made progress in clinical implementation of a VMATTBI technique. One recent study describes a novel rotating immobilization sys-tem, allowing for the 90 degree rotation of a patient between an upper body VMATtreatment at standard SSD in three arcs and a lower body treatment of two or threepairs of anterior-posterior fields [105]. This reduces the chances of error due topatient positioning as only one setup is required at the beginning of the VMATtreatment. Compared to their previous standing POP technique with a linac at 550cm, the mean dose to the body increased from 11.5 Gy to 12.8 Gy in changingfrom a conventional technique to VMAT, and the mean dose to the lungs decreasedfrom 8.8 Gy to 7.9 Gy. The treatment was determined to be robust to position-ing uncertainties of 5 mm. Another study has described how to use a clinicallyavailable diode array, ArcCHECKTM (Sun Nuclear Inc, Melbourne, FL, USA), toperform quality assurance (QA) of field junctioning in the context of VMAT TBIwith multiple arcs, recognizing the trend towards these conformal TBI treatments36and a growing need for suitable QA tools [106].3.2.6 Total marrow irradiation and total lymphatic irradiationIt has been acknowledged that a different conditioning radiotherapy technique, to-tal marrow irradiation (TMI), is naturally suited for VMAT treatment planning.This treatment targets only bones with substantial red bone marrow content in thepatient rather than treating the whole body. On the same line of thought, total lym-phatic irradiation (TLI) has surfaced where only the lymphatic system is treated, aswell as total marrow and lymphatic irradiation (TMLI) where the PTV is the sumof the TLI and TMI PTVs. Other papers discuss combining a standard TBI treat-ment with a TMI boost to treat the entire body with additional dose only to bones,allowing for dose escalation beyond the traditional 12 Gy prescription to minimizethe chance of relapse without increasing the dose to critical organs, leading to im-proved outcomes without increased rates of toxicities.TMI target volumes include the marrow-containing skeleton, such as the upperextremities, pelvic bones, and thoracic bones. A group at the City of Hope NationalMedical Centre in Duarte, California has published several reports regarding bothTMI and TMLI using an HT technique, and had treated 120 patients with TMLIas of 2012 [107]. An earlier feasibility study in 2007 showed that TMI patients re-ceived 7.2 Gy or less to 84% of the lung volume while maintaining TBI-equivalentdoses elsewhere in the body [108]. MVCTs are used for positioning, and differentplanning CTs are taken during shallow breathing, inspiration, and expiration to de-termine margins for the ribs [109]. The contouring process, including contouringOARs, took up to 8 hours for TMI plans and 12-16 hours for TMLI plans in theseinitial studies. The lower extremities are treated with an anterior-posterior POPtreatment junctioned with the HT plan [108, 110]. A study was also undertakenusing VMAT planning with eight arcs and found similar TMLI dose conformitywith only 10.5 minutes of beam-on time compared to 18.7 minutes of beam-ontime for HT, although the VMAT plan required more patient positioning and setuptime [111]. One risk of using TMLI had been that by delivering less dose to thewhole body, there was a higher chance of extramedullary relapse (that is, relapseof the disease outside of the bone marrow region), but a retrospective study from37this clinic found there was no higher risk of extramedullary relapse compared toconventional TBI [107], using dose escalation up to 15 Gy. Most recently, a 2017paper from City of Hope describes a clinical trial of VMAT TMLI to treat 51 pa-tients with relapsed or refractory acute leukemia, using several dose escalation lev-els between 12 Gy and 20 Gy [112]. For prescriptions above 13.5 Gy, the brain andliver doses were limited to 12 Gy. The overall survival at one year was 55.5% with88% of patients in complete remission at 30 days, including 100% of the patientstreated at 20 Gy. Non-relapse mortality rates were 3.9% at 100 days and 8.1% atone year. Overall, the trial was well-tolerated and a phase II clinical trial is now inprogress.A group in Minneapolis has extensive TMI experience as well, using an HTtechnique for the full body. A feasibility study was included in a previously men-tioned work that also considered HT for TBI [94]. For TMI, this required a total of31 minutes of beam-on time, and 70 minutes of treatment time total in a first pa-tient experience including 10-15 minutes of MVCT collection time [108]. Breath-ing motion was accommodated for by adding a 1 cm margin around the CTV. Adose escalation study published ten years later [113] found that escalation to 15 Gywith TMI was feasible, while escalation to 18 Gy saw three out of six patients suc-cumbing to treatment-related mortality. A technical issue called the thread effectcan arise in TMI due to helical field junctioning, which can manifest as a dose in-homogeneity in peripheral doses away from the PTV in the shape of a ripple [114].The group found that the DVH parameters of the bones of the upper arms wereinfluenced by this, but this could be remedied by using different couch pitch valuesand the dose to the whole PTV was not significantly affected.Using TMI as a boost in addition to TBI has been performed in one clinic bydelivering 2 Gy in TMI on the day after a standard 12 Gy in 6 fractions deliveredtwice-daily [115]. Anticipating dosimetry challenges at the field junction region,they chose the level of the knees for the field junction where there is no bonemarrow present. A 4 mm margin was added to all of the bones in an isotropicexpansion in planning. Beam-on time ranged from 24 to 35 minutes for the TMIfraction.In efforts to decrease the time resources required for TMI planning, one studyfound that a treatment planning system called Voxel-Less OptimizationTM (Accu-38ray Inc, Sunnyvale, CA, USA) could plan and optimize TMI plans over four timesmore quickly than Eclipse [116], while an international multi-centre study has eval-uated the use of 2D tomograms for patient postioning instead of full MVCTs whichcan be produced in less than a minute instead of the usual ten to fifteen minutes[117].3.2.7 Treatment planning systems and in-vivo dosimetryCommercially available treatment planning systems (TPSs) such as EclipseTM (Var-ian Medical Systems, Palo Alto, CA, USA) or Pinnacle are able to perform dosecalculations given a set of beam configurations and a patient CT image. Thesedose calculations are usually designed for treatments delivered within a standardSSD range and have known limitations even within this range, particularly for tis-sue inhomogeneities and surface doses. At extended SSD, these calculations maybecome even less accurate, and precautions must be taken before making clinicaldecisions based on the output of a TPS calculation for TBI. This also emphasizesthe need for in-vivo dosimetry for quality assurance (QA) purposes.One comparison of TPS approaches for TBI planning was made by a group inQuebec, using different options in Pinnacle to determine the most accurate calcu-lation model despite the above challenges [118]. The TBI technique is a movingcouch technique with variable couch velocity. Two different beam commissioningapproaches were studied, one using default settings and the other using a TBI-specific beam model in Pinnacle that incorporates features such as large field sizeand extended SSD. Two different dose calculation algorithms, a 3D pencil beamalgorithm and a superposition-convolution algorithm, were used, and TPS doseswere compared to ion chamber measurements. Except for the build-up region andat depths beyond 20 cm, where errors were as high as 27% at some points, the max-imum discrepancy was 2% between TPS and measurement, using the TBI-specificbeam model in Pinnacle and the superposition-convolution algorithm. A correctionfactor was determined and applied which brought this error down to -0.68%.As mentioned before, several in-vivo dosimetry techniques have been appliedto TBI. One group with a translating couch TBI technique used MOSFETs on thepatient surface to measure entrance and exit doses at several points in the patient39[119]. After the first treatment fraction, if MOSFET readings varied by more than10% from the prescribed dose, the velocity of the couch is modulated appropri-ately in the remaining fractions to ensure the total prescription dose is reached. Ina retrospective study of 161 patients treated this way, while a few cases saw pa-tients with very large dose differences (over 35% in one region), most discrepan-cies were resolved in subsequent fractions with the immediate MOSFET feedback.This technique is effective as long as the treatment includes multiple fractions; forexample, a similar overdosage in a single 2 Gy fraction treatment would not beable to be compensated.There are similar retrospective studies in the literature of clinics using semicon-ductor diodes for in-vivo dosimetry, with two examples being a study in London,UK of 363 patients across three hospitals using bilateral 10 MV beams [120], andanother in Melbourne, Australia of 86 patients treated with either an 18 MV tech-nique or a combined 6 MV and 18 MV technique with bilateral beams [121]. Theadvantage of dosimetry using semiconductors includes being able to use severaldosimeters at a time with one electrometer system, immediate readout for anal-ysis during treatment fractions allowing for treatment adjustment if needed, andreusability of equipment, although they must be calibrated in TBI conditions, in-cluding but not limited to a low dose rate. The Australian study found a consistentdiscrepancy where the TPS overestimated the TBI doses by about 5% over sevenyears of patient data, for prescriptions of either 12 Gy or 13.2 Gy to whole body,with the worst agreement in the lung.Another group compared TPS calculations to dosimetry results, using the col-lapsed cone convolution (CCC) algorithm in Pinnacle and alanine dosimeters whichmake up part of their routine TBI treatments, in an anthropomorphic phantom at400 cm SSD and 18 MV [66]. Alanine dosimeters use a spectroscopy techniquewhere the brightness of a central line of the spectrum from the dosimeter is propor-tional to the dose delivered. Pinnacle gave doses lower than the alanine dosimeters,2% lower throughout the body axis and 3% lower in the lung region. While a smalldifference, this can and should be accounted for. TLDs and MOSFETs have alsobeen used in combination to compare dose calculations of the XiOTM planningsystem (CMS, St. Louis, MO, USA), using either fast-Fourier transform convolu-tion or multigrid superposition algorithms [62], which would be expected to face40similar challenges as other TPS calculations for TBI. Deviations between physicaldosimeters and the TPS were found in the lung region, although the data collectedfrom this dosimetry experiment allowed for a conversion factor to be determinedfor lung dose in treatment planning.Similarly, the field-in-field technique described earlier compared MOSFETreadings to calculations in Eclipse with pencil beam convolution and Batho in-homogeneity corrections [72]. MOSFET readings were 4-5.5% higher than TPScalculations for the lungs and shoulder regions. Further studies of TBI calculationsin Eclipse comparing the analytic anisotropic algorithm (AAA) and AcurosTM al-gorithm found that while the relative doses given by two treatment planning al-gorithms in Eclipse were accurate at 400 cm SSD, absolute dose was sometimeswrong in the TPS by over 10% compared to ion chamber measurements in bothhomogeneous and heterogeneous phantoms, reinforcing the fact that while TPScalculations may assist in TBI treatment planning, they cannot be the only methodof dose calculation [122].A recent report demonstrates the use of OSLDs for TBI dosimetry [123]. OSLDsare convenient dosimeters, although they are calibrated at standard dose rates of 6Gy per minute rather than the low dose rates customary for TBI. It was shown thatOSLDs can still be used for TBI purposes as they show negligible differences insensitivity between these irradiation conditions, and that the OSLDs simply needto be pre-irradiated to work around a known short-term signal fading effect. A cor-rection table is provided to compare readings between standard conditions and TBIconditions for up to 7 days following OSLD irradiation, along with other practicalconsiderations.3.2.8 Monte Carlo dosimetryMonte Carlo (MC) dosimetry is the gold standard for dosimetry in radiotherapy,described in detail in Chapter 4, but is not routinely applied to TBI. This is for sev-eral reasons, including complicated treatment configurations with custom-designedcompensators resting away from the patient, difficulty in accounting for beamspoilers or flattening filters, a lack of complete CT imaging in clinical protocols,difficulty in modelling treatment sources, challenges in simulating a moving pa-41tient or source, and perhaps most importantly, insufficient computational resourcesto handle the large number of histories required to reach acceptably low statisticaluncertainties in very large phantoms.One group produced a MC simulation of a static Cobalt-60 beam using acustom-designed MC code written in the C programming language, attemptingto simplify the simulation to reach appropriate computation times [124]. The sim-ulations were about ten times more efficient compared to a similar simulation inGEANT4 (a widely used set of MC codes from CERN), with results agreeingwithin 2%. Dose distributions in an anthropomorphic phantom and beam profilesin a water phantom showed good agreement between simulation and measurement,although the method requires refining before it can be applied to patients for routineuse. In the theoretical paper described earlier using protons for a TBI treatment,these calculations were also done with Monte Carlo simulations in GEANT4, buthere the focus of the work was on the feasbility of a new treatment design and noton the dosimetry of currently used techniques [75].A Canadian group treating TBI patients with a static Cobalt-60 beam has re-cently used MC simulations to design patient-specific compensators [125]. With-out organ dose compensators, doses in the MC simulations were strongly inho-mogeneous with doses ranging from -5% to +25% of the prescription dose, withnotable differences in the neck and lung regions. A deformable registration tool,VelocityAITM (Varian Medical Systems, Palo Alto, CA, USA), was used to de-form the prone dose distribution to the supine image and add the two distributionstogether. Based on this distribution produced without compensators, a new com-pensator of variable thickness was designed for the supine and prone treatmentsof one patient on a voxel-by-voxel basis to reduce the doses to prescription level.Such compensators have not yet been physically constructed, but the authors con-sider that they could be 3D printed. Only one patient was shown in this conceptualstudy and dose-volume information is not reported. The disadvantage to this tech-nique is the requirement to run two sets of simulations for each patient, one withand one without the compensators.A group in Sweden produced two papers, in 2013 and 2014, describing a MCapproach to studying TBI [64, 126]. Their TBI technique uses a 15 MV linacwith bilateral beams at an SSD of 460 cm, where the patient couch is rotated 18042degrees between treatments, obviating the need for repositioning or a second plan-ning image. Compensators are used to improve dose homogeneity and the arms arepositioned in such a way to assist in the reduction of lung dose. Dose profiles wereproduced for five patients with MC calculations, with the results of two simulationspresented to illustrate improvements in dose homogeneity with the use of compen-sators. Dose-volume information for individual organs is not presented. The 2014paper is technically focused and employs these MC techniques to determine whatthe optical thickness of a beam spoiler should be, as well as the most appropriatesource-to-spoiler distance, with respect to TBI beam characteristics and superficialdose.A report described earlier of a novel positioning scheme for a VMAT TBI tech-nique reports preliminary Monte Carlo organ doses and DVHs [105]. While itis clear that the team has achieved Monte Carlo dosimetry of their pre-existingPOP technique with organ compensators, comprehensive organ doses were not de-scribed as the focus of the report was the new positioning technique.Finally, a 1988 textbook written by pioneers in the application of MC tech-niques to medical physics describes an initial MC simulation of TBI from a Cobalt-60 source [127]. 5,000 histories per voxel of a phantom of approximately 30,000voxels were simulated, which took about 30 hours. The phantom included onlythoracic slices that also included lung tissue. The measurements were relativeonly, and showed that doses in the lung and at the sides of the body were morethan 120% of the dose at the center of the thorax. These results remain valid and,viewed together with the results presented in Chapter 5 of this thesis, elegantlydemonstrate 30 years of progress in MC for medical physics.3.3 TBI at the Vancouver Cancer CentreThe TBI technique in use at the VCC has been in use for at least 23 years, andwas updated in 2011 from a film-based planning technique to a CT-based planningtechnique [59, 128]. The most common prescription at our center is 12 Gy in 6fractions to the whole body delivered twice daily, with at least six hours betweenfractions, with other dose prescriptions at our clinic including 13.5 Gy in 9 frac-tions, 6 Gy in 4 fractions, 4 Gy in 2 fractions, and 2 Gy in 1 fraction. Between 3043and 50 patients are currently treated per year.A Cobalt-60 unit (Theratron 780C, Atomic Energy of Canada Ltd., Ottawa,Canada) with a head swivel feature is used to sweep a treatment field over thepatient at an extended SSD. Each fraction is comprised of two fields, with thepatient lying supine for one field and prone for the other. Because of the inversesquare loss of radiation intensity with distance, this technique would lead to muchlarger dose being delivered to the center of the patient, which is closest to thesource, compared to the extremities in the superior and inferior directions. Tomaintain a uniform dose profile across the patient, a PMMA flattening filter ispositioned below the treatment head and centered on the beam axis. Additionally,because lung tissue has a lower density than other tissues, the dose delivered wouldbe higher than prescription without any sort of compensation. To maintain the lungdose at prescription levels, lead lung block compensators are designed on a patient-by-patient basis and are placed on a thin plastic tray above the patient. A schematicof the treatment unit is shown in Figure 3.1.The treatment planning process for each patient begins with a simulation ap-pointment. A CT image volume is taken over the torso of the patient, spanningfrom a few centimetres below the umbilicus to the level of the neck. This is donefor both supine and prone treatment positions. Three treatment marks are made oneach side of the patient: one below the umbilicus to mark the beam central axisposition (reference mark t0), one at the sternum (reference mark t1), and one at thelevel of the nipples which represents a point at depth between the lungs (referencemark t2). Reference marks t1 and t2 are used to position the lung compensators ineach treatment fraction. Radio-opaque markers are placed on the patient at eachmark in each CT image for planning purposes, as shown in Figure 3.2. During pa-tient simulation, the patient SSD is measured in each treatment position. Patientswith a lower height may be positioned on styrofoam mats to reduce the SSD, whichallows for slightly shorter treatment times due to reduced inverse square losses.Based on the CT images collected, an oncologist delineates the contours of thelung compensators. With these contours and the measurements collected duringsimulation, a physicist determines the appropriate thickness for the lung compen-sators using a ray-tracing program that measures tissue deficits for rays traversinglung tissue through the lung contour from the source position [128]. This procedure44Figure 3.1: Schematic representation of a TBI treatment at the VCC. Patientis shown in the supine position. Labelled distances are drawn to scale,measured from the location of the source.also produces a MC phantom file for the patient CT which is used and describedin Chapter 4. The thickness of the compensators as calculated from the ray-tracingprogram is rounded up to the next 0.5 mm increment and the compensators are thenproduced by machinists in our department. Typical thicknesses for compensatorsrange from 1.0 to 3.0 mm.The treatment time for each field is determined as a function of the measuredSSD, the time elapsed between the treatment date and the source installation, andthe patient separation (thickness below reference mark t0) measured during simu-45(a)(b)Figure 3.2: CT images of a TBI patient, in the (a) supine and (b) prone posi-tions, with three radio-opaque markers visible in each scan at the patientsurface (t0, t1, and t2, from right to left).46lation. A reference dose, DR, is determined based on the prescribed dose per fieldDTx and the PDD at mid-separation depth d,DR =DTxPDD(d,SSD)(3.1)where the PDD is tabulated for an SSD of 160 cm and corrected with a Mayneordfactor for the actual SSD for the treatment.The dose rate at mid-separation below t0, D˙R, is calculated asD˙R = I FT BI FISL (3.2)where FT BI and FISL are factors accounting for the dose delivered on the beamcentral axis by a sweeping beam compared to a static beam, and the inverse squarelosses, respectively. I is an output factor based on the activity of the source on thetreatment start date,I = Io RDF exp(− t ln2t1/2) (3.3)where t1/2 is the half-life of Cobalt-60, 1921.5 days, and t is the time in dayselapsed since the last source change. The relative dose factor RDF is a ratio con-verting the measured intensity at 80 cm and a 10x10 cm2 field size to an extendedSSD with 35x35 cm2 field size as used in TBI. The final treatment time per field isobtained by dividing the reference dose by the dose rate,TTx =DRD˙R. (3.4)The TBI treatment pipeline was commissioned to deliver a mean body dosewithin 10% of the prescription dose (e.g. between 10.8-13.2 Gy for a 12 Gy pre-scription) based on this point dose calculation at patient mid-separation. A semi-conductor diode is placed in the beam at 100 cm from the source during treatmentsfor in-vivo dosimetry to ensure that treatments are delivered as planned.Given the half-life of Cobalt-60 of roughly five years and three months, treat-ment times double over the five-year clinical lifetime of the source. The previoustwo source installations at our centre occurred in July 2010 and August 2015. Typ-ical treatment times increase from around 6.5 minutes per field within two months47of a source change to around 13.0 minutes per field within two months prior to asource change. The source replacement cost is on the order of $300,000.A medical physicist is present at the first fraction of each treatment. The role ofthe physicist is to ensure proper positioning of the lung compensators. As shownin Figure 3.1, a digital imaging panel can be placed underneath the treatment bed.With this in place, very short beam exposures are delivered in each treatment posi-tion with the lung compensators in place. Based on the images acquired, the lungcompensator positions may be adjusted slightly at the discretion of the physicist.Bolus material is used between the legs, around the head, and between thearms and torso of the patient where possible. This allows for a full scatter dosecontribution to all parts of the body. During each treatment, radiation therapistscount the number of beam sweeps delivered to the patient. Treatments are onlydelivered in integer numbers of sweeps which may not necessarily subdivide tocontain the entire treatment time. At the final fraction, an additional sweep isadded if necessary to reduce the error caused by the discrete number of sweepscompared to the continuously-measured treatment time. There is also a timer errorinvolved in cobalt teletherapy units due to the finite amount of time required forthe source to move from its shielding unit to the beam opening during a treatment.For treatments longer than three minutes, which encompasses most TBI treatmentsat our clinic, 0.72 seconds are subtracted from the treatment time to accommodatefor this.48Chapter 4Monte Carlo dosimetry of the TBItechnique in Vancouver4.1 Introduction to Monte Carlo4.1.1 HistoryIn 1945, the first electronic computer was completed at the University of Penn-sylvania. The Electronic Numerical Integrator and Computer, or ENIAC, was awartime investment that accelerated routine military calculations. This was one ofmany key developments leading to the bombing of Hiroshima and Nagasaki and,consequently, the end of the Second World War [129].Researchers with the Manhattan Project at the Los Alamos Laboratory, includ-ing John von Neumann, were invited to see the ENIAC in Philadelphia before thewar had ended. Using the capability of the ENIAC to perform calculations thatwould have been intractable to compute by hand, these researchers were able todesign a computational model of a thermonuclear problem, the results of whichwere presented in 1946. Stanislaw Ulam, a mathematician with a keen interest inrandom processes, was in the audience and saw clearly the potential of this ma-chine to solve problems by statistical sampling. For instance, to determine theprobability of a shuffled deck yielding a successful outcome in a game of solitaire,would it not be better to lay out a hundred games and count the successes, ratherthan to spend countless hours on combinatorial calculations? [130] Such samplingtechniques had been conceived in mathematics before, but had not been pursueddue to the tediousness involved in manual calculations.49Ulam and von Neumann soon began collaborating on a statistical technique tostudy neutron chain reactions that would help predict the explosive behaviours ofnuclear weapons that were being designed [130]. Nicholas Metropolis suggestednaming this sampling technique Monte Carlo, after the Monte Carlo Casino inMonaco where an uncle of Ulam’s often gambled with borrowed money [129].The first paper describing the MC method was published in 1949 [131], and sincethen, the technique has been applied in problems across finance, molecular struc-ture, systems biology [132], and most importantly for this report, medical physics,where MC simulations have become the gold standard for dosimetry.It may seem strange that the machine can simulate the production ofa series of random numbers, but this is indeed possible.— Metropolis and Ulam, The Monte Carlo Method, 1949 [131]4.1.2 Monte Carlo simulations in radiation oncology physicsIn medical physics, the photon and electron interactions of interest are inherentlyprobabilistic in nature, as explained in Chapter 2. The interaction probabilities,amount of transferred energy, and initial states of created particles depends on theenergy of the incident photons and the density and composition of the materialbeing traversed. Random number generators provide a natural way of exploringthese processes where analytical calculations would be out of the question, giventhe complex shapes and compositions of patients and treatment geometries.An example of how random numbers are useful in medical physics is given byEquation 2.2 representing the number of photons attenuated from a beam traversinga material. The probability distribution of having a photon interact in a slab ofthickness dx after travelling distance x is given byp(x)dx = µe−µxdx. (4.1)The cumulative probability distribution P(x) giving the probability of interacting50at some point between x = 0 and a distance x is found by integrating,P(x) =∫p(x)dx (4.2a)P(x) =∫ xoµe−µx′dx′ (4.2b)P(x) = 1− e−µx (4.2c)and then, solving for the inverse function givesx =− 1µln(1−P(x)) (4.3a)x =− 1µln(R). (4.3b)These equations allow us to sample the distribution of photon distances travelledbefore interacting. R is a random number between 0 and 1, as this is equivalent tosampling one minus the value and saves a computational step. By substituting ran-dom numbers between 0 and 1 for the probability P(x), an appropriate distributionbetween 0 and ∞ is obtained. This is an example of the transformation method,one type of sampling technique used in MC simulations to make use of uniformdistributions of random numbers in physically meaningful ways.Each particle history in a MC simulation begins with an incident particle. Anincident high-energy photon can give rise to many other photons and charged par-ticles which are all tracked as part of the same particle history. Each particle istracked with a three-dimensional position, a three-dimensional direction of travel,an energy, a particle type, and other flags may be implemented as well to countadditional properties (e.g. the number of times the particle has interacted). Thesecoordinates of a high-dimensional phase space are updated incrementally as thetracked particle travels through the treatment geometry as visualized in Figure 4.1.In short, a distance of travel is selected before an interaction occurs, the type ofinteraction is determined based on the interaction cross-sections (provided the dis-tance travelled did not bring the particle outisde the treatment geometry), and theresulting particle types, energies, and directions are determined for each interac-tion, all of which use the production of random numbers in their calculations. This51Figure 4.1: Flowchart describing the basic elements of a Monte Carlo simu-lation of charged particle transport. The process continues until all newparticles and their subsequent generations leave the geometry or reachvery low energies.continues until all particles are removed from the simulation geometry or the par-ticle energies fall below set thresholds.Charged particles undergo very many more interactions than photons, and socondensed history techniques and approximations have been devised to simulatethese interactions to an appropriate level of physical accuracy within an accept-able allotment of computational time [133, 134]. A simple example of this is thecontinuous slowing down approximation (CSDA) where electrons are modelled astransferring energy continuously along their path of motion, rather than computingdiscrete chunks and collisions.524.1.3 EGSnrcGiven the roots of the MC method in nuclear physics, it is unsurprising that photontransport was among the early problems to harness it. A set of photon and neutrontransport codes was released in 1973 by Los Alamos researchers [135]. Manymore sets of MC codes have since been developed around the world; one such set,called EGSnrc, was developed in Ottawa to simulate coupled photon and electrontransport and is in routine use around the world for simulations in medical physics(EGS for “electron gamma shower” and nrc for the National Research Council ofCanada) [133].For the purposes of this thesis, each particle simulation in EGSnrc can bethought of as two sub-simulations: one for the treatment head, where the radio-therapy source produces photons that travel through and exit the head of the unit,and one for the scoring of dose in a given voxel geometry. These two processesare carried out in systems named BEAMnrc and DOSXYZnrc, respectively. Thetreatment geometry is given in an “egsphant” file (short for EGS phantom), wherea coordinate system is specified and each voxel has both a medium number (aninteger from 0-9) as well as a density value. A portion of an egsphant file for alung compensator is shown in Figure 4.2. The output file, called a “3ddose” file,has the same geometry and coordinate system as the egsphant file and contains, ineach voxel, both a dose value and the associated statistical uncertainty.Doses in DOSXYZnrc are scored in units of Gy per particle history. As a re-sult, as the number of simulated histories increases, it is not the dose values thatincrease; rather, the dose values converge while the statistical uncertainties de-crease, reflecting improved statistics from a larger amount of sampling. The initial3ddose file produced by DOSXYZnrc thus has doses on the order of 10−16 Gywhich must be converted to clinical values based on the amount of radiation deliv-ered by the treatment machines. An advantage to this technique is that the resultsof the MC simulation can easily be scaled to any prescription dose. Even withvery large numbers of histories, the statistical uncertainty can and must be reducedfurther by using several variance reduction techniques available in EGSnrc. Theseinclude range rejection, interaction forcing, bremsstrahlung splitting, and RussianRoulette, which are described in detail elsewhere [136–138]. Statistical uncer-53Figure 4.2: A screenshot from an egsphant file modelling a lung compen-sator. Here, 1 indicates an air voxel and 6 indicates a lead voxel. In thesame file, there is a corresponding matrix of material density values.tainties are also decreased further by using Savitzky-Golay filtering, described insection 4.3.2.4.2 Phantom production for TBI simulationsPart of the TBI treatment planning at our clinic already includes the production ofan egsphant file for each patient in the supine and prone positions using ctcreate, aspart of the calculation where the thickness of the lung compensators is determined(described in section 3.3). These patient egsphants, matching the patient CT scans,span from several centimetres below the umbilicus to the level of the neck. The54egsphants are routinely produced at a resolution of 0.25x0.25x0.25 cm3, and sothis resolution was used throughout this MC simulation pipeline.For a standard MC simulation, a phantom produced from a patient CT scanalone would be sufficient. However, for TBI, the phantom must include three ad-ditional components:1. Lead lung compensators, specific to the treatment field2. A thin piece of Perspex that holds these compensators in place above thepatient3. The polystyrene flattening filter below the treatment unit head.To construct a complete TBI MC phantom, egsphant files are first produced foreach of these components individually.The outlines of the lung compensators are stored as a set of points in thetreatment planning system as part of the treatment planning routine. However,in Eclipse, these points are given as coordinates centered over the lowest treat-ment reference marker, t0, and are scaled to a distance of 100 cm from the Cobaltsource. A Python (Python Software Foundation, Beaverton, OR, USA) script isused to scale and shift these contour points to reflect their actual positions over t2,139.7 cm from the source, and to convert these points to an appropriate file format.Next, the egsphants for both the lung compensators and the Perspex support areproduced by a MatlabTM (The Mathworks, Inc., Natick, MA, USA) script, usingthe same coordinate system as the existing patient egsphant. The Perspex supportfor standard TBI cases is 6 mm thick and 24.7 cm wide in the patient superior-inferior direction, centered over a point 1 cm superior to t2, and covers the patiententirely in the left-right direction.An egsphant for the polystyrene flattening filter was produced by CT scanningit and using ctcreate, carried out by a previous student. The coordinate grid of thefilter phantom is shifted for each patient so that its center is directly above the t0reference point with the bottom of the filter positioned 56.5 cm below the Cobaltsource. The filter is about 70 cm long in the superior-inferior direction, which islonger than the planning CT used. To ensure the entire filter phantom fits above the55patient phantom in simulation, most patient phantoms were padded with 35 cm ofair voxels in the inferior direction with a C++ script.To produce the final MC phantom for the patient, these four phantoms arestacked using a series of C++ scripts. Using measurements from the treatmentroom and points from the planning CT images, phantoms are produced with ap-propriate distances between each component to reflect the treatment setup. Toreduce computational times, single large air voxels are placed between the compo-nents rather than filling a grid with air between them. A visual representation ofthis phantom for a patient is shown in Figure 4.3, with an axial and sagittal slice ofthe full geometry and a coronal slice to show the shape of the lung compensators.The display from the EGSnrc simulation software is unable to handle variable res-olutions, so Figure 4.3a does not accurately represent the air gaps between thecomponents. However, the simulation calculations use the correct geometry.A point of interest in the phantom production pipeline is that we did not usethe SSD values measured by radiation therapists at the patient simulation appoint-ments. Rather, we measured the SSD from the patient CT scans based on theseparation of the patient (measured below point t0), the number of styrofoam padsplaced below the patient (each 3.7 cm thick), and the fixed distance of 184.5 cmfrom the source to the treatment bed surface. This is because the patient may moveslightly while being measured in treatment simulation, and the measurements arebeing collected in a clinical setting where the number of tasks to be accomplishedin a limited time make it impractical to measure the treatment SSD with precisionbetter than ± 1 cm. By looking at CT images, we obtained an SSD that reflects theactual position of the patient CT scan with respect to the source during treatment.The measured SSDs from the CT scans were typically 0-2 cm larger than the valuesprovided in the patient treatment specifications.This entire procedure was performed twice for each patient, once for the supinetreatment and once for the prone treatment, resulting in two complete TBI phan-toms for each patient. It currently takes roughly 1.5 hours to complete this processfor each patient. For a typical adult patient, the initial CT egsphant contains around6.0∗106 voxels with a filesize around 50 MB, and the complete TBI egsphant con-tains around 1.2∗107 voxels with a filesize around 125 MB.56(a)(b)Figure 4.3: Slices of a MC phantom. In (a), the flattening filter (1), thelung compensators (2), the plastic tray (3), and the patient (4) are allshown in both axial and sagittal views. In (b), the lung compensatorsare shown. The visualization software in EGSnrc is unable to displayvariable resolution which is used in the beam direction, but the distancesare correctly to scale in the simulations.574.3 Simulations4.3.1 Cobalt source modelAs described in section 4.1.3, MC simulated particles first travel through a sourcein BEAMnrc, and then are scored in a phantom with DOSXYZnrc. The Cobaltsource in BEAMnrc was commissioned by Dr. Tony Teke and collaborators and isdescribed in his Ph.D thesis work [139]. In brief, the Cobalt unit head is modelledin a cylindrical geometry with sets of tungsten and lead slabs reflecting the con-struction of the source housing container. The source itself is modelled based on aCobalt spectrum measured by a group in Ottawa [30].To ensure proper commissioning of the Cobalt source in BEAMnrc, several val-idation measurements were conducted in water tank measurements, also describedin Dr. Teke’s thesis. Lateral beam profiles at four depths in a water tank and aPDD curve from the top of the tank to 25 cm of depth along the beam axis weremeasured with an ionization chamber, treating at an SSD of 80 cm and a field sizeof 35x35 cm2 at 80 cm (which is the field size used in TBI treatments). An identi-cal treatment geometry for the water tank was modelled in DOSXYZnrc. Both thebeam profile shapes and absolute dose values at several points in the tank showedexcellent agreement between simulation and measurement.4.3.2 Simulation procedure and de-noisingFor our sweeping beam TBI technique, we require a Monte Carlo source that isable to sweep over the patient geometry. Two such sources in DOSXYZnrc, calledSource 20 and Source 21, were invented by Popescu and Lobo to model contin-uously varying beam configurations [140]. Source 21 is applied to the followingsimulations to harness this capability. The 90 degree sweep of the Cobalt unit headis modelled with twenty-one control points spaced equally over an arc in the patientphantom with coordinates depending on the patient treatment markers and SSD,yi = t0y +SSDcos(90−θi) (4.4)zi = t0z−SSDsin(90−θi) (4.5)58where t0y and t0z correspond to the y and z coordinates of reference marker t0 andθ is the angle of the treatment head with respect to the horizontal (that is, an angleof zero degrees would represent the treatment head pointing directly at a wall).In each patient simulation, two billion histories were simulated. This num-ber is justified in the following section. The default cutoff energies in EGSnrc forelectrons and photons (0.7 MeV and 0.01 MeV respectively) were used. A factorof 1000 was used for directional bremsstrahlung splitting [136]. MC simulationsat our centre are run in parallel, spread across as many as 256 jobs operating si-multaneously. The Condor High-Throughput Computing software housed at theKamloops Data Centre serves as the platform for our simulations. A typical simu-lation for one treatment position with two billion histories in an adult-sized patient,split into 250 jobs, takes approximately 75 minutes to complete.Recall that the phantoms used in MC simulations are composite phantoms in-cluding several materials in addition to the patient. The first step in processing thesimulation data is to strip the 3ddose files to remove the doses to the flattening fil-ter, lung compensators, and Perspex tray, as well as the 35 cm of air that was oftenadded inferior to the patient CT scan to accommodate the full flattening filter. Thestripped 3ddose file is thus restored to a uniform resolution of 0.25x0.25x0.25 cm3with the same geometry as the initial patient egsphant. The initial dose file is over300 MB in size, and after stripping and processing, the file is around 125 MB insize.Given the nature of a MC simulation, there is bound to be a level of statisticalnoise present in the data, whereas a real dose distribution would be continuousin the patient. To reduce the statistical uncertainties of the dose files, a three-dimensional Savitzky-Golay filter is applied. This smoothing filter passes throughthe data in a 7x7x7 region of voxels at a time, but a smoothing fit for the data isaccepted in each region only if a statistical test is passed that determines whetherthe fit is harmlessly minimizing the statistical noise or erasing a true feature of thedose distribution [141, 142].After Savitzky-Golay de-noising, the 3ddose file is converted from Gy per par-ticle values to real dose values in Gy. The calibration factor was determined byan ion chamber measurement described in section 4.3.4 and is adjusted for eachpatient based on the date of their treatment and their treatment time.594.3.3 OptimizationThe very large size of these TBI MC phantoms compared to those used for standardMC dosimetry procedures exacerbates the issue of statistical noise in the simula-tions. To decrease statistical uncertainties and mitigate the issue of hot and coldspikes in the dose distributions, we can run several independent MC simulations foreach phantom and then combine these sub-simulation results voxel-by-voxel with aweighted mean, rather than use a boundlessly increasing number of histories [142].This strategy then raises the following question: in the interest of producing thehighest possible quality of data in the least amount of computational time, what isthe optimal number of sub-simulations to use for each simulation, and what is theoptimal number of histories to use for each one? To answer this question system-atically, for one supine patient phantom, twenty-five MC simulations were carriedout: five simulations each of 1.0∗109, 1.5∗109, 2.0∗109, 2.5∗109, and 3.0∗109histories, using different seeds in the egsinp files to ensure the random numbersequences were different for each simulation. Then, dose distributions were pro-duced by taking the weighted mean of either two, three, four, or five of the simu-lation results of a given number of particle histories. All dose distributions werestripped and de-noised individually, and then the composite dose distribution aftertaking the weighted mean was de-noised a second time. Simulations of differentparticle history numbers were not mixed, and individual simulations were includedfor comparison as well. The total number of final dose files for comparison wastwenty-five: five individual simulation results, and twenty composites from two tofive sub-simulations.The composite simulation of five times three billion histories was taken as thehighest quality for comparison with the other simulations. The goal was to deter-mine the smallest number of histories and simulations that could be used to obtainsimilar organ dose results as this composite. The results of the simulations werecompared by looking at the dose to the left and right lungs of the patient as well asa spherical contour 2.5 cm in diameter centered at a depth of 5 cm below treatmentmarker t0. Both the mean and standard deviation of the doses within the contourswere considered and an abridged version of the results is given in Table 4.1. Itis seen that there is a noticeable decrease in the standard deviations and thus im-60Histories persub-simulationNumber ofsub-simulationsSpherical contour Left lung Right lungMean (Gy) SD Mean (Gy) SD Mean (Gy) SD1∗109 1 6.76 5.8% 5.18 12.2% 5.38 11.7%2∗109 2 6.76 3.1% 5.16 9.5% 5.36 8.5%3∗109 5 6.75 2.8% 5.16 8.9% 5.37 8.6%Table 4.1: Abridged results of an optimization study. Dose information areshown from MC simulations of the same supine treatment (6 Gy toprescription point) with different numbers of particle histories and sub-simulations. The simulation with the most histories was taken as the bestpossible simulation quality. While the dose results remain similar withdecreasing numbers of particles, the images can be seen to have morestatistical noise as shown in Figures 4.4.proved homogeneity when using more than one simulation, but the improvementfrom two times two billion histories to five times three billion histories was not asstrong, despite the large increase in total histories used. The distributions were alsocompared qualitatively, and three cases are shown in Figures 4.4. Based on theseresults, it was decided that two times two billion histories would give dose dataof acceptable quality in a reasonable amount of computation time. The statisticalerror of a composite dose file from two times two billion particle histories for astandard adult-sized patient is on the order of 2% in the lung and 1% in the rest ofthe body.It is worth noting that while the organ doses are visibly noisier with fewersimulated particle histories, the organ doses are not dramatically different in thesecases. This is because the statistical errors cancel out in the large organ contours.In cases where a point dose will need to be examined, a larger number of sub-simulations and histories should be used (e.g. five times three billion histories), butin cases where only organ doses will be considered, smaller simulation numberscan be used to obtain the same results.4.3.4 CalibrationFor linear accelerators, the amount of radiation delivered is parameterized by mon-itor units (MUs). The conversion of DOSXYZnrc outputs to clinically meaningful61(a) One simulation, one billion histories.(b) Two simulations, two billion histories each.(c) Five simulations, three billion histories each.Figure 4.4: Isodoses for the three dose distributions corresponding to thethree conditions listed in Table 4.1. The distributions become less noisywith more simulated particle histories at the expense of computationaltime.62dose values for linear accelerator MC simulations is outlined by Popescu et al[143]. For a Cobalt unit, however, this method does not work, as the treatment isparameterized by time rather than MUs. In order to scale our simulation resultsfrom Gy per particle to physical dose in Gy, a ratio between measured dose andsimulated dose at a point was used for a radiation delivery resembling a TBI treat-ment. The measurement was made with an ion chamber in a block of solid waterat a depth of 10 cm on the beam axis. The solid water phantom had dimensions40x40x20 cm3, centered beneath the treatment head at an SSD of 160.2 cm. Theflattening filter was positioned between the treatment head and solid water. Thesolid water was irradiated during five full sweeps of the gantry with a field size of35x35 cm2 at 80 cm from the source. A picture of the setup is shown in Figure 4.5.A MC simulation was executed with the same treatment geometry, appendingthe flattening filter phantom above the solid water phantom at an appropriate dis-tance. Five simulations of three billion particles each were used, as the desiredmeasurement was a point dose at the location of the ion chamber, rather than somecontour where statistical uncertainties would cancel out. A visualization of thephantom is given in Figure 4.6. As before, the visualization software does notaccurately depict the large air voxel placed between the filter and solid water phan-toms, but appropriate spacing is in place for the simulations.Using the ratio of the measured dose rate and simulated dose, with units ofGy per minute and Gy per particle respectively, a conversion factor of the numberof particles per minute emitted by the Cobalt source is obtained. This value is tocorrect for source decay over the time between the measurement and the calibrationin August 2015, giving 1.17∗1015 particles emitted per minute. The reason fordialling the dose rate back to the calibration date is to provide a more convenientvalue that can be adjusted for each patient based on the day that they were treated.Therefore, for each patient simulated in Chapter 5, the conversion factor iscalculated by multiplying the above value by three factors: an exponential decayfactor corresponding to the treatment date, the treatment time for the patient perfield, and the number six to reflect the six treatments given for a 12 Gy treatment.This could easily be be applied to other prescription doses as well. The supineand prone dose files obtained this way are ultimately added together, described insection 4.4.63Figure 4.5: Setup of the calibration measurement. The clinical flattening fil-ter is in place and an ion chamber is inserted at a depth of 10 cm in a40x40x20 cm3 block of solid water. The treatment was delivered overfive sweeps of the beam.4.3.5 Validation measurementsInitial validation measurements were undertaken when the MC Cobalt source wascommissioned by a previous graduate student as described in section 4.3.1. We per-formed a new set of validation measurements to ensure that the strong fit betweenmeasured and simulated doses holds true for a more realistic patient simulation andnot only in a homogeneous tank of water.Two patients, who are not included in the study in Chapter 5, received treat-64Figure 4.6: Visualization of the simulated dose distribution corresponding tothe measurements shown in Figure 4.5. The coordinate system in thesimulation is rotated; here, the beam axis is from left to right acrossthe page. Most dose is delivered to the flattening filter with a relativelysmall amount delivered to water. The dose to the point correspondingto the ion chamber location in solid water was recorded.ments on a different treatment protocol with non-standard lead lung compensators.One patient had a single set of 12 mm blocks, while another patient had a standardset of 2 mm blocks with a second narrower set of 11 mm blocks to sit on top of thefirst set. We used these non-standard lung blocks in both a simulation of a treat-ment and a measurement. To create the phantom, blocks of solid water were used tocreate a shape roughly the size and depth of a patient. Two sets of lung-equivalentfoam were used as well to provide a non-homogeneous tissue distribution to mimicthe non-uniform patient geometry. A picture of the setup is shown in Figures 4.7.65The phantom was treated at extended SSD and ion chamber measurements weretaken in a piece of solid water between the two pieces of lung foam. Sets of threeOSLDs were used as well, in two places: one set at the patient surface underneath6 mm of paraffix wax buildup, and one set beneath the first lung foam block, ontop of solid water. The beam was turned on for 20 sweeps of the unit head.After these measurements, the phantom was transported to a CT simulator andwas rebuilt on the CT couch. The phantom was scanned (without an ion chamberin place, leaving an air gap in the phantom) and converted to a MC phantom withctcreate. Five sub-simulations of three billion histories each were produced for twoMC phantoms, which were identical apart from the two different sets of lung com-pensators that were used. Statistical uncertainties were below 1% and simulatedisodoses for one of the treatments is shown in Figure 4.8.The results of both measurement and simulated doses are given in Table 4.2. Insimulation, the lung blocks were intentionally offset slightly from the tip of the ionchamber insert in the simulation so that the penumbra from the lung blocks wouldoverlap with solid water at an appropriate depth instead of the air gap in the phan-tom from the ion chamber insert. Additionally, there were no OSLDs present inthe CT scan of the phantom. To obtain the simulated doses at points representativeof where the OSLDs were placed during the measurement from the Cobalt unit, wemeasured at a depth of 6 mm below the top of the phantom for the surface OSLDset to accommodate for the thickness of the OSLD and the buildup provided by theparaffin wax, and a depth of 1 mm into the solid water beneath the first lung block.The results show strong agreement at most points of comparison. Reasons for dif-ferences between points could be due to several factors: changes in the positions ofphantom components when transporting the phantom from the Cobalt unit to theCT scanner, and when transferring from the TBI treatment couch to the CT couch;differences in positioning of the top pieces of solid water and lung foam with andwithout the OSLDs present, which produced small air gaps in the measurementthat were not present in the MC phantom; inaccuracies in the approximation usedto compare measured OSLD doses with doses at depth in solid water (e.g. this didnot account for small inverse square effects as a result of collecting data furtherfrom the source). The goal of these measurements was not necessarily to see 100%agreement, but rather to test the TBI MC procedure in a complex, heterogeneous66(a)(b)Figure 4.7: Experimental setup used for the validation measurements attreatment conditions. The flattening filter is also in place as shown inFigure 4.5. (a) The interior of the phantom includes two pieces of lung-equivalent foam to mimic a heterogeneous patient environment. A set ofOSLDs is placed below the first piece of foam. (b) Solid water is placedabove the top foam block, and another OSLD is placed on top of this.An ion chamber is placed at depth at a position within the penumbra ofthe lung compensators.67Figure 4.8: Isodoses for the simulation of the experimental setup shown inFigure 4.7. The treatment corresponded to a 1 Gy prescription to mid-separation in the phantom. The simulated phantom includes two lung-equivalent blocks (1), an air gap in the ion chamber insert (2), and vari-ous blocks of solid water (rest of phantom).geometry resembling a treatment. Given these results in tandem with the earlierwater tank measurements when the MC Cobalt source was commissioned, we aresatisfied and confident in the doses reported by this MC dosimetry technique.4.4 Deformable registration of dose distributionsSo far, the pipeline for simulating a single MC phantom has been described ineither the prone or supine positions. To obtain organ dose data, the prone andsupine distributions must be combined. The shape of the patient changes betweenthe supine and prone positions for reasons including differential compression ofmobile tissues in the two positions, or different positions of the arms or head rela-tive to the torso. Additionally, the CT scanning boundaries on the patient may beslightly different between the two scans. As a result, the dose files from the supineand prone treatments cannot simply be overlaid, as the geometry is not the same.Instead, an intensity-based deformable registration tool in MIM MaestroTM v6.6.8(MIM Software Inc., Cleveland, OH, USA) is used to add the doses together basedon the positioning of the tissues in each image. The prone dose data is deformed to68Case # 1 Case # 2Simulation (Gy) Measurement (Gy) Simulation (Gy) Measurement (Gy)Ion chamber 0.59 0.54 0.55 0.55OSLD (surface) 0.68 0.63 0.61 0.61OSLD (lung) 0.64 0.63 0.59 0.65Table 4.2: Results of a validation study. Measured and simulated doses arepresented for the two OSLD measurements and ion chamber measure-ment pictured in Figure 4.7 compared with the simulated doses at thosepoints in the corresponding MC phantom. Given the uncertainties in-volved in producing an MC phantom exactly identical to the treatmentdelivered on a different unit, as well as the very close agreement betweenmeasurement and simulation in earlier homogeneous water tank studies,the agreement here between measurement and simulation is acceptable.the supine image volume, and these deformed data can then be added to the supinedata after they have been registered to a common geometry. A sample set of supinedose, prone dose, and total dose distributions acquired with this technique is givenin Figure 4.9.The exact details of the deformation algorithms of commercially available soft-ware are often obscured from users for competitive reasons, although some studieshave been undertaken to study the effectiveness of the registration in MIM Mae-stro alongside other similar commercially available tools [144, 145]. Deformableregistration is used clinically for cases including patients being treated in the sametreatment position on different dates who have lost weight since an earlier treat-ment, or have tumour shrinkage during a radiation therapy course with a largenumber of fractions, or for changes in internal geometry, such as organ resectionsor changes in bladder filling levels between scans. Additionally, the tools can beused when patients are being treated with a second course of radiotherapy to deter-mine which regions have already received dose in the first course. A full validationof the deformable registration algorithm to measure the uncertainties borne by theregistration would be a highly nontrivial and challenging task beyond the scopeof this thesis. However, given the large CT images with a large amount of bonyanatomy present to guide the registration, and by visual inspection of the results69Figure 4.9: Final results of a MC simulation for a patient. The three imagescorrespond to the supine treatment, prone treatment, and final distribu-tion. The prone treatment is displayed upside-down for ease of compar-ison with the other figures.70and from the experience of physicists using MIM Maestro clinically in the depart-ment, we are confident in the results obtained from this technique.It is possible to improve the quality of the deformable registration in MIMMaestro using a tool called RegRefine. With this tool, an expert user can place“locks” over certain areas of the CT image at intermediate stages when promptedby the software, to place more importance on certain regions and to reiterate reg-istrations over challenging areas until a satisfactory result is produced. When thisTBI MC technique is implemented clinically, locks could be applied to our tech-nique, but in the interest of reproducibility of the results in the next chapter, nomanual intervention was involved in the deformable registrations for this study.71Chapter 5Retrospective dose analysis of a TBIpatient cohort5.1 Purpose and aimThe TBI technique used in our clinic is described in detail in Chapter 3 alongsideapproaches in place at other institutions. Our technique, like the majority of tech-niques in place around the world, is a POP treatment with custom-designed lungblocks for each patient.Our technique is effective, being commissioned to deliver a dose within 10%of prescription and has been in place for over 23 years [59]. However, given theadvances in technology and radiation oncology physics in recent years, it is nowpossible to achieve a higher degree of conformity than the current 10% limit; formost other radiation treatments, the allowed variation from a prescription dose is-5% to +7%.There are some features of our current treatment that could be improved in anew technique. For many patients, the experience of being lowered to a floor-leveltreatment bed for an extended period of time, especially for the portion spent in theprone position, can be uncomfortable. A treatment that allows for the patient to betreated on a couch of standard height would represent a shift towards an improvedTBI patient experience. Moreover, if the patient can be treated in one positioninstead of two, then there is less time spent positioning the patient and less oppor-tunity for patient positioning errors. Additionally, the process of producing lungblocks for each patient is resource-intensive and time-consuming, and lung blocksof a single shape and thickness cannot fully account for the three-dimensional vol-72ume of the lung in the patient. The lung compensators also block the tissue anteriorand posterior to the lungs. This technique was commissioned to maintain the lungdoses at prescription level, but it is not trivial to adjust the level of shielding if anew protocol required further lung dose reductions or other organ dose reductions.A more conformal TBI technique that does not use physical lung blocks could pro-vide the same amount of organ sparing with a simpler clinical workflow, and wouldallow for different protocols to be enacted should they become of interest.These challenges could be overcome by a technique such as helical tomother-apy or VMAT, as described in Chapter 3. Our clinic has a high level of physicistexpertise in delivering VMAT treatments, making this a logical choice in upgradingour TBI technique. In VMAT, organ dose constraints must be specified in advance,and the treatment is inverse planned from these constraints. To reproduce our ex-isting treatment on a new platform, we thus require organ doses from our sweepingCobalt-60 technique, which have not been available. With the novel MC approachdescribed in Chapter 4, we can now obtain organ doses from TBI patients to satisfythis goal and bring our clinic closer to implementing a new technique. Moreover,our clinical TBI experience is based on the sweeping-beam method, and changecannot be implemented to delivered doses without a good understanding of the im-pact on clinical outcomes, further justifying this study. In this study, we simulatedthe treatments of recently treated TBI patients in our clinic and present organ dosedata from this population.5.2 Methods: patient selection and contouring20 patients were selected who had received 12 Gy in 6 fractions. Patients couldonly be simulated if the treatment took place after 2011 when our technique wasupdated from a film-based method to the current CT-based method [128]. Forconvenience in calculations, we only chose patients who had been treated sincethe last Cobalt source replacement in August 2015. Patients with a range of bodysizes were selected, as measured by the anterior-posterior separation beneath thet0 reference mark, (range: 10.1 cm - 23.6 cm). The patients include 10 males and10 females, including pediatric and adult patients (age range at treatment: 3 years -51 years). Patients with resected organs were excluded from the study, as were any73patients whose data records were non-standard for any reason (e.g. uneven spacingbetween rows in the CT phantoms, lung compensators saved in an unusual format).Based on the toxicities that are most of interest in TBI as outlined in section3.2.4, the contoured organs for this study were the thyroid gland, the lungs, thekidneys, and the liver. These were contoured by the author in MIM Maestro withsample contours for a patient shown in Figures 5.1.A body contour was produced for each patient with the four most superiorslices (2 cm) and inferior-most eight slices (4 cm) subtracted from the contour. Inthese two regions, because the CT image volume does not include superior and in-ferior portions of the body that would otherwise provide scatter dose contributionsto these slices in our phantom, we subtract these areas to remove slices from themean body dose calculation that are known to be underdosed. A previous graduatestudent in our department studied this underdosing effect for TBI MC simulations,measuring dose at the prescription point beneath reference mark t0 [139]. It wasfound that, in this direction, having only 4 cm of scatter gave 8% less dose thana measurement with 16 cm of scatter, with the dose difference falling off dramati-cally with less than 4 cm of scatter. This explains our choice to remove 4 cm of CTslices from the inferior edge of the body contour. We assume that in the superiordirection, the same effect is present, but less pronounced because the missing headoccupies a smaller volume than the missing torso region and would thus provideless scatter if it was present in the CT volumes. The thyroid gland of one patientresided completely within the top four slices that are expected to have underesti-mated dose, and part of the gland was truncated by the anatomical location of theCT cutoff in the superior direction. As a result, this thyroid is excluded from theorgan results in Figure 5.5 in the next section.For some patients, the prone and supine CT images were not cut off at thesame anatomical locations in treatment planning, and the deformable registrationwas not able to produce a summed dose in voxels that did not have a correspondingpoint in the other image. Where it was clear that this happened for a patient, fur-ther crops were made from the body contour at the superior-inferior edges of theimage. Additionally, the arms were removed from the body contour up to the levelof the shoulder. The primary reason was due to the challenge of the deformableregistration tool in these areas, as the arms are located at different positions relative74(a)(b)Figure 5.1: Contours shown in MIM Maestro for a patient. (a) The bodycontour (red), lungs (blue and purple), kidneys (orange and brown), andliver (green). (b) The body contour (red) and thyroid gland (yellow).75to the rest of the body in the supine and prone positions. Moreover, as we did notmodel bolus in our simulations that is present in treatment to provide scatter dose tothe arms, and because of CT artifacts in the arms that would interfere with MonteCarlo dose calculations in these regions (visible in Figure 5.1a), the true body dosewas not accurately represented in the arms in our simulations.Within each contour, the mean, minimum, and maximum doses, along with thestandard deviation of the dose in each contour, were determined with MIM Mae-stro. Additionally, a homogeneity index (HI) was calculated for the body contours,HI =D5−D95D50(5.1)to provide another metric of dose conformality within the target volume. Dx is thedose that is delivered to the x% of voxels within the contour receiving the highestdoses. There are different definitions for HI in the literature [146]; here, D5 andD95 were chosen for robustness against hot or cold spots due to statistical noise. Inthis definition of HI, a perfectly conformal plan would have an HI of zero.5.3 Results and discussionAggregate results representing the organ doses and the standard deviations of organdoses across all 20 patients are shown in Tables 5.1 and 5.2. This data provides ourclinical team with the range of doses that are delivered to each organ in the interestof implementing VMAT constraints on a future technique.However, these data do not account for the size of the patient. In Figure 5.2,it is demonstrated that there is a strong relationship between the size of the patientand the mean body dose as measured within our body contours (R2 = 0.91 witheither linear or exponential models). All mean body doses fell within the 12 Gy ±10% prescription dose, although the smaller patients received doses on the loweredge of the threshold while only the largest patient received a mean dose of 12Gy. One reason why most patients are receiving less dose than prescription is theblocking of tissue anterior and inferior to the lungs by the lead compensators.A factor that explains the variation of mean body dose with patient size is thechanging shape of POP dose distributions with increasing patient size. This effect76Mean SD Min MaxLeft lung 10.99 0.29 10.44 11.62Right lung 10.48 0.30 10.48 11.49Left kidney 11.65 0.33 10.98 12.25Right kidney 11.73 0.42 10.90 12.29Liver 11.24 0.27 10.73 11.69Thyroid 11.79 0.45 10.84 12.58Body contour 11.46 0.33 10.93 12.01Table 5.1: Aggregate organ dose results (in Gy) across the 20 simulated pa-tients. The values represent the mean and standard deviation of the aver-age doses delivered to each organ. The maximum and minimum averagedoses delivered to patients for each organ are shown.Mean of SD Min SD Max SDLeft lung 0.49 0.35 0.67Right lung 0.48 0.29 0.68Left kidney 0.26 0.16 0.40Right kidney 0.28 0.19 0.47Liver 0.37 0.25 0.53Thyroid 0.20 0.14 0.29Body contour 0.78 0.64 0.86Table 5.2: Standard deviation information of the doses delivered to each or-gan are shown (in Gy) to illustrate the level of dose homogeneity in the fi-nal simulated distributions. The average of the standard deviations acrossthe 20 patients is given for each organ, as well as the minimum and max-imum standard deviations.is illustrated in Figure 5.3 with Cobalt-60 PDD curves for three different amountsof water at extended SSD. These curves are normalized to 12 Gy at mid-separation,mimicking the prescription of a TBI patient. With an increasing depth of water,higher peaks are seen in the PDD curves, and the average doses for these threesizes are 12.0 Gy, 12.1 Gy, and 12.3 Gy, respectively.Another contribution to the lower mean body dose in smaller patients is due tolower scatter dose contributions throughout the body. While patients are positioned77Figure 5.2: Variation of TBI mean body dose with patient size for the 20 pa-tients. Size is measured by the separation depth at reference mark t0.The error bars show the standard deviation of the dose delivered to vox-els inside the body contour. All mean doses are within the prescriptiontolerance. It can be seen that the mean doses increase with patient size.with bolus around the head and the legs, smaller patients would still receive a lowerscatter dose contribution, simply from having less tissue for scatter. This variationwith patient size does not appear in the standard deviations of the voxels within thebody contours. That is, while the doses delivered to the smaller patients were lower,the conformality of the treatments was the same as for larger patients. Accordingly,the HI values are not impacted by patient size, with ranges from 0.14 to 0.22 andan average value of 0.19 across the patient population, shown in Figure 5.4.Because of the increasing mean body dose with patient size, the organ doses78Figure 5.3: PDD curves for POP beams in water for three patient widths.Tabulated Cobalt-60 PDD data [147] (collimator setting 35x35 cm2,SSD 160 cm calculated using Mayneord factors) are added together andscaled to 12 Gy at mid-separation.are better represented relative to the mean body dose for each patient as shownin Figure 5.5. The straight line in each of these figures represents the mean bodydose; points above these lines represent organs receiving more than the mean bodydose for a given patient, and vice-versa. From these figures, we can see that thelungs and liver consistently receive lower doses than the mean body dose in eachpatient, and vice-versa for the kidneys and thyroid. A sample dose-volume his-togram (DVH) is shown in Figure 5.6 for an individual patient. An ideal DVH fora patient would approach the shape of a rectangle with 100% of the PTV receivingthe prescription dose, although real treatment plans do not achieve this.The results from this retrospective study, in addition to serving as a basis forVMAT organ dose constraint implementation, suggest that the treatment time for79Figure 5.4: The homogeneity indices of the dose delivered to the body con-tour for each patient plotted against patient size. There is no clear trend,indicating that the homogeneity of the dose does not change with size.smaller patients could be increased if it is desired for them to receive doses closerto 12 Gy. This would not be challenging to implement. For individual organ doses,on the other hand, while there exists some flexibility in our technique to modulateorgan doses (such as modified lung shielding or shielding for additional organs),it is much more challenging to meaningfully adjust these without upgrading to atechnique such as VMAT or helical tomotherapy.It is important to recall that the mean body doses in all patients are likely to besystematically underdosed by a few percent because of the missing scatter in eachdirection, even with the precautions in the body contours described in section 5.2.Additionally, the bolus material above the shoulders and around the head and neckare not included in our simulations at present, and could lead to further simulatedunderdosing of the patient.80Figure 5.5: Organ dose data plotted against mean body dose. Each subplotincludes a line of equality for ease of interpretation. Points above theline indicate that the organ dose for the patient was above the meanbody dose, and vice-versa. The lungs and liver are consistently belowthe mean body dose, and the kidneys and thyroid are consistently above.The error bars are the standard deviation of the dose delivered to voxelswithin the organ contours.81Figure 5.6: Dose-volume histogram for a simulated patient.82Chapter 6Future work6.1 Ongoing investigationsThis thesis presents a novel Monte Carlo dosimetry technique for TBI as well asthe findings from a retrospective study. This project has given rise to several moreopportunities for further research, and investigations to either use or improve thistechnique continue at present. Briefly outlined below are the areas for continuedresearch.Implementation for QA purposes: Several points must be considered beforethis MC technique can be implemented for routine clinical use. The tech-nique would have to be implemented without significant disruption to theworkflow and workload distribution of the clinical medical physicists. Firstand foremost, the amount of manual time involved in using the techniquemust be minimized. This includes the time involved in producing a pair ofMC phantoms for a patient, converting MC simulation dose files to real dosefiles, importing to these files to MIM Maestro for registration, and contour-ing. Many of the steps involved in producing MC phantoms are already auto-matic, and with some additional modifications, this part of the pipeline maybe able to run without manual intervention for standard treatments. Most ofthe steps in converting a MC dose file to a real dose file are also automaticalready, and with the help of expert MIM Maestro users, these dose filesmay be able to be imported and deformed more efficiently. Contouring isa task that could be shared with other physicists, oncologists, or therapistsdepending on the concerns of the clinical team.Additionally, computational time and resources must be considered. Cur-83rently, producing two sets of simulations with two billion particle historiestakes around five hours using all 256 nodes of the clinical MC cluster, de-pending on the size of the patient. This is not only time consuming, butdraws resources away from clinical MC and could clash with other researchefforts in the clinic.An important adjustment to this method for clinical implementation is to de-crease the resolution of egsphant files to 0.5x0.5x0.5 cm3 resolution, ratherthan the current 0.25x0.25x0.25 cm3 resolution. The high resolution wasused because this has been the default resolution for determination of lungcompensator thickness (as described in section 3.3. By reducing the resolu-tion, we will retain an acceptable level of accuracy in the data, decrease thecomputational time by an order of magnitude, and reduce the sizes of theegsphant and 3ddose files.Finally, it may be possible to use even fewer histories to obtain clinicallyuseful data. The optimization study in section 4.3.3 describes only a singletreatment simulation and does not consider the fact that two simulations (onesupine, one prone) may serve to de-noise each other out without running twosimulations per position. If only one simulation is used per side, this cuts thecomputational cost of the technique by another 50%.Changing the CT imaging protocol for TBI to include more critical organs: Itwould be valuable to collect dose-volume information from several OARsthat were not included in this research because our existing TBI imagingprotocol does not allow for the retrieval of these data. The most interest-ing missing organ would be the eyes, as cataractogenesis is a well-knownand well-studied late toxicity of TBI, especially for pediatric patients. Thebrain and gonads would also be useful. Additionally, some team membershave expressed interest in the dose delivered to the bones. If all of thepelvic bones, which contain bone marrow, were included in the CT imagingprocess, then meaningful data could be obtained representing all marrow-containing bones, rather than only the bones that happen to be included inthe current CT protocol. At the time of writing this thesis, the protocol ischanging to include a full-length CT (as far as possible in a single scan) and84this supplementary dose information would become available to the clinicalteam in the near future using this MC technique.Additionally, the issue of missing scatter at superior-inferior edges of the pa-tient, as described in section 5.2, would be solved with complete CT imagevolumes. Currently, we have worked around this issue by manually remov-ing inferior and superior slices from our body contours, but this eliminatesour ability to study these regions, such as the neck where overdosing is ex-pected due to the smaller thickness. On a related note, we are currentlydiscussing how we may incorporate bolus into a MC simulation. This maybe too cumbersome to implement for routine QA, but if simulations withbolus were executed for one or two model patients, this would allow for anestimation of how far off our MC results are without accounting for bolus.Using a full-body CT will also mandate the previous suggestion of using alower resolution for the MC phantoms because DOSXYZnrc has an inherentsize limit of 512 slices in each dimension.Testing the robustness to uncertainty in lung compensator positioning: A briefstudy is underway to assess the robustness of our current TBI techniqueagainst lung compensator positioning errors. By incrementally shifting thelung compensators several millimetres in superior-inferior and left-right di-rections and examining the resulting isodoses and organ dose information,we can get an idea for how forgiving our treatment is to this source of hu-man error.Validation of the deformable registration: This task is nontrivial and beyondthe scope of this thesis. Expert users are able to qualitatively observe andguide the registration in MIM Maestro using the RegReveal and RegRefinetools, but meaningful metrics to describe the validity of this registration areunavailable, and would be very useful for this full-body prone-to-supine reg-istration.Exploring dose reduction within our current technique: We are currently in-vestigating the extent to which further lung dose reduction may be achievedin our current technique, without a shift to VMAT or HT. Our TBI tech-85nique was commissioned to obtain a mean body dose of 12 Gy throughoutthe whole body. Many investigators have become interested in reducing thelung dose, although many TBI techniques are based on point dose calcu-lations and organ dose information has not been obtainable. With our MCtechnique, we can explore the design of thicker compensators or differentshapes of contours, adding a very useful degree of freedom to our treatment.Developing and implementing a conformal VMAT technique: Finally, we re-state the long-term purpose of this research. While our TBI technique is ef-fective, we hope to follow suit with several centers in the world in upgradingto a more conformal technique using the extensive VMAT experience at ourcentre. Discussions are underway to modify an existing field junctioningtechnique that is in place for craniospinal radiotherapy treatments for appli-cation to TBI, as well as discussions to treat the patient in a single positionwithout needing to reposition the patient between beam arcs, similar to anapproach described by Ouyang et al [105]. The organ dose data provided inChapter 5 will serve as a basis for the development of this technique. In addi-tion to the physics challenges of implementing a more conformal technique,the logistical challenges of moving a relatively simple planning process to anextremely advanced planning process, by moving from a point dose calcula-tion to a full-body VMAT optimization, will need to be considered as well,in terms of physicist hours and linac availability in a clinic that is alreadyoperating at capacity.6.2 ConclusionA years-long effort at our centre to achieve MC simulations of TBI had been hin-dered by computational resources and technical challenges, but has finally beenachieved owing to the patience and dedication of several physicists over the pastdecade. This MC technique has been validated and has already been informingclinical decisions about both individual treatment cases and big-picture changesto our treatment workflow. It has been shown that smaller TBI patients receivesmaller doses than larger patients while still falling within tolerance levels from86prescription, and a basis for VMAT organ dose constraints is provided by the ret-rospective study. While the technique is currently optimized for retrospective datacollection where time constraints are not an issue, it can and will be optimized forroutine quality assurance purposes as well.87Bibliography[1] P Mazzarello. A unifying concept: the history of cell theory. Nature CellBiology, 1(1):E13–E15, 1999. → page 3[2] A Bongso and M Richards. History and perspective of stem cell research.Best Practice & Research Clinical Obstetrics and Gynaecology,18(6):827–842, 2004. → pages 3, 4[3] Anatomy and Physiology. OpenStax, Rice University, 2013.https://openstax.org/details/books/anatomy-and-physiology. → pages4, 5, 6, 7[4] M Ogawa, A C LaRue, and M Mehrotra. Hematopoietic stem cells arepluripotent and not just “hematopoietic”. Blood Cells, Molecules andDiseases, 51(1):3–8, 2013. → page 4[5] A V Hoffbrand and P A H Moss. Hoffbrand’s Essential Haematology.Wiley-Blackwell, 7th edition, 2016. → pages 4, 5, 6, 9, 17[6] S Standring. Gray’s Anatomy: The Anatomical Basis of Clinical Practice.Elsevier Limited, 41st edition, 2016. → pages 4, 7[7] B Clarke. Normal bone anatomy and physiology. Clinical Journal of theAmerican Society of Nephrology, 3(Suppl 3):S131–S139, 2008. → page 4[8] R S Taichman. Blood and bone: two tissues whose fates are intertwined tocreate the hematopoietic stem-cell niche. Blood, 105(7):2631–2639, 2005.→ page 5[9] D Hanahan and R A Weinberg. The hallmarks of cancer. Cell,(100):57–70, 2000. → pages 7, 8[10] S Pelengaris and M Khan. Molecular Biology of Cancer: A Bridge fromBench to Bedside. Wiley-Blackwell, 2nd edition, 2013. → page 8[11] S Mukherjee. The Emperor of All Maladies. Simon and Schuster, Inc.,2010. → pages 8, 12[12] D Hanahan and R A Weinberg. Hallmarks of cancer: the next generation.Cell, 144(5):646–674, 2011. → page 888[13] Canadian Cancer Society. Cancer types. http://www.cancer.ca/en/cancer-information/cancer-type/see-all/?region=bc. Accessed February 2018. →pages 9, 12[14] Dana-Farber Cancer Institute. Hematologic oncology cancer types andprograms. http://www.dana-farber.org/hematologic-oncology-treatment-center/cancer-types-and-programs/. Accessed February 2018. → pages9, 12, 13[15] A Chavez-Gonzalez, B Bakhshinejad, K Pakravan, M L Guzman, andS Babashah. Novel strategies for targeting leukemia stem cells: soundingthe death knell for blood cancer. Cellular Oncology, 40(1):1–20, 2017. →page 9[16] M Dowling, M Kelly, and T Meenaghan. Multiple myeloma: managing acomplex blood cancer. British Journal of Nursing, 25(16):S18–S28, 2016.→ page 10[17] M Mohty and J Harousseau. Handbook of Multiple Myeloma. SpringerInternational Publishing, 2015. → page 10[18] D A Arber, A Orazi, R Hasserjian, J Thiele, M J Borowitz, M M Le Beau,C D Bloomfield, M Cazzola, and J W Vardiman. The 2016 revision to theWorld Health Organization classification of myeloid neoplasms and acuteleukemia. Blood, 127(20):2391–2405, 2016. → page 10[19] S H Swerdlow, E Campo, S A Pileri, N L Harris, H Stein, R Siebert,R Advani, M Ghielmini, G A Salles, A D Zelenetz, and E S Jaffe. The2016 revision of the World Health Organization classification of lymphoidneoplasms. Blood, 127(20):2375–2390, 2016. → page 10[20] Canadian Cancer Society’s Advisory Committee on Cancer Statistics.Canadian cancer statistics 2017. Available at cancer.ca/Canadian-Cancer-Statistics-2017-EN.pdf. → page 10[21] M L Sorror, B M Sandmaier, B E Storer, G N Franke, G G Laport, T RChauncey, E Agura, R T Maziarz, A Langston, P Hari, et al. Long-termoutcomes among older patients following nonmyeloablative conditioningand allogeneic hematopoietic cell transplantation for advanced hematologicmalignancies. Journal of the American Medical Association,306(17):1874–1883, 2011. → page 1289[22] L L Popplewell and S J Forman. Is there an upper age limit for bonemarrow transplantation? Bone Marrow Transplantation, 29(4):277–284,2002. → page 12[23] H Zhang, J Chen, and W Que. Allogeneic peripheral blood stem cell andbone marrow transplantation for hematologic malignancies: meta-analysisof randomized controlled trials. Leukemia Research, 36(4):431–437, 2012.→ page 13[24] H E Johns and J R Cunningham. The Physics of Radiology. Charles CThomas, 4th edition, 1983. → pages 13, 18[25] J T Bushberg, J A Seibert, E M Leidholdt, and J M Boone. The EssentialPhysics of Medical Imaging. Lippincott Williams and Wilkins, 4 edition,2012.[26] F H Attix. Introduction to Radiological Physics and Radiation Dosimetry.John Wiley and Sons, 1st edition, 2004. → pages 13, 20[27] E J Hall and A J Giaccia. Radiobiology for the Radiologist. LippincottWilliams and Wilkins, 7th edition, 2012. → page 17[28] S Braunstein and J L Nakamura. Radiotherapy-induced malignancies:review of clinical features, pathobiology, and evolving approaches formitigating risk. Frontiers in Oncology, 3:73, 2013. → page 17[29] E Hutton. The atom bomb that saves lives. Maclean’s Magazine, 1952.February 15 issue. → pages 18, 19[30] G M Mora, A Maio, and D W O Rogers. Monte Carlo simulation of atypical 60-Cobalt therapy source. Medical Physics, 26(11):2494–2502,1999. → pages 18, 58[31] H E Johns, L M Bates, E R Epp, D V Cormack, S O Fedoruk, A Morrison,W R Dixon, and C Garrett. 1,000-Curie Cobalt-60 units for radiationtherapy. Nature, 168:1035–1036, 1951. Letter to the editor. → page 19[32] The Official Website of the Nobel Prize. Nobel prizes and laureates.https://www.nobelprize.org/nobel prizes/. Accessed February 2018. →pages 21, 22[33] J A del Regato. Radiological Oncologists: The Unfolding of a MedicalSpecialty. Radiology Centennial, Inc., 1993. → page 2190[34] A Barrett. Total body irradiation. Reports of Practical Oncology andRadiotherapy, 4(3):47–64, 2000. → pages 21, 22[35] C A Barker, T J LoSasso, and S L Wolden. Total body irradiation.https://oncohemakey.com/total-body-irradiation-2/. Clinical search enginefor hematology and oncology. Accessed February 2018. → page 21[36] A C Heublein. A preliminary report on continuous irradiation of the entirebody. Radiology, 18(6):1051–1062, 1932. → page 21[37] Advisory committee on human radiation experiments : final report, 1995.Produced by an advisory committee struck by US President Clinton.Official US Government document, publically available. → page 22[38] J P Merrill, J E Murray, J H Harrison, A Friedman, J B Dealy, and G JDammin. Successful homotransplantation of the kidney betweennonidentical twins. The New England Journal of Medicine,262(25):1251–1260, 1960. → page 22[39] E D Thomas. A history of haemopoietic cell transplantation. BritishJournal of Haematology, 105(2):330–339, 1999. → page 22[40] C E Hill-Kayser, J P Plastaras, Z Tochner, and E Glatstein. TBI during BMand SCT: review of the past, discussion of the present and consideration offuture directions. Bone Marrow Transplantation, 46(4):475–484, 2011. →pages 22, 23[41] H J Deeg, R Storb, G Longton, T C Graham, H M Shulman, F Appelbaum,and E D Thomas. Single dose or fractionated total body irradiation andautologous marrow transplantation in dogs: effects of exposure rate,fraction size, and fractionation interval on acute and delayed toxicity.International Journal of Radiation Oncology Biology Physics,15(3):647–653, 1988. → page 23[42] R Storb, R Raff, H J Deeg, T Graham, F R Appelbaum, F G Schuening,H Shulman, K Seidel, and W Leisenring. Dose rate-dependent sparing ofthe gastrointestinal tract by fractionated total body irradiation in dogs givenmarrow autografts. International Journal of Radiation Oncology BiologyPhysics, 40(4):961–966, 1998. → page 23[43] J A O’Donoghue. Fractionated versus low dose-rate total body irradiation.radiobiolgical considerations in the selection of regimes. Radiotherapy andOncology, 7(3):241–247, 1986. → page 2391[44] L Peters. Discussion: The radiobiology bases of TBI. International Journalof Radiation Oncology Biology Physics, 6(6):785–787, 1980. → page 23[45] R A Clift, D Buckner, F R Appelbaum, S I Bearman, F B Petersen, L DFisher, C Anasetti, P Beatty, W I Bensinger, K Doney, et al. Allogeneicmarrow transplantation in patients with acute myeloid leukemia in firstremission: a randomized trial of two irradiation regimens. Blood,76(9):1867–1871, 1990. → page 23[46] M M Bortin, H E M Kay, R P Gale, and A A Rimm. Factors associatedwith interstitial pneumonitis after bone-marrow transplantation for acuteleukaemia. The Lancet, 319(8269):437–439, 1982. → page 23[47] S Sampath, T E Schultheiss, and J Wong. Dose response and factors relatedto interstitial pneumonitis after bone marrow transplant. InternationalJournal of Radiation Oncology Biology Physics, 63(3):876–884, 2005. →pages 23, 30[48] S A Carruthers and M M Wallington. Total body irradiation andpneumonitis risk: a review of outcomes. British Journal of Cancer,90(11):2080–2084, 2004. → pages 23, 30[49] American College of Radiology (ACR) and American Society forRadiation Oncology (ASTRO). ACR-ASTRO practice parameter for theperformance of total body irradiation. 2017. Clinical practice statement,revised regularly, publically available. → page 23[50] A Bacigalupo, K Ballen, D Rizzo, S Giralt, H Lazarus, V Ho, J Apperley,S Slavin, M Pasquini, B M Sandmaier, et al. Defining the intensity ofconditioning regimens: working definitions. Biology of Blood and MarrowTransplantation, 15(12):1628–1633, 2009. → page 24[51] D R Adkins and J F DiPersio. Total body irradiation before an allogeneicstem cell transplantation: is there a magic dose? Current Opinion inHematology, 15(6):555–560, 2008. → page 24[52] P H Wiernik, J M Goldman, J P Dutcher, and R A Kyle. Neoplasticdiseases of the blood. Cambridge University Press, 5th edition, 2013. →page 24[53] J L Mikell, E K Waller, J M Switchenko, S Rangaraju, Z Ali, M Graiser,W A Hall, A A Langston, N Esiashvili, H J Khoury, et al. Similar survivalfor patients undergoing reduced-intensity total body irradiation (TBI)92versus myeloablative TBI as conditioning for allogeneic transplant in acuteleukemia. International Journal of Radiation Oncology Biology Physics,89(2):360–369, 2014. → page 24[54] S L McAfee, S N Powell, C Colby, and T R Spitzer. Dose-escalated totalbody irradiation and autologous stem cell transplantation for refractoryhematologic malignancy. International Journal of Radiation OncologyBiology Physics, 53(1):151–156, 2002. → page 24[55] S Altouri, M Sabloff, D Allan, H Atkins, L Huebsch, D Maze, R Samant,and C Bredeson. Total body irradiation without chemotherapy asconditioning for an allogeneic hematopoietic cell transplantation for adultacute myeloid leukemia. Case Reports in Hematology, 2016. Article ID1257679. → page 24[56] R C N Studinski, D J Fraser, R S Samant, and M S MacPherson. Currentpractice in total-body irradiation: results of a canada-wide survey. CurrentOncology, 24(3):181–186, 2017. → pages 24, 25[57] S Giebel, L Miszczyk, K Slosarek, L Moukhtari, F Ciceri, J Esteve,N Gorin, M Labopin, A Nagler, C Schmid, and M Mohty. Extremeheterogeneity of myeloablative total body irradiation techniques in clinicalpractice. Cancer, 120(17):2760–2765, 2014. → pages 24, 25[58] S Peca. Informal survey conducted via email; results not published, sharedvia private communication. → page 25[59] S Hussein and E El-Khatib. Total body irradiation with a sweeping60-Cobalt beam. International Journal of Radiation Oncology BiologyPhysics, 33(2):493–497, 1995. → pages 26, 43, 72[60] M D C Evans, R Larouche, M Olivares, P Le´ger, J Larkin, C R Freeman,and E B Podgorsak. Total body irradiation with a reconditioned cobaltteletherapy unit. Journal of Applied Clinical Medical Physics, 7(1):42–51,2006. → page 26[61] R Ravichandran, J P Binukumar, C A Davis, S S Sivakumar,K Krishnamurty, Z Al Mandhari, and B Rajan. Beam configuration andphysical parameters of clinical high energy photon beam for total bodyirradiation (TBI). Physica Medica, 27(3):163–168, 2011. → pages 26, 28[62] E J Bloemen-van Gurp, B J Mijnheer, T A M Verschueren, and P Lambin.Total body irradiation, toward optimal individual delivery: evaluation with93metal oxide field effect transistors, thermoluminescence detectors, and atreatment planning system. International Journal of Radiation OncologyBiology Physics, 69(4):1297–1304, 2007. → pages 28, 40[63] R Yao, D Bernard, J Turian, R A Abrams, and W Sensakovic. A simplifiedtechnique for delivering total body irradiation (TBI) with improved dosehomogeneity. Medical Physics, 39(4):2239–2248, 2012. → pages 26, 27[64] R Chakarova and M Krantz. A Monte Carlo evaluation of beamcharacteristics for total body irradiation at extended treatment distances.Journal of Applied Clinical Medical Physics, 15(3):182–189, 2014. →pages 27, 28, 42[65] S V Harden, D S Routsis, A R Geater, S J Thomas, P J Taylor, R E Marcus,and M V Williams. Total body irradiation using a modified standingtechnique: a single institution 7 year experience. The British Journal ofRadiology, 74(887):1041–1047, 2001. → page 27[66] B Schaeken, S Lelie, P Meijnders, D Van den Weyngaert, H Janssens, andD Verellen. Alanine/EPR dosimetry applied to the verification of a totalbody irradiation protocol and treatment planning dose calculation using ahumanoid phantom. Medical Physics, 37(12):6292–6299, 2010. → pages27, 40[67] L Papiez˙, J Montebello, C DesRosiers, and E Papiez˙. The clinicalapplication of dynamic shielding and imaging in moving table total bodyirradiation. Radiotherapy and Oncology, 51(3):219–224, 1999. → page 27[68] M Chre´tien, C Coˆte´, R Blais, L Brouard, L Roy-Lacroix, and M Larochelle.A variable speed translating couch technique for total body irradiation.Medical Physics, 27(5):1127–1130, 2000. → pages 27, 28[69] A Hussain, J E Villarreal-Barajas, P Dunscombe, and D W Brown.Aperture modulated, translating bed total body irradiation. MedicalPhysics, 38(2):932–941, 2011. → page 27[70] S Ahmed, D Brown, S B S Ahmed, M B Kakakhel, W Muhammad, andA Hussain. Translating bed total body irradiation lung shielding and doseoptimization using asymmetric MLC apertures. Journal of Applied ClinicalMedical Physics, 17(2):112–122, 2016. → page 27[71] A Jahnke, L Jahnke, F Molina-Duran, M Ehmann, S Kantz, V Steil,F Wenz, G Glatting, F Lohr, and M Polednik. Arc therapy for total body94irradiation - a robust novel treatment technique for standard treatmentrooms. Radiotherapy and Oncology, 110(3):553–557, 2014. → pages27, 28[72] C Onal, A Sonmez, G Arslan, S Sonmez, E Efe, and E Oymak. Evaluationof field-in-field technique for total body irradiation. International Journalof Radiation Oncology Biology Physics, 83(5):1641–1648, 2012. → pages28, 41[73] N Kirby, M Held, O Morin, S Fogh, and J Pouliot. Inverse-plannedmodulated-arc total-body irradiation. Medical Physics, 39(5):2761–2764,2012. → page 28[74] S Park, J Kim, Y H Joo, J C Lee, and J M Park. Total body irradiation witha compensator fabricated using a 3D optical scanner and a 3D printer.Physics in Medicine and Biology, 62(9):3735–3756, 2017. → page 28[75] M Zhang, N Qin, X Jia, W J Zou, A Khan, and N J Tue. Investigation onusing high-energy proton beam for total body irradiation (TBI). Journal ofApplied Clinical Medical Physics, 17(5):90–98, 2016. → pages 29, 42[76] A Buchali, P Feyer, J Groll, G Massenkeil, R Arnold, and V Budach.Immediate toxicity during fractionated total body irradiation asconditioning for bone marrow transplantation. Radiotherapy andOncology, 54(2):157–162, 2000. → page 30[77] A Della Volpe, A J M Ferreri, C Annaloro, P Mangili, A Rosso,R Calandrino, E Villa, G Lambertenghi-Deliliers, and C Fiorino. Lethalpulmonary complications significantly correlate with individually assessedmean lung dose in patients with hematologic malignancies treated withtotal body irradiation. International Journal of Radiation OncologyBiology Physics, 52(2):483–488, 2002. → page 30[78] C R Kelsey, M E Horwitz, J P Chino, O Craciunescu, B Steffey, R J Folz,N J Chao, D A Rizzieri, and L B Marks. Severe pulmonary toxicity aftermyeloablative conditioning using total body irradiation: an assessment ofrisk factors. International Journal of Radiation Oncology Biology Physics,81(3):812–818, 2011. → page 30[79] M Abugideiri, R H Nanda, C Butker, C Zhang, S Kim, K Chiang, E Butker,M K Khan, A E Haight, Z Chen, and N Esiashvili. Factors influencingpulmonary toxicity in children undergoing allogeneic hematopoietic stemcell transplantation in the setting of total body irradiation-based95myeloablative conditioning. International Journal of Radiation OncologyBiology Physics, 94(2):349–359, 2016. → page 30[80] R A Schneider, J Schultze, J M Jensen, D Hebbinghaus, and R M Galalae.Long-term outcome after static intensity-modulated total body radiotherapyusing compensators stratified by pediatric and adult cohorts. InternationalJournal of Radiation Oncology Biology Physics, 70(1):194–202, 2008. →page 30[81] O I Craciunescu, B A Steffey, C R Kelsey, N A Larrier, C J Paarz-Largay,R G Prosnitz, N Chao, J Chute, C Gasparetto, M Horwitz, et al. Renalshielding and dosimetry for patients with severe systemic sclerosisreceiving immunoablation with total body irradiation in the scleroderma:cyclophosphamide or transplantation trial. International Journal ofRadiation Oncology Biology Physics, 79(4):1248–1255, 2011. → page 31[82] J Gerstein, A Meyer, K Sykora, J Fru¨hauf, J H Karstens, and M Bremer.Long-term renal toxicity in children following fractionated total-bodyirradiation (TBI) before allogeneic stem cell transplantation (SCT).Strahlentherapie und Onkologie, 185(11):751–755, 2009. → page 31[83] J C Cheng, T E Schultheiss, and J Y C Wong. Impact of drug therapy,radiation dose, and dose rate on renal toxicity following bone marrowtransplantation. International Journal of Radiation Oncology BiologyPhysics, 71(5):1436–1443, 2008. → page 31[84] H B Kal and M L van Kempen-Harteveld. Induction of severe cataract andlate renal dysfunction following total body irradiation: dose-effectrelationships. Anticancer Research, 29(8):3305–3309, 2009. → page 31[85] Y Belkace´mi, M Ozsahin, F Pe`ne, B Rio, J Laporte, V Leblond, E Touboul,M Schlienger, N Gorin, and A Laugier. Cataractogenesis after total bodyirradiation. International Journal of Radiation Oncology Biology Physics,35(1):53–60, 1996. → page 31[86] Y Belkace´mi, M Labopin, J Vernant, H G Prentice, A Tichelli,A Schattenberg, M A Boogaerts, P Ernst, A Della Volpe, A H Goldstone,et al. Cataracts after total body irradiation and bone marrow transplantationin patients with acute leukemia in complete remission: a study of theEuropean Group for Blood and Marrow Transplantation. InternationalJournal of Radiation Oncology Biology Physics, 41(3):659–668, 1998. →page 3196[87] M L van Kempen-Harteveld, R Brand, H B Kal, L F Verdonck, P Hofman,A V Schattenberg, R W van der Maazen, J J Cornelissen, W M HEijkenboom, J P van der Lelie, et al. Results of hematopoietic stem celltransplantation after treatment with different high-dose total-bodyirradiation regimens in five dutch centers. International Journal ofRadiation Oncology Biology Physics, 71(5):1444–1454, 2008. → page 31[88] L M Vrooman, H R Millard, R Brazauskas, N S Majhail, M Battiwalla,M E Flowers, B N Savani, G Akpek, M Aljurf, R Bajwa, et al. Survival andlate effects after allogeneic hematopoietic cell transplantation forhematologic malignancy at less than three years of age. Biology of Bloodand Marrow Transplantation, 23(8):1327–1334, 2017. → page 32[89] M Faraci, S Barra, A Cohen, E Lanino, F Grisolia, M Miano, F Foppiano,O Sacco, M Cabria, R De Marco, et al. Very late nonfatal consequences offractionated TBI in children undergoing bone marrow transplant.International Journal of Radiation Oncology Biology Physics,63(5):1568–1575, 2005. → page 32[90] I Flandin, O Hartmann, J Michon, R Pinkerton, C Coze, J L Stephan,B Fourquet, D Valteau-Couanet, C Bergeron, T Philip, and C Carrie.Impact of TBI on late effects in children treated by megatherapy for stageIV neuroblastoma. a study of the French Society of Pediatric Oncology.International Journal of Radiation Oncology Biology Physics,64(5):1424–1431, 2006. → page 32[91] D L Friedman, A Rovo, W Leisenring, A Locasciulli, M E D Flowers,A Tichelli, J E Sanders, H J Deeg, and G Socie. Increased risk of breastcancer among survivors of allogeneic hematopoietic cell transplantation: areport from the FHCRC and the EBMT-Late Effect Working Party. Blood,111(2):939–944, 2008. → page 32[92] Mackie T R. History of tomotherapy. Physics in Medicine and Biology,51(13):R427–R453, 2006. → page 33[93] K Otto. Volumetric modulated arc therapy: IMRT in a single gantry arc.Medical Physics, 35(1):310–317, 2008. → page 33[94] S K Hui, J Kapatoes, J Fowler, D Henderson, G Olivera, R R Manon,B Gerbi, T R Mackie, and J S Welsh. Feasibility study of helicaltomotherapy for total body or total marrow irradiation. Medical Physics,32(10):3214–3224, 2005. → pages 33, 3897[95] A H Zhuang, A Liu, T E Schultheiss, and J Y C Wong. Dosimetric studyand verification of total body irradiation using helical tomotherapy and itscomparison to extended SSD technique. Medical Dosimetry,35(4):243–249, 2010. → pages 33, 34[96] M Chao, J Pen˜agarı´cano, Y Yan, E G Moros, P Corry, andV Ratanatharathorn. Voxel-based dose reconstruction for total bodyirradiation with helical tomotherapy. International Journal of RadiationOncology Biology Physics, 82(5):1575–1583, 2012. → page 34[97] J A Pen˜agarı´cano, M Chao, F Van Rhee, E G Moros, P M Corry, andV Ratanatharathorn. Clinical feasibility of TBI with helical tomotherapy.Bone Marrow Transplantation, 46(7):929–935, 2011. → pages 34, 35[98] T Magome, A Haga, Y Takahashi, K Nakagawa, K E Dusenbery, and S KHui. Fast megavoltage computed tomography: a rapid imaging method fortotal body or marrow irradiation in helical tomotherapy. InternationalJournal of Radiation Oncology Biology Physics, 96(3):688–695, 2016. →page 34[99] R Takenaka, A Haga, H Yamashita, and K Nakagawa. Adequate targetvolume in total-body irradiation by intensity-modulated radiation therapyusing helical tomotherapy: a simulation study. Journal of RadiationResearch, 58(2):210–216, 2017. → page 34[100] S Chakraborty, S Cheruliyil, R Bharathan, and G Muttath. Total bodyirradiation using VMAT (RapidArc): A planning study of a novel treatmentdelivery method. International Journal of Cancer Therapy and Oncology,3(2), 2015. → pages 34, 35[101] A Gruen, W Ebell, W Wlodarczyk, O Neumann, J S Kuehl, C Stromberger,V Budach, and S Marnitz. Total body irradiation (TBI) using helicaltomotherapy in children and young adults undergoing stem celltransplantation. Radiation Oncology, 8(92), 2013. → page 34[102] A Springer, J Hammer, E Winkler, C Track, R Huppert, A Bo¨hm,H Kasparu, A Weltermann, G Aschauer, A L Petzer, et al. Total bodyirradiation with volumetric modulated arc therapy: Dosimetric data and firstclinical experience. Radiation Oncology, 11(46), 2016. → pages 34, 36[103] R Takenaka, H Yamashita, T Toya, A Haga, S Shibata, M Kurokawa,K Ootomo, and K Nakagawa. Unique radiation dermatitis related to total98body irradiation by helical tomotherapy. The Journal of Dermatology,43(11):1376–1377, 2016. → page 35[104] J A Molloy. Statistical analysis of dose heterogeneity in circulating blood:implications for sequential methods of total body irradiation. MedicalPhysics, 37(11):5568–5578, 2010. → page 36[105] L Ouyang, M Folkerts, Y Zhang, B Hrycushko, R Lamphier, P Lee,E Chambers, E Ramirez, R Reynolds, Y Yan, et al. Volumetric modulatedarc therapy based total body irradiation: Workflow and clinical experiencewith an indexed rotational immobilization system. Physics and Imaging inRadiation Oncology, 4:22–25, 2017. → pages 36, 43, 86[106] C T Morrison, K L Symons, S J Woodings, and M J House. Verification ofjunction dose between VMAT arcs of total body irradiation using a SunNuclear ArcCHECK phantom. Journal of Applied Clinical MedicalPhysics, 18(6):177–182, 2017. → page 37[107] J H Kim, A Stein, N Tsai, T E Schultheiss, J Palmer, A Liu, J Rosenthal,S J Forman, and J Y C Wong. Extramedullary relapse following totalmarrow and lymphoid irradiation in patients undergoing allogeneichematopoietic cell transplantation. International Journal of RadiationOncology Biology Physics, 89(1):75–81, 2014. → pages 37, 38[108] S K Hui, M R Verneris, P Higgins, B Gerbi, B Weigel, SK Baker, C Fraser,M Tomblyn, and K Dusenbery. Helical tomotherapy targeting total bonemarrow–first clinical experience at the University of Minnesota. ActaOncologica, 46(2), 2007. → pages 37, 38[109] T E Schultheiss, J Wong, A Liu, G Olivera, and G Somlo. Image-guidedtotal marrow and total lymphatic irradiation using helical tomotherapy.International Journal of Radiation Oncology Biology Physics,67(4):1259–1267, 2007. → page 37[110] J Y C Wong, J Rosenthal, A Liu, T Schultheiss, S Forman, and G Somlo.Image-guided total-marrow irradiation using helical tomotherapy inpatients with multiple myeloma and acute leukemia undergoinghematopoietic cell transplantation. International Journal of RadiationOncology Biology Physics, 73(1):273–279, 2009. → page 37[111] C Han, T E Schultheiss, and J Y C Wong. Dosimetric study of volumetricmodulated arc therapy fields for total marrow irradiation. Radiotherapyand Oncology, 102(2):315–320, 2012. → page 3799[112] A Stein, J Palmer, N Tsai, M M Al Malki, I Aldoss, H Ali, A Aribi,L Farol, C Karanes, S Khaled, et al. Phase I trial of total marrow andlymphoid irradiation transplantation conditioning in patients withrelapsed/refractory acute leukemia. Biology of Blood and Marrowtransplantation, 23(4):618–624, 2017. → page 38[113] S Hui, C Brunstein, Y Takahashi, T DeFor, S G Holtan, V Bachanova,C Wilke, D Zuro, C Ustun, D Weisdorf, et al. Dose escalation of totalmarrow irradiation in high-risk patients undergoing allogeneichematopoietic stem cell transplantation. Biology of Blood and MarrowTransplantation, 23(7):1110–1116, 2017. → page 38[114] Y Takahashi, M R Verneris, K E Dusenbery, C T Wilke, G Storme, D JWeisdorf, and S K Hui. Peripheral dose heterogeneity due to the threadeffect in total marrow irradiation with helical tomotherapy. InternationalJournal of Radiation Oncology Biology Physics, 87(4):832–839, 2013. →page 38[115] M Zeverino, S Agostinelli, G Taccini, F Cavagnetto, S Garelli, M Gusinu,S Vagge, S Barra, and R Corvo`. Advances in the implementation of helicaltomotherapy-based total marrow irradiation with a novel field junctiontechnique. Medical Dosimetry, 3(37):314–320, 2012. → page 38[116] A Nalichowski, D G Eagle, and J Burmeister. Dosimetric evaluation oftotal marrow irradiation using 2 different planning systems. MedicalDosimetry, 41(3):230–235, 2016. → page 39[117] Y Takahashi, S Vagge, S Agostinelli, E Han, L Matulewicz, K Schubert,R Chityala, V Ratanatharathorn, K Tournel, J A Pen˜agarı`cano, et al.Multi-institutional feasibility study of a fast patient localization method intotal marrow irradiation with helical tomotherapy: A global health initiativeby the international consortium of total marrow irradiation. InternationalJournal of Radiation Oncology Biology Physics, 91(1):30–38, 2015. →page 39[118] M Lavalle´e, L Gingras, M Chre´tien, S Aubin, C Coˆte´, and L Beaulieu.Commissioning and evaluation of an extended SSD photon model forPINNACLE3: an application to total body irradiation. Medical Physics,36(8):3844–3855, 2009. → page 39[119] Y Akino, K P McMullen, and I J Das. Patterns of patient specific dosimetryin total body irradiation. Medical Physics, 40(4), 2013. → page 40100[120] R P Patel, A J Warry, D J Eaton, C H Collis, and I Rosenberg. In vivodosimetry for total body irradiation: five-year results and techniquecomparison. Journal of Applied Clinical Medical Physics, 15(4):306–315,2014. → page 40[121] C M Lancaster, J C Crosbie, and S R Davis. In-vivo dosimetry from totalbody irradiation patients (2000-2006): results and analysis. AustralasianPhysics & Engineering Sciences in Medicine, 31(3):191–195, 2008. →page 40[122] N Lamichhane, V N Patel, and M T Studenski. Going the distance:validation of Acuros and AAA at an extended SSD of 400 cm. Journal ofApplied Clinical Medical Physics, 17(2):63–73, 2016. → page 41[123] M Butson, M Haque, L Smith, E Butson, D Odgers, D Pope, T Gorjiana,M Whitaker, J Morales, A Hong, et al. Practical time considerations foroptically stimulated luminescent dosimetry (OSLD) in total bodyirradiation. Australasian Physical & Engineering Sciences in Medicine,40(1):167–171, 2017. → page 41[124] Z Liu, D Lack, J T Rakowski, C Knill, and M Snyder. Fast Monte Carlosimulation for total body irradiation using a 60-Co teletherapy unit. Journalof Applied Clinical Medical Physics, 14(3):133–149, 2013. → page 42[125] M Serban, J Seuntjens, E Roussin, A Alexander, J-R Tremblay, andW Wierzbicki. Patient-specific compensation for Co-60 TBI treatmentsbased on Monte Carlo design: A feasibility study. Physica Medica,32(1):67–75, 2016. → page 42[126] R Chakarova, K Mu¨ntzing, M Krantz, E Hedin, and S Hertzman. MonteCarlo optimization of total body irradiation in a phantom and patientgeometry. Physics in Medicine and Biology, 58(8):2461–2469, 2013. →page 42[127] A Ito. Three-dimensional dose calculation for total body irradiation. InMonte Carlo Transport of Electrons and Photons, chapter 27. PlenumPress, New York, 1988. → page 43[128] J Lucido, C Yuen, M Pearson, and I Popescu. New CT-based planningmethod for TBI treatment. 2011. Abstract presented at the 2011 meeting ofthe American Association of Medical Physicists and CanadianOrganization of Medical Physicists (AAPM-COMP). → pages 43, 44, 73101[129] N Metropolis. The beginning of the Monte Carlo method. Los AlamosScience, 15:125–130, 1987. → pages 49, 50[130] R Eckhardt. Stan Ulam, John von Neumann, and the Monte Carlo method.Los Alamos Science, 15:131–137, 1987. → pages 49, 50[131] N Metropolis and S Ulam. The Monte Carlo Method. Journal of theAmerican Statistical Association, 44(247):335–341, 1949. → page 50[132] W K Chan. Theory and Applications of Monte Carlo Simulations. InTech,2013. → page 50[133] I Kawrakow. Accurate condensed history Monte Carlo simulation ofelectron transport. I. EGSnrc, the new EGS4 version. Medical Physics,27(3):485–498, 2000. → pages 52, 53[134] I J Chetty, B Curran, J E Cygler, J J DeMarco, G Ezzell, B A Faddegon,I Kawrakow, P J Keall, H Liu, and C M C others Ma. Report of the AAPMTask Group No. 105: Issues associated with clinical implementation ofMonte Carlo-based photon and electron external beam treatment planning.Medical Physics, 34(12):4818–4853, 2007. → page 52[135] G D Doolen and J Hendricks. Monte Carlo at work. Los Alamos Science,15:142–143, 1987. → page 53[136] I Kawrakow, D W O Rogers, and B R B Walters. Large efficiencyimprovements in BEAMnrc using directional bremsstrahlung splitting.Medical Physics, 31(10):2883–2898, 2004. → pages 53, 59[137] A F Bielajew. Fundamentals of the Monte Carlo method for neutral andcharged particle transport. 2016. Unpublished book, arising from coursenotes from a course the author taught at The University of Michigan, AnnArbor, Michigan, USA.[138] I Kawrakow and M Fippel. Investigation of variance reduction techniquesfor Monte Carlo photon dose calculation using XVMC. Physics inMedicine and Biology, 45(8):2163–2183, 2000. → page 53[139] T Teke. Monte carlo techniques for patient specific verifications of complexradiation therapy treatments including TBI, VMAT and SBRT lung. 2012.Ph.D thesis at the University of British Columbia. → pages 58, 74[140] J Lobo and I A Popescu. Two new DOSXYZnrc sources for 4D MonteCarlo simulations of continuously variable beam configurations, with102applications to RapidArc, VMAT, TomoTherapy and CyberKnife. Physicsin Medicine and Biology, 55(16):4431–4443, 2010. → page 58[141] A Savitzky and M J E Golay. Smoothing and differentiation of data bysimplified least squared procedures. Analytical Chemistry, 36(8):1627–39,1964. → page 59[142] I Kawrakow. On the de-noising of Monte Carlo calculated dosedistributions. Physics in Medicine and Biology, 47(17):3087–3103, 2002.→ pages 59, 60[143] I A Popescu, C P Shaw, S F Zavgorodni, and W A Beckham. Absolutedose calculations for Monte Carlo simulations of radiotherapy beams.Physics in Medicine and Biology, 50(14):3375–3392, 2005. → page 63[144] P B Johnson, K R Padgett, K L Chen, and N Dogan. Evaluation of the tool“Reg Refine” for user-guided deformable image registration. Journal ofApplied Clinical Medical Physics, 17(3):158–170, 2016. → page 69[145] K Nie, J Pouliot, E Smith, and C Chuang. Performance variations amongclinically available deformable image registration tools in adaptiveradiotherapy – how should we evaluate and interpret the result? Journal ofApplied Clinical Medical Physics, 17(2):328–340, 2016. → page 69[146] T Kataria, K Sharma, V Subramani, K P Karrthick, and S S Bisht.Homogeneity index: An objective tool for assessment of conformalradiation treatments. Journal of Medical Physics, 37(4):207, 2012. → page76[147] A L McKenzie. Cobalt-60 gamma-ray beams. British Journal of RadiologySupplement, 25, 1996. → page 79103

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0365701/manifest

Comment

Related Items