UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

18F-FDOPA positron emission tomography and diffusion tensor imaging for radiation therapy of high-grade… Kosztyla, Robert 2014

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

Item Metadata

Download

Media
24-ubc_2014_spring_kosztyla_robert.pdf [ 5.75MB ]
Metadata
JSON: 24-1.0166880.json
JSON-LD: 24-1.0166880-ld.json
RDF/XML (Pretty): 24-1.0166880-rdf.xml
RDF/JSON: 24-1.0166880-rdf.json
Turtle: 24-1.0166880-turtle.txt
N-Triples: 24-1.0166880-rdf-ntriples.txt
Original Record: 24-1.0166880-source.json
Full Text
24-1.0166880-fulltext.txt
Citation
24-1.0166880.ris

Full Text

18F-FDOPA Positron Emission Tomography andDiffusion Tensor Imaging for Radiation Therapy ofHigh-Grade Gliomas with Dose PaintingbyRobert KosztylaB.Sc., The University of Waterloo, 2007B.Ed., Queen?s University, 2007M.Sc., The University of British Columbia, 2009A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinThe Faculty of Graduate and Postdoctoral Studies(Physics)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)March 2014? Robert Kosztyla 2014AbstractIn the radiation therapy of high-grade gliomas, T1-weighted magnetic reso-nance imaging (MRI) with contrast enhancement does not accurately repre-sent the extent of the tumour. Functional imaging techniques, such as positronemission tomography (PET) and diffusion tensor imaging (DTI), can potentiallybe used to improve tumour localization and for biologically-based treatmentplanning. This project investigated tumour localization using 3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine (18F-FDOPA) PET and interhemispheric differenceimages obtained from DTI, and determined whether radiation therapy of high-grade gliomas using dose painting was feasible with volumetric modulated arctherapy (VMAT). First, radiation therapy target volumes obtained from five ob-servers using 18F-FDOPA PET and MRI were compared with the location of re-currences following radiotherapy. It was demonstrated with simultaneous truthand performance level estimation that high-grade glioma radiation therapy tar-get volumes obtained with PET had similar interobserver agreement to MRI-based volumes. Although PET target volumes were significantly larger thanvolumes obtained using MRI, treatment planning using the PET-based volumesmay not have yielded better treatment outcomes since all but one central recur-rence extended beyond the PET abnormality. The second study characterizedthe distribution of fractional anisotropy (FA) and mean diffusivity (MD) valuesobtained from DTI, as well as FA and MD interhemispheric differences. It wasiidemonstrated that FA, MD, and interhemispheric differences approached thoseof contralateral normal brain as the distance from the tumour increased, consis-tent with the expectation of a gradual and decreasing presence of tumour cells.Lastly, a treatment planning study compared VMAT for high-grade gliomas ob-tained from dose painting using 18F-FDOPA PET images. Dose constraints foreach contour were specified by a radiobiological model. VMAT planning usingdose painting for high-grade gliomas was achieved using biologically-guidedthresholds of 18F-FDOPA uptake with no significant change in the dose deliv-ered to critical structures.iiiPrefaceA version of Chapter 2 has been published (Kosztyla R, Chan EK, Hsu F, Wil-son D, Ma R, Cheung A, Zhang S, Moiseenko V, Benard F, Nichol A. High-gradeglioma radiation therapy target volumes and patterns of failure obtained frommagnetic resonance imaging and 18F-FDOPA positron emission tomography de-lineations from multiple observers. Int J Radiat Oncol Biol Phys. 2013;87(5):1100?6. doi:10.1016/j.ijrobp.2013.09.008.). The principal investigator for thisstudy was A. Nichol. Radiation therapy target volumes were outlined by F. Hsu,D. Wilson, R. Ma, A. Cheung, and A. Nichol. Image fusion in treatment plan-ning software was done by S. Zhang. Recurrence image sets were identified byE. K. Chan. I conducted all data analysis, including analysis with simultaneoustruth and performance level estimation, and wrote the manuscript. All authorsassisted in editing the manuscript.A version of Chapter 3 has been submitted for publication (Kosztyla R, Reins-berg SA, Moiseenko V, Toyota B, Nichol A. Interhemispheric difference imagesfrom postoperative diffusion tensor imaging of gliomas.) The diffusion tensorfitting method was selected by S. Reinsberg and myself. I devised the methodof obtaining interhemispheric difference images from DTI parameters and com-pleted all image registration, diffusion tensor fitting, image analysis, and statis-tical interpretation. Clinical guidance was provided by A. Nichol.ivIn Chapter 4, the treatment planning technique for dose painting was de-vised by myself, V. Moiseenko, and S. Reinsberg, with clinical guidance fromA. Nichol. S. Zhang imported images into treatment planning software. Vol-umes of interest for dose escalation were identified by A. Nichol. I completedfusion of treatment planning CT and PET/CT images, obtained contours fromPET images for dose escalation, and completed all treatment planning and dataanalysis.The work reported in this thesis was completed with approval from the Uni-versity of British Columbia BC Cancer Agency research ethics board, certificatenumbers H08-02314 for the work in Chapter 2 and H10-02888 for the work inChapters 3 and 4.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiList of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xx1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 High-Grade Gliomas . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 External Beam Radiation Therapy . . . . . . . . . . . . . . . . . . . 21.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2.2 X-ray Production and Photon Interactions with Matter . . 31.2.3 Electron Interactions with Matter . . . . . . . . . . . . . . . 61.2.4 Absorbed Dose . . . . . . . . . . . . . . . . . . . . . . . . . . 101.2.5 Linear Accelerators . . . . . . . . . . . . . . . . . . . . . . . 101.3 Imaging for Radiation Therapy Planning . . . . . . . . . . . . . . . 15vi1.3.1 Computed Tomography . . . . . . . . . . . . . . . . . . . . . 151.3.2 Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . 161.3.3 Positron Emission Tomography . . . . . . . . . . . . . . . . 201.3.4 Diffusion Imaging . . . . . . . . . . . . . . . . . . . . . . . . 221.4 Radiation Therapy Planning . . . . . . . . . . . . . . . . . . . . . . 241.4.1 Planning Volumes and Dose Prescription . . . . . . . . . . 241.4.2 Treatment Planning Techniques . . . . . . . . . . . . . . . . 261.4.3 Treatment Plan Optimization . . . . . . . . . . . . . . . . . 281.4.4 Radiobiological Models . . . . . . . . . . . . . . . . . . . . . 291.5 Project Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342 Contouring with 18F-FDOPA Positron Emission Tomography . . . . . 362.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362.2 Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . 382.2.1 Patient Characteristics and Treatment Planning Imaging . 382.2.2 Radiation Therapy Planning . . . . . . . . . . . . . . . . . . 402.2.3 Consensus Contours . . . . . . . . . . . . . . . . . . . . . . . 412.2.4 Interobserver Variability . . . . . . . . . . . . . . . . . . . . 442.2.5 Recurrence Imaging . . . . . . . . . . . . . . . . . . . . . . . 462.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503 Interhemispheric Differences from Diffusion Tensor Imaging . . . . 583.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583.2 Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . 603.2.1 Patient Selection . . . . . . . . . . . . . . . . . . . . . . . . . 603.2.2 Treatment Planning Imaging . . . . . . . . . . . . . . . . . . 61vii3.2.3 Diffusion Tensor Imaging . . . . . . . . . . . . . . . . . . . . 613.2.4 Interhemispheric Difference Images . . . . . . . . . . . . . 673.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734 Biologically-Guided Volumetric Modulated Arc Therapy . . . . . . . 814.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 814.2 Methods and Materials . . . . . . . . . . . . . . . . . . . . . . . . . . 844.2.1 Patients and Imaging . . . . . . . . . . . . . . . . . . . . . . 844.2.2 Volume Delineation . . . . . . . . . . . . . . . . . . . . . . . 844.2.3 Radiation Therapy Planning . . . . . . . . . . . . . . . . . . 884.2.4 Evaluation of Treatment Plans . . . . . . . . . . . . . . . . . 914.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 924.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 955 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . . . 1035.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035.2 Future Work and Other Applications . . . . . . . . . . . . . . . . . 104Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109viiiList of Tables1.1 One-year, two-year, five-year, and ten-year relative survival ratesfor selected high-grade glioma histologies in the United States,from 1995?2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2 Physical properties of isotopes used in positron emission tomog-raphy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211.3 DVH constraints used for intensity-modulated radiation therapyof high-grade gliomas. . . . . . . . . . . . . . . . . . . . . . . . . . . 292.1 Patient, tumour, and therapy characteristics . . . . . . . . . . . . . 392.2 Consensus target volumes obtained from MRI, PET, and MRI-PET,as well as p-values from paired t-tests . . . . . . . . . . . . . . . . 482.3 Number of central recurrences, by tumour grade, that are outsideconsensus MRI, PET, and MRI-PET GTV structures . . . . . . . . . 543.1 Patient characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . 613.2 Mean interhemispheric differences using images that were spa-tially filtered with a mean filter using spheres of diameter 0 mm,10 mm, and 20 mm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 784.1 Dose volume histogram constraints used for volumetric modu-lated arc therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90ix4.2 Characteristics of patients planned for volumetric modulated arctherapy with dose painting . . . . . . . . . . . . . . . . . . . . . . . 924.3 Dosimetric comparison of organs at risk for volumetric modu-lated arc therapy plans with and without dose painting . . . . . . 964.4 Comparison of equivalent uniform doses, in Gy, of organs at riskfor volumetric modulated arc therapy plans with and withoutdose painting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96xList of Figures1.1 The bremsstrahlung spectrum obtained from Monte Carlo simu-lation of a Varian 6 MV linear accelerator . . . . . . . . . . . . . . 31.2 Schematic diagrams of Raleigh scattering, photoelectric absorp-tion, Compton scattering, and pair production . . . . . . . . . . . 51.3 The mass attenuation coefficients for Raleigh scattering, photo-electric absorption, Compton scattering, and pair production forcarbon and lead . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.4 A schematic diagram of the scattering of an electron by an atom,where a is the classical atomic radius and b is the classical impactparameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.5 A linear accelerator with the electronic portal imaging device andkilovoltage cone-beam computed tomography system shown intheir extended positions . . . . . . . . . . . . . . . . . . . . . . . . . 111.6 A schematic diagram of a medical linear accelerator . . . . . . . . 121.7 A schematic diagram of the treatment head of a linear acceleratorwhen used to generate photon beams . . . . . . . . . . . . . . . . . 131.8 A schematic diagram of the treatment head of a linear acceleratorwhen used to generate electron beams . . . . . . . . . . . . . . . . 141.9 A Varian 120-leaf multileaf collimator . . . . . . . . . . . . . . . . 14xi2.1 The gross tumour volume, clinical target volume, and planningtarget volume obtained from extent of gadolinium contrast en-hancement on magnetic resonance imaging and positron emis-sion tomography uptake contoured by one observer . . . . . . . . 422.2 The consensus volume obtained from the positron emission to-mography gross tumour volume contours of five observers usingsimultaneous truth and performance level estimation . . . . . . . 452.3 The definition of the common and encompassing volumes is il-lustrated using two contours . . . . . . . . . . . . . . . . . . . . . . 462.4 The mean interobserver volume overlap, and STAPLE sensitivityand specificity values are shown for the gross tumour volume,clinical target volume, and planning target volume delineated onmagnetic resonance imaging (MRI), positron emission tomogra-phy (PET), and both MRI-PET . . . . . . . . . . . . . . . . . . . . . 472.5 Linear regressions of the volumes of the consensus positron emis-sion tomography and magnetic resonance imaging gross tumourvolume, clinical target volume, and planning target volume . . . 492.6 The magnetic resonance imaging (MRI), positron emission to-mography (PET)/computed tomography, fused MRI and PET, andMRI at time of recurrence are shown to compare the MRI andMRI-PET target volumes for a case with a central recurrence . . 512.7 The magnetic resonance imaging (MRI), positron emission to-mography (PET)/computed tomography, fused MRI and PET, andMRI at time of recurrence are shown to compare the MRI andMRI-PET target volumes for a second case with a central recur-rence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52xii2.8 The magnetic resonance imaging (MRI), positron emission to-mography (PET)/computed tomography, fused MRI and PET, andMRI at time of recurrence are shown to compare the MRI andMRI-PET target volumes for a case with an outside recurrence . 533.1 An example of T1-weighted magnetic resonance imaging withgadolinium contrast enhancement and T2-weighted fluid atten-uated inversion recovery, and the fractional anisotropy and meandiffusivity images obtained from diffusion tensor imaging for asample patient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633.2 Region of interest contours are shown on T1-weighted magneticresonance imaging with gadolinium contrast enhancement andT2-weighted fluid attenuated inversion recovery, and the frac-tional anisotropy and mean diffusivity images obtained from dif-fusion tensor imaging for a sample patient . . . . . . . . . . . . . . 643.3 An example of the Montreal Neurological Institute 152 standardspace T1-weighted average structure template image and the T1-weighted with gadolinium contrast enhancement, fractional an-isotropy, and mean diffusivity images registered to this standardspace for a sample patient . . . . . . . . . . . . . . . . . . . . . . . . 653.4 Region of interest contours on the Montreal Neurological Insti-tute 152 standard space T1-weighted average structure template,T1-weighted with gadolinium contrast enhancement, fractionalanisotropy, and mean diffusivity images registered to this stan-dard space for a sample patient . . . . . . . . . . . . . . . . . . . . 66xiii3.5 The distribution of fractional anisotropy and mean diffusivity val-ues in the gross tumour volume, peritumoural regions of interest,and normal brain tissue for one patient . . . . . . . . . . . . . . . . 693.6 The patient-averaged fractional anisotropy and mean diffusivityvalues for tumour, peritumoural, and normal brain regions ofinterest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 713.7 An example of the fractional anisotropy and mean diffusivity in-terhemispheric difference images for one patient obtained usingunfiltered images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723.8 An example of the fractional anisotropy and mean diffusivity in-terhemispheric difference images for one patient obtained us-ing images that were spatially filtered with a mean filter usinga sphere of diameter 10 mm . . . . . . . . . . . . . . . . . . . . . . 723.9 An example of the fractional anisotropy and mean diffusivity in-terhemispheric difference images for one patient obtained us-ing images that were spatially filtered with a mean filter usinga sphere of diameter 20 mm . . . . . . . . . . . . . . . . . . . . . . 733.10 The distribution of fractional anisotropy and mean diffusivity in-terhemispheric differences in the gross tumour volume and per-itumoural regions of interest for one patient obtained using un-filtered images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743.11 The distribution of fractional anisotropy and mean diffusivity in-terhemispheric differences in the gross tumour volume and per-itumoural regions of interest for one patient obtained using im-ages spatially filtered using a mean filter with a sphere of diam-eter 10 mm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75xiv3.12 The distribution of fractional anisotropy and mean diffusivity in-terhemispheric differences in the gross tumour volume and per-itumoural regions of interest for one patient obtained using im-ages spatially filtered using a mean filter with a sphere of diam-eter 20 mm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763.13 The patient-averaged fractional anisotropy and mean diffusiv-ity interhemispheric difference for the gross tumour volume andperitumoural regions of interest obtained using images that werespatially filtered with a mean filter using a sphere of diameter 0mm, 10 mm, and 20 mm . . . . . . . . . . . . . . . . . . . . . . . . 774.1 The dose that was prescribed for dose painting as a function ofimage intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864.2 Biological target volumes shown on computed tomography, mag-netic resonance imaging, and 18F-FDOPA positron emission to-mography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 874.3 The field arrangement for volumetric modulated arc therapy isshown for axial and three-dimensional views . . . . . . . . . . . . 894.4 A dose volume histogram for the planning target volume (PTV)without dose escalation, and the PTV and biological target vol-umes with dose painting for a sample patient . . . . . . . . . . . . 934.5 Isodose lines are shown for a volumetric modulated arc therapy(VMAT) plan without dose escalation on computed tomography(CT) and 18F-FDOPA positron emission tomography (PET), anda VMAT plan with dose painting on CT and 18F-FDOPA PET . . . 94xv4.6 The dose distribution from the case in Figure 4.5 is shown forvolumetric modulated arc therapy plans without dose escalationand with dose painting . . . . . . . . . . . . . . . . . . . . . . . . . . 954.7 Dose volume histograms for the brainstem and optic chiasm andnerves for volumetric modulated arc therapy plans without doseescalation and with dose painting for a sample patient . . . . . . 974.8 Dose volume histograms for the left and right retinas for volu-metric modulated arc therapy plans without dose escalation andwith dose painting for a sample patient . . . . . . . . . . . . . . . 984.9 Dose volume histograms for the left and right anterior chambersfor volumetric modulated arc therapy plans without dose escala-tion and with dose painting for a sample patient . . . . . . . . . . 99xviList of Abbreviations11C-MET 11C-methionine18F-FDOPA 3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine18F-FDG 18F-fluorodeoxyglucose18F-FET 18F-fluoroethyltyrosine3D-CRT three-dimensional conformal radiation therapyADC apparent diffusion coefficientBED biologically equivalent doseBTV biological target volumeCI conformity indexCSF cerebrospinal spinal fluidCT computed tomographyCTV clinical target volumeDTI diffusion tensor imagingDVH dose volume histogramEUD equivalent uniform doseFA fractional anisotropyFLAIR fluid attenuated inversion recoveryFSL Oxford Centre for Functional MRI of the Brain SoftwareLibraryGTV gross tumour volumexviiGy grayHI homogeneity indexHU Hounsfield unitICRU International Commission on Radiation Units andMeasurementsIMRT intensity modulated radiation therapyL-DOPA 3,4-dihydroxy-L-phenylalaninelinac linear acceleratorMD mean diffusivityMLC multileaf collimatorMNI Montreal Neurological InstituteMRI magnetic resonance imagingMRSI magnetic resonance spectroscopic imagingMU monitor unitMV megavoltageNTCP normal tissue complication probabilityNTD normalized total doseOAR organ at riskPET positron emission tomographyPRV planning organ at risk volumePTV planning target volumeRF radiofrequencyROI region of interestRTOG Radiation Therapy Oncology GroupSF surviving fractionSPECT single photon emission computed tomographyxviiiSTAPLE simultaneous truth and performance level estimationSUV standardized uptake valueTCP tumour control probabilityTE echo timeTI inversion timeTR repetition timeTSE turbo spin echoVMAT volumetric modulated arc therapyWHO World Health OrganizationxixAcknowledgementsFirst of all, I would like to thank my research supervisor Dr. Vitali Moiseenko(University of California, San Diego) and my academic supervisor Dr. StefanReinsberg (University of British Columbia) for their support and assistance asI have completed my studies and research. I would also like to thank Dr. AlanNichol at the British Columbia Cancer Agency for his helpfulness and clinicalinsights. I also am grateful for the guidance and advice of my other supervisorycommittee members, Drs. Anna Celler and Alex Mackay at the University ofBritish Columbia.I would also like to extend my appreciation to other members of the researchteam for these projects: Drs. Fred Hsu, Don Wilson, Roy Ma, Arthur Cheung,Michael McKenzie, Susan Zhang, and Francois Benard at the British ColumbiaCancer Agency, Drs. Brian Toyota and Talia Vertinsky at Vancouver General Hos-pital, and Dr. Elisa Chan at Saint John Regional Hospital. I gratefully acknowl-edge that this research has been completed with generous funding from theNatural Sciences and Engineering Research Council of Canada Alexander Gra-ham Bell Canada Graduate Scholarship and the University of British ColumbiaFour Year Doctoral Fellowship. The work described in Chapter 2 was also sup-ported by the Hershey and Yvette Porte Neuro-Oncology Endowment Fund ofthe BC Cancer Foundation. The work described in Chapters 3 and 4 was alsoxxsupported by the Brain Tumour Foundation of Canada and the Hershey andYvette Porte Neuro-Oncology Endowment FundI would like to express my thanks to my parents, sister, and my entire ex-tended family for their continued support of my education and research. It isgreatly appreciated.xxiChapter 1Introduction1.1 High-Grade GliomasGliomas are cancers that arise from neuroglia. These cells, such as astrocytes,oligodendroglia, microglia, and ependymal cells, provide support and protec-tion in the central nervous system (1). Astrocytes form a supporting networkin the brain and spinal cord, attach neurons to blood vessels, and help regu-late nutrients and ions that are needed by nerve cells. Oligodendroglia supportthe semirigid tissue rows between neurons in the central nervous system andproduce the fatty myelin sheath of neurons. Microglia are small cells that pro-tect the central nervous system by engulfing and destroying cellular debris andmicrobes, such as bacteria. Ependymal cells line the cavities of the brain andspinal cord, produce cerebrospinal spinal fluid (CSF), and with cilia move thefluid throughout the central nervous system.Gliomas account for 28% of all primary brain and central nervous systemtumours and 80% of malignant tumours (2). High-grade gliomas are those clas-sified as either World Health Organization (WHO) grade III (e.g., anaplasticastrocytoma and oligodendroglioma) or grade IV tumours (e.g., glioblastoma)(3, 4). Despite recent technological advances, the prognosis for patients di-agnosed with high-grade glioma is poor (4?7). From 1995?2010, the two-yearrelative survival rate for patients with anaplastic astrocytomas, anaplastic oligo-1Table 1.1. One-year, two-year, five-year, and ten-year relative survival rates for selectedhigh-grade glioma histologies in the United States, from 1995?2010. Data from ref. 2.Histology Grade Relative Survival (%)1-year 2-year 5-year 10-yearAnaplastic astrocytoma III 60.1 42.1 26.5 18.1Anaplastic oligodendroglioma III 80.6 67.7 50.7 37.3Glioblastoma IV 35.0 13.7 4.7 2.4dendroglioma, and glioblastoma in the United States were 42.1%, 67.7%, and13.7% (Table 1.1) (2).The first treatment for high-grade gliomas is surgery?to make a patholog-ical diagnosis and remove as much of the tumour as is deemed feasible andsafe?followed by radiation therapy to the residual tumour (8, 9). The additionof concomitant chemotherapy to radiation therapy improves patient survival(10). Imaging plays an important role in the planning of radiation therapy. Theradiation target volume is identified using postoperative computed tomography(CT) and magnetic resonance imaging (MRI) (11).1.2 External Beam Radiation Therapy1.2.1 IntroductionIn the radiation therapy of cancer, the goal is to eradicate tumour cells usingionizing radiation while minimizing the radiation dose delivered to surroundingnormal tissue. At the British Columbia Cancer Agency, radiation therapy ofhigh-grade gliomas is delivered by external beam radiation therapy, a techniquewhere a target in a patient is irradiated using a radiation beam located outsideof the patient (12). Today, these beams are most commonly megavoltage (MV)x-ray photons or electrons from a medical linear accelerator (linac), although2Figure 1.1. The bremsstrahlung spectrum obtained from Monte Carlo simulation of aVarian 6 MV linear accelerator. Data from ref. 14.gammas rays from teletherapy units (e.g., using the isotope cobalt-60), protons,neutrons, and heavy ions are also used (13).1.2.2 X-ray Production and Photon Interactions with MatterX-ray photons are obtained from bremsstrahlung, or breaking radiation, pro-duced by the deceleration of a high-energy charged particle near the Coulombfield of the nucleus of a target atom with large atomic number (e.g., tungsten).The intensity I of bremsstrahlung resulting from a charged particle of mass mand charge ze incident on a target nuclei with charge Ze is proportional toI?Z2z4e6m2. (1.1)An example of the bremsstrahlung spectrum obtained from a linac is shown inFigure 1.1.3As photons travel through a medium, they are exponentially attenuated:I = I0e??x , (1.2)where ? is the linear attenuation coefficient. The attenuation of photons inmatter is caused by five interactions: Rayleigh (coherent) scattering, photo-electric absorption, Compton scattering, pair production, and photonuclear in-teractions. In Rayleigh scattering (Figure 1.2(a)), the photon is scattered by anatom such that the photon is redirected with its energy unchanged. Ionizationdoes not occur and no energy is transferred from the photon to the medium.In photoelectric absorption (Figure 1.2(b)), the photon is absorbed by abound (e.g., K-shell) electron, which is subsequently ejected from an atom. Theenergy of the ejected electron E isE = h?? Eb (1.3)where h? is the energy of photon and Eb is the binding energy of electron. Theatom returns to the ground state by a cascade of electron transitions, resultingin the production of characteristic x-rays and Auger electrons.In Compton scattering (Figure 1.2(c)), the photon is scattered by an outer-shell or free electron?the binding energy of electron is much less than theenergy of the photon. By applying the laws of conservation of energy and mo-mentum, the energy h?? of the scattered photon is:h?? = h?11 +? (1? cos?), (1.4)4(a) (b)(c) (d)Figure 1.2. Schematic diagrams of (a) Raleigh scattering, (b) photoelectric absorption,(c) Compton scattering, and (d) pair production.the energy E of the ejected electron is:E = h?? h?? = h?? (1? cos?)1 +? (1? cos?), (1.5)and the scattering angle ? of the electron is:cos? = (1 +?) tan?2. (1.6)5h? is the incident photon energy, ? is the scattering angle of the photon, ? =h?/m0c2, and m0c2 = 0.511 MeV is the rest energy of an electron.The pair production of an electron and positron results from the absorptionof photon near the Coulomb field of a nucleus (Figure 1.2(d)). It requires aminimum photon energy of 2m0c2 = 1.022 MeV. For photon energies greaterthan 4m0c2 = 2.044 MeV, pair production can occur in the Coulomb field of anelectron. This is referred to as triplet production since the resulting energy isshared between three particles: the original electron and the electron-positronpair. In addition, the photonuclear interaction of a high-energy photon with thenucleus of an atom can lead to a nuclear reaction and the emission of a nucleon,such as a proton or neutron (15).The quantity ?/? is known as the mass attenuation coefficient, where ?is the density of the medium the photon is traveling through. The total massattenuation coefficient ?/? is the sum of the mass attenuation coefficients fromeach photon interaction. Neglecting photonuclear interactions:??=?coh?+??+??+??, (1.7)where ?coh/?, ?/?, ?/?, and ?/? are the mass attenuation coefficients forRaleigh scattering, the photoelectric effect, Compton (incoherent) scattering,and pair (and triplet) production, respectively. The mass attenuation coeffi-cients for carbon and lead are shown in Figure 1.3.1.2.3 Electron Interactions with MatterElectrons that are set into motion by photons are scattered by the Coulomb elec-tric force fields of atoms. These interactions can be characterized in terms of therelative size of the classical impact parameter b versus the classical atomic ra-6Figure 1.3. The mass attenuation coefficients for Raleigh (coh.) scattering, photoelec-tric absorption, Compton (incoh.) scattering, and pair production for (a) carbon and(b) lead. Data from ref. 16.7Figure 1.4. A schematic diagram of the scattering of an electron by an atom, where ais the classical atomic radius and b is the classical impact parameter.dius a for electron collisions with atoms (Figure 1.4). Soft collisions (b a) oc-cur when the electron is a considerable distance from the atom and its Coulombfield interacts with the atom as a whole, causing an excitation or ionization. Theenergy transferred is on the order of a few eV. They are the most probable in-teraction and account for roughly half the energy transferred to the medium.Hard (or knock-on) collisions (b ? a) occur when the electron interacts with asingle atomic electron. These interactions are responsible for delta-ray produc-tion. These interactions are less probable, but account for approximately halfof the energy transferred to the medium due to the larger energy transferred(a few keV to MeV) per interaction. Soft and hard collisions are inelastic as en-ergy is transferred to the absorbing medium. If the Coulomb interaction takesplace within the nucleus (b  a), the electron can be scattered elastically oran inelastic radiative interaction occurs that deflects the electron and results inthe production of bremsstrahlung.8Inelastic energy losses of an electron moving through a medium of density? are described by the total mass-energy stopping power S/?:S?=1?dEdx, (1.8)where dE is the kinetic energy lost per unit path length dx (17). S/? consists oftwo components, the collisional stopping power Scol/? and the radiative stop-ping power Srad/?:S?=Scol?+Srad?. (1.9)The collisional stopping power characterizes the kinetic energy lost by softand hard (knock-on) collisions. The total collision stopping power Scol/? forelectrons is given by:Scol?= 2pir20Ne?0?2?lnE2 (E + 2?0)2?0 I2+E2/8? (2E +?0)?0 ln2(E +?0)2+ 1? ?2 ???2CZ?, (1.10)where E is the energy of the electron, r0 is the classic electron radius, Ne theelectron density of the material, I is the average ionization energy, ?0 = m0c2is the rest energy of the electron, ? = v/c is the ratio of the electron?s speed vto the speed of light in a vacuum c, ? is a density correction, and C/Z is a shellcorrection (17, 18).The radiative stopping power characterizes bremsstrahlung production re-sulting from electron-nucleus interactions, and can be calculated by:Srad?= 4r20NeZE137?ln2 (E +?0)?0?13?. (1.11)91.2.4 Absorbed DosePhotons do not directly transfer energy to a medium. The electrons that areset in motion by interactions of photons in matter transfer energy to a mediumthrough inelastic collisions. The kinetic energy transferred to the medium, orkerma K , is defined as:K =dEtrdm, (1.12)where dEtr is the net energy transferred to charged particles per unit mass dm.The kerma can be divided into two components: the collisional kerma Kcol,which is the net energy transferred leading to the production of electrons thatdissipate their energy as ionizations near the electron tracks in the medium, andthe radiative kerma Krad, which is the net energy transferred that leads to theproduction of bremsstrahlung x-rays.The energy absorbed in the medium per unit mass is quantified by the ab-sorbed dose D, which is defined by the International Commission on RadiationUnits and Measurements (ICRU) asD =d??dm, (1.13)where d?? is the mean energy imparted by ionizing particles per unit mass dm(19). The unit for absorbed dose is the gray (Gy), defined as 1 Gy = 1 J/kg.1.2.5 Linear AcceleratorsA typical clinical linac (Figure 1.5) provides two photon beams (e.g., 6 MVand 18 MV) and several monoenergetic electron beams (e.g., 6, 9, 10, 12, 16,and 22 MeV) (13). A 6-MV photon beam will consist of a bremsstrahlung x-rayphoton energy spectrum with a maximum photon energy of 6 MeV (Figure 1.1).10Figure 1.5. A linear accelerator with the electronic portal imaging device and kilovolt-age cone-beam computed tomography system shown in their extended positions.The average photon energy is approximately one third of the maximum photonenergy (15).The modern linac consists of a gantry, gantry stand, modulator cabinet,treatment couch, and control console. A schematic diagram of a linac is shownin Figure 1.6. The control console communicates with the modulator cabinet,stand, gantry, and treatment couch. High voltage pulses from the modulatorare fed to a radiofrequency (RF) power generation system: a klystron (a mi-crowave amplifier) or magnetron (a source of high power microwaves). Themodulator cabinet triggers the electron gun to fire when the microwaves enterthe accelerating structure. The accelerating structure is either a traveling waveor standing wave linear accelerator (13). Electrons fired by the electron gun en-ter the accelerating waveguide, in which a high vacuum is maintained. Cavitiesin the accelerating structure produce an electric field pattern that acceleratesthe electrons (13).11Figure 1.6. A schematic diagram of a linear accelerator. Abbreviation: AFC = auto-matic frequency controller.High-energy electrons emerge from the accelerating structure in the formof a pencil beam. This electron beam is then transported to the treatment headof the linac in a straight-through design for low-energy machines (up to 6 MV)using a traveling wave linear accelerator, or via a bending magnet for higher-energy machines, which deflects the beam through 90? or 270? before it entersthe treatment head (15).In the treatment head (Figure 1.7), the electron beam enters a primary col-limator which has a retractable x-ray target at its centre. Bremsstrahlung x-raysare produced when electrons hit the target. A flattening filter is inserted in thebeam to produce a beam with a uniform-intensity field. When a linac is usedto produce electron beams, the target and flattening field are removed fromthe beam line and an electron scattering foil is used to spread out the electronpencil beam and produce a beam with a uniform intensity across the field (Fig-ure 1.8). The beam passes through a dose-monitoring system, which consists of12Figure 1.7. A schematic diagram of the treatment head of a linear accelerator whenused to generate photon beams. Abbreviation: MLC = multileaf collimator. Adaptedfrom ref. 15.dual transmission ionization chambers. The ionization chambers monitor doserate, integrated dose, and field symmetry (15). The beam then passes througha secondary collimator. Newer machines also include a multileaf collimator(MLC) (Figure 1.9) that consists of motorized leaves that provide customizedfield shaping. The photon or electron beam then exits the machine. A slot onthe linac treatment head allows physical wedges, blocks, or compensators thatmodify the beam, or electron collimation systems (i.e., electron applicators) tobe attached to the machine.Modern linacs are constructed so that the gantry rotates about a horizontalaxis, and the secondary collimator and treatment couch rotate about a vertical13Figure 1.8. A schematic diagram of the treatment head of a linear accelerator whenused to generate electron beams. Abbreviation: MLC = multileaf collimator. Adaptedfrom ref. 15.Figure 1.9. A Varian 120-leaf multileaf collimator. Image courtesy of Varian MedicalSystems, Inc. All rights reserved.14axis. The point of intersection of the rotation axes of the gantry and collimatoris called the isocentre (15). The isocentre is typically located 100 cm from theradiation source (i.e., x-ray target). Patients are initially setup for treatmentusing a laser marking system which intersects at the isocentre. Patient positioncan then verified using electronic portal imagers and kilovoltage cone-beam CT(shown in Figure 1.5).Monitor units (MUs) are the measure of the machine output of a linac thatis obtained from the dose-monitoring system. The dose output of a linac isusually calibrated for 1 MU to be equivalent to a dose 1 cGy under standardconditions; for example, 10?10 cm2 field size at a reference depth in water (orwater-equivalent phantom), with the reference point at the isocentre (source-axis-distance set-up).1.3 Imaging for Radiation Therapy Planning1.3.1 Computed TomographyCT imaging plays an important role in radiation therapy planning. It allows forthe localization of internal structures, provides information on the location andsize of target volumes and critical organs, as well as quantitative information ofinhomogeneities within the body. The basic principle of CT is that a narrow fanbeam of x-rays scans across the patient to obtain x-ray projection images whichare then reconstructed into a three dimensional volume.Image reconstruction algorithms generate images with CT numbers, whichare related to linear attenuation coefficients by the equation:CT number =???water?water? 1000 HU . (1.14)15The unit used for CT numbers is the Hounsfield unit (HU). A change in CT num-ber of 1 HU is equivalent to the change of 0.1% of the attenuation coefficientof water.A CT simulator is a CT system that has been specially equipped for use inradiation therapy planning. A CT simulator has a flat-top surface for patientpositioning which is identical to the linac treatment couch table top. In addi-tion, a laser marking system is used to link the coordinate system of the CTsimulator, and thus the location of the treatment isocentre, with the surface ofa patient (e.g., by tattoo markings on the skin). In addition, virtual simulatorsoftware (at the CT simulator workstation or in treatment planning software)allows a user to identify the treatment isocentre and digitally reconstruct radio-graphs. This allows the design of treatment fields, the transfer of patient datato the treatment planning system, and the production of an image for treatmentverification (13).1.3.2 Magnetic Resonance ImagingMRI has developed along with CT as an important modality for radiation ther-apy target localization (15). MRI is based on the principle of nuclear magneticresonance (20, 21). The total angular momentum of a nucleus is often referredto as a nuclear spin. In the presence of a strong magnetic field B0 = B0z?, whereB0 is the magnetic field strength and z? is a unit vector in the z-direction, nuclearspins align parallel (a low-energy state) or antiparallel (a high-energy state) tothe magnetic field. This results in a small net macroscopic magnetization M, de-fined as the total magnetic moment in a unit volume, parallel to the magneticfield. The magnetization precesses around the field at a frequency of?0 = ?B0,called the Larmor frequency. The gyromagnetic ratio ? is ratio of the magnetic16dipole moment to its angular momentum for a given atom or system. MRI ismost often performed for hydrogen nuclei (protons) since the high concentra-tion of hydrogen in the body and its intrinsic high sensitivity lead to a strongdetected signal.Application of an RF pulse produces a magnetic field perpendicular to B0.The frequency of the RF pulse must match the Larmor frequency in order fornuclear spins in the low-energy state to transition to the high-energy state. Thiscauses the magnetization to precess around the new magnetic field. The flipangle describes the angle which the magnetization rotates through while theRF pulse is applied. For example, a 90? pulse will rotate a magnetization that isinitially aligned parallel to B0 into the x y-plane. The flip angle can be changedby varying the duration or strength of the RF pulse.Spin dynamics are described by the Bloch equation:dMd t= ?M?B0 ?Mx x?+ My y?T2?(Mz ?M0) z?T1, (1.15)where M = Mx x? + My y? + Mz z?. Following the application of an RF pulse, thenuclear spins will return to their original alignment. This process is called re-laxation, and is described by the following solution to the Bloch equation (withappropriate boundary conditions):M? = M0e?t/T2 , (1.16)Mz = M01? e?t/T1. (1.17)M? is the transverse and Mz is the longitudinal component of the magnetiza-tion. M0 is the magnetization at time t = 0. The transverse component of themagnetization produces an induction signal which can be measured by a nearby17receiver coil. The time constant T1 characterizes spin-lattice interactions: thetime required for spins in the high-energy state to transfer their energy to theenvironment or lattice. The time constant T2 characterizes spin-spin interac-tions: the time for dephasing of spins following the RF pulse. In practice, de-phasing of spins is much faster due to magnetic field inhomogeneities, externalfields, and local fields induced in the sample being measured. This relaxationis characterized by the time constant T2?.Imaging is achieved by applying gradient magnetic fields that vary linearlyin spatial coordinates, produced by RF coils in three orthogonal directions (15).This varies the Larmor frequency spatially. Localization of a slice (e.g., in thez-direction) is achieved by a slice selection gradient which is applied at thesame time as an RF pulse. Localization within a slice is achieved by frequencyencoding, the application of transverse gradient in the x-direction during signalacquisition, and phase encoding, the application of a gradient in y-directionprior to signal acquisition. Images are obtained by inverse Fourier transforms.By varying the timing and strength of RF pulses and gradients, differentimages can be produced. Using a single pulse (e.g., 90?), the signal measuredis called a free induction decay curve. The spin echo sequence consists of a 90?pulse followed by a 180? pulse, which produces an echo of the free inductiondecay curve at a time that is twice the time between the pulses (22). This timeis called the echo time (TE). The time between each 90? pulse is called therepetition time (TR). The signal S from a spin echo sequence can be written as:S = S01? e?TR/T1e?TE/T2 . (1.18)Different types of images can be achieved using different TE and TR times. T1-weighted images are obtained with TE T2 and TR? T1, T2-weighted images18are obtained with TR T1 and TE ? T2, and spin-density (or proton-density)images are obtained with TR T1 and TE T2.The inversion recovery sequence adds a 180? pulse before the 90? pulse of aspin echo sequence. The time between these pulses is called the inversion time(TI). Assuming a long TR, the longitudinal magnetization Mz at TI is:Mz = M01? 2e?TI/T1. (1.19)Using the inversion recovery sequence, signals for substances with known T1values, such as fat, blood, and CSF, can be suppressed by setting TI = T1 ln2.The usefulness of T2-weighted imaging of the brain is limited by the fact thatwhite and gray matter signals decays much more rapidly than the signal of CSFand motion of the high-intensity CSF signal can cause image artifacts (23). Thefluid attenuated inversion recovery (FLAIR) pulse sequence selective suppressessignal from CSF using TI ? 2300 ms. This allows clinical interpretation of T2-weighted brain images with reduced image degradation from partial volumeeffects and motion artifacts (24).Contrast agents, such as metal chelates of gadolinium, are introduced to en-hance relaxation for clinical applications. In the radiation therapy of high-gradegliomas, the target volume is identified using CT and MRI with gadolinium con-trast enhancement (11). Enhancement of high-grade gliomas on T1-weightedimages arises from blood-brain barrier disruptions which give abnormal ves-sel permeability to gadolinium (11, 25). Surrounding edema is also apparenton T2-weighted FLAIR images. However, pathological studies have shown thatglioma cells can be identified infiltrating the brain beyond the area of contrastenhancement (26). Contrast enhancement is also a nonspecific sign of blood-brain barrier disruptions and cannot accurately differentiate nonspecific post-19surgical changes from residual tumour (27). There is a need for other imagingtechniques that can improve glioma target localization.1.3.3 Positron Emission TomographyPositron emission tomography (PET) can potentially be used in conjunction withCT and MRI to improve localization of malignant tissue. The main principle ofPET is ?+-decay, the decay of a proton into a neutron, positron (antielectron),and electron neutrino:p? n + e+ + ?e . (1.20)In a typical PET study, a tracer labeled with a ?+-emitter is introduced to apatient by injection or inhalation (28). The radiotracer enters the bloodstreamand is accumulated in the organ of interest. A positron that is emitted travelsa short distance in tissue and then annihilates at its first encounter with anelectron. The annihilation process results in the emission of two nearly colinearphotons with an energy of 511 keV each. Detectors placed at 180? are used torecord the arrival of these photons in coincidence (e.g., within a 5?20 ns timingwindow). The lines of response from coincidence detections form projectionswhich are reconstructed into images.Modern systems combine PET and CT scanners. The CT scanner allowsanatomical localization of the metabolic PET data. Most isotopes used for PETare produced in cyclotrons. Some common isotopes used for PET are shown inTable 1.2. Radiotracers are obtained by attaching these isotopes to clinically-useful biomarkers. PET is therefore a functional imaging technique, since itsimages are representative of in vivo biological processes. For example, 18F-fluorodeoxyglucose (18F-FDG) provides a biomarker for glucose metabolism.18F-FDG was the first studied radiotracer for PET imaging of brain tumours.20Table 1.2. Physical properties of isotopes used in positron emission tomography.Isotope Symbol Half-life Ave. Energy Max. Energy Range in(min) (MeV) (MeV) water (mm)Carbon-11 11C 20.3 0.39 0.96 1.1Nitrogen-13 13N 9.97 0.49 1.19 1.4Oxygen-15 15O 2.0 0.73 1.7 1.5Fluorine-18 18F 109.8 0.24 0.63 1.0Abbreviations: Ave. = average; Max. = maximum.However, tumour visualization with 18F-FDG is difficult since high glucose up-take in the normal cortex gives low tumour-to-background contrast (29, 30).Since facilitated transport of amino acids is upregulated in gliomas, PETwith radiolabeled amino acids, such as 11C-methionine (11C-MET) and 18F-fluoroethyltyrosine (18F-FET), has been investigated for glioma imaging (31).These radiotracers have been shown to have a more sensitive signal than 18F-FDG, and have uptake outside the diseased volume identified with conventionalMRI (29, 30, 32?34).The radiotracer 3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine (18F-FDOPA)has also been investigated for PET imaging (35). 18F-FDOPA is an analog to3,4-dihydroxy-L-phenylalanine (L-DOPA), which is the immediate precursor ofdopamine, a neurotransmitter that is predominantly found in the nigrostriatalregion of the central nervous system (36). Since defects in this region corre-spond to neurodegenerative and movement disorders, 18F-FDOPA has been usedfor studies of Parkinson?s disease (36?39). In addition, some studies have sug-gested the use of 18F-FDOPA for the functional evaluation of brain tumours,since the amino acid transport system is highly expressed in brain tumourspathologically, causing an increased uptake of amino acids (29, 36). 18F-FDOPAPET has been shown to have better sensitivity and specificity than 18F-FDG PET21when evaluating both low-grade and high-grade gliomas (29, 38, 40, 41). Thereis an increase of uptake in high-grade gliomas and the uptake identifies diseasenot visible on conventional MRI (40, 41). However, unlike 11C-MET and 18F-FET PET, there is intense uptake of 18F-FDOPA in the normal basal ganglia (37).1.3.4 Diffusion ImagingMRI can also characterize diffusion of water in the brain. Diffusion is the ran-dom motion of molecules due to their thermal energy. The diffusion coefficientD is characterized by Einstein?s equation (42, 43):D =??r2?2n?t, (1.21)where ??r2? is the mean-square distance traveled by a particle in time ?t; nis the number of dimensions. For pure water at 20?C, D = 2.0? 10?3 mm2/s.In the absence of boundaries, diffusion in three dimensions follows a Gaussianprobability distribution:p(?r,?t) =1?(2piD?t)3exp??r24D?t. (1.22)To measure diffusion with MRI, a spin-echo pulse sequence can be used withtwo gradient pulses added before and after the 180? pulse. This is known as theStejskal-Tanner experiment (44). The first gradient dephases the spins and thesecond gradient recombines them. The gradients have no effect in the absenceof motion. A spin that is moving will accumulate additional phase. For simpleisotropic Gaussian motion, the signal measured from a diffusion weighted imageis:S = S0e?bD , (1.23)22where,b = ?2?2G2????3?, (1.24)and G is the strength of the gradient pulse, ? is the duration of the gradientpulse, and ? is the timing between the pulses.Anisotropic diffusion can be characterized with diffusion tensor imaging(DTI) (45). The probability distribution for anisotropic diffusion is:p(?r,?t) =1?(2pi?t)3 |D|exp??r?D?1?r4?t, (1.25)where D the diffusion tensor, a symmetric second-order tensor. To measure thefull diffusion tensor, Stejskal-Tanner measurements must be made in at least sixnoncolinear directions. For a given direction g?i , the signal strength Si is:Si = S0 exp?bi g??i Dg?i. (1.26)The six independent elements of the diffusion tensor are estimated using mul-tiple linear least squares methods or nonlinear modeling using apparent diffu-sivity maps Dapp,i:Dapp,i =lnSi ? lnS0bi, (1.27)where Si is an image with diffusion weighting bi and S0 is an image with nodiffusion weighting (42).The diffusion tensor is diagonalized to find its principle axes and their re-spective eigenvalues: ?1, ?2, and ?3. Several measures are used to characterizediffusion. The mean diffusivity (MD) is:MD =?1 +?2 +?33. (1.28)23The MD is often referred to as the apparent diffusion coefficient (ADC). It re-flects the magnitude of diffusion (i.e., how far a water molecule diffuses duringmeasurement). The fractional anisotropy (FA) is:FA =???32?(?1 ?MD)2 + (?2 ?MD)2 + (?3 ?MD)2?21 +?22 +?23. (1.29)It is a rotationally-invariant measure of the degree to which the diffusivities area function of the diffusion-weighting encoding direction (42, 46). Some studiesalternatively will define the axial diffusivity ??, which the largest eigenvalueof the diffusion tensor (e.g., ?? = ?1), and radial diffusivity ??, which is theaverage of the two smaller eigenvalues (?? = (?2 +?3)/2).1.4 Radiation Therapy Planning1.4.1 Planning Volumes and Dose PrescriptionImages obtained during CT simulation for radiation therapy planning are im-ported into treatment planning systems and the outlines of the patient surface,the tumour volume that will be treated, and tissues and organs that need tobe avoided are contoured. Although the contouring process is CT-based, otherimages (MRI and PET) can also be imported, fused, and used for the treatmentplanning process. The following treatment planning volumes defined by theICRU Reports 50 and 62 (47, 48) are used for radiation therapy planning:? Gross tumour volume (GTV): The visible, palpable, or demonstrable ex-tent or location of the tumour.? Clinical target volume (CTV): The GTV plus a margin to account for thesuspected microscopic spread of disease. The CTV is an anatomical and24clinical volume, which is independent of the treatment modality or tech-nique selected.? Planning target volume (PTV): The CTV plus margins to account for ex-pected physiologic movements and variations in size, shape, and positionof the CTV during therapy (the internal margin), and uncertainties in pa-tient positioning and alignment of beams (the set-up margin). The PTVis a geometrical concept, which depends on the modality and techniqueused for treatment. The dose variation in the PTV is generally requiredto be between 95% to 107% of the prescribed dose for conventional frac-tionation schemes (e.g., 2 Gy/fraction).? Organs at risk (OARs): Normal tissues whose radiation sensitivity maysignificantly influence treatment planning and/or prescribed dose. Forcritical organs, a margin to account for set-up errors may be also appliedto create a planning organ at risk volume (PRV).The target volumes for high-grade glioma radiation therapy are identifiedusing CT and MRI: T1-weighted images with gadolinium contrast enhancementand T2-weighted images (11, 25). However, it is known that tumour cells infil-trate beyond the area of gadolinium contrast enhanced on T1-weighted images(26). It is believed that the edema volume overestimates the tumour volume.Thus, two protocols have emerged for target volume definition and dose pre-scription.The European Organisation for Research and Treatment of Cancer protocol26052-22053 (49) defines the GTV as the contrast-enhancing tumour visibleon T1-weighted MRI. The CTV is defined as a 2-cm expansion of the GTV andpostsurgical tumour bed. Edema visible on T2-weighted images is included in25the CTV. The PTV is a 0.5-cm expansion of the CTV, and a dose of 60 Gy in 30daily fractions is prescribed.The Radiation Therapy Oncology Group (RTOG) 0825 protocol (50) de-fines two GTVs: GTV1 and GTV2. GTV1 is defined as the abnormality seen onT2-weighted images. GTV2 is defined by the contrast-enhancement seen on T1-weighted images. The clinical target volumes CTV1 and CTV2 are defined by2-cm expansions of GTV1 and GTV2, respectively. The planning target volumesPTV1 and PTV2 are additional expansions of 3 to 5 mm, depending on the local-ization method and set-up reproducibility of CTV1 and CTV2. The prescribeddoses are 46 Gy in 23 fractions for PTV1 and 60 Gy in 30 fractions for PTV2.The appropriate contouring protocol for glioma radiation therapy remains un-clear. Several authors have advocated using smaller margins than those in theRTOG guidelines since currently observed outcomes have failed to demonstratea preferred method of prescription (51?53).1.4.2 Treatment Planning TechniquesWith three-dimensional anatomical information available from CT, modern ra-diation treatments of high-grade gliomas are delivered with three-dimensionalconformal radiation therapy (3D-CRT), intensity modulated radiation therapy(IMRT), or volumetric modulated arc therapy (VMAT) (9, 49, 54). In general,3D-CRT refers to treatments that use three-dimensional anatomical informationto obtain dose distributions which conform as closely as possible to the targetvolume while minimizing dose in normal tissues (15).Optimization of 3D-CRT plans is completed in a forward planning process.In treatment planning systems, after the target volume and normal tissues aredelineated, a user selects the beam arrangement, shapes fields using beam?s-26eye-view visualizations, and selects beam energies, beam weights, and any mod-ifiers (e.g., wedges or compensators) that will be placed in the field. A three-dimensional dose calculation is then performed. If the treatment plan doesnot meet required constraints, the user manually changes the parameters andrecalculates the dose. This process is iterated in a trial-by-error fashion untilan acceptable plan is produced. For complex cases, this process can be labourintensive.The flattening filter of a linac generates beam profiles that are uniform. Ingeneral, IMRT refers to the use of radiation beams with non-uniform fluencesto deliver a conformal dose to the target. This can be achieved using dynamicMLCs and scanned elementary beams of variable intensity (15). Modern IMRTplans are optimized by inverse planning, where each beam is divided into alarge number of beamlets and the beamlet weights or intensities are optimizedby treatment planning systems to satisfy dose distribution criteria for the targetvolume and OARs (15, 55).The computer-controlled MLC can produce intensity modulation in a num-ber of ways. In step-and-shoot IMRT, the treatment fields are sub-divided intosubfields, which are irradiated with uniform intensity (56). The subfields aredelivered sequentially by the linac without operator intervention, with the beamoff while the MLC moves between subfields. The sum of the dose delivered bythese subfields results in intensity modulation. Dynamic delivery, also referredto as a sliding window, leaf-chasing, camera-shutter, or sweeping-variable gap,refers to a technique where corresponding (moving) leaves sweep across thefield simultaneously and unidirectionally, each leaf traveling with a differenttime-dependent velocity (15). Tomotherapy is an IMRT technique in which a pa-tient is treated slice-by-slice, analogous to a CT scanner, by intensity-modulated27beams (57). The slice size is variable and intensity modulation is achieved usinga binary MLC.Intensity modulation can also be achieved with arc therapies, where thebeam is delivered while the gantry rotates around the patient. Intensity mod-ulated arc therapy was the first technique in which the MLC was used dynam-ically to shape the treatment fields while the gantry rotated (58, 59). Devel-oped by Otto (60), VMAT also obtains intensity modulation of a field in a sin-gle treatment arc with a dynamic MLC with the plan inversely optimized bydirect-aperture optimization (61). VMAT is the precursor to the RapidArc sys-tem available commercially from Varian Medical Systems, Inc.1.4.3 Treatment Plan OptimizationIMRT treatment plan optimization is achieved through dose-volume constraintsobtained from dose volume histograms (DVHs), each assigned its own priority.A DVH is a plot of the differential or cumulative volume of a target or normalstructure that receives a given dose. Dose and volume data in DVHs is rep-resented in either absolute terms (e.g., Gy and cm3) or as percentages of theprescribed dose or total structure volume. Vx (or Vx%) is the volume of a struc-ture that receives a dose of at least of x (x% of the prescribed dose). Dx (orDx%) is the minimum dose received by the hottest x (x%) volume of a struc-ture. Care must be taken in interpreting DVH data since it does not give anyinformation on the spatial distribution of dose. In addition, DVH data expressedas percentages must be examined carefully for cases where an entire structuremay not have been contoured. Dose-volume parameters used for high-gradeglioma treatment planning are shown in Table 1.3.28Table 1.3. DVH constraints used for intensity-modulated radiation therapy of high-grade gliomas.Structure Type DVH ConstraintPTV Target V95% ? 98%Optic Chiam and Nerve OAR V54 Gy ? 1%Brainstem OAR Dmax = 60 GyRetina OAR Dmax = 45 GyAnterior Chamber OAR Dmax = 10 GyAbbreviations: DVH = dose volume histogram; max = maximum; OAR = organ at risk.Dose-volume constraints used for IMRT optimization are driven by treat-ment outcomes. The goal is to obtain an IMRT plan will all constraints ful-filled. This is done by minimization of an objective function. Minimization ofthe mean square deviations between the actual dose distribution and the dose-volume constraints is one the most common objective function used for IMRTplanning, which penalizes voxels in the dose distribution that do not meet thedose-volume constraints placed on the OARs and target volumes (62).1.4.4 Radiobiological ModelsDose-volume constraints used in IMRT plan optimization are surrogates for pa-tient outcomes following radiation therapy. Radiobiological models can also beused to evaluate treatment plans. Radiobiological models for normal tissuesaccount for organ architecture. Radiobiological tumour models account for theeffects of proliferation; more advanced tumour models account for the tumourmicroenvironment (e.g., hypoxia). Biological response can ultimately be con-nected to cell kill.Cell survival after irradiation can be described using the linear quadraticmodel (63). The general expression for the surviving fraction (SF) of cells after29receiving a dose D isSF = e??D??GD2+?T , (1.30)where G is the Lea-Catcheside factor, T is the treatment time, ? = ln 2/Td ,and Td is time it takes for the number of tumour cells to double. The factors ?and ? are radiosensitivity parameters which characterize yields of lethal lesionsproduced through single-track or double-track pathways following radiation-induced damage in deoxyribonucleic acid. Clinical data for gliomas suggeststhat ?= 0.06? 0.05 Gy?1 and ?/? = 10.0? 15.1 Gy (64).The Lea-Catcheside factor G accounts for changes in cell lethality due tofractionation or protracted irradiation (65). Assuming exponential repair,G =2D? ???D?(t) d t? t??e??(t?t?)D?(t ?) d t ? , (1.31)where 1/? is the typical lifetime for a sublethal lesion. For fractionated treat-ments, G = 1/n f where n f is the number of fractions. This assumes that thattime between fractions is  1/? (each fraction acts independently) and thetime to deliver the dose for one fraction is 1/? (an acute exposure); i.e., norepair of sublethal lesions while the beam is delivered. Neglecting tumour cellproliferation, SF for fractionated treatments is:SF = e??D??D2/n f = e??D??dD , (1.32)where d = D/n f is the dose delivered per fraction. In terms of the survivingfraction SF2 of cells after receiving a single dose of 2 Gy and ?/? ,SF = SFDiDref??/?+Di/n f?/?+Dref2 . (1.33)30A limitation of the linear quadratic model is that it predicts that the survivalcurve continuously bends, which does not agree with experimental data (66,67). For an acute irradiation, the linear quadratic model overpredicts cell killabove a certain cell-type dose threshold; for mammalian cells, a threshold of12?15 Gy is reasonable (68, 69).The biologically equivalent dose (BED), also referred to as the extrapolatedresponse dose, is used to compare the biological effectiveness of radiation ther-apy schemes. The BED is the dose that if delivered in a protracted fashion willlead to same cell survival as the actual dose delivery scheme:BED = ?ln SF?. (1.34)For fractionated treatments without accounting for proliferation:BED = D?1 +d?/??. (1.35)If two fractionation schemes lead to the same BED, it is assumed that they leadto the same biological response.A related quantity is the normalized total dose (NTD), which is the totaldose if given in 2-Gy fractions that is biologically equivalent to a total dose Ddelivered in fractions of size d:NTD = D??/? + d?/? + 2 Gy?. (1.36)The NTD is often used since dose response is well established for fractionationschemes of 2 Gy per fraction.Biological response to treatments can be estimated using the tumour controlprobability (TCP), the proportion of patients who show local control for a given31dose, and the normal tissue complication probability (NTCP), the proportion ofpatients who show complication of a certain grade. NTCP is popularly charac-terized with the Lyman-Kutcher-Burman model (70). TCP can be characterizedby a Poisson-based model or logistic equation (71).In addition, the equivalent uniform dose (EUD) can be used for assessmentof radiation therapy plans. The EUD is the dose that if given uniformly to a struc-ture will give the same biological effect as the actual heterogeneous dose distri-bution delivered, assuming that the same dose per fraction is maintained duringtreatment (72). Using a cell killing-based model with the linear quadratic for-malism, the EUD can be calculated from:N?i=1Vi ??i ? SFEUDDref??/?+EUD/n f?/?+Dref2 =N?i=1Vi ??i ? SFDiDref??/?+Di/n f?/?+Dref2 . (1.37)Solving for EUD givesEUD =n f2?????+???????2+ 4 ?Drefn f????+ Dref??lnAln SF2?? , (1.38)whereA=N?i=1Vi ??i ? SFDiDref??/?+Di/n f?/?+Dref2,N?i=1Vi ??i (1.39)and Di is the total dose delivered in n f fractions to a given voxel with volumeVi and tumour cell density ?i . N is the total number of dose matrix voxels inthe structure. The surviving fraction SF2 of cells after receiving a single dose ofDref = 2 Gy and the radiosensitivity parameter ?/? are assumed to be the samefor all voxels.32The EUD in Eq. 1.38 is applicable to target volumes. A generalized EUD fortarget volumes and OARs is given by (73):EUD =? N?i=1viDai?1/a, (1.40)where vi is the relative volume of a structure that receives a dose Di , obtainedfrom a differential DVH. The parameter a describes the volume-effect for or-gans. To minimize the effect of cold spots (regions of low dose), a is taken asnegative (a < ?10) for tumours. For serial organs, such as the spinal cord andoptic nerve, whose complications depends highly on the maximum dose deliv-ered, a is large (a > 10). For parallel organs, such as lung and kidney, the meandose is predictive of normal tissue complication (a ? 1) (74). Therefore, theEUD for parallel organs approaches the mean dose, whereas for serial organs itapproaches maximum dose.Advances in patient imaging may allow the incorporation of biological in-formation into radiation therapy planning. Ling et al. (75) first proposed thisapproach as multidimensional radiotherapy, suggesting that biological informa-tion from functional images can be introduced to radiation therapy planning byintroduction of a biological target volume (BTV). The BTV can be used to iden-tify regions within the target volume, such as regions of greater tumour celldensity, cell proliferation, or radioresistance, which may benefit from radiationdose boosts. IMRT allows the planned dose distribution to a BTV to be optimizedby dose painting, where the dose distribution is optimized to conform to thebiological information obtained from functional imaging techniques (75, 76).Biological models can also be incorporated into IMRT treatment planning byusing an objective function based on EUD, TCP, or NTCP (73, 77).331.5 Project OutlineBiological information from functional imaging techniques, such as PET andDTI, can potentially be used to improve localization of malignant tissue aswell as for biologically-based treatment planning. This project investigated theutility of target localization using uptake of 18F-FDOPA and interhemisphericdifference images obtained by DTI, and explored the possibility for a uniquebiologically-guided, physically-based radiation therapy planning technique forhigh-grade gliomas.First, a contouring study was completed to determine the utility of 18F-FDOPA PET for radiation therapy planning of high-grade gliomas. PET with18F-FDOPA visualizes glioma that is not clearly identified on MRI. In a study of19 patients, consensus target volumes obtained from five observers using 18F-FDOPA PET and MRI were compared with the location of recurrences follow-ing radiotherapy. Simultaneous truth and performance level estimation (STA-PLE) was used to calculate consensus target volumes from the observers? delin-eations and interobserver variations of MRI-based and PET-based delineationswere quantified. Recurrence volumes following radiotherapy were contouredby each observer and consensus recurrence volumes calculated using STAPLEwere compared with the consensus target volumes.The second study aimed to determine if FA or MD images obtained frompostoperative DTI can be used to improve radiation therapy target localizationof gliomas. This was done first by characterizing the distribution of FA andMD in the tumour target volumes, obtained from T1-weighted and T2-weightedMRI images. In addition, the distribution of FA and MD was characterized in re-gions of interest (ROIs) that were peritumoural shells around the target volume,and were compared with the values of contralateral normal brain. In addition,34a method was implemented to automatically calculate FA and MD interhemi-spheric difference images for potential use for radiation therapy planning.Lastly, a treatment planning study compared VMAT treatment plans for high-grade gliomas using a unique biologically-guided, physically-based treatmentplanning technique. This was accomplished by dose painting by contours using18F-FDOPA PET with dose constraints for each contour specified by a radiobio-logical model. This method was compared with VMAT plans obtained with con-ventional MRI-based target volume contours to determine the potential benefit,if any, can be realized with biological-based treatment planning of high-gradegliomas, while maintaining acceptable avoidance of critical structures in thebrain.35Chapter 2Contouring with 18F-FDOPAPositron Emission Tomography2.1 IntroductionAs standard practice in the radiation therapy of high-grade gliomas, the targetvolume is identified using postoperative computed tomography (CT) and mag-netic resonance imaging (MRI) with intravenous contrast. The area of gadolin-ium contrast enhancement on T1-weighted MRI is often used to define gross tu-mour extent radiologically. However, several pathological studies have shownthat the contrast enhancement may not represent the outer tumor border sinceinfiltrating glioma cells can be identified well beyond the area of enhancement(26). Gadolinium contrast enhancement is also a nonspecific sign of blood-brainbarrier disruptions and cannot accurately differentiate nonspecific postsurgicalchanges from residual tumour (27).Positron emission tomography (PET) can potentially be used in conjunctionwith CT and MRI to improve localization of malignant tissue. The amino acidtracer 3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine (18F-FDOPA) is taken up bycells using a specific amino acid transport system, rather than breakdown ofthe blood-brain barrier (31). Increased protein synthesis and upregulation ofamino acid transport in the supporting vasculature of brain tumour tissue is re-36sponsible for increased 18F-FDOPA uptake in tumour cells. 18F-FDOPA is alsotaken up by the dopaminergic system, resulting in increased activity seen inthe basal ganglia. Amino acid tracers, such as 11C-methionine (11C-MET), 18F-fluoroethyltyrosine (18F-FET), and 18F-FDOPA, are more useful for PET imagingof brain tumours than 18F-fluorodeoxyglucose (18F-FDG) because of high uptakein tumour tissue and low uptake in normal brain tissue, giving better tumour-to-normal-tissue contrast (30, 32, 33, 40, 78). 18F-FDOPA PET/CT also providesaccurate anatomic localization of newly diagnosed and recurrent, low-gradeand high-grade gliomas when fused with MRI and treatment planning CT (41).It has also been shown to identify regions of high cellular density and higher-grade disease in newly diagnosed and recurrent astrocytomas (79). Thus, theaddition of 18F-FDOPA PET/CT to treatment planning may translate into clini-cally significant improvements in radiation therapy outcomes.Interobserver variations in target volume contouring also have importantimplications for radiation therapy delivery and outcome. Recent brain segmen-tation studies (80, 81) have used simultaneous truth and performance level es-timation (STAPLE) to estimate consensus volumes from delineations of multipleobservers. STAPLE iteratively calculates a probabilistic estimate of the consen-sus volume by an expectation maximization algorithm that for each observer op-timizes the sensitivity (the relative frequency that an observer includes a voxelin their contour when that voxel is inside the consensus volume) and specificity(the relative frequency that an observer does not include a voxel when it isoutside the consensus volume) (82).The objective of this study was to compare the post-radiation therapy delin-eation of recurrences with pre-radiation therapy delineation of newly diagnosedhigh-grade gliomas from the gadolinium contrast enhancement of T1-weighted37MRI and 18F-FDOPA PET/CT using contours from five observers by calculatingconsensus target volumes using STAPLE.2.2 Methods and Materials2.2.1 Patient Characteristics and Treatment Planning ImagingNineteen patients with newly diagnosed high-grade gliomas underwent radia-tion therapy treatment planning with postoperative CT and T1-weighted MRIwith gadolinium contrast enhancement. 18F-FDOPA PET/CT images were alsoobtained at the time of treatment planning. Institutional research ethics boardapproval was obtained for all image protocols and all subjects provided writteninformed consent. Eligible patients had a new diagnosis of World Health Orga-nization (WHO) grade III or IV glioma and Karnofsky Performance Status of 60or better. Patients were ineligible if they were on medications for Parkinson?sdisease or had contraindications to contrast-enhanced MRI or radiation therapy.T2-weighted fluid attenuated inversion recovery (FLAIR) images were also ob-tained for 16 patients (84%). Patient, tumour, and therapy characteristics areshown in Table 2.1. The time between surgery, imaging, and radiation therapystart date is also shown in Table 2.1.MRI was obtained with a 1.5-T Siemens Magnetom Symphony Tim system(Siemens Healthcare, Erlangen, Germany). For 16 patients (84%), T1-weightedimages with gadolinium contrast enhancement were obtained with the turbospin echo (TSE) sequence (echo time (TE) = 14 ms, pixel resolution = 1 mm,slice thickness = 3 mm) and T2-weighted FLAIR images were obtained (TE =97 ms, pixel resolution = 0.5 mm, slice thickness = 3 mm). For the remainingthree patients (16%), only T1-weighted images with gadolinium contrast en-38Table 2.1. Patient, tumour, and therapy characteristics.No. of Patients 19SexFemale 8 (42%)Male 11 (58%)Age, median (range) 52 (19?74) yearsHistologyGlioblastoma 12 (63%)Anaplastic astrocytoma 4 (21%)Anaplastic oligodendroglioma 3 (16%)Extent of resectionGross total resection 9 (47%)Partial resection 6 (32%)Biopsy-only 4 (21%)Chemotherapy 17 (89%)Dose prescription and fractionation60 Gy in 30 fractions 10 (53%)59.4 Gy in 33 fractions 8 (42%)40 Gy in 15 fractions 1 (5%)Time, median (range)Surgery to MRI 23 (9?48) daysSurgery to PET/CT 28 (10?48) daysSurgery to radiation therapy start date 34 (20?55) daysAbbreviations: CT = computed tomography; MRI = magneticresonance imaging; PET = positron emission tomography.hancement were obtained using the magnetization-prepared rapid acquisitionwith gradient echo sequence (TE = 3.5 ms, pixel spacing = 1 mm, slice thick-ness = 1 mm). Radiation therapy planning CT images were also obtained forall patients with a 3-mm slice thickness.18F-FDOPA was synthesized using a previously published procedure (83).The PET/CT images were obtained with a Siemens Biograph-16 Hi-Rez PET/CTsystem (Knoxville, TN). All patients fasted for a minimum of six hours prior tointravenous injection of 3.5 MBq/kg of 18F-FDOPA. The patient?s head was im-mobilized on the scanner table and dedicated noncontrast CT and 15 minute39three-dimensional emission PET images were obtained of the brain 40 minutesfollowing injection. The attenuation corrected PET data was reconstructed us-ing an iterative ordered-subset expectation maximization algorithm (matrix:336 ? 336, brain mode, zoom: 2.5, subsets: 8, iterations: 6, Gaussian filter:2 mm). Reconstructed CT and PET image sets were exported to Siemens andSegami (Segami Corporation, Columbia, MD) workstations for clinical inter-pretation. All image sets (planning CT, MRI, and PET/CT) were imported intoEclipse treatment planning system (Varian Medical Systems, Palo Alto, CA) andfused using the Eclipse registration package. The planning CT images werefused both to the MRI and the CT image set of the PET/CT.2.2.2 Radiation Therapy PlanningAll study subjects received standard radiation therapy to MRI-defined targetvolumes contoured by their attending radiation oncologist. Twelve patients(63%) were treated with three-dimensional conformal radiation therapy (3D-CRT) with 6-MV or 10-MV photon beams. Coplanar intensity modulated radia-tion therapy (IMRT) plans were implemented for treatment for one patient (5%)with five fields and for three patients (16%) with seven fields of 6-MV photons.Three patients (16%) were planned and treated with volumetric modulated arctherapy (VMAT) using 6-MV photons. The IMRT and VMAT treatment planningtechniques and objectives have been described earlier (49). The radiation ther-apy dose prescription was 60 Gy in 30 fractions for 10 patients (53%) and 59.4Gy in 33 fractions for 8 patients (42%). One patient (5%) received an abbre-viated course of 40 Gy in 15 fractions. Seventeen (89%) patients also receivedstandard concurrent and adjuvant chemotherapy with temozolomide.40Five qualified radiation neuro-oncologists delineated the gross tumour vol-ume (GTV) using T1-weighted MRI and 18F-FDOPA PET while blinded to theoriginal planning contours used for treatment. The PET imaging display win-dow and level were determined by matching the 18F-FDOPA uptake in the basalganglia with anatomic MRI-based contours of the basal ganglia for each sub-ject. Observers were not permitted to change the PET imaging display windowand level. They each received training before the study and were advised abouthow to minimize uncertainty of contouring the PET GTVs in the vicinity of thebasal ganglia and in the presence of postoperative inflammation. Each observercontoured an MRI-based GTV, defined as the volume of gadolinium contrast en-hancement excluding the surgical cavity. A PET-based GTV was defined as thevolume of 18F-FDOPA PET uptake excluding the surgical cavity. A combinedMRI-PET GTV was defined as the union of both MRI GTV and PET GTV. Thisdefinition was included since the study design anticipated that a clinician wouldnot omit an enhancing abnormality on MRI from a PET-defined volume duringradiation therapy treatment planning. Clinical target volumes (CTVs) were de-fined as a 2-cm isotropic expansion of the union of the GTV and the surgicalcavity, cropped to exclude any portion lying outside the brain (10). Planningtarget volumes (PTVs) were defined as 0.5-cm isotropic expansions of the CTVs.Examples of these structures are shown in Figure 2.1.2.2.3 Consensus ContoursConsensus contours were obtained by STAPLE (82) using in-house software cre-ated with MATLAB (version 7.0.0.19920; The MathWorks, Inc., Natick, MA).The contours made by each observer were described by a binary decision ma-trix D, where Di j = 1 if image voxel i was included inside observer j?s contour41(a) (b)Figure 2.1. The gross tumour volume (blue), clinical target volume (cyan), and plan-ning target volume (red) obtained from (a) extent of gadolinium contrast enhancementon T1-weighted magnetic resonance imaging and (b) 18F-FDOPA positron emission to-mography uptake contoured by one observer.and Di j = 0 if the voxel was not included inside observer j?s contour. STAPLEestimated the performance level parameters p j and q j of each observer?s seg-mentation by finding the parameters which maximized a log likelihood function(p?, q?) = arg maxp,qln f (D,T|p,q) , (2.1)where Ti was the hidden binary true segmentation for each voxel, p j was thesensitivity for an observer?s segmentation (relative frequency that Di j = 1 whenTi = 1), and q j was the specificity for an observer?s segmentation (relative fre-quency than Di j = 0 when Ti = 0).The complete log likelihood function ln f (D,T|p,q) was not known sinceT was unknown. The performance level parameters were found using an ex-pectation maximization algorithm. The details of these steps are discussed by42Warfield et al. (82). In brief, first a conditional expectation of the complete loglikelihood function was computed and then the parameters that maximize thisfunction were identified.The conditional expectation of the complete log likelihood function wascomputed (E-step) using:W (k?1)i = f (Ti = 1|Di,p(k?1),q(k?1)) (2.2a)=a(k?1)ia(k?1)i + b(k?1)i(2.2b)wherea(k)i = f (Ti = 1)?j:Di j=1p(k)j?j:Di j=0(1? p(k)j ) , (2.3a)b(k)i = f (Ti = 0)?j:Di j=0q(k)j?j:Di j=1(1? q(k)j ) . (2.3b)Wi was the probability that the Ti = 1; it was the normalized product of a priorprobability that the voxel was inside the true segmentation, the sensitivity of allobservers that included that voxel inside their segmentation, and 1? sensitivityof all observers that excluded the voxel from their segmentation. The priorestimates of the sensitivity and specificity were initialized with values very closebut not equal to one:p(0)j = q(0)j = 0.9999 . (2.4)For simplicity, a uniform prior probability was used:f (Ti = 1) =1RNR?j=1N?i=1Di j , (2.5)43where R = 5 was the number of observers and N was the total number of voxelsin a region of interest (ROI) around the contours. The ROI was chosen to be thesmallest box that encompassed all contours to ensure that most voxels that allobservers did not include in their contour were not included in the calculations.The performance level parameters that maximized the conditional expecta-tion of the complete log likelihood function were then calculated (M-step) usingthe equations:p(k)j =?i:Di j=1 W(k?1)i?i W(k?1)i, (2.6a)q(k)j =?i:Di j=0?1?W (k?1)i??i?1?W (k?1)i? . (2.6b)The E-step (Eq. 2.2) and M-step (Eq. 2.6) were iterated until the sumSk =N?i=1W (k)i , (2.7)converged (Sk ? Sk?1 < 10?12).Consensus contours were then obtained using a voxel-wise maximum like-lihood approach (84). An example of the consensus PET GTV contour obtainedfrom the contours of five observers is shown in Figure 2.2.2.2.4 Interobserver VariabilityInterobserver contour variability was quantified by the percentage of volumeoverlap defined as the ratio of the volume common to all contours to the volume44Figure 2.2. The consensus volume (blue wash) obtained from the positron emissiontomography gross tumour volume contours of five observers (red, blue, green, cyan,and magenta) using simultaneous truth and performance level estimation.encompassed by all contours:volume overlap =common volumeencompassing volume? 100% . (2.8)The definition of the common and encompassing volumes is illustrated in Fig-ure 2.3. In the case of two overlapping contours, the volume overlap is equiv-alent to the Jaccard index (85, 86). The STAPLE sensitivity and specificity(Eq. 2.6) values were also used to quantify interobserver contour variability.Differences between MRI and PET interobserver volume overlap were testedusing a two-sided paired t-test (?= 0.05).Differences in the volume of the consensus MRI, PET, and MRI-PET contourswere quantified using linear regression and statistically tested using two-sidedpaired t-tests (? = 0.05). The overlap of consensus MRI CTVs and PET GTVs45Figure 2.3. The definition of the common and encompassing volumes is illustratedusing two contours.were also compared to determine if the standard 2-cm margin from MRI GTVto MRI CTV encompassed 18F-FDOPA uptake. This was done to determine ifradiation therapy planning with only MRI would result in a geographic miss ofthe PET GTV.2.2.5 Recurrence ImagingRecurrence imaging was available for 12 patients (10 MRI and 2 CT), witha median follow-up time of 4.6 months (range, 2.3?20 months). Recurrenceimage sets were fused to the planning CT volume and recurrences followingradiation therapy were contoured by each observer with each observer blindedto the previous contours. Consensus recurrence volumes were obtained usingSTAPLE. Patterns of failure were classified as central (>95% of consensus re-currence volume within 95% isodose), in-field (80?95% within 95% isodose),marginal (20?80% within 95% isodose), or outside (<20% within 95% isodose)(87). Consensus recurrence volumes were compared with margins of 0 cm, 1.5cm, 2 cm, and 2.5 cm on the consensus MRI and PET GTV and the percentageof recurrence volume extending outside the consensus MRI and PET GTV wascalculated.46Figure 2.4. The mean interobserver volume overlap, and STAPLE sensitivity and speci-ficity values are shown for the gross tumour volume (GTV), clinical target volume(CTV), and planning target volume (PTV) delineated on magnetic resonance imaging(MRI) (white), positron emission tomography (PET) (light gray), and both MRI-PET(dark gray). Standard deviations are shown by error bars.2.3 ResultsInterobserver volume overlap and STAPLE sensitivity and specificity values areshown for the MRI, PET, and MRI-PET target volumes in Figure 2.4. The meaninterobserver volume overlap of PET GTV contours (42%? 22%) was not sig-nificantly different from the mean interobserver volume overlap of MRI GTVcontours (41% ? 22%, p = 0.67). The mean interobserver volume overlap ofPET CTV (80% ? 12%) and MRI CTV (82% ? 11%) contours were not signif-icantly different (p = 0.25). Similarly, the difference in mean interobservervolume overlap of PET PTV (82%? 11%) and MRI PTV (85%? 10%) contourswas not statistically significant (p = 0.50).Consensus target volumes obtained from MRI, PET, and MRI-PET, as wellas paired statistics are shown in Table 2.2. The mean volume of the consensusPET GTV was 58.6 ? 52.4 cm3 and consensus MRI GTV was 30.8 ? 26.0 cm347Table 2.2. Consensus target volumes obtained from MRI, PET, and MRI-PET, as well asp-values from paired t-tests.ConsensusStructureVolume (cm3) p-valueMRI PET MRI-PET MRI & MRI & PET &PET MRI-PET MRI-PETRecurring cases (n = 12)GTV 23.1? 12.2 46.0? 32.1 51.2? 30.5 0.010 0.001 0.04CTV 214? 73 265? 98 270? 98 0.003 0.001 0.04PTV 325? 97 389? 123 395? 123 0.001 <0.001 0.02Non-recurring cases (n = 7)GTV 44.0? 37.9 80.2? 74.2 84.6? 73.1 0.10 0.06 0.15CTV 309? 169 375? 193 383? 195 0.04 0.015 0.08PTV 444? 220 521? 245 532? 248 0.03 0.011 0.09All cases (n = 19)GTV 30.8? 26.0 58.6? 52.4 63.5? 51.0 0.003 <0.001 0.009CTV 249? 123 306? 146 312? 148 <0.001 <0.001 0.006PTV 369? 159 437? 183 445? 185 <0.001 <0.001 0.004Abbreviations: CTV = clinical target volume; GTV = gross tumour volume; MRI =magnetic resonance imaging; PET = positron emission tomography; PTV = planningtarget volume.(p = 0.003). The mean consensus PET CTV volume was 306?146 cm3 and themean consensus MRI CTV volume was 249? 123 cm3 (p < 0.001). The meanconsensus PET PTV volume was 437? 183 cm3 and the mean consensus MRIPTV volume was 369? 159 cm3 (p < 0.001). This is consistent with linear re-gression of the consensus PET and MRI target volumes, as shown in Figure 2.5.In addition, the consensus MRI CTV margin did not encompass the consensusPET GTV in two cases (11%). In these cases, 1.6 and 2.9 cm3 of the PET GTVextended outside the MRI CTV.Uptake of 18F-FDOPA was identified outside the edema volume apparent onT2 FLAIR images in 14/16 cases (88%). In addition, the apparent edema fromT2 FLAIR images did not detect the 18F-FDOPA uptake outside the MRI CTV forthe case where the PET GTV extended outside the MRI CTV by 2.9 cm3.48(a) (b)(c)Figure 2.5. Linear regressions of the volumes of the consensus positron emission to-mography (PET) and magnetic resonance imaging (MRI) (a) gross tumour volume(GTV), (b) clinical target volume (CTV), and (c) planning target volume (PTV). Un-certainties shown for the slope and intercept are 95% confidence intervals.49Consensus recurrence volumes were analyzed for 12 patients. Eleven (92%)recurrences were central. Glioblastomas accounted for seven (64%) central re-currences and anaplastic tumours accounted for four (36%) central recurrences.The mean volume of the central recurrences was 36.2 ? 37.1 cm3. The inter-observer volume overlap was 55%? 19%. Examples of central recurrences areshown in Figures 2.6 and 2.7. Central recurrences extended outside the con-sensus MRI GTV in all cases and extended outside the consensus PET GTV in10 cases (91%). Table 2.3 shows the number of central recurrences by tumourgrade that extended outside margins of 0 cm, 1.5 cm, 2 cm, and 2.5 cm on theconsensus MRI and PET GTV. The GTV plus 2-cm margin is equivalent to theCTV contour. Two (18%) central recurrences extended outside a 2-cm marginon the MRI GTV, one (9%) central recurrence extended outside a 2-cm marginon the PET GTV and the percentage of the recurrence volume that extended be-yond the PET GTV (52%?28%) was significantly less than the percentage thatextended beyond the MRI GTV (62%? 20%, p = 0.04). One (8%) recurrencewas outside (Figure 2.8), an anaplastic oligodendroglioma with a volume of 2.6cm3 and interobserver volume overlap of 47%.2.4 DiscussionThis study illustrates use of the STAPLE method to assess interobserver variabil-ity and calculate consensus contours. The interobserver contour variability ofMRI-based contours in this study is similar to results previously reported in theliterature. The mean STAPLE sensitivity reported previously for four high-gradeastrocytoma cases was 0.914 (range, 0.753?0.981) and STAPLE specificity was0.999 for all cases (80). In a retrospective study of seven patients with glioblas-toma or anaplastic oligodendroglioma (88), the interobserver volume overlap50(a) (b)(c) (d)Figure 2.6. The (a) magnetic resonance imaging (MRI), (b) positron emission tomog-raphy (PET)/computed tomography, (c) fused MRI and PET, and (d) MRI at time ofrecurrence are shown to compare the MRI and MRI-PET target volumes for a case witha central recurrence. Contours shown: consensus MRI gross tumour volume (GTV)(blue), MRI clinical target volume (CTV) (cyan), MRI planning target volume (PTV)(red), MRI-PET GTV (green) MRI-PET CTV (yellow), MRI-PET PTV (orange), and re-currence (magenta), and the 95% isodose curve (light green).51(a) (b)(c) (d)Figure 2.7. The (a) magnetic resonance imaging (MRI), (b) positron emission tomog-raphy (PET)/computed tomography, (c) fused MRI and PET, and (d) MRI at time ofrecurrence are shown to compare the MRI and MRI-PET target volumes for a secondcase with a central recurrence. Contours shown are the same as in Figure 2.6.52(a) (b)(c) (d)Figure 2.8. The (a) magnetic resonance imaging (MRI), (b) positron emission tomog-raphy (PET)/computed tomography, (c) fused MRI and PET, and (d) MRI at time ofrecurrence are shown to compare the MRI and MRI-PET target volumes for a case withan outside recurrence. Contours shown are the same as in Figure 2.6.of delineations from five observers using CT and MRI registered with surfacematching was on average 47% (range, 21%?72%).In this study, interobserver variability of 18F-FDOPA PET-based target vol-umes was not significantly different than variability of MRI-based target vol-umes. Interobserver volume overlap and STAPLE sensitivity values for CTVsand PTVs were larger than the values of GTVs. Since the CTV and PTV wereisotropic expansions of the GTV, any interobserver variation appeared small rel-ative to the large CTV and PTV volumes since volume overlap, sensitivity, andspecificity are relatively insensitive to interobserver contour differences whenthese differences have a small impact on the total volume (81). Diffusion tensor53Table 2.3. Number of central recurrences, by tumour grade, that are outside consensusMRI, PET, and MRI-PET GTV structures.Structure MRI PET MRI-PETGrade III tumours (n = 4)GTV 4 (100%) 3 (75%) 3 (75%)GTV + 1.5 cm 1 (25%) 2 (50%) 1 (25%)GTV + 2 cm (CTV) 1 (25%) 0 (0%) 0 (0%)GTV + 2.5 cm 0 (0%) 0 (0%) 0 (0%)Grade IV tumours (n = 7)GTV 7 (100%) 7 (100%) 7 (100%)GTV + 1.5 cm 2 (29%) 3 (43%) 1 (14%)GTV + 2 cm (CTV) 1 (14%) 1 (14%) 1 (14%)GTV + 2.5 cm 1 (14%) 1 (14%) 1 (14%)All recurrences (n = 11)GTV 11 (100%) 10 (91%) 10 (91%)GTV + 1.5 cm 3 (27%) 5 (45%) 2 (18%)GTV + 2 cm (CTV) 2 (18%) 1 (9%) 1 (9%)GTV + 2.5 cm 1 (9%) 1 (9%) 1 (9%)Abbreviations: CTV = clinical target volume; GTV = gross tumour volume; MRI =magnetic resonance imaging; PET = positron emission tomography.imaging (DTI) may also be used to define anisotropic, patient-specific GTV-to-CTV margins (89, 90). In this case, there may be more interobserver variabilityof the CTV and PTV contours than that presented in Figure 2.4 and the relativedifference between MRI and PET volumes for the CTV and PTV may be largerthan that presented in 2.5(b) and (c).Consensus PET target volumes were significantly larger than consensus MRItarget volumes. The standard 2-cm margin from consensus MRI GTV-to-CTVdefinition led to geographic miss of the consensus PET GTV in 11% (2/19) ofcases. Recurrence imaging was available for two of these cases and showedthat recurrence volumes were within the 95% isodose surface and contained bya 2-cm margin on the MRI GTV in both cases. Even when reviewed retrospec-tively, the 18F-FDOPA PET image set did not detect the one case of an outside54recurrence with a drop metastasis. Overall, it is unclear if treatment planningusing the PET GTV would yield better treatment outcomes since all but one re-currence extended beyond the PET GTV and most were contained by a 2-cmmargin on the MRI GTV. Despite the small number of cases, the proportion ofrecurrences of grade III and IV tumours that were contained by a 2-cm mar-gin on the MRI GTV and PET GTV were similar (Table 2.3). It is importantto note that any potential differences in treatment outcome could be a resultof either target delineation or treatment technique. However, 18F-FDOPA PETmay detect glioma that is not detectable on MRI. Ledezma et al. (41) reporteda case where 18F-FDOPA PET uptake was in a region with a small amount ofcontrast enhancement on MRI, which could have been attributable to a resid-ual tumour or postsurgical changes, was the site of recurrence three monthsafterward. There is a potential role of 18F-FDOPA PET to include all areas oftumour while leading to a reduction of the GTV-to-CTV margin. This could re-duce the potential for long-term neurocognitive toxicity of large irradiated brainvolumes.Amino acid radiotracers, such as 11C-MET, 18F-FET, and 18F-FDOPA, havebeen comparatively studied and produce very similar quality images for braintumours (91), except that there is normal, physiological uptake of 18F-FDOPA inthe basal ganglia. 11C-MET and 18F-FET do not accumulate in normal anatomicstructures of the brain. Other studies have reported the use of 11C-MET and18F-FET in similar settings to this study. In a prospective study of the treatmentof primary glioblastomas, poor coverage of 11C-MET uptake (which was notused for target volume delineation) was associated with an increased risk ofnoncentral recurrence (92). However, a prospective study of patients treatedusing 18F-FET PET target volumes showed the observed pattern of failure to bepredominantly central (93).55Biologically-based treatment planning with 18F-FDOPA PET may also be pos-sible since there is a correlation between 18F-FDOPA uptake, cell density, andcell proliferation in newly diagnosed tumours (94). Uptake of 18F-FET has beenused to delineate radiation boost volumes. In a phase II trial of dose escala-tion for glioblastomas, a simultaneous integrated boost of 72 Gy in 30 fractionsdelivered to a 18F-FET PET-defined PTV, concurrent with 60 Gy in 30 fractionsdelivered to a traditional MRI-defined PTV, did not lead to a survival benefit(95). 18F-FDOPA uptake can potentially be used for dose painting by numbers.The feasibility of dose painting by numbers for proton treatments has been in-vestigated using 18F-FET PET (96). It is unknown if these techniques will lead toimproved outcomes. Further research is warranted to determine if biologically-based treatment planning with 18F-FDOPA PET/CT can improve outcomes forhigh-grade gliomas. However, it is important to note that twelve patients in thisstudy had a recurrence in about five months, which stresses the need for betterlocal control.This study was limited in several respects, first of all, by the relatively smallsample size (n = 19), with only 63% (12/19) having recurrence imaging avail-able. For a larger sample of patients, the location of recurrent disease couldbe compared to consensus MRI and PET GTVs and this large dataset could beused to derive more accurate GTV-to-CTV margins and also compare resultingnormal tissue involvement in the high dose region.In addition, the intense uptake of 18F-FDOPA in the normal basal gangliamay have made delineation of tumour using 18F-FDOPA uptake more difficultfor tumours located adjacent to these structures, possibly adding uncertainty tothe study contours. Another difficulty with using 18F-FDOPA PET is that post-surgical changes around the resection cavity can exhibit tracer uptake becauseof high levels of amino acid transport by activated macrophages or 18F-FDOPA56leakage due to disruption of the blood-brain barrier (97). However, the studyanticipated these challenges and the neuro-oncologists were advised on how tocontour near the basal ganglia and how to omit regions of inflamed brain fromthe PET GTVs. It is not known how successful the observers were in follow-ing these instructions, because there was no gold standard available to definethe glioma volumes. The true location of postoperative gliomas was unknownand the STAPLE consensus contours obtained from MRI and 18F-FDOPA PETpresented in this study were estimates of the true location of gliomas based onconsensus contours by expert observers.57Chapter 3Interhemispheric Differencesfrom Diffusion Tensor Imaging3.1 IntroductionT1-weighted magnetic resonance imaging (MRI) with contrast enhancementdoes not accurately represent the extent of the tumour for radiation therapyplanning (26, 27). The addition of postoperative diffusion tensor imaging (DTI)to conventional MRI to treatment planning may potentially help improve local-ization of high-grade gliomas and therefore improve treatment outcome sinceanisotropic infiltration of glioma cells has been attributed in part to proliferationdirected along structures in the brain, such as white matter tracks (98?103).Diffusion, the random motion of molecules due to their thermal energy, canbe accurately characterized by MRI. DTI is used to map and characterize thethree-dimensional distribution of anisotropic diffusion of water in the brain. Itshas been widely investigated for several pathologies, such as ischemia, myeli-nation, axonal damage, inflammation, and edema (42). DTI measures oftenreported are the mean diffusivity (MD) and the fractional anisotropy (FA). MD,often referred to as the apparent diffusion coefficient (ADC), reflects the mag-nitude of diffusion (i.e., how far water molecules diffuse during measurement).58FA is a rotationally-invariant measure of the degree to which diffusivities are afunction of the diffusion-weighting encoding direction (42, 46).The role of DTI for the localization of radiation therapy target volumes forhigh-grade gliomas continues to be actively studied. Recent studies have shownthat DTI may indicate glioma cell infiltration that is not visible on conventionalMRI with gadolinium contrast enhancement (104, 105). In addition, DTI hasbeen used to define anisotropic, patient-specific gross tumour volume (GTV)-to-clinical target volume (CTV) margins (89, 90). There is also potential thatDTI may be used to identify regions within the target volume which may benefitfrom radiation dose boosts, such as regions of greater cell density and increasedcell proliferation. FA has been suggested as a predictor of glioma cell densityand proliferation activity (106?109).FA and MD images can be difficult to interpret for the purpose of radiationtherapy planning. Although edema is apparent on MD images (42), FA valuesaround the tumour bed are smaller than in contralateral normal white mat-ter (106). A potential method to improve interpretation of FA and MD imagesis by the calculation of interhemispheric difference images. In the method pro-posed by Aubert-Broche et al. (110), interhemispherical difference images wereused to identify regions of functional interhemispheric asymmetries from brain99mTc-exametazime and 99mTc-ethyl cysteinate dimer single photon emissioncomputed tomography (SPECT) images of the brain using anatomical informa-tion from MRI. In their method, each MRI voxel was matched to its anatomicallyhomologous voxel on the contralateral side. By mapping these voxels to theSPECT image, it was possible to compute SPECT interhemispheric differenceimages.Therefore, the purpose of this study was to determine if FA or MD images ob-tained from postoperative DTI can be used to improve radiation therapy target59localization of gliomas. This was done by first characterizing the distributionof FA and MD in the GTV obtained from conventional T1-weighted and T2-weighted images. In addition, the distribution of FA and MD was characterizedin regions of interest (ROIs) that were shells 0?5 mm, 5?10 mm, 10?15 mm, 15?20 mm, and 20?25 mm regions outside the GTV, and were then compared withvalues in contralateral normal brain. In addition, a method was implementedto automatically calculate FA and MD interhemispheric difference images forpotential use in radiation therapy planning.3.2 Methods and Materials3.2.1 Patient SelectionSeven patients with histologically-confirmed, newly diagnosed glioma under-went radiation therapy planning with postoperative MRI and computed tomog-raphy (CT). Excision of the tumour was performed with MRI and 3,4-dihy-droxy-6-[18F]fluoro-L-phenylalanine (18F-FDOPA) positron emission tomogra-phy (PET) neuronavigation. Institutional research ethics board approval wasobtained for all image protocols and all subjects provided written informed con-sent. Eligible patients were at least 18 years of age, had a contrast-enhancingmass on diagnostic brain CT or MRI that strongly suggested a diagnosis ofWorld Health Organization (WHO) grade III or IV glioma prior to surgery, hada Karnofsky Performance Status of 70 or greater, and had a glomerular filtra-tion rate of 60 mL/min or greater. Exclusion criteria were indication for urgentcraniotomy to relieve mass effect, T1 enhancement or T2 signal that involvedthe basal ganglia, previous intracranial malignancy or any invasive malignancyunless free of disease at least five years, prior cranial irradiation, were taking60Table 3.1. Patient characteristics.No. of Patients 7SexFemale 2 (29%)Male 5 (71%)Age, median (range) 51 (34?80) yearsHistologyOligoastrocytoma (grade II) 1 (14%)Anaplastic astrocytoma (grade III) 1 (14%)Glioblastoma (grade IV) 5 (71%)Time, median (range)Surgery to MRI 6 (2?27) daysAbbreviation: MRI = magnetic resonance imaging.medication for the treatment of Parkinson?s disease (e.g., levodopa), or allergiesor contraindications to contrast MRI or radiation therapy. Patient and tumourcharacteristics, as well as the timing between surgery and postoperative MRI,are shown in Table 3.1.3.2.2 Treatment Planning ImagingMRI was obtained using a 1.5-T Siemens Magnetom Symphony Tim system(Siemens Healthcare, Erlangen, Germany). T1-weighted images with gadolin-ium contrast enhancement were obtained with the turbo spin echo (TSE) se-quence (echo time (TE) = 14 ms, pixel resolution = 1 mm, slice thickness = 3mm) and T2-weighted fluid attenuated inversion recovery (FLAIR) images wereobtained (TE = 97 ms, pixel resolution = 0.5 mm, slice thickness = 3 mm).3.2.3 Diffusion Tensor ImagingDTI was obtained using single-shot echo planar imaging for 20 noncolinear di-rections with b = 1000 s/mm2 and one additional image with b = 0 (TE = 9861ms, repetition time (TR) = 3800 ms, 128? 128 acquisition matrix, pixel reso-lution = 1.95 mm, slice thickness = 5 mm, slice spacing = 1 mm). Diffusionimaging was repeated four times for signal averaging.Images were processed using the Oxford Centre for Functional MRI of theBrain Software Library (FSL) (version 5.0; Oxford, UK) (111?113). The brainextraction tool was used to obtain a brain mask of the T1-weighted image (114).DTI data was corrected for eddy current distortions and then FA (Eq. 1.29) andMD (Eq. 1.28) images were obtained from diffusion tensor fitting (115?119).The following ROIs were delineated: the GTV delineated by the patient?sattending radiation oncologist on T1-weighted and T2-weighted images ex-cluding the surgical cavity, and regions 0?5 mm, 5?10 mm, 10?15 mm, 15?20 mm, and 20?25 mm outside the GTV. ROIs were cropped to ensure thatthey were within the brain mask. All images were linearly registered using a12-parameter affine model with mutual information cost function (120, 121).Figure 3.1 shows an example of the registered T1-weighted, T2-weighted, FA,and MD images for a sample patient. ROI contours for this case are shown inFigure 3.2.In order to calculate a normal brain ROI and interhemispheric difference im-ages (discussed below in Section 3.2.4), all images and ROIs were then alignedto the Montreal Neurological Institute (MNI) 152 standard-space T1-weightedaverage structure template image (122) using the FSL nonlinear registrationtool (123, 124). It used a b-spline representation of the registration warp field(125). A normal brain ROI was defined as the mirror image of the GTV in theleft-right direction in the standard space. An example of these images alignedto the standard space are shown in Figure 3.3. ROI contours in the standardspace are shown in Figure 3.4.62(a) T1 GAD (b) T2 FLAIR(c) FA (d) MDFigure 3.1. An example of (a) T1-weighted magnetic resonance imaging (MRI) withgadolinium contrast enhancement and (b) T2-weighted fluid attenuated inversion re-covery MRI, and the (c) fractional anisotropy (FA) and (d) mean diffusivity (MD) im-ages obtained from diffusion tensor imaging for a sample patient.63(a) T1 GAD (b) T2 FLAIR(c) FA (d) MDFigure 3.2. Region of interest contours are shown on (a) T1-weighted magnetic res-onance imaging (MRI) with gadolinium contrast enhancement and (b) T2-weightedfluid attenuated inversion recovery MRI, and the (c) fractional anisotropy (FA) and (d)mean diffusivity (MD) images obtained from diffusion tensor imaging for a sample pa-tient. Contours shown are the gross tumour volume (GTV) (blue), and regions regions0?5 mm (red), 5?10 mm (green), 10?15 mm (cyan), 15?20 mm (yellow), and 20?25mm (magenta) outside the GTV.64(a) MNI 152 (b) T1 GAD(c) FA (d) MDFigure 3.3. An example of the (a) Montreal Neurological Institute (MNI) 152 stan-dard space T1-weighted average structure template image and the (b) T1-weightedwith gadolinium contrast enhancement, (c) fractional anisotropy (FA), and (d) meandiffusivity (MD) images registered to this standard space for a sample patient.65(a) MNI 152 (b) T1 GAD(c) FA (d) MDFigure 3.4. Region of interest contours on the (a) Montreal Neurological Institute(MNI) 152 standard space T1-weighted average structure template image, (b) T1-weighted with gadolinium contrast enhancement, (c) fractional anisotropy (FA), and(d) mean diffusivity (MD) images registered to this standard space for a sample patient.Contours shown are the gross tumour volume (GTV) (blue), and regions regions 0?5mm (red), 5?10 mm (light green), 10?15 mm(cyan), 15?20 mm (yellow), and 20?25mm (magenta) outside the GTV, and normal brain (orange).66The distribution of FA and MD values inside the ROIs (GTV, 0?5 mm, 5?10mm, 10?15 mm, 15?20 mm, and 20?25 mm outside the GTV, and normal brain)were obtained using the images aligned in the standard space with MATLAB(version 7.0.0.19920; The MathWorks, Inc., Natick, MA). Differences betweenthe mean FA and MD in normal brain and the mean FA and MD in the GTV,and 0?5 mm, 5?10 mm, 10?15 mm, 15?20 mm, and 20?25 mm regions outsidethe GTV were tested for statistical significance using two-sided paired t-tests(?= 0.05).In addition, the mean FA in the GTV, regions 0?5 mm, 5?10 mm, 10?15mm, 15?20 mm, and 20?25 mm outside the GTV, and normal brain ROI werecalculated using a standard FA atlas: the FMRIB58_FA standard space imageavailable in FSL (126). This image is a high-resolution average of 58 well-aligned good quality FA images for healthy male and female subjects (aged,20?60 years). The original DTI resolution for these images was 2? 2? 2 mm.The mean FA obtained from the FMRIB58_FA image was then compared withthe measured mean FA using two-sided paired t-tests (?= 0.05).3.2.4 Interhemispheric Difference ImagesUsing the FA and MD standard space images, interhemispheric difference im-ages for FA (?FA) and MD (?MD) were calculated using the equations:?FA(x , y, z) = FA(x , y, z)? FA(?x , y, z) , (3.1a)?MD(x , y, z) = MD(x , y, z)?MD(?x , y, z) . (3.1b)where FA(x , y, z) and MD(x , y, z) were the mean FA and MD images obtainedusing a mean filter with a sphere centered at voxel (x , y, z) in the standard67space. Interhemispheric difference images were calculated with spheres of di-ameter 0 mm, 10 mm, and 20 mm.The distribution of FA and MD interhemispheric differences were character-ized in the GTV and the regions 0?5 mm, 5?10 mm, 10?15 mm, 15?20 mm,and 20?25 mm outside the GTV. Mean interhemispheric differences were sta-tistically compared using two-sided t-tests (?= 0.05).3.3 ResultsExamples of the distribution of FA and MD in the GTV, regions 0?5 mm, 5?10 mm, 10?15 mm, 15?20 mm, and 20?25 mm outside the GTV, and normalbrain ROIs are shown in Figure 3.5. The distribution of FA and MD suggestedthat FA values tended to increase and MD values tended to decrease as thedistance outside the GTV increased. These values approached those of normalbrain tissue. This is consistent with the expectation of a gradual and decreasingpresence of tumour cells.The mean FA and MD values in each ROI are shown in Figure 3.6. The meanFA in the GTV (0.12 ? 0.03; p = 0.004) and in the regions 0?5 mm (0.15 ?0.03; p = 0.02), 5?10 mm (0.17?0.03; p = 0.07), 10?15 mm (0.20?0.04; p =0.8), and 15?20 mm (0.21? 0.04; p = 0.2) outside the GTV were smaller thanthe mean FA in normal brain tissue (0.20? 0.04). The mean FA in the region20?25 mm (0.24? 0.05; p = 0.02) outside the GTV was larger than the meanvalue in normal brain.The mean FA obtained from the FMRIB58_FA standard space image in theGTV was 0.212 ? 0.006, which was significantly larger than mean measuredFA in the GTV (p = 0.001). The FMRIB58_FA standard space image FA in theregions 0?5 mm, 5?10 mm, 10?15 mm, 15?20 mm, and 20?25 mm outside the68(a)(b)Figure 3.5. The distribution of fractional anisotropy (FA) and mean diffusivity (MD)values in the gross tumour volume (GTV), peritumoural regions of interest, and normalbrain tissue for one patient.69GTV was 0.206 ? 0.006 (p = 0.04), 0.237 ? 0.006 (p = 0.01), 0.246 ? 0.006(p = 0.003), 0.251 ? 0.006 (p = 0.01), and 0.261 ? 0.006 (p = 0.033). Thenormal brain mean FA in the FMRIB58_FA standard space image (0.205?0.006)was not significantly different than the measured FA in that ROI (p = 0.7).The mean MD (?103 mm2/s) was significantly larger in the GTV (1.48 ?0.19; p = 0.01) and regions 0?5 mm region (1.15 ? 0.26; p = 0.10) 5?10mm (1.05 ? 0.11; p = 0.10), 0?15 mm (1.01 ? 0.10, p = 0.01), 15?20 mm(0.99? 0.10, p = 0.009), and 20?25 mm (0.96? 0.10; p = 0.02) outside theGTV than the mean MD in normal brain (0.93? 0.09).Examples of FA and MD interhemispheric difference images that were ob-tained with a mean filter using a sphere of diameter 0 mm, 10 mm, and 20mm are shown in Figures 3.7?3.9 for a sample patient. The distribution of FAand MD interhemispheric differences in Figures 3.10?3.12 a sample patient isshown for images that were obtained with a mean filter using a sphere of di-ameter 0 mm, 10 mm, and 20 mm. In general, the distribution of FA and MDinterhemispheric differences followed the same trends as FA and MD values.This is consistent with the examples shown in Figures 3.7?3.9.The mean and standard error of FA and MD interhemispheric differencesare shown in Figure 3.13 for images that were spatially filtered with a meanfilter using a sphere of diameter 0 mm, 10 mm, and 20 mm. The mean andstandard error of FA and MD interhemispheric differences, along with p-valuesfrom two sided t-tests, for images that were spatially filtered with a mean filterusing a sphere of diameter 0 mm, 10 mm, and 20 mm are shown in Table 3.2.The mean FA and MD interhemispheric differences values were significantlylarger than zero in the GTV for all cases. However, most mean FA and MDinterhemispheric difference values were not significantly larger than zero forthe ROIs outside the GTV.70(a)(b)Figure 3.6. The patient-averaged (a) fractional anisotropy (FA) and (b) mean diffusiv-ity (MD) values for tumour, peritumoural, and normal brain regions of interest. Errorbars show the standard error.71FA diff. MD diff.Figure 3.7. An example of the fractional anisotropy (FA) and mean diffusivity (MD)(?10?3 mm2/s) interhemispheric difference images for one patient obtained using un-filtered images. The gross tumour volume contour is shown in yellow.FA diff. MD diff.Figure 3.8. An example of the fractional anisotropy (FA) and mean diffusivity (MD)(?10?3 mm2/s) interhemispheric difference images for one patient obtained using im-ages that were spatially filtered with a mean filter using a sphere of diameter 10 mm.The gross tumour volume contour is shown in yellow.72FA diff. MD diff.Figure 3.9. An example of the fractional anisotropy (FA) and mean diffusivity (MD)(?10?3 mm2/s) interhemispheric difference images for one patient obtained using im-ages that were spatially filtered with a mean filter using a sphere of diameter 20 mm.The gross tumour volume contour is shown in yellow.3.4 DiscussionThis study is the first of its kind to characterize FA and MD values obtained frompostoperative DTI of high-grade glioma following surgery guided by 18F-FDOPAPET. The trend in DTI parameters is similar to those presented elsewhere. In astudy of DTI of glioblastoma prior to CT-guided stereotactic biopsy, the meanvalues of the FA in the corpus collosum, subcortical white matter, and glioblas-toma lesion were 0.70? 0.05, 0.32? 0.04, and 0.24? 0.05, with mean valuessignificantly different among all three ROIs (p < 0.05) (106). In another study,fiber density mapping and magnetic resonance spectroscopy of 48 patients withgrade II?IV glioma showed similar FA and MD values (127). For grade IV tu-mors, the FA values in the tumour, peritumoural region, and normal appearingwhite matter were 0.224? 0.043, 0.385? 0.043, and 0.469? 0.069, and MDvalues (?10?3 mm2/s) were 1.360? 0.164, 1.033? 0.107, and 0.822? 0.173.73(a)(b)Figure 3.10. The distribution of (a) fractional anisotropy (FA) and (b) mean diffusivity(MD) interhemispheric differences in the gross tumour volume (GTV) and peritumouralregions of interest for one patient obtained using images spatially filtered using unfil-tered images.74(a)(b)Figure 3.11. The distribution of fractional anisotropy (FA) and mean diffusivity (MD)interhemispheric differences in the gross tumour volume (GTV) and peritumoural re-gions of interest for one patient obtained using images spatially filtered using a meanfilter with a sphere of diameter 10 mm.75(a)(b)Figure 3.12. The distribution of fractional anisotropy (FA) and mean diffusivity (MD)interhemispheric differences in the gross tumour volume (GTV) and peritumoural re-gions of interest for one patient obtained using images spatially filtered using a meanfilter with a sphere of diameter 20 mm.76(a)(b)Figure 3.13. The patient-averaged (a) fractional anisotropy (FA) and (b) mean diffu-sivity (MD) interhemispheric difference for the gross tumour volume (GTV) and peri-tumoural regions of interest obtained using images that were spatially filtered with amean filter using a sphere of diameter (a) 0 mm (white), 10 mm (dark gray) and 20mm (light gray). Error bars show the standard error.77Table 3.2. Mean interhemispheric differences using images that were spatially filteredwith a mean filter using spheres of diameter 0 mm, 10 mm, and 20 mm.Structure ?FA ?MD (?10?3 mm2/s)Mean ? SE p Mean ? SE p0-mm filter (unfiltered)GTV 0.073? 0.044 0.004 0.543? 0.219 0.0110?5 mm 0.054? 0.044 0.04 0.203? 0.183 0.035?10 mm 0.050? 0.044 0.07 0.110? 0.142 0.1110?15 mm 0.032? 0.042 0.05 0.089? 0.114 0.0815?20 mm 0.026? 0.041 0.07 0.080? 0.104 0.1220?25 mm 0.020? 0.042 0.07 0.050? 0.098 0.0710-mm filterGTV 0.074? 0.031 0.004 0.550? 0.175 0.0080?5 mm 0.055? 0.030 0.03 0.224? 0.147 0.025?10 mm 0.049? 0.030 0.06 0.116? 0.109 0.0810?15 mm 0.033? 0.030 0.05 0.092? 0.085 0.0815?20 mm 0.025? 0.030 0.06 0.079? 0.078 0.1020?25 mm 0.019? 0.030 0.06 0.051? 0.073 0.0620-mm filterGTV 0.073? 0.022 0.005 0.544? 0.141 0.0060?5 mm 0.054? 0.021 0.02 0.256? 0.115 0.0175?10 mm 0.047? 0.022 0.05 0.133? 0.088 0.0410?15 mm 0.034? 0.023 0.05 0.096? 0.069 0.0615?20 mm 0.025? 0.023 0.06 0.077? 0.065 0.0920?25 mm 0.018? 0.021 0.05 0.051? 0.059 0.05Abbreviations: FA = fractional anisotropy; GTV = gross tumour volumes; MD = meandiffusivity; SE = standard error.However, in the present study the FA in the GTV and normal brain werelower: 0.12?0.03 and 0.20?0.04, respectively. This is due the fact that differentROI definitions were used in this study. In the aforementioned studies, theglioblastoma region of interest was placed at the enhancing central region of thetumour prior to surgery. The definition of the normal brain ROIs in this studyincluded both white and gray matter. However, comparison of the FA values78with FA in the FMRIB58_FA standard space image showed that contralateralnormal brain values were similar to those in healthy patients.The interhemispheric difference images presented here provide guidance onhow DTI images may be utilized for radiation therapy planning. Larger valuesof interhemispheric MD differences likely corresponded to edema. In addition,the area around the tumour with abnormal FA smaller than contralateral nor-mal brain values were more easily visualized than with the original FA images.In Figure 3.6(a), the mean FA in the 20?25 mm region outside the GTV wassignificantly larger than the mean FA in the normal brain ROI. This is incon-sistent with the trend of FA in the other peritumoural shells. This increase ofFA in the 20?25 mm region occurred since this shell was more likely to includethe corpus callosum. Interhemispheric differences in FA in this shell were notsignificantly different from zero (Figure 3.13(a)), consistent with the trend ofinterhemispheric FA differences in the other peritumoural shells.However, for these images to be used to treatment planning, it is necessaryto establish by some method, such as stereotactic biopsy or comparison withthree-dimensional patterns of failure, that these difference values correlate withtumour cell density or proliferation. There is no clear physical mechanism thatdescribes the expected FA and MD values for gliomas. In fact, there have beenconflicting reports of the correlation of FA and MD values with tumour cell den-sity and proliferation. Beppu et al. (106) and Kinoshita et al. (107) found thatFA positively correlated and MD negatively correlated with the cell density in thetumour core, while Stadlbauer et al. (108) and Lee et al. (109) reported that FAnegatively correlated and MD positively correlated. Moreover, a recent study of15 patients with high-grade glioma found minimal anatomical overlap of the theminimum ADC value, a marker of tumour cellularity, obtained from diffusionweighted imaging (maximum b=3000 s/mm2) and the maximum 18F-FDOPA79PET standardized uptake value (SUV) ratio, a marker of tumour infiltration andproliferation (128).In this study, isotropic margins were used to define the regions of interest0?5 mm, 5?10 mm, 10?15 mm, 15?20 mm, and 20?25 mm. Since gliomainfiltration is anisotropic (98?103), the use of anisotropic margins from the DTI(89, 90) could also be combined with the interhemispheric difference imagespresented here to improve radiation therapy target localization. In addition,fiber tractography may also be useful for glioma identification. For example,a study by Stadlbauer et al. (129) of the fiber density mapping of 20 patientswith grade II and III glioma showed a strong negative correlation between fiberdensity and both the logarithm of tumour cell number and the percentage oftumour cell infiltration.Despite the small sample size (n = 7), this study demonstrates that inter-hemispherical difference images obtained from DTI may potentially be usefulfor radiation therapy target volume localization of gliomas. As mentioned ear-lier, for these images to be used clinically it is necessary to establish if eitherFA and/or MD interhemispherical difference values correlate with tumour celldensity or proliferation. In addition, any clinical interpretation of these imagesmust take into account that this method is based on the nonlinear registration ofthe patient images with unhealthy tissue that is being mapped to a standardizedimage of a healthy patient.80Chapter 4Biologically-Guided VolumetricModulated Arc Therapy4.1 IntroductionIntensity modulated radiation therapy (IMRT) and volumetric modulated arctherapy (VMAT) offer a dosimetric advantage to three-dimensional conformalradiation therapy (3D-CRT) in the radiation therapy planning of high-gradegliomas (9, 49, 54, 130?132), maintaining dose coverage to the planning tar-get volume (PTV) while reducing dose to organs at risk (OARs). Clinical studieswith IMRT have reported good patient outcomes (54, 133?138). Although radi-ation therapy of high-grade gliomas with concomitant chemotherapy improvespatient survival (10), patterns of failure following radiation therapy are central(i.e., near the resection margin) (52, 92, 139?142).Functional imaging can potentially be used in conjunction with computedtomography (CT) and conventional magnetic resonance imaging (MRI) to im-prove localization of malignant tissue. It can also be used to identify regionswithin the target volume, such as regions of greater tumour cell density, cell pro-liferation, or radioresistance, which may benefit from radiation dose escalation.The feasibility of such dose boosts has been reported. The Radiation TherapyOncology Group (RTOG) 98-03 study reported that dose escalation to 84 Gy in812-Gy fractions using 3D-CRT for newly diagnosed glioblastoma did not result inany dose-limiting central nervous system toxicities (143). Positron emission to-mography (PET) images have been used to identify regions for dose escalation.In a study of the hypofractionated treatment of glioblastoma, a simultaneousintegrated boost of a 11C-methionine (11C-MET)-defined gross tumour volume(GTV) to 68 Gy in eight fractions, delivered with helical tomotherapy, showedefficacy in controlling tumour cells without evidence of normal tissue toxicity(138).In addition, the intensity modulation achievable with IMRT and VMAT canallow the planned dose distribution to the target to be optimized by dose paint-ing, where the dose distribution is optimized to conform to the functional in-formation obtained from imaging techniques (75, 76, 144). Dose escalationin this manner can be achieved by dose painting by contours or dose paintingby numbers. Dose painting by contours refers to the use of functional imagesto identify subvolumes for radiation boosts, with the dose in each subvolumehomogeneously prescribed (145). Dose painting by numbers refers to the voxel-by-voxel prescription of dose to biological information from images (145?148).Dose painting by contours has been notably suggested for the treatmentof prostate and lung cancer, among others (149?152). Piroth et al. (153) haveused automatically contoured boost volumes of glioblastoma using a 18F-fluoro-ethyltyrosine (18F-FET) PET tumour-to-brain ratio threshold of? 1.6. However,a simultaneous integrated boost of 72 Gy in 30 fractions delivered to those vol-umes with IMRT, concurrent with 60 Gy in 30 fractions delivered to a conven-tional MRI-defined PTV, did not lead to a survival benefit (95). A treatmentplanning study of glioblastoma patients for a simultaneous integrated boost of72 Gy to three-dimensional magnetic resonance spectroscopic imaging (MRSI)-82defined volume delivered with IMRT did not increase dose to normal tissues(154).Dose painting by numbers has been demonstrated in planning studies with18F-fluoromisonidazole PET, a biomarker of hypoxia (155), and clinical trialswith 18F-fluorodeoxyglucose (18F-FDG) PET for head and neck cancers (148,156). For brain tumours, the feasibility of dose painting by numbers using 18F-FET PET for IMRT and intensity-modulated proton therapy has been demon-strated in treatment planning studies (96, 157).3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine (18F-FDOPA) PET may also beappropriate for dose painting in the management of high-grade glioma since18F-FDOPA PET identifies regions of high tumour cell density and higher-gradedisease (79). While other dose escalation strategies have not shown a sur-vival benefit for patients, dose painting may improve tumour control by direct-ing escalated dose to regions of high tumour cell density or high proliferation.Thus, the purpose of this study was to determine the feasibility of dose paint-ing obtained from 18F-FDOPA uptake for the treatment of high-grade gliomasusing VMAT. Dose painting is achieved by contouring biological target volumes(BTVs) using 18F-FDOPA PET for five patients. The dose prescribed to eachBTV was specified by a radiobiological model. This method was compared withVMAT plans obtained without dose escalation to conventional MRI-based con-tours to determine the potential benefit, if any, can be obtained with biological-based treatment planning of high-grade gliomas, while maintaining acceptableavoidance of critical structures in the brain.834.2 Methods and Materials4.2.1 Patients and ImagingFive patients from the cohort of patients described in Chapter 3 were selected.Postoperative planning CT and MRI, T1-weighted images with gadolinium con-trast enhancement obtained with the turbo spin echo (TSE) sequence and T2-weighted fluid attenuated inversion recovery (FLAIR) images, were obtained.Preoperative and postoperative 18F-FDOPA PET were also obtained. Technicaldetails for these imaging modalities are discussed Sections 2.2.1 and 3.2.2. Allimage sets (planning CT, MRI, and PET/CT) were imported into Eclipse treat-ment planning system (Varian Medical Systems, Palo Alto, CA) and fused usingthe Eclipse auto-matching registration algorithm. The planning CT images werefused to both the MRI and the CT image set of the 18F-FDOPA PET/CT.4.2.2 Volume DelineationTarget volume and OAR delineations were imported from each patient?s treatedradiation therapy plan. The GTV was contoured by each patient?s attendingradiation oncologist as the contrast-enhancing tumour on T1-weighted and T2-weighted MRI, excluding the surgical cavity. The clinical target volume (CTV)was defined as a 2-cm expansion of the GTV and surgical cavity. The PTV wasdefined as a 0.5-cm expansion of the CTV. The dose prescribed in the PTV was60 Gy in 30 fractions (10, 49). OARs used for treatment planning were thebrainstem, optic nerves and chiasm, anterior chambers, and retinas. Marginswere not added to OAR contours. Normal brain was not used as an OAR plan-ning constraint in this study.84A radiation oncologist experienced with 18F-FDOPA PET images delineateda volume of interest for dose painting on either the preoperative or postopera-tive 18F-FDOPA PET. Seven BTVs (BTV62.5, BTV65, BTV67.5, BTV70, BTV72.5,BTV75, and BTV75.5) inside the volume of interest were delineated by thresh-olding of the 18F-FDOPA uptake. These volumes corresponded to dose boostsof 62.5 Gy to 77.5 Gy, in 2.5 Gy steps. The image intensity threshold I for eachBTV was calculated using a linear quadratic model that assumes that the imageintensity is linearly related to the tumour cell density ? in each image voxel.For the number of surviving cells in each voxel to be the same:? ? SFDDref??/?+D/n f?/?+Dref2 = ?0 ? SFD0Dref??/?+D0/n f?/?+Dref2 , (4.1)where D is the prescribed dose for the BTV, D0 = 60 Gy is the prescribed dose inthe PTV, n f = 30 is the number of fractions, ?/? = 10 Gy, and SF2 is the surviv-ing fraction of cells after a single dose of Dref = 2 Gy (64, 72, 158). Assumingthat I/I0 = ?/?0:lnII0= ? lnSF2?D?/? + D/n f? D0?/? + D0/n fDref (?/? + Dref)?. (4.2)SF2 was chosen so that that a boost of D = 80 Gy would correspond with themaximum image intensity Imax in the 18F-FDOPA volume of interest in eachpatient. I0 was the minimum image intensity for dose boosting. For this study,it was chosen only to escalate dose in regions where the image intensity waslarger than the image intensity at the anatomic border of the basal ganglia.An example of the dose painting model is shown in Figure 4.1(a). The doseescalation steps for this model are shown in Figure 4.1(b)?(d). An example ofthe target volumes is shown in Figure 4.2.85(a) (b)(c) (d)Figure 4.1. (a) The dose that was prescribed for dose painting as a function of imageintensity (I/I0). The maximum dose was fixed at 80 Gy. The dose painting levelsthat were used to derive biologically-based volumetric modulated arc therapy plans ofhigh-grade gliomas are shown using (b) one dose escalation step (i.e., simultaneousintegrated boost of 70 Gy), (c) three dose escalation steps (65 Gy, 70 Gy, and 75 Gy),and (d) seven dose escalation steps (62.5 Gy to 77.5 Gy, in steps of 2.5 Gy).86(a) (b)(c) (d)Figure 4.2. Biological target volumes shown on (a) computed tomography, (b) mag-netic resonance imaging, and (c) 18F-FDOPA positron emission tomography. Contoursshown are the BTV62.5 (blue), BTV65 (cyan), BTV67.5 (green), BTV70 (magenta),BTV72.5 (orange), BTV75 (red), and BTV75.5 (yellow), GTV (light green), CTV (vio-let), and PTV (pink). Panel (d) shows a close up of panel (c).874.2.3 Radiation Therapy PlanningFor each patient, four VMAT plans were generated using Eclipse treatment plan-ning software using the Anisotropic Analytical Algorithm (version 11.031) fordose calculations and Progressive Resolution Optimization (version 11.031) forVMAT optimization. Plans were obtained using a single 360? arc of a 6-MVphoton beam from a Varian TrueBeam linear accelerator with High Definition120-leaf multileaf collimator (MLC). The field arrangement is shown in Fig-ure 4.3. First, a treatment plan was obtained without any dose escalation. Thedose volume histogram (DVH) planning objectives that were used for treatmentplanning are shown in Table 4.1. In the case of overlapping target and OARs,plan optimization was done with non-overlapping PTVs (PTV60, the portionof the PTV that does not overlap with OARs; PTV54, the portion of the PTVthat overlaps with the optic nerve or chiasm; and PTVb, the portion of the PTVthat overlaps with the brainstem) and OARs (brainstem_opti and optic_opti,the portion of the brainstem and optic structures that do not overlap the PTV).In addition, it was attempted to reduce the mean dose in the OARs to as smallas possible while maintaining dosimetric coverage of the PTV.Dose painting was achieved by progressively adding dose constraints to theBTVs on the treatment plan without any dose escalation. Dose constraints wereadded with the expectation the mean dose in the BTVs would be similar to theprescribed dose escalation. The second treatment plan added a simultaneousintegrated boost of 70 Gy to the BTV70 structure. Coverage inside BTV70 re-quired that V66.5 Gy ? 98% (coverage of 95% of the prescribed dose boost) anda maximum dose of 72.5 Gy. For the third treatment plan, constraints for BTV65(V61.8 Gy ? 98%), BTV70 (V66.5 Gy ? 98%), and BTV75 (V71.3 Gy ? 98% and amaximum dose of 77.5 Gy). For the final treatment plan, the dose in the BTVs88(a)(b)Figure 4.3. The field arrangement for volumetric modulated arc therapy is shown for(a) axial and (b) three-dimensional views. The planning target volume is shown in red.89Table 4.1. Dose volume histogram constraints used for volumetric modulated arc ther-apy.Structure Type ConstraintNo overlapping volumesPTV Target V95% ? 98% and Dmax = 107%Optic Chiam and Nerve OAR V54 Gy ? 1%Brainstem OAR Dmax = 60 GyRetina OAR Dmax = 45 GyAnterior Chamber OAR Dmax = 10 GyOverlapping target and OARsPTV54 Target V54 Gy ? 1% and V51.3 Gy ? 98%PTV60 Target V95% ? 98% and Dmax = 107%PTVb Target V95% ? 98% and Dmax = 60 GyOptic_opti OAR V54 Gy ? 1%Brainstem_opti OAR Dmax = 60 GyRetina OAR Dmax = 45 GyAnterior Chamber OAR Dmax = 10 GyDose escalation volumes for dose paintingBTV62.5 BTV V59.4 Gy ? 98%BTV65 BTV V61.8 Gy ? 98%BTV67.5 BTV V64.1 Gy ? 98%BTV70 BTV V66.5 Gy ? 98%BTV72.5 BTV V68.9 Gy ? 98%BTV75 BTV V71.3 Gy ? 98%BTV77.5 BTV V73.6 Gy ? 98% and Dmax = 80 GyAbbreviations: BTV = biological target volume; max = maximum; OAR = organ atrisk; PTV = planning target volume.was constrained so that at least 98% of each volume was covered by 95% of itsprescribed dose boost and that the maximum dose in BTV77.5 was 80 Gy (Ta-ble 4.1). For all dose escalation schemes, it was also attempted to reduce themean dose to the OARs. These progressive dose painting steps are illustratedin Figure 4.1(b)?(d).904.2.4 Evaluation of Treatment PlansThe number of monitor units (MUs) required to deliver a 2-Gy fraction wasrecorded for each treatment plan. In addition, a dosimetric comparison of theVMAT plans without dose escalation and with all dose painting steps was per-formed using cumulative DVHs and dosimetric parameters for the PTV, BTVs,and OARs, including mean and maximum doses. In addition, conformity of thePTV was quantified using the conformity index (CI), which is defined as:CI =95% isodose surface volumePTV volume. (4.3)The ideal value for the CI in this study is 0.98. Homogeneity of the dose in thePTV was quantified using the homogeneity index (HI), which is defined as:HI =Max. % dose in PTV95%. (4.4)For plans with dose escalation, dose homogeneity was calculated using the por-tion of the PTV excluding the BTVs.In addition, the equivalent uniform dose (EUD) (Eq. 1.38) was calculatedfor each OAR. The EUD was calculated using the generalized EUD (74):EUD =? N?i=1vi NTDai?1/a(4.5)where a = 1/n. Values for the volume dependence parameter n for each OARwere obtained from the Quantitative Analysis of Normal Tissue Effects in theClinic review papers (159?161) and those reported by Burman et al. (70). Thenormalized total dose (NTD) (Eq. 1.36) was calculated with ?/? = 2 Gy for allOARs (158, 162).91Table 4.2. Characteristics of patients planned for volumetric modulated arc therapywith dose painting.No. Age Histology PET Image Volume (cm3)PTV BTV62.51 61 Glioblastoma Preop 343 20.22 80 Glioblastoma Preop 152 6.83 51 Glioblastoma Postop 329 5.74 60 Glioblastoma Postop 439 1.45 31 Anaplastic astrocytoma Postop 320 3.2Abbreviations: BTV = biological target volume; PET = positron emission tomography;Postop: postoperative; Preop: preoperative; PTV = planning target volume.Statistical comparisons of dose-volume metrics and EUDs were performedusing MATLAB (version 7.0.0.19920; The MathWorks, Inc., Natick, MA) usingtwo-sided paired t-tests (?= 0.05).4.3 ResultsPatient characteristics are shown in Table 4.2. Patient 2 had a PTV that over-lapped with the brainstem and optic nerve. The target volumes did not overlapwith OARs for all other patients. The mean PTV volume was 317 cm3 (range,152?439 cm3). The mean dose escalation volumes were 7.4 cm3 (1.4?20.3cm3) for BTV62.5, 5.7 cm3 (1.1?17.2 cm3) for BTV65, 4.5 cm3 (0.9?14.6 cm3)for BTV67.5, 3.4 cm3 (0.7?11.8 cm3) for BTV70, 2.4 cm3 (0.5?9.0 cm3) forBTV72.5, 1.3 cm3 (0.2?5.5 cm3) for BTV75, and 0.2 cm3 (0.0?0.8 cm3) forBTV77.5.It was possible to produce dose painting plans for all cases without sacrific-ing dose conformity within the PTV (Figure 4.4). Examples of the distributionof isodose lines for VMAT plans without dose escalation and with dose paint-92Figure 4.4. A dose volume histogram for the planning target volume (PTV) withoutdose escalation, and the PTV (excluding the dose escalation region) and biologicaltarget volumes (BTVs) with dose painting for a sample patient (Patient 1).ing are shown in Figure 4.5. Dose distributions are illustrated in Figure 4.6.The mean number of MUs to treat a 2-Gy fraction was 380 (range, 353?405)for VMAT plans without dose escalation and 430 (401?472) for VMAT planswith dose painting. This difference was significant (p = 0.01). The conformityof the 95% isodose surface between plans was similar. The mean CI was 1.16(1.05?1.48) without dose escalation and 1.20 (1.05?1.63) with dose painting(p = 0.32). The HI inside the PTV was 1.12 (1.12?1.14) without dose escala-tion and 1.30 (1.25?1.33) with dose painting (p < 0.001). This was expectedsince dose escalation by design will increase the HI.The dosimetric comparison of OARs for plans with and without dose paint-ing is shown in Table 4.3. For all cases, OAR mean and maximum doses werenot significantly different between VMAT plans without dose escalation andwith dose painting. In addition, dose painting VMAT plans produced EUDs thatwere similar to those for plans produced with no dose escalation in all cases93(a) (b)(c) (d)Figure 4.5. Isodose lines are shown for a volumetric modulated arc therapy (VMAT)plan without dose escalation on (a) computed tomography (CT) and (b) 18F-FDOPApositron emission tomography (PET), and a VMAT plan with dose painting on (c) CTand (d) 18F-FDOPA PET. The planning target volumes is shown in red and the BTV70structure is shown in blue.94(a) (b)Figure 4.6. The dose distribution from the case in Figure 4.5 is shown for volumetricmodulated arc therapy plans (a) without dose escalation and (b) with dose painting.The planning target volume is shown in red and the BTV70 structure is shown in blue.(Table 4.4). The DVHs for the brainstem, optic nerve and chiasm, retinas, andanterior chambers are shown for most complex case (Patient 2) with overlap-ping PTV and OARs are shown in Figures 4.7?4.9. Even in this complex case,similar DVHs were obtained for all OARs for plans without dose escalation andwith dose painting.4.4 DiscussionThis study has shown that dose painting with 18F-FDOPA PET contours waspossible with VMAT planning of high-grade gliomas without increasing the dosedelivered to critical structures. This is the first study to show the feasibility ofVMAT dose painting for gliomas. Dosimetric data of glioma VMAT in this studyis similar to those reported by others (49, 163). However, this study did notuse a clinically-implemented protocol and normal brain was not used as an95Table 4.3. Dosimetric comparison of organs at risk for volumetric modulated arc ther-apy plans with and without dose painting.Structure Original* Dose Painting* p-valueMean Dose (Gy)Brainstem 25.0 (11.6?42.5) 24.3 (10.7?43.1) 0.30Left retina 4.7 (2.1?7.2) 3.8 (1.9?6.2) 0.10Right retina 4.1 (2.2?7.6) 3.9 (1.5?7.1) 0.52Left anterior chamber 3.4 (1.8?4.5) 2.9 (1.7?4.1) 0.14Right anterior chamber 3.2 (1.8?5.2) 3.0 (1.3?4.8) 0.33Optic nerve & chiasm 19.3 (9.3?34.7) 18.6 (8.9?34.0) 0.10Maximum Dose (Gy)Brainstem 54.6 (45.5?60.1) 52.1 (39.2?59.9) 0.11Left retina 9.6 (5.6?15.1) 7.6 (4.8?11.8) 0.05Right retina 8.5 (3.5?16.7) 8.2 (2.2?16.7) 0.48Left anterior chamber 6.1 (3.8?8.1) 5.1 (3.3?6.9) 0.18Right anterior chamber 5.5 (3.2?7.8) 5.3 (2.3?8.1) 0.58Optic nerve & chiasm 35.1 (16.3?53.9) 34.4 (14.5?53.8) 0.33*Mean values are shown, with the range in parentheses.Table 4.4. Comparison of equivalent uniform doses, in Gy, of organs at risk for volu-metric modulated arc therapy plans with and without dose painting.Structure n* Original** Dose Painting** p-valueBrainstem 0.16 33.8 (18.7?48.9) 32.2 (15.8?48.7) 0.08Left retina 0.2 3.1 (1.3?4.9) 2.4 (1.2?4.0) 0.06Right retina 0.2 2.8 (1.2?5.5) 2.7 (0.8?5.3) 0.46Left anterior chamber 0.25 1.9 (1.0?2.7) 1.6 (0.9?2.3) 0.11Right anterior chamber 0.25 1.8 (1.0?2.8) 1.7 (0.7?2.6) 0.23Optic nerve & chiasm 0.3 19.7 (6.1?37.8) 19.3 (5.8?37.6) 0.07*Volume parameter used for equivalent uniform dose calculation.**Mean values are shown, with the range in parentheses.96(a)(b)Figure 4.7. Dose volume histograms for the (a) brainstem and (b) optic chiasm andnerves for volumetric modulated arc therapy plans without dose escalation (gray line)and with dose painting (black line) for a sample patient (Patient 2).97(a)(b)Figure 4.8. Dose volume histograms for the (a) left and (b) right retinas for volumetricmodulated arc therapy plans without dose escalation (gray line) and with dose painting(black line) for a sample patient (Patient 2).98(a)(b)Figure 4.9. Dose volume histograms for the (a) left and (b) right anterior chambers forvolumetric modulated arc therapy plans without dose escalation (gray line) and withdose painting (black line) for a sample patient (Patient 2).99OAR planning constraint in this study. This may result in larger normal braindoses as compared to plans obtained with IMRT (49). Further reduction innormal tissue doses may be possible if normal brain was included in the planoptimization process. This study also only used a single treatment arc. It hasbeen reported that the use of noncoplanar VMAT reduced dose to the lowercontralateral temporal lobe dose for patients with fronto-temporal high-gradeglioma (163). The use of noncoplanar VMAT for dose painting may potentiallyreduce normal brain doses while allowing dose escalation to abnormalities on18F-FDOPA PET.The unique biologically-guided choice of dose painting thresholds in thisstudy allows for dose painting with multiple contours to be performed withclinically-available treatment planning software. However, this method reliedon the choice of radiobiological parameters for the assignment of dose paintingthresholds. Qi et al. (64) reported the radiosensitivity parameters for gliomasfrom clinical outcomes data to be ? = 0.06 ? 0.05 Gy?1 and ?/? = 10.0 ?15.1 Gy. From these values, SF2 = 0.87, suggesting strongly radioresistant cells.This value is similar to the SF2 values used in this study. For example, SF2 =0.94 was used to obtain the curve in Figure 4.1(a) with the assumption ?/? =10 Gy. The choice of radiosensitivity parameters may be crucial to potentialimprovement of patient outcomes. Furthermore, the functional relationshipbetween 18F-FDOPA uptake and tumour cell density is not known in relative orabsolute terms. The impact of image acquisition parameters and reconstructionalgorithms on dose painting also needs to be established. Research is neededto provide this data.Tumour-to-background ratios are often used to define simultaneous inte-grated boost volumes. Piroth et al. (153) used the 18F-FET PET tumour-to-brainratio of ? 1.6 since it was assumed that this value would to give a 5-mm safety100margin to ratios of 2.0?2.6 which have been reported to identify malignantcells (164?166). But use of this threshold to define a simultaneous integratedboost of 72 Gy in 30 fractions did not lead to a survival benefit (95). Using three-dimensional MRSI, Ken et al. (154) defined a simultaneous integrated boost toa GTV defined by a choline-to-N-acetyl-aspartate ratio > 2. 11C-MET-definedGTVs have also been defined using a tumour-to-brain ratio of 2 (138). Recentlyreported biopsy data suggested that an optimal 18F-FDOPA PET tumour-to-brainratio of> 2 as a threshold that identifies high-grade disease for newly diagnosedand recurrent glioma (79). This suggests a useful threshold for dose painting(i.e., I0 in this study), although more clarity is needed before clinical implemen-tation. In addition, an appropriate margin would need to be added to the BTVsfor clinical implementation of the biologically-guided technique presented here.The treatment technique would need to be chosen (i.e., small set-up margin) toensure that the dose distribution as closely as possible matches the biologicalinformation from PET.Planning studies which report dose painting by numbers for gliomas haveused linear models to map image intensity or standardized uptake value (SUV)to dose (96, 157). The model presented in this study can also be used to addradiobiological-guidance for dose painting by numbers. Other biological modelscan also be employed. For example, normal tissue complication probability(NTCP) and tumour control probability (TCP) have been investigated for dosepainting subvolumes for prostate treatments (77). For head and neck cancer,TCP-based models have proposed (148, 167).Multimodality approaches have shown that different imaging techniquescan show different dose escalation volumes. In a recent study by Houwelinget al (168), 18F-FDG PET and apparent diffusion coefficient (ADC)-based tar-get volumes for head and neck cancer patients showed minimal overlap. This101has important implications dose painting treatment techniques since there waspoor dose coverage inside ADC-based target volumes for dose painting plansobtained using a PET PTV. For high-grade gliomas, a recent study found min-imal overlaps of the ADC value and the maximum 18F-FDOPA PET SUV ratio(128). Multimodality approaches for dose painting are needed to fully assessdose painting treatment plans and compare these imaging parameters with pat-terns of failure.Despite the small number of patients (n = 5), this study has demonstrateda unique, biologically-guided and physically-based VMAT planning method forhigh-grade glioma treatments. However, the dose escalation volume for mostpatients was small since 18F-FDOPA PET was used for neuronavigation duringsurgery. This study did not include patients with tumours near the basal gan-glia since there is intense normal uptake of 18F-FDOPA in these structures (37).Dose painting techniques for tumours near these structures may be more dif-ficult. As noted in Chapter 2, another difficulty with using 18F-FDOPA PET isthat postsurgical changes around the resection cavity can exhibit tracer uptakebecause of high levels of amino acid transport by activated macrophages or 18F-FDOPA leakage due to disruption of the blood-brain barrier (97). This study an-ticipated this challenge by having a radiation oncologist with experience with18F-FDOPA PET images delineate volumes of interest for dose escalation. Inaddition, care must be taken with any automatic segmentation methods thatmay include regions of normal 18F-FDOPA uptake in basal ganglia or in doseescalation volumes.102Chapter 5Conclusions and Future Work5.1 ConclusionsThis project has demonstrated that biological information from functional imag-ing techniques, such as positron emission tomography (PET) and diffusion ten-sor imaging (DTI), can potentially be used to improve localization of malig-nant tissue and for biologically-based treatment planning. First, it was demon-strated that high-grade glioma radiation therapy target volumes obtained with3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine (18F-FDOPA) PET had similar in-terobserver agreement than volumes obtained with magnetic resonance imag-ing (MRI) with gadolinium contrast enhancement. Although consensus targetvolumes obtained using 18F-FDOPA PET were significantly larger than volumesobtained using MRI, treatment planning using the PET-based volumes may nothave yielded better treatment outcomes since all but one central recurrence ex-tended beyond the PET gross tumour volume (GTV) and most were containedby a 2-cm margin on the MRI GTV.Radiation therapy target localization for high-grade gliomas was potentiallypossible using DTI. Fractional anisotropy (FA) were significantly smaller andmean diffusivity (MD) was significantly larger in the GTV as compared to con-tralateral normal brain tissue. FA and MD values, as well as FA and MD inter-hemispheric differences approached those of normal brain tissue as the distance103from the GTV increased, consistent with the expectation of a gradual and de-creasing presence of tumour cells. Interhemispheric difference images obtainedfrom DTI provide images that may allow easier interpretation of DTI data, ascompared to images of FA and MD. Further research is warranted to determineif treatment planning using interhemispheric difference images images can beused to improve target delineation or be used for biologically-based treatmentplanning, potentially improving treatment outcomes.Lastly, volumetric modulated arc therapy (VMAT) planning using dose paint-ing for high-grade gliomas was achieved using biological target volumes (BTVs)obtained from biologically-guided thresholds of 18F-FDOPA PET images. Therewas no significant difference in the dose delivered to critical normal structuresbetween plans without dose escalation and plans with dose painting. Whilefurther research is needed to clarify radiosensitivity parameters and thresholdsfor dose escalation, dose painting using 18F-FDOPA can be implemented usingclinically-available VMAT optimization techniques.5.2 Future Work and Other ApplicationsIn this thesis, the radiation therapy applications of 18F-FDOPA PET and DTI havebeen discussed. However, the addition of these imaging techniques to neuro-surgical resection planning may improve patient survival since the extent ofresection is associated with survival in patients with high-grade disease (169).Incorporation of 18F-FDOPA PET images may allow neurosurgeons to identifyhigher-grade and higher-density disease than that which could be identifiedfrom T1-weighted contrast-enhancement MRI (79, 94).18F-FDOPA PET has other applications as well, such as the detection of re-current glioma. While the diagnostic accuracies of the detection of recurrent104tumours with 18F-FDOPA PET and T1-weighted MRI are similar, 18F-FDOPA PETis more specific than MRI for recurrent glioma detection (170). Karunanithi etal. (171) have reported a multivariate analysis which shows that tumour sizeon MRI and the standardized uptake value (SUV) ratio of tumour to contralat-eral normal brain from 18F-FDOPA PET were independent predictors of patientsurvival for recurrent glioma. 18F-FDOPA PET parametric response maps (voxel-wise changes in 18F-FDOPA uptake in time) may also be a useful biomarker ofprogression free survival and overall survival for patients with recurrent gliomatreated with bevacizumab (172).The use of DTI for neurosurgical resection of glioma is well established(173, 174). In a prospective, randomized controlled trial, Wu et al. (175) re-ported that there was a significant survival benefit for patients with cerebralhigh-grade glioma with pyramidal tract involvement who underwent surgerywith DTI-based neuronavigation (21.2 months) compared to those operatedon with routine neuronavigation (14.0 months) (p = 0.048). Mathematicalmodeling of glioma growth has also been investigated using DTI to predictanisotropic pathways for glioma cell invasion and identify potential locationsfor recurrence (90, 98, 101, 176, 177).Many studies have also investigated the use of preoperative diffusion imag-ing to distinguish glioma from other pathologies, such as brain metastases, de-myelinating diseases, and radiation-induced injury (178?184). Axial and radialdiffusivities have been suggested as biomarkers to distinguish low-grade andhigh-grade gliomas (185). The decrease of FA in peritumoural white matter hasbeen reported to be significantly different for patients with glioma and menin-gioma (186). In another study, FA and MD values were significantly smallerin cerebral lymphoma than the values for glioblastoma (187). Despite theseefforts, the gold standard remains histological analysis (174).105Other imaging techniques may also be appropriate for radiation therapyplanning. For example, arterial spin labeling has been shown to improve thediagnostic accuracy of preoperative glioma grading (188). The high fractionalcerebral blood volume obtained from T2?-weighted dynamic susceptibility con-trast MRI, may identify regions that are radioresistant and thus benefit from ra-diation dose escalation (189). Pharmacokinetic parameters from T1-weighteddynamic contrast enhanced MRI have been suggested to differentiate tumourfrom radiation-induced injury and surrounding brain tissue (190, 191). Thecholine-to-N-acetyl-aspartate ratio obtained from magnetic resonance spectro-scopic imaging (MRSI) can identify abnormal metabolically active regions thatcan be used to define target volumes for primary and boost volumes (192, 193).Single photon emission computed tomography (SPECT)/computed tomography(CT) with 99mTc-glucoheptonate has been suggested as a low-cost alternative to18F-FDOPA PET for detection of recurrent glioma (194). Sodium concentrationsobtained from 23Na-MRI may potentially be used to differentiate between low-grade and high-grade glioma (195, 196).The dose painting technique proposed in this thesis may ultimately allowfor radiation treatment plans for high-grade gliomas to conform to the biologi-cal information obtained from 18F-FDOPA PET or DTI, as originally envisionedby the concept of multidimensional radiotherapy introduced by Ling et al. (75).While the feasibility of dose painting by contours (95, 153, 154) and by num-bers (96, 157) for glioma treatments has been investigated, there is a need formore research. Although dose escalation up to 84 Gy can be delivered withoutincreased incidence radionecrosis or normal tissue effects (143, 154), there isnot yet any evidence that dose escalation will improve patient outcomes (95).In this project, it has been demonstrated that 18F-FDOPA PET and DTI can poten-tially be used for biological-guidance in radiation therapy and that dose painting106can be achieved using commercially-available VMAT optimization without in-creased dose to normal tissues. However, there is the need for more researchbefore these techniques can be implemented clinically. The choice of imagethresholds used for dose painting, or the function used to prescribed dose fromimage intensities for dose painting by numbers, are crucial to any biologically-guided technique. While there is limited evidence for appropriate thresholds for18F-FDOPA PET (79), more research is needed to determine appropriate thresh-old for biologically-based radiation therapy of gliomas. In addition it must benoted that the threshold values may also depend on target-to-background inten-sity ratios, reconstruction algorithms, and the type of scanner used (197, 198).There is also a need for more clinical investigations to determine the role ofbiological guidance in the treatment of high-grade glioma. Currently, there is noconsistent data to provide a rationale for the use of heterogeneous, biologically-based treatment planning (74). Preclinical studies with animal models areneeded to provide this data. In addition, any biologically-guided treatmentplanning technique will require a robust acceptance and commissioning, qual-ity assurance, and treatment plan evaluation procedures, such as those outlinedin the report of Task Group 166 of the American Association of Physicists inMedicine (74). It is important that any biologically-based treatment plan beevaluated using established dose volume histogram (DVH) criteria. The reviewof three-dimensional dose distributions is also essential to ensure that qualityradiation therapy plans are obtained. Moreover, the effect of cold spots in theGTV may be underestimated for plans optimized by dose painting and care mustbe taken to ensure that hot spots within the planning target volume (PTV) arelocated within the GTV (74). Any validation of such techniques must ultimatelybe based on the assessment of three-dimensional patterns of failure followingtreatment (197).107This project has demonstrated the feasibility of biologically-guided radiationtherapy of high-grade gliomas through the use of 18F-FDOPA PET or interhemi-spheric difference images of DTI-based parameters. While much work muststill be done before these techniques can be introduced clinically, the use of 18F-FDOPA PET and DTI during radiation therapy planning may improve tumourlocalization and be used for VMAT dose painting, thereby potentially improvingsurvival for patients with high-grade glioma.108Bibliography1. Rizzo DC. Fundamentals of anatomy and physiology. 1st ed. Clifton Park(NY): Delmar Cengage Learning; 2000.2. Ostrom QT, Gittleman H, Farah P, Ondracek A, Chen Y, Wolinsky Y, et al.CBTRUS statistical report: primary brain and central nervous system tu-mors diagnosed in the United States in 2006-2010. Neuro Oncol. 2013;15Suppl 2:ii1?ii56.3. Louis DN, Ohgaki H, Wiestler OD, Cavenee WK, Burger PC, Jouvet A, et al.The 2007 WHO classification of tumours of the central nervous system.Acta Neuropathol. 2007;114(2):97?109.4. Daumas-Duport C, Scheithauer B, O?Fallon J, Kelly P. Grading of astrocy-tomas. A simple and reproducible method. Cancer. 1988;62(10):2152?65.5. Anderson E, Grant R, Lewis SC, Whittle IR. Randomized Phase III con-trolled trials of therapy in malignant glioma: where are we after 40 years?Br J Neurosurg. 2008;22(3):339?49.6. Smith JS, Jenkins RB. Genetic alterations in adult diffuse glioma:occurrence, significance, and prognostic implications. Front Biosci.2000;5:D213?D231.1097. Laws ER, Parney IF, Huang W, Anderson F, Morris AM, Asher A, et al. Sur-vival following surgery and prognostic factors for recently diagnosed ma-lignant glioma: data from the Glioma Outcomes Project. J Neurosurg.2003;99(3):467?73.8. Perrin RG, Bernstein M. Current treatment of malignant gliomas: anevidence-based review. Oncol Exch. 2004;3(3):28?32.9. MacDonald SM, Ahmad S, Kachris S, Vogds BJ, DeRouen M, GittlemanAE, et al. Intensity modulated radiation therapy versus three-dimensionalconformal radiation therapy for the treatment of high grade glioma: adosimetric comparison. J Appl Clin Med Phys. 2007;8(2):47?60.10. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJB,et al. Radiotherapy plus concomitant and adjuvant temozolomide forglioblastoma. N Engl J Med. 2005;352(10):987?96.11. Thornton AF, Sandler HM, Ten Haken RK, McShan DL, Fraass BA, La Vi-gne ML, et al. The clinical utility of magnetic resonance imaging in 3-dimensional treatment planning of brain neoplasms. Int J Radiat OncolBiol Phys. 1992;24(4):767?75.12. Podgorsak EB. External photon beams: Physical aspects. In: PodgorsakEB, editor. Radiation oncology physics: A handbook for teachers and stu-dents. Vienna: International Atomic Energy Agency; 2005. p. 161?218.13. Podgorsak EB. Treatment machines for external beam radiotherapy. In:Podgorsak EB, editor. Radiation oncology physics: A handbook for teach-ers and students. Vienna: International Atomic Energy Agency; 2005. p.123?60.11014. Sheikh-Bagheri D, Rogers DWO. Monte Carlo calculation of nine me-gavoltage photon beam spectra using the BEAM code. Med Phys.2002;29(3):391?402.15. Khan FM. The physics of radiation therapy. 3rd ed. Philadelphia: Lippin-cott Williams & Wilkins; 2003.16. Hubbell JH, Seltzer SM. Tables of X-ray mass attenuation coefficients andmass energy-absorption coefficients from 1 keV to 20 MeV for elements Z= 1 to 92 and 48 additional substances of dosimetric interest [Internet];1996 [cited 19 Jan 2013]. Available from: http://www.nist.gov/pml/data/xraycoef/.17. Podgorsak EB. Basic radiation physics. In: Podgorsak EB, editor. Radi-ation oncology physics: A handbook for teachers and students. Vienna:International Atomic Energy Agency; 2005. p. 1?44.18. Attix FH. Introduction to radiological physics and radiation dosimetry.New York: Wiley; 1986.19. Report 85: Fundamental quantities and units for ionizing radiation. JICRU. 2011;11(1):1?31.20. Bloch F. Nuclear Induction. Phys Rev. 1946;70(7-8):460?74.21. Bloch F, Hansen W, Packard M. Nuclear Induction. Phys Rev. 1946;69(3-4):127.22. Hahn E. Spin Echoes. Phys Rev. 1950;80(4):580?94.23. Hajnal JV, De Coene B, Lewis PD, Baudouin CJ, Cowan FM, PennockJM, et al. High signal regions in normal white matter shown by heav-111ily T2-weighted CSF nulled IR sequences. J Comput Assist Tomogr.1992;16(4):506?13.24. Hajnal JV, Bryant DJ, Kasuboski L, Pattany PM, De Coene B, Lewis PD,et al. Use of fluid attenuated inversion recovery (FLAIR) pulse sequencesin MRI of the brain. J Comput Assist Tomogr. 1992;16(6):841?4.25. Stack JP, Antoun NM, Jenkins JP, Metcalfe R, Isherwood I. Gadolinium-DTPA as a contrast agent in magnetic resonance imaging of the brain.Neuroradiology. 1988;30(2):145?54.26. Kelly PJ, Daumas-Duport C, Kispert DB, Kall Ba, Scheithauer BW, Illig JJ.Imaging-based stereotaxic serial biopsies in untreated intracranial glialneoplasms. J Neurosurg. 1987;66(6):865?74.27. Gross MW, Weber WA, Feldmann HJ, Bartenstein P, Schwaiger M, MollsM. The value of F-18-fluorodeoxyglucose PET for the 3-D radiation treat-ment planning of malignant gliomas. Int J Radiat Oncol Biol Phys.1998;41(5):989?95.28. Podgorsak EB, Podgorsak MB. Special procedures and techniques in ra-diotherapy. In: Podgorsak EB, editor. Radiation oncology physics: Ahandbook for teachers and students. Vienna: International Atomic EnergyAgency; 2005. p. 505?48.29. Heiss WD, Wienhard K, Wagner R, Lanfermann H, Thiel A, Herholz K,et al. F-Dopa as an amino acid tracer to detect brain tumors. J Nucl Med.1996;37(7):1180?2.30. Stadlbauer A, Prante O, Nimsky C, Salomonowitz E, Buchfelder M, Kuw-ert T, et al. Metabolic imaging of cerebral gliomas: spatial correlation112of changes in O-(2-18F-fluoroethyl)-L-tyrosine PET and proton magneticresonance spectroscopic imaging. J Nucl Med. 2008;49(5):721?9.31. Miyagawa T, Oku T, Uehara H, Desai R, Beattie B, Tjuvajev J, et al. ?Facili-tated? amino acid transport is upregulated in brain tumors. J Cereb BloodFlow Metab. 1998;18(5):500?9.32. Pirotte B, Goldman S, Massager N, David P, Wikler D, Vandesteene A, et al.Comparison of 18F-FDG and 11C-methionine for PET-guided stereotacticbrain biopsy of gliomas. J Nucl Med. 2004;45(8):1293?8.33. Weber DC, Zilli T, Buchegger F, Casanova N, Haller G, Rouzaud M, et al.[(18)F]Fluoroethyltyrosine- positron emission tomography-guided radio-therapy for high-grade glioma. Radiat Oncol. 2008;3(18):44.34. Niyazi M, Geisler J, Siefert A, Schwarz SB, Ganswindt U, Garny S, et al.FET-PET for malignant glioma treatment planning. Radiother Oncol.2011;99(1):44?8.35. National Center for Biotechnology Information. PubChem CompoundDatabase [Internet]; 2013 [cited 16 Oct 2013]. CID:56494. Avail-able from: http://pubchem.ncbi.nlm.nih.gov/summary/summary.cgi?cid=56494.36. Nanni C, Fanti S, Rubello D. 18F-DOPA PET and PET/CT. J Nucl Med.2007;48(10):1577?9.37. Garnett ES, Firnau G, Nahmias C. Dopamine visualized in the basal gan-glia of living man. Nature. 1983;305(5930):137?8.38. Becherer A, Karanikas G, Szab? M, Zettinig G, Asenbaum S, Marosi C, et al.Brain tumour imaging with PET: a comparison between [18F]fluorodopa113and [11C]methionine. Eur J Nucl Med Mol Imaging. 2003;30(11):1561?7.39. Agid Y. Parkinson?s disease: pathophysiology. Lancet. 1991;337(8753):1321?4.40. Chen W, Silverman DHS, Delaloye S, Czernin J, Kamdar N, Pope W, et al.18F-FDOPA PET imaging of brain tumors: comparison study with 18F-FDGPET and evaluation of diagnostic accuracy. J Nucl Med. 2006;47(6):904?11.41. Ledezma CJ, Chen W, Sai V, Freitas B, Cloughesy T, Czernin J, et al. 18F-FDOPA PET/MRI fusion in patients with primary/recurrent gliomas: ini-tial experience. Eur J Radiol. 2009;71(2):242?8.42. Alexander AL, Lee JE, Lazar M, Field AS. Diffusion tensor imaging of thebrain. Neurotherapeutics. 2007;4(3):316?29.43. Einstein A. [On the movement of small particles suspended in station-ary liquids required by the molecular-kinetic theory of heat]. Ann Phys.1905;322(8):549?60. German.44. Stejskal EO, Tanner JE. Spin diffusion measurements: spin echoesin the presence of a time-dependent field gradient. J Chem Phys.1965;42(1):288?92.45. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy andimaging. Biophys J. 1994;66(1):259?67.46. Basser PJ, Pierpaoli C. Microstructural and physiological features of tis-sues elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B.1996;111(3):209?19.11447. International Commission on Radiation Units and Measurements. Pre-scribing, recording and reporting photon beam therapy. Bethesda (MD):International Commission on Radiation Units and Measurements; 1993.Report No.: 50.48. International Commission on Radiation Units and Measurements. Pre-scribing, recording and reporting photon beam therapy (supplement toICRU report 50). Bethesda (MD): International Commission on RadiationUnits and Measurements; 1999. Report No.: 62.49. Shaffer R, Nichol AM, Vollans E, Fong M, Nakano S, Moiseenko V, et al.A comparison of volumetric modulated arc therapy and conventionalintensity-modulated radiotherapy for frontal and temporal high-gradegliomas. Int J Radiat Oncol Biol Phys. 2010;76(4):1177?84.50. Radiation Therapy Oncology Group. RTOG 0825: Phase III double-blindplacebo-controlled trial of conventional concurrent chemoradiation andadjuvant temozolomide plus bevacizumab versus conventional concur-rent chemoradiation and adjuvant temozolomide in patients with newlydiagnosed glioblastoma [Internet]; 2011 [cited 11 Nov 2013]. Avail-able from: http://www.rtog.org/ClinicalTrials/ProtocolTable/StudyDetails.aspx?action=openFile&FileID=4664.51. Champ CE, Siglin J, Mishra MV, Shen X, Werner-Wasik M, Andrews DW,et al. Evaluating changes in radiation treatment volumes from post-operative to same-day planning MRI in High-grade gliomas. Radiat Oncol.2012;7(1):220.11552. McDonald MW, Shu HKG, Curran WJ, Crocker IR. Pattern of failure afterlimited margin radiotherapy and temozolomide for glioblastoma. Int JRadiat Oncol Biol Phys. 2011;79(1):130?6.53. Chang EL, Akyurek S, Avalos T, Rebueno N, Spicer C, Garcia J, et al.Evaluation of peritumoral edema in the delineation of radiotherapy clin-ical target volumes for glioblastoma. Int J Radiat Oncol Biol Phys.2007;68(1):144?50.54. Narayana A, Yamada J, Berry S, Shah P, Hunt M, Gutin PH, et al. Intensity-modulated radiotherapy in high-grade gliomas: clinical and dosimetricresults. Int J Radiat Oncol Biol Phys. 2006;64(3):892?7.55. Bortfeld T. Optimized planning using physical objectives and constraints.Semin Radiat Oncol. 1999;9(1):20?34.56. Bortfeld TR, Kahler DL, Waldron TJ, Boyer AL. X-ray field compensationwith multileaf collimators. Int J Radiat Oncol Biol Phys. 1994;28(3):723?30.57. Mackie TR, Holmes T, Swerdloff S, Reckwerdt P, Deasy JO, Yang J, et al.Tomotherapy: a new concept for the delivery of dynamic conformal ra-diotherapy. Med Phys. 1993;20(6):1709?19.58. Yu CX. Intensity-modulated arc therapy with dynamic multileaf collima-tion: an alternative to tomotherapy. Phys Med Biol. 1995;40(9):1435?49.59. Yu CX, Li XA, Ma L, Chen D, Naqvi S, Shepard D, et al. Clinical implemen-tation of intensity-modulated arc therapy. Int J Radiat Oncol Biol Phys.2002;53(2):453?63.11660. Otto K. Volumetric modulated arc therapy: IMRT in a single gantry arc.Med Phys. 2008;35(1):310?7.61. Bergman AM, Bush K, Milette MP, Popescu IA, Otto K, Duzenli C. Directaperture optimization for IMRT using Monte Carlo generated beamlets.Med Phys. 2006;33(10):3666?79.62. Chin E. A four dimensional volumetric modulated arc therapy planningsystem for stereotactic body radiation therapy in lung cancers. Universityof British Columbia; 2013.63. Fowler JF. The linear-quadratic formula and progress in fractionated ra-diotherapy. Br J Radiol. 1989;62(740):679?94.64. Qi XS, Schultz CJ, Li XA. An estimation of radiobiologic parameters fromclinical outcomes for radiation treatment planning of brain tumor. Int JRadiat Oncol Biol Phys. 2006;64(5):1570?80.65. Lea DE, Catcheside DG. The mechanism of the induction by radiation ofchromosome aberrations in Tradescantia. J Genet. 1942;44(2-3):216?45.66. Puck TT, Marcus PI. Action of x-rays on mammalian cells. J Exp Med.1956;103(5):653?66.67. Park C, Papiez L, Zhang S, Story M, Timmerman RD. Universal sur-vival curve and single fraction equivalent dose: useful tools in under-standing potency of ablative radiotherapy. Int J Radiat Oncol Biol Phys.2008;70(3):847?52.68. Brenner DJ, Hlatky LR, Hahnfeldt PJ, Huang Y, Sachs RK. The linear-quadratic model and most other common radiobiological models result117in similar predictions of time-dose relationships. Radiat Res. 1998;150(1):83?91.69. Brenner DJ. The linear-quadratic model is an appropriate methodology fordetermining isoeffective doses at large doses per fraction. Semin RadiatOncol. 2008;18(4):234?9.70. Burman C, Kutcher GJ, Emami B, Goitein M. Fitting of normal tissuetolerance data to an analytic function. Int J Radiat Oncol Biol Phys.1991;21(1):123?35.71. Moiseenko V, Deasy JO, Van Dyk J. Radiobiological modeling for treatmentplanning. In: Van Dyk J, editor. The modern technology of radiation on-cology: A compendium for medical physicists and radiation oncologists.vol. 2. Madison (WI): Medical Physics Publishing; 2005. p. 185?220.72. Niemierko A. Reporting and analyzing dose distributions: A concept ofequivalent uniform dose. Med Phys. 1997;24(1):103?10.73. Wu Q, Mohan R, Niemierko A, Schmidt-Ullrich R. Optimization ofintensity-modulated radiotherapy plans based on the equivalent uniformdose. Int J Radiat Oncol Biol Phys. 2002;52(1):224?35.74. Li XA, Alber M, Deasy JO, Jackson A, Ken Jee KW, Marks LB, et al. Theuse and QA of biologically related models for treatment planning: shortreport of the TG-166 of the therapy physics committee of the AAPM. MedPhys. 2012;39(3):1386?409.75. Ling CC, Humm J, Larson S, Amols H, Fuks Z, Leibel S, et al. Towards mul-tidimensional radiotherapy (MD-CRT): biological imaging and biologicalconformality. Int J Radiat Oncol Biol Phys. 2000;47(3):551?60.11876. Galvin JM, De Neve W. Intensity modulating and other radiation therapydevices for dose painting. J Clin Oncol. 2007;25(8):924?30.77. Kim Y, Tom? Wa. Dose-painting IMRT optimization using biological pa-rameters. Acta Oncol. 2010;49(8):1374?84.78. Kracht LW, Miletic H, Busch S, Jacobs AH, Voges J, Hoevels M, et al. De-lineation of brain tumor extent with [11C]L-methionine positron emissiontomography: local comparison with stereotactic histopathology. Clin Can-cer Res. 2004;10(21):7163?70.79. Pafundi DH, Laack NN, Youland RS, Parney IF, Lowe VJ, Giannini C, et al.Biopsy validation of 18F-DOPA PET and biodistribution in gliomas forneurosurgical planning and radiotherapy target delineation: results of aprospective pilot study. Neuro Oncol. 2013;15(8):1058?67.80. Archip N, Jolesz FA, Warfield SK. A validation framework for brain tumorsegmentation. Acad Radiol. 2007;14(10):1242?51.81. Deeley MA, Chen A, Datteri R, Noble JH, Cmelak AJ, Donnelly EF, et al.Comparison of manual and automatic segmentation methods for brainstructures in the presence of space-occupying lesions: a multi-expertstudy. Phys Med Biol. 2011;56(14):4557?77.82. Warfield SK, Zou KH, Wells WM. Simultaneous truth and performancelevel estimation (STAPLE): an algorithm for the validation of image seg-mentation. IEEE Trans Med Imaging. 2004;23(7):903?21.83. Namavari M, Bishop A, Satyamurthy N, Bida G, Barrio JR. Regiose-lective radiofluorodestannylation with [18F]F2 and [18F]CH3COOF: a119high yield synthesis of 6-[18F]Fluoro-L-dopa. Int J Rad Appl Instrum A.1992;43(8):989?96.84. McGurk RJ, Bowsher J, Lee Ja, Das SK. Combining multiple FDG-PET radiotherapy target segmentation methods to reduce the effect ofvariable performance of individual segmentation methods. Med Phys.2013;40(4):042501.85. Jaccard P. [Comparative study of the distribution of flora in a portion ofthe Alps and the Jura]. Bull Soc Vaud Sci Nat. 1901;37(142):547?79.French.86. Jaccard P. The distribution of the flora in the alpine zone. New Phytol.1912;11(2):37?50.87. Lee SW, Fraass Ba, Marsh LH, Herbort K, Gebarski SS, Martel MK, et al.Patterns of failure following high-dose 3-D conformal radiotherapy forhigh-grade astrocytomas: a quantitative dosimetric study. Int J RadiatOncol Biol Phys. 1999;43(1):79?88.88. Cattaneo GM, Reni M, Rizzo G, Castellone P, Ceresoli GL, Cozzarini C,et al. Target delineation in post-operative radiotherapy of brain gliomas:interobserver variability and impact of image registration of MR(pre-operative) images on treatment planning CT scans. Radiother Oncol.2005;75(2):217?23.89. Jena R, Price SJ, Baker C, Jefferies SJ, Pickard JD, Gillard JH, et al. Dif-fusion tensor imaging: possible implications for radiotherapy treatmentplanning of patients with high-grade glioma. Clin Oncol (R Coll Radiol).2005;17(8):581?90.12090. Bondiau PY, Konukoglu E, Clatz O, Delingette H, Frenay M, Paquis P. Bio-computing: numerical simulation of glioblastoma growth and comparisonwith conventional irradiation margins. Phys Med. 2011;27(2):103?8.91. Chen W. Clinical applications of PET in brain tumors. J Nucl Med.2007;48(9):1468?81.92. Lee IH, Piert M, Gomez-Hassan D, Junck L, Rogers L, Hayman J, et al.Association of 11C-methionine PET uptake with site of failure after con-current temozolomide and radiation for primary glioblastoma multiforme.Int J Radiat Oncol Biol Phys. 2009;73(2):479?85.93. Weber DC, Casanova N, Zilli T, Buchegger F, Rouzaud M, Nouet P,et al. Recurrence pattern after [(18)F]fluoroethyltyrosine-positron emis-sion tomography-guided radiotherapy for high-grade glioma: a prospec-tive study. Radiother Oncol. 2009;93(3):586?92.94. Fueger BJ, Czernin J, Cloughesy T, Silverman DH, Geist CL, Walter Ma,et al. Correlation of 6-18F-fluoro-L-dopa PET uptake with proliferationand tumor grade in newly diagnosed and recurrent gliomas. J Nucl Med.2010;51(10):1532?8.95. Piroth MD, Pinkawa M, Holy R, Klotz J, Schaar S, Stoffels G, et al. Inte-grated boost IMRT with FET-PET-adapted local dose escalation in glioblas-tomas. Results of a prospective phase II study. Strahlenther Onkol.2012;188(4):334?9.96. Rickhey M, Mor?vek Z, Eilles C, Koelbl O, Bogner L. 18F-FET-PET-based dose painting by numbers with protons. Strahlenther Onkol.2010;186(6):320?6.12197. Walter F, Cloughesy T, Walter Ma, Lai A, Nghiemphu P, Wagle N, et al. Im-pact of 3,4-dihydroxy-6-18F-fluoro-L-phenylalanine PET/CT on managingpatients with brain tumors: the referring physician?s perspective. J NuclMed. 2012;53(3):393?8.98. Painter KJ, Hillen T. Mathematical modelling of glioma growth: the use ofDiffusion Tensor Imaging (DTI) data to predict the anisotropic pathwaysof cancer invasion. J Theor Biol. 2013;323:25?39.99. Giese A, Kluwe L, Laube B, Meissner H, Berens ME, Westphal M. Migrationof human glioma cells on myelin. Neurosurgery. 1996;38(4):755?64.100. Clatz O, Sermesant M, Bondiau PY, Delingette H, Warfield SK, MalandainG, et al. Realistic simulation of the 3-D growth of brain tumors in MRimages coupling diffusion with biomechanical deformation. IEEE TransMed Imaging. 2005;24(10):1334?46.101. Bondiau PY, Clatz O, Sermesant M, Marcy PY, Delingette H, Frenay M,et al. Biocomputing: numerical simulation of glioblastoma growth usingdiffusion tensor imaging. Phys Med Biol. 2008;53(4):879?93.102. Konukoglu E, Clatz O, Bondiau PY, Delingette H, Ayache N. Extrapolatingglioma invasion margin in brain magnetic resonance images: suggestingnew irradiation margins. Med Image Anal. 2010;14(2):111?25.103. Jbabdi S, Mandonnet E, Duffau H, Capelle L, Swanson KR, P?l?grini-IssacM, et al. Simulation of anisotropic growth of low-grade gliomas usingdiffusion tensor imaging. Magn Reson Med. 2005;54(3):616?24.122104. Kallenberg K, Goldmann T, Menke J, Strik H, Bock HC, Stockhammer F,et al. Glioma infiltration of the corpus callosum: early signs detected byDTI. J Neurooncol. 2013;112(2):217?22.105. Deng Z, Yan Y, Zhong D, Yang G, Tang W, L? F, et al. Quantitative anal-ysis of glioma cell invasion by diffusion tensor imaging. J Clin Neurosci.2010;17(12):1530?6.106. Beppu T, Inoue T, Shibata Y, Yamada N, Kurose A, Ogasawara K, et al. Frac-tional anisotropy value by diffusion tensor magnetic resonance imaging asa predictor of cell density and proliferation activity of glioblastomas. SurgNeurol. 2005;63(1):56?61.107. Kinoshita M, Hashimoto N, Goto T, Kagawa N, Kishima H, Izumoto S, et al.Fractional anisotropy and tumor cell density of the tumor core show posi-tive correlation in diffusion tensor magnetic resonance imaging of malig-nant brain tumors. Neuroimage. 2008;43(1):29?35.108. Stadlbauer A, Ganslandt O, Buslei R, Hammen T, Gruber S, Moser E,et al. Gliomas: histopathologic evaluation of changes in directionality andmagnitude of water diffusion at diffusion-tensor MR imaging. Radiology.2006;240(3):803?10.109. Lee EJ, Lee SK, Agid R, Bae JM, Keller A, Terbrugge K. Preoperative grad-ing of presumptive low-grade astrocytomas on MR imaging: diagnosticvalue of minimum apparent diffusion coefficient. AJNR Am J Neurora-diol. 2008;29(10):1872?7.123110. Aubert-Broche B, Grova C, Jannin P, Buvat I, Benali H, Gibaud B. Detectionof inter-hemispheric asymmetries of brain perfusion in SPECT. Phys MedBiol. 2003;48(11):1505?17.111. Jenkinson M, Beckmann CF, Behrens TEJ, Woolrich MW, Smith SM. FSL.Neuroimage. 2012;62(2):782?90.112. Woolrich MW, Jbabdi S, Patenaude B, Chappell M, Makni S, Behrens T,et al. Bayesian analysis of neuroimaging data in FSL. Neuroimage.2009;45(1 Suppl):S173?S186.113. Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TEJ,Johansen-Berg H, et al. Advances in functional and structural MR im-age analysis and implementation as FSL. Neuroimage. 2004;23 Suppl1:S208?S219.114. Smith SM. Fast robust automated brain extraction. Hum Brain Mapp.2002;17(3):143?55.115. Behrens TEJ, Woolrich MW, Jenkinson M, Johansen-Berg H, Nunes RG,Clare S, et al. Characterization and propagation of uncertainty indiffusion-weighted MR imaging. Magn Reson Med. 2003;50(5):1077?88.116. Behrens TEJ, Johansen-Berg H, Woolrich MW, Smith SM, Wheeler-Kingshott CAM, Boulby PA, et al. Non-invasive mapping of connectionsbetween human thalamus and cortex using diffusion imaging. Nat Neu-rosci. 2003;6(7):750?7.117. Behrens TEJ, Berg HJ, Jbabdi S, Rushworth MFS, Woolrich MW. Proba-bilistic diffusion tractography with multiple fibre orientations: What canwe gain? Neuroimage. 2007;34(1):144?55.124118. Johansen-Berg H, Behrens TEJ, Robson MD, Drobnjak I, Rushworth MFS,Brady JM, et al. Changes in connectivity profiles define functionally dis-tinct regions in human medial frontal cortex. Proc Natl Acad Sci U S A.2004;101(36):13335?40.119. Jbabdi S, Sotiropoulos SN, Savio AM, Gra?a M, Behrens TEJ. Model-basedanalysis of multishell diffusion MR data for tractography: how to get overfitting problems. Magn Reson Med. 2012;68(6):1846?55.120. Jenkinson M, Smith S. A global optimisation method for robust affineregistration of brain images. Med Image Anal. 2001;5(2):143?56.121. Jenkinson M, Bannister P, Brady M, Smith S. Improved optimization forthe robust and accurate linear registration and motion correction of brainimages. Neuroimage. 2002;17(2):825?41.122. Grabner G, Janke AL, Budge MM, Smith D, Pruessner J, Collins DL. Sym-metric atlasing and model based segmentation: an application to thehippocampus in older adults. Med Image Comput Comput Assist Interv.2006;9(Pt 2):58?66.123. Andersson JLR, Jenkinson M, Smith S. Non-linear registration aka spatialnormalisation. Oxford (UK): Oxford Centre for Functional MRI of theBrain; 2007. Report No.: TR07JA2.124. Andersson JLR, Jenkinson M, Smith S. Non-linear optimisation. Oxford(UK): Oxford Centre for Functional MRI of the Brain; 2007. Report No.:TR07JA1.125125. Rueckert D, Sonoda LI, Hayes C, Hill DL, Leach MO, Hawkes DJ. Non-rigid registration using free-form deformations: application to breast MRimages. IEEE Trans Med Imaging. 1999;18(8):712?21.126. Oxford Centre for Functional MRI of the Brain. FMRIB58_FA [Internet];2012 [cited 22 Nov 2013]. Available from: http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FMRIB58_FA.127. Stadlbauer A, Hammen T, Buchfelder M, Bachmair J, D?rfler A, NimskyC, et al. Differences in metabolism of fiber tract alterations in gliomas:a combined fiber density mapping and magnetic resonance spectroscopicimaging study. Neurosurgery. 2012;71(2):454?63.128. Rose S, Fay M, Thomas P, Bourgeat P, Dowson N, Salvado O, et al. Corre-lation of MRI-derived apparent diffusion coefficients in newly diagnosedgliomas with [18F]-fluoro-L-dopa PET: what are we really measuring withminimum ADC? AJNR Am J Neuroradiol. 2013;34(4):758?64.129. Stadlbauer A, Buchfelder M, Salomonowitz E, Ganslandt O. Fiber den-sity mapping of gliomas: histopathologic evaluation of a diffusion-tensorimaging data processing method. Radiology. 2010;257(3):846?53.130. Chan MF, Schupak K, Burman C, Chui CS, Ling CC. Comparison ofintensity-modulated radiotherapy with three-dimensional conformal ra-diation therapy planning for glioblastoma multiforme. Med Dosim.2003;28(4):261?5.131. Hermanto U, Frija EK, Lii MJ, Chang EL, Mahajan A, Woo SY. Intensity-modulated radiotherapy (IMRT) and conventional three-dimensional con-formal radiotherapy for high-grade gliomas: does IMRT increase the in-126tegral dose to normal brain? Int J Radiat Oncol Biol Phys. 2007;67(4):1135?44.132. Thilmann C, Zabel A, Grosser KH, Hoess A, Wannenmacher M, Debus J.Intensity-modulated radiotherapy with an integrated boost to the macro-scopic tumor volume in the treatment of high-grade gliomas. Int J Cancer.2001;96(6):341?9.133. Floyd NS, Woo SY, Teh BS, Prado C, Mai WY, Trask T, et al. Hypofrac-tionated intensity-modulated radiotherapy for primary glioblastoma mul-tiforme. Int J Radiat Oncol Biol Phys. 2004;58(3):721?6.134. Sultanem K, Patrocinio H, Lambert C, Corns R, Leblanc R, Parker W, et al.The use of hypofractionated intensity-modulated irradiation in the treat-ment of glioblastoma multiforme: preliminary results of a prospectivetrial. Int J Radiat Oncol Biol Phys. 2004;58(1):247?52.135. Iuchi T, Hatano K, Narita Y, Kodama T, Yamaki T, Osato K. Hypofraction-ated high-dose irradiation for the treatment of malignant astrocytomasusing simultaneous integrated boost technique by IMRT. Int J Radiat On-col Biol Phys. 2006;64(5):1317?24.136. Panet-Raymond V, Souhami L, Roberge D, Kavan P, Shakibnia L, MuanzaT, et al. Accelerated hypofractionated intensity-modulated radiotherapywith concurrent and adjuvant temozolomide for patients with glioblas-toma multiforme: a safety and efficacy analysis. Int J Radiat Oncol BiolPhys. 2009;73(2):473?8.137. Nakamatsu K, Suzuki M, Nishimura Y, Kanamori S, Koike R, Shibata T,et al. Treatment outcomes and dose-volume histogram analysis of simul-127taneous integrated boost method for malignant gliomas using intensity-modulated radiotherapy. Int J Clin Oncol. 2008;13(1):48?53.138. Miwa K, Matsuo M, Shinoda J, Oka N, Kato T, Okumura A, et al. Simultane-ous integrated boost technique by helical tomotherapy for the treatment ofglioblastoma multiforme with 11C-methionine PET: report of three cases.J Neurooncol. 2008;87(3):333?9.139. Chan JL, Lee SW, Fraass BA, Normolle DP, Greenberg HS, Junck LR,et al. Survival and failure patterns of high-grade gliomas after three-dimensional conformal radiotherapy. J Clin Oncol. 2002;20(6):1635?42.140. Petrecca K, Guiot MC, Panet-Raymond V, Souhami L. Failure patternfollowing complete resection plus radiotherapy and temozolomide isat the resection margin in patients with glioblastoma. J Neurooncol.2013;111(1):19?23.141. Kosztyla R, Chan EK, Hsu F, Wilson D, Ma R, Cheung A, et al. High-gradeglioma radiation therapy target volumes and patterns of failure obtainedfrom magnetic resonance imaging and (18)F-FDOPA positron emission to-mography delineations from multiple observers. Int J Radiat Oncol BiolPhys. 2013;87(5):1100?6.142. Oppitz U, Maessen D, Zunterer H, Richter S, Flentje M. 3D-recurrence-patterns of glioblastomas after CT-planned postoperative irradiation. Ra-diother Oncol. 1999;53(1):53?7.143. Tsien C, Moughan J, Michalski JM, Gilbert MR, Purdy J, Simpson J, et al.Phase I three-dimensional conformal radiation dose escalation study in128newly diagnosed glioblastoma: Radiation Therapy Oncology Group Trial98-03. Int J Radiat Oncol Biol Phys. 2009;73(3):699?708.144. Stewart RD, Li XA. BGRT: Biologically guided radiation therapy?Thefuture is fast approaching! Med Phys. 2007;34(10):3739?51.145. Walter F, la Foug?re C, Belka C, Niyazi M. Technical Issues of [(18)F]FET-PET Imaging for Radiation Therapy Planning in Malignant Glioma Patients- A Review. Front Oncol. 2012;2:130.146. Bentzen S. Theragnostic imaging for radiation oncology: dose-paintingby numbers. Lancet Oncol. 2005;6(2):112?7.147. Bentzen S, Gregoire V. Molecular imaging-based dose painting: anovel paradigm for radiation therapy prescription. Semin Radiat Oncol.2011;21(2):101?10.148. Thorwarth D, Eschmann SM, Paulsen F, Alber M. Hypoxia dose painting bynumbers: a planning study. Int J Radiat Oncol Biol Phys. 2007;68(1):291?300.149. Niyazi M, Bartenstein P, Belka C, Ganswindt U. Choline PET based dose-painting in prostate cancer?modelling of dose effects. Radiat Oncol.2010;5:23.150. W?rschmidt F, Petersen C, Wahl A, Dahle J, Kretschmer M. [18F]fluoroethylcholine-PET/CT imaging for radiation treatment planning ofrecurrent and primary prostate cancer with dose escalation to PET/CT-positive lymph nodes. Radiat Oncol. 2011;6:44.151. Chang JH, Lim Joon D, Lee ST, Gong SJ, Anderson NJ, Scott AM, et al.Intensity modulated radiation therapy dose painting for localized prostate129cancer using ??C-choline positron emission tomography scans. Int J RadiatOncol Biol Phys. 2012;83(5):e691?e696.152. Meijer G, Steenhuijsen J, Bal M, De Jaeger K, Schuring D, Theuws J. Dosepainting by contours versus dose painting by numbers for stage II/III lungcancer: practical implications of using a broad or sharp brush. RadiotherOncol. 2011;100(3):396?401.153. Piroth MD, Pinkawa M, Holy R, Stoffels G, Demirel C, Attieh C, et al.Integrated-boost IMRT or 3-D-CRT using FET-PET based auto-contouredtarget volume delineation for glioblastoma multiforme?a dosimetric com-parison. Radiat Oncol. 2009;4:57.154. Ken S, Vieillevigne L, Franceries X, Simon L, Supper C, Lotterie JA, et al.Integration method of 3D MR spectroscopy into treatment planning sys-tem for glioblastoma IMRT dose painting with integrated simultaneousboost. Radiat Oncol. 2013;8:1.155. Duprez F, De Neve W, De Gersem W, Coghe M, Madani I. Adaptive dosepainting by numbers for head-and-neck cancer. Int J Radiat Oncol BiolPhys. 2011;80(4):1045?55.156. Madani I, Duprez F, Boterberg T, Van de Wiele C, Bonte K, Deron P, et al.Maximum tolerated dose in a phase I trial on adaptive dose painting bynumbers for head and neck cancer. Radiother Oncol. 2011;101(3):351?5.157. Rickhey M, Koelbl O, Eilles C, Bogner L. A biologically adapted dose-escalation approach, demonstrated for 18F-FET-PET in brain tumors.Strahlenther Onkol. 2008;184(10):536?42.130158. Jones B, Sanghera P. Estimation of radiobiologic parameters and equiva-lent radiation dose of cytotoxic chemotherapy in malignant glioma. Int JRadiat Oncol Biol Phys. 2007;68(2):441?8.159. Mayo C, Yorke E, Merchant TE. Radiation associated brainstem injury. IntJ Radiat Oncol Biol Phys. 2010;76(3 Suppl):S36?S41.160. Mayo C, Martel MK, Marks LB, Flickinger J, Nam J, Kirkpatrick J. Radi-ation dose-volume effects of optic nerves and chiasm. Int J Radiat OncolBiol Phys. 2010;76(3 Suppl):S28?S35.161. Lawrence YR, Li XA, el Naqa I, Hahn CA, Marks LB, Merchant TE, et al.Radiation dose-volume effects in the brain. Int J Radiat Oncol Biol Phys.2010;76(3 Suppl):S20?S27.162. Wigg D. Applied radiobiology and bioeffect planning. Madison (WI):Medical Physics Publishing; 2001.163. Panet-Raymond V, Ansbacher W, Zavgorodni S, Bendorffe B, Nichol A,Truong PT, et al. Coplanar versus noncoplanar intensity-modulated ra-diation therapy (IMRT) and volumetric-modulated arc therapy (VMAT)treatment planning for fronto-temporal high-grade glioma. J Appl ClinMed Phys. 2012;13(4):3826.164. Pauleit D, Floeth F, Hamacher K, Riemenschneider MJ, Reifenberger G,M?ller HW, et al. O-(2-[18F]fluoroethyl)-L-tyrosine PET combined withMRI improves the diagnostic assessment of cerebral gliomas. Brain.2005;128(Pt 3):678?87.165. Rachinger W, Goetz C, P?pperl G, Gildehaus FJ, Kreth FW, Holtmannsp?t-ter M, et al. Positron emission tomography with O-(2-[18F]fluoroethyl)-l-131tyrosine versus magnetic resonance imaging in the diagnosis of recurrentgliomas. Neurosurgery. 2005;57(3):505?11.166. P?pperl G, G?tz C, Rachinger W, Gildehaus FJ, Tonn JC, Tatsch K. Valueof O-(2-[18F]fluoroethyl)- L-tyrosine PET for the diagnosis of recurrentglioma. Eur J Nucl Med Mol Imaging. 2004;31(11):1464?70.167. Thorwarth D, Soukup M, Alber M. Dose painting with IMPT, helicaltomotherapy and IMXT: a dosimetric comparison. Radiother Oncol.2008;86(1):30?4.168. Houweling AC, Wolf AL, Vogel WV, Hamming-Vrieze O, van Vliet-Vroegindeweij C, van de Kamer JB, et al. FDG-PET and diffusion-weightedMRI in head-and-neck cancer patients: Implications for dose painting. Ra-diother Oncol. 2013;106(2):250?4.169. Sanai N, Berger MS. Glioma extent of resection and its impact on patientoutcome. Neurosurgery. 2008;62(4):753?64.170. Karunanithi S, Sharma P, Kumar A, Khangembam BC, BandopadhyayaGP, Kumar R, et al. Comparative diagnostic accuracy of contrast-enhanced MRI and (18)F-FDOPA PET-CT in recurrent glioma. Eur Radiol.2013;23(9):2628?35.171. Karunanithi S, Sharma P, Kumar A, Gupta DK, Khangembam BC, Ballal S,et al. Can (18)F-FDOPA PET/CT predict survival in patients with suspectedrecurrent glioma? A prospective study. Eur J Radiol. 2014;83(1):219?25.172. Harris RJ, Cloughesy TF, Pope WB, Nghiemphu PL, Lai A, Zaw T, et al. 18F-FDOPA and 18F-FLT positron emission tomography parametric response132maps predict response in recurrent malignant gliomas treated with beva-cizumab. Neuro Oncol. 2012;14(8):1079?89.173. Coenen VA, Krings T, Mayfrank L, Polin RS, Reinges MH, Thron A, et al.Three-dimensional visualization of the pyramidal tract in a neuronaviga-tion system during brain tumor surgery: first experiences and technicalnote. Neurosurgery. 2001;49(1):86?92.174. Abdullah KG, Lubelski D, Nucifora PGP, Brem S. Use of diffusion tensorimaging in glioma resection. Neurosurg Focus. 2013;34(4):E1.175. Wu JS, Zhou LF, Tang WJ, Mao Y, Hu J, Song YY, et al. Clinical evalua-tion and follow-up outcome of diffusion tensor imaging-based functionalneuronavigation: a prospective, controlled study in patients with gliomasinvolving pyramidal tracts. Neurosurgery. 2007;61(5):935?48; discussion948?9.176. Trepanier PY, Fortin I, Lambert C, Lacroix F. A Monte Carlo basedformalism to identify potential locations at high risk of tumor recur-rence with a numerical model for glioblastoma multiforme. Med Phys.2012;39(11):6682?91.177. Gooya A, Pohl KM, Bilello M, Cirillo L, Biros G, Melhem ER, et al. GLISTR:glioma image segmentation and registration. IEEE Trans Med Imaging.2012;31(10):1941?54.178. Zeng QS, Li CF, Liu H, Zhen JH, Feng DC. Distinction between recur-rent glioma and radiation injury using magnetic resonance spectroscopyin combination with diffusion-weighted imaging. Int J Radiat Oncol BiolPhys. 2007;68(1):151?8.133179. Wang S, Kim S, Chawla S, Wolf RL, Zhang WG, O?Rourke DM, et al. Dif-ferentiation between glioblastomas and solitary brain metastases usingdiffusion tensor imaging. Neuroimage. 2009;44(3):653?60.180. Xu JL, Li YL, Lian JM, Dou SW, Yan FS, Wu H, et al. Distinction betweenpostoperative recurrent glioma and radiation injury using MR diffusiontensor imaging. Neuroradiology. 2010;52(12):1193?9.181. Giussani C, Poliakov A, Ferri RT, Plawner LL, Browd SR, Shaw DWW, et al.DTI fiber tracking to differentiate demyelinating diseases from diffusebrain stem glioma. Neuroimage. 2010;52(1):217?23.182. Byrnes TJD, Barrick TR, Bell BA, Clark CA. Diffusion tensor imaging dis-criminates between glioblastoma and cerebral metastases in vivo. NMRBiomed. 2011;24(1):54?60.183. Wang S, Kim S, Chawla S, Wolf RL, Knipp DE, Vossough A, et al.Differentiation between glioblastomas, solitary brain metastases, andprimary cerebral lymphomas using diffusion tensor and dynamic sus-ceptibility contrast-enhanced MR imaging. AJNR Am J Neuroradiol.2011;32(3):507?14.184. Hua C, Merchant TE, Gajjar A, Broniscer A, Zhang Y, Li Y, et al. Braintumor therapy-induced changes in normal-appearing brainstem measuredwith longitudinal diffusion tensor imaging. Int J Radiat Oncol Biol Phys.2012;82(5):2047?54.185. Chen F, Zhang X, Li M, Wang R, Wang HT, Zhu F, et al. Axial diffusivityand tensor shape as early markers to assess cerebral white matter damage134caused by brain tumors using quantitative diffusion tensor tractography.CNS Neurosci Ther. 2012;18(8):667?73.186. Provenzale JM, McGraw P, Mhatre P, Guo AC, Delong D. Peritumoralbrain regions in gliomas and meningiomas: investigation with isotropicdiffusion-weighted MR imaging and diffusion-tensor MR imaging. Radi-ology. 2004;232(2):451?60.187. Toh CH, Castillo M, Wong AMC, Wei KC, Wong HF, Ng SH, et al. Primarycerebral lymphoma and glioblastoma multiforme: differences in diffusioncharacteristics evaluated with diffusion tensor imaging. AJNR Am J Neu-roradiol. 2008;29(3):471?5.188. Kim MJ, Kim HS, Kim JH, Cho KG, Kim SY. Diagnostic accuracy and in-terobserver variability of pulsed arterial spin labeling for glioma grading.Acta Radiol. 2008;49(4):450?7.189. Cao Y, Tsien CI, Nagesh V, Junck L, Ten Haken R, Ross BD, et al. Survivalprediction in high-grade gliomas by MRI perfusion before and during earlystage of RT [corrected]. Int J Radiat Oncol Biol Phys. 2006;64(3):876?85.190. Bisdas S, Naegele T, Ritz R, Dimostheni A, Pfannenberg C, Reimold M,et al. Distinguishing recurrent high-grade gliomas from radiation injury:a pilot study using dynamic contrast-enhanced MR imaging. Acad Radiol.2011;18(5):575?83.191. Dujardin MI, Sourbron SP, Chaskis C, Verellen D, Stadnik T, de Mey J,et al. Quantification of cerebral tumour blood flow and permeability withT1-weighted dynamic contrast enhanced MRI: a feasibility study. J Neu-roradiol. 2012;39(4):227?35.135192. Laprie A, Catalaa I, Cassol E, McKnight TR, Berchery D, Marre D, et al.Proton magnetic resonance spectroscopic imaging in newly diagnosedglioblastoma: predictive value for the site of postradiotherapy relapsein a prospective longitudinal study. Int J Radiat Oncol Biol Phys.2008;70(3):773?81.193. Pirzkall A, Li X, Oh J, Chang S, Berger MS, Larson Da, et al. 3D MRSI for re-sected high-grade gliomas before RT: tumor extent according to metabolicactivity in relation to MRI. Int J Radiat Oncol Biol Phys. 2004;59(1):126?37.194. Karunanithi S, Bandopadhyaya GP, Sharma P, Kumar A, Singla S, MalhotraA, et al. Prospective comparison of 99mTc-GH SPECT/CT and 18F-FDOPAPET/CT for detection of recurrent glioma: A pilot study. Clin Nucl Med.2014;39(2):e121?e128.195. Nielles-Vallespin S, Weber MA, Bock M, Bongers A, Speier P, Combs SE,et al. 3D radial projection technique with ultrashort echo times for sodiumMRI: clinical applications in human brain and skeletal muscle. Magn Re-son Med. 2007;57(1):74?81.196. Weber MA, Henze M, T?ttenberg J, Stieltjes B, Meissner M, Zimmer F,et al. Biopsy targeting gliomas: do functional imaging techniques identifysimilar target areas? Invest Radiol. 2010;45(12):755?68.197. Tanderup K, Olsen DR, Grau C. Dose painting: art or science? RadiotherOncol. 2006;79(3):245?8.198. Daisne JF, Sibomana M, Bol A, Doumont T, Lonneux M, Gr?goire V. Tri-dimensional automatic segmentation of PET volumes based on measured136source-to-background ratios: influence of reconstruction algorithms. Ra-diother Oncol. 2003;69(3):247?50.137

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.24.1-0166880/manifest

Comment

Related Items