This thesis establishes normal age-related changes in the magnetic resonance (MR) T1 and T2 relaxation time constants using data collected as part of the National Institutes of Health (NIH) MRI Study of Normal Brain Development. This ongoing multi-centre study of normal brain and behaviour development provides both longitudinal and cross-sectional data and has enabled us to investigate the relaxation time constant evolution in several brain regions for children within the range of 0-4.5 years. Due to the multi-centre nature of the study and the extended period of data collection, periodically scanned inanimate and human phantoms were used to assess intra and inter-site variability. The main finding of this thesis is the parametrization of the mono-exponential behaviour of both the T1 and T2 relaxation time constants from birth until 4.5 years of age. This behaviour is believed to reflect the rapid changes in water content as well as myelination processes observable during neonatal brain development. These results, comprising over 200 subject scans, represents a subset of a publicly available normative pediatric MRI database, providing a basis for comparison for studies assessing normal brain development and deviation due to various neurological disorders.

This thesis presents two techniques for enhancing the volume coverage and temporal resolution of functional magnetic resonance brain imaging on conventional clinical scanners, namely, echo-shifting and retrospective temporal resolution selection (RTRS). The techniques are compared with conventional 2D gradient echo imaging on the bases of speed and sensitivity to functional changes, and the necessary theory is reviewed to develop physical models explaining the different properties observed. / Acquisition, reconstruction, and analysis software packages for functional brain mapping are presented and demonstrated with visual stimulation. The echo-shifted sequence permitted acquisition of 3D maps of brain activity which could be better correlated with local anatomy than 2D maps, and the RTRS method provided physiological response curves with greatly increased temporal resolution.

A computer program was developed during the course of this work to calculate electron arc dose distributions with the angle $ beta$ concept. The angle $ beta$ uniquely describes the dependence of radial percentage depth doses an electron arc therapy on the nominal field width, isocenter depth, and virtual source-axis distance. The $ beta$ concept can be used in clinical situations to determine the field width when the isocenter depth and the required radial percentage depth dose are known. This thesis presents an overview of the physical properties of electron arc therapy and describes in detail the angle $ beta$ pseudo-arc technique used at McGill. A description of the algorithms used in the computer program is given the $ beta$ technique is compared to measurements and other calculation methods.

Current brachytherapy dosimetry protocols assume that a physical source may be approximated by a point source. A new brachytherapy dosimetry protocol, recently proposed by the American Association of Physicists in Medicine Task Group 43, has the advantage of using functions derived solely from measurements performed in the medium and uses a more realistic source geometry than the point source approximation. The aim of this work is to obtain the dosimetric functions required by this new protocol for both a low and a high dose-rate Iridium-192 brachytherapy source through dose measurements in a water-equivalent phantom. / Dose measurements have been performed using lithium fluoride thermoluminescent detectors positioned in a polystyrene phantom at distances from the source that vary from 1 cm to 10 cm, with 1-cm intervals, and at angles that vary from 0$ sp circ$ to 170$ sp circ$ with 10$ sp circ$ intervals. / Our experimental results have clearly shown that the point-source approximation model can overestimate the dose to water, especially for the high dose-rate source, where we have found that differences between point-source estimates and exact measured values can differ by almost 30% for points along the longitudinal axis of the source.

Computer methods have proven very useful in many medical imaging studies and applications. The advantages of using computers for the analysis of the images of different modalities have raised the interest of researchers from different fields of science and medicine. / An image analysis application which aims to quantify Multiple Sclerosis (MS) lession in human brains based on Magnetic Resonance Imaging (MRI) has been developed. In this thesis, image analysis based on statistical theory applied in medical imaging studies is reviewed. A three-dimensional method has been developed to compare the classification probability distribution of two groups of subjects. Using random field theory, regional difference of two three-dimensional classification probability maps is assessed statistically. The method can be applied to the comparison of probability distributions of multiple sclerosis lesion in human brains under different conditions. The sensitivity of the method is presented, and it is shown that the method can effectively and reliably detect the regional difference between two populations. In addition, a novel algorithm is introduced to automatically register a T 2-weighted MRI into a T1-weighted MRI acquired from the same human brain.

Many of the problems faced in magnetic resonance imaging and angiography are due to hardware limitations of the scanners used. The use of multidimensional excitation pulses, and new, fast acquisition sequences such as echo-planar and spiral imaging, put demands on the gradient systems and although technology is progressing in these areas, performance must be carefully monitored to ensure artefact-free images. This thesis deals with four distinct aspects of magnetic resonance imaging (MRI). / A simple hardware modification to decrease the minimum achievable echo-time of our MRI scanner was designed, built, programmed and tested and was found to improve the quality of inflow angiograms significantly. Further improvements were demonstrated with the use of on-resonance (binomial) magnetization transfer saturation contrast enhancement pulses. / A method for measuring the k-space trajectories of gradient waveforms was adopted, modified, validated and used to measure the k-space trajectories of gradient waveforms used for selective excitation and spiral image acquisition. Distortion was observed even when gradient waveforms were designed within specifications of the manufacturer. / The literature reports the application of the k-space model, usually associated with image acquisition, to the design of multidimensional selective excitation pulses. This thesis demonstrates theoretically and experimentally the modification of the design procedure to compensate RF envelopes for distortion of the k-space trajectories of the accompanying gradient waveforms by using measured k-space data. / The correction of spiral image reconstruction algorithms to compensate for k-space trajectory distortion was also demonstrated theoretically and experimentally.

This thesis is concerned with the creation, verification and implementation of a computer programme for simulating photon transport in diagnostic radiology. The programme is based on the Monte Carlo technique in which probabilistic problems are solved using random numbers. For this reason mathematical number generators along with a few standard tests which enable the randomness of the numbers to be evaluated are introduced. A discussion on sampling from probability distributions is presented with emphasis on the physical aspects of the Monte Carlo technique applied to low energy photon transport. The validity of the computer programme is established by comparisons with previously published Monte Carlo work, with predictions from an analytical photon transport model and with experimental measurements. The application of Monte Carlo simulations to specific radiographic problems are illustrated by the modeling of percent depth doses.

Biological effects and prostate motion were studied for prostate cancer patients treated with external beam radiotherapy. The prostate motion was determined using an ultrasound-based patient positioning system (BAT: B-mode Acquisition and Targeting) just after conventional patient setup. The changes in planned biological effects due to prostate motion were calculated for the prostate target organ and for the rectum and bladder normal tissues using TCP (tumor control probability) and NTCP (normal tissue complication probability) calculations, respectively. The prostate TCP calculations were carried out for both prostate PTV and GTV structures, whereas the rectum and bladder wall structures were used for the NTCP calculations.

Current challenges facing us in developing dedicated positron emission tomography (PET) systems for metabolic breast mammography (PEM) and small animal (ANIPET) are to achieve high spatial resolution (less than 2 mm) and high efficiency. It is also crucial to extend the sensitive areas of PEM detectors to their periphery in order to overcome the difficulty in imaging near a patient's chest wall. This limitation of the periphery dead region was revealed in the clinical trials of our previously developed PEM-I system. / In the new study, we developed prototype detectors by using position-sensitive photomultiplier tubes (PS-PMTS) and pixelated bismuth germanate (BGO) crystals with depth encoding scheme to detect and localize gamma rays. We used the following methods in crystal processing: (1) Crystal cutting---each crystal block was cut by diamond saw into small elements of 2.1 mm x 2.1 mm (2.2 mm pitch) on two opposite faces. The elements on one face of the block offset by half the crystal pitch from those on the opposite face in both X and Y dimensions. The depths of two layers were 11.5 mm and 6.5 mm, respectively. The middle solid space between the two layers was 2 mm. (2) Crystal polishing and encapsulating---The very roughly cut surfaces were chemically polished by acid etching method and the cut slots were encapsulated with an epoxy-compound mixture. (3) Crystal separating---the pseudo discrete crystal blocks were cut along the middle solid space into two discrete segments. (4) Crystal coupling---the two crystal segments were glued together and optically coupled to the PS-PMTs window. / We also developed front-end electronic circuits including high-voltage dividers, anode resistor chains, position readout circuits, and last-dynode timing circuits. Methods for combining four PS-PMTs with simple X+, X-, Y+, Y- outputs have been developed to further simplify the position recording. The detectors were constructed in the structure of arrays (two in the system)---modules (four in each array)---units (four in each module). The basic unit of one crystal and one PS-PMT was formed as field replaceable unit. The acquired list-mode data were analyzed with MATLAB and C. Different methods to generate distortion look-up-table were examined and evaluated. / Our new prototype detectors have spatial resolutions of 1.8 mm (vs. 2.8 mm in PEM-I), timing resolution of 10.3 ns (vs. 12 ns in PEM-I), and a field-of-view of 88 mm x 88 mm (vs. 64 mm x 56 mm in PEM-I). Our analysis shows that the design improves the spatial resolution, enhances the detector field-of-view, and significantly reduces the peripheral dead regions.

In a large multi-center study where many groups collaborate to produce data, the conditions under which the data are collected can vary significantly. In the case of the NIH funded MRI study of normal brain development (NIHPD), 6 centers across the US collect MRI scans of the developing child brain. In all, four different models of scanners from 2 manufacturers are used. The subject of this thesis is to quantify differences in acquired data that are due to scanner differences and thus enable improved biological study. / The aim of this study is to analyze the scanner-induced differences across sites using a specialized phantom recommended by the American College of Radiology (ACR). Together with the 7 tests listed by the ACR MRI accreditation manual, phantom data from each site can be analyzed for parameters that are characteristic of image quality. The measurements are done automatically using programs written at the MNI, that use MINC tools and the MNI autoreg package to compute the necessary parameters. / The results demonstrate that, in general, all data obtained complied with the 7 tests of the ACR MRI manual. Statistically significant variations over time and between sites were found in the data. However, the magnitude of the variations is of the order of 1%-2% indicating stable data acquisitions and comparability of data between sites.