Magnetic Resonance Imaging (MRI) and Fricke-gelatin dosimetry are used to measure absorbed dose distributions from high dose rate (HDR) brachytherapy treatments. The high activity of the HDR brachytherapy source is suitable for the prompt delivery of the high doses required to give changes in the Fricke-gelatin Nuclear Magnetic Resonance (NMR) characteristics which are readily detected by MRI. The MR images can map 3D dose distributions deposited in the Fricke-gelatin matrix. Spin-lattice relaxation times (T1) and rates (R1) are computed from MR images of irradiated Fricke gelatin phantoms in order to study the dose response relationship for the Fricke gelatin system. It was confirmed that the Fricke gelatin system's R1 varies linearly with absorbed dose to a saturation limit. A fast approach for determining radiation dose from MR image intensity using a calibration curve is described. Dose distributions generated from MR images via the calibration curve show good agreement with expected distributions generated from a computerized treatment plan. As well, dose data generated using R1 maps agreed well with those generated by calibration curve. The use of MRI with Fricke-gel dosimetry is shown to be a viable means of verifying the dose distributions from high activity brachytherapy sources.

X-ray mammography cannot always distinguish between benign and malignant breast lesions. This leads to unnecessary biopsies, costs, and stress for the patient. Positron Emission Mammography (PEM) provides images of increased glucose metabolism in malignant tumours compared with healthy tissue. After injection of a radioactively-labelled glucose analog, cancerous tumours appear as bright spots on the breast image. / Quantitative analysis of PEM images consists in comparing the amount of activity absorbed in both breasts of a patient. Based on ROC analysis of 15 subjects, an asymmetry of 10% in the number of counts detected from each breast was taken as a sign of cancer. The application of the count asymmetry method seems to results in a 22% improvement of PEM accuracy (from 64% to 86%). It is particularly useful for the detection of big or diffuse tumours. Quantitative data will also provide tools for future applications of PEM technology, such as follow-up of patients after cancer therapy.

A novel approach to correcting for intensity non-uniformity in MR data is described that achieves high performance without requiring supervision. By making relatively few assumptions about the data, the method can be applied at an early stage in an automated data analysis, before a tissue intensity or geometric model is available. Described as Non-parametric Non-uniform intensity Normalization (N3), the method is independent of pulse sequence and insensitive to pathological data that might otherwise violate model assumptions. To eliminate the dependence of the field estimate on anatomy, an iterative approach is employed to estimate both the multiplicative bias field and the distribution of the true tissue intensities. The performance of this method is evaluated using both real and simulated MR data. Preprocessing of MR data using N3 is shown to substantially improve the accuracy of anatomical analysis techniques such as tissue classification and cortical surface extraction.

This thesis describes an environment in which two angiographic methods, Digital Subtraction Angiography (DSA) and Magnetic Resonance Angiography (MRA), can be objectively compared and analyzed. / Such an environment requires that both angiographic projections be displayed from the same view-point and with the same projection geometry. The two angiograms are displayed side by side and several points on the vascular structure are identified in both modalities. These points are used to estimate, using a Least Squares Minimization, the Homogeneous Transformation Matrix (HTM) characterizing the projection of the DSA image. The resulting HTM is used to generate a corresponding Maximum Intensity Projection (MIP) of the MRA dataset. The number and location of the required homologous point-pairs are determined empirically. / Other alternatives to MIP are presented as well, and their performance relative to DSA is discussed. Images from each modality are displayed stereoscopically to reflect the three dimensional nature of the vascular tree.

In modern radiotherapy, 3D conformal dose distributions are achieved using several beam ports each having pre-calculated planar distributions of photon beam intensity. The intensity matrix for a given beam port is generated by independent motion of the leaves of a multileaf collimator (MLC). In this thesis, we have used the step-and-shoot approach to intensity-modulated beam delivery, the safest and most popular approach at the moment. The first component of this thesis was to write a leaf sequence algorithm to control the MLC fitted to our Clinac 2300 C/D linear accelerator. Our algorithm is more efficient than other published step-and-shoot type algorithms, and takes into account the MLC transmission, MLC penumbra, and change in scatter conditions with field size. / Although sophisticated means to calculate and deliver these spatially-modulated beams have been developed by our group as well as by other medical physics research centres, means to verify their actual delivery are definitely the most problematic at the moment, making equipment and treatment quality assurance difficult to enforce. The second (and major) component of this thesis has been to investigate the use of a new portal imaging device for dosimetric verification purposes. We show that an electronic portal imaging device of the scanning liquid ionization chamber type yields images which, once calibrated from a previously-determined calibration curve, provide highly-precise planar maps of the incident dose rate distribution. For verification of an intensity-modulated beam delivered in the segmented approach with an MLC, a portal image is acquired for each subfield of the leaf sequence. Subsequent to their calibration, the images are multiplied by their respective associated monitor unit settings, and summed to produce a planar dose distribution at the measurement depth in phantom. The excellent agreement of our portal imager measurements with calculations of our treatment planning system and measurements with a one-dimensional beam profiler attests to the usefulness and relative simplicity of this method for the planar verification of intensity-modulated fields, which are produced in the segmented approach on a computerized linear accelerator equipped with an MLC.

This thesis presents the development and implementation of a quantitative multi-slice cerebral perfusion imaging technique using magnetic resonance imaging. An acquisition sequence capable of acquiring up to 9 slices was designed and implemented into two final pulse sequences: an interleaved perfusion/BOLD (blood oxygenation level dependent) sequence and a perfusion-only sequence. A number of practical imaging issues were addressed and resolved, including the design of an appropriate inversion pulse for labelling of arterial spins, spatial offsetting of this pulse for use in the arterial spin labelling technique chosen for implementation, and the design of various saturation pulses necessary for quantification of the technique. Experimental validation of the quantitative multi-slice perfusion technique was performed by measuring visual cortex cerebral blood flow (CBF) values in a group of 8 subjects using a block-design visual stimulus paradigm. Results indicated good sequence stability and CBF measurements agreed well with quantitative values found in the literature.

A fully automatic procedure for brain tissue classification from 3D magnetic resonance head images (MRI) is described. The procedure uses feature space proximity measures, and does not make any assumptions about the tissue intensity data distributions. As opposed to existing methods for automatic tissue classification, which are often sensitive to anatomical variability and pathology, the proposed procedure is robust against morphological deviations from the model. A novel method for automatic generation of classifier training samples, using a minimum spanning tree graph-theoretic approach, is proposed in this thesis. Starting from a set of samples generated from prior tissue probability maps (the "model") in a standard, brain-based coordinate system ("stereotaxic space"), the method reduces the fraction of incorrectly labelled samples in this set from 25% down to 2%. The corrected set of samples is then used by a supervised classifier for classifying the entire 3D image. Validation experiments were performed on both real and simulated MRI data; the kappa similarity measure increased from 0.90 to 0.95.

A Monte Carlo simulation system was enhanced in order to analyze photon transport in tapered geometries. This system was used to evaluate the performance of new tapered collimator designs, for multi-slice positron emission tomography (MS-PET) and positron volumetric imaging systems (PVI-Systems). For each imaging modality, the performance of the new collimators was compared to those of the current collimators of choice. Collimators were evaluated based upon their: (1) trues detection efficiencies; (2) scatter detection efficiencies; (3) scatter fractions; and (4) total singles to trues events ratios. In MS-PET, collimators with a diamond-like cross section--double tapered collimators--were introduced and their performance compared to those of single tapered and cylindrical collimators. Double tapered collimators yielded better performance than single tapered ones. In PVI systems, external tapered collimators were simulated and their performance compared to those of cylindrical ones. External tapered collimators did not improve the scanner performance.

An efficient and productive radiation treatment planning (RTP) system must make use of both appropriate visualization techniques and good user interface design. The suitability of several visualization techniques have been examined in the context of 3-D radiation treatment planning. These techniques include wire frame, surface rendering, volume rendering and a subset of volume rendering: reformatting of data. A rudimentary computerized RTP system was written using the most appropriate visualization techniques examined earlier. These techniques were used to display the anatomical data acquired from computed tomography (CT) scanners, the beam position within the anatomy, and finally, the dose distributions resulting from the entered plan. The program was written in ANSI C and runs on a Silicon Graphics Personal Iris UNIX workstation. The system makes use of effective user interface tools and efficient code which results in an efficient and interactive system. The accuracy of the system is verified by comparing dose profiles obtained with film dosimetry and from the computer calculations.

Functional magnetic resonance imaging (fMRI) is a technique allowing the study of hemodynamic changes in the brain. Having a number of advantages over positron emission tomography (PET), the current gold standard, fMRI appears like an attractive alternative to study brain function. However, the measurements obtained with fMRI have not been rigorously validated. / This thesis describes a study comparing cerebral blood flow (CBF) changes measured using a flow-sensitive alternating inversion recovery (FAIR) fMRI perfusion method, to the ones obtained using PET. We scanned 10 healthy normal volunteers under identical experimental conditions during presentation of 4 levels of visual stimulation and one level of hypercapnia. The CBF changes were compared in 4 regions-of-interest. / Good correspondence was found in the locations of the CBF changes. FAIR CBF changes had a higher signal-to-noise ratio and presented a monotonic increase with stimulation intensity, absent in PET measurements. FAIR measurements were correlated to PET but slightly lower. Statistical analysis of the data did not show that FAIR measurements were significantly different from PET ones.