Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal

Iterative CT reconstruction algorithms are superior to the standard convolution backpropagation (CBP) methods when reconstructing from a small number of views (hence less radiation), but are computationally costly. To reduce the execution time, this work implements and tests a parallel approach to iterative algorithms using a cluster of workstations, which is a low cost system found in many offices and non-academic sites. A previous implementation showed little speedup because of the significant cost of inter-processor communication. In this thesis, several data partitioning methods are examined, including some image tiling methods that exploit the spatial locality demonstrated by local CT. Using these methods, computation can proceed locally, without the need for inter-processor communication during every iteration. A relative speedup of up to 17 times is obtained using 25 processors, demonstrating that good performance can be obtained running computationally intensive CT reconstruction algorithms on distributed memory hardware.

A plethora of anthropological studies have been undertaken on the skull, including many analyses of sexual dimorphism. Sexual dimorphism reflected in the eye orbits has not always demonstrated consistent or reliable results. However, recent studies (Pretorius, Steyn, & Scholtz, 2006; Ji et al., 2010) suggest some positive results utilizing geometric morphometrics to predict sex. Utilizing 97 post-mortem CT (computed tomography) scans, established morphological and metric techniques for sex determination were assessed from 3D rendered models of the crania. In addition, landmark data were collected on the orbital margin to evaluate the accuracy of sex determination using geometric morphometric techniques. Traditional methods demonstrated poor levels of accuracy for prediction of sex, however, utilizing generalised procrustes analysis and discriminant function analysis on 3D landmark data resulted in 94.95% overall accuracy. Application of recent methodological advances, including geometric morphometrics, should continue to be developed as it increases the ability to assess sexual dimorphism which will allow for greater identification of unknown remains.

Pressboard, a high density cellulose-based material used for insulation in high voltage power transformers, exhibits stress relaxation during compressive loading. Investigating the micro-mechanical mechanisms responsible for the relaxation can lead to modifications of the production process to control the behaviour of the material. This investigation can be done using Synchrotron X-ray micro Computed Tomography which provides sufficient temporal and spatial resolutions to capture the stress relaxation process. In the present thesis, a compression rig for in-situ mechanical loading during X-ray micro Computed Tomography was designed and constructed. Local tomography scans with sub-micrometre resolution were obtained at the TOMCAT beamline at the Swiss Light Source, Paul Scherrer Institut, Switzerland. Several fibre segmentation techniques are analysed, together with Optical Flow and Digital Volume Correlation (DVC), methods used for estimating displacement, strain and velocity vector fields. Suitability of the tested methods is evaluated, and it is found that segmentation of individual fibres in a cellulose material of such a high density is probably not possible using currently available segmentation techniques. The movements during relaxation are measurable at the used resolution, and can be estimated using Optical Flow. Further work into correction of image shift due to rig movement between scans, as well as image artefact reduction should allow for measurement and comparisons of displacement during relaxation as well as DVC-computed strain measurements during compression, recreating earlier results.

Objective: To evaluate three-dimensional position of the TMJ condyle within the glenoid fossa in different types of skeletal patterns. Materials and methods: Ninety CBCT images were consecutively selected and divided into skeletal class I, class II and class III. The images were analyzed locating landmarks in the different areas of the condyle and glenoid fossa. All landmarks presented acceptable reliability. The mean results were compared using ANOVA and Bonferroni post-hoc test (p < 0.05). Results: There was a tendency for the anterior joint space to be smaller than the posterior joint space. Statistical analysis, however, evidenced no significant differences between the anterior, superior and posterior joint spaces and the different skeletal patterns or between sides. Conclusion: There was non-concentricity of the condyle for all the groups studied, and no particular direction was statistically significantly favored. It is unclear whether the differences found would be clinically significant, considering anatomical individual variations.

Iterative CT reconstruction algorithms are superior to the standard convolution backpropagation (CBP) methods when reconstructing from a small number of views (hence less radiation), but are computationally costly. To reduce the execution time, this work implements and tests a parallel approach to iterative algorithms using a cluster of workstations, which is a low cost system found in many offices and non-academic sites. A previous implementation showed little speedup because of the significant cost of inter-processor communication. In this thesis, several data partitioning methods are examined, including some image tiling methods that exploit the spatial locality demonstrated by local CT. Using these methods, computation can proceed locally, without the need for inter-processor communication during every iteration. A relative speedup of up to 17 times is obtained using 25 processors, demonstrating that good performance can be obtained running computationally intensive CT reconstruction algorithms on distributed memory hardware.

A plethora of anthropological studies have been undertaken on the skull, including many analyses of sexual dimorphism. Sexual dimorphism reflected in the eye orbits has not always demonstrated consistent or reliable results. However, recent studies (Pretorius, Steyn, & Scholtz, 2006; Ji et al., 2010) suggest some positive results utilizing geometric morphometrics to predict sex. Utilizing 97 post-mortem CT (computed tomography) scans, established morphological and metric techniques for sex determination were assessed from 3D rendered models of the crania. In addition, landmark data were collected on the orbital margin to evaluate the accuracy of sex determination using geometric morphometric techniques. Traditional methods demonstrated poor levels of accuracy for prediction of sex, however, utilizing generalised procrustes analysis and discriminant function analysis on 3D landmark data resulted in 94.95% overall accuracy. Application of recent methodological advances, including geometric morphometrics, should continue to be developed as it increases the ability to assess sexual dimorphism which will allow for greater identification of unknown remains.

X-ray computed tomography (CT) polymer gel dosimetry (PGD) is an attractive tool for three-dimensional (3D) radiation therapy (RT) treatment verification due to the availability of CT scanners in RT clinics. Nevertheless, wide-spread use of the technique has been hindered by low signal-to-noise CT images largely resulting from gel formulations with low radiation sensitivity. However, a new gel recipe with enhanced dose sensitivity was recently introduced that shows great promise for use with CT readout. This dissertation describes development of an CT PGD system for 3D verification of RT treatments using the new gel formulation. The work is divided into three studies: gel characterization, commissioning of a multislice CT scanner and investigation of a dose rate dependence observed during gel characterization. The first component of this work examines the dosimetric properties of the new gel formulation. The response of the gel is found to be stable between 15 - 36 hours post-irradiation and excellent batch reproducibility is seen for doses between 0 - 28 Gy. A dose rate dependence is found for gels irradiated between 100 - 600 MU/min, indicating machine dose rate must be consistent for calibration and test irradiations to avoid dosimetric error. An example clinical application is also presented using an IMRT treatment verification that demonstrates the potential of the system for use in modern RT. The second component of this work focuses on commissioning a multislice CT scanner for CT PGD. A new slice-by-slice background subtraction technique is introduced to account for the anode heel effect. Additional investigations show recommendations for optimizing image quality in CT PGD using a single slice machine also apply to multislice scanners. In addition, the consistency of CT numbers across the multislice detector array is found to be excellent for all slice thicknesses. Further work is performed to assess the tube load characteristics of the scanner and develop a scanning protocol for imaging large gel volumes. Finally, images acquired throughout the volume of an unirradiated active gel show variations in CT data across each image on the order of 7 HU. However, these variations are not expected to greatly influence gel measurements as they are consistent throughout the gel volume. The third component of this work examines the dose rate dependence found during gel characterization. Studies using gel vials and 1 L cylinders indicate the response of the gel does not depend on changes in mean dose rate on the order of seconds to minutes. However, the machine dose rate remains, indicating variations in dose rate on the order of milliseconds influence the response of the gel. An attempt is made to mitigate the effect by increasing the concentration of antioxidant in the gel system but results in reduced overall response. Further work is performed to determine if self-crosslinking of one of the gel components contributes to the observed machine dose rate dependence. In summary, this dissertation has significantly advanced the field of gel dosimetry by providing a prototype CT PGD system with enhanced dose resolution for complex RT treatment verification. / Graduate / 0992 / 0495 / 0756 / holly.johnston@utsouthwestern.edu

Circular source and detector trajectories in cone-beam computed tomography (CT) are known to collect insufficient data for accurate object reconstruction. One model predicts that the lacking information corresponds to a shift-variant cone of missing spatial frequency components in the local Fourier domain. These predictions were experimentally verified by imaging small, localized objects and observing their Fourier transforms. Measurements indicated that the internal angle of the ‘missing cone’ varies as the angle of locally intersecting x rays with respect to the horizontal plane, as expected. Object recovery was also found to depend greatly on the distribution of the object’s frequency spectrum relative to the missing cone, as predicted. Findings agreed with more anatomically relevant phantoms, which showed preferential intensity discrepancies at gradients oriented within or near the missing cone. Methods for artefact correction are in general limited to approximation unless a priori information is incorporated.

Circular source and detector trajectories in cone-beam computed tomography (CT) are known to collect insufficient data for accurate object reconstruction. One model predicts that the lacking information corresponds to a shift-variant cone of missing spatial frequency components in the local Fourier domain. These predictions were experimentally verified by imaging small, localized objects and observing their Fourier transforms. Measurements indicated that the internal angle of the ‘missing cone’ varies as the angle of locally intersecting x rays with respect to the horizontal plane, as expected. Object recovery was also found to depend greatly on the distribution of the object’s frequency spectrum relative to the missing cone, as predicted. Findings agreed with more anatomically relevant phantoms, which showed preferential intensity discrepancies at gradients oriented within or near the missing cone. Methods for artefact correction are in general limited to approximation unless a priori information is incorporated.