Temporal phenomena in a range of disciplines are more naturally modelled in continuous-time than coerced into a discrete-time formulation. Differential systems form the mainstay of such modelling, in fields from physics to economics, geoscience to neuroscience. While powerful, these are fundamentally limited by their determinism. For the purposes of probabilistic inference, their extension to stochastic differential equations permits a continuous injection of noise and uncertainty into the system, the model, and its observation. This thesis considers Bayesian filtering for state and parameter estimation in general non-linear, non-Gaussian systems using these stochastic differential models. It identifies a number of challenges in this setting over and above those of discrete time, most notably the absence of a closed form transition density. These are addressed via a synergy of diverse work in numerical integration, particle filtering and high performance distributed computing, engineering novel solutions for this class of model. In an area where the default solution is linear discretisation, the first major contribution is the introduction of higher-order numerical schemes, particularly stochastic Runge-Kutta, for more efficient simulation of the system dynamics. Improved runtime performance is demonstrated on a number of problems, and compatibility of these integrators with conventional particle filtering and smoothing schemes discussed. Finding compatibility for the smoothing problem most lacking, the major theoretical contribution of the work is the introduction of two novel particle methods, the kernel forward-backward and kernel two-filter smoothers. By harnessing kernel density approximations in an importance sampling framework, these attain cancellation of the intractable transition density, ensuring applicability in continuous time. The use of kernel estimators is particularly amenable to parallelisation, and provides broader support for smooth densities than a sample-based representation alone, helping alleviate the well known issue of degeneracy in particle smoothers. Implementation of the methods for large-scale problems on high performance computing architectures is provided. Achieving improved temporal and spatial complexity, highly favourable runtime comparisons against conventional techniques are presented. Finally, attention turns to real world problems in the domain of Functional Magnetic Resonance Imaging (fMRI), first constructing a biologically motivated stochastic differential model of the neural and hemodynamic activity underlying the observed signal in fMRI. This model and the methodological advances of the work culminate in application to the deconvolution and effective connectivity problems in this domain.

Bone marrow oedema is a term used to define the appearance of regions of low signal on T1 weighted and high signal on T2 weighted fat-suppressed magnetic resonance images. The potential association between bone marrow oedema and prognosis in pathologies such as osteoarthritis is becoming increasingly recognised through clinical studies. A limited number of clinical studies have linked bone marrow oedema to altered bone density or altered bone marrow perfusion. The principal aims of this study were to investigate these findings in vitro, using the equine forelimb. The presence of bone marrow oedema within the equine forelimb was initially confirmed by undertaking magnetic resonance imaging scans. Bone samples were selected from 10 animals, 5 exhibiting the presence of bone marrow oedema-type abnormalities (BMOA) at the distal metacarpal. Raman microspectroscopy was used to determine the chemical composition of bone and projection radiography to provide a measure of bone density. Micro computed x-ray tomography was undertaken on a subset of three bone samples exhibiting BMOA. A second component of the study utilised contrast enhanced magnetic resonance imaging to enable comparison of perfusion to bone marrow with and without evidence of oedema. A saline flushing agent containing Evan’s blue was used so that subsequent sectioning of the bone would enable visualisation of the distribution of contrast agent as part of a histological examination of the oedematous region. An initial observation was that the majority of bone marrow oedema that was observed in the distal metacarpal appeared in a consistent location, namely the postero-inferior aspect of the bone, corresponding to the point of greatest load thereby suggesting a potential relationship to forces upon the joint. The principal observations were that there appears to be increased bone volume densities in those bone samples with evidence of bone marrow oedema. The Raman microspectroscopy did not demonstrate any statistically significant differences in the chemical composition of bone. Hence the overall impression is that bone marrow oedema is associated with a greater volume of bone, although of similar maturity and composition. There was limited evidence of increased perfusion (suggestive of increased vascularity and / or hyperpermeability) in those samples with bone marrow oedema. This work suggests that these particular bone marrow oedema lesions are associated with bone changes and potentially vascular changes although the aetiology is currently unclear. Further work is needed to determine the clinical significance and prognosis associated with these particular lesions, and whether these findings can be replicated for bone marrow oedema demonstrated at other anatomical locations.

3T MRI provides higher signal-to-noise ratio images compared to lower field machines. However, a major drawback of 3T MRI is a higher B1 transmission-field inhomogeneity across the field-of-view compared to imaging at lower fields. B1-field mapping was performed on volunteers using a Philips 3.0T MR scanner and a typical head-first prone patient positioning technique. The B1-field transmitted in the breasts was found to be reduced towards the right side of the body. In some volunteers, the B1-field was reduced to about one-half of the nominal field in the right breast. To minimize the B1 inhomogeneity artefacts, a saturation recovery snapshot FLASH (SRSF) imaging sequence was proposed. Different saturation techniques were assessed. The best saturation efficiency was produced by Hoffmann’s saturation method. By using Hoffmann’s SRSF sequence, the error in the enhancement ratio (ER) can be reduced to about one half compared to imaging obtained using typical FLASH sequence in the presence of a 50% B1-field reduction. Other techniques i.e. bilateral power optimization and a dedicated patient support system were also tested. Both of these approaches produced substantial reductions of the B1 inhomogeneity seen with the standard technique. To address the effects of the native T1 (T10) of different tissues on DCE-MRI, novel enhancement factor indices calculated using SRSF sequence images were introduced and assessed. Computer simulations and gel phantom experiments showed that less error was observed in the indices calculated compared to the ER calculated using the conventional and widely used FLASH sequence. Furthermore, the effect of B1-field inhomogeneity on the novel indices is also reduced. One of the indices proposed is directly related to the contrast agent concentration. The theory and results presented show that the SRSF pulse sequence and the quantification techniques proposed have the potential to improve the accuracy of breast DCE-MRI at 3T.

Résumé: La majorité des femmes ayant un cancer du sein en stade précoce sont traitées par radiothérapie impliquant souvent l’irradiation du sein entier. Malgré l’efficacité de cette modalité de traitement, la dose de radiation n’est pas optimale pour éliminer toutes les cellules cancéreuses résiduelles, mais plutôt pour obtenir les meilleurs résultats à long terme avec le moins de complications possibles. Les effets secondaires observés résultent tous de processus inflammatoires engendrés par la radiation. L’augmentation de l’expression et de l’activité de molécules inflammatoires, notamment la cyclooxygénase-2, dans les tissus normaux et malins stimulent l’invasion et l’angiogenèse tumorales, deux mécanismes importants menant à l’établissement de métastases. Le but global de ce projet de recherche est d’améliorer la radiothérapie en tentant de réduire la récurrence du cancer du sein. Les objectifs spécifiques étaient de déterminer grâce à l’imagerie par résonance magnétique que l’irradiation du stroma sain pouvait augmenter in vivo la capacité d’invasion des cellules cancéreuses du sein, stimuler la néovascularisation tumorale et qu’une co-administration à la radiothérapie d’un agent anti-inflammatoire inhibiteur sélectif de la cyclooxygénase-2 pouvait prévenir l’augmentation radio-induite de cette invasion. Dans notre étude, nous avons établi une méthode utilisant l’imagerie par résonance magnétique pour mesurer rapidement l’angiogenèse in vivo chez la souris dans des implants de Matrigel. Cette méthode servira ultérieurement à analyser l’effet de la radiation sur l’angiogenèse tumorale. Nous avons également suivi chez un modèle de souris l’invasion des cellules cancéreuses mammaires implantées après irradiation du tissu sain. Nous avons démontré que l’irradiation du tissu sain augmente l’invasion des cellules cancéreuses mammaires. L’invasion radio-induite est stimulée par une irradiation unique de 30 Gy tout comme avec un protocole d’irradiations fractionnées de 5x7,5 Gy se rapprochant plus des doses utilisées en clinique. Ensuite, un traitement avec un inhibiteur sélectif de cyclooxygénase-2, soit le NS-398, a été effectué. Le NS-398 limite l’augmentation radio-induite de l’invasion. Ces résultats supporteraient le développement de nouveaux traitements basés sur des inhibiteurs de COX-2 pour augmenter l’efficacité de la radiothérapie chez les femmes ayant un cancer du sein. / Abstract: Most women with early breast cancer are treated with radiotherapy to the whole breast. Despite the efficiency of this treatment, the dose of radiation is not calculated to eliminate all the residual cancer cells, but rather to obtain the best long-term results with minimal side-effects. The observed side-effects all result from inflammatory processes caused by radiation. Increase of inflammatory molecules expression and activity, such as cyclooxygenase-2, in normal and malignant tissues induce invasion and tumour angiogenesis. Both of these important mechanisms lead to metastasis formation. The general aim of this research project is to improve radiotherapy by decreasing breast cancer recurrence. Specific objectives were to determine with magnetic resonance imaging that irradiation of normal tissues could increase breast cancer cells invasiveness in vivo, stimulate tumour neovascularization and prevent radiation-enhanced invasion by the administration of an anti-inflammatory agent inhibiting selectively the cyclooxygenase-2 during radiotherapy. In this study, we have developed a new assay to monitor angiogenesis in Matrigel plugs in live mice using magnetic resonance imaging. This method would be a promising tool to test the effect of radiation on tumour angiogenesis. We also followed in a mouse model the invasion of mammary cancer cells implanted post-irradiation of healthy tissues. We demonstrated that irradiation of healthy tissues leads to an increase in mammary cancer cells invasion. Radiation-induced invasion was observed with a unique 30 Gy dose as well as with a more clinically-relevant fractionated protocol consisting in 5 irradiations of 7.5 Gy. Then, mice were treated with NS-398, a selective inhibitor of cyclooxygenase-2. NS-398 limits the increase of invasion stimulated by radiation. These results could support new treatments development based on COX-2 inhibition to increase radiotherapy efficiency for women with breast cancer.

Chronic liver disease is a worldwide health problem, and hepatic fibrosis (HF) is one of the hallmarks of the disease. The current reference standard for diagnosing HF is biopsy followed by pathologist examination; however, this is limited by sampling error and carries a risk of complications. Pathology diagnosis of HF is based on textural change in the liver as a lobular collagen network that develops within portal triads. The scale of collagen lobules is characteristically in the order of 1 to 5 mm, which approximates the resolution limit of in vivo gadolinium-enhanced magnetic resonance imaging in the delayed phase. We use MRI of formalin-fixed human ex vivo liver samples as phantoms that mimic the textural contrast of in vivo Gd-MRI. We have developed a local texture analysis that is applied to phantom images, and the results are used to train model observers to detect HF. The performance of the observer is assessed with the area-under-the-receiver-operator-characteristic curve (AUROC) as the figure-of-merit. To optimize the MRI pulse sequence, phantoms were scanned with multiple times at a range of flip angles. The flip angle that was associated with the highest AUROC was chosen as optimal for the task of detecting HF. (C) The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.

It is more than a decade that Ionic Polymer-Metal Composites (IPMCs) have been known as an exciting class of smart materials and attracted growing worldwide attention. IPMCs are soft and flexible, and can generate large and reversible strains in response to electrical stimulus. Thus, they have potential applications in industrial and biomedical fields, as actuators. Before these applications can be realized , however, the performance of IPMCs must be understood and improved through improvement of component characteristics and of preparation methods. In general, the aim of this thesis is to gain a fundamental understanding of the chemical and structural factors that affect the electromechanical performance of IPMCs. To this end, a multi-technique investigation is applied to correlate the electrochemical and electromechanical behavior of IPMCs, during operation, with their chemistry, microstructure and nanostructure. Researchers have suggested several plausible mechanical and mathematical models to reveal that ion transport occurs within IPMCs and that it is an important factor in their actuation performance. However, there is still a need for further experimental studies to help refine our understanding of the actuation mechanism of these materials. In this work, the powerful, non-invasive and non-destructive technique of Magnetic Resonance Imaging (MRI) is employed to study the internal structure and water content distribution in Nafion membranes and also IPMCs. Moreover, MRI is also applied to image electrically-induced diffusion of ions with their associated water molecules in real time, in operating IPMC actuators. This forms the major part of this project and, to the best of our knowledge, it is the first recorded electrochemical experiment of this kind. The size and dimensions of IPMCs can affect their actuation performance. Thus, in this work, model IPMC actuators are prepared based on the available commercial Nafion membrane and fabricated cast Nafion membrane and their electromechanical behaviors are compared. The effect of parameters such as electrode composition and Nafion thickness on actuation behavior is also studied by measuring displacement and force generation of the IPMC actuators during actuation cycles. Simultaneous current and electrochemical measurements are made to correlate electrochemical processes with actuation behavior directly. Scanning electron microscopy (SEM) is also used to study the internal and surface structure of IPMCs.

Duchenne muscular dystrophy (DMD) is a fatal, muscle-wasting disorder due to mutations/deletions in the dystrophin gene. Whilst improvements in palliative care have increased the life expectancy of patients, cardiomyopathy and respiratory complications are still the leading causes of death. A potential therapy for the treatment of DMD is antisense oligonucleotides (AOs), which modulate dystrophin pre-mRNA splicing to restore the dystrophin reading frame and generate a truncated functional protein. Conjugation of AOs to cell penetrating peptides (CPP), such as Pip5e-, significantly improves delivery to skeletal muscles and to the heart, which is imperative given the impact of cardiomyopathy to mortality. However, it should be noted that the contribution of skeletal muscles, such as the core respiratory muscle, the diaphragm, in dystrophic cardiopulmonary function is poorly understood. The specific aims of the work in this thesis were to (i) understand the effect of the diaphragm on cardiac function using magnetic resonance imaging (MRI), (ii) screen a number of derivatives of Pip5e (Pip6) in an effort to discover further promising peptides and define the properties integral to heart penetrating capacity, and (iii) assess whether Pip6-PMOs restore cardiac function (MRI) following a repeat, low dose regimen. In short, the specific restoration of dystrophin in the diaphragm of the dystrophic mouse model, the mdx mouse, did not improve cardiac function, highlighting the importance of a body-wide therapy. The screening of multiple Pip5e-PMO derivatives revealed 3 promising peptides with improved cardiac splicing capacity; however, serial deletions of amino acids from the central core resulted in the diminution of dystrophin restoration, possibly due to a reduction in hydrophobicity. Finally, the Pip6-PMO treatment regimen substantially restored dystrophin protein (28% in heart) and stabilised cardiac function, even with an increased work load. In conclusion, this study illustrates the importance of a body-wide treatment, such as the CPP strategy (Pip-PMO). These Pip-PMO conjugates demonstrate high dystrophin restoration in a number of muscles, including cardiac muscle, and have a beneficial effect on cardiac function.

Magnetic resonance imaging plays an important yet underutilized role in determining the natural history and prognosis of oral carcinoma. Depth of tumour invasion is an emergent factor in the oral cancer literature. However, problems exist with the definition of cut-points suitable for inclusion in TNM staging criteria. Statistical methodology represents a possible explanation but is underexplored. In this work, a review of the depth of invasion literature is conducted with emphasis on statistical technique. As well, statistical simulation is used to explore the implications of the of the minimum p-value method. The results demonstrate that the use of continuous variable categorization and multiple testing is widespread, and contributes to cut-point variability and false-positive tests. Depth, as a predictor of OCLNM and survival, must be questioned. The volume of tumour invasion is a promising prognostic factor that has not been fully investigated in the oral carcinoma literature. In this work, the volume of tumour invasion is measured on MRI and compared to thickness and maximum diameter in its capacity to predict 2-year all-cause, disease-related and disease-free survival, as well as occult cervical lymph node metastasis prediction. As part of a comprehensive approach, morphometric factors are incorporated into multifactor predictive models using regression, artificial neural networks and recursive partitioning. It is evident that MRI-based volume is superior all other linear measurements for both occult cervical lymph node metastasis and survival prediction. Artificial neural networks wee superior to all other techniques for survival prediction. There is a case for a unified artificial neural networks model for survival prediction that uses volume, midline invasion and N-stage to determine prognosis. This model can be used to determine individualized probabilities of 2-year survival. The lateral extrinsic muscles of the tongue lie just beneath the surface of the lateral tongue, yet their invasion is a criterion for T4 classification using the TNM staging system. In this work, the Visible Human Female is used to conduct an anatomic study of the extrinsic muscles of the tongue. Linear measurement is used to quantify the distance from the surface mucosa to the most superficial muscle fibres of the styloglossus and genioglossus. Further, the lateral extrinsic muscles are poorly demonstrated on MRI. An anatomic atlas of the tongue is fused with MRI images of oral carcinoma to demonstrate lateral muscle invasion. The results demonstrate that the styloglossus and hyoglossus lie very close to the surface of the lateral tongue, in some cases passing within 1 mm of the surface mucosa. These extrinsic muscles are readily invaded by even small tumours of the lateral tongue. Strict application of the TNM T4a criteria leads to unnecessary upstaging as these carcinomas do not warrant the prognosis and aggressive treatment of Stage IV disease. Extrinsic muscle invasion should be removed as a T4a criterion for the oral cavity. A separate category, T4a (oral tongue) specifying invasion of the genioglossus is also recommended. This work presented in this thesis is an original contribution to the field of oral cavity cancer research and has determined that there is capacity for improvement in current efforts to determine the natural history and prognosis of oral cavity squamous cell carcinoma. This thesis is the first to examine the role of statistical methodology in oral carcinoma depth of invasion cut-point variability. Further, this work presents an original approach to the prediction of regional metastasis and survival using advanced multivariate modeling techniques. No other work explored MRI-measured volume using the substantial sample size gathered in this thesis. Finally, this work is the first to demonstrate that lateral extrinsic muscle invasion is an unnecessary component of the T4a (oral cavity) classification criteria and should be reconsidered.

Magnetic resonance imaging (MRI) technology has facilitated capturing the dynamics of speech production at fine temporal and spatial resolutions, thus generating substantial quantities of images to be analysed. Manual processing of large MRI databases is labour intensive and time consuming. Hence, to study articulation on large scale, techniques for automatic feature extraction are needed. This thesis investigates approaches for automatic information extraction from an MRI database of dynamic articulation. We first study the articulation by observing the pixel intensity variations in image sequences. The correspondence between acoustic segments and images is established by forced alignment of speech signals recorded during the articulation. We obtain speaker-specific typical phoneme articulations that represent general articulatory configurations in running speech. Articulation dynamics are parametrised by measuring the magnitude of change in intensities over time. We demonstrate a direct correlation between the dynamics of articulation thus measured and the energy of the generated acoustic signals. For more sophisticated applications, a parametric description of vocal tract shape is desired. We investigate different shape extraction techniques and present a framework that can automatically identify and extract the vocal tract shapes. The framework incorporates shape prior information and intensity features in recognising and delineating the shape. The new framework is a promising new tool for automatic identification of vocal tract boundaries in large MRI databases, as demonstrated through extensive assessments. The segmentation framework proposed in this thesis is, to the best of our knowledge, novel in the field of speech production. The methods investigated in this thesis facilitate automatic information extraction from images, either for studying the dynamics of articulation or for vocal tract shape modelling. This thesis advances the state-of-the-art by bringing new perspectives to studying articulation, and introducing a segmentation framework that is automatic, does not require extensive initialisation, and reports a minimum number of failures.

The advent of hyperpolarized magnetic resonance (MR) has provided new potential for real-time visualization of in vivo metabolic processes. The aim of the work in this thesis was to use hyperpolarized substrates to study rapid metabolic processes occurring in the healthy and diseased rat heart. Initial work, described in Chapter 2, optimized the hyperpolarization process to reproducibly generate tracers. Chapter 3 describes use of hyperpolarized 1-13C-pyruvate to investigate in vivo flux through the regulatory enzyme pyruvate dehydrogenase (PDH). Cardiac PDH activity was altered in several physiological and pathological states, namely fasting, type 1 diabetes, and high-fat feeding, and in vivo flux through PDH was measured using hyperpolarized MR. These measurements correlated with measurements of in vitro PDH activity obtained using a validated biochemical assay. The work in Chapter 4 investigated the physiological interaction between hyperpolarized tracer and cardiac tissue. The effect of hyperpolarized 1-13C-pyruvate concentration on its in vivo metabolism was analyzed using modified Michaelis-Menten kinetics. It was found that hyperpolarized MR could non-invasively follow mechanisms of metabolic regulation, in addition to reporting enzyme activity. In Chapter 5, hyperpolarized MR was incorporated into the isolated perfused rat heart. 1-13C-pyruvate in normal and ischaemic hearts revealed significant differences in lactate metabolism, and provided the foundation for a novel intracellular pH probe. Infusion of 2-13C-pyruvate in the isolated rat heart enabled the first real-time visualization of Krebs cycle intermediates. In summary, the work in this thesis has highlighted the potential of hyperpolarized MR to reveal novel information on heart disease.