The feasibility of making spatially resolved measurements of blood velocity using a pulsed photoacoustic Doppler technique has been investigated. Doppler time shifts were quantified via cross-correlation of photoacoustic waveform pairs. The waveforms were generated within a blood-simulating phantom using pairs of light pulses and detected using an ultrasound transducer. Two types of blood-simulating phantom were investigated. The first was a rotating wheel phantom consisting of micron-scale absorbers imprinted on an acetate sheet and moved at known velocities; this simulated plug flow. A time-correlation data processing scheme was used to quantify velocities in the range 0.15 to 1.5 m/s with accuracies as low as 1% and a measurement resolution <4%. The transducer beam width determines a maximum measurable velocity |Vmax| beyond which correlation is lost due to absorbers moving out of the focal beam between the two laser pulses. Resolution and |Vmax| can be scaled to much lower velocities such as those encountered in microvasculature (< 50 mm/s). Velocities in this range were investigated for the second type of phantom comprising absorbers, such as red blood cells or microspheres, flowing in a suspension within a transparent tube; this demonstrated non-plug flow. The absorber-filled tube could also be manually shifted for direct comparison between the plug and non-plug flow cases. Laminar flow gave rise to under-reading of the known velocities, which was exacerbated by increasing absorber concentrations and tube diameters, presumably due to inadequate light penetration into the tube. A novel signal processing scheme (“waveform segmentation”) was developed to surmount this difficulty, and also adds the potential for mapping out the flow velocity profile across the tube. The results show that the absorber spatial heterogeneity can be resolved even using a relatively low frequency detector, and thus pave the way for applying the cross-correlation technique to make blood velocity measurements in vivo.

The hippocampus is one of the most extensively studied structures of the brain, and yet its diverse set of cognitive functions is still being identified. It is particularly associated with episodic memory in humans and spatial processing in animals, but more recently it has also been implicated in processes beyond memory, including imagination of fictitious and future experiences and even visual perception. The relationship between these cognitive functions, and the underlying involvement of the hippocampus, is not well understood. In this dissertation, I draw upon and extend the hypothesis that the feature common to episodic memory, imagination and visual perception, which necessitates them all to engage the hippocampus, is scene construction – the creation of internal representations of spatially coherent scenes. In a series of experiments, I improved our understanding of scene construction by directly comparing the imagination of scenes (the hippocampus being driven endogenously) against scene perception (being driven by visual stimuli), and further by comparing imagination against memory recall, using functional Magnetic Resonance Imaging (fMRI). Two experiments had the added advantage of using high-resolution structural MRI (0.5mm3) and fMRI (1.5mm3), for the first time making inferences about scene processing at the level of the sub-units of the hippocampus, termed the hippocampal subfields. Furthermore, by capitalising on recent developments in connectivity analysis (Generalized Psychophysiological Interactions and Stochastic Dynamic Causal Modelling), I examined the interactions of the hippocampus with other brain regions, and inferred the flow of information between the hippocampal subfields using fMRI. Together, my findings provide new insights into the relationship between scene perception, imagination and memory, and develop our understanding of the heterogeneous functional anatomy of the hippocampus. This has important implications for understanding why patients with hippocampal lesions may lose their ability to construct scenes in their imagination and recall episodes from their past.

Dynamic magnetic resonance imaging (MRI) consists of collecting multiple MR images in time, resulting in a spatio-temporal signal. However, MRI intrinsically suffers from long acquisition times due to various constraints. This limits the full potential of dynamic MR imaging, such as obtaining high spatial and temporal resolutions which are crucial to observe dynamic phenomena. This dissertation addresses the problem of the reconstruction of dynamic MR images from a limited amount of samples arising from a nuclear magnetic resonance experiment. The term limited can be explained by the approach taken in this thesis to speed up scan time, which is based on violating the Nyquist criterion by skipping measurements that would be normally acquired in a standard MRI procedure. The resulting problem can be classified in the general framework of linear ill-posed inverse problems. This thesis shows how low-dimensional signal models, specifically lowrank and sparsity, can help in the reconstruction of dynamic images from partial measurements. The use of these models are justified by significant developments in signal recovery techniques from partial data that have emerged in recent years in signal processing. The major contributions of this thesis are the development and characterisation of fast and efficient computational tools using convex low-rank and sparse constraints via proximal gradient methods, the development and characterisation of a novel joint reconstruction–separation method via the low-rank plus sparse matrix decomposition technique, and the development and characterisation of low-rank based recovery methods in the context of dynamic parallel MRI. Finally, an additional contribution of this thesis is to formulate the various MR image reconstruction problems in the context of convex optimisation to develop algorithms based on proximal splitting methods.

Non-rigid registration of serial head and neck FDG PET/CT images from a combined scanner can be problematic. Registration techniques typically rely on similarity measures calculated from voxel intensity values; CT-CT registration is superior to PET-PET registration due to the higher quality of anatomical information present in this modality. However, when metal artefacts from dental fillings are present in a pair of CT images, a nonrigid registration will incorrectly attempt to register the two artefacts together since they are strong features compared to the features that represent the actual anatomy. This leads to localised registration errors in the deformation field in the vicinity of the artefacts. Our objective was to develop a registration technique which overcomes these limitations by using combined information from both modalities. To study the effect of artefacts on registration, metal artefacts were simulated with one CT image rotated by a small angle in the sagittal plane. Image pairs containing these simulated artifacts were then registered to evaluate the resulting errors. To improve the registration in the vicinity where there were artefacts, intensity information from the PET images was incorporated using several techniques. A well-established B-splines based non-rigid registration code was reworked to allow multicomponent registration. A similarity measure with four possible weighted components relating to the ways in which the CT and PET information can be combined to drive the registration of a pair of these dual-valued images was employed. Several registration methods based on using this multicomponent similarity measure were implemented with the goal of effectively registering the images containing the simulated artifacts. A method was also developed to swap control point displacements from the PET-derived transformation in the vicinity of the artefact. This method yielded the best result on the simulated images and was evaluated on images where actual dental artifacts were present.

The immense challenge of annotating the entire mouse genome has stimulated development of cutting-edge imaging technologies in a drive for novel information. These techniques promise to improve our understanding of the genes involved in embryo development, at least one third of which have been shown to be essential. Aligning advanced imaging technologies with biological needs will be fundamental to maximising the number of phenotypes discovered in the coming years. International efforts are underway to meet this challenge through an integrated and sophisticated approach to embryo phenotyping, which will include advanced imaging tools. This thesis investigates advanced imaging methodologies and computational image analysis techniques for mouse embryo phenotyping using magnetic resonance imaging (MRI). Additionally, the novel application of an emerging method called photoacoustic imaging is demonstrated for imaging mouse embryos in utero. First, the lack of tissue staining capabilities that currently limits embryo MR imaging was addressed by investigating the MRI staining properties of two readily available contrast agents and their underlying contrast enhancement mechanisms. A methodological framework was developed for high-throughput screening of embryos using diffusion MRI and implemented to study the splotch mouse model of human neural tube defects. A validation study was carried out to comprehensively assess the accuracy of volumetric measurements generated using a computational image analysis method called segmentation propagation. Finally, an all-optical photoacoustic scanner and novel time-reversal image reconstruction algorithm were developed, enabling photoacoustic imaging of whole embryos in utero. Overall, this thesis presents advanced imaging methodologies and computational image analysis techniques that may form an essential part of the toolkit available for annotating the mouse genome and facilitate identification of novel phenotypes in the coming years.

Lateral flow immunoassays have large applications in the diagnostic and food industry. A biosensor made of electrostatic self-assembled multilayers of polyphenol oxidase-polyallylamine was studied in order to assess its relevance to a future application in immunoassays. Electrostatic self-assembled multilayers composed only of polyelectrolytes were also investigated to provide a model. This model system allowed understanding the internal properties of this structure. Variations of solution pH and modification of the internal local pH and surface charge density by electrochemistry provided interesting information about the mechanisms involved in multilayer stability to desorption. This stability depends on both multilayers surface charge density and interpenetration of each layer in its neighbouring ones for multilayers with more than nine layers. Enzyme-polyelectrolyte multilayers were more sensible to pH variation as polyallylamine only participated in layer interpenetration. Enzyme activity was also studied by chronoamperometry. Enzymatic kinetics was determined as well as the effect of enzyme activity on multilayer structure. A classic lateral flow sandwich immunoassay (such as a pregnancy test) was studied. A mathematical model describing lateral flow assay was developed and adapted to the studied system. Assay characteristics and properties were described by using different antigen concentrations, flow velocities, particle nature and sizes. Reaction kinetics was determined with the described model. Several factors such as particle size and nature, antibody concentration on the test line, membrane porosity and solution pH were varied in order to optimise signal intensity. Surface treatments of nitrocellulose membrane were also done to enhance the signal. Particle binding on the test line was also investigated to better understand the front edge effect as it is a major problem in lateral flow immunoassays and can involve a wrong signal interpretation.

Prion protein (PrP) is the only factor known to be essential in the pathogenesis of the transmissible spongiform encephalopathies (TSEs) or prion diseases. The cellular isoform (PrPc), a GPI-anchored sialoglycoprotein of unknown function, has an identical primary structure to the disease-associated conformer (PrPSc). Thus, animals are tolerant to PrPSc and TSEs do not trigger a classical immune response. Vaccine development for human TSEs requires elucidation of the immunodominant human T cell epitopes within PrP. Further, successful immunotherapy requires that the function of PrPc in lymphocytes is understood, as therapeutic targeting of prion protein risks interfering with immune function. Peripheral blood leukocytes from healthy donors were cultured with PrP sequence peptides to elicit proliferative and cytokine responses. Responses were seen to peptides clustered around the position 129 polymorphism and the C-terminus, in accordance with a predictive algorithm. The substitution of methionine by valine at position 129 altered both epitope immunogenicity and cytokine profile. Studies in murine T cell activation models demonstrated transcriptional and late surface protein upregulation of PrPc. Memory T cells expressed higher PrPc levels than naive cells and there was also a strong correlation at both protein and transcriptional levels between expression of PrPc and the regulatory T cell marker, Foxp3. Embryonic deletion of Prnp did not lead to deficits in T cell conjugation, proliferation or cytokine production, although memory cell numbers were slightly reduced. In PrP*7" mice regulatory T cells developed normally but may have enhanced suppressor function. However, neither PrP ablation nor anti-PrP monoclonal antibodies altered the phenotype of T cell mediated autoimmune disease. These findings demonstrate that tolerance to PrP is not complete in humans and raise the prospect of generating protective immunity through vaccination. However, PrPc is a potentially important memory, regulatory and T cell activation antigen, therapeutic disruption of which may precipitate immunopathology.

The spectrum of diseases caused by mitochondrial dysfunction is very broad and encompasses the archetypal mtDNA mutation diseases, mutations of nuclear genesencoding mitochondrial proteins (including those of the oxidative phosphorylation system), and a variety predominantly neurodegenerative diseases in which the primary cause of mitochondrial dysfunction remains undefined. The last two decades have seen an explosion in our understanding of the archetypal i mitochondrial disorders. Attention has now focused on the nuclear encoded mitochondrial disorders. Furthermore, nuclear factors may be of significance in the pathogenesis of the archetypal disorders associated with mitochondrial DNA mutations. These conditions are typified by their clinical diversity and poor phenotype-genotype correlation. One of several potential explanations for this is that nuclear genes determine the fate of mtDNA mutations, or that secondary mtDNA mutations have a modulating effect upon the expression of the primary mutation. In this thesis I have sought to address several aspects of the biochemical and clinical features of mitochondrial diseases. In chapter 3 cell cybrids have been used to study the role of the nuclear genome on the biochemical expression of mtDNA mutations in an attempt to understand potential influences on phenotypic expression. An extension of this was the use of xenomitochondrial cybrids to analyse nuclear-mitochondrial interactions and the function of the respiratory chain. At the biochemical/clinical interface, skeletal muscle from patients with focal dystonia has been used as a model to investigate the role that mitochondrial dysfunction might play in this movement disorder. Finally, the clinical role of therapy for mitochondrial disorders has been investigated in the context of Friedreich's ataxia (FRDA). Existing rating scales have been assessed and new ones developed to lay a firm foundation for evaluating disease-modifying therapies. These have been piloted in a long term intervention trial for FRDA.

Attenuation correction is essential for reliable interpretation of emission tomography; however the use of transmission measurements to generate attenuation maps is limited by availability of equipment and potential mismatches between the transmission and emission measurements. This work investigates the possibility of estimating an attenuation map using measured scatter data without a transmission scan. A scatter model has been developed that predicts the distribution of photons which have been scattered once. The scatter model has been used as the basis of a maximum likelihood gradient ascent method (SMLGA) to estimate an attenuation map from measured scatter data. The SMLGA algorithm has been combined with an existing algorithm using photopeak data to estimate an attenuation map (MLAA) in order to obtain a more accurate attenuation map than using either algorithm alone. Iterations of the SMLGA-MLAA algorithm are alternated with iterations of the MLEM algorithm to estimate the activity distribution. Initial tests of the algorithm were performed in 2 dimensions using idealised data before extension to 3 dimensions. The basic algorithm has been tested in 3 dimensions using projection data simulated using a Monte Carlo simulator with software phantoms. All soft tissues within the body have similar attenuation characteristics and so only a small number of different values are normally present. A Level-Set technique to restrict the attenuation map to a piecewise constant function has therefore been investigated as a potential way to improve the quality of the reconstructed attenuation map. The basic SMLGA-MLAA algorithm contains a number of assumptions; the effect of these has been investigated and the model extended to include the effect of photons which are scattered more than once and scatter correction of the photopeak. The effect of different phantom shapes and activity distributions has been assessed and the final algorithm tested using data acquired using a physical phantom.

Malignant melanoma represents the most aggressive form of skin cancer. Although early stage disease is treatable through surgical excision alone, late stage tumours frequently metastasise to the liver, at which point treatment options remain limited. Migration of melanoma towards metastatic sites has been shown to be associated with the CXCR4-CXCL12 chemokine axis. The chemokine receptor CXCR4 is expressed by melanoma cells and the chemokine CXCL12 is secreted by the liver. Expression of CXCL12 has been shown to be increased in liver fibrosis and therefore it was hypothesized that cells involved in liver damage may promote melanoma metastasis to this organ. CXCR4 and CXCL12 expression in melanoma and liver cells in vitro and in vivo was examined by RT-PCR, Western blotting and immunohistochemical staining. Chemotaxis assays were performed to test the ability of AMD11070 to inhibit migration of melanoma cells. Quantitative RT-PCR and Western blotting determined the influence of different fibrosis models (Carbon tetrachloride (CCl4), Bile Duct Ligation (BDL) and Methapyrilene (MP)) on CXCL12 expression. Furthermore, the migration of melanoma was examined in animal models of liver injury. Results showed that melanoma cells and different liver cell types (myofibroblasts and biliary epithelial cells) express both CXCR4 and CXCL12. CXCR4 expression in melanoma promoted migration of tumour cells towards CXCL12 secreting liver cells and AMD11070 inhibited this. CXCR4 and CXCL12 proteins of varying sizes were observed in vivo suggesting that post translational modifications of these proteins may occur. CXCL12 expression increased in three models of chronic liver injury; CCl₄, BDL and MP. In an animal model, murine melanoma cells metastasized to the lungs and to both the fibrotic and normal liver. These findings suggest that the reduction of liver cells secreting CXCL12 may help to reduce melanoma metastasis to this organ.