Multimodal imaging of soft tissues across length scales

Soft tissues possess an intricate hierarchical architecture whose organization underlies both physiological function and disease. As no single imaging modality can capture every relevant length scale or contrast mechanism, this thesis assesses the effectiveness of multimodal imaging --- combining synchrotron-based X-ray scattering with optical microscopy, X-ray fluorescence, and immunofluorescence --- for investigating soft tissue structure and pathology. Two methodological developments are addressed to begin with: first, a segmentation approach based on principal component analysis and unsupervised K-means clustering is introduced to manage the complexity and volume of multimodal datasets, and second, the influence of soft tissue processing and embedding strategies on its structure is evaluated using X-ray scattering, to optimize sample preparation method. The latter investigation revealed that all methods alter the tissue to varying degrees and that fibrillar collagen is markedly more resilient to processing than the myofibrillar structure. Multimodal imaging is then applied to two pathological systems: (1) murine breast cancer, and (2) failed bioprosthetic heart valves. In the former, the collagen architecture of metastatic and non-metastatic tumours proved highly heterogeneous, with pronounced inter-sample variability that limited definitive conclusions but provided methodological insight into the characterization of heterogeneous tissues. In failed bioprosthetic heart valves, multimodal X-ray characterization revealed calcification-associated alterations in collagen structure and a spatially evolving hydroxyapatite phase that shed light on the mineralization process, alongside a crystalline lamellar lipid phase co-aligned with collagen and found only in HAp-mineralized samples. Finally, a novel polarised-light technique for extracting out-of-plane optical anisotropy is applied to mouse brain tissue, recovering three-dimensional fibre orientation in agreement with known neuroanatomy. Collectively, this thesis demonstrates that multimodal imaging is a powerful but context-dependent tool whose performance depends strongly on tissue heterogeneity, data complexity, and sample preparation. Use of appropriate preliminary characterization, preferably non-destructive techniques together with the incorporation of three-dimensional methods is key to its broader application in the study of soft tissue structure and pathology.