Robustness of diffusion MRI spherical deconvolution analyses in patients with white matter lesions: from fiber orientations to connectivity

Diffusion Magnetic Resonance Imaging (dMRI) enables the non-invasive investigation of the microstructural organization of brain tissue by measuring the diffusion of water molecules. Among the models derived from this technique, spherical deconvolution has emerged as one of the most accurate methods for resolving complex fiber configurations in white matter, forming the basis for tractography and brain connectivity analyses. While originally developed for the study of healthy brains, these methods are increasingly applied in clinical and pathological contexts. However, the presence of structural abnormalities, such as white matter hyperintensities (WMHs), may distort the diffusion signal and consequently affect the accuracy of fiber orientation estimates and their downstream applications. This thesis explores how WMHs influence diffusion MRI spherical deconvolution and tractography in patients with genetic small vessel disease. The study is based on multi-shell diffusion data acquired at baseline and after a two-year follow-up in a cohort of 33 subjects. White matter voxels were classified into four distinct categories according to lesion evolution over time—chronic lesions, newly formed lesions, peri-lesional areas, and normal-appearing white matter. Within this framework, three complementary analyses were designed: (1) the impact of WMHs on the microstructural features of the fiber orientation distributions (FODs), such as the number, amplitude, and orientation of diffusion peaks; (2) the effect of lesions on tractography reconstruction, evaluating trajectory consistency across time and sensitivity to different amplitude thresholds; and (3) the influence of WMHs on whole-brain structural connectivity derived from graph-based analysis. Through these analyses, the study aims to provide insight into the robustness and potential limitations of diffusion-based models in lesioned brains. Preliminary findings indicate that, although local diffusion properties within WMHs are altered—showing reduced FOD amplitude, the macro-scale fiber architecture remains preserved over time. Moreover, the results show high sensitivity to the applied FOD amplitude threshold, underscoring the importance of careful parameter selection when interpreting diffusion-derived metrics in pathological tissue. Overall, this work contributes to a better understanding of how pathological white matter changes can bias diffusion-derived metrics, ultimately guiding more reliable applications of tractography and connectivity analysis in clinical research.