Neurodegenerative and neuroinflammatory disorders often involve complex pathophysiological mechanisms that are â to this date â only partially understood. A more comprehensive understanding of those microstructural processes and their characterization in clinical practice are key to ensure effective and individualized patient care. Magnetic resonance imaging (MRI) has become an essential tool to assess neurological pathologies in vivo thanks to its excellent soft tissue contrast. In the last decades, a wide range of MRI clinical decision support tools have been developed, either aimed at replacing tedious and time-consuming radiological reading tasks, or at providing new insights into tissue pathology. However, the clinical adoption of these tools is limited, mainly due to the lack of robustness as a result of the large heterogeneity seen in MRI data.
Among the different strategies aimed at overcoming this limitation, image harmonization techniques have the potential to improve the reliability of automated tools by generating a more homogeneous dataset.
Additionally, the use of population-averaged atlases can also contribute to reducing the inter-site variability that typically characterizes complex acquisition and image processing techniques. Furthermore, by measuring a specific physical parameter, quantitative MRI (qMRI) techniques reduce inter-site variability while potentially providing additional insights into tissue pathology.

This thesis aims at developing new imaging biomarkers for neurodegenerative disorders motivated by real-world clinical challenges and while keeping clinical applicability in mind. To this end, we first explore the use of conditional generative adversarial networks in an image harmonization task and investigate the effect of different reconstruction losses both on image similarity and volumetric consistency using an automated brain morphometry tool. Further pursuing the goal of reducing inter-patient variability, we propose the use of a population-averaged tractography atlas to study structural brain connectivity and validate the clincal relevance of connectivity biomarkers in three different multiple sclerosis (MS) cohorts. Focusing on MS, we further propose novel qMRI-based biomarkers quantifying the extent of microstructural alterations in white matter pathways, and their ability to explain current disability and future progression. The methodological framework established in this thesis work is then complemented with a technique using multi-parametric qMRI alteration maps to differentiate MS lesion subtypes, which provides new insights in microstructural tissue pathology.