My research lies at the intersection of statistical learning and differential geometry with applications in neuroscience and clinical trial designs. We develop statistical methods for the analysis of structured data such as images or meshes. In particular, we invented the technique named "disease course mapping" for the analysis of longitudinal data, e.g. repeated data of the same subjects over time. This technique allowed the construction of personalised models of disease progression for several neurodegenerative diseases: Alzheimer, Parkinson or Huntington. These models are used to spot the optimal moment to test new therapies in each patient.