Magnetic Resonance Image (MRI) and diffusion MRI in particular have become very useful for assessing brain pathologies and trauma, and for diagnosing neurodegenerative disorders. Although it provides sensitive information pertaining to tissue microstructure and structural connectivity, diffusion MRI has been used in clinical context only for population studies and using either its microstructure information or its connectivity information but not both. In addition, the information is always reduced into a small features vector as statistics are otherwise not available. This results in a lack of specificity and a failure to provide meaningful patient-specific biomarkers for clinical follow-up. We hypothesise that providing statistical methods that account for the whole combined information of tissue microstructure and structural connectivity, and compute patient-specific abnormality detection maps will significantly impact clinical routine. We thus propose to develop and validate novel methods along three major axes: (i) robust estimation of tissue microstructure and structural connectivity from diffusion MRI clinical data acquired in limited amount of time by adding priors, (ii) appropriate mathematical representations of microstructure-augmented structural connectivity embedded in meaningful metric spaces practical for statistics and (iii) extension of two-sample hypothesis testing to the case where one sample is of size 1 to detect patient-specific microstructure and/or connectivity abnormalities. We will apply these new methodologies to neuro-traumatology. This project will allow patient-specific characterisation of brain damage and its influence, providing crucial insights for improving further patient care and treatment adaptation. Methods will be easily applicable to other pathologies. We will facilitate their clinical translation by providing appropriate visualisation software in close collaboration with the clinicians.