Normative Modelling of Brain Volume for Diagnostic and Prognostic Stratification in Multiple Sclerosis

Interpreting brain structure at the individual level remains a major challenge in neuroimaging, as population variability across age and sex limits the clinical utility of group-level findings. Here, we develop large-scale normative models of regional cortical and subcortical brain volumes from more than 62,000 healthy individuals across the lifespan and apply them to multiple sclerosis (MS) to enable individualised, reference-based assessments of grey matter morphology. We identify a temporally stable yet heterogeneous morphometric phenotype of MS, expressed as concordant deviations from age- and sex-adjusted reference values. Individual deviation profiles are clinically informative: both the magnitude and cumulative burden of lower-than-reference volumes are associated with disability cross-sectionally and longitudinally. Moreover, the profiles can be translated into interpretable stratification rules linked to disability trajectories and relapse dynamics. This work reframes known structural abnormalities into stable, individual-level deviation profiles, demonstrating how normative modelling can move neuroimaging beyond group averages toward clinically interpretable inference. Together, these findings establish a generalisable framework for translating population-level neuroimaging data into individual-level phenotypes with potential beyond multiple sclerosis.