Longitudinal MRI follow-up of 300 MS patients was conducted in several Belgian centers equipped with 14 MR systems. Baseline MRI scans and follow-up scans about 1 year later were considered, as well as clinical scores (EDSS, 25FWT, 9HPT and SDMT) at baseline and about 2 years later.

Automated image analysis was conducted in clinical practice with icobrain ms to estimate brain and lesion volumes, and indicate changes over time.

Patients were classified as ‘worse outcome’ after 2 years according to clinical scores (1 increase in EDSS for baseline EDSS up to 5, 0.5 increase in EDSS if baseline EDSS above 5, 20% increase in 25FWT, 20% increase in 9HPT, 10% or 5 points decrease in SDMT), or ‘stable’ (no worsening in clinical scores). MRI parameters were compared between groups, both individually and using multivariate logistic regression.

Sixty patients had worsening in at least one of the clinical parameters (‘worse outcome’ group), while 124 patients had constant or better scores after 2 years (‘stable’ group).

The brain volumes at baseline in worse outcome group were more extreme with respect to icobrain’s healthy population than in the stable group, with median percentiles of 3.7 versus 6.7 for whole brain volume, 3.1 versus 6.5 for gray matter and 93.7 versus 91 for lateral ventricles. New lesion volumes and enlarging lesion volumes had higher median in worse outcome (new: 0.04ml; enlarging: 0.55ml) compared to stable group (0.03ml; 0.27ml).

Multivariate classification with logistic regression, using new and enlarging lesion volumes as most predictive univariate parameters, reached 75% accuracy (AUC 0.66).