Depicting multiple sclerosis disease course using lesion parenchymal fraction: a quantified expression of the...

The topographical model of multiple sclerosis proposes that functional reserve compensates for multiple sclerosis lesions, and that disability accumulation is the result of an insidious unmasking of deficits referable to lesion burden. To utilize topographical principles to establish an imaging metric-lesion parenchymal fraction (LPF)-defined as lesion volume divided by parenchymal volume in the same regional compartment, and to assess the relationship of LPF with disability. One hundred patients with relapsing-remitting or secondary progressive multiple sclerosis were evaluated; clinical and MRI data were longitudinally acquired from 2011-19. Lesion and parenchymal volumes in brain and cervical cord were processed using icobrain/icospine pathways, parsed into topographical compartments, and regional LPF was computed. Performance of LPF-based linear models was evaluated using Pearson correlation and root mean squared error between measured and estimated disability scores (expanded disability status score). To establish relative contributions of LPF in cervical, infratentorial and cerebral compartments, a density plot of weight distributions was generated. Individual disability scores and compartment-weighted LPF trajectories were rendered using matplotlib in Python. MRI and clinical data from 78 patients were sufficient for modelling: 39 remaining relapsing-remitting and 39 progressing to secondary progressive multiple sclerosis. The LPF model had the best performance in decoding the disability score using root mean squared error (1.638) and ranked second in Pearson correlation (0.275). Setting the mean coefficient of cerebral LPF to 1, the cervical compartment had the largest coefficient (3.8), followed by infratentorial (2.5). Compartment-weighted cumulative LPF values depict multiple sclerosis disease trajectory longitudinally on a per-patient basis. This visualization is shown for patients who transitioned from relapsing-remitting to secondary progressive phenotypes; non-progressing patients; and outliers where the LPF model does not approximate the disability score trajectory. We developed and evaluated LPF as an MRI-quantified expression of the topographical model of multiple sclerosis, a first effort to operationalize this model to depict individual disease course. That LPF from spinal cord and infratentorial compartments conferred respectively 3.8 and 2.5 more impact on the expanded disability status score than the cerebral hemispheres emphasizes the importance of lesion topography. Implications of outliers are instructive regarding current model limitations; refinement using additional clinical and imaging metrics could allow application to individual patients.