Premium
Lesion features on magnetic resonance imaging discriminate multiple sclerosis patients
Author(s) -
Hurtado Rúa Sandra M.,
Kaunzner Ulrike W.,
Pandya Sneha,
Sweeney Elizabeth,
Tozlu Ceren,
Kuceyeski Amy,
Nguyen Thanh D.,
Gauthier Susan A.
Publication year - 2022
Publication title -
european journal of neurology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.881
H-Index - 124
eISSN - 1468-1331
pISSN - 1351-5101
DOI - 10.1111/ene.15067
Subject(s) - medicine , magnetic resonance imaging , lesion , multiple sclerosis , expanded disability status scale , hierarchical clustering , radiology , nuclear medicine , pathology , cluster analysis , artificial intelligence , psychiatry , computer science
Background Magnetic resonance imaging (MRI) provides insight into various pathological processes in multiple sclerosis (MS) and may provide insight into patterns of damage among patients. Objective We sought to determine if MRI features have clinical discriminative power among a cohort of MS patients. Methods Ninety‐six relapsing remitting and seven progressive MS patients underwent myelin water fraction (MWF) imaging and conventional MRI for cortical thickness and thalamic volume. Patients were clustered based on lesion level MRI features using an agglomerative hierarchical clustering algorithm based on principal component analysis (PCA). Results One hundred and three patients with 1689 MS lesions were analyzed. PCA on MRI features demonstrated that lesion MWF and volume distributions (characterized by 25th, 50th, and 75th percentiles) accounted for 87% of the total variability based on four principal components. The best hierarchical cluster confirmed two distinct patient clusters. The clustering features in order of importance were lesion median MWF, MWF 25th, MWF 75th, volume 75th percentiles, median individual lesion volume, total lesion volume, cortical thickness, and thalamic volume (all p values <0.01368). The clusters were associated with patient Expanded Disability Status Scale (EDSS) ( n = 103, p = 0.0338) at baseline and at 5 years ( n = 72, p = 0.0337). Conclusions These results demonstrate that individual MRI features can identify two patient clusters driven by lesion‐based values, and our unique approach is an analysis blinded to clinical variables. The two distinct clusters exhibit MWF differences, most likely representing individual remyelination capabilities among different patient groups. These findings support the concept of patient‐specific pathophysiological processes and may guide future therapeutic approaches.