Machine learning integration of MRI and gait reveals mobility phenotypes in multiple sclerosis

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Mobility impairment is a key indicator of multiple sclerosis (MS) progression, but its diverse nature and underlying causes remain incompletely understood.
• Conventional gait assessments in MS patients often lack the precision to fully capture complex gait abnormalities.
• Integrating advanced quantitative gait analysis, MRI-derived metrics, and machine learning (ML) offers a promising approach to uncover novel mobility phenotypes and their links to disease mechanisms.

Purpose of the Study:

• To identify and characterize distinct mobility phenotypes in people with MS (pwMS) using a combination of spatiotemporal gait parameters and MRI-derived features.
• To leverage unsupervised ML clustering to reveal patterns of gait abnormalities and their association with disease characteristics.
• To explore how quantitative MRI metrics and gait analysis can be integrated to better phenotype clinical impairments in pwMS.

Main Methods:

• 1026 pwMS underwent comprehensive gait assessments and quantitative MRI between 2018 and 2023.
• Principal component analysis was used for dimensionality reduction, followed by k-means clustering to identify distinct gait phenotypes.
• Clusters were statistically compared using demographic, clinical, and MRI features via Kruskal-Wallis and Chi-square tests.

Main Results:

• Four distinct gait clusters were identified, ranging from efficient gait with high grey matter fractions to severely unstable gait with profound disability.
• An intermediate cluster (Cluster 3) exhibited increased cadence and shorter strides, correlated with higher lesion burden and lower brain parenchymal fraction, suggesting compensatory gait patterns.
• Clinical measures, including progressive MS, disability scores, and self-reported impairment, strongly aligned with the identified gait instability clusters.

Conclusions:

• Quantitative MRI metrics and spatiotemporal gait analysis, when integrated, can effectively phenotype clinical impairments in pwMS.
• ML-driven analysis revealed a novel intermediate mobility phenotype characterized by specific gait and MRI abnormalities, potentially indicating adaptive mechanisms.
• This ML approach can detect subtle mobility alterations in MS that may not be apparent through conventional clinical evaluation.