Smartphone- and Smartwatch-Based Remote Characterisation of Ambulation in Multiple Sclerosis During the Two-Minute...

Area of Science:

• Biomedical Engineering
• Digital Health
• Neurology

Background:

• Remote monitoring of multiple sclerosis (MS) using consumer technology offers objective insights into disease progression.
• Gait characteristics can be effectively modeled using machine learning (ML) to differentiate MS patient subgroups from healthy controls (HC).

Purpose of the Study:

• To explore the feasibility of using smartphone and smartwatch sensors to assess physical function in people with MS (PwMS).
• To characterize gait-related features for ML-based classification of PwMS subgroups against HC.

Main Methods:

• Collected 24-week Two-Minute Walk Test (2MWT) data from 97 subjects (24 HC, 52 mildly disabled PwMS, 21 moderately disabled PwMS) using smartphones and smartwatches.
• Extracted 156 signal-based gait features and applied LASSO for feature selection.
• Utilized Logistic Regression, Support Vector Machines (SVM), and Random Forest models for classification, comparing smartphone-only, smartwatch-only, and fused data.

Main Results:

• Smartphone-based models achieved the highest classification performance, demonstrating the potential of a single device for remote ambulatory assessment.
• Support Vector Machine (SVM) with Radial Basis Function (RBF) distinguished moderately disabled PwMS from HC (Acc. 82.2%) and mildly disabled PwMS (Acc. 82.3%).
• Mildly disabled PwMS exhibited HC-like gait, making them less distinguishable (Acc. 66.4%).

Conclusions:

• Consumer smartphones provide a viable tool for objective, remote assessment of gait characteristics in people with MS.
• ML models, particularly SVM, can effectively differentiate MS patient subgroups based on sensor-derived gait data.
• Subject-specific gait variability highlights the need for personalized remote monitoring strategies in MS care.