Exploring the Personalisation of Digital Cognitive Rehabilitation in Multiple Sclerosis Through Wearable Data and Machine Learning.

Cognitive impairment is a frequent, disabling symptom of multiple sclerosis (MS). Digital interventions may help people with MS maintain or improve cognition, but effectiveness varies between individuals. This exploratory study examines the feasibility of using machine learning models trained on Fitbit-derived real-world data to predict responsiveness to a computerised cognitive training intervention. Linear support vector machine classifiers achieved a mean balanced accuracy of 0.71 in distinguishing responders from non-responders, while SHAP analysis identified sleep and cardiac activity features as key behavioural predictors. y linking real-world behavioural metrics with cognitive outcomes, this work shows how artificial intelligence and wearables may personalise neurorehabilitation and bridge the gap between data and precision care in MS.