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

Area of Science:

• Neuroscience
• Digital Health
• Artificial Intelligence

Background:

• Cognitive impairment is a significant challenge for individuals with multiple sclerosis (MS).
• Digital interventions offer potential for cognitive improvement in MS, but individual responses vary.
• Personalized approaches are needed to optimize neurorehabilitation for MS patients.

Purpose of the Study:

• To explore the feasibility of using machine learning models to predict responsiveness to computerized cognitive training in MS.
• To identify real-world behavioral predictors of cognitive training response using wearable data.
• To investigate the potential of AI and wearables for personalized neurorehabilitation in MS.

Main Methods:

• Utilized machine learning, specifically linear support vector machine classifiers, to analyze data from Fitbit wearables.
• Trained models on real-world behavioral data to distinguish between responders and non-responders to cognitive training.
• Employed SHAP analysis to identify key features influencing prediction accuracy.

Main Results:

• Achieved a mean balanced accuracy of 0.71 in predicting patient response to cognitive training.
• Identified sleep patterns and cardiac activity metrics as significant behavioral predictors of training responsiveness.
• Demonstrated a link between real-world behavioral data and cognitive outcomes.

Conclusions:

• Machine learning models trained on wearable data can predict cognitive training response in MS patients.
• Wearable-derived sleep and cardiac data are valuable for personalizing neurorehabilitation strategies.
• This approach highlights the potential of AI and digital health to advance precision care in MS.