Personalized Longitudinal Assessment of Multiple Sclerosis Using Smartphones

Area of Science:

• Neurology
• Biomedical Engineering
• Digital Health

Background:

• Personalized longitudinal assessment is crucial for managing multiple sclerosis (MS).
• Identifying individual disease trajectories and profiles aids in optimal treatment adaptation.
• Current methods may lack automated, continuous monitoring capabilities.

Purpose of the Study:

• To develop a novel, automated longitudinal model for assessing multiple sclerosis (MS) disease trajectories.
• To utilize smartphone sensor data for remote, personalized MS monitoring.
• To identify digital markers for predicting MS progression.

Main Methods:

• Collected smartphone sensor data on gait, balance, and upper extremity function.
• Employed data imputation techniques for missing values.
• Utilized generalized estimation equations to discover MS markers.
• Ensembled parameters from multiple datasets to create a unified predictive model.
• Incorporated subject-specific fine-tuning for improved accuracy in severe cases.

Main Results:

• The proposed model demonstrates promise for personalized longitudinal MS assessment.
• Features related to gait, balance, and upper extremity function were identified as potential digital markers.
• Remote data collection via smartphone assessments can effectively predict MS over time.
• The model achieved accurate forecasting in previously unseen individuals with MS.

Conclusions:

• Smartphone-based sensor data can be leveraged for effective, personalized longitudinal monitoring of MS.
• Gait, balance, and upper extremity function are valuable digital biomarkers for predicting MS progression.
• The developed model offers a promising tool for automated and adaptive MS management.