Utilizing Aerobic Capacity Data for EDSS Score Estimation in Multiple Sclerosis: A Machine Learning Approach.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• The Expanded Disability Status Scale (EDSS) is crucial for monitoring multiple sclerosis (MS) progression and treatment efficacy.
• Inconsistent EDSS scoring among physicians poses a challenge to reliable patient assessment.
• Developing autonomous solutions is necessary to improve the objectivity and reliability of EDSS evaluations.

Purpose of the Study:

• To propose a machine learning (ML) approach for predicting EDSS scores in patients with MS (PwMS).
• To utilize aerobic capacity and cardiovascular data as input features for ML models.
• To enhance the reliability of EDSS assessments and mitigate complications from scoring discrepancies.

Main Methods:

• Collected cardiovascular and aerobic capacity data, including ventilation, heart rate, and oxygen consumption, from PwMS.
• Employed machine learning algorithms: CatBoost, Gradient Boosting Machine (GBM), Extreme Gradient Boosting (XGBoost), and Decision Trees (DT).
• Evaluated model performance using metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared.

Main Results:

• The XGBoost algorithm demonstrated the highest accuracy in predicting EDSS scores.
• Achieved performance metrics: MAE of 0.26, RMSE of 0.4, and R-squared of 0.68.
• The ML approach effectively predicted EDSS scores based on physiological data.

Conclusions:

• Machine learning, particularly XGBoost, offers a reliable method for predicting EDSS scores in PwMS.
• Aerobic capacity and cardiovascular parameters are significant predictors of MS disability.
• This ML-based approach can improve the consistency and objectivity of EDSS assessments.