Machine Learning-Based Prediction of Multiple Sclerosis Diagnosis and Progression

Abstract

Multiple sclerosis (MS) is an autoimmune disorder of the central nervous system that is characterized by the destruction of the myelin sheath and disruption of neurological function. The diagnosis and prognosis of MS continue to be a challenge because of the diverse clinical course and presentation of the disease. The aim of this study is to explore the potential of machine learning algorithms in predicting diagnosis and progression of MS using clinical and demographic information. A preprocessed dataset of medical and lifestyle information of MS patients was used, and some machine learning models like Random Forest, Support Vector Machine (SVM), CatBoost, and XGBoost were trained and evaluated. Results on the test set showed the accuracy of 99% for gradient boosting-based models (CatBoost and XGBoost) and their superiority over the traditional logistic regression and deep learning models mentioned in previous works. The results show that models developed using machine learning algorithms have the potential to make significant contributions to the prediction of MS, as well as to identifying and progressing an early diagnosis and individual treatment approach. Additional future studies will be aimed at using more biomarkers and online patient monitoring information to increase generalizability in models.