Related Experiment Video
Updated: Aug 11, 2026
05:44
Modeling Multiple Sclerosis in the Two Sexes: MOG35-55-Induced Experimental Autoimmune Encephalomyelitis
Published on: October 13, 2023
Manual federated simulation for multiple sclerosis integrating XGBoost algorithm with SHAP explanation
Hagar E Ghazy1,2, Zainab H Ali3,4, Tamer Medhat5
1Department of Artificial Intelligence, Faculty of Artificial Intelligence, Delta University for Science and Technology, Gamasa, Dakahlia, 35712, Egypt. hagar.saleh@deltauniv.edu.eg.
Scientific Reports
|July 30, 2026
Summary
This study developed a privacy-preserving machine learning model for predicting multiple sclerosis (MS) progression in Clinically Isolated Syndrome (CIS) patients. The explainable AI framework achieved high accuracy, supporting early diagnosis and enhancing clinical trust.
Area of Science:
- Neurology
- Artificial Intelligence
- Medical Informatics
Background:
- Multiple sclerosis (MS) is a chronic central nervous system autoimmune disorder requiring early diagnosis.
- Clinically Isolated Syndrome (CIS) is an early stage of MS, necessitating predictive models for disease progression.
- Current diagnostic and prognostic methods for MS can be improved with advanced computational approaches.
Purpose of the Study:
- To develop a privacy-preserving, federated, and explainable Machine Learning (ML) framework for predicting MS progression in CIS patients.
- To enhance the accuracy and transparency of MS progression prediction models.
- To facilitate secure data collaboration in decentralized clinical settings for MS research.
Main Methods:
- Utilized Multivariate Imputation by Chained Equations (MICE) for handling missing data while preserving feature dependencies.
- Employed the Extreme Gradient Boosting (XGBoost) algorithm for classification of MS progression.
- Integrated Explainable Artificial Intelligence (XAI) techniques, specifically Shapley Additive Explanations (SHAP), for model interpretability.
- Implemented an in silico federated learning (FL) framework to ensure data confidentiality and simulate decentralized environments.
Main Results:
- The ML model demonstrated strong predictive performance with high accuracy (up to 96.7%) and ROC-AUC (up to 99%) during training.
- Validation and test sets showed robust accuracy (92.5% and 81.8%, respectively) with significant AUC values (88%).
- The federated learning simulation maintained competitive performance, achieving 76.3% accuracy and 83.9% AUC.
Conclusions:
- The proposed privacy-preserving, explainable ML framework effectively predicts MS progression in CIS patients.
- The approach enhances clinical trust through model interpretability and supports secure data collaboration via federated learning.
- This contributes to more informed clinical decision-making, transparent patient care, and improved outcomes for individuals with MS.