Federated GNNs for EEG-Based Stroke Assessment

Machine learning (ML) has become essential in supporting clinical decision- making processes, offering enhanced diagnostic capabilities and personalized treatment plans. However, outsourcing medical records to train ML models using patient data raises legal, privacy, and security concerns. Federated learning has emerged as a promising paradigm for collaborative ML, meeting healthcare insti- tutions’ requirements for robust models without sharing patient data. This study proposes a novel method that combines Federated Learning and Graph Neural Networks (GNNs) to predict stroke severity using EEG signals across distributed medical institutions. Our approach enables multiple hospitals to jointly train a shared GNN model on their local EEG data without exchanging patient information. It leverages masked self-attention mechanism to capture salient brain connectivity patterns and EdgeSHAP to provide post-hoc explanations of the neurological states after a stroke. We evaluated our method on EEG recordings from four institutions, achieving a mean absolute error (MAE) of 3.23 in predicting stroke severity, close to the average error made by human experts (MAE of 3.0). This demonstrates the method’s effectiveness in providing accurate and explainable predictions while maintaining data privacy.