Dynamic Graphs Analysis of EEG

In this study, we investigate the role of temporal dynamics in brain connectivity for the classification of electroencephalography (EEG) signals using dynamic Graph Neural Networks (GNNs). The proposed methods are evaluated on several large-scale EEG datasets focusing on abnormality and epilepsy detection. The implemented models achieve competitive performance on unseen test subjects across all datasets, outperforming existing graph-based baselines in terms of accuracy and F1 score. We explore multiple GNN architectures specifically designed to capture temporal variations in graph-structured data, demonstrating their effectiveness in modeling dynamic brain activity. Beyond classification, we employ graph theoretical metrics such as network efficiency and node degree to analyze temporal changes in brain networks across EEG time windows. The aim is to characterize differences between pathological and healthy groups at both the node and network levels. In particular, we examine epileptic and healthy subject groups, highlighting differences in local network efficiency and node degree, with statistical significance confirmed via F-tests.