کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Exploring Graph Biomarkers and Connectivity in Epilepsy Through Graph Learning

نویسندگان :

Ali Khosravipour¹ Sepideh Hajipour Sardouie²

1- دانشگاه صنعتی شریف 2- دانشگاه صنعتی شریف

کلمات کلیدی :

Epileptic Seizures،EEG،Brain Connectivity،Power Spectral Density (PSD)،Graph Biomarkers،Graph Learning،Network Measures،Seizure Detection

چکیده :

Epileptic seizures are characterized by abnormal neural activity that disrupts brain connectivity, causing significant and transient changes in interactions between different brain regions. In this paper, we propose two potential graph biomarkers to assist in seizure detection. Our analysis consists of two main parts. First, we construct graphs representing the three distinct seizure phases: preictal, ictal, and postictal, using data from the CHB-MIT EEG dataset. This process involves three steps: calculating the power spectral density (PSD) of the EEG signals as features, learning weighted graphs based on these features, and sparsifying the graphs by retaining only the edges with statistically significant weight changes across the three phases. In the second part, we investigate two key network measures, characteristic path length (CPL) and node strength, assessing their potential as graph biomarkers by analyzing their behavior across the graphs for all three phases. The results show a substantial decrease in both CPL and strength values during the ictal phase compared to the preictal and postictal phases, reflecting impaired integration and disrupted connectivity in the brain during seizures. These findings show that CPL and node strength could be useful graph biomarkers for seizure detection, offering helpful insights for improving epilepsy diagnosis and treatment.