tsta

فارسی

Home / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

EEG Graph Construction: A Comparative Analysis for Classification Application

Authors :

Kiana Kalantari¹ Mohammad Bagher Shamsollahi²

1- دانشگاه صنعتی شریف 2- دانشگاه صنعتی شریف

Keywords :

Electroencephalogram (EEG)،Graph Signal Processing (GSP)،Graph Construction،Classification،Schizophrenia

Abstract :

Electroencephalogram (EEG) is a widely used tool for studying brain function due to its non-invasiveness and high temporal resolution, but traditional analysis methods often struggle to capture the complex spatiotemporal dependencies in EEG data. Graph signal processing (GSP) offers a principled framework to model EEG as signals on graphs, enabling the quantification of neural interactions beyond conventional spectral or linear methods. In this study, we compare three graph construction strategies for representing EEG signals in the context of schizophrenia classification: a Gaussian kernel-based similarity graph, a functional-causal fusion graph, and a Semilocal graph. EEG recordings from healthy controls and schizophrenia patients were preprocessed and segmented, graphs were constructed according to each method, and features, tailored for each graph construction method, were extracted. Classification was performed using a support vector machine with stratified cross-validation. Results show that the functional–causal fusion graph achieved the highest classification accuracy, outperforming both the Gaussian kernel and Semilocal graphs. These findings demonstrate that the choice of graph construction method critically influences classification performance.