Schizophrenia Detection Using Interconnected Graph-Based Features From EEG Signals

Schizophrenia (Sz) is a complex mental disorder characterized by disruptions in thought processes, perceptions, and emotional regulation. It is imperative to identify schizophrenia at an early stage with precision due to its significance in medical contexts. This research article introduces a graph-based approach for schizophrenia detection using the graph Fourier transform (GFT) and spectral graph wavelet transform (SGWT) applied to multichannel electroencephalogram (EEG) signals. Unlike traditional methods that focus on individual signal components and often neglect the functional connectivity within the brain in cases of brain-related disorders, this research article introduces an innovative approach rooted in graph signal processing (GSP). In this article, each EEG signal channel is linked to a node in the graph, with the EEG time series for each node represented within the graph structure. This approach considers the interconnections between different brain regions to capture the functional connectivity among these nodes, allowing for a more nuanced understanding of neural network abnormalities associated with schizophrenia. To assess the efficacy of the suggested approach, various machine learning models, including support vector machine (SVM), decision tree (DT), random forest (RF), k-nearest neighbors (kNNs), and extreme gradient boosting (XGBoost), are employed on a dataset consisting of 14 individuals with good health and 14 individuals diagnosed with schizophrenia. Through the utilization of graph signal representation and the inclusion of the SGWT feature, a classification accuracy of 97.22% using SVM classifiers has been attained.