A hybrid CNN-GNN approach for enhanced early detection of schizophrenia using EEG data

Schizophrenia, a complex neuropsychiatric disorder, poses significant challenges in early detection due to its impact on cognitive, emotional, and social functioning. Hybrid approach integrating CNNs and Graph Neural Networks has shown promise in addressing this challenge. CNNs excel at capturing local spatial-temporal features from EEG signals, while GNNs are designed to model the brain’s connectivity patterns, providing a holistic representation of brain function. This study presents a hybrid CNN-GNN model that extracts both spatial-temporal features and models global brain connectivity for schizophrenia detection. Using EEG data from the Lomonosov Moscow State University (LMSU) dataset, this model achieves high accuracy (99.12%), sensitivity (98.83%), and specificity (99.35%). An improvement in schizophrenia detection, the CNN-GNN hybrid model gives doctors a more precise and trustworthy instrument for early diagnosis.