Early Diagnosis of Schizophrenia in EEG Signals Using One Dimensional Transformer Model

Schizophrenia (SZ) is a complex mental disorder, hallmarked by symptoms including delusions, hallucinations, disorganized speech, cognitive impairments, and diminished motivation. Electroencephalography (EEG) recordings have become a critical tool for clinicians and psychologists in diagnosing SZ. Nonetheless, interpreting EEG data to diagnose SZ presents significant challenges for specialists, leading to increased interest in leveraging artificial intelligence (AI) for early detection. This study introduces a novel approach for SZ detection from EEG signals utilizing a transformer-based architecture. The methodology encompasses dataset selection, preprocessing, feature extraction, and classification phases. The RepOD dataset was employed for all simulations. Preprocessing entails filtering, normalization, and segmenting into time windows. Following this, a one-dimensional (1D) transformer architecture, incorporating various activation functions, is applied to extract features from the preprocessed EEG signals. In the architecture’s final layer, the Softmax activation function is utilized for classifying the data. The performance of the proposed model is assessed using a K-Fold cross-validation strategy, with K set to 10. The proposed method achieved a maximum accuracy of \(97.62 \%\) in diagnosing schizophrenia (SZ), underscoring its potential efficacy in SZ diagnosis.