EEG-based schizophrenia classification using attention-integrated deep convolutional networks

Schizophrenia is a complex psychiatric disorder marked by cognitive and perceptual disruptions, for which electroencephalography (EEG) provides a valuable non-invasive biomarker. In this study, we propose a convolutional attention-based deep learning framework for the automatic detection of schizophrenia from EEG signals. The model integrates spatial feature extraction via convolutional layers with an attention mechanism that adaptively focuses on discriminative temporal patterns within the EEG. We have conducted experiments on two publicly available datasets, the Moscow EEG dataset and the IBIB PAN dataset. Performance on the Moscow dataset was lower, with an accuracy of 73.98%, likely due to age-related neural variability and limited recording duration. The model achieved a classification accuracy of 98.45% on the IBIB PAN dataset, demonstrating strong generalization and discriminative capability. These results highlight the potential of attention-augmented convolutional networks for schizophrenia detection, while also underscoring the challenges of generalizing across datasets with differing demographic and acquisition characteristics.