A Deep Learning Approach to Schizophrenia Diagnosis: Leveraging GRU and Attention Mechanisms for Improved EEG Analysis

Schizophrenia (SZ) is a chronic psychological disorder marked by significant instabilities in cognition, emotion, and behavior, affecting millions worldwide. Early and accurate diagnosis is crucial to improve patient outcomes, yet conventional methods, such as clinical interviews and symptom assessment, often lack objectivity and precision, leading to delayed intervention. Recent advances in artificial intelligence, especially deep learning (DL), offer promising diagnostic improvements by analyzing neurophysiological data, such as electroencephalograms (EEG). However, existing EEG-based approaches for SZ diagnosis struggle with inadequate temporal pattern recognition and feature extraction, resulting in suboptimal accuracy. This study proposes a novel hybrid DL model that integrates a Gated Recurrent Unit (GRU) network with an attention mechanism to enhance the diagnosis of SZ by capturing crucial temporal patterns in EEG signals. The method utilizes GRU’s ability to process sequential EEG data, while the attention mechanism selectively focuses on relevant EEG segments, isolating SZ-related neural signatures from noise. The model utilized publicly available EEG dataset from Kaggle, comprising of recordings from SZ patients and healthy controls across multiple electrode sites. The proposed model attained an impressive 98.02% accuracy, with an Area Under the Curve (AUC) of 0.96, 97.53% precision, 96.29% recall, and 96.90% F1-score. These outcomes demonstrate the model’s strong ability to differentiate SZ cases from healthy controls with high sensitivity and specificity. Overall, this study underscores the potential of GRU-attention models in providing an automated, non-invasive, and objective tool for SZ diagnosis, with significant implications for clinical application.