A CNN-Transformer Fusion Model for Proactive Detection of Schizophrenia Relapse from EEG Signals

Area of Science:

• Neuroscience and Artificial Intelligence
• Computational Psychiatry
• Biomedical Signal Processing

Background:

• Schizophrenia relapse detection is challenging due to complex neurophysiological dynamics.
• Existing models lack sensitivity, especially in early relapse stages, often using shallow architectures or single data sources.

Purpose of the Study:

• To develop a hybrid CNN-Transformer model for enhanced schizophrenia relapse prediction.
• To integrate electroencephalogram (EEG) signals with clinical and sentiment data for a comprehensive analysis.

Main Methods:

• A CNN-Transformer fusion model was developed for joint spatial-temporal modeling of relapse indicators.
• A multi-resource data fusion pipeline was implemented, integrating EEG, clinical, and sentiment features.
• The model was evaluated against leading baselines using experimental data.

Main Results:

• The proposed model achieved a superior prediction accuracy of 97%.
• Significant improvements in recall and F1-score were observed compared to existing methods.
• The model demonstrated a notable reduction in false negatives, crucial for timely intervention.

Conclusions:

• The CNN-Transformer fusion model offers a robust and interpretable approach to schizophrenia relapse detection.
• This framework advances personalized relapse prevention and continuous mental health monitoring.
• The findings pave the way for scalable, real-world clinical applications.