Detecting Schizophrenia Patients Using Deep Learning Models

Schizophrenia is a chronic psychiatric illness affecting around 1% of the population worldwide. Early detection and diagnosis are crucial for prompt treatment and improved outcomes. Electroencephalography (EEG) provides insights into abnormal brain activity associated with schizophrenia. Deep learning techniques like convolutional neural networks (CNNs) show promise for automated analysis of EEG patterns. This study evaluated 5 CNN architectures—Attention-based CNN, ResNet50, Inception-V4, EfficientNetB0, and Squeeze and Excitation Net—for classifying schizophrenia based on EEG graphs from 14 patients and 14 healthy controls. The attention-based CNN achieved the best performance with 99.39% training accuracy and 98.95% validation accuracy. However, variability across models and overfitting on the small dataset were key limitations. Overall, this research demonstrates the potential of an automated EEG-based screening tool for schizophrenia using deep learning. Given the promising results, further research with larger datasets could enable the deployment of a CNN tool for earlier diagnosis and treatment.