Reproduce “Automatic Diagnosis of Schizophrenia in EEG Signals Using CNN-LSTM Models” on OpenResearch

About this paper

Afshin Shoeibi, Delaram Sadeghi, Parisa Moridian, Navid Ghassemi, Jonathan Heras, Roohallah Alizadehsani, Ali Khadem, Yinan Kong and 3 more

This study developed a deep learning system for automated schizophrenia diagnosis from EEG signals, employing a hybrid CNN-LSTM model that achieved 99.25% accuracy. The system, leveraging combined z-score and L2 normalization with ReLU activation, demonstrated superior performance over conventional machine learning methods.

Key insights

• Hybrid CNN-LSTM architectures are exceptionally effective for classifying complex time-series EEG data, outperforming both conventional machine learning and simpler deep learning models by adeptly extracting both spatial and temporal features.
• The ReLU activation function consistently yields superior performance in these deep learning models compared to Leaky ReLU and seLU for schizophrenia diagnosis.
• Combining z-score and L2 normalization techniques on EEG input data significantly enhances model stability and generalization, leading to higher diagnostic accuracy.

Summary by alphaXiv.