Enhancing Schizophrenia Diagnosis Through Multi-View EEG Analysis: Integrating Raw Signals and Spectrograms in a Deep...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Schizophrenia is a chronic mental disorder with complex diagnostic challenges.
• Early detection of schizophrenia is vital for improved patient outcomes.
• Electroencephalogram (EEG) data shows promise for identifying neural patterns associated with schizophrenia.

Purpose of the Study:

• To develop and evaluate a deep learning framework for automated schizophrenia detection using multi-channel EEG signals.
• To integrate raw EEG data and spectrograms to capture both temporal dynamics and frequency-specific features.
• To assess the efficacy of depth-wise convolution for combining spatial dependencies in EEG data.

Main Methods:

• A novel two-branch deep learning model was designed to process raw EEG signals and their spectrograms.
• Depth-wise convolution was employed to efficiently integrate spatial information across EEG channels.
• The model was trained and validated on two independent EEG datasets.

Main Results:

• The deep learning framework achieved high classification accuracies of 0.985 and 0.994 on two separate datasets.
• Combining raw EEG signals with time-frequency representations proved effective for precise schizophrenia detection.
• An ablation study confirmed the significant contributions of individual architectural components.

Conclusions:

• The proposed framework demonstrates superior performance compared to existing methods for schizophrenia detection.
• Utilizing multi-view EEG data, including raw signals and spectrograms, enhances diagnostic accuracy.
• This approach holds potential for developing more effective clinical diagnostic tools for schizophrenia.