EEG-based deep learning model for the automatic detection of clinical depression

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Clinical depression diagnosis relies on subjective assessments, with Electroencephalography (EEG) offering objective but complex signal analysis.
• Existing automated depression diagnosis strategies using EEG signals are limited by signal complexity and individual variations, leading to diagnostic inaccuracies.

Purpose of the Study:

• To develop and evaluate a deep learning model for accurate depression detection by analyzing Electroencephalography (EEG) signals.
• To leverage the strengths of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) for enhanced feature extraction and pattern recognition in EEG data.

Main Methods:

• A hybrid deep learning architecture combining CNN for local feature extraction and LSTM for sequential pattern learning was designed.
• CNN filters processed EEG signals to generate feature maps, which were then fed into LSTM layers for temporal dependency analysis.
• Classification was performed using fully connected layers, with LSTM's memory cells aiding in retaining essential long-term features.

Main Results:

• The integrated CNN-LSTM model achieved high diagnostic accuracies: 99.07% for the right hemisphere and 98.84% for the left hemisphere EEG signals.
• The model demonstrated effective learning of local characteristics and complex temporal patterns within EEG signals.
• Random splitting was employed for robust model testing and validation.

Conclusions:

• The proposed deep learning model integrating CNN and LSTM shows significant promise for accurate and reliable automated depression detection using EEG signals.
• This approach addresses the limitations of previous methods by effectively handling the complexity and variability of EEG data in depressed individuals.
• The high accuracies achieved suggest a potential for clinical application in objective depression diagnosis.