Electroencephalography-Based Depression Detection Using Multiple Machine Learning Techniques.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Data Science

Background:

• Electroencephalography (EEG) is increasingly used for depression diagnosis, but signal complexity and individual variations pose challenges.
• Demographic factors like age and gender influence depression incidence and EEG signals, necessitating their inclusion in diagnostic models.

Purpose of the Study:

• To develop an algorithm for recognizing depression patterns using electroencephalography (EEG) data.
• To investigate the efficacy of integrating demographic data with EEG signals for enhanced depression detection.

Main Methods:

• Multiband analysis of resting-state EEG signals from a 128-channel cap.
• Application of machine learning and deep learning techniques, specifically Convolutional Neural Networks (CNNs), for automated depression detection.
• Utilized the multi-modal open dataset MODMA, classifying patients into major depressive disorder (MDD) and healthy control groups.

Main Results:

• A CNN model achieved 97% accuracy in detecting MDD after 25 training epochs.
• The study considered two primary categories: major depressive disorder (MDD) and healthy controls, with MDD encompassing various sub-classifications.
• Integration of EEG signals with demographic data demonstrated a promising approach for depression diagnosis.

Conclusions:

• Combining EEG signals with demographic data offers a robust strategy for improving the accuracy and generalizability of depression detection systems.
• This multimodal approach holds significant potential for advancing the automated diagnosis of major depressive disorder (MDD).