Deep Depression Detection with Resting-State and Cognitive-Task EEG.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Psychiatry

Background:

• Depression is a prevalent mental disorder impacting overall health and functioning.
• Accurate and objective diagnosis of depression is currently challenging.
• Electroencephalography (EEG) offers a low-cost, high-performance solution for monitoring brain activity.

Purpose of the Study:

• To investigate the efficacy of deep neural networks in detecting depression using EEG-based neural patterns.
• To analyze neural patterns during resting states and cognitive tasks in individuals with and without depression.
• To develop objective biomarkers for depression detection through advanced computational models.

Main Methods:

• Collected EEG signals from 33 depressed patients and 40 healthy controls using wearable dry electrodes.
• Employed Attentive Simple Graph Convolutional Network and Transformer neural network models for depression detection.
• Designed four experimental stages: two resting states and two cognitive tasks (Continuous Performance Test-Identical Pairs, Stroop Color Word Test).

Main Results:

• The Transformer model achieved an Area Under the Curve (AUC) of 0.94 on cognitive tasks (sensitivity: 0.87-0.93, specificity: 0.88-0.91).
• The Transformer model achieved an AUC of 0.89 on resting states (sensitivity: 0.85-0.87, specificity: 0.88-0.90).
• Observed decreased neural energy and impaired performance in sustained attention and response inhibition in depressed patients.

Conclusions:

• EEG-based neural patterns, analyzed by deep learning models, demonstrate significant potential for objective depression detection.
• The findings offer new insights into depression mechanisms and the development of EEG-based depression biomarkers.
• This approach holds promise for both clinical practice and home-use applications due to its performance and cost-effectiveness.