An Improved Classification Model for Depression Detection Using EEG and Eye Tracking Data.

Area of Science:

• Neuroscience
• Psychiatry
• Biomedical Engineering

Background:

• Depression is a significant global health burden with diagnostic challenges including low accuracy and subjectivity.
• Objective and reliable methods for depression detection are crucial for effective diagnosis and treatment.
• Electroencephalogram (EEG) and eye movements (EMs) offer non-invasive data acquisition for depression assessment.

Purpose of the Study:

• To propose and validate a novel content-based ensemble method (CBEM) for enhanced depression detection.
• To evaluate the performance of both static and dynamic CBEM approaches.
• To compare CBEM's accuracy against traditional classification methods.

Main Methods:

• Developed a content-based ensemble method (CBEM) utilizing electroencephalogram (EEG) and eye movement (EM) datasets.
• EEG/EM datasets were segmented based on experimental context.
• A majority vote strategy was employed for subject label determination within the ensemble framework.

Main Results:

• CBEM achieved high accuracies of 82.5% on an eye-tracking dataset and 92.65% on a resting-state EEG dataset.
• The proposed CBEM demonstrated superior performance compared to conventional classification techniques.
• Validation was conducted on two distinct datasets comprising 36 and 34 subjects, respectively.

Conclusions:

• The content-based ensemble method (CBEM) offers an effective solution for improving depression identification accuracy.
• CBEM provides a reliable and objective approach for depression detection, outperforming traditional methods.
• Findings suggest CBEM's potential utility as an auxiliary diagnostic tool for depression in clinical settings.