Automated detection of clinical depression based on convolution neural network model

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Depression is a prevalent mental disorder with significant societal impact.
• Current depression detection methods have limitations, necessitating objective and efficient approaches.
• Advancements in AI and automation are driving interest in computer-aided diagnosis for mental health.

Purpose of the Study:

• To explore the potential of deep learning (DL) for depression detection.
• To develop and evaluate a diagnostic model for depression using electroencephalogram (EEG) data and convolutional neural networks (CNNs).

Main Methods:

• EEG recordings were collected from 80 subjects (40 with depression, 40 healthy controls) using three electrodes (Fp1, Fz, Fp2) in a resting state.
• Three CNN structures (EEGNet, DeepConvNet, ShallowConvNet) were applied to analyze EEG data for dichotomizing depression patients and healthy controls.
• Performance was evaluated by comparing classification accuracy, sensitivity, and specificity.

Main Results:

• EEGNet achieved high classification performance, with 93.74% accuracy, 94.85% sensitivity, and 92.61% specificity.
• Optimization of EEGNet parameters further improved accuracy to 94.27%.
• The study demonstrated effective depression detection using the developed DL models.

Conclusions:

• EEGNet shows significant potential for accurate and efficient depression detection.
• This DL-based approach using EEG offers a valuable alternative to traditional diagnostic methods.
• The findings support the use of computer-aided diagnosis for mental health applications.