EEG-derived brainwave patterns for depression diagnosis via hybrid machine learning and deep learning frameworks.

Area of Science:

• Neuroscience
• Computer Science
• Medical Informatics

Background:

• Artificial intelligence (AI), machine learning (ML), and deep learning (DL) are advancing rapidly in medicine.
• Depression is a leading cause of disability, necessitating early and objective diagnostic methods.
• Electroencephalogram (EEG) signals offer a patient-friendly and cost-effective approach for neurological assessments.

Purpose of the Study:

• To develop ML and DL techniques for diagnosing depression using EEG signals.
• To evaluate the efficacy of different classifiers and signal conditions for depression detection.

Main Methods:

• Extracted statistical features from EEG signals of 34 Major Depressive Disorder (MDD) patients and 30 healthy controls.
• Implemented and compared three classifiers: 1D Convolutional Neural Network (1DCNN), Support Vector Machine (SVM), and Logistic Regression (LR).
• Tested models on EEG data collected during Task, Eye Close (EC), and Eye Open (EO) conditions.

Main Results:

• The 1DCNN model achieved the highest accuracy (90.21%) using TASK signals, followed by SVM (89.3%) and LR (88.4%).
• TASK condition signals consistently yielded higher classification accuracies across all tested models compared to EC and EO conditions.
• The proposed methods demonstrated statistically significant improvements (p < 0.05) over existing approaches.

Conclusions:

• EEG-based ML and DL approaches show significant potential for the accurate and reliable clinical diagnosis of depression.
• The TASK condition provides a promising data acquisition protocol for enhanced depression detection via EEG.
• Further research into EEG signal analysis can lead to improved diagnostic tools for mental health.