A major depressive disorder classification framework based on EEG signals using statistical, spectral, wavelet,...

Area of Science:

• Neuroscience
• Computational Psychiatry
• Biomedical Engineering

Background:

• Major Depressive Disorder (MDD) diagnosis relies on subjective questionnaires, often leading to inaccuracies.
• Electroencephalogram (EEG) signals offer a potential objective biomarker for MDD detection.
• Machine learning (ML) techniques are increasingly explored to analyze complex EEG data for psychiatric disorders.

Purpose of the Study:

• To propose and validate a novel ML framework for accurate MDD diagnosis using EEG signals.
• To investigate the efficacy of diverse EEG-derived features and feature selection methods for MDD classification.
• To identify the optimal ML classifier for distinguishing MDD patients from healthy controls based on EEG data.

Main Methods:

• EEG data from MDD patients and healthy subjects were analyzed.
• Features were extracted using statistical, spectral, wavelet, functional connectivity, and nonlinear methods.
• Sequential Backward Feature Selection (SBFS) and various classifiers, including RBFSVM, were employed for model development and selection.

Main Results:

• The proposed ML framework achieved high diagnostic performance, with the RBFSVM classifier yielding an average accuracy of 99%.
• Key performance metrics included 98.4% sensitivity, 99.6% specificity, 98.9% F1-score, and 0.4% false discovery rate.
• The method demonstrated superior performance compared to existing EEG-based MDD classification approaches.

Conclusions:

• The developed ML framework provides a highly accurate method for MDD diagnosis using EEG signals.
• The findings support the potential of this approach for developing clinical computer-aided diagnosis (CAD) tools.
• Objective EEG-based diagnosis could significantly improve the clinical management of major depressive disorder.