EEG Artifact Removal System for Depression Using a Hybrid Denoising Approach

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) analysis is crucial for early depression diagnosis.
• Artifacts in EEG signals hinder the accuracy of computer-aided diagnosis systems.
• Developing robust denoising techniques is essential for reliable EEG-based depression detection.

Purpose of the Study:

• To propose a novel denoising method for EEG signals to enhance depression diagnosis.
• To improve the accuracy of computer-aided diagnosis systems for depression.
• To address artifact contamination in EEG signal processing.

Main Methods:

• Empirical Mode Decomposition (EMD) for signal decomposition.
• Detrended Fluctuation Analysis (DFA) for mode selection.
• Wavelet Packet Decomposition (WPD) for signal denoising.
• Classification using Random Forest (RF) and Support Vector Machine (SVM).

Main Results:

• The proposed EMD-DFA-WPD technique significantly improved signal-to-noise ratio (SNR) and reduced Mean Absolute Error (MAE).
• Achieved high classification accuracy: 98.51% for RF and 98.10% for SVM.
• Demonstrated superior performance compared to EMD-DFA and EMD-DWT methods.

Conclusions:

• The developed denoising system enhances the classification of depressed individuals.
• The proposed method offers a more accurate and reliable EEG-based system for depression diagnosis.
• Further research can explore solutions for the mode mixing problem in EMD.