GVICA: A Multi-Channel EEG Hierarchical Noise Reduction Framework Based on GWO Dynamically Optimized VMD-ICA Fusion

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) signal acquisition is frequently contaminated by artifacts, primarily from ocular (EOG) and electromyographic (EMG) sources.
• These artifacts obscure genuine neural activity, complicating accurate interpretation and analysis of brain signals.

Purpose of the Study:

• To develop and validate an automated, multi-level denoising method (GVICA) for robust artifact removal from multi-channel EEG.
• To ensure maximal preservation of underlying neurophysiological information during the artifact removal process.

Main Methods:

• GVICA integrates optimized Variational Mode Decomposition (VMD) with Independent Component Analysis (ICA) in a multi-level framework.
• The Grey Wolf Optimizer (GWO) adaptively tunes VMD parameters, followed by anomaly detection and correction of intrinsic mode functions (IMFs).
• An ICA-based entropy module objectively selects relevant sub-signals in the frequency domain for artifact removal.

Main Results:

• GVICA demonstrated superior artifact removal on simulated and real EEG data, evidenced by lower RMSE and higher PCC across various SNRs.
• The method successfully and automatically removed both EMG and EOG artifacts simultaneously.
• GVICA achieved more accurate restoration of brain responses compared to existing artifact removal techniques.

Conclusions:

• The GVICA method offers a significant improvement in multi-channel EEG signal quality via its adaptive, multi-level denoising approach.
• It provides a reliable tool for cleaning EEG data in clinical settings, particularly for analyzing motor execution in stroke rehabilitation.