AnEEG: leveraging deep learning for effective artifact removal in EEG data

Area of Science:

• Neuroscience
• Clinical Diagnostics
• Biomedical Engineering

Background:

• Electroencephalography (EEG) is vital for neural activity monitoring but suffers from artifacts (muscle, eye blinks, environmental interference).
• Artifacts obscure crucial information, hindering accurate analysis and diagnosis.
• Deep learning shows promise in mitigating EEG artifacts and enhancing signal quality.

Purpose of the Study:

• To introduce AnEEG, a novel deep learning method for artifact removal in EEG signals.
• To quantitatively evaluate AnEEG's effectiveness in improving EEG signal quality.
• To compare AnEEG's performance against established artifact removal techniques like wavelet decomposition.

Main Methods:

• Development of a novel deep learning model named AnEEG.
• Application of AnEEG to eliminate artifacts from EEG signals.
• Quantitative assessment using metrics: Normalized Mean Squared Error (NMSE), Root Mean Squared Error (RMSE), Correlation Coefficient (CC), Signal-to-Noise Ratio (SNR), and Signal-to-Artifact Ratio (SAR).

Main Results:

• AnEEG demonstrated superior performance compared to wavelet decomposition techniques.
• The model achieved lower NMSE and RMSE, indicating better signal fidelity.
• Higher CC values confirmed stronger linear agreement with ground truth signals.
• Significant improvements were observed in SNR and SAR, signifying enhanced signal clarity.

Conclusions:

• The proposed AnEEG deep learning method effectively removes artifacts from EEG signals.
• AnEEG offers a promising approach to enhance EEG data quality for research and clinical applications.
• The quantitative results validate AnEEG's superiority over traditional methods, paving the way for more reliable neural activity analysis.