Multiple time-lag canonical correlation analysis for removing muscular artifacts in EEG

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Electroencephalogram (EEG) recordings are susceptible to muscular artifacts, which can obscure neural activity.
• Existing methods for artifact removal may not fully address the complexities of joint source separation across multiple time lags.

Purpose of the Study:

• To develop a novel joint blind source separation (BSS) approach for enhanced muscular artifact removal from EEG data.
• To leverage canonical correlation analysis (CCA) across multiple time lags for improved artifact identification and separation.

Main Methods:

• A new joint BSS method utilizing CCA on datasets at multiple time lags was developed.
• Sources are jointly extracted in decreasing order of between-set correlations, isolating low-correlation artifactual sources.
• The theoretical performance of the proposed CCA-based method was analyzed.

Main Results:

• The proposed method successfully separates sources by prioritizing those with the lowest between-set correlations, effectively targeting muscular artifacts.
• Theoretical analysis demonstrates superior BSS performance under more relaxed conditions compared to conventional CCA.
• Experimental validation on real EEG data confirms the effectiveness of the approach in removing muscular artifacts.

Conclusions:

• The novel CCA-based joint BSS approach offers improved performance for muscular artifact removal in EEG.
• This method provides a more robust and effective solution for cleaning EEG signals compared to traditional techniques.