A Novel Technique for Selecting EMG-Contaminated EEG Channels in Self-Paced Brain-Computer Interface Task Onset

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electromyography (EMG) artifacts are a significant challenge in electroencephalography (EEG) studies, impacting brain-computer interfaces (BCIs), brain mapping, and clinical applications.
• Existing Blind Source Separation (BSS) techniques for artifact removal can inadvertently eliminate crucial EEG data, leading to information loss.

Purpose of the Study:

• To introduce a novel method for statistically selecting EEG channels contaminated with class-dependent EMG (EMG-CCh).
• To reduce the loss of useful EEG information during artifact removal processes.

Main Methods:

• Developed a technique to select EMG-contaminated EEG channels (EMG-CCh) based on correlations between EEG and facial EMG signals.
• Compared artifact removal efficacy using the proposed EMG-CCh selection with traditional BSS methods applied to all channels.
• Validated the method using a custom dataset and BCI competition data.

Main Results:

• The proposed EMG-CCh selection method significantly improved class separation compared to applying BSS to all channels, particularly with Independent Component Analysis (ICA), Principal Component Analysis (PCA), and BSS-Canonical Correlation Analysis (BSS-CCA).
• Improvements in class separation were observed in 79% of cases for ICA, 53% for PCA, and 11% for BSS-CCA on the custom dataset.
• The method demonstrated improved class separation in 60% of cases for ICA and BSS-CCA using BCI competition data.

Conclusions:

• The novel EMG-CCh selection method effectively reduces useful information loss during artifact removal in EEG studies.
• This approach offers a significant improvement in class separation for BCIs and other EEG applications.
• The method is unique, as no prior techniques utilize the correlation between EEG and EMG channels for EMG artifact removal, and it can be used independently or combined with existing methods.