Removal of EOG artifacts from EEG recordings using stationary subspace analysis

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Ocular artifacts are common and significant sources of noise in electroencephalography (EEG) recordings.
• Existing blind source separation methods often struggle with nonstationary signals and limited electrode data.
• Accurate artifact removal is crucial for reliable EEG analysis in clinical and research settings.

Purpose of the Study:

• To develop and validate an effective approach for removing ocular artifacts from raw EEG signals.
• To leverage Stationary Subspace Analysis (SSA) for improved artifact separation compared to traditional methods.
• To demonstrate the method's efficacy across various challenging recording conditions.

Main Methods:

• Applied Stationary Subspace Analysis (SSA), a blind source separation technique, to raw EEG data.
• Utilized SSA's ability to consider distribution changes (mean and covariance) without assuming source independence or uncorrelation.
• Projected identified artifact components back and subtracted them from the EEG signal to obtain clean data.

Main Results:

• The proposed SSA-based method successfully removed ocular artifacts from both simulated and real EEG data.
• Demonstrated superior performance in scenarios with limited electrode usage.
• Showed effectiveness even when EEG signals were highly nonstationary and sources were not independent or uncorrelated.

Conclusions:

• Stationary Subspace Analysis provides an effective and robust method for ocular artifact removal in EEG.
• This approach offers significant advantages over conventional methods, especially for complex and limited-data scenarios.
• The validated technique enhances the reliability of EEG data for subsequent analysis.