Sparse Logistic Regression-Based EEG Channel Optimization Algorithm for Improved Universality across Participants

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) channel optimization is crucial for reducing data redundancy and improving brain-computer interface (BCI) performance.
• Assessing the generalizability of optimized EEG channels across diverse participants is essential for practical BCI applications.

Purpose of the Study:

• To investigate the universality of EEG channel optimization across different participants.
• To propose and evaluate a novel sparse logistic regression (SLR)-based algorithm for EEG channel selection.

Main Methods:

• Developed an SLR-based EEG channel optimization algorithm utilizing a non-zero model parameter ranking method.
• Evaluated the algorithm's performance through individual and group analyses on raw EEG data.
• Compared the proposed method against the conventional correlation coefficients (CCS) channel selection technique.

Main Results:

• The SLR algorithm effectively filtered 75-96.9% of redundant EEG channels.
• Achieved a 1.65-5.1% increase in EEG decoding accuracy compared to CCS.
• Demonstrated satisfactory group-level decoding accuracy using only 2-15 common EEG electrodes across participants.

Conclusions:

• The proposed SLR-based EEG channel optimization algorithm exhibits superior universality for EEG decoding.
• This method enhances BCI real-world applicability by reducing data acquisition burden and improving cross-participant performance.