A Study of the Effects of Electrode Number and Decoding Algorithm on Online EEG-Based BCI Behavioral Performance

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Motor imagery-based brain-computer interfaces (BCIs) decode movement intentions using electroencephalography (EEG).
• Channel selection and decoding algorithms significantly impact BCI performance.
• Limited online studies compare these factors across multiple training sessions.

Purpose of the Study:

• To assess the effect of channel number and decoding method on BCI performance progression across multiple training sessions.
• To investigate associated neurophysiological changes during BCI control.
• To compare Laplacian Filtering (LAP/S) and Common Spatial Pattern (CSP) algorithms with varying channel configurations.

Main Methods:

• Forty-five subjects were divided into three groups: LAP/S (9 channels), CSP/L (40 channels), and CSP/S (9 channels).
• Online BCI performance and EEG data were recorded across multiple training sessions.
• Statistical analysis compared performance metrics and neurophysiological correlates between groups.

Main Results:

• CSP/L showed higher initial performance than LAP/S, with no significant difference compared to CSP/S.
• LAP/S performance improved over the first three sessions, while CSP methods did not show significant improvement.
• LAP/S generated control signals more correlated with target location, with higher R-square values in later sessions.

Conclusions:

• A large EEG montage (40 channels) did not yield significantly better online BCI performance after the first session compared to a smaller montage (9 channels).
• The LAP/S method with a small EEG montage facilitated skill improvement across sessions.
• The CSP method showed limited improvement across sessions, irrespective of channel number.