Performance of common spatial pattern under a smaller set of EEG electrodes in brain-computer interface on chronic stroke patients: A multi-session dataset study

Abstract

Brain-computer interface (BCI) uses non-muscular channel of the nervous system for communication. Common Spatial Pattern (CSP) is a popular spatial filtering method used to reduce the effect of volume conduction on EEG signals. It is thought that CSP requires a large number of electrodes to be effective. Using a 20-session dataset of motor imagery BCI usage by 5 stroke patients, we demonstrated that after channel selection, CSP can still maintain a high accuracy with low number of electrodes using a newly proposed channel selection method called CSP-rank (higher than 90% with 8 electrodes). The results showed that using only the first session for channel selection, a high accuracy can be maintained in subsequent sessions. CSP-rank has been compared to the popular support vector machine recursive feature elimination (SVM-RFE). The results showed that the CSP-rank required less electrodes to maintain accuracy higher than 90% (a minimum of 8 compared to 12 of SVM-RFE) and it attained a higher maximum accuracy (91.7% compared with 90.7% of SVM-RFE). This could support clinicians to apply more BCI in routine rehabilitation.