Inter-channel Connectivity of Motor Imagery EEG Signals for a Noninvasive BCI Application

Noninvasive brain-computer interfaces (BCIs) based on motor imagery translate brain activity into motor execution commands to control external devices. They have largely relied on the measurement of the sensorimotor rhythms (SMR) and the beta rhythms in electroencephalography (EEG). However, most BCIs of this type have exploited SMR and beta rhythms observed from a few EEG channels over the sensorimotor area. They also extracted movement-related information from each channel independently, without considering connectivity between channels. In this study, we aim to investigate whether we can obtain useful information of movements from the connectivity measures across a wide range of EEG channels, over the whole brain. To address this question, we evaluated a simple connectivity measure - cross-correlation coefficients (CCs) - for twenty-two EEG channels distributed over different brain regions to differentiate four different motor imagery states, including left hand, right hand, both feet, and tongue. The tem oral variations of CCs across twenty-two channels exhibited distinct patterns as to four motor imagery states. It suggests that we may use connectivity as a useful source to derive noninvasive BCIs.