Analysis of correlated EEG activity during motor imagery for brain-computer interfaces

Overall accuracy of noninvasive brain-computer interfaces (BCIs) based on motor imagery electroencephalography (EEG) is highly dependent on the extraction of features from the oscillation of sensorimotor rhythms (SMRs) during imagination of movements. In this study, we statistically evaluated whole-brain connectivity using the measurement of linear correlation coefficients (CCs) between EEG channel pairs instead of using conventional spectral analysis. We showed distinct patterns of temporal variations of CCs of all channel pairs and significant channel connections for four motor imageries, including left hand, right hand, both feet, and tongue, in two subjects. Contralateral connectivity was observed in the motor imagery of left and right hands, whereas central connectivity was observed in the motor imagery of both feet. Our results suggest to the implementation of the state-of-the-art BCIs based on whole-brain channel connectivity in motor imagery.