Classification of motor-related brain states using high frequency information from muscle recordings

Abstract

Motor neural interfaces use recordings from the nervous system to extract control signals used to interact with the environment. Muscle signals are becoming an increasingly popular choice for motor interfaces, especially when used to estimate the neural motor commands driving muscle contractions and eventually movement. Here we study the possibility of using electromyography (EMG) to classify motor-related cortical states associated with the cancellation of movement in a GO/NO-GO task, which has been previously linked to changes in cortical beta oscillations. We show that beta (13-30Hz) and low gamma (30-45Hz) frequency bands have the most predictive power in electroencephalography (EEG) and EMG recordings. Moreover, we found comparable accuracy in cancellation state decoding from EEG (average of 74%) and EMG (average of 77%) recordings, which supports the concept of using peripheral signals to predict cortical activity associated to specific motor-related brain states.