Improving motor imagery classification during induced motor perturbations

Abstract Objective. Motor imagery is the mental simulation of movements. It a common paradigm to design brain-computer interfaces (BCIs) that elicits modulation brain oscillatory activity similar real, passive and induced In this study, we used peripheral stimulation provoke movements one limb during performance motor tasks. Unlike other works, in which are support BCI operation, our goal was test improve robustness based systems perturbations caused by artificially generated Approach. We performed session with ten participants who carried out three limbs. some trials, arms moved neuromuscular stimulation. analysed 2-class classifications without movement perturbations. investigated decrease produced these disturbances designed different computational strategies attenuate observed classification accuracy drop. Main results. When not coincident classes, extracting sources imagination tasks resulted being control (undisturbed) condition; when also involved tasks, drop significantly alleviated spatially filtering neural noise show loss accompanied weaker power sensorimotor rhythm. Importantly, residual could be predict whether user will perform sufficient under disturbances. Significance. provide methods ameliorate even eliminate related afferent This can help improving reliability current systems.