Enhancing brain-computer interface-based neurorehabilitation using adaptive sensory electrical stimulation: a feasibility study

Stroke is the leading cause of long-term disability and effective treatment strategies are lacking, urging for the discovery of new approaches. Brain Computer Interface (BCI)-based technologies are able to restore upper limb movement in stroke survivors, by converting brain activity related to the intention to move into the actual movement of a robotic exoskeleton, and thus represent a promising neurotechnology-based rehabilitation alternatives for stroke survivors with highly restricted arm and hand function. However, BCI illiteracy, namely the inability to establish a reliable control over this technology, is an important issue in BCI applications: indeed, it is estimated that about 15-30% of users are not able to control a BCI. Understanding physiological mechanisms underpinning BCI learning and control could facilitate the translation of BCI-based therapies in stroke neurorehabilitation. Neuromodulation techniques are a promising approach to unravel the role of specific frequency onto motor learning. In particular, integrating alpha-tuned median nerve stimulation (MNS) and motor-BCI may allow the direct modulation of alpha neural oscillations and the investigation of their causal role on BCI motor learning. However, the feasibility of the integration of frequency-tuned electrical stimulation of the median nerve and motor-BCI has not yet been investigated. Thus, this study aimed at evaluating the feasibility of such an approach in terms of residual current artifacts in the electroencephalographic (EEG) signal, reliable BCI classification and users' perceived control. In particular, after a preliminary investigation of the effect of alpha-tuned MNS on neural activity, a new calibration approach and a new feedback control were conceptualized. The efficacy of these approaches was then evaluated by combining a motor-BCI task and MNS. As predicted, the stimulation induced no artifacts in the data, the accuracy has not been reduced and no change in perceived feedback control were found, consequently the results confirm the feasibility of the integration of MNS and motor-BCI and pave the way for future researches investigating the role of alpha-oscillations on motor learning.