Using error-related brain activity to identify mismatches during virtually simulated and electrically stimulated upper-limb movements

Patients recovering from motor impairments can benefit from rehabilitation therapy provided using functional electrical stimulation (FES). Yet, its clinical adoption has been limited, due to poor repeatability and inconsistencies in performance. To overcome these challenges, several closed-loop techniques have been proposed that rely on position or force sensors. The use of additional sensors not only increases the device's weight and cost, but it is also difficult to deploy in a clinical environment. Studies involving EEG report error-related potentials (ErrPs) that occur when a person perceives a mismatch or error in their own actions or that of others. This study investigates whether we can detect movement errors when using FES. Two experimental paradigms are presented: one generates errors during discrete stepwise movements, while other generates errors during continuous or sweeping motion of the upper arm. Within the continuous movement experiment, ErrPs arise when observing movement of a virtual Avatar's arm or when using electrical stimulation. Using an ensemble classifier based on decision trees, ErrPs could be successfully detected with an accuracy of 96 ± 3% for discrete and 91 ± 13% for continuous movement errors, respectively. This study demonstrates that ErrPs arise even for errors generated by a rehabilitation device like FES, paving the way for a feedback controller FES using ErrPs.