Analysis of Cortical Excitability During Brain-Computer Interface Stroke Rehabilitation of Upper and Lower Extremity

Brain-computer interfaces (BCls) based on functional electrical stimulation and visual feedback show promise for both upper and lower extremity motor rehabilitation after stroke. Additionally., these BCls allow for the unique opportunity to investigate neurophysiological dynamics during patients” rehabilitation process. Here., we explore these dynamics in 19 stroke patients who underwent 25 upper., followed by 25 lower extremity BCI therapy sessions. We performed two analyses per BCI treatment. In the first analysis., we focused on changes in patients” cortical excitability during paretic side motor imagery between early and late therapy sessions (i.e.., first and last 5 sessions., respectively). To this end., we employed within-patient classification models to investigate if event-related (de)synchronization (ERD/S) in early and late therapy sessions can be distinguished. In the second analysis., we used regression models to explore the prediction of patients” motor function based on their ERD/S during paretic side MI. The within-patient classification models achieved median accuracies of 78.7% and 76.7% for upper and lower extremity BCI treatment., respectively. Furthermore., these ERD/S patterns allowed us to predict patients” upper extremity motor function both in early and late therapy sessions (p<O.05., respectively). In contrast., for lower extremity this was only possible for late therapy sessions (p<O.05). The results of the current study show that patients” ERD/S patterns during paretic side motor imagery undergo robust changes throughout both upper and lower extremity BCI treatment. Additionally., patients” ERD/S patterns do not merely correlate with motor function but allow for prediction., as shown in cross-validated regression analyses.