Influence of VR-based Brain-Computer Interfaces Training in Brain Activity and Clinical Outcome in Chronic Stroke: A Longitudinal Study of Single Cases

Abstract

Background: Restorative Brain-Computer Interfaces (BCI's) can provide an alternative non-muscular channel for patients who lack volitional movement by reinforcing the modulation of sensorimotor rhythms through motor-imagery (MI). MI practice can be augmented through embodied feedback delivered by virtual reality (VR). Nonetheless, detailed information concerning the impact of VR-BCI training in clinical outcome is still largely missing. This paper investigates the long-term effect of VR-BCI training in brain activity patterns measured by the electroencephalogram (EEG) and clinical scales of stroke patients. Methods: Four chronic stroke patients underwent a 3-week VR-BCI intervention. Fugl-Meyer Assessment (FMA) was conducted pre-, post- intervention, and a follow-up, one-month later. The VR-BCI training consisted of a bimanual rowing task (NeuRow) involving MI and motor observation. From the EEG, the Event-related Desynchronization (ERD) was extracted from the Mu band (8-12 Hz). Finally, the BCI performance was computed from the BCI training session. Results: All patients showed significantly increased Mu ERD compared to baseline, however, the affected side showed reduced ERD during the contralateral MI. Significant ERD differences originating from both the lesioned and the healthy hemisphere were found when comparing post- with pre- intervention in all patients, and decreasing ERD was found for patients with no clinical improvement. Patients with decreased lateralization of ERD had no improvement of the FMA score. Conclusion: Stroke patients in the chronic phase can modulate their brain activity patterns using MI in a VR-BCI task, with observable patient-specific differences between brain activity and clinical outcome. Nonetheless, only patients with increased ERD lateralization showed clinical improvement.