Hand motor recovery from post-stroke chronic severe hemiplegia with brain-computer interface based neurofeedback exercise: A randomized controlled trial

BackgroundPost-stroke hemiplegia is a common neurological condition with a significant impact on the daily lives of patients, their families, and society. Since motor imagery and execution are supposed to share common neural networks, repeated motor imagery has been used as a common exercise for motor recovery. However, recent meta-analysis has reported inconsistent results. In the present study, we investigated whether real-time feedback of cortical activities estimated from scalp electroencephalogram (EEG) could enhance the efficacy of motor imagery-based rehabilitation in post-stroke hemiplegic patients. MethodsForty post-stroke chronic hemiplegic patients received 10 days of 40-minute mental practice with motor imagery of the paralyzed hand in addition to 40-minute standard occupational therapy. Patients were randomly allocated to REAL and CTRL groups. In the REAL group, the EEG sensorimotor rhythm recorded over the ipsilesional sensorimotor area were fed back during motor imagery though combination of visual cursor movement, neuromuscular electrical stimulation, and robotic movement support. The patients in the CTRL group wore the same EEG headset as the REAL group, but no feedback was provided. The primary outcome was the upper-extremity subscales of the Fugl-Meyer assessment (FMA). Primary outcomes were tested after 28 days of intervention using analysis of covariance models. This trial is registered with UMIN Clinical Trial Registry, UMIN000026372. FindingsFMA scores showed a significant main effect of time (p = 0{middle dot}027) with an ANCOVA model considering baseline but not a significant between-group difference. However a significant interaction between time and group (p = 0*037) and greater functional gain in the REAL group (p < 0*001) were found with another model considering patient heterogeneity. The improvement was accompanied with the significant changes in the modulation index of finger extensor muscles in the affected hand in the REAL group. Moreover, imagery-related cortical activation in the sensorimotor area was significantly greater in the REAL group than in the CTRL group. No moderate or serious adverse events were confirmed in both groups. InterpretationEEG-based brain-computer interface neurofeedback may enhance the efficacy of mental practice with motor imagery and augment motor recovery in poststroke patients with chronic and severe hemiplegia. FundingSponsored by AMED under Grant Number JP18hk0102032 and JSPS KAKENHI Grant Number 20H05923.