Closed-loop motor imagery brain-computer interface-assisted training for upper limb rehabilitation after subacute...

Closed-loop motor imagery brain-computer interface (MI-BCI) training may support post-stroke upper-limb rehabilitation by coupling motor intention with contingent multisensory feedback. This randomized pilot trial examined its feasibility, safety, short-term clinical effects, and exploratory EEG correlates in patients with subacute stroke.

In this single-center, assessor-blinded, two-arm pilot trial, 40 patients with first-ever subcortical stroke in the subacute phase were randomized 1:1 to a BCI group or an active control group after a 2-day motor imagery familiarization phase. Both groups received routine medical management, standardized conventional rehabilitation, and dose-matched motor imagery-based hand training for 4 weeks. The BCI group received EEG-contingent closed-loop MI-BCI-assisted training with a soft rehabilitation glove, whereas the control group received non-EEG-contingent glove-assisted motor imagery training under matched training duration, task instructions, device exposure, and multisensory feedback. The primary outcome was the Fugl-Meyer Assessment for the Upper Extremity (FMA-UE). Secondary outcomes included the Action Research Arm Test (ARAT) and Modified Barthel Index (MBI). Exploratory EEG outcomes included FFT%α and FFT%β during motor imagery. Clinical and EEG outcomes were analyzed using baseline-adjusted ANCOVA models, with week-4 values as dependent variables and corresponding baseline values as covariates.

All randomized participants completed the 4-week assessment. In baseline-adjusted ANCOVA models, the BCI group showed higher week-4 scores than the control group for FMA-UE (adjusted mean difference, 13.40 points; 95% CI, 10.71-16.08; p < 0.001), ARAT (7.31 points; 95% CI, 4.55-10.07; p < 0.001), and MBI (12.21 points; 95% CI, 8.55-15.87; p < 0.001). Exploratory EEG analyses also showed higher week-4 FFT%α and FFT%β in the BCI group, with adjusted mean differences of 6.78 percentage points (95% CI, 5.22-8.34; p < 0.001) and 3.95 percentage points (95% CI, 2.53-5.36; p < 0.001), respectively. No serious adverse events occurred.

Closed-loop MI-BCI-assisted training was feasible and well tolerated in selected patients with subacute stroke. The observed short-term improvements in upper-limb impairment and activity capacity provide preliminary signals of potential benefit beyond dose-matched non-EEG-contingent feedback training. Exploratory EEG findings suggest task-related modulation of alpha- and beta-band sensorimotor rhythmic activity, but should be interpreted as hypothesis-generating rather than confirmatory evidence of neural reorganization. Larger multicenter trials with longer follow-up, rigorous neurophysiological analyses, and real-world upper-limb use outcomes are needed.

ChiCTR2400083992. https://www.chictr.org.cn/showproj.html?proj=229529.