Development and feasibility of a motor imagery-based brain-computer interface-controlled closed-loop functional electrical stimulation system for swallowing rehabilitation.

Abstract

Conventional swallowing functional electrical stimulation (FES) is usually delivered in open loop or triggered by peripheral signals, which may not align precisely with voluntary swallowing intention. We developed a motor imagery-based brain-computer interface (MI-BCI)-controlled closed-loop swallowing FES system for post-stroke dysphagia (PSD) and investigated its neurophysiological basis, decoding performance, and short-term feasibility. Two experiments were conducted. In Experiment 1, swallowing motor imagery (SMI)-related electroencephalography (EEG) features were identified in healthy controls (HC, n = 15), patients with PSD (n = 15), and post-stroke patients without dysphagia (PSND, n = 15). A threshold-based decoder based on Fp1 spectral power ratios was then validated in an independent HC cohort (n = 10). In Experiment 2, 10 patients with PSD received 10 sessions of MI-BCI-controlled closed-loop swallowing FES over 2 weeks, and feasibility, usability, and safety were assessed. During SMI, Fp1 spectral power ratios decreased relative to rest. The δ/α ratio decreased significantly in all three groups, whereas the δ/(α + β) ratio and the (δ + θ)/(α + β) ratio decreased significantly in HC and PSND and showed the same downward trend in PSD. Patients with PSD also showed higher θ-band power at T3 than HC and PSND (P = 0.0382). The decoder achieved a mean classification accuracy of 71.5% in the independent validation cohort. In Experiment 2, adherence was 100%, with 29.8 ± 6.2 successful closed-loop triggers per session, a mean System Usability Scale score of 72.8 ± 4.2, and no serious adverse events. These findings support the technical feasibility of the proposed system, indicate acceptable short-term usability, and show no major safety concerns during the intervention period. 
Trial registration: ChiCTR2400079388.