Brain–computer interface stimulation shows clearer recovery signal after stroke, review finds

A new systematic review and network meta-analysis has compared two brain–computer interface approaches used in post-stroke motor rehabilitation, and the results point to a meaningful difference between them. The analysis, led by researchers affiliated with the 964th Hospital of the Joint Logistic Support Force and The Chinese University of Hong Kong, was published in Frontiers in Neuroscience.

The research team set out to answer a question that has lingered in the neurorehabilitation field: when a patient’s brain signals are decoded to drive a therapy, does it matter whether those signals trigger functional electrical stimulation of muscles or control a robotic rehabilitation device? The relative effects of different brain-computer interface output modalities for post-stroke motor rehabilitation have remained unclear, the authors wrote, motivating the synthesis.

To address the question, the researchers searched six databases: PubMed, Web of Science, Embase, IEEE Xplore, the Cochrane Library, and MEDLINE. They identified sixteen randomized controlled trials published from January 2020 to October 2025. The interventions covered in the analysis included brain-computer interface-triggered functional electrical stimulation, brain-computer interface-controlled robotic rehabilitation, and conventional rehabilitation alone.

The primary outcome was motor recovery as measured by the Fugl-Meyer Assessment. The team performed a frequentist random-effects network meta-analysis, a statistical approach that allows comparison of multiple interventions even when some have never been tested against each other directly. Certainty of the evidence was assessed using the CINeMA framework, which grades confidence in estimates from network meta-analyses.

The structural finding of the review is as important as the numbers. No trial directly compared the two active brain-computer interface modalities against each other, which produced what statisticians call a star-shaped evidence network: each active intervention was compared only with conventional rehabilitation, and any comparison between the two active modalities rests entirely on indirect, statistical inference.

Within that network, brain-computer interface-triggered functional electrical stimulation showed a clearer signal of benefit. Compared with conventional rehabilitation, it improved motor recovery with a standardized mean difference of 0.57, with a 95 percent confidence interval running from 0.12 to 1.01. Because that interval excludes zero, the estimate suggests a positive effect, though the lower bound sits close to the null.

Brain-computer interface-controlled robotic rehabilitation, by contrast, showed an uncertain effect against conventional rehabilitation, with a standardized mean difference of 0.17 and a 95 percent confidence interval of −0.25 to 0.60. That interval crosses zero, meaning the data cannot rule out no effect, a negative effect, or a benefit of the size seen with the stimulation approach.

The indirect comparison between the two active modalities numerically favored brain-computer interface-triggered functional electrical stimulation over the robotic approach, yielding a standardized mean difference of 0.39. Crucially, however, the 95 percent confidence interval ran from −0.22 to 1.01, crossing the null. The authors were explicit that this comparison has very low confidence and does not establish comparative superiority between the two technologies.

Certainty in the findings was limited by three factors the authors named: indirectness, imprecision, and heterogeneity. Indirectness stems from the star-shaped network and the absence of head-to-head trials. Imprecision reflects the wide confidence intervals surrounding the estimates. Heterogeneity indicates that the included trials varied in ways that complicate pooling. The authors also noted that P-score rankings, a common way of ordering interventions in network meta-analyses, should be treated as exploratory pending direct head-to-head trials.

For clinicians and researchers watching the brain–computer interface rehabilitation field mature, the review offers a measured takeaway rather than a verdict. Brain-computer interface-triggered functional electrical stimulation currently carries the clearer signal of benefit relative to conventional rehabilitation, while the robotic modality’s effect remains uncertain. But the central comparison between the two active approaches depends wholly on indirect evidence from a network in which no trial has yet pitted them against each other. The authors’ conclusion points squarely at the next step: direct head-to-head randomized trials will be needed before any claim of superiority between these brain–computer interface output modalities can be made.

Subject of Research: Comparing brain-computer interface output modalities for motor recovery after stroke

Article Title: Effects of brain–computer interface output modalities on motor recovery after stroke: a systematic review and network meta-analysis

Article References: Jiang, H., Zhang, X., Xu, J., Fu, Y., Jin, N., Xiao, Y., & Fang, L. (2026). Effects of brain–computer interface output modalities on motor recovery after stroke: a systematic review and network meta-analysis. Frontiers in Neuroscience, 20, 1978309. https://doi.org/10.3389/fnins.2026.1978309

Image Credits: AI Generated

DOI: 10.3389/fnins.2026.1978309

Keywords: brain-computer interface, stroke, motor recovery, functional electrical stimulation, robotic rehabilitation, network meta-analysis, systematic review, Fugl-Meyer Assessment, rehabilitation, neuroprosthetics, Effects, brain