Resting-state fMRI correlates of self-balancing robot-assisted lower-limb rehabilitation after stroke: a pilot randomized controlled neuroimaging study

Background Lower-limb weakness and poor balance are major barriers to independent walking after stroke. Self-balancing robot-assisted training permits supported upright practice, repeated weight shifting and task-oriented lower-limb loading, but its training-related neural correlates remain unclear. Methods In this pilot randomized controlled neuroimaging study, 24 patients with subcortical stroke received 4 weeks of REX-assisted training plus conventional rehabilitation or dose-matched active upright lower-limb training plus conventional rehabilitation (n = 12 per group). The Berg Balance Scale (BBS) was the primary clinical outcome, the Fugl-Meyer Assessment of the Lower Extremity (FMA-LE) was the key secondary outcome, and the Postural Assessment Scale for Stroke (PASS) and Functional Ambulation Category (FAC) were exploratory outcomes. Resting-state functional MRI was acquired before and after treatment; 20 participants were retained for imaging analyses (REX, n = 11; control, n = 9). Group-by-time analyses examined amplitude of low-frequency fluctuations (ALFF), affected-side precentral-gyrus seed-based functional connectivity (FC), and interhemispheric correlation (IHC); local and global correlation measures were exploratory. Results Baseline-adjusted analyses favored REX for post-intervention BBS scores (adjusted difference, 12.75; 95% CI, 1.45–24.04; P = 0.029) and FMA-LE scores (adjusted difference, 7.19; 95% CI, 3.71–10.67; P < 0.001). FAC was also favorable in an exploratory analysis, whereas PASS did not show a statistically definitive between-group difference. Group-by-time effects were observed in ALFF, seed-based FC, and IHC analyses, involving the affected-side precuneus, non-lesioned-side association cortices, the affected-side precentral–middle temporal connectivity, and bilateral precuneus/posterior cingulate homotopic coordination. Local and global correlation interactions were not significant. Conclusion Four weeks of REX-assisted lower-limb rehabilitation was associated with favorable pilot clinical outcomes and selective resting-state fMRI changes involving posterior-midline, visual-association, frontal-control, and motor-association systems. These findings represent candidate training-related neural correlates and require confirmation in larger randomized studies. Trial registration: Chinese Clinical Trial Registry, ChiCTR2300079122; (registered on 26 December 2023, https://www.chictr.org.cn/)