Balance function and cortical activation patterns following speed- and duration-adapted robotic exoskeleton training in people with subacute stroke

Abstract Background Robotic exoskeleton training (RET) parameters in stroke rehabilitation exhibit variability, highlighting the need for optimization to improve therapeutic efficacy. This study quantified the effects of speed- and duration-adapted RET on balance function and cortical activation patterns in people with subacute stroke and explored the associated neurophysiological mechanisms. Methods Sixty participants were randomized into four groups using a 2 x 2 factorial design (speed: 0.2 m/s vs. 0.5 m/s; duration: 20 min vs. 50 min), training 5 days/week for 3 weeks. Outcome measures included the Berg Balance Scale (BBS), Timed Up and Go Test (TUGT) and cortical activation indices assessed using functional near-infrared spectroscopy (fNIRS), including changes in oxygenated hemoglobin concentration (ΔHbO₂) and laterality index (LI) in the primary motor cortex (M1), premotor cortex/supplementary motor area (PMC/SMA), and dorsolateral prefrontal cortex (DLPFC). Results All groups exhibited significant improvements in BBS and TUGT scores after the intervention. Compared with the low-speed / short-duration group, the high-speed / long-duration and low-speed / long-duration group showed significantly greater improvements in BBS scores, with no significant TUGT differences. Analysis of cortical activation indices analysis revealed that the high-speed / long-duration group exhibited significantly increased ΔHbO₂ in the affected M1 (A-M1), and more symmetrical interhemispheric activation pattern in the LI PMC / SMA and LI DLPFC . Correlation analysis confirmed that improvements in BBS scores were significantly positively correlated with ΔHbO 2 A−PMC / SMA and negatively correlated with changes in LI PMC / SMA recovery. Conclusion Combining higher speed and longer duration in RET may promote balance recovery and is associated with changes in cortical activation patterns. These findings provide preliminary evidence for optimizing RET parameters and offer a potential neurophysiological basis for personalized stroke rehabilitation.