A data-informed framework for individualized exercise prescription in post-stroke gait rehabilitation: a pilot randomized feasibility study

Most individuals with stroke experience walking difficulties, while standardized exercise training may not address heterogeneous gait deviations. This pilot feasibility study evaluated a data-informed intervention for individuals with post-stroke lower-extremity impairments. Participants were randomized to a data-informed intervention or control group. In the experimental group, exercises were prescribed using a predefined biomechanical mapping framework based on lower-limb gait deviations identified through inertial measurement units and Statistical Parametric Mapping. The control group received standardized task-oriented training. Both groups completed five exercises per session across 12 supervised sessions over 6 weeks. Feasibility and acceptance were evaluated based on attendance, safety, and participant feedback. Gait, clinical, and sEMG outcomes were analyzed exploratorily using aligned rank transform models, supplemented by selected sensitivity and participant-level analyses. Seventeen individuals registered, sixteen met the eligibility criteria, and all enrolled participants completed the intervention and post-intervention assessment. All participants attended every scheduled session, with no dropouts or intervention-related adverse events. Participant feedback indicated good acceptance of the intervention. For the exploratory outcomes, the experimental group showed a greater numerical reduction in Gait Profile Score, although the group-by-time interaction was not significant ( p = 0.187). Exploratory group-by-time interactions favoring the experimental group were observed for the composite gait deviation score ( p = 0.014), FMA-LE ( p = 0.032), TUG ( p = 0.008), and TIS ( p = 0.006). No significant interactions were observed for SIPSO, SPPB, BBS, rectus femoris S-iEMG ( p = 0.294), or tibialis anterior S-iEMG ( p = 0.596). Overall, this pilot feasibility study demonstrated that the data-informed assessment-to-prescription workflow was feasible, safe, and acceptable under supervised laboratory conditions. This was supported by complete attendance, no dropouts, full intervention completion, and positive participant feedback. The exploratory gait, clinical, and sEMG findings warrant further investigation in a larger, adequately powered trial.