Comparative efficacy of motor imagery augmented with central non-invasive brain stimulation versus peripheral electrical stimulation for upper extremity rehabilitation post-stroke: a systematic review and network meta-analysis

Upper limb dysfunction is a common and debilitating consequence of stroke, severely affecting patients’ activities of daily living and quality of life. Motor imagery (MI) has emerged as a promising rehabilitation technique, and its combination with various forms of non-invasive stimulation, both central (e.g., repetitive transcranial magnetic stimulation, rTMS; transcranial direct current stimulation, tDCS) and peripheral (e.g., functional electrical stimulation, FES), has been increasingly investigated. While previous meta-analyses have confirmed the general benefit of combined interventions, the relative efficacy of different MI-based combination strategies remains unclear. This systematic review and network meta-analysis aimed to directly and indirectly compare the effectiveness of MI augmented with different non-invasive central or peripheral stimulation modalities for upper extremity recovery post-stroke. We registered the study on PROSPERO (CRD420251131264) and followed the PRISMA guidelines. Randomized controlled trials (RCTs) were searched in PubMed, Cochrane Library, EMBASE, Scopus, CNKI, and Wanfang databases from inception until August 4, 2025. The included RCTs involved adult stroke patients with upper limb dysfunction receiving MI combined with any non-invasive stimulation. The primary outcome was the change in upper limb motor function measured by the Fugl-Meyer Assessment (FMA or FMA-UE). A frequentist network meta-analysis was performed using random-effects models. Risk of bias was assessed using the Cochrane RoB 2 tool. Subgroup, sensitivity, and meta-regression analyses were conducted to explore heterogeneity. Seventeen RCTs involving 846 participants were included in the systematic review, with 13 studies forming the network for meta-analysis, comparing 9 intervention strategies. Network meta-analysis for the FMA outcome showed that MI combined with low-frequency rTMS (MI-LF-rTMS) showed a statistically significant difference compared to conventional rehabilitation alone (Standardized Mean Difference, SMD = 1.755, 95% CI 0.631 to 2.879, p = 0.002). No other intervention, including MI-tDCS, MI-FES, or any single therapy, showed a statistically significant difference compared to conventional rehabilitation. MI-LF-rTMS also showed a statistically significant difference in upper limb functional activity (Action Research Arm Test). Subgroup analyses indicated that the statistically significant difference for MI-LF-rTMS was also observed across intervention durations ≤ 4 weeks, disease stages ≤ 3 months post-stroke, and in protocols not using brain-computer interface technology. Meta-regression identified that the use of a brain-computer interface, publication year, and patient mean age were significant sources of heterogeneity. Among the intervention strategies evaluated in this network meta-analysis, motor imagery combined with low-frequency repetitive transcranial magnetic stimulation (MI-LF-rTMS) showed a statistically significant difference compared to conventional rehabilitation. This regimen integrates central neuromodulation with cognitive training and may be a clinically feasible option, particularly for patients in the early phase after stroke. Future research should focus on parameter optimization, mechanistic exploration, and validation in larger, more diverse populations.