Integrating Physiological Signals With Technology-Driven Stroke Rehabilitation: A Systematic Review

Stroke remains a critical public health concern as the world’s leading cause of adult disability and the second leading cause of mortality, with the burden disproportionately affecting low-and middle-income countries (LMICs). Current stroke rehabilitation approaches often fall short due to inconsistent and insufficient therapy, lack of patient-specific data-driven protocols, limited access to specialized care, and a global shortage of trained professionals. This systematic review aims to critically appraise technology-driven post stroke rehabilitation trends and methodologies to identify gaps, limitations, and research opportunities. The goal is to inform the development of novel scalable, patient-centered solutions that can optimize both motor and cognitive recovery, considering the patient’s unique physical needs and adaptability and leveraging the short post stroke window of neuroplasticity. A decade-spanning literature search (2015–2025) was conducted based on ”Preferred Reporting Items for Systematic Reviews and Meta-Analyses” (PRISMA) guidelines across Scopus, Web of Science, and IEEE Xplore databases. Predefined keywords included stroke rehabilitation, AI, robotics, VR, and physiological signals (e.g., electromyography (EMG), electroencephalogram (EEG), electrocardiogram (ECG), electrodermal activity (EDA), and restrictive spirometric pattern (RSP)). Studies were included if they addressed post-stroke rehabilitation using one or more physiological signals and applied technology-driven rehabilitation systems such as robotics, brain-computer interface (BCI), virtual environments, or artificial intelligence (AI). Data extraction focused on intervention characteristics, technology integration, and clinical outcomes. Quality assessment considered test environment, protocol completeness, data type, cross-validation, and adaptability. Bibliometric analysis was performed using VOSviewer to map keyword co-occurrence and country-specific contributions. Of the 13,661 records initially screened, only 21 studies met the eligibility criteria. The bibliometric analysis highlighted the United States, China, and Italy as the most active research contributors, with “stroke rehabilitation,” “robotics,” and “virtual reality” emerging as dominant keywords. Overall, multi-modal physiological signal-based rehabilitation protocols demonstrated greater potential for adaptability and personalization as compared to unimodal approaches. However, key gaps were identified in usability, real-time system responsiveness, and standardized signal acquisition. A Pareto analysis revealed that a small subset (19.04%) of studies contributed most methodological rigor, indicating room for improvement in sensing/data acquisition technologies, data processing pipelines, and multi-domain integration. AI-driven, physiological signal modulated robotic and VR-based rehabilitation platforms, especially those based on multimodal signal fusion, show great promise towards enhancing the effectiveness, personalization, and accessibility of current post-stroke therapy. Future work should focus on improving user safety and comfort, reducing latency, and establishing accessible, cost-effective, standardized rehabilitation frameworks. This review offers a roadmap for designing these next-generation rehabilitation platforms based on integrating network physiology with technology-driven rehabilitation systems to deliver reliable, scalable, adaptive, and patient-specific care.