Artificial Intelligence Applications in Home-Based Stroke Rehabilitation: A Scoping Review of Technologies, Patient Experiences, and Implementation Barriers

Background: Stroke is a leading cause of long-term adult disability worldwide. Rehabilitation services are progressively shifting from inpatient settings to home and community environments, where artificial intelligence (AI) technologies are increasingly applied — either integrated with conventional rehabilitation devices (e.g., intelligent rehabilitation robotics, AI-enabled wearable monitoring, intelligent virtual reality training systems) or as standalone AI algorithms/software (e.g., machine-learning-based motor assessment algorithms, risk prediction models, rehabilitation guidance applications, and large-language-model-assisted decision tools). No review has simultaneously integrated technology applications, patient/caregiver/clinician experiences, and implementation barriers specifically for home-based stroke rehabilitation under a strict AI operational definition. Objective: To map the evidence on AI applications in home-based stroke rehabilitation across three axes: (1) technology types and applications; (2) patient, caregiver and clinician experiences, perceptions, and acceptance; (3) implementation barriers and facilitators. Methods: A scoping review following the JBI methodology and reported per PRISMA-ScR and PRISMA-S. Seven English-language databases (PubMed, Embase, CINAHL, Cochrane Library, Web of Science, PsycINFO, IEEE Xplore) will be searched from 1 January 2020 to the date of search execution (planned immediately after registration approval). The search strategy uses four AND-connected concept blocks (stroke × rehabilitation × AI [mandatory] × home/community) with a strict AI operational definition (integration of ML/DL or equivalent algorithms, or rule-based reasoning/expert systems, used for patient outcome inference). Records will be screened independently by two reviewers in two stages; data will be charted using a JBI-based form with three-axis coding; MMAT (2018) will be used for descriptive quality appraisal; results will be synthesised using the PAGER framework. English only. Registration statement: This protocol was prospectively registered on OSF prior to the literature search (DOI: to be assigned).