Functional and motoric outcome of AI-assisted stroke rehabilitation: A meta-analysis of randomized controlled trials

Abstract

Introduction: Stroke is the primary contributor to disability worldwide, causing a high economic burden due to its morbidity. Due to the application of artificial intelligence (AI), stroke rehabilitation has been revolutionized, resulting in significant improvement. Implementing AI also enables home-based care, thus helping stroke patients who generally have ambulatory difficulties. Methods: This research was a systematic review from Pubmed, ScienceDirect, and ProQuest, including randomized controlled trials (RCT) published from 2009 to 2024. Meta-analysis included seven studies discussing the functional and motoric outcomes of AI-assisted stroke rehabilitation. Results: Six studies included post-stroke patients within 3 to 6 months after the stroke occurred. AI models used were varied, ranging from end-effector or exoskeleton robots to a combination of both and virtual reality (VR). Overall, the included studies had a low risk of bias. Standard mean differences (SMDs) of the Barthel Index and Motricity Index were 0.16 and 0.60. No significant difference between AI-assisted stroke rehabilitation and conventional stroke rehabilitation for both outcomes. Non-inferiority trials showed that the AI-assisted method was not inferior to the conventional method of stroke rehabilitation. Discussion: Considering its feasibility, personalization, and flexible rehabilitation program, AI-assisted was non-inferior to the conventional method. A comprehensive guideline is needed to facilitate its usage in clinical practice. Conclusion: AI-assisted stroke rehabilitation was not inferior to conventional stroke rehabilitation.