Artificial intelligence for gait and balance in neurological disorders: a scoping review of clinical applications and technologies

Clinical Snapshot

Key Findings

• Effect Size: Not applicable (scoping review)

• Primary Outcome: Mapping of AI applications in neurological gait and balance rehabilitation

• Nnt Or Sensitivity: 72.22% of studies used AI for prognostic purposes; 88.9% employed machine learning approaches

Clinical Application

Limited by lack of prospective validation and methodological barriers to clinical translation Relevant to Australian neurological rehabilitation services, but requires local validation studies and consideration of TGA approval pathways for AI-based medical devices Adults with stroke, Parkinson's disease, or multiple sclerosis requiring gait and balance assessment or rehabilitation

Abstract

BACKGROUND: Artificial intelligence (AI) has rapidly emerged within healthcare systems and neurological rehabilitation with the potential to revolutionize clinical decision-making and therapeutic strategies. However, a comprehensive understanding of how AI is currently applied to gait and balance rehabilitation in stroke, Parkinson's disease (PD), and multiple sclerosis (MS) is still lacking. OBJECTIVE: To map the current use of AI in neurological rehabilitation, focusing on clinical purposes, geographical distribution, and applied technologies for gait and balance rehabilitation. METHODS: Following the PRISMA-ScR statement, we conducted a literature search through MEDLINE, Cochrane CENTRAL, EMBASE, and Google Scholar up to July 2025 to identify studies applying AI-based methods to gait and balance outcomes in adults with stroke, PD, or MS. Study characteristics, AI methods, validation strategies, clinical purpose, and motor outcomes were extracted and synthesized narratively. RESULTS: Eighteen studies published from 2009 to 2025 were included. Most studies were conducted in Asia (50%) and involved people with stroke (77.8%). AI was predominantly used for prognostic purposes (72.22%), such as predicting falls, gait recovery, or treatment response and diagnostic applications (33.3%). Machine learning was the most common approach (88.9%) with Random Forest, Support Vector Machine, logistic regression, and eXtreme Gradient Boosting being the most frequently applied algorithms. None of the included studies performed prospective or external validation on independent datasets. CONCLUSIONS: This scoping review provides a comprehensive overview of current AI applications, highlighting promising but still immature approaches in the neurological rehabilitation of gait and balance. However, substantial methodological limitations remain major barriers to clinical translation.

References

1. 1.Pegorini, C., Cattaneo, D., Meotti, M., Baglio, F., Mannini, A., Cordani, C., & Gervasoni, E. (2026). Artificial intelligence for gait and balance in neurological disorders: a scoping review of clinical applications and technologies. Journal of Neurology. https://doi.org/10.1080/10749357.2019.1574417