Objective: To systematically review training protocols integrating machine learning (ML) models in neurological rehabilitation, focusing on balance recovery predictions and their clinical applicability.

Background: Balance impairment is a major disability factor in neurological disorders such as stroke, Parkinson’s disease, and cerebral palsy. Machine learning has emerged as a promising tool for predicting recovery patterns, optimizing rehabilitation strategies, and improving patient outcomes. However, the integration of ML-based approaches into clinical training protocols remains inconsistent, necessitating a systematic evaluation.

Method: A comprehensive literature search was conducted across PubMed, Web of Science, Scopus, and other databases up to March 2024. Inclusion criteria encompassed studies implementing ML models to predict balance recovery in neurological rehabilitation. Data extraction focused on study design, ML algorithms, training protocols, and clinical outcomes. The quality of studies was assessed using PRISMA guidelines, and a meta-analysis was performed where applicable.

Results: A total of 27 studies met inclusion criteria, employing supervised learning (64%), reinforcement learning (21%), and hybrid models (15%). Recurrent neural networks (RNNs) and support vector machines (SVMs) demonstrated high predictive accuracy (>85%) for balance recovery in post-stroke and Parkinson’s disease rehabilitation. Training protocols incorporating ML-driven feedback mechanisms improved patient adherence and recovery rates. However, variability in dataset size, feature selection, and validation strategies limited generalizability.

Conclusion: ML-assisted balance recovery prediction is a rapidly evolving field with significant clinical implications. Despite promising results, further research is needed to standardize ML integration in rehabilitation protocols, enhance model interpretability, and ensure real-world applicability. Future studies should emphasize large-scale, multi-center trials to validate findings and optimize training strategies for personalized rehabilitation.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/harnessing-machine-learning-to-predict-balance-recovery-a-systematic-review-of-training-protocols-in-neurological-disorders/