To develop a low-cost, AI-driven rehabilitation exergame that assesses upper-limb motor function and provides therapy in post-stroke individuals.
Key Findings:
Strong correlations between extracted features and FMA scores, indicating the model's effectiveness.
High predictive performance of the regression model (Spearman ρ = 0.92, R² = 0.89, RMSE = 4.42), suggesting reliability in assessments.
Severity classification accuracy ranged from 86% to 93%, highlighting the model's potential for clinical use.
Interpretation:
The AI-driven exergame provides an interpretable and efficient method for assessing upper-limb function, reducing the need for specialized personnel and potentially transforming rehabilitation practices.
Limitations:
Study involved a small sample size of twelve post-stroke individuals, which may limit the generalizability of the findings.
Datasets are not publicly available at this stage, which could hinder further research.
Conclusion:
The proposed framework is scalable, sensor-free, and offers immediate feedback, making it suitable for telerehabilitation and remote monitoring, with potential for future research to expand its application.
