Lower Limb Motor Function Assessment With Muscle Synergy and Interpretable Machine Learning: A Pilot Study in Healthy Individuals and Stroke Patients

Lower limb motor dysfunction severely impacts stroke patients’ daily activities and quality of life. Traditional assessment methods often rely on subjective clinical scales, which are time-consuming and can lack sensitivity to subtle functional changes. Recently, automatic limb motor function evaluation using wearable sensors and machine learning has gained significant attention. However, this new field faces challenges like difficulties in creating personalized feature representations and poor interpretability of evaluation results. Consequently, this paper presents a novel lower limb motor function evaluation method that considers both motion specificity and individual differences. Firstly, an adaptive muscle synergy analysis framework was developed to quantify inter-individual differences in muscle synergies and activation patterns across different groups, offering quantitative assessment indicators. Then, a multidimensional surface Electromyography (sEMG) feature set was constructed with redundant features removed, aiming to fully reveal neuromuscular activation information in sEMG signals. Subsequently, demographic and clinical features were incorporated to further construct feature representations reflecting individual differences. On this basis, machine learning algorithms were combined to automatically predict lower limb motor function scores or functional impairment degrees. Finally, SHapley Additive exPlanation analysis technology was introduced to perform interpretable analysis of evaluation results from both local and global perspectives. Experimental validation was conducted on a self-constructed HI-SP dataset comprising data from multiple healthy individuals and stroke patients. Results indicate that the proposed method achieves superior performance in two lower limb motor function evaluation tasks. The study demonstrates strong clinical relevance by providing an objective, quantitative, and interpretable tool that can support clinicians in developing personalized rehabilitation treatment plans and tracking patient progress with high precision.