Evaluation of Stroke Rehabilitation Using Smart Pressure Insoles and Ensemble Learning Algorithm

Stroke is the second leading cause of disability and mortality worldwide. Extensive researches on rehabilitation assessment for stroke survivors using wearable plantar pressure sensors often suffer from troublesome processes and unsatisfactory results due to excessive sensors and sparse pressure sensing structures. To address this, we developed an automatic assessment system for the recovery stage of stroke patients based solely on smart high-density flexible wearable plantar pressure insoles and an ensemble learning classification algorithm (ELCA). This system quantitatively evaluates the stroke patient's recovery progress. The data from 37 stroke patients were collected using self-developed smart insoles, from which multidimensional high-order features were extracted. ELCA employing seven basic classifiers was then developed to quantitatively assess recovery stages. Finally, the leave-one-out method was used to evaluate the performance of classifiers. The results show the ELCA has the best performance, achieving an accuracy, specificity, and sensitivity of up to 91.890/0, 97.30%, and 91.89% respectively, preliminarily demonstrating the system possesses excellent capabilities in quantitatively assessing rehabilitation progress. This solution is expected to provide an objective and practical method for long-term home rehabilitation monitoring of stroke patients.