Quantitative Assessment of Lower Limb Spasticity Using sEMG Data and Bayesian-Optimised SVM

Evaluation of lower limb spasticity plays a pivotal role in enhancing rehabilitation outcomes for stroke patients. Traditional assessment methods, such as Modified Ashworth Scale (MAS), are prone to subjective biases, resulting in inconsistent evaluations among clinical practitioners. On the contrary, neurophysiological approaches, particularly the analysis of surface electromyography (sEMG) data, offer a more objective and sensitive alternative. While previous research has predominantly focused on upper limb spasticity, there exists a notable gap in the evaluation of lower limb spasticity. In this study, we introduce a Bayesian-optimized Support Vector Machine (SVM) classifier for the objective assessment of lower limb spasticity using sEMG data. The experiment involved seven participants at four MAS levels, and the results showed an average recognition accuracy exceeding 75% with optimal model and sEMG features. This innovative method holds promising potentials in the enhancement of objectivity and precision in lower limb spasticity assessment, effectively addressing the limitations associated with subjective biases in traditional evaluation approaches.