An integrated machine learning approach for spasticity assessment combining passive and active movements

Spasticity is a common sequela of neurological disorders and a highly disabling condition often associated with pain, joint contractures and reduced limb functionality. Assessing spasticity is essential for monitoring progress and optimizing rehabilitation. Currently, evaluation primarily relies on passive movements and clinical scales, which may not fully capture the impact of spasticity on motor functions. Utilizing active movements and objective methods can offer valuable insights into the underlying motor impairments. The aim of this study was to propose an objective method for evaluating spasticity, utilizing passive and active movements. Fifteen participants were involved in this study, comprising 10 individuals with spasticity and 5 healthy controls. Participants performed 15 passive and 15 active movements while surface electromyographic (sEMG) signals were recorded from the biceps and triceps brachii. Machine Learning (ML) algorithms were applied to classify the groups based on their spasticity severity and their classification performance was evaluated, based on the level of individuals' impairment. The results showed that the combination of data from passive and active movements provided superior accuracy compared to when the movements were analyzed separately, reaching 93.22% using the Support Vector Machine (SVM) algorithm. In conclusion, the combination of active and passive movements for spasticity assessment, along with the application of ML, that demonstrated high performance in this task, presents a promising alternative for clinical practice. This approach can supports therapists in making more accurate evaluations and tailoring treatments to individual patients, thereby promoting a patient-centered approach and enhancing rehabilitation outcomes for individuals with spasticity.