Performance Analysis of Quantitative Muscle Spasticity Model in Flexion and Extension Passive Movements

Stretching is a fundamental physical therapy technique for assessing spasticity through passive flexion and extension movements. The conventional procedure required clinical tools such as the Modified Ashworth Scale (MAS) to assess the spasticity level of joints according to the extent of muscle resistance, resulting in assessment inconsistencies that could impact the efficacy of the rehabilitation process. There is no evidence to suggest that either passive flexion or extension movements are independently used in measuring muscle spasticity during assessments. This study aims to explore the differentiation between passive flexion and extension movements in measuring muscle spasticity during assessments, as these movements may exhibit distinct characteristics using Mechanomyography (MMG) signals from forearm muscles mapped to MAS levels. The findings demonstrate the need for two distinct models to accurately assess spasticity in both movements. Data were collected from 30 patients with upper limb spasticity, including 29 stroke cases and one with cerebral palsy, all of whom provided informed consent under ethical approval. The dataset included 48 features extracted from ACC-MMG signals along the x, y, and z axes during flexion and extension movements of the biceps and triceps of the limb. Statistical analyses, including linear regression and one-way MANDVA, were conducted to examine the characteristics of MMG signal features for each movement. To improve spasticity detection, Decision Trees (DTs), K-Nearest Neighbour (KNN), Linear Discriminant Analysis (LDA), and Support Vector Machines (SVM) classifiers were applied. The KNN-based classifier attained outstanding results, with an accuracy of 90.12% for flexion and 86.42% for extension, showing flexion as the most effective movement for measuring spasticity.