Data-Driven Model for Upper Limb Spasticity Detection

Abstract

Upper limb spasticity (ULS) is a common pathophysiological changes manifest by a structural damage towards the central nervous system (CNS) that includes brain and spinal cord. The current clinical practice of spasticity assessment utilizes Modified Ashworth Scale (MAS) as a subjective tool to measure the severity of spasticity. Lack of objective value, poor sensitivity in detecting minimal changes, and dependency to the interpretation by the assessing clinicians are the several reasons of the inter and intra-rater variability of the measurement using MAS. These limit the use of MAS in diagnosing, treating, and monitoring spasticity especially in inexperienced clinicians, hence leading to inadequate spasticity management. To overcome this problem, a study is carried out to quantify and develop a data-driven model of ULS detection based on MAS. The characteristics that detect the existence of ULS according to MAS are identified and adopted to train the machine learning models for smart diagnosis purpose to assist the physicians to effectively manage spasticity.