Impact Factor 3.877
2017 JCR, Clarivate Analytics 2018

The world's most-cited Neurosciences journals

This article is part of the Research Topic

Machine Learning in Neuroscience

Original Research ARTICLE Provisionally accepted The full-text will be published soon. Notify me

Front. Neurosci. | doi: 10.3389/fnins.2019.00398

A Regression-Based Framework for Quantitative Assessment of Muscle Spasticity Using Combined EMG and Inertial Data from Wearable Sensors

 Xu Zhang1*,  Xiao Tang2, Xiaofei Zhu2, Xiaoping Gao3 and  Xiang Chen2
  • 1Department of Electronic Science and Technology, University of Science and Technology of China, China
  • 2University of Science and Technology of China, China
  • 3First Affiliated Hospital of Anhui Medical University, China

There have always been practical demands for objective and accurate assessment of muscle spasticity beyond its clinical routine. A novel regression-based framework for quantitative assessment of muscle spasticity is proposed in this paper using wearable surface electromyogram (EMG) and inertial sensors combined with a simple examination procedure. Sixteen subjects with elbow flexor or extensor (i.e., biceps brachii muscle or triceps brachii muscle) spasticity and eight healthy subjects were recruited for the study. The sEMG and inertial data were recorded from each subject when a series of passive elbow stretches with different stretch velocities were conducted. In the proposed framework, both lambda model and kinematic model were constructed from the recorded data, and biomarkers were extracted respectively from the two models to describe the neurogenic component and biomechanical component of the muscle spasticity, respectively. Subsequently, three evaluation methods using supervised machine learning algorithms including single-/multi-variable linear regression and support vector regression (SVR) were applied to calibrate biomarkers from each single model or combination of two models into evaluation scores. Each of these evaluation scores can be regarded as a prediction of the modified Ashworth scale (MAS) grades for spasticity assessment with the same meaning and clinical interpretation. In order to validate performance of three proposed methods within the framework, a 24-fold leave-one-out cross validation was conducted for each of all subjects. Both methods with each individual model achieved satisfactory performance, with low mean square error (MSE, 0.14 and 0.47) between the resultant evaluation score and the MAS. By contrast, the method using SVR to fuse biomarkers from both models outperformed other two methods with the lowest MSE at 0.059. The experimental results demonstrated the usability and feasibility of the proposed framework, and it provides an objective, quantitative and convenient solution to spasticity assessment, suitable for clinical, community and home-based rehabilitation.

Keywords: spasticity assessment, machine learning, Surface electromyogram, Stroke, Home-based rehabilitation

Received: 19 Dec 2018; Accepted: 08 Apr 2019.

Edited by:

Ali Ghazizadeh, Sharif University of Technology, Iran

Reviewed by:

Zhong Yin, University of Shanghai for Science and Technology, China
Chris A. McGibbon, University of New Brunswick Fredericton, Canada  

Copyright: © 2019 Zhang, Tang, Zhu, Gao and Chen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

* Correspondence: Dr. Xu Zhang, University of Science and Technology of China, Department of Electronic Science and Technology, Hefei, China, xuzhang90@ustc.edu.cn