Auto Detection of Parkinson’s Disease based on Objective Measurement of Gait Parameters using Wearable Sensors

Abstract

Neurological diseases such as Parkinson’s disease (PD) are becoming a major problem worldwide due to increasing longevity and the resulting aging population. Since PD is progressive in nature, subjective assessment and monitoring of PD patients at different stages is cumbersome and time consuming. The innovation of low-cost wearable sensors and their suitability for objective gait parameter measurements facilitate the personalization of clinical treatment. With the development of machine learning techniques, it has also become possible to detect PD automatically based on gait parameters. In this study, we proposed a method for measuring gait parameters using wearable sensors and identified PD patients automatically based on machine learning techniques. The subjects of this study were 40 patients with PD and 40 control patients; the experimental set-up included three-dimensional (3D) motion analysis. Wearable devices were placed on the knees and ankles. Spatiotemporal gait parameter data were collected via 3D motion analysis. A performance comparison was conducted using different classification techniques with reduced feature sets obtained by the random feature elimination and principal component analysis methods. We obtained a maximum accuracy of 88.89% using a support vector machine with a radial basis function combined with a random feature elimination set. Our results demonstrate that the proposed technique can assist medical practitioners to distinguish PD patients from control groups using gait data. This method is recommended to medical practitioners for clinical implementation to personalize the monitoring and treatment of PD at different stages. © 2018 SERSC Australia.