A Wearable Accelerometer System for Unobtrusive Monitoring of Parkinson's Diease Motor Symptoms

Abstract

Parkinson's disease is a complex condition currently monitored at home with paper diaries which rely on subjective and unreliable assessment of motor function at nonstandard time intervals. We present an innovative wearable and unobtrusive monitoring system for patients which can help provide physicians with significantly improved assessment of patients' responses to drug therapies and lead to better-targeted treatment regimens. In this paper we describe the algorithmic development of the system and an evaluation in patients for assessing the onset and duration of advanced PD motor symptoms.

Automatic assessment of medication states of patients with Parkinson's disease using wearable sensors

“…The methods are proposed to detect and assess the severity of tremor [9][10][11][12][13][14][15][16][17], bradykinesia [9,11,16,18,19], and dyskinesia [10,12,13,17,18,[20][21][22][23][24][25][26]. Some of these methods accurately detect PD motor impairments during constrained and unconstrained activities.…”

“…There are a variety of approaches in the published literature that are applied to classify medication treatment status (ON versus OFF) [9,17,24,[27][28][29][30][31][32]. They use accelerometer and gyroscope sensors that are worn on different parts of the body (wrist [9,27,31,32], leg [9,27,28] or/and waist or trunk [9,17,24,[27][28][29][30]), and they use different windows sizes in the range of 3 seconds to 10 minutes with or without overlap for signal segmentation. For feature extraction, [9,27] extract temporal features, whereas [17,24,[28][29][30][31][32] extract both temporal and spectral features.…”

“…Multiple supervised classification methods are used: linear discriminant (thresholdbased) [24,[27][28][29], Support Vector Machine (SVM) [17,30]), Artificial Neural Network (ANN) [32], logistic regression [9], and deep learning [31].…”

“…Some of these approaches are activity-independent and providing continuous monitoring but they achieved accuracy lower than 80% [9,17,27,31,32] which is not adequate to be used in the routine clinical practice. Other methods trade continuous monitoring with performance and they achieved accuracy over 90%, but only when patients' activities are constrained [24,[28][29][30].…”

“…In addition, the majority of the approaches to date have involved classification algorithms that are trained using data from a group of well-characterized patients, and then generalized to individual patients for testing [9,17,24,31,32]. In this classification problem, these approaches may not generalize well because the accuracy is likely impacted by the significant variability between patients with respect to:…”