Measuring Tremor Progression in Free-Living Conditions Using Wearable Sensors

Keywords

Loading...
Thumbnail Image

Issue Date

2022-06-30

Language

en

Document type

Journal Title

Journal ISSN

Volume Title

Publisher

Title

ISSN

Volume

Issue

Startpage

Endpage

DOI

Abstract

Wearable sensors have been used successfully to detect tremor in patients, but most studies have been conducted in highly controlled laboratory conditions, and the patients were only monitored for a short amount of time. Objectives: This project aims to investigate (1) whether we can use wearable sensors to accurately and objectively detect tremor in daily life and (2) if we can measure the progression of tremor over time. Methods: 74 features based on existing literature are extracted from the raw signal of the triaxial accelerometer of a smartwatch.We train and validate a logistic regression classifier on the labelled dataset Parkinson@Home. Subsequently, we apply the trained classifier to the unlabelled longitudinal dataset of the Personalised Parkinson Project (PPP). PPP is unprecedented in cohort size (650 PD patients) and length (2 years) of the study. Results and Conclusion: The results on PD@Home indicate that the classifier can accurately detect tremor with a sensitivity of 65:90% given a 94:99% specificity. When applied to PPP, the predicted daily tremor strongly correlates with the MDS-UPDRS tremor items. However, there is no apparent progression measurable. We conclude that two years might not be enough to measure the progression of tremor sufficiently and a study even longer than PPP is required to provide more definite evidence.

Description

Citation

Faculty

Faculteit der Sociale Wetenschappen