Reducing a complex two-sided smartwatch examination for Parkinson's Disease to an efficient one-sided examination preserving machine learning accuracy

TL;DR

This study investigates reducing a comprehensive two-sided smartwatch assessment for Parkinson's Disease to a simpler one-sided version, aiming to maintain machine learning classification accuracy for easier home-based screening.

Contribution

It introduces a method to simplify two-sided smartwatch assessments to one-sided ones without losing classification accuracy in Parkinson's Disease detection.

Findings

One-sided assessments maintain classification accuracy.

Largest PD sample size with two-hand smartwatch data.

Potential for easy home-based PD screening.

Abstract

Sensors from smart consumer devices have demonstrated high potential to serve as digital biomarkers in the identification of movement disorders in recent years. With the usage of broadly available smartwatches we have recorded participants performing technology-based assessments in a prospective study to research Parkinson's Disease (PD). In total, 504 participants, including PD patients, differential diagnoses (DD) and healthy controls (HC), were captured with a comprehensive system utilizing two smartwatches and two smartphones. To the best of our knowledge, this study provided the largest PD sample size of two-hand synchronous smartwatch measurements. To establish a future easy-to use home-based assessment system in PD screening, we systematically evaluated the performance of the system based on a significantly reduced set of assessments with only one-sided measures and assessed, whether we can maintain classification accuracy.