•  
  •  
 

COMPARISON OF RAW ACCELERATION FROM CONSUMER WEARABLES AND ACTIGRAPH ACCELEROMETERS USING A MECHANICAL SHAKER TABLE

Abstract

James W. White III, Nick Tindall, Olivia Finnegan, Kasey Hansen, Meghan Bastyr, Hannah Parker, Roddrick Dugger, Elizabeth L. Adams, Sarah Burkart, Bridget Armstrong, Michael W. Beets, R. Glenn Weaver. University of South Carolina, Columbia, SC.

BACKGROUND: Though the proprietary signal processing of acceleration output from consumer wearables limits their use for research on physical activity (PA) and sleep assessment, it may be possible to develop open-source prediction equations for estimating PA and sleep based on raw acceleration estimates from these devices. Thus, the aim of this study was to compare raw acceleration output from ActiGraph wGT3X-BT (ActiGraph) and consumer wearables (i.e., Garmin Vivoactive 4S [Garmin] and Apple Watch Series 7 [Apple]) using a mechanical shaker table (Scientific Industries; Mini-300 Orbital Genie, Model 1500). METHODS: A total of 30 devices, including 10 ActiGraph accelerometers and 10 of each consumer wearable were analyzed in this study. Validity of raw acceleration estimates from consumer wearables was tested against a criterion of ActiGraph. Devices were mounted directly to the twin ratcheting clamps of the shaker table and were oscillated at various speeds (i.e., 0.6 Hz, 1.0 Hz, 1.5 Hz, 1.9 Hz, 2.4 Hz, 2.8 Hz, and 3.2 Hz) for 2-minutes each (i.e., 7 speeds for 2 minutes each) until all consumer wearables were compared to all ActiGraph devices. The raw acceleration values for the x, y, and z axes were extracted from the middle minute of each 2-minute speed, and the maximum vector magnitude was calculated for each second. Pearson product moment (r) and Lin’s concordance correlation coefficients (CCC) were calculated. Bland-Altman plots were also constructed with mean bias and 95% limits of agreement. RESULTS: The correlations of Garmin and Apple with Actigraph were r=0.881 and r=0.933, respectively. CCC from raw acceleration estimates for Garmin and Apple were 0.763 and 0.918, respectively. Bland-Altman plots (consumer wearable minus ActiGraph) revealed mean differences 0.044 (95% CI: -0.054, 0.142) between Garmin and ActiGraph and -0.002 (95% CI: -0.097, 0.094) between Apple and ActiGraph. CONCLUSIONS: There was moderate concordance and strong correlation between raw acceleration estimates from Garmin and ActiGraph, while there was strong concordance and correlation between raw acceleration estimates from Apple compared to ActiGraph. Garmin and Apple provide comparable estimates of raw acceleration compared to ActiGraph, suggesting that raw acceleration estimates from consumer wearables can be used to develop open-source prediction equations for estimating PA and sleep. Grant or funding information: Research reported in this abstract was supported by the National Institute of Diabetes and Digestive Kidney Diseases of the National Institutes of Health under Award Number R01DK129215. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

This document is currently not available here.

Share

COinS