MACHINE LEARNING MODELS FOR PREDICTING LABORATORY-BASED PHYSICAL ACTIVITY TYPE FROM CONSUMER WEARABLES IN CHILDREN

Authors

James W. White III, University of South CarolinaFollow
R. Glenn Weaver, University of South CarolinaFollow
Olivia Finnegan, University of South CarolinaFollow
Aliye B. Cepni, University of South CarolinaFollow
Xuanxuan Zhu, University of South CarolinaFollow
Lauren von Klinggraeff, University of South CarolinaFollow
Hannah Parker, University of South CarolinaFollow
Meghan Bastyr, University of South CarolinaFollow
Nicholas Niako, University of South CarolinaFollow
Massimiliano de Zambotti, University of South CarolinaFollow
Gregory J. Welk FACSM, Iowa State UniversityFollow
Srihari Nelakuditi, University of South CarolinaFollow
Yuan Wang, University of South CarolinaFollow
Sarah Burkart, University of South CarolinaFollow
Rahul GhosalFollow

Abstract

BACKGROUND: Accurate classification of physical activity type could improve estimates of free-living (i.e., occurring in everyday environments) physical activity energy expenditure (PAEE) in children. Studies have shown pattern recognition approaches trained on accelerometer data from research-grade devices can predict children’s free-living activities. However, there has been limited examination of the potential of using accelerometry (accel) and heart rate (HR) data from consumer wearables for increasing the accuracy of children’s activity type prediction. Thus, this study evaluated underlying accel and HR features from consumer wearables to predict children’s activities compared to accel and HR data from research-grade devices. METHODS: One hundred ninety-one children (5-12 years, 53% male, 57% White) completed a 60-minute protocol consisting of simulated free-living activities (e.g., walking, running, soccer, etc.). These activities were directly observed and categorized into four activity classes: Lying Down, Enrichment, Fundamental Movement Skills (FMS), and Sports/Games. Using the underlying accel and HR data from wrist-placed consumer wearables (i.e., Apple Watch Series 7 and Garmin Vivoactive 4S) and the combined accel and HR data (i.e., ActiGraph accel+Actiheart HR) from a wrist-placed, research-grade accelerometer (i.e., ActiGraph GT9X) and a chest-placed research-grade device (i.e., Actiheart 5), 13 time and frequency domain features were extracted at each second and 21 additional features were extracted using 60-second sliding windows for Random Forest model training. Leave-one-subject-out cross validation was used to evaluate the performance of each model. RESULTS: Underlying data from Apple exhibited the highest accuracy (90.8%, 95%CI: 90.7%, 90.9%) followed by ActiGraph+Actiheart (87.6%, 95%CI: 87.6%, 87.7%), and Garmin (85.5%, 95%CI: 85.4%, 85.6%). Apple also exhibited the highest sensitivity (87%, 91%, 92%, 78%) and specificity (98%, 95%, 94%, 97%) across the four activity classes, respectively. CONCLUSIONS: These results demonstrate the potential of consumer wearables to predict children’s activities with similar (or better) precision than research-grade devices. Future studies should evaluate whether features from these devices can accurately predict children’s activities in free-living environments. Grant or funding information: Research reported in this abstract was supported by the National Institute of Diabetes and Digestive and Kidney Diseases under Award Numbers F31DK136205 and R01DK129215 of the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Recommended Citation

White III, James W.; Weaver, R. Glenn; Finnegan, Olivia; Cepni, Aliye B.; Zhu, Xuanxuan; von Klinggraeff, Lauren; Parker, Hannah; Bastyr, Meghan; Niako, Nicholas; de Zambotti, Massimiliano; Welk, Gregory J. FACSM; Nelakuditi, Srihari; Wang, Yuan; Burkart, Sarah; and Ghosal, Rahul (2024) "MACHINE LEARNING MODELS FOR PREDICTING LABORATORY-BASED PHYSICAL ACTIVITY TYPE FROM CONSUMER WEARABLES IN CHILDREN," International Journal of Exercise Science: Conference Proceedings: Vol. 16: Iss. 3, Article 185.
Available at: https://digitalcommons.wku.edu/ijesab/vol16/iss3/185