Development and validation of a bone loading model for tracking adolescent physical activity bone scores: a preliminary report
Ren, Sicong
This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.
Permalink
https://hdl.handle.net/2142/115555
Description
Title
Development and validation of a bone loading model for tracking adolescent physical activity bone scores: a preliminary report
Author(s)
Ren, Sicong
Issue Date
2022-04-21
Director of Research (if dissertation) or Advisor (if thesis)
Zhu, Weimo
Doctoral Committee Chair(s)
Zhu, Weimo
Committee Member(s)
Chiu, Chung-Yi
Hernandez, Manuel
Sydnor, Synthia
Department of Study
Kinesiology & Community Health
Discipline
Kinesiology
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
Ph.D.
Degree Level
Dissertation
Keyword(s)
accelerometer
bone loading
osteoporosis
physical activity
machine learning
Abstract
Osteoporosis is associated with aging and menopause; however, predisposing factors are identified during childhood and adolescence. Achieving high peak bone mass via physical activity (PA) in adolescents has been found to be beneficial in preventing osteoporosis as they age. Mechanical loading produced in PA plays an important role in bone formation. Therefore, objective free-living measurement of external loading experienced by bones supports clinicians in helping adolescents optimize peak bone mass. Although wearable sensors provide relatively strong validity evidence in laboratory settings, their psychometric quality in free-living settings needs to be established. The primary goal of this study was to develop a bone loading score model that can track free-living PA in adolescents. The secondary goal was to cross-validate the optimal bone loading score model using accelerometer-based tracking data in free-living settings.
In Study I, accelerometer-based features (i.e., time- and frequency-domain features) were extracted from accelerometer data at the waist, wrists, and ankles. Statistical learning algorithms were used to model the relationship between acceleration features and normalized peak vertical ground reaction forces. There were five statistical models used in this study: multiple linear regression, least absolute shrinkage and selection operator regression, k-nearest neighbors, decision tree, and random forest, all of which were used to predict normalized peak vertical ground reaction forces in various movements as well as to select significant features. Each model was trained on the features from all combinations of the five site-worn accelerometers.
In Study II, data from the 2011-2012 National Health and Nutrition Examination Survey (NHANES) were used to cross-validate the bone loading model. This study tested the external validity of the bone loading model through (a) testing the differences in bone loading scores between males and females as well as between normal weight and overweight/obese groups; and (b) computing the correlations between bone loading scores and bone health outcomes by sex and overall.
The results showed that there were significant differences in bone loading scores between males and females as well as between normal weight and overweight/obese groups (p < 0.05). The bone loading model based on waist-worn (R-squared = 0.63 – 0.78) and left wrist-worn (R-squared = 0.37 – 0.65) accelerometers provided higher prediction accuracy than from other single-site sensors (R-squared = 0.28 – 0.62). Random forest as an embedded algorithm offered the most accurate prediction of normalized peak vertical ground reaction forces among all regressors (R-squared = 0.59 – 0.80). The waist-based bone loading model embedded with random forest and left wrist-based bone loading model embedded with random forest were found to be potentially optimal combinations for tracking free-living physical activity due to their small errors (RMSE = 0.63, 0.80). The cross-validation results confirmed that the left wrist-based bone loading model embedded with the random forest moderately reflected adolescent bone mineral contents at the (a) total body without head, (b) left arm, and (c) both legs (r = 0.42 – 0.45); and between bone loading scores and bone geometry areas at the (a) trunk, (b) left arm, and (c) both legs (r = 0.44 – 0.50).
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.