Machine Learning to Improve Energy Expenditure Estimation in Children With Disabilities: A Pilot Study in Duchenne Muscular Dystrophy

Abstract

BackgroundChildren with physical impairments are at a greater risk for obesity and decreased physical activity. A better understanding of physical activity pattern and energy expenditure (EE) would lead to a more targeted approach to intervention.ObjectiveThis study focuses on studying the use of machine-learning algorithms for EE estimation in children with disabilities. A pilot study was conducted on children with Duchenne muscular dystrophy (DMD) to identify important factors for determining EE and develop a novel algorithm to accurately estimate EE from wearable sensor-collected data.MethodsThere were 7 boys with DMD, 6 healthy control boys, and 22 control adults recruited. Data were collected using smartphone accelerometer and chest-worn heart rate sensors. The gold standard EE values were obtained from the COSMED K4b2 portable cardiopulmonary metabolic unit worn by boys (aged 6-10 years) with DMD and controls. Data from this sensor setup were collected simultaneously during a series of concurrent activities. Linear regression and nonlinear machine-learning–based approaches were used to analyze the relationship between accelerometer and heart rate readings and COSMED values.ResultsExisting calorimetry equations using linear regression and nonlinear machine-learning–based models, developed for healthy adults and young children, give low correlation to actual EE values in children with disabilities (14%-40%). The proposed model for boys with DMD uses ensemble machine learning techniques and gives a 91% correlation with actual measured EE values (root mean square error of 0.017).ConclusionsOur results confirm that the methods developed to determine EE using accelerometer and heart rate sensor values in normal adults are not appropriate for children with disabilities and should not be used. A much more accurate model is obtained using machine-learning–based nonlinear regression specifically developed for this target population.

Automatic energy expenditure measurement for health science

“…Devices such as Apple Watch, Microsoft Band, Nike+ Fuelband, Jawbone Up, Fitbit apply machine learning algorithms and utilize from the accelerometer sensors. However, a recent study showed that methods which use accelerometer values in conjunction with heart rates cannot be used for children with disabilities [23]. Therefore, models which work for normal adults might not provide good performance for some cases.…”

A Survey on Energy Expenditure Estimation Using Wearable Devices

“…As can be seen, the subject number varies from 10 (the minimum considered number) to a hundred. The age range is also diverse, normally adults from 25 to 45 years old, but there are several papers related to EEE in children [16,45,48,75,80,94,109,120,123] and in older adults [28,41,113], interesting because their EE can differ from that of young adults. Obese individuals' EE was also studied [20,97,128] although the number of activities that they had to perform in the validation process was normally very reduced.…”

“…Not all of the datasets include a health and functional screening evaluation, but normally all the subjects were supposed to be free of major chronic conditions. There are, however, some studies related to special groups such as pregnant women [124], wheelchair users [57], Duchenne muscular dystrophy patients [94,136] or chronic obstructive pulmonary disease patients [30,99,125].…”

Validation of a method for the estimation of energy expenditure during physical activity using a mobile device accelerometer

“…The classification of the activities influences the estimation of energy expenditure, and several authors created some methods for the identification of the types of activities. In [60], the authors created a method with linear (regression) and nonlinear (machine-learning-based) models for the estimation of energy expenditure, comparing the accelerometer values with the values obtained with Cosmed K4b2, reporting a correlation of 91%.…”