Related Experiment Video
Updated: Mar 1, 2026
09:18
Measurements of Motor Function and Other Clinical Outcome Parameters in Ambulant Children with Duchenne Muscular Dystrophy
Published on: January 12, 2019
10.9K
Machine Learning to Improve Energy Expenditure Estimation in Children With Disabilities: A Pilot Study in Duchenne
Amit Pande1, Prasant Mohapatra1, Alina Nicorici2
1University of California Davis, Department of Computer Science, Davis, CA, United States.
JMIR Rehabilitation and Assistive Technologies
|June 6, 2017
Summary
A new machine learning model accurately estimates energy expenditure (EE) in boys with Duchenne muscular dystrophy (DMD), outperforming existing methods for children with disabilities. This advancement aids in understanding physical activity and developing targeted interventions.
Area of Science:
- Biomedical Engineering
- Pediatric Physical Therapy
- Machine Learning in Healthcare
Background:
- Children with physical impairments face higher risks of obesity and reduced physical activity.
- Understanding energy expenditure (EE) patterns is crucial for effective interventions in this population.
Purpose of the Study:
- To evaluate machine learning algorithms for estimating EE in children with disabilities.
- To develop a novel algorithm for accurate EE estimation using wearable sensor data in boys with Duchenne muscular dystrophy (DMD).
Main Methods:
- Collected data from 7 boys with DMD, 6 healthy boys, and 22 adults using accelerometers and heart rate sensors.
- Used COSMED K4b2 as the gold standard for EE measurement during concurrent activities.
- Applied linear regression and nonlinear machine learning approaches to analyze sensor data against gold standard EE values.
Main Results:
- Existing models showed low correlation (14%-40%) with actual EE in children with disabilities.
- The proposed ensemble machine learning model for boys with DMD achieved a 91% correlation with measured EE.
- The novel model demonstrated a low root mean square error of 0.017.
Conclusions:
- Standard EE estimation methods for healthy adults are unsuitable for children with disabilities.
- A specialized machine learning-based nonlinear regression model significantly improves EE accuracy for children with DMD.
- This tailored approach is vital for precise physical activity assessment and intervention planning in pediatric populations with physical impairments.
