Energy Expenditure Estimation in boys with Duchene muscular dystrophy using accelerometer and heart rate sensors

Accurate Energy Expenditure (EE) Estimation is very important to monitor physical activity of healthy and disabled population. In this work, we examine the limitations of applying existing calorimetry equations and machine learning models based on sensor data collected from healthy adults to estimate EE in disabled population, particularly children with Duchene muscular dystrophy (DMD). We propose a new machine learning-based approach which provides more accurate EE estimation for boys living with DMD. Existing calorimetry equations obtain a correlation of 40% (93% relative error in linear regression) with COSMED indirect calorimeter readings, while the non-linear model derived for normal healthy adults (developed using machine learning) gave 37% correlation. The proposed model for boys with DMD give a 91% correlation with COSMED values (only 38% relative absolute error) and uses ensemble meta-classifier with Reduced Error Pruning Decision Trees methodology.