Neural Network for Estimating Energy Expenditure in Paraplegics from Heart Rate

Abstract

The aim of the present study is to obtain models for estimating energy expenditure based on the heart rates of people with spinal cord injury without requiring individual calibration. A cohort of 20 persons with spinal cord injury performed a routine of 10 activities while their breath-by-breath oxygen consumption and heart rates were monitored. The minute-by-minute oxygen consumption collected from minute 4 to minute 7 was used as the dependent variable. A total of 7 features extracted from the heart rate signals were used as independent variables. 2 mathematical models were used to estimate the oxygen consumption using the heart rate: a multiple linear model and artificial neural networks. We determined that the artificial neural network model provided a better estimation (r = 0.88, MSE = 4.4 ml.kg(-1).min(-1)) than the multiple linear model (r = 0.78; MSE = 7.63 ml.kg(-1).min(-1)). The goodness of fit with the artificial neural network was similar to previous reported linear models involving individual calibration. In conclusion, we have validated the use of the heart rate to estimate oxygen consumption in paraplegic persons without individual calibration and, under this constraint, we have shown that the artificial neural network is the mathematical tool that provides the better estimation.