Prediction method of oxygen uptake during exercise based on oxygen uptake efficiency slope theory

Measurement of oxygen uptake is the primary method to evaluate the energy expenditure of the human during exercise, typically acquired through cardiopulmonary exercise test (CPET). However, CPET is time-consuming and inefficient, and can not quickly accumulate a large number of energy expenditure evaluation data. In this study, we propose a prediction method of oxygen uptake, the data of CPET within the first 2.5 minutes can be used to predict oxygen uptake in the next 7.5 minute. Firstly, the data of CPET are augmented by by resampling and introducing noise. Then, a novel OEAL (OUES embedding augmented LSTM) network is established based on oxygen uptake efficiency slope (OUES) theory. This network extracts a one-dimensional linear feature, representing the logarithm of ventilation per minute, which is fused with the hidden layer of LSTM network. The mean absolute percentage error (MAPE) between the predicted results and source data is 3.37%, proving that this method can effectively predict oxygen uptake and significantly shorten the time of CPET.