A study of Cardiac reference values and environmental factors using big data and deep learning methods

Abstract

AbstractAimsThe purpose of this study was to simulate the Left Ventricular Ejection Fraction (LVEF) reference values with heart rate and environmental data using LSTM deep learning methods and to derive the spatial distribution of LVEF reference values in China.MethodsHeart rate and environmental factors were used as independent variables, and LVEF indicator values were used as the dependent variable. After the sample data were randomly sampled, the block acquisition data were converted into sample sequence data. Once the sequence data were processed, the sample data were input into the long- and short-term memory networks for deep learning parameter training. After repeated weighting parameter adjustment training and optimization, the optimal prediction model for adult LVEF reference values were derived.ResultsThe LVEF reference values of normal adults showed a downward trend from moving from the south to north in China. LVEF is negatively correlated with heart rate and annual air temperature range; while positively correlated with annual mean air temperature, annual mean relative humidity, annual precipitation.ConclusionsThe LVEF reference values are related to heart rate and environmental factors. According to the predicted LSTM model, if the environmental factors of a region is known, combined with the heart rate data from big data, the model can be used to obtain a more accurate LVEF reference value prediction model. A predicted LSTM model is more capable of taking geographical and individual differences into account.