Intermittent blood pressure prediction via multiscale entropy and ensemble artificial neural networks

Abstract

This study evaluates the correlation between the intermittent blood pressure (BP) and the photoplethysmography (PPG). This study of a total of twenty-five cases is started by the partitioning of the PPG signal into a 5-minute segment. The segmented PPG is filtered by ensemble empirical mode decomposition (EEMD). The feature extraction method, multiscale entropy (MSE) is utilized for the purified signal to achieve some information. The seventy-five scale of MSE is taken into the input of the artificial neural network (ANN) modeling. The output of this system are the intermittent diastolic and systolic blood pressure. Originally, a thousand models areis created. The best model is chosen for the best single ANN model. In advanced, the ensemble artificial neural network (EANN) model is initiated to observe the testing data. The result, compared to the best single ANN model, shows that the EANN model recognizes better the testing data by producing lower mean absolute error (MAE).