A medical comparative study evaluating electrocardiogram signal-based blood pressure estimation

Abstract

In general, blood pressure (BP) is measured using standard methods (medical monitors), which are widely used, or from physiological sensor data, which is a difficult task usually solved by combining several signals. In recent research, electrocardiogram (ECG) signals alone have been used to estimate blood pressure. The authors present a comparative study that evaluates ECG signal-based blood pressure estimation using complexity analysis to extract features, comparing the results obtained with a random forest regression model as well as with the combination of a stacking-based classification module and a regression module. It was determined that the best result obtained is a mean absolute error range of 3.73 mmHg with a standard deviation of 5.19 mmHg for diastolic blood pressure (DBP) and 5.92 mmHg with a standard deviation of 7.23 mmHg for systolic blood pressure (PAS).