Development of Software for Generating Synthetic Fetal Electrocardiograms

Abstract

Some of the most common and fatal birth defects are those related to the heart. In adults, abnormal heart conditions are often identified through the use of an electrocardiogram (ECG). However, due to the presence of other signals and noise in the recording, fetal electrocardiography has not yet proven effective in diagnosing these defects. This paper develops a mathematical model of three-dimensional heart vector trajectories, which we use to generate synthetic maternal and fetal ECG signals. This dipole vector model simulates the electrical activity of the heart as a single time-varying vector originating at the center of the body. We use a system of ordinary differential equations whose numerical solution approximates the cardiac dipole vector. This system of equations is modified to account for noise, heart rate variability, and other physical and physiological considerations. Using this model, we have built a database of realistic, synthetic fECG signals using a variety of parameter values and noise levels. Currently, algorithms to extract the fetal signal from the fECG are tested on databases of clinical recordings. Our synthetic database allows further algorithm effectiveness testing to be performed on a broader set of data with knowledge of what the algorithm should output. We have also built a publicly accessible, interactive user interface for our model.