Fetal QRS Detection Based on Convolutional Neural Networks in Noninvasive Fetal Electrocardiogram

Abstract

Detection of fetal QRS complexes in a noninvasive fetal electrocardiogram (NI-FECG) signal is an important task to check fetal conditions and to prevent birth defects. However, the detection is not easy because the NI-FECG signal contains a maternal ECG signal that has greater amplitude than that of a fetal ECG signal. This paper proposes an algorithm to detect the fetal QRS complexes in the NI-FECG signal. The proposed algorithm is based on convolutional neural networks (CNN) and can reliably detect the fetal QRS complexes without separating the maternal ECG signal. To verify the algorithm, NI-FECG data (PhysioNet/computing in cardiology challenge 2013) were used. The proposed algorithm showed the average sensitivity of 89.06 % and positive predictive value of 92.77 %. The proposed algorithm can help to check fetal conditions and to prevent birth defects.