Fetal ECG Extraction from Abdominal Signals Using Empirical Wavelet Transform
Ionuț Manea, Dragoş Daniel Ţarălungă
Abstract
Monitoring the fetal heart activity during pregnancy is essential in assessing its health state. An early diagnosis can prevent intrauterine deaths or malformations that cannot be treated after birth. The noninvasive fetal electrocardiogram (fECG) extracted from abdominal signals recorded from the maternal abdomen, overcomes the limitations of conventional fetal monitoring methods. However, it is difficult to be used due to multiple sources of interference (motion noise, abdominal electromyogram, electrical activity of the uterus), the most impactful factor being the maternal electrocardiogram (mECG). To overcome these limitations, the present paper aims to extract fECG and fetal heart rate (fHR) parameters in 2 steps: (1) preprocessing that combines the advantages of Empirical Mode Decomposition (EMD) and Empirical Wavelet Transform (EWT) to accurately approximate powerline interference (PLI), baseline wander (BW) and motion artifacts (MA), and (2) the processing step where the mECG is estimated in order to be removed. Hence, the fECG is isolated to determine fetal R peak locations. Using the Fractional Fourier Transform (FrFT) and the Maximum Likelihood Estimation (MLE), the mECG is precisely estimated and eliminated from the signal. To enhance the fetal QRS, a bandpass filter is applied. The algorithm is evaluated on the Physionet Challenge 2013 database.