Fetal Electrocardiogram Detection And Analysis From Maternal Abdominal Ecg For Diagnosis Of Fetal Heart Rhythm Abnormalities

Fetal cardiac arrhythmias are present in about 1–3% of pregnancies and account for about 10–20% of the referrals to fetal cardiology around the world. Most fetal arrhythmias are benign; however, some can cause fetal hydrops, preterm delivery, and higher perinatal morbidity and mortality. The evolution and treatment of fetal arrhythmia depend on a timely and complete diagnosis. However, conventional methods used in clinical practice for fetal arrhythmia diagnosis are limited since they do not reflect the primary electrophysiological conduction processes in the myocardium. Fetal electrocardiography (fECG) has the potential to better support fetal arrhythmia diagnosis through the continuous analysis of the beat-to-beat variation of the fetal heart rate (FHR) and morphological analysis of the ECG waves.To date, however, the acquisition and analysis of fECG during pregnancy are considered a challenging problem for obstetricians. This is mainly due to the lack of technology to separate the fECG signal from maternal abdominal recordings while preserving its morphology. Fetal ECG extraction is currently limited to FHR estimation in clinical applications. This limitation is due to the fact that fECG from abdominal signals is mixed with the maternal electrocardiogram (mECG), and artifacts. These make it difficult to extract the fECG and to preserve its morphology.This study presents an efficient hybrid algorithm for fECG extraction from abdominal multichannel signal recordings based on independent component analysis (ICA), template subtraction, and wavelet denoising. Here, the ICA is based on the approximations of negentropy. The performance is measured with the estimation of sensitivity (SE), positive predictive value (PPV), and F1 score. QRS-peak detection accuracy is SE= 97.4%, PPV=97.2% and F1=97.29%. In addition, an ECG morphology analysis for P-wave detection based on a multiresolution analysis of the maximal overlap discrete wavelet transform is presented. The P-wave detection accuracy in signals under arrhythmic conditions is SE= 99.4%, PPV=98.5% and F1=98.94%. The main contributions of this study are a fECG extraction algorithm from non-invasive ECG recordings that preserves the morphology of the P-wave, and an algorithm that enhances and localizes the extracted signals' P-waves.