Impact of Linearization in Abdominal ECG for Non-Causal Filtering Structure in Fetal ECG Extraction

Abstract

Extraction of fetal electrocardiogram (FECG) plays a major role in monitoring fetal growth and in the diagnosis of fetal heart disorder. This research study proposes a linearization process in a Non-Causal filtering structure for extracting the fetal ECG. The linearization aims to match the baselines of abdominal and thorax ECG which improves the performance in adaptive filtering. The method initially detects the maternal R-peaks from the thorax ECG (TECG) recordings and abdominal ECG (AECG) recordings. The slope is then estimated using the amplitude of the detected peak points. Based on the estimated slopes, the abdominal ECG is then linearized which is then fed as the primary input for the non-casual adaptive filter. The non-casual filter uses both the future and past samples to extract the fetal ECG at sample index n. The metrics like, fetal to maternal signal-to-noise ratio (FmSNR), Correlation coefficient, peak root mean square difference (PRD), and R-peak detection accuracy (RPDA) were used to evaluate the algorithm performance. The datasets namely Daisy and Physionet are used for analysis. The method provides an FmSNR, correlation coefficient, PRD, and RPDA of 8.63dB, 0.9872, 81.98%, and 97.21 % respectively when evaluated on the Physionet dataset.