Online Detection and Extraction of FECG Signals Using ICA: A Comparative Study

Abstract

In this paper a new study to detect fetal heart rate (F H R) online from abdominal electrocardiogram (ECG) signal, which are extracted by three different algorithms of independent component analysis ICA (AMUSE, EVD2 and SOBI) is presented. Four stages for fetal electrocardiogram (FECG) extraction and detection is proposed. After preprocessing and (FECG) extraction by ICA, maternal QRS complex removal window is used to remove or scale down the maternal remaining peaks, and smoothed by II notch filter. 25 data sets are used to validate this method of study for fetal peak detection online from signals extracted by ICA. Two ways are used to test 25 signals firstly off line and secondly online. The average sensitivity of the ICA (AMUSE, EVD2 and SOBI) based method are 72.3%, 66.2% and 75.1% off line respectively, and 55%, 53% and 059% online respectively, while average positive predictivity are 61.4%, 61.3% and 69.7% off line respectively, while 43%, 41% and 46% online respectively. These show that the ICA based method is more successful in detecting the FHR off line than online, which is more complicated, where the automatic selection of the output signals is not a trivial task.