Fetal heart abnormality detection based on subspace separation and Wiener filtering

Early identification of possible abnormalities in fetal heart signal allows clinicians to take necessary actions to cure them with ample time in hands. This research focused on developing a system that extracts an accurate fetal heart signal from a recording of Electrocardiograms obtained from the thorax and the abdomen of pregnant mothers. The main interferences that cause difficulty in the extraction of fetal heart signal are the maternal chest signal and various other noisy signals. Initially, the removal of these various noisy signals was achieved through the Eigen analysis based subspace separation technique. However, the removal of maternal chest signal cannot be achieved through existing non-adaptive filtering techniques. Therefore, an adaptive signal processing technique based on Wiener filter was employed to overcome this issue. Further, by studying the possible abnormalities that could occur in the ST interval of the Electrocardiogram of a fetus, a synthetic signal having abnormalities was generated. Then, the system was evaluated whether it is possible to extract signals having such abnormalities. This was used to validate our system on its abnormality detection capability. Since this research is a preliminary study on abnormality detection of fetal heart signals, it is expected that the extension of this work would give a better outcome in detecting various abnormality conditions caused due to compromised cardiac performances of the fetus.