Non Invasive Foetal Electrocardiogram Extraction Using Abdominal Electrocardiogram of the Mother

Abstract

New life entering the universe always gives joy to the environment. At the same time, it is our duty to ensure that the new-born is hale and healthy. This research work focuses on monitoring the health of the foetus. Currently, in addition with other basic tastings, foetus monitoring means the monitoring of the foetus cardiac conditions, cardiac variability and the foetal ECG (fECG) which gives a lot of required information about the foetus growth and health. Intrauterine foetal growth rate calculation, weight gain, Ultrasound and Doppler technologies are few basic non-invasive tests carried out in general. To know the foetal ECG, only way available now is invasive monitoring. This kind of invasive monitoring is very r sky for mother s as we as foeta s health Therefore, in the proposed research, non-invasive way of extraction of fECG from an abdominal ECG of the mother and other possibilities of adopting it for real-time monitoring has been explored. In the proposed research work, a non-linear transformation and maximum likelihood estimation have been used in estimating the mother ECG (mECG) from a single channel abdominal ECG (abdECG) of the mother after some pre-processing. Once the mECG has been estimated, it has been removed from the pre-processed abdECG which leaves the fECG and noise as residues. Polynomial approximation and total variation method have been used to remove the noise and improves the detection of foetal ECG, its R-peaks and other required details from the fECG. Foetal R-peaks have been identified by the same process used in detecting the maternal R-peaks with required changes since mECG and fECG differ in certain aspects like the amplitude levels, the number of heartbeats per minutes. The proposed research work has been tested with different databases for validation and achieved very good results in terms of accuracy, sensitivity, positive predictivity, F1 score, mean square errors in estimated values, signal to noise ratio etc..