Comparison study of population-based methods for non-invasive fetal electrocardiography extraction.

This paper presents a comparative analysis of five popular population-based algorithms in the field of non-invasive fetal electrocardiogram (NI-fECG) extraction: (1) artificial bee colony (ABC), (2) gray wolf optimization (GWO), (3) moth flame optimization (MFO), (4) particle swarm optimization (PSO), and (5) whale optimization algorithm (WOA). The five optimization algorithms are used along with sequential analysis (SA) to extract the fetal electrocardiogram (fECG) signal that is present along with other signals in the abdominal electrocardiogram. The most prominent one is the mECG signal, which also overlaps with fECG in time and frequency domains making its extraction challenging. The extraction systems were tested on two available datasets (Labor and Pregnancy) and their efficiency was evaluated using the accuracy of the R-peak detection. The algorithms are stochastic in nature; therefore, the experiments were conducted 30 times independently to observe any potential instability. The results indicate that extraction systems with GWO, MFO, PSO, and WOA demonstrated similar performance in the task, showing comparable extraction accuracy. However, the ABC-based system performed poorly and exhibited instability. It is suggested that further investigation into hybrid approaches and further modifications to the ABC algorithm could potentially enhance its extraction accuracy.