Fetal QRS Detection from Single-Channel Abdominal ECG by Adaptive Improved Clustering

Assessment of fetal development and wellness throughout pregnancy is essential in the early detection of pregnancy anomalies. Recent advances in wearable technology have led to the development of home pregnancy monitoring systems, which hold potentials for the continuous monitoring of expectant mothers and their unborn children through fe-tal/maternal electrocardiogram (f/mECG) technology. Given the complexity of extracting useful information from f/mECG data, these devices are often equipped with multiple channels and sophisticated signal processing algorithms. Thus, the implementation of f/mECG monitoring in daily life remains unfeasible. In this study, we propose an algorithm for extracting the fetal electrocardiogram (fECG) using a single channel. The algorithm detects the peaks of the maternal ECG (mECG) through a clustering method, followed by peak correction. The mECG signal is then reconstructed with Principal Component Analysis (PCA) before applying the Template Subtraction (TS) method to obtain fECG. Rigorous testing with various online f/mECG databases demonstrates the potential to accurately extract fECG using only a single channel. The evaluation results based on the online dataset achieved an accuracy of 99.60% and an F1 score of 99.30%. This advancement also opens opportunities to optimize home-based f/mECG monitoring, making it more compact and reliable.