Fetal ECG Extraction and QRS Detection Using Advanced Adaptive Filtering-Based Signal Decomposition and Peak Threshold Technique from Abdominal ECG Signals

The noninvasive fetal electrocardiography monitoring system consists of composite signals such as maternal ECG (MECG) and fetal ECG (FECG) acquired from abdominal ECG (AECG) signals. These records allow for the collection of relevant and reliable information that contributes to the safety of the fetus during pregnancy. However, FQRS detection and FECG extraction are challenging due to the non-stationary complexity of FECG signals, similar frequency components, low detection rates, and overlap issues that can occur for FQRS in abdominal recordings. Therefore, improved biomedical signal processing methods are required to achieve high detection accuracy with fewer errors. This paper introduces an advanced framework for FECG signal extraction and QRS detection using a signal decomposition technique and improved threshold-based detection with an adaptive noise cancelation approach (ANC-SDITD) in AECG signals. It consists of three steps: Initially, a robust VSS-WALMS (variable step size-weighted adaptive least mean square) algorithm is used for signal denoising from AECG signals with the reserved amplitude information. The underlying FECG signals could be automatically extracted from the complexity of the AECG signals by removing the MECG signal using an empirical wavelet transform and inverse scattering entropy method. Finally, one method relies on an improved self-adaptive peak threshold detection algorithm to detect fetal QRS complexes. Experimental results show that the proposed ANC-SDITD approach is effective for FECG extraction and FQRS detection when simulated on a fetal ECG synthetic database. Performance results achieve higher statistical accuracy under extraction metrics, signal-to-noise ratio (SNR) metrics, and statistical metrics compared to other algorithms. Also, it shows significant improvement compared to other techniques when the SNR ranges from 0 to 12 dB. SNR estimates good performance for the QRS complex between the noise level and the ECG channels.