Template subtraction based methods for non-invasive fetal electrocardiography extraction.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Maternal-Fetal Medicine

Background:

• Assessing fetal heart rate (fHR) non-invasively via fetal electrocardiogram (fECG) presents significant challenges.
• Accurate fECG signal extraction is crucial for effective fetal monitoring during labor and pregnancy.

Purpose of the Study:

• To compare the performance of five template subtraction (TS) methods for non-invasive fECG signal extraction.
• To evaluate the influence of maternal wavelet selection on fetal R-peak detection using continuous wavelet transform (CWT).
• To classify fECG signal quality for improved signal processing.

Main Methods:

• Compared five TS methods: standard TS, TS with singular value decomposition (TS[Formula: see text]), TS with linear prediction (TS[Formula: see text]), TS with scaling factor (TS[Formula: see text]), and sequential analysis (SA).
• Evaluated performance using the F1 score across Labor and Pregnancy datasets (total 88 signals).
• Assessed the impact of different maternal wavelets for CWT-based fetal R-peak detection.

Main Results:

• Sequential analysis (SA) achieved the highest F1 score (95.99%), closely followed by TS[Formula: see text] (95.93%).
• The 'gaus3' maternal wavelet was identified as optimal for fetal R-peak detection with the CWT detector.
• Signals were categorized into high, medium, and low quality, offering insights for signal extraction.

Conclusions:

• Advanced TS methods and SA demonstrate high efficacy in non-invasive fECG signal extraction.
• Optimal maternal wavelet selection is critical for accurate fetal R-peak detection.
• This research provides a foundation for enhancing clinical fetal monitoring through improved fECG signal processing.