Wavelet filtering of fetal phonocardiography: A comparative analysis

Area of Science:

• Biomedical Engineering
• Signal Processing
• Maternal-Fetal Medicine

Background:

• Fetal heart rate (FHR) monitoring is crucial for identifying high-risk fetuses.
• Fetal phonocardiogram (FPCG) records fetal heart sounds but is heavily noise-contaminated.
• Effective filtering is essential for clinical usability of FPCG data.

Purpose of the Study:

• To compare the performance of various Wavelet transform (WT) based filters for FPCG data.
• To identify the optimal WT filter combination for noise reduction and accurate FHR estimation.
• To evaluate filters using simulated and experimental FPCG datasets.

Main Methods:

• Developed 18 WT-based filters by combining three mother Wavelets (Coiflet, Daubechies, Symlet) with two thresholding rules (Soft, Hard) and three algorithms (Universal, Rigorous, Minimax).
• Applied filters to 37 simulated and 119 experimental FPCG datasets from PhysioNet/PhysioBank.
• Evaluated filter performance based on FHR estimation accuracy and signal-to-noise ratio (SNR) improvement.

Main Results:

• The filter combining 4th-order Coiflet mother Wavelet, Soft thresholding rule, and Universal algorithm demonstrated optimal performance.
• This optimal filter accurately maintained FHR compared to reference values in both simulated and experimental data (P > 0.05).
• Significantly improved SNR was observed with the optimal filter, increasing from 0.7 dB to 25.9 dB in simulated data and 15.6 dB to 22.9 dB in experimental data (P < 10^-14 and P < 10^-37, respectively).

Conclusions:

• The WT-based filter using 4th-order Coiflet, Soft rule, and Universal algorithm is optimal for FPCG filtering.
• This filter effectively reduces noise while preserving critical clinical information for FHR monitoring.
• The findings provide a robust method for enhancing the clinical utility of FPCG signals.