Fetal and maternal heart sounds denoising based on multi-level discrete wavelet transform

Abstract

Fetal Development and maternal health assessment during the antepartum period are extremely important for every pregnant woman to ensure a healthy pregnancy and normal childbirth. In this regard, Phonocaridiography (PCG) is a promising monitoring technology due to its distinctive ability in recording the heart sound's acoustic properties, making it better for detecting murmurs and providing additional diagnostic information regarding some Cardiovascular Diseases that cannot be found in any other monitoring method. However, PCG signals are non-stationary and have relatively small amplitudes, making them vulnerable to multiple noises while recording. This work aimed to determine the best suitable multi-level DWT parameters for denoising simultaneously recorded fetal and maternal heart sounds (fPCG and mPCG ) as efficiently as possible. Prior to that, a 4th order Butterworth Bandpass Filter was designed with a passband from 20Hz to 200Hz and 50Hz to 150Hz for fPCG and mPCG signals, respectively, in which their energy is concentrated. The performance was evaluated in terms of SNR, RMSE, and PRD performance metrics. While the fPCG Signal is denoised, the observed highest SNR value is 17.8941dB, whereas the RMSE and PRD values are 0.0065 and 12.7583%, respectively. On the other hand, while mPCG Signal is denoised, the observed values of SNR, RMSE, and PRD are 20.5330dB, 0.0049, and 9.4049%, respectively.The experiment results show that the best performance for denoising both the fetal and maternal heart sounds could be achieved when combining db10 as the mother wavelet at 6th level of decomposition and SURE as the denoiser along with Soft Thresholding Rule.