The comprehensive analysis of the determination of wavelet function-level pair for the decomposition and reconstruction of artificial S1 heart signals by using multi-resolution analysis. (September 2021)

• Record Type: Journal Article
• Authors: Polat, Adem
• Is Part Of: Biomedical signal processing and control. Volume 70(2021)
• Publication Date: 2021-09
• Subjects: Heart sounds -- S1 signal -- Linear chirp transient model -- Multiresolution analysis -- Wavelet transform -- Daughter wavelet functions; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.103055
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Linear chirp transient model is used to create M1&T1 components of S1 heart signal. Multiresolution analysis based on wavelet separates M1&T1 components of S1 signal. 104 daughter wavelets of bior, coif, db, sym offer the comprehensive analysis of S1. The comprehensive analysis by wavelet provides perfect reconstruction of noisy-S1. Daughter wavelet functions with a computation time of less than 1 s were identified. Abstract: Two major sounds of the normal heart sound like "lub dub". The "lub" is the first heart sound, commonly termed S1 results from mitral (M1) and tricuspid (T1) valve closure at the start of systole. In this work, the noisy-S1 heart signal was investigated to separate M1 and T1 components of it by making the comprehensive analysis of the determination of wavelet function-level pair for the decomposition and reconstruction of artificial S1 by using multiresolution analysis (MRA) and discrete wavelet transform (DWT). For this purpose, a synthetic S1 and its three different noisy-S1 signals were created by using the linear chirp transient model and then decomposed to their approximations and details at three different decomposition levels (3, 4, 5). 86 daughter wavelets of Biorthogonal, Coiflet, Daubechies, and Symlet were used to reconstruct noisy-S1 signals using comprehensive MRA&DWT. S1 and reconstructed noisy-S1 were compared qualitatively and quantitatively. For quantitative assessment, signal-to-noise-ratio (SNR), peak-SNR (PSNR), root-mean-square-error (RMSE), and structural-similarity (SSIM(%)) metrics were used for noisy-S1 signals at 3–4-5 decomposition levels. In the final evaluation of 86 daughter wavelets, db5, bior3.3, and bior3.9 performed superior results both qualitatively and quantitatively. The db5 was the superior one qualitatively at level 5, and quantitatively, the SNR values of the reconstructed signal by db5 were 8.620, 8.009, and 6.333 for %5-, %10-, and %20-noisy heart signals, respectively. The study proved that MRA&DWT provides a comprehensive analysis opportunity consisting of 86 daughter wavelets for perfect reconstruction of the S1 heart signals and detecting transients between M1 and T1 components.