Evaluation of parametric methods for the spectral analysis of Carpentier-Edwards heart valve sounds

Abstract

The authors examine the performance of modern parametric (all-pole and pole-zero) methods in estimating the spectral characteristics of Carpentier-Edwards bioprosthetic heart valve closing sounds. The parametric methods used were autoregressive techniques based on the Burg algorithm and least-squares linear predictive methods, an autoregressive-moving average technique based on singular value decomposition, as well as Prony's method. Algorithms were tested on sounds obtained from patients with Carpentier-Edwards porcine xenografts implanted in the aortic and mitral positions. Using these techniques, the spectrum produced by Carpentier-Edwards bioprostheses has been characterised. This paper represents the first publication of this finding. The paper also considers factors such as the suitability of algorithms, implementation of these algorithms, as well as the selection of the optimal number of poles and zeros.<>