Localization of heart sounds based on S-transform and radial basis function neural network

Abstract

This paper presents an original method for heart sounds localization based on S-Transform and radial basis function neural network (SRBF). The S-Transform is used to extract the features of heart sound. These features are then applied as inputs to RBF classifier. The performance of the localization is evaluated according to a data base of 50 subjects (including 25 cardiac pathologies sounds) which correspond to 1074 S1 and S2 heart sounds, selected from The University Hospital of Strasbourg and the Mars500 project. This study is made under the control of an experienced cardiologist. The SRBF was shown to have 95% sensitivity and 98% positive predictivity value. The proposed solution is compared with other existing methods and the robustness is shown against additive white Gaussian noise. © 2011 Springer-Verlag.