A modular approach to computer-aided auscultation: Analysis and parametric characterization of murmur acoustic qualities

In the present work, a modularized approach to computer-aided auscultation based on the traditional cardiac auscultation of murmur is proposed. Under such an approach, the present paper concerns the task of evaluating murmur acoustic quality character. The murmurs were analyzed in their time-series representation, frequency representation as well as time-frequency representation, allowing extraction of interpretable features based on their signal structural and spectral characters. The features were evaluated using scatter plots, receiver operating characteristic curves (ROC), and numerical experiments using a KNN classifier. The possible physiological and hemodynamical associations with the feature set are made. The implication and advantage of the modular approach are discussed.