In this paper, abnormalities in heart valves are identified using two en-tropy methods are explored, the Time-Dependent Entropy and the Spectral Entropy, calculated in the time domain and frequency domain, respectively. The entropies together with the Probability Distribution were calculated to a database that contains simultaneous recordings from the four main auscultation areas to test if the use of these channels increases de probability of detecting the abnormality in any of the heart valves and, to compare the results per area respect to signals randomly selected from the four areas. The three parameters obtained from 20 randomly selected signals of the database were used as input features for the K-Nearest Neighbour classifier, obtaining accuracies of 90% and 80% for pathologic and normal sounds classification, respectively. Finally, the parameters calculated from all the database were separat-ed and presented in each of the four auscultation areas in 3D-distributions where a visible separability is shown. Results suggest that some noise associated to valve disfunction is reflected in the entropy values. In addition, results show that information in each area is different and the four areas analysis might improve the correct classification when there is a pathology.