Heart sound segmentation based on morphological transformation

In cardiology, the study of heart sounds is an essential diagnostic tool. Heart sound segmentation is used to automatically analyse phonocardiogram signals(PCG). This provides relevant information for diagnosis and increases the accuracy of identification. This paper presents a method for automatic segmentation of S1 and S2 heart sounds based on the operations of multi-resolution morphological gradient, multi-scale morphological closing envelope, morphological opening and top-hat transformation. These operations first suppress murmurs and artefacts, then localise the boundaries of the heart sound components. Finally, the heart sounds are segmented into several parts based on the results of locating the extremities and identifying the first sound S1 and the second sound S2. The proposed method was validated using the Physionet and Pascal databases and the following results were obtained. The experimental results show that the proposed method achieves an average sensitivity of 99.75%, an average positive predictive value of 98.42%, an average accuracy of 98.17% and an average F-score of 99.08% for the segmentation of S1 and S2 sounds. These exceptional performances, superior to those of previous methods, suggest that this approach could become a reference tool for the automatic analysis of cardiac signals.