Automatic Identification of Heart Abnormalities Using PCG Signals
Résumé
A phonocardiogram (PCG) signal holds aural information generated by the heart during a cycle. A close examination of the PCG signal can reveal valuable cardiac information thereby allowing detection of abnormalities and diagnosis of heart diseases. An automation-aided analysis of PCG signals can play a vital role in the medical field, especially in remote patient monitoring, apart from being a very efficient approach. In this study, PCG signals are classified under 5 different classes based on the features extracted. The five classes are normal, mitral stenosis, mitral regurgitation, mitral valve prolapse, aortic stenosis (N, MS, MR, MVP, AS). Mel-Frequency Cepstral Coefficients (MFCCs) are extracted from the PCG audio signals and fed into a deep learning based convolutional neural network (CNN). The proposed approach achieves a maximum accuracy of 99.64% which outperforms the existing state-of-the-art approaches.
Origine | Fichiers produits par l'(les) auteur(s) |
---|