ABSTRACT
Cardiovascular disease is one of the major causes of mortality round the globe. Early detection and diagnosis are the only ways to combat it. Modern automated technical solutions are now crucial as a result of this. Artificial intelligence-based automatic diagnosis systems have proved their significance in this cause. Different image and sound modalities are being used in the diagnosis of heart diseases. This paper focuses on the prediction of heart disease using heart sounds. In this study two different architecture of deep learning based on convolutional neural with discrete wavelet transform is presented to deal with multi-labelled heart sounds dataset. First, the data are acquired from the repository and the features are extracted from the heart signal using discrete wavelet transform. These features are then fed as an input to the deep learning model for the identification of heart health. The performance of the model is validated on a phonocardiogram dataset that consists of 4 classes of heart disorder and is measured in the terms of accuracy, precision, recall and specificity. The experimental results show that the model with second architecture achieved highest accuracy with 99.5%, 100% precision, 97.36% recall and 100% specificity.
