Back to Search Start Over

An automated diagnosis model for classifying cardiac abnormality utilizing deep neural networks.

Authors :
Singh, Gurjot
Verma, Abhinav
Gupta, Lavanya
Mehta, Anant
Arora, Vinay
Source :
Multimedia Tools & Applications; Apr2024, Vol. 83 Issue 13, p39563-39599, 37p
Publication Year :
2024

Abstract

Cardiovascular diseases remain the leading cause of global mortality, resulting in the loss of 17.9 million lives annually, as reported by the World Health Organization (WHO). This study focuses on the classification of human heart-related sounds into normal or pathological categories. The PhysioNet Computing in Cardiology (CinC) 2016 and 2022 reference datasets, also known as PhysioNet 2016 and PhysioNet 2022 respectively, have been employed to examine the technique suggested in this research work. These benchmark datasets are comprised of 3,200 and 3,168 Phonocardiogram (PCG) recordings, respectively. In the current research, Mel spectrograms hold special significance in reducing the dimensions of a raw audio signal without causing any loss of important data, thus making it more manageable for processing. The work proposes a classification system based on the UNet architecture, which processes transformed spectrograms of the PCG signals. The augmented spectrograms have yielded the best results. Specifically, on the PhysioNet 2016 dataset, the proposed model has achieved an accuracy of 96.05%, specificity of 98.82%, and F1 score as 0.91. As the focus of this study has been to develop a novel architecture for classification and not data cleaning, the model has attained an accuracy of 56.80%, specificity of 59.29%, with F1 score as 0.58 on the PhysioNet 2022 dataset which is perceived to be a noisy dataset. [ABSTRACT FROM AUTHOR]

Details

Language :
English
ISSN :
13807501
Volume :
83
Issue :
13
Database :
Complementary Index
Journal :
Multimedia Tools & Applications
Publication Type :
Academic Journal
Accession number :
176408727
Full Text :
https://doi.org/10.1007/s11042-023-16930-5