Citation

Shatiswaran, Vigian and Kanesraj, Ramasamy and Abdullah, Junaidi and Hamzah, Faizal Amri and Thanjappan, Sivasutha (2026) Deep Learning–Based Analysis of Phonocardiogram Signals for Heart Murmur Detection Using Dense-Net. In: 3rd IEEE International Conference on Trends in Quantum Computing and Emerging Business Technologies, TQCEBT 2026, 9 April 2026 - 11 April 2026, Pune.

Text 7.pdf - Published Version Restricted to Repository staff only Download (362kB)

Abstract

Traditionally, cardiac auscultation which calls for specific expertise and experience is widely used to diagnose heart murmurs. Using phonocardiograms (PCGs) acquired at various auscultation positions, the study aims to detect the presence or absence of heart murmurs and predict the clinical outcomes of patients (normal or abnormal). In this paper, a semi-supervised model specifically designed for PCG classification is presented with the aim of enhancing performance through the use of timefrequency deep features. Using the Stockwell transform to translate the PCG signal into two-dimensional time-frequency mappings (TFMs), the study starts by examining how PCGs behave in the time-frequency domain. These TFMs are then converted into deep feature sets using a deep network called Dense-net where redundancy is removed during feature reduction, and there are fewer deep features overall signals phonocardiograms (PCGs) acquired at various auscultation positions on the study to gain the best accuracy. Feature reduction streamlines the feature set by removing redundancies and reducing the number of deep features. Three criteria are used to x`x how effective the extracted features are. various classifiers with the dataset from the ICBHI dataset. The suggested method obtained an average sensitivity of 91%, accuracy of 93% and F1- score of 91% for Task I, which is concerned with the detection of cardiac murmurs. The methodology demonstrated a clinical outcome cost of 5290 heart sounds, surpassing the standard set by the challenge's top methods, according to Task II of the CinC/Physionet challenge 2022 [10],[11].

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Acoustic signal analysis, cardiomyopathy
Subjects:	R Medicine > R Medicine (General) > R858-859.7 Computer applications to medicine. Medical informatics
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	30 Oct 2026 02:41
Last Modified:	30 Oct 2026 02:41
URII:	http://shdl.mmu.edu.my/id/eprint/17028

Downloads

Edit (login required)