Date of Award

January 2026

Document Type

Open Access Thesis

Degree Name

Master of Public Health (MPH)

Department

School of Public Health

First Advisor

Brian Wahl

Abstract

Rheumatic Heart Disease (RHD), caused by Group A Streptococcus (GAS), remains a significant global health burden, particularly among children in low- and middle-income countries. Untreated GAS infection can lead to acute rheumatic fever and subsequent chronic valvular damage of RHD, such as mitral valve (MV) regurgitation. Hence, early detection is critical for the administration of Secondary Antibiotic Prophylaxis to prevent disease progression. Although deep learning models demonstrate high diagnostic accuracy, limited explainability restricts clinical adoption, especially in resource-limited settings. In this proof-of-concept study, we evaluated Shannon energy (SE) transformation as a proxy to differentiate holosystolic murmurs associated with RHD-related mitral regurgitation from normal heart sounds. Pediatric phonocardiograms from Northeast Brazil (2014-2015) in the CirCor DigiScope dataset (version 1.0.3) were analyzed, including cardiologist-validated holosystolic MV murmurs and healthy controls. Signals were filtered, denoised, and transformed into SE envelopes, which were integrated with clinician-defined cardiac phase boundaries to derive cumulative energy (area under the curve, AUC) and energy intensity (AUC normalized by systolic duration) at the systolic and participant levels. Participant-level metrics were obtained by averaging systolic-level values. Binary classification was performed using kernel density estimation. Thirteen pediatric participants (9 controls, 4 cases) were analyzed. Wavelet denoising achieved median signal-to-noise ratios (IQR) of 8.68 dB (4.29) for cases and 13.59 dB (3.47) for controls. At the systolic level, energy intensity outperformed cumulative energy, improving accuracy (95.22% vs 86.60%), sensitivity (97.37% vs 92.11%), and specificity (93.98% vs 83.46%). At the participant level, both metrics achieved 100% performance. These preliminary findings support SE-derived systolic energy as an interpretable, low-cost, and scalable approach for RHD screening and triage in resource-limited settings, with further validation in larger studies warranted.

Comments

This is an Open Access Thesis.

Download
Open Access

This Article is Open Access

Share

COinS