Related Experiment Video
Updated: Jul 18, 2026

Fetal Mouse Cardiovascular Imaging Using a High-frequency Ultrasound (30/45MHZ) System
Published on: May 5, 2018
Accuracy of deep learning-based heart sound analysis as screening tool for congenital heart disease: A diagnostic
Sebastian Emmanuel Willyanto1, Imke Maria Del Rosario Puling1, Nyoman Deva Pramana Giri1
1Faculty of Medicine, Clinical Clerkship Program, Universitas Brawijaya - Saiful Anwar Regional General Hospital, Malang, Indonesia.
Background:
With over 260,000 yearly mortality, 87% from low socio-economic countries, congenital heart disease (CHD) stands as a major global health crisis. Echocardiography, the current diagnostic standardized approach, often delays diagnosis due to interpretation variability and limited availability in low socioeconomic countries. Hence, exploring alternative screening approaches is crucial. The deep learning (DL) method coupled with heart sound analysis is a promising diagnostic approach for early CHD detection.
Objectives:
This study focuses on the innovation of DL-based heart sound analysis in enhancing the diagnosis of CHD patients.
Methods:
The outcomes of interest of the study were the sensitivity, specificity, and area under the curve (AUC) of various diagnostic tools and DL methods used in diagnosing CHD. Quality appraisal was done using Quality Assessment of Diagnostic Accuracy Studies 2, while the diagnostic meta-analysis used Stata MP 17.
Results:
A comprehensive search across seven databases yielded 19 articles for analysis. Among these, 13 were deemed low-risk and six unclear-risk of bias. The results were observed through meta-analysis with sensitivity at 91% (95% confidence interval [CI] = 0.85-0.94, P = 0.001), specificity at 92% (95% CI = 0.86-0.95, P = 0.001), and an AUC of 96% (95% CI = 0.94-0.98).
Conclusions:
The adoption of DL-based heart sound analysis for CHD diagnosis shows promise, demonstrating good diagnostic accuracy and potential as an adjunctive tool to support triage or pre-screening, although further validation is required before it can be considered alongside echocardiography.
