AI model detects pediatric heart disease with 92% accuracy from stethoscope recordings

Researchers have developed a novel method for detecting pediatric congenital heart disease (CHD) using phonocardiogram (PCG) recordings from digital stethoscopes. This approach integrates deep learning features with handcrafted ones to improve diagnostic accuracy, especially in low-resource settings where traditional echocardiography is not readily available. The model achieved high performance metrics, including 92% accuracy and a 96% AUROC, on data from 751 pediatric subjects in Bangladesh. This technology holds promise as a cost-effective screening tool for real-time remote detection of CHDs. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Offers a potential low-cost, remote screening tool for congenital heart disease in underserved regions.

RANK_REASON This is a research paper detailing a new method for disease detection using machine learning.

Read on arXiv cs.CV →

• arXiv
• Bangladesh
• CHD
• PCG

• paper
• other