Research & Publications

RESEARCH & INNOVATION

Building evidence for more accessible heart screening

Every child deserves a healthy heart. One Heart Health works with clinical and research partners to explore how digital heart sounds, remote specialist review, and machine learning can support earlier identification of congenital heart disease (CHD) in communities with limited access to pediatric cardiac care.

Explore the published research informing this approach. Results below describe individual studies, populations, and methods; they should not be read as a single measure of the current One Heart App’s performance.

Published papers  |  Collaborate with us

Published papers

Maiby’s Algorithm: A Two-Stage Deep Learning Approach for Murmur Detection in Mel Spectrograms for Automatic Auscultation of Congenital Heart Disease

Araujo M, Zeng D, Palotti J, Xi X, Shi Y, Pyles L, Ni Q.
Computing in Cardiology, 2022;49. Conference proceedings paper.
DOI: 10.22489/CinC.2022.249

This study developed a two-stage deep learning approach to identify murmurs from digital heart sounds for the 2022 George B. Moody PhysioNet Challenge. On the official hidden test set, the model achieved a weighted accuracy score of 0.729 for murmur detection, ranking 17th of 40 teams. This challenge-specific score is distinct from ordinary diagnostic accuracy and does not establish performance in routine clinical practice.

View publication

Initial Field Test of a Cloud-Based Cardiac Auscultation System to Determine Murmur Etiology in Rural China

Pyles L, Hemmati P, Pan J, et al.
Pediatric Cardiology, 2017;38:656–662. Journal article.
DOI: 10.1007/s00246-016-1563-8

This foundational proof-of-concept study evaluated remote review by a pediatric cardiologist using the HeartLink system. Among 7,993 children screened in rural China, 149 with detected murmurs received digital recordings and echocardiography. Remote review identified 11 of 14 children with echo-confirmed CHD; reported sensitivity was 78.5% and specificity 92.6%. The selected murmur cohort and study limitations matter when interpreting these results. This was specialist tele-auscultation, rather than an evaluation of the current AI model.

Read abstract on PubMed  |  View journal article

From research to responsible screening

Remote auscultation and automated analysis are being studied as tools to support clinical assessment and referral. Screening findings are not confirmed diagnoses. These studies do not establish that automated analysis can replace specialist evaluation or echocardiography.

Help advance the research

Interested in clinical research, digital auscultation, or expanding access to heart screening? Contact One Heart Health to discuss collaboration.

Explore our approach  |  Support our work