Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Apr 1;85(12):1302-1313.
doi: 10.1016/j.jacc.2025.01.030.

Ensemble Deep Learning Algorithm for Structural Heart Disease Screening Using Electrocardiographic Images: PRESENT SHD

Lovedeep S Dhingra  1 , Arya Aminorroaya  1 , Veer Sangha  2 , Aline F Pedroso  1 , Sumukh Vasisht Shankar  1 , Andreas Coppi  3 , Murilo Foppa  4 , Luisa C C Brant  5 , Sandhi M Barreto  6 , Antonio Luiz P Ribeiro  5 , Harlan M Krumholz  7 , Evangelos K Oikonomou  1 , Rohan Khera  8
Affiliations

Ensemble Deep Learning Algorithm for Structural Heart Disease Screening Using Electrocardiographic Images: PRESENT SHD

Lovedeep S Dhingra et al. J Am Coll Cardiol. .

Abstract

Background: Identifying structural heart diseases (SHDs) early can change the course of the disease, but their diagnosis requires cardiac imaging, which is limited in accessibility.

Objectives: The purpose of this study was to leverage images of 12-lead electrocardiograms (ECGs) for automated detection and prediction of multiple SHDs using an ensemble deep learning approach.

Methods: We developed a series of convolutional neural network models for detecting a range of individual SHDs from images of ECGs with SHDs defined by transthoracic echocardiograms performed within 30 days of the ECG at the Yale New Haven Hospital (YNHH). SHDs were defined as left ventricular ejection fraction <40%, moderate-to-severe left-sided valvular disease (aortic/mitral stenosis or regurgitation), or severe left ventricular hypertrophy (interventricular septal diameter at end-diastole >1.5 cm and diastolic dysfunction). We developed an ensemble XGBoost model, PRESENT-SHD (Practical scREening using ENsemble machine learning sTrategy for SHD detection), as a composite screen across all SHDs. We validated PRESENT-SHD at 4 U.S. hospitals and the prospective, population-based ELSA-Brasil (Brazilian Longitudinal Study of Adult Health) cohort, with concurrent protocolized ECGs and transthoracic echocardiograms. We also used PRESENT-SHD for risk stratification of new-onset SHD or heart failure (HF) in clinical cohorts and the population-based UK Biobank.

Results: The models were developed using 261,228 ECGs from 93,693 YNHH patients and evaluated on a single ECG from 11,023 individuals at YNHH (19% with SHD), 44,591 across external hospitals (20%-27% with SHD), and 3,014 in the ELSA-Brasil (3% with SHD). In the held-out test set, PRESENT-SHD demonstrated an area under the receiver-operating characteristic curve (AUROC) of 0.886 (95% CI: 0.877-894), 90% sensitivity, and 66% specificity. At hospital-based sites, PRESENT-SHD had AUROCs ranging from 0.854 to 0.900, with sensitivities and specificities of 93% to 96% and 51% to 56%, respectively. The model generalized well to ELSA-Brasil (AUROC 0.853 [95% CI: 0.811-0.897], 88% sensitivity, 62% specificity). PRESENT-SHD demonstrated consistent performance across demographic subgroups, novel ECG formats, and smartphone photographs of ECGs from monitors and printouts. A positive PRESENT-SHD screen portended a 2- to 4-fold higher risk of new-onset SHD/heart failure, independent of demographics, comorbidities, and the competing risk of death across clinical sites and UK Biobank, with high predictive discrimination.

Conclusions: We developed and validated PRESENT-SHD, an AI-ECG tool identifying a range of SHD using images of 12-lead ECGs, representing a robust, scalable, and accessible modality for automated SHD screening and risk stratification.

Keywords: artificial intelligence; cardiovascular screening; deep learning; echocardiography; electrocardiograms; predictive modeling; structural heart disease.

PubMed Disclaimer

Conflict of interest statement

Funding Support and Author Disclosures Dr Brant is supported in part by CNPq (307329/2022-4). Dr Ribeiro is supported in part by the National Council for Scientific and Technological Development - CNPq (grants 465518/2014-1, 310790/2021-2, 409604/2022-4 e 445011/2023-8). Dr Krumholz is the Editor-in-Chief of JACC; works under contract with the Centers for Medicare and Medicaid Services to support quality measurement programs; is associated with research contracts through Yale University from Janssen, Kenvue, and Pfizer; in the past 3 years has received options for Element Science and Identifeye and payments from F-Prime for advisory roles; and is a co-founder of and holds equity in Hugo Health, Refactor Health, and Ensight-AI. Dr Oikonomou was supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health (under award F32HL170592); has been a consultant for Caristo Diagnostics Ltd and Ensight-AI Inc; and has received royalty fees from technology licensed through the University of Oxford, outside the submitted work. Drs Oikonomou and Khera are cofounders of Evidence2Health, a precision health platform to improve evidence-based cardiovascular care. Dr Khera was supported by the National Institutes of Health (under awards R01AG089981, R01HL167858, and K23HL153775) and the Doris Duke Charitable Foundation (under award 2022060); is an Associate Editor of JAMA; has received support from the Blavatnik Foundation through the Blavatnik Fund for Innovation at Yale; has received research support, through Yale, from Bristol Myers Squibb, BridgeBio, and Novo Nordisk; and is a coinventor of U.S. Pending Patent Applications WO2023230345A1, US20220336048A1, 63/484,426, 63/508,315, 63/580,137, 63/606,203, 63/619,241, and 63/562,335. Dr Khera and Mr Sangha are the coinventors of U.S. Provisional Patent Application No. 63/346,610, “Articles and methods for format-independent detection of hidden cardiovascular disease from printed electrocardiographic images using deep learning”; and are cofounders of Ensight-AI. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the paper; and decision to submit the paper for publication. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

Update of

References

    1. Steinberg DH, Staubach S, Franke J, Sievert H. Defining structural heart disease in the adult patient: current scope, inherent challenges and future directions. Eur Heart J Suppl. 2010;12:E2–E9.
    1. Picano E. Economic and biological costs of cardiac imaging. Cardiovasc Ultrasound. 2005;3. - PMC - PubMed
    1. Vitola JV, Shaw LJ, Allam AH, et al. Assessing the need for nuclear cardiology and other advanced cardiac imaging modalities in the developing world. J Nucl Cardiol. 2009;16:956–961. - PubMed
    1. Alkhouli M, Alqahtani F, Holmes DR, Berzingi C. Racial disparities in the utilization and outcomes of structural heart disease interventions in the United States. J Am Heart Assoc. 2019;8. - PMC - PubMed
    1. Samad Z, Sivak JA, Phelan M, Schulte PJ, Patel U, Velazquez EJ. Prevalence and outcomes of left-sided valvular heart disease associated with Chronic kidney disease. J Am Heart Assoc. 2017;6. - PMC - PubMed