Electrocardiogram screening for aortic valve stenosis using artificial intelligence
Abstract
Aims: Early detection of aortic stenosis (AS) is becoming increasingly important with a better outcome after aortic valve replacement in asymptomatic severe AS patients and a poor outcome in moderate AS. We aimed to develop artificial intelligence-enabled electrocardiogram (AI-ECG) using a convolutional neural network to identify patients with moderate to severe AS. Methods and results: Between 1989 and 2019, 258 607 adults [mean age 63 ± 16.3 years; women 122 790 (48%)] with an echocardiography and an ECG performed within 180 days were identified from the Mayo Clinic database. Moderate to severe AS by echocardiography was present in 9723 (3.7%) patients. Artificial intelligence training was performed in 129 788 (50%), validation in 25 893 (10%), and testing in 102 926 (40%) randomly selected subjects. In the test group, the AI-ECG labelled 3833 (3.7%) patients as positive with the area under the curve (AUC) of 0.85. The sensitivity, specificity, and accuracy were 78%, 74%, and 74%, respectively. The sensitivity increased and the specificity decreased as age increased. Women had lower sensitivity but higher specificity compared with men at any age groups. The model performance increased when age and sex were added to the model (AUC 0.87), which further increased to 0.90 in patients without hypertension. Patients with false-positive AI-ECGs had twice the risk for developing moderate or severe AS in 15 years compared with true negative AI-ECGs (hazard ratio 2.18, 95% confidence interval 1.90–2.50). Conclusion: An AI-ECG can identify patients with moderate or severe AS and may serve as a powerful screening tool for AS in the community.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Valvular and composite SHD AI-ECG models are promising for referral enrichment.
“Valvular and composite SHD models are promising for referral enrichment”
From AI-ECG for Echocardiography Triage in Structural Heart Disease: Evidence, Implementation, and Future Directions (2026) · cites this paper as [19]
At present, AI-ECG is better interpreted as a tool for risk enrichment and follow-up prioritization in selected screening settings than as a stand-alone mandate for structured surveillance or immediate pathway escalation.
“At present, AI-ECG is better interpreted as a tool for risk enrichment and follow-up prioritization in selected screening settings than as a stand-alone mandate for structured surveillance or immediate pathway escalation.12,19,44,45”
From AI-ECG for Echocardiography Triage in Structural Heart Disease: Evidence, Implementation, and Future Directions (2026) · cites this paper as [19]
In aortic stenosis, patients with positive AI-ECG results but no moderate-to-severe AS at baseline had a higher long-term risk of subsequently developing moderate-to-severe AS, and recent community validation showed a fourfold higher risk of future AS hospitalization among false-positive screens.
“Analogous findings have been reported in aortic stenosis: patients with positive AI-ECG results but no moderate-to-severe AS at baseline had a higher long-term risk of subsequently developing moderate-to-severe AS, and recent community validation likewise showed a fourfold higher risk of future AS hospitalization among false-positive screens.19,23”
From AI-ECG for Echocardiography Triage in Structural Heart Disease: Evidence, Implementation, and Future Directions (2026) · cites this paper as [19]
In the Mayo Clinic study, a convolutional neural network detected moderate-to-severe AS with an AUC of 0.85, and patients with positive AI-ECG results but no moderate-to-severe AS at baseline had a 2.18-fold higher long-term risk of subsequently developing clinically significant AS.
“In the Mayo Clinic study, a convolutional neural network detected moderate-to-severe AS with an AUC of 0.85; importantly, patients with positive AI-ECG results but no moderate-to-severe AS at baseline had a 2.18-fold higher long-term risk of subsequently developing clinically significant AS.19”
From AI-ECG for Echocardiography Triage in Structural Heart Disease: Evidence, Implementation, and Future Directions (2026) · cites this paper as [19]
Valvular AI-ECG has moved beyond proof of concept and shows meaningful referral enrichment, but prospective pathway validation remains limited.
“From a deployment perspective, however, the practical conclusion is narrower: valvular AI-ECG has moved beyond proof of concept and shows meaningful referral enrichment, but prospective pathway validation remains limited.19,23”
From AI-ECG for Echocardiography Triage in Structural Heart Disease: Evidence, Implementation, and Future Directions (2026) · cites this paper as [19]
Patients with apparent false-positive AI-ECG results for aortic stenosis may still represent a higher-risk group, as some later developed clinically significant aortic stenosis.
“As with the ejection-fraction models, patients with apparent false-positive AI-ECG results here may still represent a higher-risk group, as some later developed clinically significant aortic stenosis [30].”
From Deep Learning Applied to 12-Lead ECGs for Detection of Structural, Metabolic and Systemic Disease (2026) · cites this paper as [30]
AI-ECG screening for moderate-to-severe aortic stenosis in a large Mayo Clinic cohort had a sensitivity of 78% and specificity of 74%.
“Cohen-Shelly et al. later evaluated AI-ECG screening for moderate-to-severe aortic stenosis in a large Mayo Clinic cohort and reported a sensitivity of 78% and specificity of 74%.”
From Deep Learning Applied to 12-Lead ECGs for Detection of Structural, Metabolic and Systemic Disease (2026) · cites this paper as [30]
Deep learning has been shown to be beneficial in detecting specific heart diseases from the 12-lead ECG, including aortic stenosis, low LVEF, low left ventricular hypertrophy and composites of valvular diseases.
“The application of deep learning, a subset of artificial intelligence (AI), has been shown to be beneficial in the detection of specific heart diseases from the 12-lead electrocardiogram (ECG). This includes conditions such as aortic stenosis, low left ventricular ejection fraction (LVEF) and low left ventricular hypertrophy, as well as composites of many valvular diseases3,17–21.”
From Detecting structural heart disease from electrocardiograms using AI (2025) · cites this paper as [17]
Previously, applications of artificial intelligence for electrocardiograms have shown potential to detect signatures of structural heart diseases.
“Previously applications of artificial intelligence for electrocardiograms (AI-ECG) have shown potential to detect signatures of SHDs.10-18”
From Ensemble Deep Learning Algorithm for Structural Heart Disease Screening Using Electrocardiographic Images (2025) · cites this paper as [12]
A key challenge of AI-ECG models in detecting specific cardiac disorders using ECGs is low precision driven by the low prevalence of individual disorders.
“A key challenge of AI-ECG models in detecting specific cardiac disorders using ECGs is the low precision driven by the low prevalence of individual disorders.10-12”
From Ensemble Deep Learning Algorithm for Structural Heart Disease Screening Using Electrocardiographic Images (2025) · cites this paper as [12]
The low prevalence of individual structural heart diseases limits the potential implementation of models for broad screening because of the low PPVs of the proposed models.
“While these models provide a strong foundation for the role of ECG-based detection of SHDs, the low prevalence of these individual diseases, their potential implementation for broad screening is limited by the low PPVs of the proposed models.10-12,16,32-34”
From Ensemble Deep Learning Algorithm for Structural Heart Disease Screening Using Electrocardiographic Images (2025) · cites this paper as [12]
Cites
DeLong, E. R., DeLong, D. M., & Clarke-Pearson, D. L. (1988) · Biometrics · doi:10.2307/2531595 · Cited by 22,649 · see citing works
Kuhn, M., & Johnson, K. (2013) · doi:10.1007/978-1-4614-6849-3 · Cited by 6,788 · see citing works
Shin, H.-C., Roth, H. R., Gao, M., Lu, L., Xu, Z., Nogues, I., Yao, J., Mollura, D., & Summers, R. M. (2016) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2016.2528162 · Cited by 5,759 · see citing works
Bergmark, B. A., O’Donoghue, M. L., Murphy, S. A., Kuder, J. F., Ezhov, M. V., Ceška, R., Gouni-Berthold, I., Jensen, H. K., Tokgozoglu, S. L., Mach, F., et al. (2020) · JAMA Cardiology · doi:10.1001/jamacardio.2020.0728 · Cited by 137 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
Armoundas, A. A., Narayan, S. M., Arnett, D. K., Spector-Bagdady, K., Bennett, D. A., Celi, L. A., Friedman, P. A., Gollob, M. H., Hall, J. L., Kwitek, A. E., et al. (2024) · Circulation · doi:10.1161/cir.0000000000001201 · Cited by 347 · see citing works
Self-citationDetecting structural heart disease from electrocardiograms using AI
Poterucha, T. J., Jing, L., Ricart, R. P., Adjei-Mosi, M., Finer, J., Hartzel, D., Kelsey, C., Long, A., Rocha, D., Ruhl, J. A., et al. (2025) · Nature · doi:10.1038/s41586-025-09227-0 · Cited by 116 · see citing works
Chung, C. T., Lee, S., King, E., Liu, T., Armoundas, A. A., Bazoukis, G., & Tse, G. (2022) · International Journal of Arrhythmia · doi:10.1186/s42444-022-00075-x · Cited by 79 · see citing works
Hanneman, K., Playford, D., Dey, D., van Assen, M., Mastrodicasa, D., Cook, T. S., Gichoya, J. W., Williamson, E. E., Rubin, G. D., & on behalf of the American Heart Association Council on Cardiovascular Radiology and Intervention; and Council on Lifelong Congenital Heart Disease and Heart Health in the Young (2024) · Circulation · doi:10.1161/cir.0000000000001202 · Cited by 72 · see citing works
Adedinsewo, D. A., Morales-Lara, A. C., Afolabi, B. B., Kushimo, O. A., Mbakwem, A. C., Ibiyemi, K. F., Ogunmodede, J. A., Raji, H. O., Ringim, S. H., Habib, A. A., et al. (2024) · Nature Medicine · doi:10.1038/s41591-024-03243-9 · Cited by 63 · see citing works
Self-citationDhingra, L. S., Aminorroaya, A., Sangha, V., Pedroso, A. F., Shankar, S. V., Coppi, A., Foppa, M., Brant, L. C. C., Barreto, S. M., Ribeiro, A. L. P., et al. (2025) · Journal of the American College of Cardiology · doi:10.1016/j.jacc.2025.01.030 · Cited by 53 · see citing works
Harmon, D. M., Mangold, K., Baez Suarez, A., Scott, C. G., Murphree, D. H., Malik, A., Attia, Z. I., Lopez-Jimenez, F., Friedman, P. A., Dispenzieri, A., et al. (2023) · JACC: Advances · doi:10.1016/j.jacadv.2023.100612 · Cited by 47 · see citing works
Self-citationLiang, Y., Sau, A., Zeidaabadi, B., Barker, J., Patlatzoglou, K., Pastika, L., Sieliwonczyk, E., Whinnett, Z., Peters, N. S., Yu, Z., et al. (2025) · European Heart Journal · doi:10.1093/eurheartj/ehaf448 · Cited by 26 · see citing works
Mitigating bias in clinical machine learning models
Perez-Downes, J. C., Tseng, A. S., McConn, K. A., Elattar, S. M., Sokumbi, O., Sebro, R. A., Allyse, M. A., Dangott, B. J., Carter, R. E., & Adedinsewo, D. (2024) · Current Treatment Options in Cardiovascular Medicine · doi:10.1007/s11936-023-01032-0 · Cited by 23 · see citing works
Self-citationAlexandrino, F. B., Schlesinger, R., Bird, J., Lee, E., Deshmukh, A. J., Nkomo, V. T., Oh, J. K., Noseworthy, P. A., Pellikka, P. A., Attia, Z., et al. (2026) · American Heart Journal · doi:10.1016/j.ahj.2025.107337 · Cited by 5 · see citing works
Self-citationSegar, M. W., Lambeth, K. D., Postalian, A., Coulter, S., Gilge, J., Razvi, N., Saeed, M., Paisley, R. D., Pandey, A., & Razavi, M. (2026) · European Heart Journal - Digital Health · doi:10.1093/ehjdh/ztag018 · Cited by 2 · see citing works
Tang, Q., Deng, K., Cui, Y., Liu, H., & Qian, B. (2026) · International Journal of General Medicine · doi:10.2147/ijgm.s610298
“19,23 It is therefore best regarded as a translationally promising referral tool rather than a deployable pre-echocardiographic standard.”
InfluentialDeep Learning Applied to 12-Lead ECGs for Detection of Structural, Metabolic and Systemic Disease
Shaulian, S., Zeltser, R., & Makaryus, A. N. (2026) · Diagnostics · doi:10.3390/diagnostics16172834
- Authors
- Cohen-Shelly, M., Attia, Z. I., Friedman, P. A., Ito, S., Essayagh, B. A., Ko, W.-Y., Murphree, D. H., Michelena, H. I., Enriquez-Sarano, M., Carter, R. E., Johnson, P. W., Noseworthy, P. A., Lopez-Jimenez, F., & Oh, J. K.
- Affiliations
- Department of Cardiovascular Medicine, Mayo Clinic , 200 First St SW, Rochester, MN 55905, USA
- Mayo Clinic (US)
- Corresponding author
- Oh, Jae K
- ORCID
- Venue
- European Heart Journal
- Publisher
- Oxford University Press (OUP)
- Published
- 2021-03-22
- Volume
- 42
- Issue
- 30
- Pages
- 2885-2896
- ISSN
- 0195-668X (ISSN-L)
- 1522-9645
- OpenAlex
- W3136799519
- Cited by
- 252
- References
- 4
- Citations by year
- 2026: 45 · 2025: 66 · 2024: 57 · 2023: 41 · 2022: 29 · 2021: 14
- FWCI
- 23.63
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model (vor)
- Locations
- European Heart Journal (publishedVersion)
- PubMed (publishedVersion)
- Topics
- Health Sciences › Medicine › Cardiology and Cardiovascular Medicine › Cardiac Valve Diseases and Treatments
- Health Sciences › Medicine › Cardiology and Cardiovascular Medicine › ECG Monitoring and Analysis
- Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
- Keywords
- aortic stenosis, electrocardiogram screening, artificial intelligence, convolutional neural networks, early detection, echocardiography, sex differences, age effects, hypertension, false positives, community-based screening
- MeSH terms
- Adult, Aged, Aortic Valve, Aortic Valve Stenosis, Artificial Intelligence, Electrocardiography, Female, Humans, Male, Mass Screening, Middle Aged, Retrospective Studies, Neural Networks, Computer
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
