{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,23]],"date-time":"2026-10-23T22:25:34Z","timestamp":1792794334112,"version":"4.5.0"},"reference-count":10,"publisher":"Oxford University Press (OUP)","issue":"4","license":[{"start":{"date-parts":[[2025,5,21]],"date-time":"2025-05-21T00:00:00Z","timestamp":1747785600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","award":["1R56HL161475"],"award-info":[{"award-number":["1R56HL161475"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","award":["DP2HL174046"],"award-info":[{"award-number":["DP2HL174046"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"name":"University of California Noyce Initiative","id":[{"id":"https:\/\/ror.org\/00pjdza24","id-type":"ROR","asserted-by":"crossref"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,7,22]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>In healthcare, scarcity of high-quality human-adjudicated labelled data may limit the potential of deep neural networks (DNNs). Foundation models provide an efficient starting point for deep learning that can facilitate effective DNN training with fewer labelled training examples. In this study, we leveraged cardiologist-confirmed labels from a large dataset of 1.6 million electrocardiograms (ECGs) acquired as part of routine clinical care at UCSF between 1986 and 2019 to pre-train a convolutional DNN to predict 68 common ECG diagnoses. To our knowledge, this model is one of the most comprehensive ECG DNN models to date, demonstrating high performance with a median area under the receiver operating curve (AUC) of 0.978, median sensitivity of 0.937, and median specificity of 0.923. We then demonstrate the model\u2019s utility as a foundation model by additionally training (fine-tuning) the DNN to detect three novel ECG diagnoses with relatively small datasets: carcinoid syndrome, pericardial constriction, and rheumatic doming of the mitral valve. Fine-tuning training of the foundation model achieved an AUC of 0.772 (95% CI 0.723\u20130.816) for carcinoid syndrome, 0.883 (0.863\u20130.906) for pericardial constriction, and 0.826 (95% CI 0.802\u20130.854) for rheumatic doming, compared to 0.492 (95% CI 0.434\u20130.558), 0.689 (95% CI 0.656\u20130.720), and 0.701 (95% CI 0.657\u20130.745), respectively, for DNNs trained from scratch on the same small datasets. Our results demonstrate that the ECG foundation model learned a flexible representation of ECG waveforms and can improve performance of fine-tuned downstream models, particularly in data-limited settings.<\/jats:p>","DOI":"10.1093\/ehjdh\/ztaf051","type":"journal-article","created":{"date-parts":[[2025,5,21]],"date-time":"2025-05-21T08:53:25Z","timestamp":1747817605000},"page":"619-623","source":"Crossref","is-referenced-by-count":8,"title":["A deep foundation model for electrocardiogram interpretation: enabling rare disease detection through transfer learning"],"prefix":"10.1093","volume":"6","author":[{"given":"Stephanie M","family":"Hu","sequence":"first","affiliation":[{"name":"Department of Medicine, Division of Cardiology, University of California, San Francisco , 555 Mission Bay Blvd South, San Francisco, CA 94158, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Joshua P","family":"Barrios","sequence":"additional","affiliation":[{"name":"Department of Medicine, Division of Cardiology, University of California, San Francisco , 555 Mission Bay Blvd South, San Francisco, CA 94158, USA"},{"name":"Bakar Computational Health Sciences Institute, University of California, San Francisco , 480 16th St, San Francisco, CA 94158, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0310-3326","authenticated-orcid":false,"given":"Geoffrey H","family":"Tison","sequence":"additional","affiliation":[{"name":"Department of Medicine, Division of Cardiology, University of California, San Francisco , 555 Mission Bay Blvd South, San Francisco, CA 94158, USA"},{"name":"Bakar Computational Health Sciences Institute, University of California, San Francisco , 480 16th St, San Francisco, CA 94158, USA"},{"name":"Center for Biosignal Research, University of California, San Francisco , 555 Mission Bay Blvd South, San Francisco, CA 94158, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2025,5,21]]},"reference":[{"key":"2025072210423622200_ztaf051-B1","doi-asserted-by":"crossref","first-page":"65","DOI":"10.1038\/s41591-018-0268-3","article-title":"Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network","volume":"25","author":"Hannun","year":"2019","journal-title":"Nat Med"},{"key":"2025072210423622200_ztaf051-B2","doi-asserted-by":"crossref","first-page":"1760","DOI":"10.1038\/s41467-020-15432-4","article-title":"Automatic diagnosis of the 12-lead ECG using a deep neural network","volume":"11","author":"Ribeiro","year":"2020","journal-title":"Nat Commun"},{"key":"2025072210423622200_ztaf051-B3","doi-asserted-by":"crossref","first-page":"19615","DOI":"10.1038\/s41598-022-24254-x","article-title":"Development and validation of deep learning ECG-based prediction of myocardial infarction in emergency department patients","volume":"12","author":"Gustafsson","year":"2022","journal-title":"Sci Rep"},{"key":"2025072210423622200_ztaf051-B4","doi-asserted-by":"crossref","first-page":"1285","DOI":"10.1001\/jamacardio.2021.2746","article-title":"Performance of a convolutional neural network and explainability technique for 12-lead electrocardiogram interpretation","volume":"6","author":"Hughes","year":"2021","journal-title":"JAMA Cardiol"},{"key":"2025072210423622200_ztaf051-B5","doi-asserted-by":"crossref","first-page":"19","DOI":"10.4258\/hir.2021.27.1.19","article-title":"Effectiveness of transfer learning for deep learning-based electrocardiogram analysis","volume":"27","author":"Jang","year":"2021","journal-title":"Healthc Inform Res"},{"key":"2025072210423622200_ztaf051-B6","doi-asserted-by":"crossref","first-page":"1345","DOI":"10.1109\/TKDE.2009.191","article-title":"A survey on transfer learning","volume":"22","author":"Pan","year":"2010","journal-title":"IEEE Trans Knowl Data Eng"},{"key":"2025072210423622200_ztaf051-B7","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1214\/aos\/1176344552","article-title":"Bootstrap methods: another look at the jackknife","volume":"7","author":"Efron","year":"1979","journal-title":"Ann Stat"},{"key":"2025072210423622200_ztaf051-B8","doi-asserted-by":"crossref","first-page":"62","DOI":"10.1016\/j.cvdhj.2020.08.005","article-title":"A comprehensive artificial intelligence-enabled electrocardiogram interpretation program","volume":"1","author":"Kashou","year":"2020","journal-title":"Cardiovasc Digit Health J"},{"key":"2025072210423622200_ztaf051-B9","first-page":"156","article-title":"3KG: contrastive learning of 12-lead electrocardiograms using physiologically-inspired augmentations","volume":"158","author":"Gopal","year":"2021","journal-title":"Proc Mach Learn Res"},{"key":"2025072210423622200_ztaf051-B10","doi-asserted-by":"crossref","first-page":"5251","DOI":"10.1038\/s41598-021-84374-8","article-title":"Transfer learning for ECG classification","volume":"11","author":"Weimann","year":"2021","journal-title":"Sci Rep"}],"container-title":["European Heart Journal - Digital Health"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/ehjdh\/advance-article-pdf\/doi\/10.1093\/ehjdh\/ztaf051\/63266097\/ztaf051.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/ehjdh\/article-pdf\/6\/4\/619\/63266097\/ztaf051.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/ehjdh\/article-pdf\/6\/4\/619\/63266097\/ztaf051.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,22]],"date-time":"2025-07-22T14:42:44Z","timestamp":1753195364000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/ehjdh\/article\/6\/4\/619\/8140118"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,5,21]]},"references-count":10,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2025,5,21]]},"published-print":{"date-parts":[[2025,7,22]]}},"URL":"https:\/\/doi.org\/10.1093\/ehjdh\/ztaf051","relation":{},"ISSN":["2634-3916"],"issn-type":[{"value":"2634-3916","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2025,7]]},"published":{"date-parts":[[2025,5,21]]}}}