{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,10]],"date-time":"2026-09-10T06:06:55Z","timestamp":1789020415876,"version":"build-2803163510"},"reference-count":46,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,9,10]],"date-time":"2026-09-10T00:00:00Z","timestamp":1788998400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Cardiovasc. Med."],"abstract":"<jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Electrocardiography (ECG) is the most widely deployed cardiac screening modality, but demographic, acquisition and label-space heterogeneity across cohorts obstructs clinical translation. Cross-population studies rarely quantify which conditions transfer, how much target labelling is needed, or whether the adapted score stays calibrated and clinically useful.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We harmonised 12,521 recordings from PTB-XL (Germany; 7,371 records, 6,643 patients), Chapman-Ningbo (China; 4,574) and the MIT-BIH Arrhythmia Database (USA; 576 segments, 48 recordings) into 13 SNOMED-CT super-classes. A 38-dimensional interpretable feature set, computed in physical units and standardised on the source training partition alone, fed a compact multi-task backbone (30,400 parameters) trained on patient-separated partitions. It was evaluated zero-shot on both targets and adapted with 5%\u201350% of Chapman-Ningbo labels by linear probing on the frozen embedding, with partial and full fine-tuning as comparators; prevalence-aware sampling made four target classes (NORM, STD, LVH, TINV) evaluable. The composite diagnostic score was recalibrated on a dedicated split and assessed on untouched test data using bootstrap intervals, Holm-corrected comparisons and subgroup interaction tests.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      The source backbone reached macro-AUROC 0.866 (0.837\u20130.891) across 13 classes. Zero-shot transfer recovered 88% of the four-class source reference on Chapman-Ningbo (0.754, 0.727\u20130.779) and 0.642 (0.546\u20130.735, record-clustered) on MIT-BIH. Linear probing reached 0.795 with 5% of target labels and 0.829 with 50%, below the four-class source reference of 0.859; after Holm correction the gain over zero-shot was significant for NORM (+0.135), STD (+0.093) and TINV (+0.065) but not LVH (+0.004). Full fine-tuning added 0.021 at 50% labels but was inferior at 5% and less well calibrated. The recalibrated composite score attained AUROC 0.791 (0.743\u20130.836), Brier score 0.055 and expected calibration error 0.018 on Chapman-Ningbo, where 97% of positives were left-ventricular hypertrophy, and 0.834 on the seven-condition PTB-XL composite. Discrimination varied across age strata (0.708\u20130.863; interaction\n                      <jats:italic>p<\/jats:italic>\n                      \u2009=\u20090.012).\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>Cross-population transfer of ECG models is feasible with label-efficient fine-tuning of a compact, interpretable feature backbone, and the resulting composite diagnostic score remains well-calibrated in a geographically distinct population. Subgroup heterogeneity nonetheless persists, with lower discrimination in participants aged 65 years and over and a sex-dependent calibration slope, and requires further validation.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fcvm.2026.1928354","type":"journal-article","created":{"date-parts":[[2026,9,10]],"date-time":"2026-09-10T05:31:44Z","timestamp":1789018304000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Label-efficient cross-population transfer learning for electrocardiographic risk stratification: evaluation across German, Chinese and US cohorts"],"prefix":"10.3389","volume":"13","author":[{"given":"Yan","family":"Wang","sequence":"first","affiliation":[{"name":"Department of Functional Examination, The Fifth Affiliated Hospital of Anhui University of Chinese Medicine","place":["Lu\u2019an, Anhui, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yan","family":"Lu","sequence":"additional","affiliation":[{"name":"Department of Functional Examinations, Hubei Third People\u2019s Hospital","place":["Wuhan, Hubei, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,9,10]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"2982","DOI":"10.1016\/j.jacc.2020.11.010","article-title":"Global burden of cardiovascular diseases and risk factors, 1990\u20132019: update from the GBD 2019 study","volume":"76","author":"Roth","year":"2020","journal-title":"J Am Coll Cardiol"},{"key":"B2","doi-asserted-by":"publisher","first-page":"1306","DOI":"10.1161\/CIRCULATIONAHA.106.180200","article-title":"Recommendations for the standardization and interpretation of the electrocardiogram: part I","volume":"115","author":"Kligfield","year":"2007","journal-title":"Circulation"},{"key":"B3","doi-asserted-by":"publisher","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":"B4","doi-asserted-by":"publisher","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":"B5","doi-asserted-by":"publisher","first-page":"70","DOI":"10.1038\/s41591-018-0240-2","article-title":"Screening for cardiac contractile dysfunction using an artificial intelligence-enabled electrocardiogram","volume":"25","author":"Attia","year":"2019","journal-title":"Nat Med"},{"key":"B6","doi-asserted-by":"publisher","first-page":"1909","DOI":"10.1056\/NEJMoa1901183","article-title":"Large-scale assessment of a smartwatch to identify atrial fibrillation","volume":"381","author":"Perez","year":"2019","journal-title":"N Engl J Med"},{"key":"B7","doi-asserted-by":"publisher","first-page":"195","DOI":"10.1186\/s12916-019-1426-2","article-title":"Key challenges for delivering clinical impact with artificial intelligence","volume":"17","author":"Kelly","year":"2019","journal-title":"BMC Med"},{"key":"B8","doi-asserted-by":"publisher","first-page":"135","DOI":"10.1016\/j.neucom.2018.05.083","article-title":"Deep visual domain adaptation: a survey","volume":"312","author":"Wang","year":"2018","journal-title":"Neurocomputing"},{"key":"B9","volume-title":"Dataset Shift in Machine Learning","author":"Qui\u00f1onero-Candela","year":"2009"},{"key":"B10","first-page":"1180","article-title":"Unsupervised domain adaptation by backpropagation","author":"Ganin","year":"2015"},{"key":"B11","first-page":"7167","article-title":"Adversarial discriminative domain adaptation","author":"Tzeng","year":"2017"},{"key":"B12","doi-asserted-by":"publisher","first-page":"105114","DOI":"10.1016\/j.compbiomed.2021.105114","article-title":"Self-supervised representation learning from 12-lead ECG data","volume":"141","author":"Mehari","year":"2022","journal-title":"Comput Biol Med"},{"key":"B13","doi-asserted-by":"publisher","first-page":"154","DOI":"10.1038\/s41597-020-0495-6","article-title":"PTB-XL, a large publicly available electrocardiography dataset","volume":"7","author":"Wagner","year":"2020","journal-title":"Sci Data"},{"key":"B14","doi-asserted-by":"publisher","first-page":"48","DOI":"10.1038\/s41597-020-0386-x","article-title":"A 12-lead electrocardiogram database for arrhythmia research covering more than 10,000 patients","volume":"7","author":"Zheng","year":"2020","journal-title":"Sci Data"},{"key":"B15","doi-asserted-by":"publisher","first-page":"45","DOI":"10.1109\/51.932724","article-title":"The impact of the MIT-BIH arrhythmia database","volume":"20","author":"Moody","year":"2001","journal-title":"IEEE Eng Med Biol Mag"},{"key":"B16","first-page":"arXiv:1802.03426","article-title":"UMAP: uniform manifold approximation and projection for dimension reduction","author":"McInnes","year":"2018"},{"key":"B17","first-page":"723","article-title":"A kernel two-sample test","volume":"13","author":"Gretton","year":"2012","journal-title":"J Mach Learn Res"},{"key":"B18","first-page":"625","article-title":"Predicting good probabilities with supervised learning","author":"Niculescu-Mizil","year":"2005"},{"key":"B19","doi-asserted-by":"publisher","first-page":"565","DOI":"10.1177\/0272989X06295361","article-title":"Decision curve analysis: a novel method for evaluating prediction models","volume":"26","author":"Vickers","year":"2006","journal-title":"Med Decis Making"},{"key":"B20","doi-asserted-by":"publisher","first-page":"e215","DOI":"10.1161\/01.CIR.101.23.e215","article-title":"Physiobank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals","volume":"101","author":"Goldberger","year":"2000","journal-title":"Circulation"},{"key":"B21","doi-asserted-by":"publisher","first-page":"1519","DOI":"10.1109\/JBHI.2020.3022989","article-title":"Deep learning for ECG analysis: benchmarks and insights from PTB-XL","volume":"25","author":"Strodthoff","year":"2021","journal-title":"IEEE J Biomed Health Inform"},{"key":"B22","doi-asserted-by":"publisher","first-page":"1043","DOI":"10.1161\/01.CIR.93.5.1043","article-title":"Heart rate variability: standards of measurement, physiological interpretation, and clinical use","volume":"93","year":"1996","journal-title":"Circulation"},{"key":"B23","doi-asserted-by":"publisher","first-page":"230","DOI":"10.1109\/TBME.1985.325532","article-title":"A real-time QRS detection algorithm","volume":"32","author":"Pan","year":"1985","journal-title":"IEEE Trans Biomed Eng"},{"key":"B24","doi-asserted-by":"publisher","first-page":"70","DOI":"10.1109\/TAU.1967.1161901","article-title":"The use of fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms","volume":"15","author":"Welch","year":"1967","journal-title":"IEEE Trans Audio Electroacoust"},{"key":"B25","first-page":"1929","article-title":"Dropout: a simple way to prevent neural networks from overfitting","volume":"15","author":"Srivastava","year":"2014","journal-title":"J Mach Learn Res"},{"key":"B26","first-page":"448","article-title":"Batch normalization: accelerating deep network training by reducing internal covariate shift","author":"Ioffe","year":"2015"},{"key":"B27","first-page":"arXiv:1711.05101t","article-title":"Decoupled weight decay regularization","author":"Loshchilov","year":"2019"},{"key":"B28","first-page":"arXiv:1608.03983","article-title":"SGDR: stochastic gradient descent with warm restarts","author":"Loshchilov","year":"2017"},{"key":"B29","doi-asserted-by":"publisher","first-page":"161","DOI":"10.1016\/0002-8703(49)90562-1","article-title":"The ventricular complex in left ventricular hypertrophy as obtained by unipolar precordial and limb leads","volume":"37","author":"Sokolow","year":"1949","journal-title":"Am Heart J"},{"key":"B30","doi-asserted-by":"publisher","first-page":"230","DOI":"10.1186\/s12916-019-1466-7","article-title":"Calibration: the achilles heel of predictive analytics","volume":"17","author":"van Calster","year":"2019","journal-title":"BMC Med"},{"key":"B31","doi-asserted-by":"publisher","first-page":"g7594","DOI":"10.1136\/bmj.g7594","article-title":"Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD)","volume":"350","author":"Collins","year":"2015","journal-title":"Br Med J"},{"key":"B32","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1175\/1520-0493(1950)078<0001:VOFEIT>2.0.CO;2","article-title":"Verification of forecasts expressed in terms of probability","volume":"78","author":"Brier","year":"1950","journal-title":"Mon Weather Rev"},{"key":"B33","first-page":"1321","article-title":"On calibration of modern neural networks","author":"Guo","year":"2017"},{"key":"B34","doi-asserted-by":"publisher","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":"B35","first-page":"65","article-title":"A simple sequentially rejective multiple test procedure","volume":"6","author":"Holm","year":"1979","journal-title":"Scand J Stat"},{"key":"B36","doi-asserted-by":"publisher","first-page":"837","DOI":"10.2307\/2531595","article-title":"Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach","volume":"44","author":"DeLong","year":"1988","journal-title":"Biometrics"},{"key":"B37","doi-asserted-by":"publisher","first-page":"5","DOI":"10.1023\/A:1010933404324","article-title":"Random forests","volume":"45","author":"Breiman","year":"2001","journal-title":"Mach Learn"},{"key":"B38","first-page":"8024","article-title":"Pytorch: an imperative style, high-performance deep learning library","author":"Paszke","year":"2019"},{"key":"B39","first-page":"2825","article-title":"Scikit-learn: machine learning in python","volume":"12","author":"Pedregosa","year":"2011","journal-title":"J Mach Learn Res"},{"key":"B40","first-page":"arXiv:2202.10054","article-title":"Fine-tuning can distort pretrained features and underperform out-of-distribution","author":"Kumar","year":"2022"},{"key":"B41","doi-asserted-by":"publisher","first-page":"123","DOI":"10.1146\/annurev-biodatasci-092820-114757","article-title":"Ethical machine learning in healthcare","volume":"4","author":"Chen","year":"2021","journal-title":"Annu Rev Biomed Data Sci"},{"key":"B42","first-page":"2661","article-title":"Do better ImageNet models transfer better?","author":"Kornblith","year":"2019"},{"key":"B43","first-page":"5606","article-title":"CLOCS: contrastive learning of cardiac signals across space, time, and patients","author":"Kiyasseh","year":"2021"},{"key":"B44","first-page":"1","article-title":"Will two do? Varying dimensions in electrocardiography: the PhysioNet\/computing in cardiology challenge 2021","author":"Reyna","year":"2021"},{"key":"B45","doi-asserted-by":"publisher","first-page":"279","DOI":"10.1038\/s41597-023-02153-8","article-title":"PTB-XL+, a comprehensive electrocardiographic feature dataset","volume":"10","author":"Strodthoff","year":"2023","journal-title":"Sci Data"},{"key":"B46","doi-asserted-by":"publisher","first-page":"1","DOI":"10.13026\/4nqg-sb35","article-title":"MIMIC-IV-ECG: diagnostic electrocardiogram matched subset","volume":"10","author":"Gow","year":"2023","journal-title":"Sci Data"}],"container-title":["Frontiers in Cardiovascular Medicine"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fcvm.2026.1928354\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,10]],"date-time":"2026-09-10T05:31:46Z","timestamp":1789018306000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fcvm.2026.1928354\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,10]]},"references-count":46,"alternative-id":["10.3389\/fcvm.2026.1928354"],"URL":"https:\/\/doi.org\/10.3389\/fcvm.2026.1928354","relation":{},"ISSN":["2297-055X"],"issn-type":[{"value":"2297-055X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,10]]},"article-number":"1928354"}}