[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119239-en":3,"doc-seo-119239-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},119239,687197207639,"Asher","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","The Necessity of Multiple Data Sources for ECG-Based Machine Learning Models","Growing interest in medical machine learning highlights a persistent gap between research claims and clinical relevance, driven largely by data quality and interoperability limitations. This study analyzes how site- and study-specific differences in public 12-lead ECG datasets can destabilize trained models, even when datasets nominally match on lead definition, sampling rate, and recording duration. It compares modern network and unsupervised pattern detection performance across multiple datasets to assess generalization across single-site ECG studies.","Caring is Sharing – Exploiting the Value in Data for Health and Innovation 33  \nM. Hägglund et al. (Eds.)  \n© 2023 European Federation for Medical Informatics (EFMI) andIOS Press.  \nThis article is published online with Open Access by IOS Press and distributed under the terms of the Creative Commons Attribution Non-Commercial License 4.0 (CC BY-NC 4.0).  \ndoi:10.3233/SHTI230059  \nThe Necessity of Multiple Data Sources for ECG-Based Machine Learning Models  \nLucas PLAGWITZa,1, Tobias VOGELSANGa, Florian DOLDIb, Lucas BICKMANNa, Michael FUJARSKIa, Lars ECKARDTb, and Julian VARGHESEa  \na Institute of Medical Informatics, University of Münster, Germany b Department for Cardiology II-Electrophysiology, University Hospital Münster,  \nGermany  \nAbstract. Even though the interest in machine learning studies is growing significantly, especially in medicine, the imbalance between study results and clinical relevance is more pronounced than ever. The reasons for this include data quality and interoperability issues. Hence, we aimed at examining site-and studyspecific differences in publicly available standard electrocardiogram (ECG) datasets, which in theory should be interoperable by consistent 12-lead definition, sampling rate, and measurement duration. The focus lies upon the question of whether even slight study peculiarities can affect the stability of trained machine learning models.  \nTo this end, the performances of modern network architectures as well as unsupervised pattern detection algorithms are investigated across different datasets.  \nOverall, this is intended to examine the generalization of machine learning results  \nof single-site ECG studies.  \nKeywords. data integration, ECG, machine learning, external validation  \n1. Introduction  \nThe research trend on decision support systems via machine learning (ML) continues unabated in many disciplines. However, analysis of ML algorithms and pattern recognition for medical problems is subject to strong bias, as it consists mainly of retrospective data that are insufficient for robust clinical application and cannot adequately measure the underlying phenomenon, since they usually consider only one data source. This can be especially problematic when a data collection is not standardized, as in the case of magnetic resonance imaging, where measurements are dependent on the device and sequence [1] . A transfer of trained decision models to other datasets is therefore hardly possible. However, even with standardized data acquisition, which is given in the case of standard electrocardiograms (ECGs) with 12-channel array, 10 seconds duration, and a sampling frequency of 500 HZ, different devices and preprocessing steps potentially alter the outcome. The question arises whether this is sufficient to cause an impact on a trained model. Preliminary work already shows the broad-based data collection and interoperability problems at the level of ECG hardware,  \n1 Corresponding Author: Lucas Plagwitz, E-mail: [lucas.plagwitz@uni-muenster.de](lucas.plagwitz@uni-muenster.de).  \n34 L. Plagwitz et al. / The Necessity of Multiple Data Sources  \nFigure 1. Overview of different ECG measurements depending on the data source. Each column contains two scaled sinus rhythm recordings in lead II of the corresponding data sources (table header) .  \nsoftware, and file formats [2, 3] . Despite this, an in-depth analysis of dataset-specific signal differences is still lacking. To this end, we consider three publicly available data sources that contain quite different ECGs, as shown in Figure 1. Based on these data, we examine the transferability and vulnerability of ML algorithms between these datasets in supervised and unsupervised learning settings.  \n2. Methods  \n2.1. Datasets  \nTo investigate comparability between different ECG studies, we investigated the three largest freely available clinical 12-lead ECG datasets (10s length and sampling frequency of 500 HZ) hosted on the PhysioNet online database a","cbCaijaQDLzkEiQn","https://ap.wps.com/l/cbCaijaQDLzkEiQn","pdf",487702,1,5,"English","en",105,"# Introduction\n## Dataset bias and interoperability challenges\n# Methods\n## Datasets\n## Unsupervised analysis\n## Comparative evaluation across data sources","[{\"question\":\"Why might ECG-based machine learning results fail to generalize clinically?\",\"answer\":\"Because training often relies on retrospective data from single sources, and dataset collection can introduce quality and interoperability differences that the model cannot reliably account for.\"},{\"question\":\"How does the study test whether small dataset peculiarities affect model stability?\",\"answer\":\"By evaluating supervised and unsupervised machine learning performance across three publicly available ECG datasets and comparing transferability between them.\"},{\"question\":\"Which datasets are used in the analysis?\",\"answer\":\"The study investigates PTB-XL, a Chapman University arrhythmia database, and the Georgia dataset provided via PhysioNet.\"}]","The Necessity of Multiple Data Sources for ECG-Based Machine Learning Models | PDF",1785723250,13,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"the-necessity-of-multiple-data-sources-for-ecg-based-machine-learning-models","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/the-necessity-of-multiple-data-sources-for-ecg-based-machine-learning-models/119239/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why might ECG-based machine learning results fail to generalize clinically?","Question",{"text":75,"@type":76},"Because training often relies on retrospective data from single sources, and dataset collection can introduce quality and interoperability differences that the model cannot reliably account for.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the study test whether small dataset peculiarities affect model stability?",{"text":80,"@type":76},"By evaluating supervised and unsupervised machine learning performance across three publicly available ECG datasets and comparing transferability between them.",{"name":82,"@type":73,"acceptedAnswer":83},"Which datasets are used in the analysis?",{"text":84,"@type":76},"The study investigates PTB-XL, a Chapman University arrhythmia database, and the Georgia dataset provided via PhysioNet.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":106,"show_sort_weight":107,"slug":108},"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":46,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":21,"slug":137},19,"General","general"]