Clinically Aligned Explainable AI Improves Myocardial Infarction Detection Using Multi-Lead Electrocardiography

Clinically Aligned Explainable AI Improves Myocardial Infarction Detection Using Multi-Lead Electrocardiography | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Clinically Aligned Explainable AI Improves Myocardial Infarction Detection Using Multi-Lead Electrocardiography AmirSajjad Taleban, Rodney Sparapani, Sharone Zlochiver, Qiang Lu, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8715003/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Myocardial infarction (MI), resulting from coronary artery occlusion and myocardial necrosis, is a leading cause of global cardiovascular mortality, requiring rapid diagnosis and localization to guide timely revascularization and optimize outcomes. While artificial intelligence offers a promising avenue to expedite triage and facilitate expert-level interpretation, its clinical adoption remains limited by the ``black box'' nature of many architectures. This opacity undermines clinician trust and impedes integration into clinical workflows. To address this critical gap, we introduce a clinically interpretable deep learning framework combining diagnostic accuracy with transparent decision-making. Our specialized U-Net encoder-decoder architecture processes the 12-lead electrocardiogram as a unified, spatially structured input to preserve inter-lead dependencies essential for detecting coronary territory-specific patterns. Trained on the PTB-XL dataset of 21,799 annotated ECGs, the model performs binary MI detection and classification of major anatomical subtypes: anterior and inferior MI. To enhance clinical trust and explainability, we employed Gradient-weighted Class Activation Mapping, which explains the model's reasoning to confirm that learned features align with established pathophysiological signs. Comparative analysis reveals that the multi-lead model achieves superior discrimination with AUROC 0.97 to 0.99 compared to single-lead variants, with positive and negative predictive values near or exceeding 90%, critical for effective triage and intervention prioritization. Grad-CAM effectively localizes diagnostically relevant regions, aligning with clinical ECG criteria for anterior MI in precordial leads and inferior MI in limb leads, offering valuable visual cues that support clinical decision-making across diverse practice settings. This interpretable framework enhances trust and facilitates clinician-AI collaboration, paving the way for improved MI management through robust decision support and future prospective validation. Myocardial infarction Electrocardiography Explainable artificial intelligence Grad-CAM U-Net PTB-XL Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8715003","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":587533295,"identity":"ac9dd377-8b87-4152-8e36-69dbd6863bc7","order_by":0,"name":"AmirSajjad Taleban","email":"","orcid":"","institution":"University of Wisconsin–Milwaukee","correspondingAuthor":false,"prefix":"","firstName":"AmirSajjad","middleName":"","lastName":"Taleban","suffix":""},{"id":587533297,"identity":"8d89edd4-fbbd-41ea-b823-269eef58a7af","order_by":1,"name":"Rodney Sparapani","email":"","orcid":"","institution":"Medical College of Wisconsin","correspondingAuthor":false,"prefix":"","firstName":"Rodney","middleName":"","lastName":"Sparapani","suffix":""},{"id":587533298,"identity":"f05b4950-0856-432d-9ff7-e7373e3e8a26","order_by":2,"name":"Sharone Zlochiver","email":"","orcid":"","institution":"Baxter (United States)","correspondingAuthor":false,"prefix":"","firstName":"Sharone","middleName":"","lastName":"Zlochiver","suffix":""},{"id":587533300,"identity":"03a9e21b-16bc-4a3e-85c0-f9d33ee8ae58","order_by":3,"name":"Qiang Lu","email":"","orcid":"","institution":"China University of Petroleum, Beijing","correspondingAuthor":false,"prefix":"","firstName":"Qiang","middleName":"","lastName":"Lu","suffix":""},{"id":587533302,"identity":"b9d2057c-520e-4646-aaa1-5bc5af31c1a1","order_by":4,"name":"Michael Widlansky","email":"","orcid":"","institution":"Medical College of Wisconsin","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"Widlansky","suffix":""},{"id":587533303,"identity":"6c70ae73-3cfb-42c5-81b5-673b6f5063e2","order_by":5,"name":"Jake Luo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAuUlEQVRIiWNgGAWjYBACxgY2BoYHDAxyEC4bsVoSGBiMidcCVgXUkthAtBbm9mOJDxLbbNI33MhOYPhQdpgIh/WkHTZIbEvL3XAjdwPjjHPEaGlIb5NIbDucu+F27gZm3jZitPQ/b/8B1JJuANLylygtM9KOMQC1JIC1MBKn5VmyRMK5NMOZ999uONhzLp2wFsP+NMMPH8ps5PnOnN344EeZNRFaGpA4BwirBwJ5olSNglEwCkbByAYAHW9BKPXJMH8AAAAASUVORK5CYII=","orcid":"","institution":"University of Wisconsin–Milwaukee","correspondingAuthor":true,"prefix":"","firstName":"Jake","middleName":"","lastName":"Luo","suffix":""}],"badges":[],"createdAt":"2026-01-28 01:08:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8715003/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8715003/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105565045,"identity":"cde21ac3-d3c4-4436-a478-781c86848ae1","added_by":"auto","created_at":"2026-03-27 12:51:39","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":12969081,"visible":true,"origin":"","legend":"","description":"","filename":"ECGMImainSpringernature3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8715003/v1_covered_5e45d729-931a-4fc3-9695-8e6bc5bbd292.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Clinically Aligned Explainable AI Improves Myocardial Infarction Detection Using Multi-Lead Electrocardiography","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Myocardial infarction, Electrocardiography, Explainable artificial intelligence, Grad-CAM, U-Net, PTB-XL","lastPublishedDoi":"10.21203/rs.3.rs-8715003/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8715003/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMyocardial infarction (MI), resulting from coronary artery occlusion and myocardial necrosis, is a leading cause of global cardiovascular mortality, requiring rapid diagnosis and localization to guide timely revascularization and optimize outcomes. While artificial intelligence offers a promising avenue to expedite triage and facilitate expert-level interpretation, its clinical adoption remains limited by the ``black box'' nature of many architectures. This opacity undermines clinician trust and impedes integration into clinical workflows. To address this critical gap, we introduce a clinically interpretable deep learning framework combining diagnostic accuracy with transparent decision-making. Our specialized U-Net encoder-decoder architecture processes the 12-lead electrocardiogram as a unified, spatially structured input to preserve inter-lead dependencies essential for detecting coronary territory-specific patterns. Trained on the PTB-XL dataset of 21,799 annotated ECGs, the model performs binary MI detection and classification of major anatomical subtypes: anterior and inferior MI. To enhance clinical trust and explainability, we employed Gradient-weighted Class Activation Mapping, which explains the model's reasoning to confirm that learned features align with established pathophysiological signs. Comparative analysis reveals that the multi-lead model achieves superior discrimination with AUROC 0.97 to 0.99 compared to single-lead variants, with positive and negative predictive values near or exceeding 90%, critical for effective triage and intervention prioritization. Grad-CAM effectively localizes diagnostically relevant regions, aligning with clinical ECG criteria for anterior MI in precordial leads and inferior MI in limb leads, offering valuable visual cues that support clinical decision-making across diverse practice settings. This interpretable framework enhances trust and facilitates clinician-AI collaboration, paving the way for improved MI management through robust decision support and future prospective validation.\u003c/p\u003e","manuscriptTitle":"Clinically Aligned Explainable AI Improves Myocardial Infarction Detection Using Multi-Lead Electrocardiography","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-10 05:52:11","doi":"10.21203/rs.3.rs-8715003/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"fa224c3a-dde4-4eb7-b9a2-ff2440227a40","owner":[],"postedDate":"February 10th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-24T07:42:00+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-10 05:52:11","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8715003","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8715003","identity":"rs-8715003","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}