Decoding Tumor Phenotypes: A Radiologist-Inspired Deep Learning Framework for Breast Cancer Recurrence Prediction

Decoding Tumor Phenotypes: A Radiologist-Inspired Deep Learning Framework for Breast Cancer Recurrence Prediction | 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 Article Decoding Tumor Phenotypes: A Radiologist-Inspired Deep Learning Framework for Breast Cancer Recurrence Prediction Tao Tan, Chunyao Lu, Tianyu Zhang, Xinglong Liang, Antonio Portaluri, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9185479/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Accurate prognostication in breast cancer remains constrained by the limited resolution of conventional clinicopathological markers, often leading to overtreatment of indolent disease or missed opportunities for intensified therapy in aggressive cases. Here we present a radiologist-inspired multimodal deep learning framework that integrates whole-volume dynamic contrast-enhanced MRI (DCE-MRI), radiology reports and clinical variables to predict recurrence-free survival. The model is trained through cross-modal semantic alignment, in which report-derived representations guide the learning of prognostic imaging features. In a large development cohort from the Netherlands Cancer Institute (NKI, n = 3,266), the framework achieved strong discrimination (C-index 0.727) and identified high-risk patients with a hazard ratio of 7.32 (P < 0.0001). Robustness was further assessed in two independent public external cohorts (Duke and I-SPY 1; n = 1,054) using text-free inference, where the model maintained consistent performance (C-index 0.702 and 0.697, respectively). Prognostic accuracy remained stable across both short- and long-term horizons, with high discriminative power from 2 to 12 years after treatment (AUC 0.77–0.83). This temporal stability enabled reliable identification of ultra-low-risk subgroups (5-year Recurrence-free survival > 95%) as well as early detection of aggressive phenotypes prone to early recurrence. These findings support semantically guided multimodal learning as an interpretable and operationally robust imaging biomarker for personalized breast cancer management. Health sciences/Oncology/Cancer/Breast cancer Health sciences/Health care/Prognosis Full Text Additional Declarations There is NO Competing Interest. Supplementary Files Thesupplementaryfile.pdf The supplementary file of paper rawdata.xlsx Source data for the main figures and tables within this paper. Cite Share Download PDF Status: Under Review 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-9185479","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":622596394,"identity":"1faef63a-009c-4e43-bc06-e0299a7791ee","order_by":0,"name":"Tao Tan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1klEQVRIiWNgGAWjYPACNgZ+EJVQQIRaHgYGxgaQFskGkBYD4rUwMBgcAJNEaLFnP3z8wc8dfPbG51cnfnhgwCDPL3aAgC08aYmNvWfYErfdeLtZAugww5mzEwg5LMewgbeNLcHsxtkNIC0JBrcJaeF/Y9j4t43N3njG2c0/iNMikWPYDLSFcQN/7zYibbnxLHG2bBtb4owbvNssEgwkCPuFvT/5wMe3bcfs+fvPbr75o8JGnl+agBYoOMbAIAFWKUGUchCoYWDgP0C06lEwCkbBKBhhAADIxUMTkk9teQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-5403-0887","institution":"The Netherlands Cancer Institute","correspondingAuthor":true,"prefix":"","firstName":"Tao","middleName":"","lastName":"Tan","suffix":""},{"id":622596395,"identity":"74e2d6c3-0abb-401a-9d52-a1e4fffa3bf9","order_by":1,"name":"Chunyao Lu","email":"","orcid":"","institution":"The Netherlands Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Chunyao","middleName":"","lastName":"Lu","suffix":""},{"id":622596396,"identity":"51f125d7-1396-4598-a128-e920ac8f646c","order_by":2,"name":"Tianyu Zhang","email":"","orcid":"https://orcid.org/0009-0007-4724-719X","institution":"Radboud University Medical Centre","correspondingAuthor":false,"prefix":"","firstName":"Tianyu","middleName":"","lastName":"Zhang","suffix":""},{"id":622596397,"identity":"5b3d3aa2-8132-4f23-bebe-78a1ccf7aa05","order_by":3,"name":"Xinglong Liang","email":"","orcid":"","institution":"The Netherlands Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Xinglong","middleName":"","lastName":"Liang","suffix":""},{"id":622596398,"identity":"959a4a60-3a73-4f3b-9f36-47c02fbc166b","order_by":4,"name":"Antonio Portaluri","email":"","orcid":"","institution":"Radboud University Medical Centre","correspondingAuthor":false,"prefix":"","firstName":"Antonio","middleName":"","lastName":"Portaluri","suffix":""},{"id":622596399,"identity":"dd2d0077-de4b-4790-b6ee-bc970c00b7eb","order_by":5,"name":"Luyi Han","email":"","orcid":"https://orcid.org/0000-0003-4046-2763","institution":"The Netherlands Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Luyi","middleName":"","lastName":"Han","suffix":""},{"id":622596400,"identity":"2c86950e-53a8-458b-9a96-3837c6cc6740","order_by":6,"name":"Yaqian Chen","email":"","orcid":"","institution":"Duke University","correspondingAuthor":false,"prefix":"","firstName":"Yaqian","middleName":"","lastName":"Chen","suffix":""},{"id":622596401,"identity":"083a21c3-28b4-43b2-8ac2-524d72b4d05b","order_by":7,"name":"Nika Rasoolzadeh","email":"","orcid":"","institution":"The Netherlands Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Nika","middleName":"","lastName":"Rasoolzadeh","suffix":""},{"id":622596402,"identity":"c0458679-8ba1-40b4-b45f-e6846974a0de","order_by":8,"name":"Ruixiang Qi","email":"","orcid":"","institution":"Affiliated Hangzhou First People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ruixiang","middleName":"","lastName":"Qi","suffix":""},{"id":622596403,"identity":"2ee886eb-cbe3-492b-ab24-1a720faca70f","order_by":9,"name":"Yuan Gao","email":"","orcid":"https://orcid.org/0000-0001-6326-129X","institution":"Netherlands Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Gao","suffix":""},{"id":622596404,"identity":"e8f98719-8773-47f0-a815-7f9df2a0c4b3","order_by":10,"name":"Xin Wang","email":"","orcid":"https://orcid.org/0000-0001-9619-7503","institution":"The Netherlands Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Xin","middleName":"","lastName":"Wang","suffix":""},{"id":622596405,"identity":"d4228f9f-b535-4834-a30f-4130f8bd4a6a","order_by":11,"name":"Yaofei Duan","email":"","orcid":"https://orcid.org/0009-0007-2562-5794","institution":"Macau Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Yaofei","middleName":"","lastName":"Duan","suffix":""},{"id":622596406,"identity":"84f1f627-fb5d-43b4-b8d3-5c645274cbc6","order_by":12,"name":"Zahra Aghdam","email":"","orcid":"","institution":"The Netherlands Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Zahra","middleName":"","lastName":"Aghdam","suffix":""},{"id":622596407,"identity":"5864b26a-928d-4840-b10f-6a0cc3ab6609","order_by":13,"name":"Muzhen He","email":"","orcid":"","institution":"Fujian Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Muzhen","middleName":"","lastName":"He","suffix":""},{"id":622596408,"identity":"1e9b834f-7cd0-48aa-a326-b704ed461e6e","order_by":14,"name":"Jonas Teuwen","email":"","orcid":"https://orcid.org/0000-0002-1825-1428","institution":"The Netherlands Cancer Institute","correspondingAuthor":false,"prefix":"","firstName":"Jonas","middleName":"","lastName":"Teuwen","suffix":""},{"id":622596409,"identity":"f5a4b68c-7ec5-4689-bee3-5cd16e487923","order_by":15,"name":"Maciej Mazurowski","email":"","orcid":"","institution":"Duke University","correspondingAuthor":false,"prefix":"","firstName":"Maciej","middleName":"","lastName":"Mazurowski","suffix":""},{"id":622596410,"identity":"8ee062f6-24e5-46f0-ad33-7ecebddebca7","order_by":16,"name":"Ritse Mann","email":"","orcid":"https://orcid.org/0000-0001-8111-1930","institution":"Radboud UMC","correspondingAuthor":false,"prefix":"","firstName":"Ritse","middleName":"","lastName":"Mann","suffix":""}],"badges":[],"createdAt":"2026-03-21 11:45:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9185479/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9185479/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107482936,"identity":"ecf11fe7-811c-4ef9-a65b-92c47e0ca270","added_by":"auto","created_at":"2026-04-22 02:25:39","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2809444,"visible":true,"origin":"","legend":"Article File","description":"","filename":"Manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9185479/v1_covered_752b825c-459a-4657-a6b5-76932a7797c6.pdf"},{"id":106939540,"identity":"b531e285-bea5-49a3-a782-a8c9966bc8c2","added_by":"auto","created_at":"2026-04-15 04:41:36","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":3503829,"visible":true,"origin":"","legend":"The supplementary file of paper","description":"","filename":"Thesupplementaryfile.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9185479/v1/c663f8a318489befb6e8c796.pdf"},{"id":107480447,"identity":"350b8d93-36ae-4358-967e-d1e7efdc9b87","added_by":"auto","created_at":"2026-04-22 02:10:36","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":9969058,"visible":true,"origin":"","legend":"Source data for the main figures and tables within this paper.","description":"","filename":"rawdata.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9185479/v1/4de0292ed830c84fc325bda3.xlsx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Decoding Tumor Phenotypes: A Radiologist-Inspired Deep Learning Framework for Breast Cancer Recurrence Prediction","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"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":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9185479/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9185479/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Accurate prognostication in breast cancer remains constrained by the limited resolution of conventional clinicopathological markers, often leading to overtreatment of indolent disease or missed opportunities for intensified therapy in aggressive cases. Here we present a radiologist-inspired multimodal deep learning framework that integrates whole-volume dynamic contrast-enhanced MRI (DCE-MRI), radiology reports and clinical variables to predict recurrence-free survival. The model is trained through cross-modal semantic alignment, in which report-derived representations guide the learning of prognostic imaging features. In a large development cohort from the Netherlands Cancer Institute (NKI, n = 3,266), the framework achieved strong discrimination (C-index 0.727) and identified high-risk patients with a hazard ratio of 7.32 (P \u003c 0.0001). Robustness was further assessed in two independent public external cohorts (Duke and I-SPY 1; n = 1,054) using text-free inference, where the model maintained consistent performance (C-index 0.702 and 0.697, respectively). Prognostic accuracy remained stable across both short- and long-term horizons, with high discriminative power from 2 to 12 years after treatment (AUC 0.77–0.83). This temporal stability enabled reliable identification of ultra-low-risk subgroups (5-year Recurrence-free survival \u003e 95%) as well as early detection of aggressive phenotypes prone to early recurrence. These findings support semantically guided multimodal learning as an interpretable and operationally robust imaging biomarker for personalized breast cancer management.","manuscriptTitle":"Decoding Tumor Phenotypes: A Radiologist-Inspired Deep Learning Framework for Breast Cancer Recurrence Prediction","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-15 04:41:32","doi":"10.21203/rs.3.rs-9185479/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"0b8b3d1e-76f1-4db0-b864-864f1a7f9334","owner":[],"postedDate":"April 15th, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"This content is not available.","date":"2026-05-02T09:47:34+00:00","index":3,"fulltext":"This content is not available."}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":66253146,"name":"Health sciences/Oncology/Cancer/Breast cancer"},{"id":66253147,"name":"Health sciences/Health care/Prognosis"}],"tags":[],"updatedAt":"2026-04-15T04:41:32+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-15 04:41:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9185479","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9185479","identity":"rs-9185479","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}