High-Sensitivity Detection of Invasive Ductal Carcinoma via Domain-Specific SimCLR Pre-Training

High-Sensitivity Detection of Invasive Ductal Carcinoma via Domain-Specific SimCLR Pre-Training | 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 High-Sensitivity Detection of Invasive Ductal Carcinoma via Domain-Specific SimCLR Pre-Training Mufakir Qamar Ansari This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8031909/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 Automated detection of Invasive Ductal Carcinoma (IDC) in digital histopathology images remains challenging due to the domain gap between natural image pre-training and tissue-specific features, as well as significant class imbalance in patch-level datasets. We investigate domain-specific self-supervised learning (SSL) using the SimCLR contrastive framework to pre-train a ResNet50 encoder on 277,524 unlabeled 50 × 50 pixel IDC patches from 162 patients, employing a rigorous patient-stratified split (90% train, 5% validation, 5% test) to prevent data leakage. Following pre-training, we fine-tuned the encoder for binary classification with weighted cross-entropy loss to address the 71.6%/28.4% class imbalance. The SimCLR-pretrained model achieved an AUC-ROC of 0.950 and AUC-PR of 0.886 on the held-out test set, improvements of +0.032 and +0.044 over ImageNet transfer learning, while maintaining high sensitivity (Recall 0.932) and a low false-negative rate (47/13,876 patches). t-SNE and UMAP visualizations demonstrate superior class separation in SSL-derived embeddings, and Grad-CAM heatmaps confirm focus on histologically relevant features. These findings underscore the methodological rigor of large-scale, stratified evaluation and highlight the efficacy of domain-specific SSL for robust, interpretable computational pathology. Nuclear Medicine & Medical Imaging Cancer Biology Self-Supervised Learning SimCLR Histopathology Invasive Ductal Carcinoma Contrastive Learning Digital Pathology ResNet50 Full Text Additional Declarations The authors declare no competing interests. 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-8031909","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":540010047,"identity":"78843f64-b45e-4503-a707-8677efe6388b","order_by":0,"name":"Mufakir Qamar Ansari","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABCUlEQVRIiWNgGAWjYFACxgeMDQYwjoENkGBuYOABcQ7g0sJsgKwlDaiakRgtCN5hwlp0ZyQzfpxRcI+Bf3b7w88VBecT97M3Nj54U8Egx3cjAasWsxvJzJIbDIoZJO6cMZY8Y3A7sYfnYLPhnDMMxpI4teQfkHxgAJS8kcMg2QDSIpHYJs3bxpC4AY8tP0Fa5G+kP/7ZYHAOpKX9N1BLPR4tbECHJTAY3EgwA9pyAGwLM1BLggEuLWces1nOMEjgMbyRY2bZYJBs3HPmYLPknDMShjPPPMCu5Xgy882ePwlyckCH3Wz4Yyfb3t588MObCht5vuPYbYEBHnQBCbzKR8EoGAWjYBTgBwAfi2Sd8/N2wAAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0009-0000-8719-7456","institution":"The University of Toledo","correspondingAuthor":true,"prefix":"","firstName":"Mufakir","middleName":"Qamar","lastName":"Ansari","suffix":""}],"badges":[],"createdAt":"2025-11-04 18:59:59","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8031909/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8031909/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95225632,"identity":"fe9d6e6a-a92f-457d-8b7a-1492c3a8bcb9","added_by":"auto","created_at":"2025-11-05 16:25:19","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1587865,"visible":true,"origin":"","legend":"","description":"","filename":"HighSensitivityDetectionofInvasiveDuctalCarcinoma.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8031909/v1_covered_984b4803-e09d-43e4-90ae-2543c3186047.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eHigh-Sensitivity Detection of Invasive Ductal Carcinoma via Domain-Specific SimCLR Pre-Training\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Toledo","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":"Self-Supervised Learning, SimCLR, Histopathology, Invasive Ductal Carcinoma, Contrastive Learning, Digital Pathology, ResNet50","lastPublishedDoi":"10.21203/rs.3.rs-8031909/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8031909/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAutomated detection of Invasive Ductal Carcinoma (IDC) in digital histopathology images remains challenging due to the domain gap between natural image pre-training and tissue-specific features, as well as significant class imbalance in patch-level datasets. We investigate domain-specific self-supervised learning (SSL) using the SimCLR contrastive framework to pre-train a ResNet50 encoder on 277,524 unlabeled 50 × 50 pixel IDC patches from 162 patients, employing a rigorous patient-stratified split (90% train, 5% validation, 5% test) to prevent data leakage. Following pre-training, we fine-tuned the encoder for binary classification with weighted cross-entropy loss to address the 71.6%/28.4% class imbalance. The SimCLR-pretrained model achieved an AUC-ROC of 0.950 and AUC-PR of 0.886 on the held-out test set, improvements of +0.032 and +0.044 over ImageNet transfer learning, while maintaining high sensitivity (Recall 0.932) and a low false-negative rate (47/13,876 patches). t-SNE and UMAP visualizations demonstrate superior class separation in SSL-derived embeddings, and Grad-CAM heatmaps confirm focus on histologically relevant features. These findings underscore the methodological rigor of large-scale, stratified evaluation and highlight the efficacy of domain-specific SSL for robust, interpretable computational pathology.\u003c/p\u003e","manuscriptTitle":"High-Sensitivity Detection of Invasive Ductal Carcinoma via Domain-Specific SimCLR Pre-Training","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-05 03:12:26","doi":"10.21203/rs.3.rs-8031909/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":"f9de1c80-6c21-474e-a110-4ce663fdceef","owner":[],"postedDate":"November 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":57441911,"name":"Nuclear Medicine \u0026 Medical Imaging"},{"id":57441912,"name":"Cancer Biology"}],"tags":[],"updatedAt":"2025-11-05T03:12:26+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-05 03:12:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8031909","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8031909","identity":"rs-8031909","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}