Mammo-SAE: Interpreting Breast Cancer Concept Learning with Sparse Autoencoders

Mammo-SAE: Interpreting Breast Cancer Concept Learning with Sparse Autoencoders | 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 Mammo-SAE: Interpreting Breast Cancer Concept Learning with Sparse Autoencoders Krishna Kanth Nakka This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7614664/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 Interpretability is critical in high-stakes domains such as medical imaging, where understanding model decisions is essential for clinical adoption. In this work, we introduce Sparse Autoencoder (SAE)-based interpretability to breast imaging by analyzing {Mammo-CLIP}, a vision--language foundation model pretrained on large-scale mammogram image--radiology report pairs. We train a patch-level \texttt{Mammo-SAE} on Mammo-CLIP visual features to identify and probe latent neurons associated with clinically relevant breast concepts such as \textit{mass} and \textit{suspicious calcification}. We show that top-activated class-level latent neurons often tend to align with ground-truth regions, and also uncover several confounding factors influencing the model’s decision-making process. Furthermore, we demonstrate that finetuning Mammo-CLIP leads to larger concept separation in the latent space, improving interpretability and predictive performance. Our findings suggest that sparse latent representations offer a powerful lens into the internal behavior of breast foundation models. The code will be released at https://krishnakanthnakka.github.io/MammoSAE/. Biomedical Engineering Cancer Biology Sparse Autoencoders Breast Cancer Mammography Breast Interpretability 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-7614664","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":528259968,"identity":"df20650f-8fa0-456f-942d-1bab1c3522d4","order_by":0,"name":"Krishna Kanth Nakka","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAklEQVRIiWNgGAWjYBACCQbGBmaGAxDWgw8VQAYzcwNxWngkGJgNZ5wBaWEkpAWoBqqFTZq3DSRGQIvk7MPNnwvO2OXZSzewSfDOq43mbwdq+VGxDacWab7ENukZN5KLeWQOMFtIbjueO+MwYwNjz5nbOLXI8TC2MfN8YE7skUhgvGG47VhuA1ALM2MbXi3Nn3k+1IO0MEgkzjmWO5+QFmkexgZpnhuHQVqYJA421ORuIKRFsoexTZrnzPHEnhuJzYYNxw7kbgRqOYjPLxJn2B9/5jlWndg+I/ng4z81dbnzzh8++OBHBW4tSAAcHYfBzAPEqIeBOlIUj4JRMApGwQgBALXqW6uqo6W/AAAAAElFTkSuQmCC","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Krishna","middleName":"Kanth","lastName":"Nakka","suffix":""}],"badges":[],"createdAt":"2025-09-14 20:54:48","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-7614664/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7614664/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93443146,"identity":"81d1a2e4-407d-4b4c-b224-f9f614179b0d","added_by":"auto","created_at":"2025-10-14 00:23:40","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2126460,"visible":true,"origin":"","legend":"","description":"","filename":"MammoSAEv11.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7614664/v1_covered_3440180c-97d0-455a-958d-bc4fc686106f.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eMammo-SAE: Interpreting Breast Cancer Concept Learning with Sparse Autoencoders\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Independent work","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":"Sparse Autoencoders, Breast Cancer, Mammography, Breast Interpretability","lastPublishedDoi":"10.21203/rs.3.rs-7614664/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7614664/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eInterpretability is critical in high-stakes domains such as medical imaging, where understanding model decisions is essential for clinical adoption. In this work, we introduce Sparse Autoencoder (SAE)-based interpretability to breast imaging by analyzing {Mammo-CLIP}, a vision--language foundation model pretrained on large-scale mammogram image--radiology report pairs. We train a patch-level \\texttt{Mammo-SAE} on Mammo-CLIP visual features to identify and probe latent neurons associated with clinically relevant breast concepts such as \\textit{mass} and \\textit{suspicious calcification}. We show that top-activated class-level latent neurons often tend to align with ground-truth regions, and also uncover several confounding factors influencing the model’s decision-making process. Furthermore, we demonstrate that finetuning Mammo-CLIP leads to larger concept separation in the latent space, improving interpretability and predictive performance. Our findings suggest that sparse latent representations offer a powerful lens into the internal behavior of breast foundation models.\u0026nbsp;The code will be released at https://krishnakanthnakka.github.io/MammoSAE/.\u003c/p\u003e","manuscriptTitle":"Mammo-SAE: Interpreting Breast Cancer Concept Learning with Sparse Autoencoders","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-14 00:15:28","doi":"10.21203/rs.3.rs-7614664/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":"da0d9bc5-8d51-41b9-8b99-c3b66cbccc13","owner":[],"postedDate":"October 14th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":56149108,"name":"Biomedical Engineering"},{"id":56149109,"name":"Cancer Biology"}],"tags":[],"updatedAt":"2025-10-14T00:15:28+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-14 00:15:28","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7614664","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7614664","identity":"rs-7614664","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}