Enhancing Medical Anomaly Detection via Text-Adapted Few-Shot Learning with Visual-Language Models

Enhancing Medical Anomaly Detection via Text-Adapted Few-Shot Learning with Visual-Language Models | 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 Enhancing Medical Anomaly Detection via Text-Adapted Few-Shot Learning with Visual-Language Models Keming Mao, Shengbin Hou, Haoming Fang, Jianzhe Zhao, Xinlu Xiao This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8450850/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Medical image anomaly detection (AD) is crucial for early disease diagnosis, yet it faces challenges such as data heterogeneity and scarcity of annotated samples. This paper introduces a text-adapted few-shot training framework using CLIP, which extends the text encoder to incorporate fine-grained descriptions and introduces a text feature adapter for better alignment with image representations. A text-image feature alignment module and a contrastive learning mechanism are presented to enhance cross-modal integration and the distinction between normal and abnormal samples. Experimental evaluations on six medical imaging datasets demonstrate that our method significantly outperforms state-of-the-art techniques in both classification and segmentation tasks, achieving an average improvement of 1.13% in AUC. The implementation code is available at https://github.com/clownddd/TAFT . Few-shot learning Medical image anomaly detection Text feature adapter Text-image feature alignment Contrastive learning Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 14 Feb, 2026 Reviews received at journal 11 Feb, 2026 Reviews received at journal 18 Jan, 2026 Reviewers agreed at journal 12 Jan, 2026 Reviewers agreed at journal 08 Jan, 2026 Reviewers invited by journal 08 Jan, 2026 Editor assigned by journal 26 Dec, 2025 Submission checks completed at journal 26 Dec, 2025 First submitted to journal 25 Dec, 2025 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-8450850","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":572502716,"identity":"177aeef9-f45c-4b55-83f7-badae8c041a1","order_by":0,"name":"Keming Mao","email":"","orcid":"","institution":"Northeastern University","correspondingAuthor":false,"prefix":"","firstName":"Keming","middleName":"","lastName":"Mao","suffix":""},{"id":572502719,"identity":"e4ebacb6-3f3b-4510-a448-eb9e2b5f1aec","order_by":1,"name":"Shengbin Hou","email":"","orcid":"","institution":"Northeastern University","correspondingAuthor":false,"prefix":"","firstName":"Shengbin","middleName":"","lastName":"Hou","suffix":""},{"id":572502721,"identity":"efeab627-abf7-409d-a282-29a61f8eb71e","order_by":2,"name":"Haoming Fang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/UlEQVRIie3QMUvEMBTA8YRAi5Cj63MQv0IgUG447r5KQuEmB0GQDgcGhDh2rd/CWw7HHA86BW7N4OLi5FDxA5ztIOiQcjg55D89Aj/yEkJSqX9YwZhzfX28a34cikly/mD1vvWMPhp2IhHeSzyzjD65UwkJSiDxGZOHQwf0eaFNfr8DsnmJCtoqhdc1z8pQZUD9Whve3QLp3qKEgXLYeuBlYAOxqA1clUANRkkG2uDMCpANjuSozeX7NOEcyUCUEGRczLrhFj5NILdk+GSnIFRyrm0lLV/fzFUXJyssPvu+dqpo9q/hwy4vmhy3od/Eye/U+LrvIZVKpVJ/7wveWledsbo9VQAAAABJRU5ErkJggg==","orcid":"","institution":"Northeastern University","correspondingAuthor":true,"prefix":"","firstName":"Haoming","middleName":"","lastName":"Fang","suffix":""},{"id":572502722,"identity":"fcd5c7a6-eeba-47b1-a3e1-6f1c18378eaf","order_by":3,"name":"Jianzhe Zhao","email":"","orcid":"","institution":"Northeastern University","correspondingAuthor":false,"prefix":"","firstName":"Jianzhe","middleName":"","lastName":"Zhao","suffix":""},{"id":572502723,"identity":"dfcaf4b9-a9f1-4654-8392-ca959026a114","order_by":4,"name":"Xinlu Xiao","email":"","orcid":"","institution":"China Mobile (China)","correspondingAuthor":false,"prefix":"","firstName":"Xinlu","middleName":"","lastName":"Xiao","suffix":""}],"badges":[],"createdAt":"2025-12-25 20:23:07","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8450850/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8450850/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100078292,"identity":"eb14def1-f46c-430a-92ce-c80942dd5f2c","added_by":"auto","created_at":"2026-01-12 17:56:00","extension":"json","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6165,"visible":true,"origin":"","legend":"","description":"","filename":"2291639bfc294d39ba8d733439aa08fc.json","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/3a61b2db2c7f3bfd13b038b5.json"},{"id":100364389,"identity":"a996ba5b-7580-4762-a9a6-43c9e95e37d3","added_by":"auto","created_at":"2026-01-16 07:53:35","extension":"xml","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":91447,"visible":true,"origin":"","legend":"","description":"","filename":"2291639bfc294d39ba8d733439aa08fc1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/53b9812f3329d33a9f4b951e.xml"},{"id":100078296,"identity":"85ac655d-f58a-4a31-82ea-e25966796a5c","added_by":"auto","created_at":"2026-01-12 17:56:00","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4400834,"visible":true,"origin":"","legend":"","description":"","filename":"TAFT.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/e0a19ac6636eead78ea4d180.pdf"},{"id":100364546,"identity":"34a12d98-999f-48d4-9374-6dc72d256126","added_by":"auto","created_at":"2026-01-16 07:53:54","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":522677,"visible":true,"origin":"","legend":"","description":"","filename":"fig.2.png","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/be120a9d52c0e8f72a583bf9.png"},{"id":100363670,"identity":"b33cc8aa-271f-4349-acca-9048576616fc","added_by":"auto","created_at":"2026-01-16 07:51:04","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":311150,"visible":true,"origin":"","legend":"","description":"","filename":"figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/c2f87b22d925f54633cb52be.png"},{"id":100365071,"identity":"3138179c-bf92-46e0-843c-af5ef08d4cbd","added_by":"auto","created_at":"2026-01-16 07:54:39","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":183626,"visible":true,"origin":"","legend":"","description":"","filename":"figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/7c9c2f2dda4238b5e1348278.png"},{"id":100078300,"identity":"812d4f2c-c2b1-4589-b087-09909c0403c6","added_by":"auto","created_at":"2026-01-12 17:56:00","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40849,"visible":true,"origin":"","legend":"","description":"","filename":"figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/d9a5a7f6387bdf3327a8c033.png"},{"id":100364726,"identity":"47bfc71d-8de1-4352-a5f7-07c45bbbe52e","added_by":"auto","created_at":"2026-01-16 07:54:15","extension":"jpg","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2465862,"visible":true,"origin":"","legend":"","description":"","filename":"shot.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/61958097fac29eb1d7515fad.jpg"},{"id":100078299,"identity":"46931595-922b-4c45-8143-b9cbf7350376","added_by":"auto","created_at":"2026-01-12 17:56:00","extension":"cls","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55857,"visible":true,"origin":"","legend":"","description":"","filename":"snjnl.cls","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/0dbf67deb9173431eb6267c4.cls"},{"id":100364820,"identity":"8dc96e80-e37b-48da-ba1d-1ad17ce33891","added_by":"auto","created_at":"2026-01-16 07:54:21","extension":"bst","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64166,"visible":true,"origin":"","legend":"","description":"","filename":"snmathphysnum.bst","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/ad4e699bb07e6f7b68767cab.bst"},{"id":100078301,"identity":"d390e487-2c39-4836-95b4-8de876575de8","added_by":"auto","created_at":"2026-01-12 17:56:00","extension":"xml","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97563,"visible":true,"origin":"","legend":"","description":"","filename":"2291639bfc294d39ba8d733439aa08fc1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/57d30aa4c05ed868bb008a85.xml"},{"id":100078303,"identity":"8a24dc12-af91-4b0c-abd8-7658201b32fd","added_by":"auto","created_at":"2026-01-12 17:56:00","extension":"html","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":108883,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1/485f703f1856bc143e447395.html"},{"id":100381956,"identity":"5d98d308-8f64-4aaf-b91b-40d5fe807023","added_by":"auto","created_at":"2026-01-16 10:40:06","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4108313,"visible":true,"origin":"","legend":"","description":"","filename":"TAFT.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8450850/v1_covered_e71c1d6f-ae0c-47b9-a4b9-8a03160f9ff4.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Enhancing Medical Anomaly Detection via Text-Adapted Few-Shot Learning with Visual-Language Models","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"the-visual-computer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"tvcj","sideBox":"Learn more about [The Visual Computer](http://link.springer.com/journal/371)","snPcode":"371","submissionUrl":"https://submission.nature.com/new-submission/371/3","title":"The Visual Computer","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Few-shot learning, Medical image anomaly detection, Text feature adapter, Text-image feature alignment, Contrastive learning","lastPublishedDoi":"10.21203/rs.3.rs-8450850/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8450850/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMedical image anomaly detection (AD) is crucial for early disease diagnosis, yet it faces challenges such as data heterogeneity and scarcity of annotated samples. This paper introduces a text-adapted few-shot training framework using CLIP, which extends the text encoder to incorporate fine-grained descriptions and introduces a text feature adapter for better alignment with image representations. A text-image feature alignment module and a contrastive learning mechanism are presented to enhance cross-modal integration and the distinction between normal and abnormal samples. Experimental evaluations on six medical imaging datasets demonstrate that our method significantly outperforms state-of-the-art techniques in both classification and segmentation tasks, achieving an average improvement of 1.13% in AUC. The implementation code is available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/clownddd/TAFT\u003c/span\u003e\u003cspan address=\"https://github.com/clownddd/TAFT\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. \u003c/p\u003e","manuscriptTitle":"Enhancing Medical Anomaly Detection via Text-Adapted Few-Shot Learning with Visual-Language Models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-12 17:55:55","doi":"10.21203/rs.3.rs-8450850/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-02-14T12:57:24+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-11T18:59:53+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-19T01:25:29+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"26262322591135914892301103544207856961","date":"2026-01-12T08:34:47+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"232309210707116023637890615009825973046","date":"2026-01-08T11:22:44+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-08T10:49:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-12-27T00:05:49+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-12-26T08:17:52+00:00","index":"","fulltext":""},{"type":"submitted","content":"The Visual Computer","date":"2025-12-25T20:10:50+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"the-visual-computer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"tvcj","sideBox":"Learn more about [The Visual Computer](http://link.springer.com/journal/371)","snPcode":"371","submissionUrl":"https://submission.nature.com/new-submission/371/3","title":"The Visual Computer","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"e7c13f09-9a6f-4515-b46b-8880a3ccd81c","owner":[],"postedDate":"January 12th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-13T09:41:16+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-12 17:55:55","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8450850","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8450850","identity":"rs-8450850","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}