FedCKD: Cluster-Aware Knowledge Distillation for Heterogeneous Medical Federated Learning

FedCKD: Cluster-Aware Knowledge Distillation for Heterogeneous Medical Federated Learning | 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 FedCKD: Cluster-Aware Knowledge Distillation for Heterogeneous Medical Federated Learning Yi Xu, Kun Chen, Haoyu Luo, Xiao Liu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7250878/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 In medical knowledge systems, federated learning provides a promising paradigm for collaborative knowledge extraction while preserving data privacy. However, inherent heterogeneity in medical information—stemming from variations in disease distribution, imaging protocols, and patient demographics—severely degrades the performance of traditional federated frameworks. To address this challenge, we propose FedCKD , a knowledge-driven federated framework tailored for heterogeneous medical information systems. FedCKD introduces three key innovations: (1) a label-driven knowledge clustering mechanism that partitions medical nodes based on disease-specific knowledge representations, ensuring intra-cluster semantic consistency; (2) a two-stage adaptive aggregation strategy for knowledge-oriented model fusion within each cluster, balancing local specialization and cluster-level consistency; (3) a cross-cluster knowledge distillation protocol that enables privacy-preserving transfer of complementary knowledge across specialized medical domains via weighted teacher ensembles. By simulating interoperability in distributed medical systems, FedCKD achieves cross-domain knowledge integration while respecting statistical heterogeneity. Comprehensive experiments on multiple datasets demonstrate that FedCKD significantly outperforms state-of-the-art methods, establishing it as an effective solution for knowledge extraction and integration in privacy-sensitive, heterogeneous medical ecosystems. Federated Learning Medical Information Systems Knowledge Distillation Data Heterogeneity Knowledge Clustering 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-7250878","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":501534845,"identity":"2dd52d6b-9a97-4d70-a953-264d6406de68","order_by":0,"name":"Yi Xu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIiWNgGAWjYJCCAwkGNnb8zMwHHxCv5UFFWrJkO1uyAdFaGB+cOcy44TyPmQBRyg1u5BgeSGxjZjY+zGDGwFBjE01Qi+SMHAOgFjY+s8MMaQ8YjqXlNhDSwi8B1sLDDNRy3ICx4TBhLWwQLRKMm5sZgSQxWsC2JJwxYNzAzMxGnBbJnmcFBxIqEpIlDrMxGyQQ4xeD48mbP/4w+G/H33/+44MPNTaEtTAwcCBFYAJh5SDA/oA4daNgFIyCUTByAQDsQkBlf3NC8AAAAABJRU5ErkJggg==","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Yi","middleName":"","lastName":"Xu","suffix":""},{"id":501534846,"identity":"34eb57f8-0c71-4954-8970-1294f7a212b3","order_by":1,"name":"Kun Chen","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Kun","middleName":"","lastName":"Chen","suffix":""},{"id":501534847,"identity":"e58f5296-d589-492f-be62-ba91576f71a0","order_by":2,"name":"Haoyu Luo","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Haoyu","middleName":"","lastName":"Luo","suffix":""},{"id":501534848,"identity":"b36fdf61-2db1-45bf-8aea-8d0a30c2e20f","order_by":3,"name":"Xiao Liu","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Xiao","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2025-07-30 09:08:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7250878/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7250878/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":92628164,"identity":"df8d4a52-92f3-4ef6-9b9d-2a185d7c8650","added_by":"auto","created_at":"2025-10-02 01:31:33","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1487983,"visible":true,"origin":"","legend":"","description":"","filename":"FedCKD.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7250878/v1_covered_e6667f9e-9976-41a7-80df-094d9e641693.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"FedCKD: Cluster-Aware Knowledge Distillation for Heterogeneous Medical Federated Learning","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":"Federated Learning, Medical Information Systems, Knowledge Distillation, Data Heterogeneity, Knowledge Clustering","lastPublishedDoi":"10.21203/rs.3.rs-7250878/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7250878/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn medical knowledge systems, federated learning provides a promising paradigm for collaborative knowledge extraction while preserving data privacy. However, inherent heterogeneity in medical information\u0026mdash;stemming from variations in disease distribution, imaging protocols, and patient demographics\u0026mdash;severely degrades the performance of traditional federated frameworks. To address this challenge, we propose \u003cb\u003eFedCKD\u003c/b\u003e, a knowledge-driven federated framework tailored for heterogeneous medical information systems. FedCKD introduces three key innovations: (1) a label-driven knowledge clustering mechanism that partitions medical nodes based on disease-specific knowledge representations, ensuring intra-cluster semantic consistency; (2) a two-stage adaptive aggregation strategy for knowledge-oriented model fusion within each cluster, balancing local specialization and cluster-level consistency; (3) a cross-cluster knowledge distillation protocol that enables privacy-preserving transfer of complementary knowledge across specialized medical domains via weighted teacher ensembles. By simulating interoperability in distributed medical systems, FedCKD achieves cross-domain knowledge integration while respecting statistical heterogeneity. Comprehensive experiments on multiple datasets demonstrate that FedCKD significantly outperforms state-of-the-art methods, establishing it as an effective solution for knowledge extraction and integration in privacy-sensitive, heterogeneous medical ecosystems.\u003c/p\u003e","manuscriptTitle":"FedCKD: Cluster-Aware Knowledge Distillation for Heterogeneous Medical Federated Learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-22 05:54:52","doi":"10.21203/rs.3.rs-7250878/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":"cae0d4fc-90b3-49a4-ad4e-b3869a7b21a5","owner":[],"postedDate":"August 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-12-25T17:38:12+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-22 05:54:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7250878","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7250878","identity":"rs-7250878","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}