Edge-Compatible Domain Specific Transfer Learning Framework for Automated Knee Osteoarthritis Grading

Edge-Compatible Domain Specific Transfer Learning Framework for Automated Knee Osteoarthritis Grading | 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 Edge-Compatible Domain Specific Transfer Learning Framework for Automated Knee Osteoarthritis Grading Hara Gopal V P, Dr.Nagaraju S This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8001086/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 Accurate Kellgren–Lawrence grading of knee radiographs is essential for managing knee osteoarthritis (KOA). The key variants of convolutional neural networks pretrained on natural images in the literature may miss radiology-specific texture cues. This research presents a novel lightweight domain-adapted transfer learning framework using RadImageNet and fine-tuned on a class-balanced KOA dataset. Using 10-fold cross-validation, optuna-based hyperparameter optimisation improved transferability accuracy for DenseNet-121 and ResNet-50 with F1score = 0.9707 and F1score = 0.9762 respectively. The models are compact (≈ 9–26M parameters; state-dict sizes ≈ 35–98MB) and exhibit per-image inference latency of approximately 31–76ms (FP32, batch=1) in our experiments, indicating feasibility for deployment on edge-computing platforms and portable radiography systems. This study provides one of the first systematic comparisons of RadImageNet and ImageNet pretraining across multiple backbones for KL grading that also report deployment-oriented metrics feasible for edge-computing devices. Deep Learning Kellgren–Lawrence Grading Knee Osteoarthritis RadImageNet Transfer Learning Edge Computing 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-8001086","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":549693976,"identity":"738e91e2-09f0-48ae-881f-e0add7586c54","order_by":0,"name":"Hara Gopal V P","email":"data:image/png;base64,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","orcid":"","institution":"Pondicherry University","correspondingAuthor":true,"prefix":"","firstName":"Hara","middleName":"Gopal V","lastName":"P","suffix":""},{"id":549693977,"identity":"a5f52d5a-4e5a-4c80-9026-81f90e9fea68","order_by":1,"name":"Dr.Nagaraju S","email":"","orcid":"","institution":"Pondicherry University","correspondingAuthor":false,"prefix":"Dr.","firstName":"Nagaraju","middleName":"","lastName":"S","suffix":""}],"badges":[],"createdAt":"2025-10-31 18:53:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8001086/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8001086/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":96701246,"identity":"116198ef-2cd9-487a-a8fc-7e8af99b4cfd","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"json","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4288,"visible":true,"origin":"","legend":"","description":"","filename":"0cab31e828dc44d5a0a5f74ba52cddea.json","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/04a4458a5b6059e32dfcbc9a.json"},{"id":96701247,"identity":"cebf1a20-af19-4f79-910a-4bb66f701e5c","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"xml","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":88288,"visible":true,"origin":"","legend":"","description":"","filename":"0cab31e828dc44d5a0a5f74ba52cddea1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/e7d91e44b593ec8be9a24f59.xml"},{"id":96701248,"identity":"91551edd-ae7c-43c2-9394-82cc935f412f","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":215112,"visible":true,"origin":"","legend":"","description":"","filename":"CoverLetter.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/ff5ae1393d19f40da64d8c38.pdf"},{"id":96711256,"identity":"bb7128a6-f85c-4f89-b557-51332f84e271","added_by":"auto","created_at":"2025-11-25 10:11:48","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":219951,"visible":true,"origin":"","legend":"","description":"","filename":"QCRevision5.11.25.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/d2d0d124a314dfd88a3d88e8.pdf"},{"id":96701261,"identity":"af48bfd5-839c-4c30-8a8c-0e4fd4519730","added_by":"auto","created_at":"2025-11-25 08:27:56","extension":"pdf","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2537204,"visible":true,"origin":"","legend":"","description":"","filename":"RadXKOA.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/4a7ae706dea7a36051b6d220.pdf"},{"id":96711413,"identity":"164aeb7b-4de2-4a14-9e12-2dd70d363565","added_by":"auto","created_at":"2025-11-25 10:11:58","extension":"bst","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151345,"visible":true,"origin":"","legend":"","description":"","filename":"snapacite.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/16b5cf9ab6c9d3a7577c4f05.bst"},{"id":96711083,"identity":"76f42f15-01b9-4381-97a3-f55b25511a22","added_by":"auto","created_at":"2025-11-25 10:11:38","extension":"bst","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29828,"visible":true,"origin":"","legend":"","description":"","filename":"snaps.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/9a6f41a913e8ac6a19d14e34.bst"},{"id":96701250,"identity":"cc4995ad-9d89-469b-a921-613a44d9aec6","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"bst","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35515,"visible":true,"origin":"","legend":"","description":"","filename":"snbasic.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/606408281cdc659e6b6475fd.bst"},{"id":96711346,"identity":"a8f2acdd-b8c5-4047-ad54-e4d97061ad1b","added_by":"auto","created_at":"2025-11-25 10:11:53","extension":"bst","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":33968,"visible":true,"origin":"","legend":"","description":"","filename":"snchicago.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/488f40182e60fcf0cf9ea599.bst"},{"id":96701253,"identity":"fc23abb1-0f47-49f7-a3c5-38a2afa70dbd","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"cls","order_by":14,"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-8001086/v1/f10cd66060400d341a831aa8.cls"},{"id":96701259,"identity":"667cfe5f-803e-4c64-ab6b-8b91fb144d4b","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"bst","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64023,"visible":true,"origin":"","legend":"","description":"","filename":"snmathphysay.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/2d88dbcb1cb8d4dc1b6e126d.bst"},{"id":96711270,"identity":"50dec12e-7b69-4e77-8faf-190678abaf7d","added_by":"auto","created_at":"2025-11-25 10:11:49","extension":"bst","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64164,"visible":true,"origin":"","legend":"","description":"","filename":"snmathphysnum.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/20d917f8e83684a070a3e9f4.bst"},{"id":96701257,"identity":"03854595-dfd2-4a1a-abf8-89635544b009","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"bst","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39056,"visible":true,"origin":"","legend":"","description":"","filename":"snnature.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/589c3ea94d90cb05ffea3095.bst"},{"id":96701256,"identity":"ecb2cc5d-1f8c-4894-8a28-b121051179ea","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"bst","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39951,"visible":true,"origin":"","legend":"","description":"","filename":"snvancouveray.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/07ef251fb8de74f0878345eb.bst"},{"id":96701255,"identity":"e92611ee-1a04-4d39-a65e-42ea1f5ec5d7","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"bst","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40758,"visible":true,"origin":"","legend":"","description":"","filename":"snvancouvernum.bst","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/3e6d86fff8d9ffe26858f123.bst"},{"id":96711601,"identity":"7f7440a4-29bc-4424-8a98-b9c4c37aa57a","added_by":"auto","created_at":"2025-11-25 10:12:17","extension":"xml","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":91524,"visible":true,"origin":"","legend":"","description":"","filename":"0cab31e828dc44d5a0a5f74ba52cddea1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/9aa1ac72f7c35258a32d521e.xml"},{"id":96701260,"identity":"f61c6016-e1a3-48c2-88b0-f62c226670d8","added_by":"auto","created_at":"2025-11-25 08:27:55","extension":"html","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":106200,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1/3efb3a8fe16978ff30280280.html"},{"id":104287380,"identity":"53bbdf32-856a-423d-8ddb-d57887f4bc69","added_by":"auto","created_at":"2026-03-10 05:41:04","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1538788,"visible":true,"origin":"","legend":"","description":"","filename":"RadXKOA.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8001086/v1_covered_30cb916e-5817-4c1b-999b-f5ff7906eb5f.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Edge-Compatible Domain Specific Transfer Learning Framework for Automated Knee Osteoarthritis Grading","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":"Deep Learning, Kellgren–Lawrence Grading, Knee Osteoarthritis, RadImageNet, Transfer Learning, Edge Computing","lastPublishedDoi":"10.21203/rs.3.rs-8001086/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8001086/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAccurate Kellgren–Lawrence grading of knee radiographs is essential for managing knee osteoarthritis (KOA). The key variants of convolutional neural networks pretrained on natural images in the literature may miss radiology-specific texture cues. This research presents a novel lightweight domain-adapted transfer learning framework using RadImageNet and fine-tuned on a class-balanced KOA dataset. Using 10-fold cross-validation, optuna-based hyperparameter optimisation improved transferability accuracy for DenseNet-121 and ResNet-50 with F1score = 0.9707 and F1score = 0.9762 respectively. The models are compact (≈ 9–26M parameters; state-dict sizes ≈ 35–98MB) and exhibit per-image inference latency of approximately 31–76ms (FP32, batch=1) in our experiments, indicating feasibility for deployment on edge-computing platforms and portable radiography systems. This study provides one of the first systematic comparisons of RadImageNet and ImageNet pretraining across multiple backbones for KL grading that also report deployment-oriented metrics feasible for edge-computing devices.\u003c/p\u003e","manuscriptTitle":"Edge-Compatible Domain Specific Transfer Learning Framework for Automated Knee Osteoarthritis Grading","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-25 08:27:51","doi":"10.21203/rs.3.rs-8001086/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":"586775cc-c1cf-4177-a714-b045e3298a2e","owner":[],"postedDate":"November 25th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-10T05:40:44+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-25 08:27:51","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8001086","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8001086","identity":"rs-8001086","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}