Enhancing Diabetic Retinopathy Prediction Using Transformer-based Attention in Hybrid CNN Models

Enhancing Diabetic Retinopathy Prediction Using Transformer-based Attention in Hybrid CNN 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 Article Enhancing Diabetic Retinopathy Prediction Using Transformer-based Attention in Hybrid CNN Models Aayush Verma, Sanket Agrawal, Shreyans Jain, S Kanthimathi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6552005/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 Diabetic retinopathy (DR) is one of the major causes of blindness across the globe and hence its early and accurate detection is required to avoid drastic vision loss. In this study, we introduce a hybrid learning method that utilizes the combination of deep learning models with transformer-based attention mechanisms to make improved predictions for diabetic retinopathy. Our approach utilizes an ensemble of pre-trained models including InceptionV3, DenseNet121, VGG16, MobileNetV2, and ResNet50, which are individually renowned for their robust feature extraction ability. By incorporating self-attention and multi-head attention mechanisms with the hybrid models, we try to enhance feature representation and obtain increased classification accuracy. Our experimental findings indicate that such hybrid architectures are able to learn intricate retinal patterns and improve model performance compared to individual architectures. Surprisingly, the integration of ResNet50 and DenseNet121 with a transformer-based attention mechanism provided the most stable accuracy and robust results. This paper demonstrates the potential of hybrid deep learning models with the inclusion of attention mechanisms as a viable solution for enhanced diabetic retinopathy diagnosis. Our findings promote the advancement of sophisticated automatic medical image analysis techniques and improve clinical decision support systems for retinal disease detection. Health sciences/Diseases Health sciences/Medical research Physical sciences/Engineering Transformer-Based Attention Self Attention Multi Head Attention Transfer Learning 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-6552005","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":510278812,"identity":"08659e9a-768e-467c-bf81-26d702bce46b","order_by":0,"name":"Aayush Verma","email":"","orcid":"","institution":"Vellore institute of technology , Chennai","correspondingAuthor":false,"prefix":"","firstName":"Aayush","middleName":"","lastName":"Verma","suffix":""},{"id":510278814,"identity":"0e6b0e89-44d2-4b66-b37a-0a1546308074","order_by":1,"name":"Sanket Agrawal","email":"","orcid":"","institution":"Vellore institute of technology , Chennai","correspondingAuthor":false,"prefix":"","firstName":"Sanket","middleName":"","lastName":"Agrawal","suffix":""},{"id":510278816,"identity":"4afb7e1b-b999-4c01-9c9d-f3aba83e0919","order_by":2,"name":"Shreyans Jain","email":"","orcid":"","institution":"Vellore institute of technology , Chennai","correspondingAuthor":false,"prefix":"","firstName":"Shreyans","middleName":"","lastName":"Jain","suffix":""},{"id":510278818,"identity":"5e39618b-04b6-4e46-9672-b4aeaf35c6ef","order_by":3,"name":"S Kanthimathi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+0lEQVRIiWNgGAWjYBACCSBmZjAAsRgbGBgqGHgg4gZ4tTA2gxTwgLWcYeDhIU4LA0gLkGBsY4BZgxtItp89/rigYJu8vXRzA+PPeYdl7BmYH35gKLiDU4s0T15i8wyD24Y9MgcbmHm3HQY6jM1YgsHgGU4tcgw5hs08BrcZeyQSG5gZt6WB/GIG9Mth3Fr434C12IO0MP6cA9LC/g2vFmkJiC2JIC0MvA02QC08+G2RnPHGcDZQS3LPjcSGwzzHgFoO8xRLJODRInE+x+Azz5/btu0z0h8+/FEjYc/e3r7xw4c/uLWggANgkhmIE4jTMApGwSgYBaMABwAAOVlJ693yQkwAAAAASUVORK5CYII=","orcid":"","institution":"Vellore institute of technology , Chennai","correspondingAuthor":true,"prefix":"","firstName":"S","middleName":"","lastName":"Kanthimathi","suffix":""}],"badges":[],"createdAt":"2025-04-29 03:23:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6552005/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6552005/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":92020085,"identity":"b10a2ffe-05ca-44b0-8afd-8694765a28ab","added_by":"auto","created_at":"2025-09-23 17:30:41","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1247296,"visible":true,"origin":"","legend":"","description":"","filename":"EnhancingDiabeticRetinopathyPredictionUsingTransformerBasedAttentioninHybridCNNModels1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6552005/v1/c557ff1a367dbf76ad16074d.pdf"},{"id":92020084,"identity":"22e2d956-b6c9-4882-9768-c732ba0d1bb2","added_by":"auto","created_at":"2025-09-23 17:30:41","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5997,"visible":true,"origin":"","legend":"","description":"","filename":"90cc224ac69b4f84ad86715f2e4aec2f.json","url":"https://assets-eu.researchsquare.com/files/rs-6552005/v1/b12d74d8a33c65f9007434b6.json"},{"id":107311677,"identity":"30f86fb0-74f3-41fe-b040-b0b7d5029db1","added_by":"auto","created_at":"2026-04-20 09:14:13","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":836920,"visible":true,"origin":"","legend":"","description":"","filename":"EnhancingDiabeticRetinopathyPredictionUsingTransformerBasedAttentioninHybridCNNModels1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6552005/v1_covered_7b10a6d8-a4ca-463f-833f-3e8940be5b35.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eEnhancing Diabetic Retinopathy Prediction Using Transformer-based Attention in Hybrid CNN Models\u003c/p\u003e","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":"Transformer-Based Attention, Self Attention, Multi Head Attention, Transfer Learning","lastPublishedDoi":"10.21203/rs.3.rs-6552005/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6552005/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDiabetic retinopathy (DR) is one of the major causes of blindness across the globe and hence its early and accurate detection is required to avoid drastic vision loss. In this study, we introduce a hybrid learning method that utilizes the combination of deep learning models with transformer-based attention mechanisms to make improved predictions for diabetic retinopathy. Our approach utilizes an ensemble of pre-trained models including InceptionV3, DenseNet121, VGG16, MobileNetV2, and ResNet50, which are individually renowned for their robust feature extraction ability. By incorporating self-attention and multi-head attention mechanisms with the hybrid models, we try to enhance feature representation and obtain increased classification accuracy.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur experimental findings indicate that such hybrid architectures are able to learn intricate retinal patterns and improve model performance compared to individual architectures. Surprisingly, the integration of ResNet50 and DenseNet121 with a transformer-based attention mechanism provided the most stable accuracy and robust results. This paper demonstrates the potential of hybrid deep learning models with the inclusion of attention mechanisms as a viable solution for enhanced diabetic retinopathy diagnosis. Our findings promote the advancement of sophisticated automatic medical image analysis techniques and improve clinical decision support systems for retinal disease detection.\u003c/p\u003e","manuscriptTitle":"Enhancing Diabetic Retinopathy Prediction Using Transformer-based Attention in Hybrid CNN Models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-23 17:30:36","doi":"10.21203/rs.3.rs-6552005/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":"e4562c78-ac67-4537-975a-2f5684579476","owner":[],"postedDate":"September 23rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":54487787,"name":"Health sciences/Diseases"},{"id":54487788,"name":"Health sciences/Medical research"},{"id":54487789,"name":"Physical sciences/Engineering"}],"tags":[],"updatedAt":"2026-04-20T09:12:00+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-23 17:30:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6552005","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6552005","identity":"rs-6552005","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}