Integrating CBAM into YOLOv9c for Polyp Detection: A Systematic Cross-Dataset Evaluation

Integrating CBAM into YOLOv9c for Polyp Detection: A Systematic Cross-Dataset Evaluation | 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 Integrating CBAM into YOLOv9c for Polyp Detection: A Systematic Cross-Dataset Evaluation Shivangi Krishna, Soren Salehi, Noha Ghatwary, Omar O. Ebada, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9628010/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 and real-time detection of colorectal polyps is important for early diagnosis and prevention of colorectal cancer, yet maintaining robustness across heterogeneous colonoscopy dataset remains challenging. This study evaluates the performance and generalisability of the YOLOv9c architecture for colorectal polyp detection, with a focus on the impact of integrating Convolutional Block Attention Modules (CBAM) at different stages of the network. This work provides a systematic comparison of attention placement within the YOLOv9 architecture, highlighting how design choices influence both localisation performance and cross-data generalisation. A YOLOv9c model was optimised on the multi-centre PolypGen dataset and compared against variants incorporating CBAM within the backbone and immediately prior to detection head layer. Performance was assessed using standard object detection metrics alongside qualitative visualization. The optimised YOLOv9c baseline demonstrated strong in-domain and external performance, achieving an [email protected] of 0.887 on PolypGen and 0.898 on Kvasir-SEG. On the external Kvasir-SEG dataset, the backbone-integrated CBAM variant achieved the highest [email protected] of 0.912. In contrast, the detection-head variant showed less consistent localisation performance across evaluation metrics. These findings suggest that YOLOv9c is well-suited for polyp detection and that attention mechanisms are more effective when applied during early feature extraction rather than at later detection stages. Computer-aided diagnosis Deep learning Colorectal polyp detection YOLOv9 Attention mechanism 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-9628010","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":638110914,"identity":"9ef7dcb0-8f5a-4faa-a0ae-7ad62e05d115","order_by":0,"name":"Shivangi Krishna","email":"","orcid":"","institution":"The University of Sheffield","correspondingAuthor":false,"prefix":"","firstName":"Shivangi","middleName":"","lastName":"Krishna","suffix":""},{"id":638110915,"identity":"d6037d4d-554e-4a06-8052-1f597c6f985c","order_by":1,"name":"Soren Salehi","email":"","orcid":"","institution":"Sharif University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Soren","middleName":"","lastName":"Salehi","suffix":""},{"id":638110916,"identity":"b1a35e02-7446-4df4-83ef-fd3fd23a55aa","order_by":2,"name":"Noha Ghatwary","email":"","orcid":"","institution":"Arab Academy for Science, Technology and Maritime Transport","correspondingAuthor":false,"prefix":"","firstName":"Noha","middleName":"","lastName":"Ghatwary","suffix":""},{"id":638110917,"identity":"633cb3e9-4260-4de7-849d-67717ede0cb7","order_by":3,"name":"Omar O. Ebada","email":"","orcid":"","institution":"Arab Academy for Science, Technology and Maritime Transport","correspondingAuthor":false,"prefix":"","firstName":"Omar","middleName":"O.","lastName":"Ebada","suffix":""},{"id":638110918,"identity":"4d993791-464f-4d4e-a555-50facf17ab50","order_by":4,"name":"Neda Azarmehr","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABEUlEQVRIie2RMUvDQBiGv3JwXb7U9Uok+QtfKNh26W+50FkQXByKpgTSRTr7MwJCJocLhbqcuBYyScDJIRIQhFI8OyeNo8M93PTCw/e+HIDF8p8hYIoBeGccWCSOEXYqXBplNEyg92cF6ahEXcoYWFl9P2Xe2NV1fXVDYtBfLnewmAFp1ahMI07B/Xsxmq4vM/dBk+CYxxPYzoFeouY+CkiiKsJUOxlzkv0tF2EigJv8tWWC6lf5XhV3qcaydg7miv9mlMMpBYPYXJGkEVwnMoroJeap9mIbvGbnqghSzS9c3P5uCeNJuJ7jsGU+Pa8ePz9U4ZNmZY0LEv5qk++qr5k30LK5GWtMZfdHWiwWi+UEPxfvXAPgwLbBAAAAAElFTkSuQmCC","orcid":"","institution":"The University of Sheffield","correspondingAuthor":true,"prefix":"","firstName":"Neda","middleName":"","lastName":"Azarmehr","suffix":""}],"badges":[],"createdAt":"2026-05-06 09:23:35","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9628010/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9628010/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109076912,"identity":"55c7aa7a-ec48-483b-a224-98bafcb45c78","added_by":"auto","created_at":"2026-05-12 11:05:34","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5836869,"visible":true,"origin":"","legend":"","description":"","filename":"IntegratingCBAMintoYOLOv9cforPolypDetectionASystematicCrossDatasetEvaluation.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9628010/v1_covered_d0e1fa97-c97c-4362-915f-e75a1cc339ac.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Integrating CBAM into YOLOv9c for Polyp Detection: A Systematic Cross-Dataset Evaluation","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"Computer-aided diagnosis, Deep learning, Colorectal polyp detection, YOLOv9, Attention mechanism","lastPublishedDoi":"10.21203/rs.3.rs-9628010/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9628010/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAccurate and real-time detection of colorectal polyps is important for early diagnosis and prevention of colorectal cancer, yet maintaining robustness across heterogeneous colonoscopy dataset remains challenging. This study evaluates the performance and generalisability of the YOLOv9c architecture for colorectal polyp detection, with a focus on the impact of integrating Convolutional Block Attention Modules (CBAM) at different stages of the network. This work provides a systematic comparison of attention placement within the YOLOv9 architecture, highlighting how design choices influence both localisation performance and cross-data generalisation. A YOLOv9c model was optimised on the multi-centre PolypGen dataset and compared against variants incorporating CBAM within the backbone and immediately prior to detection head layer. Performance was assessed using standard object detection metrics alongside qualitative visualization. The optimised YOLOv9c baseline demonstrated strong in-domain and external performance, achieving an [email protected] of 0.887 on PolypGen and 0.898 on Kvasir-SEG. On the external Kvasir-SEG dataset, the backbone-integrated CBAM variant achieved the highest [email protected] of 0.912. In contrast, the detection-head variant showed less consistent localisation performance across evaluation metrics. These findings suggest that YOLOv9c is well-suited for polyp detection and that attention mechanisms are more effective when applied during early feature extraction rather than at later detection stages.\u003c/p\u003e","manuscriptTitle":"Integrating CBAM into YOLOv9c for Polyp Detection: A Systematic Cross-Dataset Evaluation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-12 10:29:07","doi":"10.21203/rs.3.rs-9628010/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":"a6032870-ab57-41bd-941b-12d592f70efd","owner":[],"postedDate":"May 12th, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvited","content":"","date":"2026-05-10T09:27:45+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-05-08T11:10:42+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-05-08T11:09:44+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2026-05-06T09:11:37+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-12T10:29:08+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-12 10:29:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9628010","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9628010","identity":"rs-9628010","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}