Advanced Deep Learning Methods for MLC Leaf Error Classification in Fluence Maps of Radiation Therapy

Advanced Deep Learning Methods for MLC Leaf Error Classification in Fluence Maps of Radiation Therapy | 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 Advanced Deep Learning Methods for MLC Leaf Error Classification in Fluence Maps of Radiation Therapy Ju Yeol Shin, Chang Heon Choi, Jung-in Kim, Jong Min Park, Wonjoong Cheon, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8429164/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 Background Multi-leaf collimator (MLC) positioning errors are a major source of delivery uncertainty in modern radiation therapy techniques such as volumetric modulated arc therapy (VMAT) and intensity-modulated radiation therapy (IMRT). Traditional quality assurance (QA) methods, particularly gamma analysis, have limited sensitivity for detecting individual leaf errors and provide only binary pass or fail outcomes without precise localization. Although artificial intelligence (AI)-based approaches have been introduced to improve error detection, most have focused on overall treatment plan assessments rather than identifying leaf-specific deviations, reducing their clinical usefulness for targeted error correction. To address these limitations, this study developed and validated a deep learning method to identify MLC leaf positioning errors with high precision. Methods Treatment plans from forty patients with prostate cancer, each containing two full arcs, were retrospectively analyzed using the Varian High Definition MLC system. Inner leaves numbered twenty-one to forty, each with a width of 2.5 mm, were selected for analysis. Fluence maps were generated from DICOM radiotherapy plan files, and systematic MLC positioning errors ranging from − 5 mm to + 5 mm were introduced, producing 121 error combinations per leaf and a total of 48,400 samples. The dataset was divided into training, validation, and test subsets in an 8:1:1 ratio. Each fluence segment was converted into a two-channel image and used to train a convolutional neural network (CNN) to classify the magnitude and direction of the deviations. Results The model achieved a final test accuracy of 97.21% and maintained consistent performance during cross-validation, detecting over 95% of leaf errors within 1 mm of the true offset. Conclusions This CNN-based framework enables accurate, leaf-specific error identification and has strong potential to enhance the efficiency and reliability of modern radiation therapy QA. Radiation therapy Multi-leaf collimator Quality assurance Deep learning Volumetric modulated arc therapy 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-8429164","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":581909407,"identity":"6e0299e8-ff0d-4dba-bb00-74a0a8aee3f5","order_by":0,"name":"Ju Yeol Shin","email":"","orcid":"","institution":"Paprica Lab. Co., Ltd","correspondingAuthor":false,"prefix":"","firstName":"Ju","middleName":"Yeol","lastName":"Shin","suffix":""},{"id":581909408,"identity":"eecccf9c-5f57-4499-9597-67a072215530","order_by":1,"name":"Chang Heon Choi","email":"","orcid":"","institution":"Paprica Lab. Co., Ltd","correspondingAuthor":false,"prefix":"","firstName":"Chang","middleName":"Heon","lastName":"Choi","suffix":""},{"id":581909409,"identity":"9328086c-4768-465f-bed3-f66076085a8e","order_by":2,"name":"Jung-in Kim","email":"","orcid":"","institution":"Paprica Lab. Co., Ltd","correspondingAuthor":false,"prefix":"","firstName":"Jung-in","middleName":"","lastName":"Kim","suffix":""},{"id":581909410,"identity":"75bb1a05-f149-4035-924a-81ea488ac729","order_by":3,"name":"Jong Min Park","email":"","orcid":"","institution":"Seoul National University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jong","middleName":"Min","lastName":"Park","suffix":""},{"id":581909411,"identity":"53cddb4c-9ad9-40ed-8e5b-2b8c7cc921a3","order_by":4,"name":"Wonjoong Cheon","email":"","orcid":"","institution":"Seoul St. Mary's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Wonjoong","middleName":"","lastName":"Cheon","suffix":""},{"id":581909412,"identity":"831bef5d-1dec-414b-97f3-2f18fa129c6b","order_by":5,"name":"So-Yeon Park","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyUlEQVRIiWNgGAWjYDADfhCRUECKFskGkBYDUrQYHACTxKg8f/iYxMe2O4mbz69O/PDAgEGeX+wAAS030tIkZ7Y9S9x24+1mCaDDDGfOTsCvRXIGj5k0b9thY7MbZzeAtCQY3Cakpf8MRIvxjLObfxClhZ8hB6xFzoC/dxtxtvBLpCVbzjh3WE7iBu82iwQDCcJ+YeM/fPDGh7LDPPz9Zzff/FFhI88vTUALELBIgCkJsEoJgspBgPkDxIkHiFI9CkbBKBgFIxAAAGBmQtun3Fh3AAAAAElFTkSuQmCC","orcid":"","institution":"Veterans Health Service Medical Center","correspondingAuthor":true,"prefix":"","firstName":"So-Yeon","middleName":"","lastName":"Park","suffix":""}],"badges":[],"createdAt":"2025-12-23 02:08:26","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8429164/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8429164/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104397685,"identity":"ab1e6208-20a7-481b-ba40-f479824b67b4","added_by":"auto","created_at":"2026-03-11 11:54:33","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":764896,"visible":true,"origin":"","legend":"","description":"","filename":"20251224ManuscriptRadiationOncologyrevised.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8429164/v1_covered_b6d4dc7b-6b8a-430d-bbf3-24a754b58960.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Advanced Deep Learning Methods for MLC Leaf Error Classification in Fluence Maps of Radiation Therapy","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":"Radiation therapy, Multi-leaf collimator, Quality assurance, Deep learning, Volumetric modulated arc therapy","lastPublishedDoi":"10.21203/rs.3.rs-8429164/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8429164/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eMulti-leaf collimator (MLC) positioning errors are a major source of delivery uncertainty in modern radiation therapy techniques such as volumetric modulated arc therapy (VMAT) and intensity-modulated radiation therapy (IMRT). Traditional quality assurance (QA) methods, particularly gamma analysis, have limited sensitivity for detecting individual leaf errors and provide only binary pass or fail outcomes without precise localization. Although artificial intelligence (AI)-based approaches have been introduced to improve error detection, most have focused on overall treatment plan assessments rather than identifying leaf-specific deviations, reducing their clinical usefulness for targeted error correction. To address these limitations, this study developed and validated a deep learning method to identify MLC leaf positioning errors with high precision.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eTreatment plans from forty patients with prostate cancer, each containing two full arcs, were retrospectively analyzed using the Varian High Definition MLC system. Inner leaves numbered twenty-one to forty, each with a width of 2.5 mm, were selected for analysis. Fluence maps were generated from DICOM radiotherapy plan files, and systematic MLC positioning errors ranging from \u0026minus;\u0026thinsp;5 mm to +\u0026thinsp;5 mm were introduced, producing 121 error combinations per leaf and a total of 48,400 samples. The dataset was divided into training, validation, and test subsets in an 8:1:1 ratio. Each fluence segment was converted into a two-channel image and used to train a convolutional neural network (CNN) to classify the magnitude and direction of the deviations.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe model achieved a final test accuracy of 97.21% and maintained consistent performance during cross-validation, detecting over 95% of leaf errors within 1 mm of the true offset.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThis CNN-based framework enables accurate, leaf-specific error identification and has strong potential to enhance the efficiency and reliability of modern radiation therapy QA.\u003c/p\u003e","manuscriptTitle":"Advanced Deep Learning Methods for MLC Leaf Error Classification in Fluence Maps of Radiation Therapy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-29 09:52:20","doi":"10.21203/rs.3.rs-8429164/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":"b962de10-fc88-4c99-a421-42e1f2008c39","owner":[],"postedDate":"January 29th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-26T14:56:35+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-29 09:52:20","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8429164","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8429164","identity":"rs-8429164","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}