Supervised vs. Unsupervised GAN for Pseudo-CT Synthesis in Brain MR-Guided Radiotherapy

Supervised vs. Unsupervised GAN for Pseudo-CT Synthesis in Brain MR-Guided Radiotherapy | 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 Supervised vs. Unsupervised GAN for Pseudo-CT Synthesis in Brain MR-Guided Radiotherapy Milad Zeinali Kermani, Mohammad Bagher Tavakoli, Amir Khorasani, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6170896/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 22 Jul, 2025 Read the published version in Physical and Engineering Sciences in Medicine → Version 1 posted 6 You are reading this latest preprint version Abstract Purpose Radiotherapy is a crucial treatment for brain tumor malignancies. To address the limitations of CT-based treatment planning, recent research has explored MR-only radiotherapy, requiring precise MR-to-CT synthesis. This study compares two deep learning approaches, supervised (Pix2Pix) and unsupervised (CycleGAN), for generating pseudo-CT (pCT) images from T1- and T2-weighted MR sequences. Methods and Materials: 3270 paired T1 and T2 weighted MRI images are collected and registered with corresponding CT images. After preprocessing a supervised pCT generative model was trained using a "pix2pix" model, and an unsupervised generative network (CycleGan), was also trained for the purpose of comparing pCT quality against the pix2pix. To assess the differences between pCT images and reference CT images, three key metrics (SSIM, PSNR and MAE) are used. Results The average of SSIM, PSNR and MAE for pix2pix on T1 images was 0.964 ± 0.03, 32.812 ± 5.21 and 79.681 ± 9.52 HU respectively. Statistical analysis revealed that Pix2Pix significantly outperformed CycleGAN in generating high-fidelity pCT images (p 0.05). Conclusion Both supervised and unsupervised methods demonstrated the capability to generate accurate pCT images from conventional T1- and T2-weighted MR sequences. While supervised methods like Pix2Pix achieve higher accuracy, unsupervised approaches such as CycleGAN offer greater flexibility by eliminating the need for paired training data, making them suitable for applications where paired data is unavailable. Deep learning radiotherapy treatment planning Generative Adversarial Networks synthetic CT Full Text Cite Share Download PDF Status: Published Journal Publication published 22 Jul, 2025 Read the published version in Physical and Engineering Sciences in Medicine → Version 1 posted Editorial decision: Major revisions 30 Apr, 2025 Reviewers agreed at journal 27 Mar, 2025 Reviewers invited by journal 26 Mar, 2025 Editor invited by journal 23 Mar, 2025 Editor assigned by journal 07 Mar, 2025 First submitted to journal 06 Mar, 2025 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-6170896","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":434119445,"identity":"ef19575a-4278-45a5-a605-e3220a464f0f","order_by":0,"name":"Milad Zeinali Kermani","email":"","orcid":"","institution":"Isfahan University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Milad","middleName":"Zeinali","lastName":"Kermani","suffix":""},{"id":434119446,"identity":"250d5048-cfdd-40fc-95c0-d83ab16ed01c","order_by":1,"name":"Mohammad Bagher Tavakoli","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIiWNgGAWjYPCCBAY+9gYGBh4gkw0sYIBPNTNECxvPAZK1SCRAtBAEug38Bz/z1KTJsUm+TnzwhmFbYh8D88MPDAX3cGoxO8DMLM1zLMeYTTp3s+EchtuJbQxsxhIMBsX4tDBIzmCrSGyTzt0mzQPWwmAG9EsCXlt+zvgH1CJ5dvtviBb2b4S0sEl8bMtJbJPg3cYM0cJDwJbDzGYWH/vSjNl4cjdLzjG4bdzGzFMskYBPy/HGxzcSviXL8bOf3fjhTcVt2fnt7Rs/fPiDWwskWuDAACqCR8MoGAWjYBSMAiIAAJo/SHG3vWaEAAAAAElFTkSuQmCC","orcid":"","institution":"Isfahan University of Medical Sciences","correspondingAuthor":true,"prefix":"","firstName":"Mohammad","middleName":"Bagher","lastName":"Tavakoli","suffix":""},{"id":434119447,"identity":"c6a786ac-5570-47a6-9a14-903a5f66a0ef","order_by":2,"name":"Amir Khorasani","email":"","orcid":"","institution":"Isfahan University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Amir","middleName":"","lastName":"Khorasani","suffix":""},{"id":434119448,"identity":"79c47148-16a0-4a28-817c-ffa409959bfc","order_by":3,"name":"Iraj Abedi","email":"","orcid":"","institution":"Isfahan University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Iraj","middleName":"","lastName":"Abedi","suffix":""},{"id":434119449,"identity":"269d7d06-d0f4-4931-8714-86c78bce1b6c","order_by":4,"name":"Vahid Sadeghi","email":"","orcid":"","institution":"Isfahan University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Vahid","middleName":"","lastName":"Sadeghi","suffix":""},{"id":434119450,"identity":"1c8e8653-feea-42c7-b2e8-dcf445198c6c","order_by":5,"name":"Alireza Amouheidari","email":"","orcid":"","institution":"Isfahan University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Alireza","middleName":"","lastName":"Amouheidari","suffix":""}],"badges":[],"createdAt":"2025-03-06 13:18:27","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6170896/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6170896/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s13246-025-01606-1","type":"published","date":"2025-07-22T15:57:37+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":87756801,"identity":"130c24fb-ffaa-445b-860a-e70f54e1ccff","added_by":"auto","created_at":"2025-07-28 16:09:28","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":696394,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6170896/v1_covered_17aa3953-2bce-4c2d-b30d-c8e41cbf6d47.pdf"}],"financialInterests":"","formattedTitle":"Supervised vs. Unsupervised GAN for Pseudo-CT Synthesis in Brain MR-Guided Radiotherapy","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"physical-and-engineering-sciences-in-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"apes","sideBox":"Learn more about [Physical and Engineering Sciences in Medicine](http://link.springer.com/journal/13246)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/apes/default.aspx","title":"Physical and Engineering Sciences in Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Deep learning, radiotherapy, treatment planning, Generative Adversarial Networks, synthetic CT","lastPublishedDoi":"10.21203/rs.3.rs-6170896/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6170896/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eRadiotherapy is a crucial treatment for brain tumor malignancies. To address the limitations of CT-based treatment planning, recent research has explored MR-only radiotherapy, requiring precise MR-to-CT synthesis. This study compares two deep learning approaches, supervised (Pix2Pix) and unsupervised (CycleGAN), for generating pseudo-CT (pCT) images from T1- and T2-weighted MR sequences.\u003c/p\u003e\u003ch2\u003eMethods and Materials:\u003c/h2\u003e \u003cp\u003e3270 paired T1 and T2 weighted MRI images are collected and registered with corresponding CT images. After preprocessing a supervised pCT generative model was trained using a \"pix2pix\" model, and an unsupervised generative network (CycleGan), was also trained for the purpose of comparing pCT quality against the pix2pix. To assess the differences between pCT images and reference CT images, three key metrics (SSIM, PSNR and MAE) are used.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe average of SSIM, PSNR and MAE for pix2pix on T1 images was 0.964\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03, 32.812\u0026thinsp;\u0026plusmn;\u0026thinsp;5.21 and 79.681\u0026thinsp;\u0026plusmn;\u0026thinsp;9.52 HU respectively. Statistical analysis revealed that Pix2Pix significantly outperformed CycleGAN in generating high-fidelity pCT images (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). There was no notable difference in the effectiveness of T1-weighted versus T2-weighted MR images for generating pCT (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eBoth supervised and unsupervised methods demonstrated the capability to generate accurate pCT images from conventional T1- and T2-weighted MR sequences. While supervised methods like Pix2Pix achieve higher accuracy, unsupervised approaches such as CycleGAN offer greater flexibility by eliminating the need for paired training data, making them suitable for applications where paired data is unavailable.\u003c/p\u003e","manuscriptTitle":"Supervised vs. Unsupervised GAN for Pseudo-CT Synthesis in Brain MR-Guided Radiotherapy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-04 12:58:31","doi":"10.21203/rs.3.rs-6170896/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Major revisions","date":"2025-04-30T06:45:43+00:00","index":"","fulltext":""},{"type":"reviewerAgreed","content":"","date":"2025-03-27T07:38:26+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-03-26T07:03:02+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"Physical and Engineering Sciences in Medicine","date":"2025-03-24T03:31:06+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-03-07T08:20:09+00:00","index":"","fulltext":""},{"type":"submitted","content":"Physical and Engineering Sciences in Medicine","date":"2025-03-06T08:17:36+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"physical-and-engineering-sciences-in-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"apes","sideBox":"Learn more about [Physical and Engineering Sciences in Medicine](http://link.springer.com/journal/13246)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/apes/default.aspx","title":"Physical and Engineering Sciences in Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"eb2cd56a-e099-463e-8a66-6113d5920d56","owner":[],"postedDate":"April 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-07-28T16:05:43+00:00","versionOfRecord":{"articleIdentity":"rs-6170896","link":"https://doi.org/10.1007/s13246-025-01606-1","journal":{"identity":"physical-and-engineering-sciences-in-medicine","isVorOnly":false,"title":"Physical and Engineering Sciences in Medicine"},"publishedOn":"2025-07-22 15:57:37","publishedOnDateReadable":"July 22nd, 2025"},"versionCreatedAt":"2025-04-04 12:58:31","video":"","vorDoi":"10.1007/s13246-025-01606-1","vorDoiUrl":"https://doi.org/10.1007/s13246-025-01606-1","workflowStages":[]},"version":"v1","identity":"rs-6170896","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6170896","identity":"rs-6170896","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}