IRMA: Machine learning-based harmonization of 18F-FDG PET brain scans in multi-center studies

IRMA: Machine learning-based harmonization of 18F-FDG PET brain scans in multi-center studies | 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 Method Article IRMA: Machine learning-based harmonization of 18 F-FDG PET brain scans in multi-center studies Sofie Lövdal, Rick van Veen, Giulia Carli, Remco Renken, Tamara Shiner, and 10 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5825843/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 17 Feb, 2025 Read the published version in European Journal of Nuclear Medicine and Molecular Imaging → Version 1 posted You are reading this latest preprint version Abstract Purpose: Center-specific effects in PET brain scans arise due to differences in technical and procedural aspects. This restricts the merging of data between centers and introduces source-specific bias. Methods: We demonstrate the use of the recently proposed machine learning method Iterated Relevance Matrix Analysis (IRMA) for harmonization of center-specific effects in brain 18 F-Fluorodeoxyglucose ( 18 F-FDG) PET scans. The center difference is learned by applying IRMA on PCA-based feature vectors of healthy controls (HC), resulting in a subspace V, representing information not comparable between centers, and the remaining subspace U, where no center differences are present. In this proof-of-concept study, we demonstrate the properties of the method using data from four centers. After center-harmonization, a Generalized Matrix Learning Vector Quantization (GMLVQ) model was trained to discriminate between Parkinson’s disease, Alzheimer’s disease and Dementia with Lewy Bodies. Results: At the initial IRMA iteration, the system was able to determine the center origin of the four HC cohorts almost perfectly. The method required six iterations, corresponding to a six-dimensional subspace V, to determine the entire center difference. An uncorrected disease classification model was highly biased to center-specific effects, creating a falsely inflated performance when applying internal (cross-) validation. The cross-validation performance of the center-harmonized model remained high, while it generalized significantly better to unseen test cohorts. Furthermore, the framework is highly transparent, providing analytic reconstructions of the correction and visualizations of the data in voxel space. Conclusion: IRMA can be used to learn and disregard center-specific information in features extracted from brain 18 F-FDG PET scans, while retaining disease-specific information. Nuclear Medicine & Medical Imaging Artificial Intelligence and Machine Learning 18F-FDG PET neuroimaging site effect harmonization machine learning neurodegeneration Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Published Journal Publication published 17 Feb, 2025 Read the published version in European Journal of Nuclear Medicine and Molecular Imaging → 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-5825843","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Method Article","associatedPublications":[],"authors":[{"id":401932758,"identity":"ed79d3b1-9647-4ae3-8caa-facea3f95c99","order_by":0,"name":"Sofie Lövdal","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIiWNgGAWjYLCCBwwMjA3szAfAHD6itCSAtDCzJYA5bCRo4TEgTot8A/sDhoSKw7L9zTzfPvP8YsgjqMXgANDwhDOHjWcc5t08m7ePoZiwFgYeBobEtrTEBqAWZt4eIJsoh4G0zD/M85g4LQwHgBYlttkkbjjMw8zM84MILQaHeQwOJJyxMd54mM2YcW6DBBEOa29/+OBDhYTsvOPNjxne/LFJ7CfoMGaQ22CAsU2CoAZ08IdkHaNgFIyCUTACAAC8Szn/lbVrDgAAAABJRU5ErkJggg==","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Sofie","middleName":"","lastName":"Lövdal","suffix":""},{"id":401932759,"identity":"23050112-5255-4f7c-879a-fe9df7089a22","order_by":1,"name":"Rick van Veen","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Rick","middleName":"van","lastName":"Veen","suffix":""},{"id":401932760,"identity":"b4af009b-c2a8-40d6-ac95-da6489fdfd28","order_by":2,"name":"Giulia Carli","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Giulia","middleName":"","lastName":"Carli","suffix":""},{"id":401932761,"identity":"0f0801ee-eb38-4b5c-9f5f-ad72dcdcbdc0","order_by":3,"name":"Remco Renken","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Remco","middleName":"","lastName":"Renken","suffix":""},{"id":401932907,"identity":"a7417072-85ef-49d7-b7be-4350d0e470e5","order_by":4,"name":"Tamara Shiner","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Tamara","middleName":"","lastName":"Shiner","suffix":""},{"id":401932762,"identity":"2d8a8480-9b8f-4310-961c-540b8d27d547","order_by":5,"name":"Noa Bregman","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Noa","middleName":"","lastName":"Bregman","suffix":""},{"id":401932763,"identity":"cc44a42a-9a7f-4f17-beab-470ad78d7ce9","order_by":6,"name":"Rotem Orad","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Rotem","middleName":"","lastName":"Orad","suffix":""},{"id":401932764,"identity":"fe95d1b4-68fc-4425-88cb-1b9653b47990","order_by":7,"name":"Dario Arnaldi","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Dario","middleName":"","lastName":"Arnaldi","suffix":""},{"id":401932765,"identity":"06342199-41e9-4e2b-aeff-77eaef20167c","order_by":8,"name":"Beatrice Orso","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Beatrice","middleName":"","lastName":"Orso","suffix":""},{"id":401932766,"identity":"6d4622f7-9a2a-4178-8538-fd79d66c550b","order_by":9,"name":"Silvia Morbelli","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Silvia","middleName":"","lastName":"Morbelli","suffix":""},{"id":401932767,"identity":"f35f1aa0-7a3e-4455-9368-684c04c30121","order_by":10,"name":"Pietro Mattioli","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Pietro","middleName":"","lastName":"Mattioli","suffix":""},{"id":401932768,"identity":"52def9a6-481b-46c1-8d29-260501686225","order_by":11,"name":"Klaus Leenders","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Klaus","middleName":"","lastName":"Leenders","suffix":""},{"id":401932769,"identity":"3fc47239-4096-4253-a698-d85efe76b321","order_by":12,"name":"Rudi Dierckx","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Rudi","middleName":"","lastName":"Dierckx","suffix":""},{"id":401932770,"identity":"b64f7568-1bac-4340-a6ee-d2f9a7403747","order_by":13,"name":"Sanne Meles","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Sanne","middleName":"","lastName":"Meles","suffix":""},{"id":401932771,"identity":"5768efd2-d789-41ac-aa56-5f95c5b15244","order_by":14,"name":"Michael Biehl","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"Biehl","suffix":""}],"badges":[],"createdAt":"2025-01-14 09:37:08","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-5825843/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5825843/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00259-025-07114-4","type":"published","date":"2025-02-18T00:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":81990171,"identity":"dead2de9-761c-47f3-bbe9-2460d3aa4740","added_by":"auto","created_at":"2025-05-05 16:25:01","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4396938,"visible":true,"origin":"","legend":"","description":"","filename":"cleanmanuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5825843/v1_covered_082c401d-0240-4627-a0ef-d2c15b679a93.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eIRMA: Machine learning-based harmonization of \u003csup\u003e18\u003c/sup\u003eF-FDG PET brain scans in multi-center studies\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[{"identity":"6011dc70-b4f2-4b59-9c28-a30fde3344d1","identifier":"10.13039/501100008383","name":"Stichting ParkinsonFonds","awardNumber":"2022/1891","order_by":0}],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"University Medical Center Groningen","isAcceptedByJournal":true,"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":"18F-FDG PET, neuroimaging, site effect, harmonization, machine learning, neurodegeneration","lastPublishedDoi":"10.21203/rs.3.rs-5825843/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5825843/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePurpose: Center-specific effects in PET brain scans arise due to differences in technical and procedural aspects. \u0026nbsp;This restricts the merging of data between centers and introduces source-specific bias. Methods: We demonstrate the use of the recently proposed machine learning method Iterated Relevance Matrix Analysis (IRMA) for harmonization of center-specific effects in brain \u003csup\u003e18\u003c/sup\u003eF-Fluorodeoxyglucose (\u003csup\u003e18\u003c/sup\u003eF-FDG) PET scans. The center difference is learned by applying IRMA on PCA-based feature vectors of healthy controls (HC), resulting in a subspace V, representing information not comparable between centers, and the remaining subspace U, where no center differences are present. In this proof-of-concept study, we demonstrate the properties of the method using data from four centers. After center-harmonization, a Generalized Matrix Learning Vector Quantization (GMLVQ) model was trained \u0026nbsp;to discriminate between Parkinson’s disease, Alzheimer’s disease and Dementia with Lewy Bodies. Results: At the initial IRMA iteration, the system was able to determine the center origin of the four HC cohorts almost perfectly. The method required six iterations, corresponding to a six-dimensional subspace V, to determine the entire center difference. An uncorrected disease classification model was highly biased to center-specific effects, creating a falsely inflated performance when applying internal (cross-) validation. The cross-validation performance of the center-harmonized model remained high, while it generalized significantly better to unseen test cohorts. Furthermore, the framework is highly transparent, providing analytic reconstructions of the correction and visualizations \u0026nbsp;of the data in voxel space. Conclusion: IRMA can be used to learn and disregard center-specific information in features extracted from brain \u003csup\u003e18\u003c/sup\u003eF-FDG PET scans, while retaining disease-specific information.\u003c/p\u003e","manuscriptTitle":"IRMA: Machine learning-based harmonization of 18F-FDG PET brain scans in multi-center studies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-15 04:38:53","doi":"10.21203/rs.3.rs-5825843/v1","editorialEvents":[{"type":"communityComments","content":1}],"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":"05572cdb-bf16-4c42-8e98-d117aa05370c","owner":[],"postedDate":"January 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":42836204,"name":"Nuclear Medicine \u0026 Medical Imaging"},{"id":42836205,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2025-05-05T16:24:46+00:00","versionOfRecord":{"articleIdentity":"rs-5825843","link":"https://doi.org/10.1007/s00259-025-07114-4","journal":{"identity":"european-journal-of-nuclear-medicine-and-molecular-imaging","isVorOnly":false,"title":"European Journal of Nuclear Medicine and Molecular Imaging"},"publishedOn":"2025-02-18 00:00:00","publishedOnDateReadable":"February 18th, 2025"},"versionCreatedAt":"2025-01-15 04:38:53","video":"","vorDoi":"10.1007/s00259-025-07114-4","vorDoiUrl":"https://doi.org/10.1007/s00259-025-07114-4","workflowStages":[]},"version":"v1","identity":"rs-5825843","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5825843","identity":"rs-5825843","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}