Detecting ovarian endometriomas from ultrasound using vision transformers and cross-modality transfer learning

Detecting ovarian endometriomas from ultrasound using vision transformers and cross-modality transfer learning | 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 Detecting ovarian endometriomas from ultrasound using vision transformers and cross-modality transfer learning Matthew Watson, Miliani Fraser-Fletcher, Tom Willshare, Molly Jowsey, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-10284032/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 Endometriosis is a chronic condition that affects around 10% of reproductive age women worldwide. It requires lifelong management and, due to symptom overlap with other gynaecological conditions and the traditional requirement for invasive laparoscopy for diagnosis, there are often severe delays between becoming symptomatic and diagnosis. During this time patients often experience debilitating symptoms that can have a severe effect on their quality of life. Recently, standardised ultrasound approaches have been developed to aid diagnosis without the need for laparoscopy. However, these rely on trained specialists to acquire and analyse the ultrasound images-leading to bottlenecks in clinical pathways. In this paper, we present the first machine learning (ML) model trained on publicly available ultrasound data for endometrioma detection, demonstrating high performance (average precision: 0.81) despite heavily imbalanced data. Acknowledging that the limited amount of publicly available ultrasound data is hampering progress in this area, we investigate transfer learning using other (non-ultrasound) radiographic imaging modalities and find that this substantially improves endometriosis detection performance, despite the remaining significant data shift to ultrasound. This extends existing transfer learning literature in medical imaging, evidencing that medical imaging (non-ultrasound) models must learn low-level feature/pattern recognition that are applicable to unseen imaging modalities and body parts. Using SHAP values, we perform exploratory analyses of the features learned by our models and find no evidence of shortcut learning or bias. This study pioneers the use of open-source data and models in this domain, and we hope it motivates future collaborative, open-source efforts to refine and evaluate machine learning approaches for endometriosis. Biological sciences/Computational biology and bioinformatics Health sciences/Diseases Health sciences/Health care Health sciences/Medical research 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-10284032","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":675186253,"identity":"9ebe0bad-d316-48ed-b255-10dca23bba2b","order_by":0,"name":"Matthew Watson","email":"data:image/png;base64,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","orcid":"","institution":"Durham University","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Matthew","middleName":"","lastName":"Watson","suffix":""},{"id":675186254,"identity":"7457187c-4206-45f6-be4d-72859c1eb0ce","order_by":1,"name":"Miliani Fraser-Fletcher","email":"","orcid":"","institution":"Durham University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Miliani","middleName":"","lastName":"Fraser-Fletcher","suffix":""},{"id":675186255,"identity":"8491129e-d4e8-4a39-a5e8-478b5f67db87","order_by":2,"name":"Tom Willshare","email":"","orcid":"","institution":"Revela Health Limited","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Tom","middleName":"","lastName":"Willshare","suffix":""},{"id":675186256,"identity":"d68b85c5-f447-4095-a59b-c19101d79f88","order_by":3,"name":"Molly Jowsey","email":"","orcid":"","institution":"Revela Health Limited","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Molly","middleName":"","lastName":"Jowsey","suffix":""},{"id":675186257,"identity":"feb45fe6-867b-4ca4-914b-44aca0be35c6","order_by":4,"name":"Noura Al Moubayed","email":"","orcid":"","institution":"Durham University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Noura","middleName":"Al","lastName":"Moubayed","suffix":""}],"badges":[],"createdAt":"2026-07-07 19:23:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-10284032/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-10284032/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":114639510,"identity":"40241a35-b46a-417d-9dac-03d8361ec1d4","added_by":"auto","created_at":"2026-07-14 05:44:27","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":986742,"visible":true,"origin":"","legend":"","description":"","filename":"latex.pdf","url":"https://assets-eu.researchsquare.com/files/rs-10284032/v1_covered_89952328-6222-4255-9422-870c8fa5a861.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Detecting ovarian endometriomas from ultrasound using vision transformers and cross-modality transfer learning","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":"","lastPublishedDoi":"10.21203/rs.3.rs-10284032/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-10284032/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Endometriosis is a chronic condition that affects around 10% of reproductive age women worldwide. It requires lifelong management and, due to symptom overlap with other gynaecological conditions and the traditional requirement for invasive laparoscopy for diagnosis, there are often severe delays between becoming symptomatic and diagnosis. During this time patients often experience debilitating symptoms that can have a severe effect on their quality of life. Recently, standardised ultrasound approaches have been developed to aid diagnosis without the need for laparoscopy. However, these rely on trained specialists to acquire and analyse the ultrasound images-leading to bottlenecks in clinical pathways. In this paper, we present the first machine learning (ML) model trained on publicly available ultrasound data for endometrioma detection, demonstrating high performance (average precision: 0.81) despite heavily imbalanced data. Acknowledging that the limited amount of publicly available ultrasound data is hampering progress in this area, we investigate transfer learning using other (non-ultrasound) radiographic imaging modalities and find that this substantially improves endometriosis detection performance, despite the remaining significant data shift to ultrasound. This extends existing transfer learning literature in medical imaging, evidencing that medical imaging (non-ultrasound) models must learn low-level feature/pattern recognition that are applicable to unseen imaging modalities and body parts. Using SHAP values, we perform exploratory analyses of the features learned by our models and find no evidence of shortcut learning or bias. This study pioneers the use of open-source data and models in this domain, and we hope it motivates future collaborative, open-source efforts to refine and evaluate machine learning approaches for endometriosis.","manuscriptTitle":"Detecting ovarian endometriomas from ultrasound using vision transformers and cross-modality transfer learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-07-13 07:00:29","doi":"10.21203/rs.3.rs-10284032/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":"f40d9e06-18f4-49ef-8b04-556af946f94d","owner":[],"postedDate":"July 13th, 2026","published":true,"recentEditorialEvents":[{"type":"editorAssigned","content":"","date":"2026-07-13T05:02:18+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-07-09T12:24:39+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Women's Health","date":"2026-07-07T19:07:27+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":71800906,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":71800907,"name":"Health sciences/Diseases"},{"id":71800908,"name":"Health sciences/Health care"},{"id":71800909,"name":"Health sciences/Medical research"}],"tags":[],"updatedAt":"2026-07-13T07:00:29+00:00","versionOfRecord":[],"versionCreatedAt":"2026-07-13 07:00:29","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-10284032","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-10284032","identity":"rs-10284032","version":["v1"]},"buildId":"7rjqhiLT3MXkJMwkYKINL","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}