Prediction of PD-L1 and EBV expression in gastric cancer from H&E-stained Whole Slide Imaging using deep learning

Prediction of PD-L1 and EBV expression in gastric cancer from H&E-stained Whole Slide Imaging using deep 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 Prediction of PD-L1 and EBV expression in gastric cancer from H&E-stained Whole Slide Imaging using deep learning Tao Wang, Hejun Zhang, Junlin Lan, Zhida Wu, Xinlin Zhang, Jianchao Wang, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5607860/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 PD-L1 inhibitors are more effective for patients with CPS ≥ 5 or EBV positivity in gastric cancer. However, determining PD-L1 and EBV expression requires costly and complex immunohistochemistry staining and in situ hybridization techniques. Therefore, there is a strong need to develop an AI system to assess PD-L1 and EBV expressions using H&E-stained slides. We developed an AI system based on H&E-stained slides, including 432 cases collected from Fujian Cancer Hospital, and 258 cases from the TCGA-STAD dataset. The backbone network was trained using self-supervised learning to enhance feature extraction and was combined with an attention network for slide-level prediction. Additionally, clinical information was integrated to improve accuracy. The proposed model demonstrated high performance, with AUCs of 0.860 for PD-L1, 0.962 for EBV, and 0.899 for the TCGA-STAD dataset. Furthermore, validation on biopsy slides yielded AUCs of 0.757 for PD-L1 and 0.918 for EBV. Our study confirmed that the proposed model can accurately predict PD-L1 and EBV expressions in gastric cancer using H&E-stained slides. The model trained on surgical specimens shows good generalization to biopsy samples. Additionally, incorporating self-supervised learning could enhance the model's feature extraction capabilities, thereby improving its robustness. Biological sciences/Cancer/Cancer therapy/Cancer immunotherapy Physical sciences/Engineering/Biomedical engineering Health sciences/Oncology/Cancer/Gastrointestinal cancer Health sciences/Oncology/Cancer/Tumour biomarkers Biological sciences/Computational biology and bioinformatics/Machine learning PD-L1 EBV Gastric Cancer Deep Learning 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-5607860","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":392420357,"identity":"0703908c-b8fa-4ee8-97d3-999a6f226217","order_by":0,"name":"Tao Wang","email":"","orcid":"","institution":"Fuzhou University","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Wang","suffix":""},{"id":392420358,"identity":"a625d12c-730a-4b9e-aeb0-5cabe78f43bc","order_by":1,"name":"Hejun Zhang","email":"","orcid":"","institution":"Fujian Provincial Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Hejun","middleName":"","lastName":"Zhang","suffix":""},{"id":392420361,"identity":"5b5fed1e-1fd9-4c1f-b365-728643a94925","order_by":2,"name":"Junlin Lan","email":"","orcid":"","institution":"Fuzhou University","correspondingAuthor":false,"prefix":"","firstName":"Junlin","middleName":"","lastName":"Lan","suffix":""},{"id":392420363,"identity":"ded8432a-b621-43b8-9429-3f26c73feb84","order_by":3,"name":"Zhida Wu","email":"","orcid":"","institution":"Fujian Provincial Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Zhida","middleName":"","lastName":"Wu","suffix":""},{"id":392420369,"identity":"0012c1b4-a0a4-42ec-a00f-e4f1b4298f05","order_by":4,"name":"Xinlin Zhang","email":"","orcid":"","institution":"Fuzhou University","correspondingAuthor":false,"prefix":"","firstName":"Xinlin","middleName":"","lastName":"Zhang","suffix":""},{"id":392420374,"identity":"e651126d-f233-4e77-8771-8f047d9e9b63","order_by":5,"name":"Jianchao Wang","email":"","orcid":"","institution":"Fujian Provincial Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jianchao","middleName":"","lastName":"Wang","suffix":""},{"id":392420375,"identity":"29b9b7d1-b559-427d-ac39-75137b01ebdf","order_by":6,"name":"Yuanbo Zhou","email":"","orcid":"","institution":"Fuzhou University","correspondingAuthor":false,"prefix":"","firstName":"Yuanbo","middleName":"","lastName":"Zhou","suffix":""},{"id":392420377,"identity":"14fb6d4a-f5d0-4b32-9afb-ee2d8640a441","order_by":7,"name":"Yuanbin Chen","email":"","orcid":"","institution":"Fuzhou University","correspondingAuthor":false,"prefix":"","firstName":"Yuanbin","middleName":"","lastName":"Chen","suffix":""},{"id":392420379,"identity":"f118d7b7-df3c-4e46-9024-772b903768de","order_by":8,"name":"Longxuan Zhao","email":"","orcid":"","institution":"Fuzhou University","correspondingAuthor":false,"prefix":"","firstName":"Longxuan","middleName":"","lastName":"Zhao","suffix":""},{"id":392420380,"identity":"db137002-c1d8-4f9b-aefa-c926f6224c89","order_by":9,"name":"Tao Tan","email":"","orcid":"","institution":"Macao Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Tan","suffix":""},{"id":392420381,"identity":"fe118f4d-f18a-4a3a-825e-05644b7dfcd6","order_by":10,"name":"Gang Chen","email":"","orcid":"","institution":"Fujian Provincial Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Gang","middleName":"","lastName":"Chen","suffix":""},{"id":392420382,"identity":"6782f6d5-e0d5-457f-8068-c84047c41863","order_by":11,"name":"Tong Tong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3klEQVRIiWNgGAWjYPACCxDB+ABI8PARUsvDBqYkJIAEswFCgEgtbCCCgaAWe/neh595KiTq+KXbr1V+zbGTYWNgfvjoBl5b2I2lec5ISEjOOVN2W3ZbMtBaNmPjHLxa2BgkZ7ZJSBjcyEm7LbmNGaiFh02agBbmnzP/QbQUS26rJ0oLm8THBpCW9GOMH7cdJkLLsTQ2iw/HJCRnzshhlmbcdpyHjZmAX9ibjzHfSKix4eeXSH/48ee2ant+9uaHj/FpQbbQgJkHRDMTpxxs4QPGH8SrHgWjYBSMghEEAP0rO0Sn/V3dAAAAAElFTkSuQmCC","orcid":"","institution":"Fuzhou University","correspondingAuthor":true,"prefix":"","firstName":"Tong","middleName":"","lastName":"Tong","suffix":""}],"badges":[],"createdAt":"2024-12-09 09:53:35","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5607860/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5607860/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":76680914,"identity":"04cf5adc-a638-4902-93c3-fc0b53d29208","added_by":"auto","created_at":"2025-02-19 15:09:42","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3126213,"visible":true,"origin":"","legend":"","description":"","filename":"submitSR.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5607860/v1_covered_cb2ed347-81e0-418e-9b43-32205feb5a39.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Prediction of PD-L1 and EBV expression in gastric cancer from H\u0026E-stained Whole Slide Imaging using deep learning","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":"PD-L1, EBV, Gastric Cancer, Deep Learning","lastPublishedDoi":"10.21203/rs.3.rs-5607860/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5607860/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"PD-L1 inhibitors are more effective for patients with CPS ≥ 5 or EBV positivity in gastric cancer. However, determining PD-L1 and EBV expression requires costly and complex immunohistochemistry staining and in situ hybridization techniques. Therefore, there is a strong need to develop an AI system to assess PD-L1 and EBV expressions using H\u0026E-stained slides. We developed an AI system based on H\u0026E-stained slides, including 432 cases collected from Fujian Cancer Hospital, and 258 cases from the TCGA-STAD dataset. The backbone network was trained using self-supervised learning to enhance feature extraction and was combined with an attention network for slide-level prediction. Additionally, clinical information was integrated to improve accuracy. The proposed model demonstrated high performance, with AUCs of 0.860 for PD-L1, 0.962 for EBV, and 0.899 for the TCGA-STAD dataset. Furthermore, validation on biopsy slides yielded AUCs of 0.757 for PD-L1 and 0.918 for EBV. Our study confirmed that the proposed model can accurately predict PD-L1 and EBV expressions in gastric cancer using H\u0026E-stained slides. The model trained on surgical specimens shows good generalization to biopsy samples. Additionally, incorporating self-supervised learning could enhance the model's feature extraction capabilities, thereby improving its robustness.","manuscriptTitle":"Prediction of PD-L1 and EBV expression in gastric cancer from H\u0026amp;E-stained Whole Slide Imaging using deep learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-20 17:06:44","doi":"10.21203/rs.3.rs-5607860/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":"bc5ef1c6-fe58-4c3c-b9cf-1ce2fdc70007","owner":[],"postedDate":"December 20th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":41831565,"name":"Biological sciences/Cancer/Cancer therapy/Cancer immunotherapy"},{"id":41831566,"name":"Physical sciences/Engineering/Biomedical engineering"},{"id":41831567,"name":"Health sciences/Oncology/Cancer/Gastrointestinal cancer"},{"id":41831568,"name":"Health sciences/Oncology/Cancer/Tumour biomarkers"},{"id":41831569,"name":"Biological sciences/Computational biology and bioinformatics/Machine learning"}],"tags":[],"updatedAt":"2025-02-19T14:53:12+00:00","versionOfRecord":[],"versionCreatedAt":"2024-12-20 17:06:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5607860","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5607860","identity":"rs-5607860","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}