Revolutionizing Corneal Staining Assessment: Advanced Evaluation through Lesion-aware Fine-Grained Knowledge Distillation

Revolutionizing Corneal Staining Assessment: Advanced Evaluation through Lesion-aware Fine-Grained Knowledge Distillation | 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 Revolutionizing Corneal Staining Assessment: Advanced Evaluation through Lesion-aware Fine-Grained Knowledge Distillation Jin Yuan, Yuqing Deng, Pujin Cheng, Ruiwen Xu, Lirong Ling, Hongliang Xue, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4274726/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 23 May, 2025 Read the published version in npj Digital Medicine → Version 1 posted 11 You are reading this latest preprint version Abstract Corneal staining is crucial for evaluating ocular surface diseases, yet existing AI models for CSS (Corneal Staining Score) assessments struggle with detailed lesion identification and lack applicability in real-world clinical settings. Moreover, the output of current AI-assist staining evaluation system only provides categories of grades, leading to potential “plateau” effect, which could misrepresent treatment response in clinical practices. Addressing these gaps, we developed the Fine-grained Knowledge Distillation Corneal Staining Score (FKD-CSS) model, which effectively distills fine-grained features into the CSS grading process and outputs continuous, nuanced scores for thorough assessments. Trained on 1471 images from 14 centers of heterogenous sources, FKD-CSS demonstrates robust accuracy with a Pearson's r of 0.898 against ground-truth and an area under the curve (AUC) of 0.881 in internal validation, rivaling senior ophthalmologists. Additionally, the model achieved expert performance with considerable Pearson's r (0.844–0.899) and AUCs (0.804–0.883) in external tes ts in six regions of China using 2376 corneal staining images of dry eye across 23 hospitals, and generalizes to multi-ocular-surface-disease test (Pearson's r: 0.816, AUC: 0.807), underscore its efficiency and explainability for CSS assessment. These results highlight FKD-CSS's potential as a precise, valuable tool for staging and outcome measurement of ocular surface diseases. Health sciences/Signs and symptoms/Eye manifestations Health sciences/Biomarkers/Diagnostic markers Health sciences/Diseases/Eye diseases/Corneal diseases Full Text Additional Declarations There is a conflict of interest Xiaoyi Li is affiliated with Zhaoke Ophthalmology Ltd, which is the sponsor of the phase III study for cyclosporine-A eye drops (ClinicalTrials.gov; identifier: NCT04541888). The images of developing, internal validation and primary external test datasets of the FKD-CSS model were mainly collected from this phase III study. The authors report no other conflicts of interest in this work. Supplementary Files Supplementaryappendix1.docx Supplementaryappendix2visualizationtoolforFKDCSS01.exe Supplementary appendix 2 visualization tool for FKD-CSS-01 Cite Share Download PDF Status: Published Journal Publication published 23 May, 2025 Read the published version in npj Digital Medicine → Version 1 posted Editorial decision: revise 01 Jul, 2024 Review # 2 received at journal 21 Jun, 2024 Review # 3 received at journal 11 Jun, 2024 Reviewer # 3 agreed at journal 02 Jun, 2024 Reviewer # 2 agreed at journal 29 May, 2024 Review # 1 received at journal 14 May, 2024 Reviewer # 1 agreed at journal 30 Apr, 2024 Reviewers invited by journal 28 Apr, 2024 Editor assigned by journal 16 Apr, 2024 Submission checks completed at journal 16 Apr, 2024 First submitted to journal 16 Apr, 2024 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-4274726","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":296492409,"identity":"9b946a99-560c-4704-a2ab-d4cc17c63525","order_by":0,"name":"Jin Yuan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA2klEQVRIiWNgGAWjYLACCQMYq0JCjp9ELWcsjCUbSLKOsa0icQMhLQY3kp89sCi4I88vffbgg5/zJBg3MDA/fHQDr5Y0cwMJg2eGM/vykg17t0kwmzOwGRvn4NWSYCYhYXA4weAMj5kE7zYJNssGHjZp/FrSv8G0mP/8O0eCx+AAQS05CFuYeRuAbEJaJM+8KQNpMZzZw2MsLXNMwkCymYBf+I6nb5OW+HNYnp+Hx/Djm5q6+n725oeP8WlROMDAwCyBIsSMRzkIyDcAI/ADAUWjYBSMglEwwgEAJdZF5nPotIUAAAAASUVORK5CYII=","orcid":"","institution":"State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University.","correspondingAuthor":true,"prefix":"","firstName":"Jin","middleName":"","lastName":"Yuan","suffix":""},{"id":296492410,"identity":"83deea2d-45fe-4808-8845-9312b3c1122b","order_by":1,"name":"Yuqing Deng","email":"","orcid":"","institution":"State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University.","correspondingAuthor":false,"prefix":"","firstName":"Yuqing","middleName":"","lastName":"Deng","suffix":""},{"id":296492411,"identity":"b3e0d0a1-ea12-4fff-9d07-857b1003e4e4","order_by":2,"name":"Pujin Cheng","email":"","orcid":"","institution":"Department of Electrical and Electronic Engineering, Southern University of Science and Technology.","correspondingAuthor":false,"prefix":"","firstName":"Pujin","middleName":"","lastName":"Cheng","suffix":""},{"id":296492412,"identity":"0e58f7b3-c8db-4aa5-bd27-1b4218c20aaf","order_by":3,"name":"Ruiwen Xu","email":"","orcid":"","institution":"State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University.","correspondingAuthor":false,"prefix":"","firstName":"Ruiwen","middleName":"","lastName":"Xu","suffix":""},{"id":296492413,"identity":"362472dd-ce6d-45d4-89f7-85a6e9cdf85c","order_by":4,"name":"Lirong Ling","email":"","orcid":"","institution":"State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University.","correspondingAuthor":false,"prefix":"","firstName":"Lirong","middleName":"","lastName":"Ling","suffix":""},{"id":296492414,"identity":"27deed0d-cbf0-4344-8e7f-b8f7e65f1718","order_by":5,"name":"Hongliang Xue","email":"","orcid":"","institution":"The Key Laboratory of Advanced Interdisciplinary Studies, The First Affiliated Hospital of Guangzhou Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hongliang","middleName":"","lastName":"Xue","suffix":""},{"id":296492415,"identity":"43df237c-3517-4b6d-8e1e-64b9ea15f4ff","order_by":6,"name":"Shiyou Zhou","email":"","orcid":"","institution":"State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University.","correspondingAuthor":false,"prefix":"","firstName":"Shiyou","middleName":"","lastName":"Zhou","suffix":""},{"id":296492416,"identity":"a126a9a1-13e2-4270-83d7-daae0b76950f","order_by":7,"name":"Yansong Huang","email":"","orcid":"","institution":"Department of Electrical and Electronic Engineering, Southern University of Science and Technology.","correspondingAuthor":false,"prefix":"","firstName":"Yansong","middleName":"","lastName":"Huang","suffix":""},{"id":296492417,"identity":"7b15e01b-868c-4543-ae65-05837b6e8576","order_by":8,"name":"Junyan Lyu","email":"","orcid":"","institution":"Department of Electrical and Electronic Engineering, Southern University of Science and Technology.","correspondingAuthor":false,"prefix":"","firstName":"Junyan","middleName":"","lastName":"Lyu","suffix":""},{"id":296492418,"identity":"851f0fbc-aeda-4a2a-9dd7-138ae7cc069d","order_by":9,"name":"Zhonghua Wang","email":"","orcid":"","institution":"Department of Electrical and Electronic Engineering, Southern University of Science and Technology.","correspondingAuthor":false,"prefix":"","firstName":"Zhonghua","middleName":"","lastName":"Wang","suffix":""},{"id":296492419,"identity":"7d632a1f-fa37-41f3-a901-ba8c23fb10e4","order_by":10,"name":"Kenneth Wong","email":"","orcid":"","institution":"Department of Electrical and Electronic Engineering, the University of Hong Kong.","correspondingAuthor":false,"prefix":"","firstName":"Kenneth","middleName":"","lastName":"Wong","suffix":""},{"id":296492420,"identity":"aba7bbe6-b017-42dc-8ad6-9546ba49be57","order_by":11,"name":"Yimin Zhang","email":"","orcid":"","institution":"State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University.","correspondingAuthor":false,"prefix":"","firstName":"Yimin","middleName":"","lastName":"Zhang","suffix":""},{"id":296492421,"identity":"2e98030b-7f87-4c8c-9510-404d65a678bd","order_by":12,"name":"Kang Yu","email":"","orcid":"","institution":"State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University.","correspondingAuthor":false,"prefix":"","firstName":"Kang","middleName":"","lastName":"Yu","suffix":""},{"id":296492422,"identity":"30fc2bcf-9d99-46de-9c9e-bcba0cb9fd6c","order_by":13,"name":"Tingting Zhang","email":"","orcid":"","institution":"State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University.","correspondingAuthor":false,"prefix":"","firstName":"Tingting","middleName":"","lastName":"Zhang","suffix":""},{"id":296492423,"identity":"3053e6b1-c5c9-4342-8feb-31b2864a32b5","order_by":14,"name":"Xiaoqing Hu","email":"","orcid":"","institution":"[email protected]","correspondingAuthor":false,"prefix":"","firstName":"Xiaoqing","middleName":"","lastName":"Hu","suffix":""},{"id":296492424,"identity":"4af577ce-2663-4d4c-8014-2c8c27f66fde","order_by":15,"name":"Xiaoyi Li","email":"","orcid":"","institution":"Zhaoke (Guangzhou) Ophthalmology Pharmaceutical Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Xiaoyi","middleName":"","lastName":"Li","suffix":""},{"id":296492425,"identity":"4d2b3b6e-5b39-417b-b299-a80d470ef531","order_by":16,"name":"Yan Lou","email":"","orcid":"","institution":"Department of Computer, School of Intelligent Medicine, China Medical University.","correspondingAuthor":false,"prefix":"","firstName":"Yan","middleName":"","lastName":"Lou","suffix":""},{"id":296492426,"identity":"c508b3c3-651e-416f-a435-21d1d2f63297","order_by":17,"name":"Xiaoying Tang","email":"","orcid":"","institution":"Department of Electrical and Electronic Engineering, Southern University of Science and Technology.","correspondingAuthor":false,"prefix":"","firstName":"Xiaoying","middleName":"","lastName":"Tang","suffix":""}],"badges":[],"createdAt":"2024-04-16 09:07:03","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4274726/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4274726/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41746-025-01706-y","type":"published","date":"2025-05-23T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":83376656,"identity":"4e2af47b-dc0d-43fa-84d6-9328c73eb857","added_by":"auto","created_at":"2025-05-24 07:11:27","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1086277,"visible":true,"origin":"","legend":"","description":"","filename":"FKDCSSdraft0416final.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4274726/v1_covered_bb308606-e345-48b4-8c94-f71e448c4afd.pdf"},{"id":55780952,"identity":"967204b0-6498-4867-aa36-76d928a6397d","added_by":"auto","created_at":"2024-05-03 04:49:42","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":186183,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Supplementaryappendix1.docx","url":"https://assets-eu.researchsquare.com/files/rs-4274726/v1/2cf8074ffd3fd483322f7bb5.docx"},{"id":55780953,"identity":"b1ec7722-bb14-4467-b1f1-ab680e6d5719","added_by":"auto","created_at":"2024-05-03 04:49:43","extension":"exe","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":10307072,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary appendix 2 visualization tool for FKD-CSS-01\u003c/p\u003e","description":"","filename":"Supplementaryappendix2visualizationtoolforFKDCSS01.exe","url":"https://assets-eu.researchsquare.com/files/rs-4274726/v1/a7dd21606b953756efe975d6.exe"}],"financialInterests":"There is a conflict of interest\nXiaoyi Li is affiliated with Zhaoke Ophthalmology Ltd, which is the sponsor of the phase III study for cyclosporine-A eye drops (ClinicalTrials.gov; identifier: NCT04541888). The images of developing, internal validation and primary external test datasets of the FKD-CSS model were mainly collected from this phase III study. The authors report no other conflicts of interest in this work.","formattedTitle":"Revolutionizing Corneal Staining Assessment: Advanced Evaluation through Lesion-aware Fine-Grained Knowledge Distillation","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":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-4274726/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4274726/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eCorneal staining is crucial for evaluating ocular surface diseases, yet existing AI models for CSS (Corneal Staining Score) assessments struggle with detailed lesion identification and lack applicability in real-world clinical settings. Moreover, the output of current AI-assist staining evaluation system only provides categories of grades, leading to potential \u0026ldquo;plateau\u0026rdquo; effect, which could misrepresent treatment response in clinical practices. Addressing these gaps, we developed the Fine-grained Knowledge Distillation Corneal Staining Score (FKD-CSS) model, which effectively distills fine-grained features into the CSS grading process and outputs continuous, nuanced scores for thorough assessments. Trained on 1471 images from 14 centers of heterogenous sources, FKD-CSS demonstrates robust accuracy with a Pearson's r of 0.898 against ground-truth and an area under the curve (AUC) of 0.881 in internal validation, rivaling senior ophthalmologists. Additionally, the model achieved expert performance with considerable Pearson's r (0.844\u0026ndash;0.899) and AUCs (0.804\u0026ndash;0.883) in external tes\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ets\u003c/span\u003e in six regions of China using 2376 corneal staining images of dry eye across 23 hospitals, and generalizes to multi-ocular-surface-disease test (Pearson's r: 0.816, AUC: 0.807), underscore its efficiency and explainability for CSS assessment. These results highlight FKD-CSS's potential as a precise, valuable tool for staging and outcome measurement of ocular surface diseases.\u003c/p\u003e","manuscriptTitle":"Revolutionizing Corneal Staining Assessment: Advanced Evaluation through Lesion-aware Fine-Grained Knowledge Distillation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-05-03 04:49:37","doi":"10.21203/rs.3.rs-4274726/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"revise","date":"2024-07-01T15:13:00+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"This content is not available.","date":"2024-06-21T20:12:07+00:00","index":2,"fulltext":"This content is not available."},{"type":"editorInvitedReview","content":"This content is not available.","date":"2024-06-11T20:42:28+00:00","index":3,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2024-06-02T16:25:18+00:00","index":3,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2024-05-30T01:11:00+00:00","index":2,"fulltext":"This content is not available."},{"type":"editorInvitedReview","content":"This content is not available.","date":"2024-05-14T06:06:18+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2024-04-30T05:49:07+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewersInvited","content":"","date":"2024-04-29T02:42:15+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-04-17T00:44:19+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-04-16T10:24:06+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Digital Medicine","date":"2024-04-16T09:04:23+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"18b0685a-df1a-40a7-9748-c41c718e9ffb","owner":[],"postedDate":"May 3rd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":31268534,"name":"Health sciences/Signs and symptoms/Eye manifestations"},{"id":31268535,"name":"Health sciences/Biomarkers/Diagnostic markers"},{"id":31268536,"name":"Health sciences/Diseases/Eye diseases/Corneal diseases"}],"tags":[],"updatedAt":"2025-05-24T07:11:19+00:00","versionOfRecord":{"articleIdentity":"rs-4274726","link":"https://doi.org/10.1038/s41746-025-01706-y","journal":{"identity":"npj-digital-medicine","isVorOnly":false,"title":"npj Digital Medicine"},"publishedOn":"2025-05-23 04:00:00","publishedOnDateReadable":"May 23rd, 2025"},"versionCreatedAt":"2024-05-03 04:49:37","video":"","vorDoi":"10.1038/s41746-025-01706-y","vorDoiUrl":"https://doi.org/10.1038/s41746-025-01706-y","workflowStages":[]},"version":"v1","identity":"rs-4274726","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4274726","identity":"rs-4274726","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}