A Multimodal Large Reasoning Model For Fair and Interpretable Dermatological Diagnosis Across Skin Tones

A Multimodal Large Reasoning Model For Fair and Interpretable Dermatological Diagnosis Across Skin Tones | 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 A Multimodal Large Reasoning Model For Fair and Interpretable Dermatological Diagnosis Across Skin Tones Juexiao Zhou, Yuhao Shen, Zhangtianyi Chen, Yuanhao He, Yan Xu, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9102583/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract The clinical translation of dermatological artificial intelligence is severely limited by opaque decision reasoning and systematic performance disparities across skin tones. Here, we introduce SkinGPT-R1, a multimodal large reasoning model explicitly designed for trustworthy, interpretable, and fairness-aware skin disease diagnosis. SkinGPT-R1 unifies chain-of-thought diagnostic reasoning with a fairness-aware mixture-of-experts architecture to enable equitable and human-readable diagnostic outputs. Trained on 334,168 clinical samples, SkinGPT-R1 generates comprehensive diagnostic reports comprising visual findings, differential reasoning, and final diagnosis. On six independent external validation datasets covering diverse dermatological conditions and imaging settings, SkinGPT-R1 achieves state-of-the-art diagnostic accuracy. On a challenging 40-class long-tail classification task, it attains 82.5% accuracy, representing an absolute improvement of 19.3% over strong baseline models. In a blinded evaluation by five board-certified dermatologists using 1,000 phenotypically balanced cases, SkinGPT-R1 achieves a mean overall score of 3.6 out of 5, with the highest ratings for safety (3.8/5) and reasoning coherence (3.6/5), confirming that its generated rationales are clinically valid, logically consistent, and suitable for supporting clinical decision-making. Critically, SkinGPT-R1 effectively mitigates algorithmic bias across the full Fitzpatrick skin tone spectrum, achieving a robust worst-group performance of 41.40% on the Fitz17k benchmark and a five-fold relative improvement in lower-bound accuracy on the DDI dataset relative to standard multimodal baselines. These results establish a generalizable framework for fair, interpretable, and clinically trustworthy AI-assisted dermatological diagnosis, addressing key obstacles to real-world clinical deployment and advancing health equity in dermatological care. Biological sciences/Cancer/Skin cancer Health sciences/Diseases/Skin diseases Dermatology AI Medical MLLM Chain-of-Thought Reasoning Algorithmic Fairness Full Text Additional Declarations There is NO Competing Interest. Supplementary Files 7demo.mp4 Supplementary Video 3SupplementarySkinGPTR1compress.pdf Supplementary Information Cite Share Download PDF Status: Under Review 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-9102583","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":609601899,"identity":"eedf6340-135d-40ba-a6be-84b9fd61fcdc","order_by":0,"name":"Juexiao Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8ElEQVRIiWNgGAWjYBACxgYgkQBiSQDxByBmYydFC+MMkBZmou0DamHmATEIaWGekXxM4kGNDYP87OZj0ja/tsnzMTMwfviYg8dhM9KSDRKOpTEwzjmWJp3bd9uwjZmBWXLmNnxacgwfJLAdZmCWyDGTzu25zQjUwsbMi1dL/ocDCf8OM7CBtFj23LYnQksO44PEtsMMPCAtDD9uJxLW0vPM2CCxL41HQiIt2bK34XZyGzNjM16/GLYnP5P88c1GTn5G8sEbP/7ctp3f3nzww0d8WhogNA/EzjYw2YBbPRDIo3L/4FU8CkbBKBgFIxQAAIJISgN5dmkyAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-6739-6236","institution":"The Chinese University of Hong Kong, Shenzhen","correspondingAuthor":true,"prefix":"","firstName":"Juexiao","middleName":"","lastName":"Zhou","suffix":""},{"id":609601900,"identity":"aa4e4dc7-df8a-4a96-a00a-d31b61ec40cf","order_by":1,"name":"Yuhao Shen","email":"","orcid":"","institution":"The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen)","correspondingAuthor":false,"prefix":"","firstName":"Yuhao","middleName":"","lastName":"Shen","suffix":""},{"id":609601901,"identity":"6d7f3694-be20-4dc2-bbf7-929fbc8a02cd","order_by":2,"name":"Zhangtianyi Chen","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Zhangtianyi","middleName":"","lastName":"Chen","suffix":""},{"id":609601902,"identity":"393eb0a8-49ca-442b-a5c6-3f046276f92b","order_by":3,"name":"Yuanhao He","email":"","orcid":"","institution":"The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen)","correspondingAuthor":false,"prefix":"","firstName":"Yuanhao","middleName":"","lastName":"He","suffix":""},{"id":609601903,"identity":"f920afa0-d245-462e-9a53-34578554ed48","order_by":4,"name":"Yan Xu","email":"","orcid":"","institution":"Department of Dermatology, Tianjin Institute of Integrative Dermatology, Tianjin Academy of Traditional Chinese Medicine Affiliated Hospital, Tianjin 300120, China.","correspondingAuthor":false,"prefix":"","firstName":"Yan","middleName":"","lastName":"Xu","suffix":""},{"id":609601904,"identity":"ec75517f-078f-4ec4-a60e-e71e1fd7a1d4","order_by":5,"name":"Shuping Zhang","email":"","orcid":"","institution":"Department of Dermatology, The First Affiliated Hospital, Shantou University Medical College, Shantou 515041, China.","correspondingAuthor":false,"prefix":"","firstName":"Shuping","middleName":"","lastName":"Zhang","suffix":""},{"id":609601905,"identity":"c9a8e4d3-f3d5-4f19-8d12-f1dd89df0758","order_by":6,"name":"Liyuan Sun","email":"","orcid":"","institution":"Beijing AnZhen Hospital, Capital Medical University","correspondingAuthor":false,"prefix":"","firstName":"Liyuan","middleName":"","lastName":"Sun","suffix":""},{"id":609601906,"identity":"9135192f-eef7-4c08-880b-415cf7d5e670","order_by":7,"name":"Zijian Wang","email":"","orcid":"","institution":"School of Data Science, The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen), Guangdong, 518172, China.","correspondingAuthor":false,"prefix":"","firstName":"Zijian","middleName":"","lastName":"Wang","suffix":""},{"id":609601907,"identity":"f511cad9-e58c-474c-9a32-6717df8685da","order_by":8,"name":"Yinghao Zhu","email":"","orcid":"https://orcid.org/0000-0002-2640-6477","institution":"The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Yinghao","middleName":"","lastName":"Zhu","suffix":""},{"id":609601908,"identity":"731dda2e-28e4-4314-b1b8-9d6340a763ba","order_by":9,"name":"Jiahe Qian","email":"","orcid":"","institution":"Northwestern University","correspondingAuthor":false,"prefix":"","firstName":"Jiahe","middleName":"","lastName":"Qian","suffix":""},{"id":609601909,"identity":"e2db0c27-e2cb-4982-a4b1-1416db0d83f8","order_by":10,"name":"Yuyuan Yang","email":"","orcid":"","institution":"School of Data Science, The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen), Guangdong, 518172, China.","correspondingAuthor":false,"prefix":"","firstName":"Yuyuan","middleName":"","lastName":"Yang","suffix":""},{"id":609601910,"identity":"2873d1e1-36eb-4fcb-b0aa-97139fed5f64","order_by":11,"name":"Ziwen Wang","email":"","orcid":"","institution":"School of Data Science, The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen), Guangdong, 518172, China.","correspondingAuthor":false,"prefix":"","firstName":"Ziwen","middleName":"","lastName":"Wang","suffix":""},{"id":609601911,"identity":"66e38bb4-f2af-417c-8843-ca57908b51bd","order_by":12,"name":"Xinyuan Zhang","email":"","orcid":"","institution":"School of Data Science, The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen), Guangdong, 518172, China.","correspondingAuthor":false,"prefix":"","firstName":"Xinyuan","middleName":"","lastName":"Zhang","suffix":""},{"id":609601912,"identity":"2727874d-e505-498d-a4f3-33d82378a611","order_by":13,"name":"Wenbin Liu","email":"","orcid":"","institution":"Department of Dermatology, Beijing Aerospace General Hospital, Beijing 100076, China.","correspondingAuthor":false,"prefix":"","firstName":"Wenbin","middleName":"","lastName":"Liu","suffix":""},{"id":609601913,"identity":"c6d2741c-2146-4e4f-aa91-68ae65a2591f","order_by":14,"name":"Zongyuan Ge","email":"","orcid":"https://orcid.org/0000-0002-5880-8673","institution":"Monash University","correspondingAuthor":false,"prefix":"","firstName":"Zongyuan","middleName":"","lastName":"Ge","suffix":""},{"id":609601914,"identity":"1ca6b3f4-f394-4c6d-aa81-8dabef46a1cf","order_by":15,"name":"Tao Lu","email":"","orcid":"","institution":"Department of Dermatology, The First Affiliated Hospital, Shantou University Medical College, Shantou 515041, China.","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Lu","suffix":""},{"id":609601915,"identity":"c6d97695-8654-4683-aa20-1c54fbafd144","order_by":16,"name":"Siyuan Yan","email":"","orcid":"https://orcid.org/0000-0002-6372-8336","institution":"Monash University","correspondingAuthor":false,"prefix":"","firstName":"Siyuan","middleName":"","lastName":"Yan","suffix":""}],"badges":[],"createdAt":"2026-03-12 09:00:30","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9102583/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9102583/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105904606,"identity":"06ae3d9a-8a82-425a-9d1e-3e3fdb009654","added_by":"auto","created_at":"2026-04-01 10:09:51","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8175848,"visible":true,"origin":"","legend":"Article File","description":"","filename":"2ManuscriptSkinGPTR1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9102583/v1_covered_0c2cdf99-c547-4d4f-9d9a-74aa83b43b81.pdf"},{"id":105823277,"identity":"a0e67cd4-c4fd-4180-949b-9626cae46923","added_by":"auto","created_at":"2026-03-31 13:36:38","extension":"mp4","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1481378,"visible":true,"origin":"","legend":"Supplementary Video","description":"","filename":"7demo.mp4","url":"https://assets-eu.researchsquare.com/files/rs-9102583/v1/3cc3c8ff2c7e3880c02beced.mp4"},{"id":105823276,"identity":"3ea6d144-41ea-4974-9665-4a34ce5dea36","added_by":"auto","created_at":"2026-03-31 13:36:38","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":13282648,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"3SupplementarySkinGPTR1compress.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9102583/v1/8e18e1b4c09177f0efb9c95e.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"A Multimodal Large Reasoning Model For Fair and Interpretable Dermatological Diagnosis Across Skin Tones","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"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":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Dermatology AI, Medical MLLM, Chain-of-Thought Reasoning, Algorithmic Fairness","lastPublishedDoi":"10.21203/rs.3.rs-9102583/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9102583/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The clinical translation of dermatological artificial intelligence is severely limited by opaque decision reasoning and systematic performance disparities across skin tones. Here, we introduce SkinGPT-R1, a multimodal large reasoning model explicitly designed for trustworthy, interpretable, and fairness-aware skin disease diagnosis. SkinGPT-R1 unifies chain-of-thought diagnostic reasoning with a fairness-aware mixture-of-experts architecture to enable equitable and human-readable diagnostic outputs. Trained on 334,168 clinical samples, SkinGPT-R1 generates comprehensive diagnostic reports comprising visual findings, differential reasoning, and final diagnosis. On six independent external validation datasets covering diverse dermatological conditions and imaging settings, SkinGPT-R1 achieves state-of-the-art diagnostic accuracy. On a challenging 40-class long-tail classification task, it attains 82.5% accuracy, representing an absolute improvement of 19.3% over strong baseline models. In a blinded evaluation by five board-certified dermatologists using 1,000 phenotypically balanced cases, SkinGPT-R1 achieves a mean overall score of 3.6 out of 5, with the highest ratings for safety (3.8/5) and reasoning coherence (3.6/5), confirming that its generated rationales are clinically valid, logically consistent, and suitable for supporting clinical decision-making. Critically, SkinGPT-R1 effectively mitigates algorithmic bias across the full Fitzpatrick skin tone spectrum, achieving a robust worst-group performance of 41.40% on the Fitz17k benchmark and a five-fold relative improvement in lower-bound accuracy on the DDI dataset relative to standard multimodal baselines. These results establish a generalizable framework for fair, interpretable, and clinically trustworthy AI-assisted dermatological diagnosis, addressing key obstacles to real-world clinical deployment and advancing health equity in dermatological care.","manuscriptTitle":"A Multimodal Large Reasoning Model For Fair and Interpretable Dermatological Diagnosis Across Skin Tones","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-31 13:36:33","doi":"10.21203/rs.3.rs-9102583/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"b62d0724-567a-4cc6-a959-ddc67bdf21c8","owner":[],"postedDate":"March 31st, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":64869278,"name":"Biological sciences/Cancer/Skin cancer"},{"id":64869279,"name":"Health sciences/Diseases/Skin diseases"}],"tags":[],"updatedAt":"2026-03-31T13:36:33+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-31 13:36:33","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9102583","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9102583","identity":"rs-9102583","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}