Bias-constrained multimodal intelligence for equitable and reliable clinical AI

Bias-constrained multimodal intelligence for equitable and reliable clinical AI | 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 Biological Sciences - Article Bias-constrained multimodal intelligence for equitable and reliable clinical AI Shanshan Wang, Cheng Li, Weijian Huang, Jiarun Liu, Hao Yang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9447370/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 integration of medical imaging and clinical text 1,2 has enabled the emergence of generalist artificial intelligence (AI) systems for healthcare 3,4 . However, pervasive biases, such as imbalanced disease prevalence, skewed anatomical region distributions, heterogeneous imaging protocols, and demographic disparities, pose significant challenges to the fairness and reliability of vision-language systems in real-world clinical settings 5,6 . Here we present BiasCareVL, a bias-aware multimodal learning framework that introduces bias control directly into model design, rather than treating it as a post hoc correction. BiasCareVL incorporates adaptive uncertainty modeling with optional human-in-the-loop refinement to regulate the influence of dominant data patterns and to promote equitable reasoning under distributional imbalance. Trained on 3.44 million samples spanning over 15 imaging modalities, the framework supports diverse clinical tasks, including visual question answering, disease classification, segmentation, and report generation within a unified representation space. Across eight public benchmarks covering dermatology, oncology, radiology, and pathology, BiasCareVL consistently outperforms 20 state-of-the-art methods, with pronounced gains in clinically challenging scenarios, including over 10% accuracy improvement in multi-class skin lesion diagnosis and more than 20% Dice improvement in small tumor segmentation. Furthermore, BiasCareVL achieves diagnostic performance exceeding human accuracy with substantially reduced time requirements when evaluated with board-certified radiologists. By open-sourcing BiasCareVL, we aim to promote a transparent, reproducible, and equitable future for AI in healthcare, paving the way for general-purpose, trustworthy, and clinically reliable AI systems. Biological sciences/Computational biology and bioinformatics/Image processing Physical sciences/Engineering/Biomedical engineering Biological sciences/Computational biology and bioinformatics/Machine learning Health sciences/Health care/Medical imaging Full Text Additional Declarations There is NO Competing Interest. Supplementary Files nrreportingsummary.pdf Reporting Summary 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-9447370","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Biological Sciences - Article","associatedPublications":[],"authors":[{"id":627170177,"identity":"7a53a10e-0d25-420c-960f-caf144476080","order_by":0,"name":"Shanshan Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/0lEQVRIiWNgGAWjYHACNoYEGPNjA4hkbDxAtBbGmQ0MEkCqgbAWGGDmBWthYMCrRb6999iDBzV19gbHzx5+bbvDpk63/TDQlhqbaFxaDM6cSzdIOHY4ccOZvDTr3DNpEmZnEoFajqXlNuDSIpFjJpHAdiDB7ECOmXFu22EJswNALYwNh3FqkZ8B0vKvzt7s/BszY0uQlvMP8WthuAHUktjGzLjtRo7xY0aQlhsEbAH6JU0ise9w4v4bb8wYe9vSJLfdANqSgMcvoBCT/PGtzl6yP8f4w882G36z8+kPH3yoscHtMAYeOItNAs5MwKkcVQvzB7wKR8EoGAWjYMQCAK+kZU6vQ9PSAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-0575-6523","institution":"Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences","correspondingAuthor":true,"prefix":"","firstName":"Shanshan","middleName":"","lastName":"Wang","suffix":""},{"id":627170178,"identity":"72cc2c5d-8a76-4e44-bbbc-fe99bb44aac9","order_by":1,"name":"Cheng Li","email":"","orcid":"https://orcid.org/0000-0001-5400-2093","institution":"Paul C. Lauterbur Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Cheng","middleName":"","lastName":"Li","suffix":""},{"id":627170179,"identity":"cffd08a4-0ac1-46be-995c-f8ed2dbd347f","order_by":2,"name":"Weijian Huang","email":"","orcid":"","institution":"Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Weijian","middleName":"","lastName":"Huang","suffix":""},{"id":627170180,"identity":"2e397c32-dbab-402b-938e-6d15fbef1819","order_by":3,"name":"Jiarun Liu","email":"","orcid":"","institution":"Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Jiarun","middleName":"","lastName":"Liu","suffix":""},{"id":627170181,"identity":"152cf199-60aa-4990-a4d9-7877393d58a3","order_by":4,"name":"Hao Yang","email":"","orcid":"","institution":"Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Yang","suffix":""},{"id":627170182,"identity":"f8f92c92-1348-49e4-ad72-cf89f23b63d7","order_by":5,"name":"Qi Yang","email":"","orcid":"","institution":"Department of Radiology, Beijing Chaoyang Hospital, Capital Medical University","correspondingAuthor":false,"prefix":"","firstName":"Qi","middleName":"","lastName":"Yang","suffix":""},{"id":627170183,"identity":"ae723453-0367-435e-adb8-698d3f078f0c","order_by":6,"name":"Song Wu","email":"","orcid":"https://orcid.org/0000-0003-3504-1630","institution":"South China Hospital, Medical School, Shenzhen University, College of Electronics and Information Engineering, Shenzhen University","correspondingAuthor":false,"prefix":"","firstName":"Song","middleName":"","lastName":"Wu","suffix":""},{"id":627170184,"identity":"e9ac237b-f78e-401e-89c7-8ba06ccc556f","order_by":7,"name":"Ye Li","email":"","orcid":"","institution":"Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Ye","middleName":"","lastName":"Li","suffix":""},{"id":627170185,"identity":"106f3f3b-18e6-468c-a198-077b21e54457","order_by":8,"name":"Hairong Zheng","email":"","orcid":"https://orcid.org/0000-0002-8558-5102","institution":"Shenzhen Institutes of Advanced Technology","correspondingAuthor":false,"prefix":"","firstName":"Hairong","middleName":"","lastName":"Zheng","suffix":""}],"badges":[],"createdAt":"2026-04-17 09:51:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9447370/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9447370/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107506042,"identity":"e6aa6612-14ba-406f-af8c-59a98cc87519","added_by":"auto","created_at":"2026-04-22 06:57:53","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1679985,"visible":true,"origin":"","legend":"","description":"","filename":"BiasCareVLmanuscript20260417.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9447370/v1_covered_ad45be90-0d34-406d-b45f-b6bc5ce09dc7.pdf"},{"id":107505882,"identity":"dffaf55e-8c98-4a23-be46-c504dc792d67","added_by":"auto","created_at":"2026-04-22 06:57:07","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1666207,"visible":true,"origin":"","legend":"Reporting Summary","description":"","filename":"nrreportingsummary.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9447370/v1/385216cf0b515c9033215332.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Bias-constrained multimodal intelligence for equitable and reliable clinical AI","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":"","lastPublishedDoi":"10.21203/rs.3.rs-9447370/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9447370/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The integration of medical imaging and clinical text\u003csup\u003e1,2\u003c/sup\u003e has enabled the emergence of generalist artificial intelligence (AI) systems for healthcare\u003csup\u003e3,4\u003c/sup\u003e. However, pervasive biases, such as imbalanced disease prevalence, skewed anatomical region distributions, heterogeneous imaging protocols, and demographic disparities, pose significant challenges to the fairness and reliability of vision-language systems in real-world clinical settings\u003csup\u003e5,6\u003c/sup\u003e. Here we present BiasCareVL, a bias-aware multimodal learning framework that introduces bias control directly into model design, rather than treating it as a post hoc correction. BiasCareVL incorporates adaptive uncertainty modeling with optional human-in-the-loop refinement to regulate the influence of dominant data patterns and to promote equitable reasoning under distributional imbalance. Trained on 3.44 million samples spanning over 15 imaging modalities, the framework supports diverse clinical tasks, including visual question answering, disease classification, segmentation, and report generation within a unified representation space. Across eight public benchmarks covering dermatology, oncology, radiology, and pathology, BiasCareVL consistently outperforms 20 state-of-the-art methods, with pronounced gains in clinically challenging scenarios, including over 10% accuracy improvement in multi-class skin lesion diagnosis and more than 20% Dice improvement in small tumor segmentation. Furthermore, BiasCareVL achieves diagnostic performance exceeding human accuracy with substantially reduced time requirements when evaluated with board-certified radiologists. By open-sourcing BiasCareVL, we aim to promote a transparent, reproducible, and equitable future for AI in healthcare, paving the way for general-purpose, trustworthy, and clinically reliable AI systems.","manuscriptTitle":"Bias-constrained multimodal intelligence for equitable and reliable clinical AI","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-22 06:55:07","doi":"10.21203/rs.3.rs-9447370/v1","editorialEvents":[],"status":"published","journal":{"display":false,"email":"[email protected]","identity":"nature","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"nature","sideBox":"Learn more about [Nature](http://www.nature.com/nature/)","snPcode":"","submissionUrl":"","title":"Nature","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"070aff50-9354-44fe-a512-4a4f80f92257","owner":[],"postedDate":"April 22nd, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"This content is not available.","date":"2026-05-13T14:47:40+00:00","index":6,"fulltext":"This content is not available."},{"type":"editorInvitedReview","content":"This content is not available.","date":"2026-05-12T22:38:49+00:00","index":7,"fulltext":"This content is not available."}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":66752508,"name":"Biological sciences/Computational biology and bioinformatics/Image processing"},{"id":66752509,"name":"Physical sciences/Engineering/Biomedical engineering"},{"id":66752510,"name":"Biological sciences/Computational biology and bioinformatics/Machine learning"},{"id":66752511,"name":"Health sciences/Health care/Medical imaging"}],"tags":[],"updatedAt":"2026-04-22T06:55:07+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-22 06:55:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9447370","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9447370","identity":"rs-9447370","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}