Mini‑LLM‑Augmented Explainable Multi‑Class Chest X‑ray Classification for COVID‑19 Clinical Decision Support

Mini‑LLM‑Augmented Explainable Multi‑Class Chest X‑ray Classification for COVID‑19 Clinical Decision Support | 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 Research Article Mini‑LLM‑Augmented Explainable Multi‑Class Chest X‑ray Classification for COVID‑19 Clinical Decision Support MD Roman Sarkar This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8433620/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 Mini‑LLM–augmented explainable AI framework for multi‑class analysis of chest X‑rays to support COVID‑19 screening/triage workflows. The pipeline uses a deep convolutional classifier (EfficientNet‑B0) trained to discriminate four clinically relevant categories (COVID‑19, Lung Opacity, Normal, and Viral Pneumonia). To improve transparency, model attention is visualized using Grad‑CAM++ heatmaps and further contextualized by overlaying available lung/lesion masks, enabling image‑level rationale inspection alongside predicted class probabilities. For human‑readable decision support, a compact instruction‑tuned language model (Phi‑3.5 Mini‑Instruct) generates concise, non‑diagnostic clinical narrative summaries conditioned on the model’s probabilistic outputs and quantitative imaging cues. Experiments on the COVID‑19 Radiography Dataset demonstrate strong multi‑class performance, achieving 96.29% accuracy, macro F1‑score of 0.9627, and ROC‑AUC of 0.9966, with consistently high per‑class F1‑scores across all categories. Compared with common CNN baselines (ResNet18, DenseNet121, EfficientNet‑B0), the proposed Mini‑LLM framework yields improved accuracy and ROC‑AUC. Overall, the approach combines high‑accuracy classification with visual and textual explainability, providing a practical pathway toward clinician‑interpretable, deployment‑friendly decision support from chest radiographs. Mini‑LLM small language model Phi‑3.5 Mini‑Instruct explainable AI (XAI) clinical decision support 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-8433620","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":565735075,"identity":"dfe9561a-4b2c-470d-81e0-31029ec7aa63","order_by":0,"name":"MD Roman Sarkar","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDUlEQVRIiWNgGAWjYPCDAwwM/CA6oYAULZINIC0GpGgxOABi4NHCL91+7QPjHpto/tntDz9XnLHLNz6/OvHDAwMGeX6xA1i1SM45UzyD4Vla7ow7Z4wlz9xIttx24+1mCaDDDGfOTsCqxeBGTjLQPYdzG27kAL3xgdnA7MbZDSAtCQa3sWuxh2mZfyP98c+GD/UGxjPObv6BT4uBRPphsJYNNxLMJBtuHDYw4O/dhtcWiRs5zAwJB9JyN97IMbNsOHPcQOIG7zaLBAMJnH7hn5H+mOHDAZvceUCH3Ww4Vm3A3392880fFTby/NLYtTAw8ADdgGoxmCuBQzkIsD9At/gAHtWjYBSMglEwEgEAn1NnpkeoiCsAAAAASUVORK5CYII=","orcid":"","institution":"Southwest University of Science and Technology","correspondingAuthor":true,"prefix":"","firstName":"MD","middleName":"Roman","lastName":"Sarkar","suffix":""}],"badges":[],"createdAt":"2025-12-23 12:23:51","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8433620/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8433620/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109087305,"identity":"b683aec1-d210-4611-b501-57a5e87daa3b","added_by":"auto","created_at":"2026-05-12 13:14:34","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2846893,"visible":true,"origin":"","legend":"","description":"","filename":"MiniLLMRevision.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8433620/v1_covered_af1c10f4-70f0-46cf-873c-fb71427362a4.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Mini‑LLM‑Augmented Explainable Multi‑Class Chest X‑ray Classification for COVID‑19 Clinical Decision Support","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":"Mini‑LLM, small language model, Phi‑3.5 Mini‑Instruct, explainable AI (XAI), clinical decision support","lastPublishedDoi":"10.21203/rs.3.rs-8433620/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8433620/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMini‑LLM\u0026ndash;augmented explainable AI framework for multi‑class analysis of chest X‑rays to support COVID‑19 screening/triage workflows. The pipeline uses a deep convolutional classifier (EfficientNet‑B0) trained to discriminate four clinically relevant categories (COVID‑19, Lung Opacity, Normal, and Viral Pneumonia). To improve transparency, model attention is visualized using Grad‑CAM++ heatmaps and further contextualized by overlaying available lung/lesion masks, enabling image‑level rationale inspection alongside predicted class probabilities. For human‑readable decision support, a compact instruction‑tuned language model (Phi‑3.5 Mini‑Instruct) generates concise, non‑diagnostic clinical narrative summaries conditioned on the model\u0026rsquo;s probabilistic outputs and quantitative imaging cues. Experiments on the COVID‑19 Radiography Dataset demonstrate strong multi‑class performance, achieving 96.29% accuracy, macro F1‑score of 0.9627, and ROC‑AUC of 0.9966, with consistently high per‑class F1‑scores across all categories. Compared with common CNN baselines (ResNet18, DenseNet121, EfficientNet‑B0), the proposed Mini‑LLM framework yields improved accuracy and ROC‑AUC. Overall, the approach combines high‑accuracy classification with visual and textual explainability, providing a practical pathway toward clinician‑interpretable, deployment‑friendly decision support from chest radiographs.\u003c/p\u003e","manuscriptTitle":"Mini‑LLM‑Augmented Explainable Multi‑Class Chest X‑ray Classification for COVID‑19 Clinical Decision Support","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-12 12:25:24","doi":"10.21203/rs.3.rs-8433620/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":"b0dd38fe-3850-4358-939f-a47c45954b7e","owner":[],"postedDate":"May 12th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-12T12:25:24+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-12 12:25:24","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8433620","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8433620","identity":"rs-8433620","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}