A Reasoning Pathway Explanation Framework for Clinical AI: Methods and Evaluation

A Reasoning Pathway Explanation Framework for Clinical AI: Methods and Evaluation | 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 A Reasoning Pathway Explanation Framework for Clinical AI: Methods and Evaluation Yunguo Yu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9418990/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 Objective : When AI predicts acute myocardial infarction, existing explanation methods identify which features mattered (e.g., elevated troponin) but not how risk factors lead to the diagnosis through biological mechanisms. We developed a framework that generates reasoning pathways—clinically grounded chains linking risk factors, pathophysiology, and evidence—to address this gap. Methods : Using a 34-node cardiology reasoning graph, MIMIC-III data, and medical ontologies (SNOMED CT, UMLS), we built an explanation engine that maps AI predictions to temporally ordered, evidence-linked pathways. We evaluated 100 AMI cases (87 expanded, 13 independent) with six structural metrics, adversarial validation, and comparison with BioBERT and SHAP. Three physicians independently rated 11 cases across five clinical quality dimensions. Results : The framework generated consistent reasoning pathways across all cases (pathwayprediction consistency 0.85 ± 0.01). Adversarial validation confirmed discriminative power for this metric (AUC-ROC 0.81) but not others. In the physician pilot, inter-rater agreement was strong (ICC = 0.83) and all three evaluators detected the deliberately flawed control case (2.13/5 vs. 3.92/5 for genuine cases). Evidence sufficiency for complex independent cases was the primary concern identified. Conclusion : Structured reasoning pathways that trace clinical logic from risk factors to diagnosis can be generated and evaluated systematically. Physicians agreed on pathway quality (ICC = 0.83) and consistently rejected flawed explanations, though evidence depth for complex cases requires improvement. Bioinformatics Artificial Intelligence Clinical Decision Support Explainable AI Reasoning Pathways Ontologies Diagnostic Reasoning MIMIC-III Full Text Additional Declarations The authors declare no competing interests. Supplementary Files supplementalmaterialsRS.pdf 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-9418990","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":623167555,"identity":"75f3a823-6430-4e42-ad00-bd9c12d445d5","order_by":0,"name":"Yunguo Yu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAw0lEQVRIiWNgGAWjYDADfhCRUECKFskGkBYDUrQYHACTxKi8kWMmXfDrsJzx+dWJHx4YMMjzix0gQsvMvsPGZjfebpYAOsxw5uwEIrTw9hxO3Hbj7AaQlgSD28Rq2Tzj7OYfxGvh+XE4cQN/7zbibJE886zYmrch3VjiBu82iwQDCcJ+4TuevPE2zx9rOf7+s5tv/qiwkeeXJqBF4QCHAQNjG5AlAVYpgV85CMg3sD9gYPgDZPEfIKx6FIyCUTAKRiYAAAlWRzOowwZ5AAAAAElFTkSuQmCC","orcid":"https://orcid.org/0009-0006-2003-3813","institution":"Zyter|TruCare","correspondingAuthor":true,"prefix":"","firstName":"Yunguo","middleName":"","lastName":"Yu","suffix":""}],"badges":[],"createdAt":"2026-04-14 19:01:19","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9418990/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9418990/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107481015,"identity":"27b835c8-6c57-456a-a5f8-56c6dfd9ce74","added_by":"auto","created_at":"2026-04-22 02:15:12","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":228328,"visible":true,"origin":"","legend":"","description":"","filename":"manuscriptRS.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9418990/v1_covered_3aee68ef-f55f-4f50-82f3-9a3550b4d795.pdf"},{"id":107098595,"identity":"79d23572-24e1-4b95-a8f3-57046b47306e","added_by":"auto","created_at":"2026-04-16 18:11:07","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":77854,"visible":true,"origin":"","legend":"","description":"","filename":"supplementalmaterialsRS.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9418990/v1/9bc985a1410ada6d2ecba31a.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eA Reasoning Pathway Explanation Framework for Clinical AI: Methods and Evaluation\u003c/p\u003e","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":"Artificial Intelligence, Clinical Decision Support, Explainable AI, Reasoning Pathways, Ontologies, Diagnostic Reasoning, MIMIC-III","lastPublishedDoi":"10.21203/rs.3.rs-9418990/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9418990/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective\u003c/strong\u003e: When AI predicts acute myocardial infarction, existing explanation methods identify which features mattered (e.g., elevated troponin) but not how risk factors lead to the diagnosis through biological mechanisms. We developed a framework that generates reasoning pathways—clinically grounded chains linking risk factors, pathophysiology, and evidence—to address this gap.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e: Using a 34-node cardiology reasoning graph, MIMIC-III data, and medical ontologies (SNOMED CT, UMLS), we built an explanation engine that maps AI predictions to temporally ordered, evidence-linked pathways. We evaluated 100 AMI cases (87 expanded, 13 independent) with six structural metrics, adversarial validation, and comparison with BioBERT and SHAP. Three physicians independently rated 11 cases across five clinical quality dimensions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: The framework generated consistent reasoning pathways across all cases (pathwayprediction consistency 0.85 ± 0.01). Adversarial validation confirmed discriminative power for this metric (AUC-ROC 0.81) but not others. In the physician pilot, inter-rater agreement was strong (ICC = 0.83) and all three evaluators detected the deliberately flawed control case (2.13/5 vs. 3.92/5 for genuine cases). Evidence sufficiency for complex independent cases was the primary concern identified.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e: Structured reasoning pathways that trace clinical logic from risk factors to diagnosis can be generated and evaluated systematically. Physicians agreed on pathway quality (ICC = 0.83) and consistently rejected flawed explanations, though evidence depth for complex cases requires improvement.\u003c/p\u003e","manuscriptTitle":"A Reasoning Pathway Explanation Framework for Clinical AI: Methods and Evaluation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-16 18:11:03","doi":"10.21203/rs.3.rs-9418990/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":"64cdfede-f8e6-4a1f-9cb0-e73bf85c24c2","owner":[],"postedDate":"April 16th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":66319504,"name":"Bioinformatics"}],"tags":[],"updatedAt":"2026-04-16T18:11:03+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-16 18:11:03","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9418990","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9418990","identity":"rs-9418990","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}