Ensuring Safety in Clinical AI: Formally Verified Deep Learning for Heart Failure Detection

Ensuring Safety in Clinical AI: Formally Verified Deep Learning for Heart Failure Detection | 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 Ensuring Safety in Clinical AI: Formally Verified Deep Learning for Heart Failure Detection Imen Chebbi, Sarra Abidi, Leila Ben Ayed This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7517125/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 A major obstacle in fields including sports, clinical rehabilitation, and workplace safety is the timely detection and prevention of physical injuries. The majority of conventional monitoring systems are reactive, depending on post-event analysis or unimodal data sources, which restricts their ability to provide proactive actions and early warnings. Furthermore, current AI-driven health systems lack rigorous validation procedures, which compromises their suitability for practical implementation in safety-critical settings. In this work, we present MHIDS (Multimodal Hybrid Injury Detection System), an integrated, AI-based diagnostic framework that combines wearable physiological sensors, computer vision, and personalized physiological modeling for real-time injury forecasting. A continuously updated digital twin is employed to capture each user’s biomechanical and physiological profile, allowing adaptive, individualized risk assessment. Unlike conventional approaches, MHIDS incorporates a closed-loop feedback mechanism that dynamically reconfigures sensing parameters and provides actionable recommendations (e.g., posture correction, intensity adjustment, or rest scheduling), thereby shifting the paradigm from passive detection to proactive prevention. To guarantee correctness and operational trustworthiness, MHIDS is formally modeled in UPPAAL as a network of timed automata, ensuring critical properties such as bounded response times (<100 ms), safety, liveness, and deadlock freedom. Experimental validation using the publicly available MHEALTH dataset demonstrates superior predictive performance, achieving an accuracy of 99.21%, precision of 98.94%, recall of 99.07%, and F1-score of 99.00%, significantly outperforming state-of-the-art baselines. Multimodal Hybrid Injury Detection AI Healthcare Monitoring 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-7517125","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":516595512,"identity":"ff1d9383-ec34-4939-955a-ef0fd41035fa","order_by":0,"name":"Imen Chebbi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABEElEQVRIiWNgGAWjYBACNhDB2MCQYMDMfOADXLiCOC1siTPgwmeAmAefVWAtDDyGxGnhEzt87MHPHTZ55uw8Hxt/VNTaM0gkMH84UHOYwV66AbvDpNPSDXvPpBVbNvNubOY5czyxQSKBTeLAsTQGHpkDOLTkmEkzth1O3HCYd/tjxrZjCQwS+d+YP7DZMPBIJODQkv8NqOU/UAvPw8af/45BHfZPAo+WHDaglgMgLYwNvA01jECHMUgcbMNnS5qZZG9bMtAvbIbNPMcOJLbxPGCTONiXxsNzA7sW+dnJzyR+ttnlmfMfftj4o6bOnp8d5LBvh+XYZ2DXgg4OQyKXgUBMIoM6YhWOglEwCkbBCAIAfblbWcCfmkIAAAAASUVORK5CYII=","orcid":"","institution":"FSEG Sfax, Manouba University","correspondingAuthor":true,"prefix":"","firstName":"Imen","middleName":"","lastName":"Chebbi","suffix":""},{"id":516595513,"identity":"3b567b6a-c189-400b-a475-3bbf9abb8e4f","order_by":1,"name":"Sarra Abidi","email":"","orcid":"","institution":"Manouba University","correspondingAuthor":false,"prefix":"","firstName":"Sarra","middleName":"","lastName":"Abidi","suffix":""},{"id":516595514,"identity":"e0f000f2-c5d7-4d97-9657-9fcb7ecb8c1d","order_by":2,"name":"Leila Ben Ayed","email":"","orcid":"","institution":"Manouba University","correspondingAuthor":false,"prefix":"","firstName":"Leila","middleName":"Ben","lastName":"Ayed","suffix":""}],"badges":[],"createdAt":"2025-09-02 11:23:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7517125/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7517125/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":92142286,"identity":"d5f18c46-6699-4656-8f06-78d90fd619eb","added_by":"auto","created_at":"2025-09-25 06:12:41","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5352340,"visible":true,"origin":"","legend":"","description":"","filename":"MHIDSrgelsiver.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7517125/v1/c10310798599e94926fe547d.pdf"},{"id":92142285,"identity":"dc696442-6dd4-479a-b478-36c13c8759a1","added_by":"auto","created_at":"2025-09-25 06:12:41","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5155,"visible":true,"origin":"","legend":"","description":"","filename":"f081a533c7564a5b9e08980ec02c74eb.json","url":"https://assets-eu.researchsquare.com/files/rs-7517125/v1/e1063602165d28773c86b9d7.json"},{"id":104658526,"identity":"8354738f-e9b1-4fdc-b82b-3ae7bf3292d3","added_by":"auto","created_at":"2026-03-15 12:10:04","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":636706,"visible":true,"origin":"","legend":"","description":"","filename":"MHIDSrgelsiver.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7517125/v1_covered_3e4f7f91-4558-4659-bb7c-c736f66de70b.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Ensuring Safety in Clinical AI: Formally Verified Deep Learning for Heart Failure Detection","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"Multimodal, Hybrid, Injury, Detection, AI, Healthcare Monitoring","lastPublishedDoi":"10.21203/rs.3.rs-7517125/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7517125/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"A major obstacle in fields including sports, clinical rehabilitation, and workplace safety is the timely detection and prevention of physical injuries. The majority of conventional monitoring systems are reactive, depending on post-event analysis or unimodal data sources, which restricts their ability to provide proactive actions and early warnings. Furthermore, current AI-driven health systems lack rigorous validation procedures, which compromises their suitability for practical implementation in safety-critical settings. In this work, we present MHIDS (Multimodal Hybrid Injury Detection System), an integrated, AI-based diagnostic framework that combines wearable physiological sensors, computer vision, and personalized physiological modeling for real-time injury forecasting. A continuously updated digital twin is employed to capture each user’s biomechanical and physiological profile, allowing adaptive, individualized risk assessment. Unlike conventional approaches, MHIDS incorporates a closed-loop feedback mechanism that dynamically reconfigures sensing parameters and provides actionable recommendations (e.g., posture correction, intensity adjustment, or rest scheduling), thereby shifting the paradigm from passive detection to proactive prevention. To guarantee correctness and operational trustworthiness, MHIDS is formally modeled in UPPAAL as a network of timed automata, ensuring critical properties such as bounded response times (\u003c100 ms), safety, liveness, and deadlock freedom. Experimental validation using the publicly available MHEALTH dataset demonstrates superior predictive performance, achieving an accuracy of 99.21%, precision of 98.94%, recall of 99.07%, and F1-score of 99.00%, significantly outperforming state-of-the-art baselines.","manuscriptTitle":"Ensuring Safety in Clinical AI: Formally Verified Deep Learning for Heart Failure Detection","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-25 06:12:36","doi":"10.21203/rs.3.rs-7517125/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":"ae138dea-c35a-41fd-b910-7f3d927c8772","owner":[],"postedDate":"September 25th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-15T12:09:12+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-25 06:12:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7517125","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7517125","identity":"rs-7517125","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}