Clinical-Grade EEG Seizure Detection Achieving 93.9% Precision: First Hospital-Deployable System Through Multi-Domain Feature Engineering and Ensemble Deep Learning

Clinical-Grade EEG Seizure Detection Achieving 93.9% Precision: First Hospital-Deployable System Through Multi-Domain Feature Engineering and Ensemble Deep Learning | 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 Clinical-Grade EEG Seizure Detection Achieving 93.9% Precision: First Hospital-Deployable System Through Multi-Domain Feature Engineering and Ensemble Deep Learning Mohammed Irshad This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7271600/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 Automated electroencephalogram (EEG) seizure detection remains limited by precision rates of 35-70%, preventing clinical deployment due to excessive false alarms. This study presents the first automated system to achieve clinical-grade precision (93.9%) on the CHB-MIT dataset, directly addressing the deployment barrier that has prevented hospital implementation. Our methodology integrates an unprecedented 2,138-dimensional multi-domain feature space combining time-frequency analysis, spectral characteristics, statistical measures , and connectivity patterns, processed through a three-model ensemble (Enhanced XGBoost, LightGBM, Transformer-GRU) with focal loss optimization. Patient-wise cross-validation demonstrates exceptional performance: F1-score of 0.948, precision of 93.9%, recall of 95.8%, and AUC of 0.998. Comprehensive ablation studies reveal that multi-domain feature engineering contributes 57% of performance gains, SMOTE class balancing adds 35%, ensemble methodology provides 16%, and focal loss optimization delivers final clinical-grade refinement. The system achieves 6.1% false positive rate (0.89 false alarms/hour) with <100ms inference latency, meeting stringent clinical deployment requirements. This breakthrough represents the first clinically-viable automated seizure detection system suitable for continuous hospital monitoring, with complete code availability ensuring reproducibility. Biological sciences/Computational biology and bioinformatics Physical sciences/Engineering Health sciences/Neurology EEG seizure detection machine learning clinical deployment ensemble learning biomedical signal processing artificial intelligence in healthcare 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-7271600","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":557640230,"identity":"16df8eb4-d0ba-48a2-a664-95c77d135853","order_by":0,"name":"Mohammed Irshad","email":"data:image/png;base64,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","orcid":"","institution":"Yenepoya Institute of Technology Visvesvaraya Technological University","correspondingAuthor":true,"prefix":"","firstName":"Mohammed","middleName":"","lastName":"Irshad","suffix":""}],"badges":[],"createdAt":"2025-08-01 13:23:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7271600/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7271600/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97982011,"identity":"0c7c56b3-6c99-4dba-9ce2-d640831ca45c","added_by":"auto","created_at":"2025-12-11 12:56:16","extension":"png","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":781172,"visible":true,"origin":"","legend":"","description":"","filename":"clinicalvalidationdashboard1.png","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1/91e98fae84d846ababe0918b.png"},{"id":97982016,"identity":"b0601c02-d9b3-474b-8526-b36fdcb1bd26","added_by":"auto","created_at":"2025-12-11 12:56:16","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3443304,"visible":true,"origin":"","legend":"","description":"","filename":"finalpaper.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1/fe8f8c212b5dbdb559af1f3d.pdf"},{"id":97982013,"identity":"4b9e4ee9-9537-4da7-87f3-364d7e532fb0","added_by":"auto","created_at":"2025-12-11 12:56:16","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":888559,"visible":true,"origin":"","legend":"","description":"","filename":"eegsystemarchitecture.png","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1/2bf02e2cd32b11858cd13db5.png"},{"id":97982017,"identity":"21e104f2-90ce-4392-913e-0777a96238b7","added_by":"auto","created_at":"2025-12-11 12:56:16","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":670969,"visible":true,"origin":"","legend":"","description":"","filename":"featureengineeringpipeline.png","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1/2180e0e61f192346ab8d4a5f.png"},{"id":97982018,"identity":"f7a5ee0a-4c2a-4190-bcbf-113fef197376","added_by":"auto","created_at":"2025-12-11 12:56:16","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":429873,"visible":true,"origin":"","legend":"","description":"","filename":"modelarchitectures.png","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1/cba8ff0bc0db39a9006e4cc1.png"},{"id":98425200,"identity":"4b3904f2-2421-492e-9c6f-8a60cef22b46","added_by":"auto","created_at":"2025-12-17 16:34:31","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":816968,"visible":true,"origin":"","legend":"","description":"","filename":"performancedashboard.png","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1/822a3f7cae6f087e402721e7.png"},{"id":98425185,"identity":"f837193b-7ce5-4e18-bf46-a22e5435959c","added_by":"auto","created_at":"2025-12-17 16:34:30","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":534570,"visible":true,"origin":"","legend":"","description":"","filename":"resultsinfographic.png","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1/cb3e7c85e62692e6b09dd53b.png"},{"id":97982012,"identity":"991c3d40-c0eb-4f41-8faa-e0482ae183df","added_by":"auto","created_at":"2025-12-11 12:56:16","extension":"json","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3589,"visible":true,"origin":"","legend":"","description":"","filename":"590c4faf81e541d8849eeba839b5b78a.json","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1/dd7a20a431d8c80fbba99963.json"},{"id":98444323,"identity":"534a7fa4-5673-43f1-a56b-e214107dbccf","added_by":"auto","created_at":"2025-12-17 17:15:34","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4694278,"visible":true,"origin":"","legend":"","description":"","filename":"finalpaper.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7271600/v1_covered_9997252a-f7cd-40ea-bb70-1500a95c0c4d.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Clinical-Grade EEG Seizure Detection Achieving 93.9% Precision: First Hospital-Deployable System Through Multi-Domain Feature Engineering and Ensemble Deep Learning","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":"EEG seizure detection, machine learning, clinical deployment, ensemble learning, biomedical signal processing, artificial intelligence in healthcare","lastPublishedDoi":"10.21203/rs.3.rs-7271600/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7271600/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Automated electroencephalogram (EEG) seizure detection remains limited by precision rates of 35-70%, preventing clinical deployment due to excessive false alarms. This study presents the first automated system to achieve clinical-grade precision (93.9%) on the CHB-MIT dataset, directly addressing the deployment barrier that has prevented hospital implementation. Our methodology integrates an unprecedented 2,138-dimensional multi-domain feature space combining time-frequency analysis, spectral characteristics, statistical measures , and connectivity patterns, processed through a three-model ensemble (Enhanced XGBoost, LightGBM, Transformer-GRU) with focal loss optimization. Patient-wise cross-validation demonstrates exceptional performance: F1-score of 0.948, precision of 93.9%, recall of 95.8%, and AUC of 0.998. Comprehensive ablation studies reveal that multi-domain feature engineering contributes 57% of performance gains, SMOTE class balancing adds 35%, ensemble methodology provides 16%, and focal loss optimization delivers final clinical-grade refinement. The system achieves 6.1% false positive rate (0.89 false alarms/hour) with \u003c100ms inference latency, meeting stringent clinical deployment requirements. This breakthrough represents the first clinically-viable automated seizure detection system suitable for continuous hospital monitoring, with complete code availability ensuring reproducibility.","manuscriptTitle":"Clinical-Grade EEG Seizure Detection Achieving 93.9% Precision: First Hospital-Deployable System Through Multi-Domain Feature Engineering and Ensemble Deep Learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-11 12:56:12","doi":"10.21203/rs.3.rs-7271600/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":"22fe1f46-a344-4e8a-b7ce-68e1a03b1d90","owner":[],"postedDate":"December 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":59351315,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":59351316,"name":"Physical sciences/Engineering"},{"id":59351317,"name":"Health sciences/Neurology"}],"tags":[],"updatedAt":"2025-12-17T11:09:15+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-11 12:56:12","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7271600","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7271600","identity":"rs-7271600","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}