HybridEEGNet: Combining Spatial Attention and Temporal Convolutions with Transformer Encoders for Automated Alzheimer's Disease Detection from EEG Signals

HybridEEGNet: Combining Spatial Attention and Temporal Convolutions with Transformer Encoders for Automated Alzheimer's Disease Detection from EEG Signals | 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 HybridEEGNet: Combining Spatial Attention and Temporal Convolutions with Transformer Encoders for Automated Alzheimer's Disease Detection from EEG Signals Lev Kung, Brandon Yee, Maximilian Rutkowski This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8751868/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 Alzheimer’s disease (AD) is a progressive neurodegenerative disorder affecting over 55 million people worldwide, with early detection being crucial for effective intervention and care planning. Electroencephalography (EEG) offers a non-invasive, cost-effective, and widely accessible approach for AD screening, but traditional analysis methods often fail to capture the complex spatial-temporal patterns indicative of cognitive decline. In this paper, we present HybridEEGNet, a novel deep learning architecture that synergistically combines spatial attention mechanisms, multi-scale temporal convolutions with dilated kernels, and Transformer encoders for automated AD detection from resting-state EEG recordings. Our approach addresses the unique challenges of EEG analysis through a three-stage processing pipeline: (1) spatial attention to learn channel-specific importance reflecting regional brain activity patterns, (2) multi-scale dilated convolutions to extract local temporal features across different frequency resolutions, and (3) self-attention mechanisms to capture longrange temporal dependencies characteristic of neural synchronization patterns. We conduct comprehensive experiments on the PhysioNet EEG dataset, systematically comparing our approach against multiple baseline architectures including 1D-CNN, 2D-CNN, LSTM, and pure Transformer models using consistent preprocessing and evaluation protocols. Our HybridCNNTransformer achieves state-of-the-art performance with 72.9% accuracy, 81.1% F1-score, 91.1% recall, and 69.3% ROCAUC, demonstrating significant improvements over all baselines. We provide extensive ablation studies quantifying the contribution of each architectural component, attention visualizations for model interpretability aligned with known AD biomarkers, cross-validation analysis for robustness assessment, and detailed computational efficiency comparisons. The high recall achieved by our model is particularly relevant for clinical screening applications where minimizing false negatives is critical. We discuss the clinical implications, limitations, and future directions for EEG-based AD detection. Our code, trained models, and experimental configurations are publicly available at https://github.com/YCRG-Labs/alzheimers-eeg to facilitate reproducibility and accelerate future research in this important domain. Bioinformatics Artificial Intelligence and Machine Learning Computational Neuroscience Alzheimer's Disease Electroencephalography Deep Learning Transformer Convolutional Neural Networks Attention Mechanisms Medical Diagnosis Neural Signal Processing Brain-Computer Interface Full Text Additional Declarations The authors declare no competing interests. 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-8751868","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":583602552,"identity":"27954c3a-9c26-4c60-b16f-a0e13341df9f","order_by":0,"name":"Lev Kung","email":"","orcid":"","institution":"Yee Collins Research Group","correspondingAuthor":false,"prefix":"","firstName":"Lev","middleName":"","lastName":"Kung","suffix":""},{"id":583602553,"identity":"d0f40b32-e9fa-4465-b8e1-5ffb8bc10c84","order_by":1,"name":"Brandon Yee","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6UlEQVRIiWNgGAWjYBACAzBZASYZDzA2EK3lDIRDghbGNlK0mDPwGD74OO+wvHn/AoYDP3fUMfBLH7+AV4tlA4+x4cxthw3n3HjAcLD3zGEGyb6cAvwOO8BjJs277XCChMT5Dwd42w4wGJzhSSCkxfz33zkgLQcYDv5tqyNKixkzYwNQC38Dw2HeNmagFvYD+P3SzFYs2XMs3XCGBAPDYdkzh3kke3jw6mAwZ2/e+OFHjbW8BP8Bxodvd9TJ8fOwP8Cvh5kDFDXNDAwSEB8AreAxwK+FAWxmHQMD/wEUkVEwCkbBKBgFcAAAlaJIQauKuysAAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0003-3948-1403","institution":"Yee Collins Research Group","correspondingAuthor":true,"prefix":"","firstName":"Brandon","middleName":"","lastName":"Yee","suffix":""},{"id":583602554,"identity":"7b236b74-71c8-4cb0-bbb4-d5166253f684","order_by":2,"name":"Maximilian Rutkowski","email":"","orcid":"","institution":"Yee Collins Research Group","correspondingAuthor":false,"prefix":"","firstName":"Maximilian","middleName":"","lastName":"Rutkowski","suffix":""}],"badges":[],"createdAt":"2026-01-31 18:15:49","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-8751868/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8751868/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101754018,"identity":"038f9d69-f2d5-4b94-8764-c25b87b0aeb0","added_by":"auto","created_at":"2026-02-03 10:41:22","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":350220,"visible":true,"origin":"","legend":"","description":"","filename":"levproject2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8751868/v1_covered_ccc86413-f86e-45b4-89b1-4d9a46a5caeb.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eHybridEEGNet: Combining Spatial Attention and Temporal Convolutions with Transformer Encoders for Automated Alzheimer's Disease Detection from EEG Signals\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Yee Collins Research Group","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":"Alzheimer's Disease, Electroencephalography, Deep Learning, Transformer, Convolutional Neural Networks, Attention Mechanisms, Medical Diagnosis, Neural Signal Processing, Brain-Computer Interface","lastPublishedDoi":"10.21203/rs.3.rs-8751868/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8751868/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAlzheimer’s disease (AD) is a progressive neurodegenerative disorder affecting over 55 million people worldwide, with early detection being crucial for effective intervention and care planning. Electroencephalography (EEG) offers a non-invasive, cost-effective, and widely accessible approach for AD screening, but traditional analysis methods often fail to capture the complex spatial-temporal patterns indicative of cognitive decline. In this paper, we present HybridEEGNet, a novel deep learning architecture that synergistically combines spatial attention mechanisms, multi-scale temporal convolutions with dilated kernels, and Transformer encoders for automated AD detection from resting-state EEG recordings. Our approach addresses the unique challenges of EEG analysis through a three-stage processing pipeline: (1) spatial attention to learn channel-specific importance reflecting regional brain activity patterns, (2) multi-scale dilated convolutions to extract local temporal features across different frequency resolutions, and (3) self-attention mechanisms to capture longrange temporal dependencies characteristic of neural synchronization patterns. We conduct comprehensive experiments on the PhysioNet EEG dataset, systematically comparing our approach against multiple baseline architectures including 1D-CNN, 2D-CNN, LSTM, and pure Transformer models using consistent preprocessing and evaluation protocols. Our HybridCNNTransformer achieves state-of-the-art performance with 72.9% accuracy, 81.1% F1-score, 91.1% recall, and 69.3% ROCAUC, demonstrating significant improvements over all baselines. We provide extensive ablation studies quantifying the contribution of each architectural component, attention visualizations for model interpretability aligned with known AD biomarkers, cross-validation analysis for robustness assessment, and detailed computational efficiency comparisons. The high recall achieved by our model is particularly relevant for clinical screening applications where minimizing false negatives is critical. We discuss the clinical implications, limitations, and future directions for EEG-based AD detection. Our code, trained models, and experimental configurations are publicly available at https://github.com/YCRG-Labs/alzheimers-eeg to facilitate reproducibility and accelerate future research in this important domain.\u003c/p\u003e","manuscriptTitle":"HybridEEGNet: Combining Spatial Attention and Temporal Convolutions with Transformer Encoders for Automated Alzheimer's Disease Detection from EEG Signals","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-03 06:22:54","doi":"10.21203/rs.3.rs-8751868/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":"cd1f2f8a-8973-41ea-8c51-256d5dc3455b","owner":[],"postedDate":"February 3rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":62093015,"name":"Bioinformatics"},{"id":62093016,"name":"Artificial Intelligence and Machine Learning"},{"id":62093017,"name":"Computational Neuroscience"}],"tags":[],"updatedAt":"2026-02-03T06:22:55+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-03 06:22:54","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8751868","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8751868","identity":"rs-8751868","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}