{"paper":{"title":"ECG-NAT: A Self-supervised Neighborhood Attention Transformer for Multi-lead Electrocardiogram Classification","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"ECG-NAT uses masked autoencoder pretraining on unlabeled signals and dual-loss fine-tuning to classify multi-lead ECG arrhythmias at 88.1 percent accuracy from only 1 percent labeled data.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Amjad Seyedi, Fardin Akhlaghian Tab, Fatemeh Daneshfar, Mahsa Gazeran, Sayvan Soleymanbaigi","submitted_at":"2026-05-13T08:50:00Z","abstract_excerpt":"Electrocardiogram (ECG) arrhythmia classification remains challenging due to signal variability, noise, limited labeled data, and the difficulty in achieving both accuracy and efficiency in models. While self-supervised learning reduces label dependency, most methods target either global contextual features or local morphological patterns, but rarely implement hierarchical multi-scale feature extraction. ECG signals require architectures that simultaneously capture fine-grained beat-level morphology and broader rhythm-level dependencies with computational efficiency. To overcome this limitatio"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ECG-NAT achieves robust performance on benchmark datasets, with 88.1% accuracy using only 1% labeled data, demonstrating strong efficacy in low-resource settings.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That generative pretraining via masked autoencoder on multiple diverse unlabeled datasets produces robust domain-invariant representations that transfer effectively to the downstream classification task under the dual-loss fine-tuning regime.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ECG-NAT combines masked autoencoder pretraining with hierarchical neighborhood attention and dual-loss fine-tuning to reach 88.1% accuracy on ECG classification using just 1% labeled data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ECG-NAT uses masked autoencoder pretraining on unlabeled signals and dual-loss fine-tuning to classify multi-lead ECG arrhythmias at 88.1 percent accuracy from only 1 percent labeled data.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"25e4a27a76020b741d0c115a6a95814d006a9bc97763846b6f8f52abf122d2cc"},"source":{"id":"2605.13194","kind":"arxiv","version":1},"verdict":{"id":"b202bc09-c4a5-4e96-8346-4df7e308ee19","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T20:37:05.749408Z","strongest_claim":"ECG-NAT achieves robust performance on benchmark datasets, with 88.1% accuracy using only 1% labeled data, demonstrating strong efficacy in low-resource settings.","one_line_summary":"ECG-NAT combines masked autoencoder pretraining with hierarchical neighborhood attention and dual-loss fine-tuning to reach 88.1% accuracy on ECG classification using just 1% labeled data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That generative pretraining via masked autoencoder on multiple diverse unlabeled datasets produces robust domain-invariant representations that transfer effectively to the downstream classification task under the dual-loss fine-tuning regime.","pith_extraction_headline":"ECG-NAT uses masked autoencoder pretraining on unlabeled signals and dual-loss fine-tuning to classify multi-lead ECG arrhythmias at 88.1 percent accuracy from only 1 percent labeled data."},"references":{"count":51,"sample":[{"doi":"","year":2022,"title":"Detection of cardiovascular diseases in ecg images using machine learning and deep learning methods.IEEE transactions on artificial intelligence, 4(2):373–382, 2022","work_id":"8100f560-9dee-4184-8a2a-e4556f333366","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2013,"title":"Bentham Science Publishers, 2013","work_id":"82a6fefd-ba1f-4c2b-b7de-971cf5a81107","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2012,"title":"Patricia Paglini-Oliva, MS Lo Presti, and H Walter Rivarola.Electrocardiography as a diagnos- tic method for Chagas disease in patients and experimental models. InTech, 2012","work_id":"97ed7986-26c9-41a8-98b4-8e3e8d01cbfd","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Convolutional neural network based automatic screening tool for cardiovascular diseases using different intervals of ECG signals","work_id":"f62d1b11-63e4-4974-afb7-fb4d6d652c17","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Chuang Han, Jiajia Sun, Yingnan Bian, Wenge Que, and Li Shi. Automated detection and localization of myocardial infarction with interpretability analysis based on deep learning.IEEE Transactions on In","work_id":"3a9a3a08-54bd-4ace-b12f-b4aec4c0ae4c","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":51,"snapshot_sha256":"27e585e8fa02a31d9f040030187d470d88c1ba3b231a52b185400103c85b4627","internal_anchors":1},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}