{"paper":{"title":"Dynamical Predictive Modelling of Cardiovascular Disease Progression Post-Myocardial Infarction via ECG-Trained Artificial Intelligence Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Pretraining ECG models with patient-specific temporal contrastive learning raises post-MI outcome prediction AUC from 0.608 to 0.794 in small-data settings.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Adelaide de Vecchi, Andrew King, Lupo Lovatelli, Oleg Aslanidi, Riccardo Cavarra, Shaheim Ogbomo-Harmitt, Shahid Aziz","submitted_at":"2026-05-13T14:05:43Z","abstract_excerpt":"Myocardial infarction (MI) is a leading cause of death, and its adverse outcomes are urgent to predict. Yet ECG-based prognostic models underperform because deep learning requires large, labelled datasets, which are scarce in medicine. Foundation models can learn from unlabelled ECGs via selfsupervision, but medically relevant training strategies remain underexplored. We propose a pretrained artificial intelligence model that combines patient-specific temporal information using contrastive learning with supervised multitask heads, then fine-tunes on post-MI outcome prediction. The proposed mod"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The proposed model outperformed a model trained from scratch (0.794 vs 0.608 AUC) showing that clinically structured ECG modelling improves classification in limited data regimes.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the contrastive pretraining objective with patient-specific temporal information extracts features that are genuinely predictive of post-MI clinical outcomes rather than dataset-specific artifacts.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A contrastive-learning ECG foundation model with multitask heads predicts post-MI outcomes better than training from scratch (AUC 0.794 vs 0.608).","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Pretraining ECG models with patient-specific temporal contrastive learning raises post-MI outcome prediction AUC from 0.608 to 0.794 in small-data settings.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"7f5929c37a869e2775addc2aeca22a820cb2e4810d73281de0ffc572e1f8fb10"},"source":{"id":"2605.13568","kind":"arxiv","version":1},"verdict":{"id":"cca5171e-8263-499e-ac84-74e823370750","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:46:01.128415Z","strongest_claim":"The proposed model outperformed a model trained from scratch (0.794 vs 0.608 AUC) showing that clinically structured ECG modelling improves classification in limited data regimes.","one_line_summary":"A contrastive-learning ECG foundation model with multitask heads predicts post-MI outcomes better than training from scratch (AUC 0.794 vs 0.608).","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the contrastive pretraining objective with patient-specific temporal information extracts features that are genuinely predictive of post-MI clinical outcomes rather than dataset-specific artifacts.","pith_extraction_headline":"Pretraining ECG models with patient-specific temporal contrastive learning raises post-MI outcome prediction AUC from 0.608 to 0.794 in small-data settings."},"references":{"count":9,"sample":[{"doi":"10.1055/a-2761-","year":2026,"title":"Digital Twins for Predictive Modelling of Thrombosis and Stroke Risk: Current Approaches and Future Directions,","work_id":"681d9365-b862-4b77-9518-5b8907e73e24","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1093/eurheartj/ehad191","year":2023,"title":"2023 ESC Guidelines for the management of acute coronary syndromes,","work_id":"d9c1cdcc-bca7-4857-850a-1363942c9365","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1093/eurheartj/ehaf784.4422","year":2025,"title":"Explainable machine learning models to improve prediction of incident stroke in atrial fibrillation patients using health records, medical imaging and ECG derived metrics,","work_id":"66fc3967-422a-4b79-b033-159d521e9ab8","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1109/lsp.2021.3114119","year":1993,"title":"CLECG: A Novel Contrastive Learning Framework for Electrocardiogram Arrhythmia Classification,","work_id":"e1b7911e-5f9e-4152-bff5-de4a0016212a","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.48550/arxiv.2002.05709","year":2020,"title":"A Simple Framework for Contrastive Learning of Visual Representations","work_id":"77d995ce-c44e-4692-9c54-cf8ce771464a","ref_index":5,"cited_arxiv_id":"2002.05709","is_internal_anchor":true}],"resolved_work":9,"snapshot_sha256":"d6af76ecb9521fa70f90f79c9c494e13d7457e4df473dd5e44148331119c5b29","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"64ff867c747381ebd39530ce5c0033afdb5021c459ba71d714f278157a2e84a0"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}