A Dataset and Benchmarks for Atrial Fibrillation Detection from Electrocardiograms of Intensive Care Unit Patients

2025-12-19

AI Summary

This study addresses the challenge of automated atrial fibrillation (AF) detection in ICU patients. We construct the first publicly available, high-quality, ICU-specific ECG dataset with expert-annotated AF episodes. Building upon this benchmark, we systematically evaluate three methodological paradigms—handcrafted feature engineering, deep learning models, and electrocardiogram foundation models (ECG-FMs)—under zero-shot and transfer learning settings. Notably, this work presents the first successful adaptation of ECG-FMs to the ICU domain, demonstrating their superiority over conventional approaches: the ECG-FM-based transfer learning method achieves an F1-score of 0.89 on the ICU test set and exhibits the strongest cross-dataset generalization. Our key contributions are: (1) releasing the first ICU-specific AF benchmark dataset; (2) establishing the first systematic comparison between ECG-FMs and traditional methods in critical care; and (3) identifying ECG-FM transfer learning as the current state-of-the-art paradigm for ICU AF detection.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Cognitive Modeling & Cognitive Systems: Affective Computing
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Large-scale security measurements
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Objective: Atrial fibrillation (AF) is the most common cardiac arrhythmia experienced by intensive care unit (ICU) patients and can cause adverse health effects. In this study, we publish a labelled ICU dataset and benchmarks for AF detection. Methods: We compared machine learning models across three data-driven artificial intelligence (AI) approaches: feature-based classifiers, deep learning (DL), and ECG foundation models (FMs). This comparison addresses a critical gap in the literature and aims to pinpoint which AI approach is best for accurate AF detection. Electrocardiograms (ECGs) from a Canadian ICU and the 2021 PhysioNet/Computing in Cardiology Challenge were used to conduct the experiments. Multiple training configurations were tested, ranging from zero-shot inference to transfer learning. Results: On average and across both datasets, ECG FMs performed best, followed by DL, then feature-based classifiers. The model that achieved the top F1 score on our ICU test set was ECG-FM through a transfer learning strategy (F1=0.89). Conclusion: This study demonstrates promising potential for using AI to build an automatic patient monitoring system. Significance: By publishing our labelled ICU dataset (LinkToBeAdded) and performance benchmarks, this work enables the research community to continue advancing the state-of-the-art in AF detection in the ICU.

Problem

• Develops a labelled ICU dataset for atrial fibrillation detection
• Compares AI approaches to identify best method for AF detection
• Publishes benchmarks to advance AF detection research in ICU

Innovation

• ECG foundation models with transfer learning
• Deep learning vs feature-based classifiers comparison
• Labelled ICU dataset for AF detection benchmarks