MIT‑Physio AFib ECG Database

The MIT-Physio AFib ECG Database is a comprehensive integrated resource that combines two of the most frequently used datasets for atrial fibrillation research: the MIT‑BIH AFib Database and the PhysioNet/Computing in Cardiology Challenge 2017 dataset. This resource includes 25 long-term 10‑hour recordings with dual-channel ECG signals (recorded at 250 Hz with 12‑bit resolution over ±10 mV) as well as short single‑lead ECG recordings (ranging from 30 to 60 seconds at 300 Hz). Each recording is accompanied by detailed rhythm annotations (.atr files) and automated beat annotations (.qrs files), with some records including manually corrected beat annotations (.qrsc files).

The dataset is designed to support robust, state‑of‑the‑art research in atrial fibrillation (AF) detection by providing high‑quality, well‑annotated data that captures the complexity and variability of ECG signals. Researchers using this resource are encouraged to address challenges such as noise contamination, inter‑patient variability, and the episodic nature of AF. When using the MIT‑Physio AFib ECG Database, please cite Moody and Mark (1983) and include the standard PhysioNet citation (Goldberger et al., 2000). This integrated dataset offers a unique and scalable platform for developing and validating advanced AF detection algorithms.