ChatPPG.
Dataset

MIT-BIH Atrial Fibrillation Database (PhysioNet)

ECG
Get data DOI Cite

The MIT-BIH Atrial Fibrillation Database contains 25 long-term ECG recordings from people with atrial fibrillation, mostly paroxysmal. Each recording is 10 hours long with two ECG channels sampled at 250 samples per second and 12-bit resolution, and rhythm annotations mark the AF episodes. It was published on PhysioNet by George Moody and Roger Mark and is a standard benchmark for AF detection.

  • Recordings: 25 x 10 h
  • Sampling: 250 Hz, 12-bit
  • Labels: rhythm annotations (AF episodes)
  • Hosted: PhysioNet (open access)
At a glance
Type
Dataset
Host
Not reported
DOI
10.13026/C2MW2D (opens in a new tab)
Licence
Not reported
Access
Not reported
Subjects
Not reported
Signals
ECG
PPG sites
None
Setting
In the wild, Clinical
Wrist
Not reported
Raw waveforms
Not reported
Tasks
AFib
Updated
Oct 16, 2026
MIT-BIH Atrial Fibrillation Database (PhysioNet): ChatPPG research summary card

How it works

How MIT-BIH Atrial Fibrillation Database (PhysioNet) works: Devices & body sites → Signals recorded → TasksDevices & body sites: None (PPG site). Signals recorded: MIT-BIH Atrial Fibrillation Database (PhysioNet), signals ECG. Tasks: AF / rhythm.Devices & body sitesNone: PPG siteNonePPG siteSignals recordedMIT-BIH Atrial Fibrillation…ECGTasksAF / rhythmAF / rhythm

No model in the hub reports results on this dataset yet.

Device roles
Clinical reference (ECG)

Models trained or evaluated on it

No model in the hub reports results on MIT-BIH Atrial Fibrillation Database (PhysioNet) yet. Be the first: results submitted with a paper are added here.

Cite

@misc{afdb2026,
  title = {{MIT-BIH Atrial Fibrillation Database (PhysioNet)}},
  year = {2026},
  doi = {10.13026/C2MW2D},
  url = {https://physionet.org/content/afdb/1.0.0/}
}

History

  1. Record updated

More on ChatPPG