Can ChatGPT accurately detect atrial fibrillation using smartwatch ECG?

Area of Science:

• Cardiology
• Artificial Intelligence
• Digital Health

Background:

• Physician confirmation is required for smartwatch-detected atrial fibrillation (AF), increasing healthcare burdens.
• The latest AI model, ChatGPT-4o (GPT-4o), offers advanced image analysis capabilities.

Purpose of the Study:

• To evaluate the diagnostic performance of GPT-4o in identifying AF from smartwatch-recorded electrocardiograms (ECGs).

Main Methods:

• Single-lead ECGs were recorded using Apple Watch Series 6® from 114 patients with AF and 60 controls with sinus rhythm (SR).
• Two cardiologists independently classified the smartwatch ECGs.
• GPT-4o analyzed all smartwatch ECG recordings.

Main Results:

• GPT-4o correctly analyzed 47.3% of AF and 71.6% of SR smartwatch ECGs.
• The Apple Watch algorithm achieved 97.3% accuracy for AF and 100% for SR.
• No specific AF characteristics predicted GPT-4o's diagnostic accuracy.

Conclusions:

• Current GPT-4o capabilities are insufficient for reliable AF diagnosis from smartwatch ECGs.
• Human expertise remains indispensable for accurate AF detection.
• Consumers should be aware of the limitations of AI in diagnosing cardiac conditions.