Artificial intelligence for individualized treatment of persistent atrial fibrillation: a randomized controlled trial.

Summary

In a multicenter double-blind RCT (n=370), AI-guided targeting of spatio-temporal dispersion areas in addition to PVI achieved higher 12-month AF freedom (88% vs 70%) than PVI alone after a single procedure, with similar safety but longer procedure times. Freedom from any atrial arrhythmia was not different, suggesting subsequent organized atrial tachycardias may arise.

Key Findings

• AI-guided tailored ablation plus PVI achieved 88% AF freedom at 12 months vs 70% with PVI alone (superiority; log-rank P<0.0001).
• No difference in freedom from any atrial arrhythmia after one ablation, implying subsequent organized atrial tachycardias.
• Safety outcomes were similar between arms, but procedure and ablation times were approximately doubled in the AI-tailored arm.

Clinical Implications

AI-guided dispersion mapping may be adopted as an adjunct to PVI for persistent AF in experienced centers, balancing improved AF freedom against longer procedures and planning for potential post-ablation atrial tachycardias.

Limitations

• Single-procedure 12-month follow-up; longer-term durability and repeat procedures not assessed
• Longer procedure times in AI arm and no difference in any atrial arrhythmia freedom

Future Directions

Evaluate long-term durability, impact on organized atrial tachycardias, cost-effectiveness, and generalizability across centers and operators; integration with other mapping modalities.