Deep learning model to generate patient-specific pulmonary vein isolation lines from successful atrial fibrillation ablation cases: a proof-of-concept study

1

A deep learning model was developed to generate patient-specific pulmonary vein isolation lines from pre-ablation 3D voltage maps.
2

The model was trained on lesion sets from 171 successful atrial fibrillation ablation cases with over one year of documented freedom from recurrence.
3

Using a U-Net-based architecture, the model achieved a mean Intersection over Union of 0.87 and a Dice score of 0.93 on the test set.
4

The study involved a retrospective analysis of 1,969 patients who underwent catheter ablation for atrial fibrillation.
5

This proof-of-concept study suggests the model can reproduce patient-specific PVI patterns, pending prospective clinical validation.

Frontiers In Cardiovascular Medicine

by Kazuo Sakamoto, Takeshi Tohyama, Hirotake Yokoyama, Tsukasa Watanabe, Tomomi Nagayama, Yasushi Mukai, Shunsuke Kawai, Daisuke Yakabe, Hiroshi Mannoji, Kazuhiro Nagaoka, Atsushi Tanaka, Mitsutaka Yamamoto, Kiyohiro Ogawa, Takeshi Mikami, Shujiro Inoue, Susumu Takase, Kei Inoue, Kazuya Hosokawa, Koji Todaka, Hiroyuki Tsutsui, Kohtaro Abe
July 6, 2026

1