Deep learning model to generate patient-specific pulmonary vein isolation lines from successful atrial fibrillation ablation cases: a proof-of-concept study - Summary - MDSpire
Coming Soon: Introducing MDSpire News. Learn more
Conexiant’s news site is now MDSpire News. Learn more

Utilizing a Deep Learning Approach to Create Individualized Pulmonary Vein Isolation Lines from Successful Atrial Fibrillation Ablation Cases: A Proof-of-Concept Investigation

  • 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

Share

Objective:

To develop a deep learning model that generates patient-specific pulmonary vein isolation (PVI) lines from pre-ablation 3D voltage maps based on successful atrial fibrillation (AF) ablation cases.

Approach:
  • Study Design: A retrospective, multicenter study involving 1,969 patients who underwent catheter ablation for AF, with data collected from April 2016 to March 2023.
  • Model Development: A U-Net-based deep learning model was trained on 513 3D voltage maps from 171 patients with documented freedom from AF recurrence for over one year.
  • Performance Metrics: The model's performance was evaluated using mean Intersection over Union (IoU) and Dice score, achieving 0.87 and 0.93, respectively.
Key Findings:
  • The deep learning model successfully reproduced anatomical and electrophysiological features of PVI lesion sets.
  • The model achieved a mean Intersection over Union of 0.87 and a Dice score of 0.93 on the test set.
Interpretation:

Limitations:
  • The study is a proof-of-concept without external validation.
  • Inter-operator variability in ground-truth annotations was not quantitatively assessed.
Conclusion:

The study demonstrates the feasibility of using deep learning to create individualized PVI lines, warranting further investigation for clinical application.

Sources:

Original Source(s)

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

  • 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

Related Content

Case report: Beyond arrhythmia: STAR-inspired single-fraction cardiac radioablation for Ewing sarcoma metastasis

May Lower RFR Signal Repeat Revascularization?

Patients with lower nonischemic resting full-cycle ratio values had more repeat revascularizations despite similar 2-year major cardiovascular event rates.

Segmental analysis of left atrial substrate severity predictors in patients undergoing persistent atrial fibrillation ablation