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

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

  • 0 min

Share

Clinical Report: Utilizing a Deep Learning Approach for PVI Lines in AF Ablation

Overview

This study developed a deep learning model to create individualized pulmonary vein isolation (PVI) lines from pre-ablation 3D voltage maps, achieving high accuracy in reproducing successful lesion sets.

Background

Pulmonary vein isolation (PVI) is a standard procedure for atrial fibrillation (AF) ablation, yet recurrence remains a significant challenge. The variability in anatomical features and operator skill can affect the success of PVI. Advances in deep learning may offer new methods to optimize PVI strategies tailored to individual patient anatomy.

Data Highlights

The model was trained on 513 maps from 171 successful AF ablation cases, achieving a mean Intersection over Union of 0.87 and a Dice score of 0.93 on the test set.

Key Findings

  • The deep learning model utilized a U-Net architecture to generate patient-specific PVI lines.
  • Ground truth was established using lesion sets from cases with documented freedom from AF recurrence for over one year.
  • The model achieved high accuracy metrics, with a mean Intersection over Union of 0.87.
  • Dice score of 0.93 indicates strong agreement between predicted and actual PVI lines.
  • Further prospective clinical validation is necessary to determine the impact on AF recurrence rates.

Clinical Implications

The development of individualized PVI lines using deep learning may enhance the precision of AF ablation procedures.

Conclusion

Future research is needed to validate these findings in clinical settings.

Related Resources & Content

  1. npj Digital Medicine, 2026 -- A deep learning model integrating structured data and clinical text for predicting atrial fibrillation recurrence
  2. Sparse Catheter Pathways for Neural Network-Based Reconstruction of the Left Atrial Structure, 2024
  3. Frontiers in Cardiovascular Medicine, 2026 -- Atrial Fibrillation Type-Specific Prediction of Recurrence After Catheter Ablation
  4. 2023 ACC/AHA/ACCP/HRS Guideline for the Diagnosis and Management of Atrial Fibrillation - American College of Cardiology
  5. Pulsed Field or Conventional Thermal Ablation for Paroxysmal Atrial Fibrillation - American College of Cardiology
  6. Frontiers in Cardiovascular Medicine — Single-catheter radiofrequency pulmonary vein isolation for atrial fibrillation: a comparative evaluation
  7. 2023 ACC/AHA/ACCP/HRS Guideline for the Diagnosis and Management of Atrial Fibrillation - American College of Cardiology
  8. Pulsed Field or Conventional Thermal Ablation for Paroxysmal Atrial Fibrillation - American College of Cardiology
  9. Effect of MRI-Guided Fibrosis Ablation vs Conventional Catheter Ablation on Atrial Arrhythmia Recurrence in Patients With Persistent Atrial Fibrillation: The DECAAF II Randomized Clinical Trial - PMC

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

ASE Plenary: Progress, Priorities, and What’s Next

Efforts to align, sustain, and engage drove a year of growth for ASE.

Self-supervised representation learning reveals explainable physiological structure in high-dimensional magnetocardiography

Overt hyperthyroidism is associated with increased dispersion of ventricular repolarization: a case-control study