Scholarlink

Atrial fibrillation is the most common heart rhythm disorder and is linked to an elevated risk of stroke. Selecting the appropriate treatment for AF remains a significant challenge, but biophysical simulations offer a promising solution by enabling the creation of patient-specific cardiac digital twins for in-silico testing of various treatments. To achieve this, it is essential to have an accurate and efficient tool for segmenting individual atrial structures, including atrial cavities and walls. We proposed a method using a residual encoder nnUNet for segmenting LGE-MRI scans. Our approach excelled in the validation stage of the MBAS challenge, achieving Dice scores of 91.25 for the left atrium cavity, 88.44 for the right atrium cavity, and 67.16 for the atrial walls.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Ensemble of 3D Residual Encoder UNet Models for Solving Multi-class Bi-atrial Segmentation Challenge

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Alexander Zolotarev,
  • Kiane Johnson,
  • Abbas Khan,
  • Gregory Slabaugh,
  • Caroline Roney

摘要

Atrial fibrillation is the most common heart rhythm disorder and is linked to an elevated risk of stroke. Selecting the appropriate treatment for AF remains a significant challenge, but biophysical simulations offer a promising solution by enabling the creation of patient-specific cardiac digital twins for in-silico testing of various treatments. To achieve this, it is essential to have an accurate and efficient tool for segmenting individual atrial structures, including atrial cavities and walls. We proposed a method using a residual encoder nnUNet for segmenting LGE-MRI scans. Our approach excelled in the validation stage of the MBAS challenge, achieving Dice scores of 91.25 for the left atrium cavity, 88.44 for the right atrium cavity, and 67.16 for the atrial walls.