Scholarlink

The aorta is the largest artery of the body, pathology of the aorta and its main branches can be immediate threats to life or limb. Efficient and accurate segmentation of the aorta and its branches is conducive to assisting doctors to develop more appropriate diagnosis and treatment plans. Based on nnU-Net, we introduce a Channel-wised Sparse Self-Attention to enhance the network ability for tubular structure, and adopt an more efficient sliding window inference strategy to perform aorta segmentation. Experiments were conducted on AortaSeg24 challenge dataset. Our method achieves the mean DSC of 0.750 on 10 unseen validation images, and 0.752 on 40 unseen testing images. Codes are at https://github.com/Kaixiang-Yang/AortaSeg24 .

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

Accurate and Efficient Multi-class Segmentation for Aortic Branches and Zones in CTA

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Kaixiang Yang,
  • Yuxuan Luo,
  • Qiang Li,
  • Zhiwei Wang

摘要

The aorta is the largest artery of the body, pathology of the aorta and its main branches can be immediate threats to life or limb. Efficient and accurate segmentation of the aorta and its branches is conducive to assisting doctors to develop more appropriate diagnosis and treatment plans. Based on nnU-Net, we introduce a Channel-wised Sparse Self-Attention to enhance the network ability for tubular structure, and adopt an more efficient sliding window inference strategy to perform aorta segmentation. Experiments were conducted on AortaSeg24 challenge dataset. Our method achieves the mean DSC of 0.750 on 10 unseen validation images, and 0.752 on 40 unseen testing images. Codes are at https://github.com/Kaixiang-Yang/AortaSeg24 .