Scholarlink

Studies have shown that Perirenal Fat (PRF) is a key potential factor in the study and treatment of Metabolic Syndrome. Very recently, the Denoising Diffusion Probabilistic Model has achieved remarkable performance and shows great potential for various image generation tasks, including medical image segmentation. In this paper, we are the first to propose applying a diffusion model to the segmentation task of PRF in CT images. In order to acquire more contextual and local information and enhance positional correlation, we propose the method of incorporating relative position encoding in the diffusion model, called RPE-Diff. Compared against other state-of-the-art methods, the experimental results show that RPE-Diff has the best performance in PRF segmentation and demonstrates the effectiveness of RPE-Diff.

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

RPE-Diff: A Relative Position Encoding Diffusion Model for Perirenal Fat Segmentation in Metabolic Syndrome

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shuai Ye,
  • Tianming Du,
  • Frank Kulwa,
  • Xiangyu Meng,
  • Md Mamunur Rahaman,
  • Marcin Grzegorzek,
  • Ning Xu,
  • Tao Jiang,
  • Hongzan Sun,
  • Chen Li

摘要

Studies have shown that Perirenal Fat (PRF) is a key potential factor in the study and treatment of Metabolic Syndrome. Very recently, the Denoising Diffusion Probabilistic Model has achieved remarkable performance and shows great potential for various image generation tasks, including medical image segmentation. In this paper, we are the first to propose applying a diffusion model to the segmentation task of PRF in CT images. In order to acquire more contextual and local information and enhance positional correlation, we propose the method of incorporating relative position encoding in the diffusion model, called RPE-Diff. Compared against other state-of-the-art methods, the experimental results show that RPE-Diff has the best performance in PRF segmentation and demonstrates the effectiveness of RPE-Diff.