Scholarlink

<p>The accurate segmentation of brain glioma subregions is an important technical prerequisite for clinical diagnosis and treatment and has important research significance. Conventional convolutional neural networks generally demonstrate limitations in explicitly modeling long-range dependency, which is not conducive to accurate segmentation. This paper proposes a brain glioma segmentation model based on the encoder-decoder structure (DEUNet), inspired by the U-Net model. The proposed model consists of a two-branch encoder, a dual-path feature fusion module based on the attention mechanism (DPAM) and a decoder. The proposed method is evaluated on the public BraTS2021 dataset. The experiments showed that the suggested model achieved a competitive dice score in comparison to recent advanced methods. The proposed method reached the Dice coefficient of 91.6%, 83.0%, and 85.1% in the WT, ET, and TC regions, respectively on the BraTS2021 dataset.</p>

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

DEUNet: a dual encoder network based on multimodal MRI images for brain glioma segmentation

  • 期刊论文

  • 发表日期：

    2025年9月23日

查看全文
  • Fuqiang You,
  • Yaohui Zhou,
  • Jingjie Liu

摘要

The accurate segmentation of brain glioma subregions is an important technical prerequisite for clinical diagnosis and treatment and has important research significance. Conventional convolutional neural networks generally demonstrate limitations in explicitly modeling long-range dependency, which is not conducive to accurate segmentation. This paper proposes a brain glioma segmentation model based on the encoder-decoder structure (DEUNet), inspired by the U-Net model. The proposed model consists of a two-branch encoder, a dual-path feature fusion module based on the attention mechanism (DPAM) and a decoder. The proposed method is evaluated on the public BraTS2021 dataset. The experiments showed that the suggested model achieved a competitive dice score in comparison to recent advanced methods. The proposed method reached the Dice coefficient of 91.6%, 83.0%, and 85.1% in the WT, ET, and TC regions, respectively on the BraTS2021 dataset.