Scholarlink

<p>Image segmentation is critical in medical image analysis, enabling accurate identification and localization of key regions within images. This accuracy is indispensable for reliable diagnosis and effective treatment planning. To achieve an outstanding model performance with moderate parameters, a context feature extraction network for medical image segmentation (abbreviated as CFSeg-Net) has been proposed. The proposed CFSeg-Net is designed to enhance the understanding of the context features. The CFSeg-Net primarily comprises two components: the context feature extraction (CFE) module and feature enhancement (FE) module. The CFE module leverages a multi-range perception scheme, channel shuffle, and multi-feature integration scheme to achieve a broader receptive field. A broader receptive field leads to a better understanding of context features and better segmentation performance. Notably, model performance can be further improved when the FE module is inserted between two CFE modules in the encoder. Extensive experiments have been conducted on four public medical image datasets, which are ISIC2018, Kvasir-SEG, BUSI, and CVC-ClinicDB, respectively. Experimental results demonstrate that the proposed CFSeg-Net can outperform other popular models, with Dice scores of 89.87% on ISIC2018, 86.31% on Kvasir-SEG, 79.15% on BUSI, and 92.73% on ClinicDB. These results highlight CFSeg-Net’s potential as a robust and efficient tool for segmentation in medical imaging. The source code will be available at <a href="https://github.com/LGD2333/CFSeg-Net.git">https://github.com/LGD2333/CFSeg-Net.git</a>.</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

Cfseg-Net: context feature extraction network for medical image segmentation

  • 期刊论文

  • 发表日期：

    2025年6月14日

查看全文
  • Guodong Li,
  • Shiren Li,
  • Yaoxue Lin,
  • Sihua Tang,
  • Wenguang Xu,
  • Kangxian Chen,
  • Guangguang Yang

摘要

Image segmentation is critical in medical image analysis, enabling accurate identification and localization of key regions within images. This accuracy is indispensable for reliable diagnosis and effective treatment planning. To achieve an outstanding model performance with moderate parameters, a context feature extraction network for medical image segmentation (abbreviated as CFSeg-Net) has been proposed. The proposed CFSeg-Net is designed to enhance the understanding of the context features. The CFSeg-Net primarily comprises two components: the context feature extraction (CFE) module and feature enhancement (FE) module. The CFE module leverages a multi-range perception scheme, channel shuffle, and multi-feature integration scheme to achieve a broader receptive field. A broader receptive field leads to a better understanding of context features and better segmentation performance. Notably, model performance can be further improved when the FE module is inserted between two CFE modules in the encoder. Extensive experiments have been conducted on four public medical image datasets, which are ISIC2018, Kvasir-SEG, BUSI, and CVC-ClinicDB, respectively. Experimental results demonstrate that the proposed CFSeg-Net can outperform other popular models, with Dice scores of 89.87% on ISIC2018, 86.31% on Kvasir-SEG, 79.15% on BUSI, and 92.73% on ClinicDB. These results highlight CFSeg-Net’s potential as a robust and efficient tool for segmentation in medical imaging. The source code will be available at https://github.com/LGD2333/CFSeg-Net.git.