Scholarlink

<p>Accurate medical image segmentation is crucial for intelligent healthcare. While convolutional neural networks (CNNs), particularly UNet and its variants, have demonstrated excellent performance, they face challenges such as insufficient adaptive capability, redundant feature information, and weak multi-scale feature sensing. To address these issues, we propose a novel medical image segmentation method called Dynamic Neighbourhood-Enhanced UNet with Interwoven Fusion (NLIE-UNet). Our approach includes a Cyclic Dynamic Convolution Block (CDCB) to adaptively capture edge contour information, a Neighbourhood Enhanced Bridge (NEB) to exploit the consistency and complementarity of different layer features, and a Hierarchical Interwoven Fusion Module (HIFM) to fuse cross-layer information effectively. Extensive experiments on five public medical image datasets demonstrate the effectiveness and superiority of our method, achieving state-of-the-art performance in terms of Intersection over Union and Dice coefficients. The code and datasets for this work are published in <a href="https://github.com/0neDawn/NLIE-UNet">https://github.com/0neDawn/NLIE-UNet</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

Dynamic neighbourhood-enhanced UNet with interwoven fusion for medical image segmentation

  • 期刊论文

  • 发表日期：

    2025年2月14日

查看全文
  • Liming Wan,
  • Lin Song,
  • Ying Zhou,
  • Chenrui Kang,
  • Shijian Zheng,
  • Guo Chen

摘要

Accurate medical image segmentation is crucial for intelligent healthcare. While convolutional neural networks (CNNs), particularly UNet and its variants, have demonstrated excellent performance, they face challenges such as insufficient adaptive capability, redundant feature information, and weak multi-scale feature sensing. To address these issues, we propose a novel medical image segmentation method called Dynamic Neighbourhood-Enhanced UNet with Interwoven Fusion (NLIE-UNet). Our approach includes a Cyclic Dynamic Convolution Block (CDCB) to adaptively capture edge contour information, a Neighbourhood Enhanced Bridge (NEB) to exploit the consistency and complementarity of different layer features, and a Hierarchical Interwoven Fusion Module (HIFM) to fuse cross-layer information effectively. Extensive experiments on five public medical image datasets demonstrate the effectiveness and superiority of our method, achieving state-of-the-art performance in terms of Intersection over Union and Dice coefficients. The code and datasets for this work are published in https://github.com/0neDawn/NLIE-UNet.