基于三端注意力机制的视网膜血管分割算法
作者:
作者单位:

中国科学院沈阳自动化研究所

作者简介:

通讯作者:

中图分类号:

TP391

基金项目:

国家自然科学基金项目(面上项目,重点项目,重大项目)


An Improved U-Net Based on three-terminal attention mechanism for Retinal Vessel Segmentation
Author:
Affiliation:

Shenyang Institute of Automation (SIA), Chinese Academy of Sciences

Fund Project:

The National Natural Science Foundation of China (General Program, Key Program, Major Research Plan)

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    视网膜血管的结构和形态是计算机辅助系统诊断眼科疾病的重要依据. 针对细小血管分割精度低的问题, 提出了一种融合残差密集模块与三端注意力模块的改进型U-Net算法. 首先, 将残差模块与密集模块相结合, 充分利用每层的特征, 提高网络提取细小血管特征的能力. 在解码阶段引入三端注意力模块, 利用空间注意力机制自适应地对特征进行空间校正, 抑制背景噪声, 突出目标区域. 同时, 通过多尺度特征融合的方式, 利用高级语义特征改善网络对细小血管的分割效果. 最后, 为了获取血管的多尺度特征, 在编码-解码网络结构中加入空洞卷积, 在不增加参数的情况下增加了感受野. 基于DRIVE和STARE数据集的实验结果表明, 所提网络的灵敏度、特异性、准确率和AUC(Area Under Curve)分别为81.26%/82.57%、98.20%/98.37%、96.70%/97.51%和98.12%/98.41%, 优于现有先进算法.

    Abstract:

    The structure and morphology of retinal blood vessels are essential for computer-aided systems to diagnose ophthalmological diseases. To solve the problem of low precision of tiny blood vessel segmentation, we propose an improved U-Net algorithm combining residual dense block and three-terminal attention gate block. First of all, we combine the residual block with the dense block to make full use of the features of each layer and improve the ability to extract the characteristics of tiny blood vessels. In the decoding stage, we introduce a three-terminal attention gate block. And the spatial attention mechanism is used to adaptively correct the features, suppress background noise and highlight the target area. At the same time, we use high-level semantic features to improve the segmentation effect of tiny blood vessels through multi-scale feature fusion. Finally, to obtain the multi-scale features of blood vessels, we introduce deformable convolution into the network structure and increase the receptive field without increasing the parameters. The experimental results based on the DRIVE and STARE data sets show that the sensitivity, specificity, accuracy and AUC (Area Under Curve) of the proposed network are 81.26%/82.57%, 98.20%/98.37%, 96.70%/97.51% and 98.12%/98.41%, which are better than existing advanced algorithms.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2021-03-16
  • 最后修改日期:2021-07-15
  • 录用日期:2021-07-19
  • 在线发布日期: 2021-08-01
  • 出版日期: