智能系统学报2023，Vol.18Issue(6)：1268-1274,7.DOI:10.11992/tis.202107063
基于残差通道注意力的视网膜血管图像分割
Retinal vascular image segmentation based on residual channel attention
王文辉 1刘彦隆1
作者信息
- 1. 太原理工大学 信息与计算机学院,山西 晋中 030600
- 折叠
摘要
Abstract
Segmentation of retinal blood vessels is an important step in the diagnosis of many early eye-related diseases.In this paper,the holistically-nested edge detection(HED)network is applied to retinal vascular image segmentation,and a series of improvements are made to the model:a new modified efficient channel attention(MECA)module is in-troduced to address the lack of ability of existing methods to identify edges and fine vessels,and a double residual struc-ture is used to deepen the model structure to extract finer vascular structures.A structured DropBlock module is intro-duced to prevent overfitting problems from model deepening.In order to further improve sensitivity of the model,a short connection structure incorporating the MECA module is added in the feature fusion phase of the HED network.Experiments show that compared with the current state-of-the-art methods,the sensitivity of the proposed network is significantly improved,which indicates that the proposed method has the state-of-the-art ability to identify retinal vessels.关键词
图像处理/视网膜血管/通道注意力/边缘检测/灵敏度/双残差结构/特征融合/深度学习Key words
image processing/retinal blood vessel/channel attenrion/edge detection/sensitivity/double residual block/feature fusion/deep learning分类
信息技术与安全科学引用本文复制引用
王文辉,刘彦隆..基于残差通道注意力的视网膜血管图像分割[J].智能系统学报,2023,18(6):1268-1274,7.
