Multiscale U-Net with Spatial Positional Attention for Retinal Vessel Segmentation.

Congjun Liu,Zhiyong Xiao,Penghui Gu,Jinshan Tang

doi:10.1155/2022/5188362

Abstract

Retinal vessel segmentation is essential for the detection and diagnosis of eye diseases. However, it is difficult to accurately identify the vessel boundary due to the large variations of scale in the retinal vessels and the low contrast between the vessel and the background. Deep learning has a good effect on retinal vessel segmentation since it can capture representative and distinguishing features for retinal vessels. An improved U-Net algorithm for retinal vessel segmentation is proposed in this paper. To better identify vessel boundaries, the traditional convolutional operation CNN is replaced by a global convolutional network and boundary refinement in the coding part. To better divide the blood vessel and background, the improved position attention module and channel attention module are introduced in the jumping connection part. Multiscale input and multiscale dense feature pyramid cascade modules are used to better obtain feature information. In the decoding part, convolutional long and short memory networks and deep dilated convolution are used to extract features. In public datasets, DRIVE and CHASE_DB1, the accuracy reached 96.99% and 97.51%. The average performance of the proposed algorithm is better than that of existing algorithms.

Highlights

Many eye-related diseases can lead to structural characteristics changes of the retinal vessel in fundus images.erefore, fundus retinal vessel segmentation plays a significant role in the detection and diagnosis of these eye diseases, such as diabetic retinopathy, hypertension, and arteriosclerosis [1]
With the development of deep neural networks, good results have been achieved in the field of medical image processing [5,6,7]. erefore, more and more algorithms based on deep learning are used for retinal vessel segmentation
We propose an improved U-Net-based fundus vessel segmentation algorithm. e main contributions of this paper are summarized as follows. (1) e improved position attention module (PA) and channel attention module (CA) were added in the jump connection part to improve the effect of vessel segmentation under low contrast

Summary

Introduction

Many eye-related diseases can lead to structural characteristics changes of the retinal vessel in fundus images. Xiao et al [3] used an improved level set method to minimize the proposed energy function to identify blood vessels in retinal images. Erefore, more and more algorithms based on deep learning are used for retinal vessel segmentation. Existing deep learning-based retinal vessel segmentation models can be classified into four groups according to network structure [8]. Li et al [10] constructed FCN with jumping connection part and introduced active learning into retinal vessel segmentation. E network model based on U-Net can capture local and global information through a connection feature graph to make better decisions, so it can obtain better segmentation results. We propose an improved U-Net-based fundus vessel segmentation algorithm.

Related Works

Method

Results and Discussion

Experimental Results