[PR2022] The Code of “EANet: Iterative Edge Attention Network for Medical Image Segmentation”
Authors: Kun Wang, Xiaohong Zhang, Xiangbo Zhang, Yuting Lu, Sheng Huang, Dan Yang
git clone https://github.com/DLWK/EANet.git
cd EANet/
Download the datasets and unzip them into data
folder
cd train/
python3 train.py
Accurate and automatic segmentation of medical images can greatly assist the clinical diagnosis and analysis. However, it remains a challenging task due to (1) the diversity of scale in the medical image targets and (2) the complex context environments of medical images, including ambiguity of structural boundaries, complexity of shapes, and the heterogeneity of textures. To comprehensively tackle these challenges, we propose a novel and effective iterative edge attention network (EANet) for medical image segmentation with steps as follows. First, we propose a dynamic scale-aware context (DSC) module, which dynamically adjusts the receptive fields to extract multi-scale contextual information efficiently. Second, an edge-attention preservation (EAP) module is employed to guide effectively removes noise and helps the edge stream focus on processing only the boundary-related information. Finally, a multi-level pairwise regression (MPR) module is designed to combine the complementary edge and region information for refining the ambiguous structure. This iterative optimization helps to learn better representations and more accurate saliency maps. Extensive experimental results demonstrate that the proposed network achieves superior segmentation performance to state-of-the-art methods in four different challenging medical segmentation tasks, including lung nodule segmentation, COVID-19 infection segmentation, lung segmentation, and thyroid nodule segmentation.
Figure 1: Overview of the proposed EANet. Our network is in an encoder-decoder style,
it consists of the VGG-19 feature encoder E1 ∼ E5, dynamic scale-aware context (DSC) module, edge attention preservation (EAP) module, the feature decoder
D1 ∼ D5 and multilevel pairwise regression (MPR) module. HFF denotes hybrid feature fusion block, which is
a component of MPR. The final prediction is the generated global map (Sg) after iterative
optimization by combining edge information and region information. The Si, {i = 2, 3, 4, 5}
indicates the feature map of the side-output laye.
Figure 2:Visual comparison of COVID-19 infection segmentation results, where the red and
green labels indicate the predicted segmentation and ground truth, respectively.
Figure 3:The visual comparison of results on the TN-SCUI dataset, where the blue and
green labels indicate the predicted segmentation and ground truth, respectively
Figure 4:Visual comparison of feature maps for showing the effect of multi-level pairwise
regression module (MPR). D5 ∼ D1 denote the feature maps of each decoder block. The odd
and even rows show the baseline results without or with MPR, respectively.
@article{wang2022eanet,
title={EANet: Iterative edge attention network for medical image segmentation},
author={Wang, Kun and Zhang, Xiaohong and Zhang, Xiangbo and Lu, Yuting and Huang, Sheng and Yang, Dan},
journal={Pattern Recognition},
volume={127},
pages={108636},
year={2022},
publisher={Elsevier}
}