A graph‐based edge attention gate medical image segmentation method

Abstract

For the purpose of solving the problems of missing edges and low segmentation accuracy in medical image segmentation, a medical image segmentation network (EAGC_UNet++) based on residual graph convolution UNet++ with edge attention gate (EAG) is proposed in the study. With UNet++ as the backbone network, the idea of graph theory is introduced into the model. First, the dropout residual graph convolution block (DropRes_GCN Block) and the traditional convolution structure in UNet++ are used as encoders. Second, EAGs are adopted so that the model pays more attention to image edge features during decoding. Finally, aiming at the imbalance problem of positive and negative samples in medical image segmentation, a new weighted loss function is introduced to enhance segmentation accuracy. In the experimental part, three datasets (LiTS2017, ISIC2018, COVID‐19 CT scans) were used to evaluate the performances of various models; multiple groups of ablation experiments were designed to verify the effectiveness of each part of the model. The experimental results showed that EAGC_UNet++ had better segmentation performance than the other models under three quantitative evaluation indicators and better solved the problem of missing edges in medical image segmentation.

Image Segmentation and Grouping Using Graph Method (SLIC Algorithm)

“…In such scenarios, parameter tuning is of utmost importance, as emphasized by Yu et al (2023). To overcome these challenges, optimizations such as integrating edge features as suggested by Hao and Li (2023) or utilizing neural networks as implemented by Meng et al (2021) could prove to be helpful.…”

Advancing Ocular Imaging: A Hybrid Attention Mechanism-Based U-Net Model for Precise Segmentation of Sub-Retinal Layers in OCT Images

“…We introduce an improved edge feature attention mechanism for retinal images to enhance edge information and address these gaps. Inspired by the approach in [ 35 ] and designed for 2D images, our edge attention (EA) block combines the structure with the Canny operator to boost edge features. In Figure 4 , represents the feature mapping output at the i th layer, characterised by feature maps with dimensions , where, is the number of channels and denotes the size of each feature map.…”

“…Attention mechanisms empower models to selectively focus on crucial features or regions, enhancing performance in tasks like image segmentation. Various studies have suggested incorporating attention blocks or modules into the U-Net’s encoder or decoder [ 35 , 36 , 37 , 38 ]. These attention mechanisms may utilise techniques such as channel attention, spatial attention, or self-attention.…”

VGA‐Net: Vessel graph based attentional U‐Net for retinal vessel segmentation

“…Yu et al [44] maintained the topology of vessels by creating edges and predicting links between various nodes, which were generated from semantic information extracted using U-Net. Hao et al [45] presented a new framework called EAGC_UNet++, a graph-based edge attention gate, for medical image segmentation. By leveraging a combination of a residual graph convolution network, UNet++ and edge attention gates (EAG), the proposed model has demonstrated significant improvements in the accuracy of medical image segmentation and tackled the problem of missing edges.…”