RVSeg-Net: An Efficient Feature Pyramid Cascade Network for Retinal Vessel Segmentation

Global Transformer and Dual Local Attention Network via Deep-Shallow Hierarchical Feature Fusion for Retinal Vessel Segmentation

“…In recent studies on retinal blood vessel segmentation, due to the limitation of computing capacity, it is difficult for deep learning models to directly segment vessels in retinal images with high resolution. To address this issue, recent methods scaled images during the training stage and restored images to the original resolutions to compute performance metrics during the testing stage [4,5,19]. In this paper, we also adopt this strategy, which all the images in the four datasets are reshaped as 512 × 512, 512 × 512, 512 × 512, and 800 × 800 resolution in the training stage respectively.…”

“…In this paper, we also adopt this strategy, which all the images in the four datasets are reshaped as 512 × 512, 512 × 512, 512 × 512, and 800 × 800 resolution in the training stage respectively. Similar to most recent methods, after obtaining preliminary segmentation results, we upsampled the obtained segmentation results to the same size as the original image in order to accurately calculate relevant performance indicators [4,5,19].…”

“…For example, Wang et al combined two identical multi-scale backbone networks with skip connections, effectively integrating the shallow and deep features of the network [18]. Wang et al also proposed a feature pyramid cascade module to capture the scale changes of vessels and alleviate the problem of discontinuous segmentation results by aggregating local and global context information [19]. Wang et al further investigated a hard attention network, which consists of an encoder and three decoders to solve the segmentation of vessel backbone, easy region, and difficult region respectively [20].…”

Study Group Learning: Improving Retinal Vessel Segmentation Trained with Noisy Labels

“…Conventionally, hand-crafted filters [12,13,16,27] like Gabor [13] and Gaussian-based ones [12] are explored to extract features for pixel selection, vessel clustering and segmentation. Recently, data-driven based methods utilize UNet-based model [17] or its variants [24,25,28,29,10] to achieve significant performance compared with traditional methods. Those deep learning methods focus on the design of UNet structures with better feature representation [28,24], or the decouple of structure and textures of retinal images [29].…”

“…Recently, data-driven based methods utilize UNet-based model [17] or its variants [24,25,28,29,10] to achieve significant performance compared with traditional methods. Those deep learning methods focus on the design of UNet structures with better feature representation [28,24], or the decouple of structure and textures of retinal images [29]. However, datadriven methods highly suffer from over-fitting issues when the given training data is insufficient.…”