Discover the SciOpen Platform and Achieve Your Research Goals with Ease.
Search articles, authors, keywords, DOI, etc.
In the process of retinal vessel segmentation, issues such as incomplete vessel segmentation or the presence of micro-vessel fractures arise due to the complex structure of small vessels and blurred boundaries. To address these challenges, we propose an edge feature enhancement U-Net (EFE U-Net) based on edge feature enhancement. By introducing an edge feature enhancement module to replace the 2D convolution in the U-Net model, the receptive field of U-Net is expanded while enhancing the edge information in retinal vessel images. The edge feature enhancement module consists of an edge information extraction module, an edge information fusion module, and a channel and spatial attention module. Experiments on the DRIVE, CHASE_DB1, and STARE datasets demonstrate that EFE U-Net achieves sensitivities of 82.64%, 84.41% and 84.45%, the F1 scores are 83.09%, 84.49% and 82.84%. The two metrics show improvemens over the baseline model. The experimental results indicate that EFE U-Net can effectively segment retinal vessels.
This is an open access article under the CC BY-NC-ND 4.0 license (https://creativecommons.org/licenses/by-nc-nd/4.0/).
Comments on this article