西北师范大学机构知识库

Semantic segmentation of the retinal vessels is a pivotal stage in the treatment of certain eye disorders. Efficient and accurate segmentation is a challenge for retinal vessel segmentation. In this article, we presented a technique which is based on a Gated Skip-connection Network (GS-CNN), which implements the simultaneous segmentation of retinal vessel. In GS-CNN, a novel skip-connection with gating is first used in extension path, which facilitates flow of information from down-sampling to up-sampling. Specifically, we use gated skip-connection between encoder and decoder to gate the lower-level information from the encoder. This can effectively remove noise and help the decoder to focus on processing the relevant boundary-related information. Secondly, multi-scale input images are constructed in UNet. Finally, we verified the GS-CNN on DRIVE, CHASE datasets. The experimental results proved the effectiveness of the GS-CNN