A nested U-shape network with multi-scale upsample attention for robust retinal vascular segmentation

作者：

Highlights：

• Propose a novel encoder-decoder framework for vessel segmentation in retinal imaging.

• A new multi-scale upsample attention module is proposed to enhance vessel segmentation in a hierarchical structure.

• Proposed model achieves state-of-the-art performance of vessel segmentation on DRIVE, STARE, CHASE_DB1, HRF and IOSTAR datasets.

• Experimental results demonstrate the robustness of our method in handling fundus image with lesions and microvessels.

摘要

•Propose a novel encoder-decoder framework for vessel segmentation in retinal imaging.•A new multi-scale upsample attention module is proposed to enhance vessel segmentation in a hierarchical structure.•Proposed model achieves state-of-the-art performance of vessel segmentation on DRIVE, STARE, CHASE_DB1, HRF and IOSTAR datasets.•Experimental results demonstrate the robustness of our method in handling fundus image with lesions and microvessels.

论文关键词：Vascular segmentation,Retinal imaging,Dense U-Net,Multi-scale attention,Deep learning

论文评审过程：Received 6 December 2019, Revised 26 January 2021, Accepted 18 April 2021, Available online 24 April 2021, Version of Record 10 July 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107998