Dual-path multi-scale context dense aggregation network for retinal vessel segmentation

Wei Zhou, Weiqi Bai, Jianhang Ji, Yugen Yi, Ningyi Zhang, Wei Cui

Abstract: Highlights•We have developed a dual-path deep learning framework aimed at achieving precise segmentation of blood vessels.•To improve feature propagation and generate discriminative feature maps, we have designed a novel network called MCDAU-Net.•A fusion module is proposed to convey contextual information and enhance the integrity and continuity of vessel structures.•Our method achieved a remarkable increase of 7.9% and 4.7% in the mean values of Se and F1 scores on three public datasets.