Orientation and Context Entangled Network for Retinal Vessel Segmentation

Most existing deep learning based methods for vessel segmentation neglect two important aspects of retinal vessels: The orientation information vessels and the contextual whole fundus region. In this paper, we propose a robust context entangled network (OCE-Net), which can extract complex from blood vessels. To achieve orientation-aware convolution, dynamic convolution (DCOA Conv) to with multiple orientations improving continuity. simultaneously capture global emphasize local information, fusion module (GLFM) model long-range dependency give sufficient attention thin A novel nonlocal (OCE-NL) is also proposed entangle together. addition, an unbalanced refining (UARM) deal pixel numbers background thick Extensive experiments were performed on several commonly used datasets (DRIVE, STARE, CHASEDB1) some more challenging (AV-WIDE, UoA-DR, RFMiD, UK Biobank). ablation study results show method’s good performance in maintaining continuity vessels, comparative experimental OCE-Net’s segmentation. Thus, framework effectively carry out