Dominant Set Clustering for Retinal Vascular Network Topology Reconstruction and Artery Vein Classification PROJECT TITLE : Retinal Vascular Network Topology Reconstruction and Artery Vein Classification via Dominant Set Clustering ABSTRACT: To understand the link between vascular alterations and a wide range of disorders, complicated network topology estimate is critical. An ophthalmologist's diagnosis and treatment of eye illness is aided by the automatic classification of retinal vascular trees into arteries and veins. Their ambiguity and minor variations in appearance, contrast and geometry during the imaging process make it difficult to capture them. Using vascular network topological features, we have developed a novel approach for distinguishing between arteries and veins (A/V) in retinal colour fundus pictures. A/V classification and retinal blood vessel topology estimates are formalised as a pairwise clustering issue in order to use the concept of dominant set clustering. Image segmentation, skeletonization and identification of relevant nodes are used to build the graph. Using the feature space of intensity, direction, curvature, diameter and entropy, the inverse Euclidean distance between the two end points of the edge is defined as the edge weight. Blood vessels are divided into arterial and veinous networks according to their intensity and shape. Accuracy of 95.1 percent, 94.2 percent; 93.8 percent; 91.1 percent; and 91.0 percent were attained by applying the proposed approach to five public databases, including INSPIRE, IOSTAR, VICAVR and DRIVE. The topologies of blood vessels have been manually annotated in the INSPIRE, IOSTAR, VICAVR, and DRIVE datasets, and these annotations have been made accessible so that researchers can use them to aid their work. Did you like this research project? To get this research project Guidelines, Training and Code... Click Here facebook twitter google+ linkedin stumble pinterest Salient Object Detection Using Residual Learning Salient Object Detection with a Reverse Attention-Based Residual Network