Detection of Hypertensive Retinopathy by Segmentation of Retinal Blood Vessels and Retinal Tortuosity Measurement

The automatic segmentation of retinal blood vessels is a crucial step in the early detection and treatment of disorders like Hypertensive Retinopathy (HR), and cardiovascular diseases. Because of the presence of minute blood vessels, segmentation will be complex work when performed manually and requires skilled and trained physicians in segmenting these minute vessels. In this work, hypertensive retinopathy is detected by segmenting the retinal blood vessels using Convolutional Neural Network (CNN) architecture like LinkNet and by using segmented image as input retinal tortuosity like fractal dimension and branching coefficient are measured. The proposed method is tested on the MESSIDOR and Ocular Disease Intelligent Recognition (ODIR) publicly available databases. After segmenting the blood vessels, retinal tortuosity is measured. As a result, this model has reached an accuracy of 96.36% for ODIR and 90.02% for MESSIDOR databases.