Extraction of retinal blood vessels and diagnosis of proliferative diabetic retinopathy

Abstract

The diagnosis and screening of retinal illnesses is one use of image processing in the medical field that is being explored in the current study. DR is an eye condition that worsens over time, eventually resulting in blindness. A more advanced stage of DR, known as PDR, is when new retinal vessels (neovascularization) begin to form. PDR is characterised by fragile new arteries developing on the retina that leak blood onto the retina's surface, resulting in significant vision loss. Automatic segmentation of retinal arteries and PDR diagnosis are presented in this article. Extraction of the blood vessel is done using matching filtering and adjusted local entropy thresholding. A new machine learning technique known as the Extreme Learning Machine (ELM) classifier is used to classify the images as either normal, NPDR, or PDR. SVM classifier performance is compared to that of the algorithm.