DETECTION OF HEMMOREGES OF DIABETIC RETINOPATHY USING MACHINE LEARNING ALGORITHMS

Abstract

Diabetic retinopathy (DR) is one of the most common and dangerous capillary consequences of diabetes. It is marked by lesions like microaneurysms, haemorrhages, exudates, and new blood vessels growing in the retina.  Haemorrhages are one of the most important signs of how bad the disease is and how fast it is spreading. They are often the first sign of severe DR stages that can cause permanent blindness if they are not found early.  So, it's important to find haemorrhages quickly and correctly so that they can be treated quickly and prevent vision loss.  Traditional ways of diagnosing depend on ophthalmologists looking at retinal fundus pictures by hand, which takes a long time and can be subjective.  These problems have led to a lot of interest in automatic methods based on picture processing and machine learning over the past few years. The main goal of this study is to use machine learning methods to find haemorrhages in diabetes retinopathy.  A set of retinal fundus images that was open to the public was used, and expert comments were used as the basis for confirmation.  The method includes several steps: preprocessing to improve the image and get rid of noise; extraction of distinguishing features such as colour, texture, and shape descriptors; and classification using both deep learning models and traditional machine learning algorithms (Support Vector Machine, Random Forest, k-Nearest Neighbour).  To find out how accurate, sensitive, and detailed the different methods were, a comparative performance review was done.  These results show that machine learning-based classifiers are much more accurate at finding haemorrhages than traditional handcrafted methods. Additionally, deep learning approaches perform better because they can automatically capture complex feature representations.