Optic Disk Extraction and Hard Exudate Identification in Fundus Images using Computer Vision and Machine Learning

The past decade has seen a surge in individuals adopting a sedentary lifestyle and this has increased a hundredfold due to lockdowns and change in work habits because of the ongoing pandemic of CoVID-19. This lack of exercise has put a toll on people's health and has resulted in an increase in the prevalence of diseases such as Diabetes Mellitus. Though known for its association with the pancreas, diabetes also affects other organs of the body including the eye. This paper discusses one such side-effect of diabetes known as Diabetic Retinopathy, wherein deposition of lipids (also known as exudates) occurs within the eye. The paper goes on to summarize a novel approach to identify the same in fundus images of the eye using computer vision and machine learning. It also elaborates on ways to eliminate the optic disk of the eye in such images as well, which is a vital step to be done prior to exudate identification. The algorithm proposed in this paper provides an accuracy of more than 0.18% compared to other popular algorithms and also features a speed improvement of 1.82 times in comparison with its predecessor methodologies. All experiments and algorithms were developed using MATLAB R2017a.