A Review on Recent Developments for the Retinal Vessel Segmentation Methodologies and Exudate Detection in Fundus Images Using Deep Learning Algorithms

摘要

Retinal image analysis is considered as a well-known non-intrusive diagnosis technique in modern opthalmology. The pathological changes which occurs due to hypertension, diabetic retinopathy and glaucoma can be viewed directly from the blood vessels in retina. The examination of the optic cup-to-disc ratio is the main parameter for detecting glaucoma in the early stages. The significant areas of the fundus images are isolated using the segmentation techniques for deciding the value of cup-to-disc ratio. The deep learning algorithms, such as the Convolutional Neural Networks (CNNs), is often used technique for the analysis of fundus images. The algorithms using the concepts of CNNs can provide better accuracy for the retinal images. This review explains the recent techniques in deep learning relevant for the analysis of exudates.