Multi feature analysis and jaya chicken swarm optimization based recurrent neural network for glaucoma detection

Abstract

newline Glaucoma is a constant and irrevocable eye disease that causes weakening in vision and affects the quality of life. The lost potential of the optic nerve cannot be recovered, but earlier determination and essential treatment are imperative to a patient affected with glaucoma to protect from vision loss. The diagnosis of glaucoma is performed based on visual field loss tests, medical history, intraocular pressure of patients, and finally manual assessment using fundus images. Earlier diagnosis of glaucoma is essential for preventing permanent damage of structure and irreversible loss of vision. The fundus images are employed in ophthalmology for visualizing the structures of the optic disc. However, accuracy is a major limitation in glaucoma detection. To improve accuracy, three major contributions are devised for attaining effective glaucoma detection. The first contribution is to devise a method for segmenting the blood vessel and optic disc detection from the retinal fundus images. newlineThis method is employed for supporting non-intrusive diagnosis in modern ophthalmology as the morphology of the blood vessel and the optic disc is an important indicator for detecting glaucoma. Initially, pre-processing is done, where the noise and artifacts contained in the images are removed followed by blood vessel segmentation and optic disc detection. The segmentation of blood vessels is performed by the Renyi-based sparking method wherein the sparking process and Renyi entropies are applied for generating the segmented blood vessel. Simultaneously, the optic disc detection is carried out using binarization and a circle fitting method for estimating the location of the optic disc. newline newline