Retinal Glaucoma Detection Using Deep Learning Algorithm

Abstract

Glaucomatic retinopathy is a degenerative eye disease that is assessed as it progresses, making it necessary to examine it more frequently. A novel method of analyzing retinal images by detecting vessels and exudates has been proposed to analyze retinal vascular disorders. In this work, noise removal is modified by a suitable nonlinear function. The modified function parameters are derived by using fast DCT (FDCT) coefficients, which enhance weak edges while eliminating noise. In this work, a minimum of 10 features such as Mean, Variance, Entropy with the data set trained images on 15 images using NN (neural network) training is implemented, and NN classifier based Normal or Abnormal is applied. Simulation is done using MATLAB Simulink and comparisons among Discrete Wavelet, Curvelet, Orthogonal transform, and fuzzy segmentation are executed and the blood vessels segmentation resulted in promising results. This work proposes deep learning-based system for glaucoma diagnosis using retinal fundus images, developed using image processing and deep learning approaches.