Early diagnosis of Glaucoma in retinal images using multi structure descriptor and hybrid neural network classifiers

Abstract

Medical imaging is extensively being used to automatically detect and diagnose various diseases. Medical imaging provides a way to diagnose the diseases in their very early stages by classifying the diseases into their different stages. Diabetes mellitus is an ailment around the world, which can cause genuine outcomes. The assessed predominance of diabetes for all age bunches worldwide was 2.8% in 2000 and is required to be 4.4% in 2030. Glaucoma is an optic neuropathy defined by characteristic damage to the optic nerve and accompanying visual field deficits. This paper proposes an image highlight portrayal strategy, in particular Multi Structure Descriptor (MSD), it uses all together change of pixels to represent the neighborhood spatial structure data for image include extraction. The component assurance step is joined into the framework to diminish the figuring multifaceted nature. A high request precision of 95.8% is cultivated using hybrid neural framework based classifiers. A glaucoma integrative record (GRI) is moreover intended to procure a trustworthy and effective system. The exploratory results are affirmed using the K-cover cross endorsement strategy. The overall classification accuracy of MSD with HNN is 93.45%, MSD with FSVM is 90.12%, MSD with SVM is 84.11 and MSD with RBF is 77.23%.