A certain investigation on glaucoma detection using deep learning algorithms

Abstract

Glaucoma is the second leading cause of blindness next to cataract. It newlinedamages the human optic nerve due to the increase in eye pressure. It is newlinecaused by low aqueous formation or blockage which results in increased newlineintraocular pressure inside the eye leading to vision loss. Hence, early newlinedetection of glaucoma is vital to avoid visual loss. The manual evaluation of newlinethe result is subjective, time-consuming and tedious. These factors make newlinemanual evaluation impractical for real-world applications. The computerized newlineevaluation of the result is objective and it consumes only less time. As a newlineresult, an objective method is utilized to automate the diagnosis. newlineThe objective of the present research is to develop an intelligent newlineglaucoma classification model for classifying images into either normal or newlineglaucoma by using machine learning and deep learning algorithms. In the newlinecurrent research, the publicly available datasets such as Online Retinal newlineFundus Image Database for Glaucoma Analysis and Research (ORIGA), newlineStructured Analysis of the Retina (STARE) and Retinal Fundus Glaucoma newlineChallenge (REFUGE) are employed. In Machine Learning, the retinal fundus newlineimages are initially preprocessed through Gaussian filter and Histogram newlineequalization. Further, image segmentation is done by using semantic newlinesegmentation approach. Features are extracted through density with newlinecorrelation based method and the Principal Component Analysis (PCA) is newlineused to find the optimal features. The machine learning algorithms such as newlineNeural Network, Support Vector Machine (SVM) and Random Forest are newlineemployed for glaucoma classification. The retinal fundus images are given to newlinethe deep learning algorithms such as VGG16, Inception V3 and Inception- newlineResNet-V2 for glaucoma classification. Further, a novel ensemble model, newlineGlaucoma Classification using Ensemble Network (GCENet), that includes newlinethree deep learning models is proposed to improve the classification accuracy. newlineIt is observed that the proposed model obtains good results and it has the newlinepotential to be used in the real world for the detection of glaucoma. newline newline