Glaucoma Eyes Disease Identification: Using Vgg16 Model throughDeep Neural Network

Abstract

Due to a rise in intraocular pressure, diabetes can potentially induce glaucoma, an optic nerve disease. Severe vision loss may result from the condition if it is not identified promptly. Artificial intelligence can be used to automate this procedure. Early illness diagnosis is therefore essential. The medical term "glaucoma" applies to several sensations rather than an actual disease. These symptoms are marked by elevated intraocular pressure (IOP) linked to an injury to the optic nerve and consequent loss or damage to retinal ganglion cells. However, elevated IOP is not the sole feature of glaucoma that requires or relates to changes in the visual field or damage to the optic nerve. Fundus image research has recently concentrated on the application of automated technology in Deep Learning (DL) based frameworks to extract manual features. Our approach involves utilizing unprocessed fundus images to train a hybrid ML and DL model how to aid experts in identifying indicators of glaucoma. A new framework from the Visual Geometry Group (VGG) is used to extract deep features. Popular traditional Machine Learning (ML) techniques for classification, involving Support Vector Machine (SVM), AdaBoost, Random Forest (RF), k Nearest Neighbour (kNN), and Multilayer Perceptron (MLP) have made use of deep features. Using the ACRIMA dataset of 705 imageries, the ML classifier and vgg16 models' performances were assessed. There are 80% training and 20% testing data in the dataset. Based on experimental data, the vgg16 emulate has the best success rate, with 94.6 sensitivity, 92.5% specificity, and 93.4% accuracy.