Determination of Early Onset Glaucoma Using OCT Image

Abstract

In order to find the glaucoma in an early stage with the help of optical coherence tomography (OCT) using deep learning and extracting the features of glaucoma, the authors are able to classify the four types of glaucoma such as CNV, DME, DRUSEN, and the normal ones with perfect accuracy by training this dataset. This dataset contained 968 images and 242 images of each type the authors trained their model by using CNN algorithm, and has greater accuracy of when compared to the determination of glaucoma using support vector machine image. The authors have good architecture constructed for the determination. They pre-trained their deep learning model in order to obtain the initial representation. The proposed system gives out 95.4% accuracy level of sensitivity, specificity, and classification. A large increase in the volume of the cup, a larger cup diameter, and a thickened lip of the neuroretina rim suggest glaucoma. These regions correspond anatomically to currently used clinical markers for glaucoma diagnosis.