Glaucoma Disease Diagnosis-Based Deep Learning Network

Abstract

Increased intraocular pressure and optic nerve damage, which may cause irreversible blindness, are the hallmarks of glaucoma. If this disease is identified early on, its severe effects can be prevented. However, among the older population, the illness is often identified at a later stage. Consequently, individuals may be spared irreversible visual loss by early identification. Ophthalmologists use a variety of expensive, time-consuming, skill-oriented techniques when manually assessing glaucoma. A definitive diagnostic method for early-stage glaucoma detection is still elusive, while a number of approaches are in the experimental stages of development. We offer an autonomous deep learning-based technique that has very high accuracy in detecting early-stage glaucoma. The detection method entails identifying patterns in the retinal pictures that physicians frequently miss. A test accuracy of 99.26% was attained in this study using the Resnet-50 network and data from the G1020 database that had been processed and transformed from RGB to RGBA.