An overview of automated glaucoma detection

Abstract

Glaucoma is a neurotic condition of optic nerve impairment and second foremost cause of vision loss worldwide. This study is meta-analysis of 37 existing automated glaucoma detection systems proposed in last 10 years. The aim of the study was to compare existing system on the base of ophthalmic imaging technology, functional and structural features of eye, Machine Learning techniques, accuracy and data set used for evaluation. Results showed that glaucoma detection has been mostly done using Fundus images and structural feature Cup to Disc ratio at final stage. There is no known cure of glaucoma at final stage. There is a need to develop automated glaucoma detection systems that are able to detect glaucoma at an early stage where it can be treated. Glaucoma detection can be effectively achieved using Optical Coherence Tomography (OCT) at an early stage where it could be treated. We found evidence that automated glaucoma detection at the last stage is at a mature level using Fundus images however there is a great margin of improvement in automated glaucoma detection using OCT. It was proposed to use a hybrid feature set consisting of both structural and texture features of retinal image for more accurate glaucoma detection.