Hybrid Deep Learning Techniques for Glaucoma detection

Abstract

In today’s world, glaucoma ranks high among the ocular conditions that often results in visual loss. There is currently no test that can reliably and specifically diagnose glaucoma on its own. Nonetheless, it has been speculated in earlier research if anatomical features of the optical whim-whams might predict glaucomatous damage. This research provides the public with a dataset that includes health data and fundus prints from the same case’s both eyes. Furthermore, this study has performed segmentations of the optical slice and cup, as well as patient labelling based on the analysis of clinical data. To distinguish between those who are healthy and those who have glaucoma, a neural network was tested on the dataset. ResNet-50 is used to classify various cases using both the linked data from each case’s two eyes as well as data from each eye individually. The results provide the required steps to further research on fast glaucoma diagnosis predicated on parallel screening among both eyes for a single subject.