Diagnosis of Glaucoma from Retinal Fundus Image Using Deep Transfer Learning

Abstract

The phenemeon of retinal condition brought about by expanding intraocular strain inside the eye is knows as Glaucoma. The existing diagnosis process of glaucoma through various careful retinal tests such as ophthalmoscopy, tonometry, perimetry, gonioscopy, and pachymetry are costly as well as time-consuming. Moreover, the diagnosis processes are fully dependent on the Ophthalmologists knowledge of test report analysis. To overcome these issues, this work aims to propose the utilization of a profound deep learning model such as ResNetl52 as well as VGG16 for the primary feature extraction qualities, specifically cup-to-circle proportions, plate obligation scale harm, and unrivaled nasal fleeting lower regions to diagnose glaucoma. Performance evaluation of the model is performed based on the accuracy matrix that shows 87% and 72% of accuracy for the ResNetl52 and VGG16 models respectively. The ResNetl52 model outperformed the VGG16 model because of the capability of extracting deep structures of the retinal image with the aid of skip connections from the previous consecutive convolutional layers.