Glaucoma Diagnosis From Smartphone Captured Fundus Images Using Deep Learning

The retina contains millions of photoreceptive cells called cones, rods, and blood vessels that nourish these cells. The optic nerve carries visual information from the retina to the brain. Glaucoma is a condition, where optic nerves get damaged often caused by high pressure within the eye. Fundus images provide rich information about the appearance of a patient’s retina and optic nerve head region. Even though there are various painless diagnosis methods, conventional fundus photography using Retcams is the most commonly used one. Examinations of the fundus images for structural abnormalities in the optic nerve head region such as increased cup-to-disc ratio, NRR thinning, and Violation of the ISNT rule provide sufficient evidence for the presence of glaucoma. However, conventional fundus imaging costs a lot concerning equipment and operation costs. Nowadays, smartphones can capture retina images and are a cheaper alternative to conventional fundus imaging using retcam devices. So our approach is to analyze smartphone-captured fundus images (SCFI) to detect the presence of glaucoma and provide accurate results at a lower cost. We designed an automated one-step approach-based glaucoma detection system that can detect the presence of glaucoma from smartphone-based fundus images with an accuracy of 93 percent. using a deep learning-based approach, The publicly available smartphone-captured fundus image dataset, BrG is used for conducting ablation studies.