Optimizing Glaucoma Detection Using Graph Neural Networks and DBSCAN Clustering Techniques

The human being perceives the world through the eyes. Glaucoma is an eye ailment that results in the loss of visual acuity. Initial signs of primary open-angle glaucoma include blurred vision and problems seeing in different lighting conditions. The symptoms of late primary open-angle glaucoma include impaired central vision and side vision loss. High ocular pressure attacks can result from angle closure glaucoma. The symptoms of this medical emergency include red eyes, headaches, dilated pupils, visual loss, nausea, vomiting, and seeing haloes around lights. Early intervention is crucial for treatment and eye preservation. However, in developing nations, around 90% of people are not aware they have the illness. Additionally, approximately 50% of patients with glaucoma in the US are unaware that they have the condition. Furthermore, the existing diagnosis processes and techniques are expensive and require skilled manpower to use. This paper proposes a novel glaucoma diagnosis method that integrates Density-Based Spatial Clustering of Applications with Noise (DBSCAN) plus Graph Neural Networks (GNNs) and Convolutional Neural Networks (CNNs). GNN with DBSCAN was able to detect minimal changes in the structure within the retina which could reflect early stages of glaucoma, achieving a 96.5% accuracy on the RIGA dataset.