A Novel Prediction Method for Glaucoma Detection Using Retinographies

Glaucoma, characterized by its damage to the retinal nerve, is one of the most statistically dominant eye diseases in the U.S. It can cause vision loss and blindness by affecting the optic nerve. As the disease progresses, it is not necessarily noticeable to patients, requiring elaborate solutions to manage this critical condition. In this paper, we propose a novel detection model that provides improved accuracy and performance in the detection of glaucoma using retinographies as input. After careful consideration, we adopted multiple features suitable for distinguishing healthy and diseased eyes. Moreover, the adoption, integration, and feature extraction of 3D meshes was a significant factor in developing our glaucoma detection system. With our acquired dataset, we compared the performance of Classification Decision Trees (CDT), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN) in classifying retinographies as being healthy or having glaucoma. Experimental results show that the proposed model methodology can efficiently predict glaucoma detection with 100%, 100%, and 83.3% accuracy from CDT, SVM, and KNN, respectively.