Automatic method for glaucoma diagnosis using a three-dimensional convoluted neural network

Glaucoma is as an abnormality of the optic system that alters the patient's vision, causing damage to the nervous system and potentially increasing intraocular pressure. Early detection is essential in glaucoma - a progressive disease - in order to initiate preventive treatment and thus avoid total vision loss in patients. Efficient glaucoma diagnosis is expensive and time consuming. Considering these aspects, computer vision techniques have been developed to obtain a rapid and cost-effective diagnosis. This paper presents a new method of classification for glaucomatous and healthy background images of the eye. Here, we propose the use of a three-dimensional convolutional neural network (3DCNN) applied to volumes constructed from a transformation, which converts two-dimensional (2D) background images of the eye. The proposed method showed favorable results, reaching 96.4% accuracy, 100% sensitivity, 93.02% specificity, a 0.965 area under the curve (AUC), and a 0.928 Kappa. (c) 2021 Elsevier B.V. All rights reserved.