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

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...

Full description

Saved in:
Bibliographic Details
Published in:Neurocomputing (Amsterdam) Vol. 438; pp. 72 - 83
Main Authors: de Sales Carvalho, Nonato Rodrigues, da Conceição Leal Carvalho Rodrigues, Maria, de Carvalho Filho, Antonio Oseas, Mathew, Mano Joseph
Format: Journal Article
Language:English
Published: Elsevier B.V 28-05-2021
Elsevier
Subjects:
2D images
3D images
3DCNN
Artificial Intelligence
Computer Science
Glaucoma
Life Sciences
Machine Learning
Medical Imaging
Volumes
Glaucoma
2D images
Volumes
3D images
3DCNN
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary: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.
ISSN:0925-2312
1872-8286
DOI:10.1016/j.neucom.2020.07.146