Abstract:
Glaucoma is one of the leading causes of blindness in people above the age of 45. It
is a group of eye conditions that deteriorate the optic nerve. Glaucoma is usually
caused by the build up of intraocular pressure around the optic disc leading to
permanent vision loss or total blindness.One of the most common approaches to
diagnosing glaucoma is through the thorough examination of the dilated pupil by
an ophthalmologist. However, this a tedious task and often results in the misdiagnosis of early onset glaucoma. This issue can be resolved using a machine learning
approach.
This paper presents a machine learning algorithm that makes use of a combination of Convolutional Neural Networks (CNNs) and Artificial Neural Networks
(ANNs) for the detection of glaucomatous and non-glaucomatous fundus images.
It focuses on extraction of spatial features using CNNs and then further performing
the classification of those images using ANNs. This process yields better results
than other methods employed for feature extraction and classification on the basis
of AUC and accuracy
