Evaluation of convolutional neural network model for classifying red and healthy eye

Abstract

Human eye is covered with mucous like thin membrane on the outer surface as well as inner surface of the eyelids. The thin membrane covering the outer surface is called as bulbar conjunctiva. Conjunctiva hyperemia refers to the redness of the conjunctiva. There are several reasons for this condition, some of which are related to pathologies like trauma, allergy, injury, prolonged use of lenses or glaucoma. Hence, it is one of the most vital parameter to diagnose these pathologies. For populous developing countries like India, ratio of ophthalmologist to citizen is highly skewed which negatively affects the public health. This gives rise to need for developing technology based solutions for initial screening of eye problems like conjunctiva hyperemia. This paper presents a framework for classifying normal versus red eyes using deep learning technique of convolution neural network (CNN). The model has shown promising results with 94% accuracy.