Classification of Eye Diseases in Fundus Images

Abstract

Eye diseases have been a severe problem worldwide, especially in developing countries where technology and finance are limited. Today, the problem is being resolved thanks to the task of classification that is part of pattern recognition. Its primary goal is to group standard features from any entity, object, phenomenon, or event belonging to the real or abstract world. Convolutional Neural Networks are a type of Artificial Neural Network used in intelligent pattern classification, Machine Learning, and Data Mining. Also, medicine and ophthalmology used these algorithms for detecting diseases in the human body. This work presents a novel intelligent pattern classification algorithm based on a Convolutional Neural network, which is validated through the K-Fold Cross Validation test. Two different groups of retinography images are given: Glaucoma and Diabetic Retinopathy. The result of accuracy percentage was 99.89%. Numerical metrics: Accuracy, Recall, Specificity Precision, and F1 score with values close to 1, and ROC curves support the suitable performance of the proposed classifier.