Glaucoma Prediction Using Deep Convolutional Neural Network

Abstract

Pressure of the fluid intraocular damages the optical nerve of eye and this disorder is known as Glaucoma eye disease. Pressure will be generated due to increase in eye fluid. It is observed that 4.5 million people are suffering from Glaucoma every year globally. More than 12% of the people among all blindness are because of glaucoma. Glaucoma is globe’s second most commonly observed reason of blindness. This type of eye disease is found in the people with age of above 40 years. In this application deep convolutional neural network (DCNN) is used for prediction of glaucoma disease from retinal fundus images. There are many machine learning algorithms have been implemented by different researchers which has lower accuracy and performance. To get improved performance deep learning algorithm has been used with a greater number of epochs and a greater number of layers. LAG dataset is used which is collected from Kaggle.com website. This dataset contains two categories’ data such as ‘Normal’ and ‘Glaucoma’. Performance is calculated using accuracy, precision, recall and F score. Confusion matrix also plotted for both SVM classifier and deep CNN classifier which represents plot of predicted class of glaucoma image by given algorithm versus actual or true class of glaucoma image present in dataset. As proposed deep CNN classifier has superior performance in all the metrics over SVM classifier for prediction of glaucoma in retinal fundus image so for further testing the deep CNN classifier is recommended for both users and ophthalmologists. In future, security and new features like internet of things (IoT) can be added to the proposed classifier results to get additional security and easy handling of data by ophthalmologists using IOT