Efficient image processing and machine learning approach for predicting retinal diseases

Abstract

As the computational technology and hadrware system improved over time, the use of neural network in image processing has become more and more prominent. Soon deep learning also caught the attention of the medical sector and started getting used in classify diseases. Lots of research are currently going on to predict retinal diseases using deep learning algorithms. However, very small amount of research have been conducted on predicting choroidal neovascularization (CNV), Diabetic Macular Edema (DME) and DRUSEN. In this paper, we have classified OCT images into 4 categories (CNV, DME, DRUSEN and natural retina) by using two deep learning algorithm (convolutional neural network and artificial neural network). Before passing the images into the neural network, we have performed a number of preprocessing methods on the images. Furthermore, we have implemented different model for each algorithms. Each model has varying numbers of hidden layer attached to it. After completing our research we have found out that, convolutional neural network with four hidden layers ou