A Convolutional Neural Network for Classification of Diabetic Retinopathy

Abstract

As the years pass by, more and more diseases are being diagnosed and there has been a gradual increase in the number of patients suffering from diabetic. With the advancing technology, medical sciences have been fast-forwarding towards machines that help in the diagnosis of the disease in every field. In our thesis, we worked with diabetic retinopathy images to classify between retinopathy eye and normal eye using a convolutional neural network. The convolutional neural network is very efficient for image classification problems as it extracts the features from the images that help us differentiate between the normal eye and retinopathy eye. The implementation of such technology has enabled the doctors to provide faster and better treatments to patients suffering from diabetic. Many diseases can be diagnosed through such systems in hospitals. Some of the hospitals have already started using these systems and the results are quite impressive. This will benefit both the people and the doctors as the treatments will be provided within a short period and the patients don’t have to suffer through diabetic retinopathy. We developed a convolutional neural network model to classify diabetic retinopathy images in [1] dataset. The dataset is an open-source diabetic retinopathy dataset that has about 35126 images of various patients appropriate for our research.