A deep neural network based cnn model for improved diabetic retinopathy detection

Abstract

Diabetic retinopathy (DR) is a prevalent eye disease that can lead to vision loss if not detected and treated in its early stages. In this research study, we propose a methodology for the detection of diabetic retinopathy using deep learning techniques. The images are pre-processed to rectify issues such as blurring and resizing, and the blood vessels are extracted to obtain accurate results. Statistical data, including mean, standard deviation, and variance, is extracted from the images. We have used Convolutional Neural Networks (CNN) in our thesis for the image classification task. Thereafter we performed various hyperparameter tunings for achieving the desired results. In this thesis, we also evaluate the results of all five stages of diabetic retinopathy and plot the confusion matrix. The proposed methodology aims to achieve improved prediction accuracy for diabetic retinopathy detection which can be useful in early diagnosis of diabetic retinopathy.