Detection of Retinopathy of Prematurity Using Convolution Neural Network

Abstract

Retinopathy of prematurity (ROP) is an eye disease that can happen in premature babies which leads to blindness. ROP can be detected by using a convolution neural network (CNN) algorithm. CNN is a deep neural network mostly applied in medical fields. CNNs are regularized versions of multilayer perceptrons. A CNN is composed of multiple building blocks, such as the convolution layer, rectified linear unit, pooling layer, and fully connected layer and is designed to automatically and adaptively learn spatial hierarchies of features through a back propagation algorithm. CNN is used in various fields such as radiology, medical imaging, etc. This work focuses on the basic concept of CNNs and their application in medical imaging, particularly focusing on ROP. ROP occurs when abnormal blood vessels grow and spread throughout the retina, the tissue that lines the back of the eye. These abnormal blood vessels are flimsy and can leak, scar the retina, and pull it out of its position. ROP affects newborn babies' vision. It is very essential to detect this problem at an early stage, which is proposed in this work. In this work a dataset of ROP image and normal images are used to train CNN algorithm and the test image is compared with the available dataset. The result is provided as the test image is of normal eye or eye with ROP..