An efficient approach for the detection of retinopathy of prematurity using deep learning networks

Retinopathy of Prematurity (ROP) is a major cause of blindness in infants. This disease is a vascular proliferative retinal disease which causes damages to low birth weight children and premature children. The Objective of this proposed system is to develop an efficient automated screening system for diagnosis of ROP by different deep learning network. First, the input color images are captured from the diabetic patients. Then, the acquired images are given as an input to the preprocessing stage for improvement. The Preprocessed noise free color images are applied to the Deep Learning Network for the detection of presence of ROP. Five different deep learning networks such as VGG-16, VGG-19, AlexNet, GoogLeNet and MobileNet are deployed for the classification of normal and abnormal images. The best system out of these five networks will be identified after calculating the evaluation parameters like Accuracy, Precision, Sensitivity, Specificity, ROC AUC. The proposed system is compared with the existing deep learning work and it is found that our proposed system outperforms than all other existing system. This developed Deep Learning CAD system will be a proven affordable technology for the ophthalmologists since it consumes less time in screening for ROP.