Catalyzing early intervention: A robust deep learning approach for detecting retinopathy of prematurity and plus diseases in infants

Retinopathy of prematurity (ROP) is a major cause of childhood blindness, requiring precise and prompt diagnosis. This paper presents a new method that uses deep convolutional neural network (DCNN) models—VGG19, ResNet101, and DenseNet169—to identify illnesses in ROP fundus images. 2776 pictures from the Al-Amal Eye Centre were included in the dataset, with an equal number of normal and plus illness cases. The models underwent thorough training and evaluation, with VGG19 emerging as the most accurate, with an impressive 97.07% accuracy. The study emphasizes the potential benefits of using these models to improve ROP screening programs by offering a consistent and effective method of diagnosis that can greatly impact clinical decision-making. This research enhances the field of neonatal ophthalmology and provides vital insights for enhancing patient care in managing ROP.