Deep Learning Assisted Retinopathy of Prematurity (ROP) Screening

Retinopathy of prematurity (ROP) is a leading cause blindness in premature infants worldwide, particularly developing countries. In this research, we propose Deep Convolutional Neural Network (DCNN) and image processing-based approach for the automatic detection retinal features, including optical disc (OD) blood vessels (BV), as well disease classification using rule-based method ROP patients. Our DCNN model uses YOLO-v5 OD either Pix2Pix, U-Net or BV segmentation. We trained our models on publicly available fundus datasets size 1,117, 288 segmentation, respectively. evaluated dataset 439 preterm neonatal images, testing Zone 6 masks. proposed system achieved excellent results, with module achieving an overall accuracy 98.94% (when IoU 0.5) segmentation 96.69% Dice coefficient between 0.60 0.64. Moreover, accurately diagnosed Zone-1 88.23% accuracy. offers promising solution accurate screening diagnosis, low-resource settings, where it has potential to improve healthcare outcomes.