Deep learning-based early detection of retinopathy of prematurity from retinal images of preterm infants

Al Husaini, Mohammed Abdulla Salim and Mahfouz, Ahmed Muhammed and Abrar, Mohammad and Al Husaini, Yousuf Nasser and Habaebi, Mohamed Hadi (2026) Deep learning-based early detection of retinopathy of prematurity from retinal images of preterm infants. In: 2025 10th International Conference on Computer and Communication Engineering (ICCCE), 26-27 August 2025, KOE, IIUM.

PDF - Published Version Restricted to Repository staff only Download (406kB) | Request a copy

Abstract

Abstract—Retinopathy of Prematurity (ROP) is a significant cause of blindness in premature infants, and early detection is crucial for preventing severe visual impairment. Traditional diagnosis methods are subjective and prone to variability. This study aims to develop an automated deep learning system for detecting and grading ROP from retinal images, aiming to improve diagnostic accuracy, speed, and accessibility. A dataset of 6100 RetCam III retinal images, annotated for lesion type, location, ROP stage, zone, and plus disease presence, was used to train the proposed deep learning system. The system employed pre-trained deep learning models like InceptionV3, ResNet50, and ResNet101. The system was trained using these models to classify ROP severity based on clinical guidelines, with results evaluated against human expert assessments. Diagnostic accuracy metrics, including sensitivity, specificity, and AUC, were calculated to assess performance. The dataset was randomly divided into training (70%), validation (20%), and testing (10%) sets. The model achieved an AUC of 0.95, an F1 score of 0.76, and significant improvements over traditional methods. The deep learning system showed strong performance in detecting ROP stages, zones, and plus disease, outpacing human experts in classification accuracy. This study demonstrates the efficacy of a deep learning-based system for diagnosing and grading ROP. The system can provide accurate, automated assessments of ROP severity, improving screening efficiency and consistency. With further optimization and larger, more diverse datasets, the system has the potential to significantly enhance early detection and treatment of ROP, particularly in underserved regions.

Item Type:	Proceeding Paper (Plenary Papers)
Uncontrolled Keywords:	Deep learning, Retinopathy of Prematurity, InceptionV3, ResNet50, ROP classification, Fundus images, AI diagnostics.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7800 Electronics. Computer engineering. Computer hardware. Photoelectronic devices > TK7885 Computer engineering
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button):	Kulliyyah of Engineering Kulliyyah of Engineering > Department of Electrical and Computer Engineering
Depositing User:	Dr. Mohamed Hadi Habaebi
Date Deposited:	29 Apr 2026 16:18
Last Modified:	29 Apr 2026 16:18
Queue Number:	2026-04-Q2958
URI:	http://irep.iium.edu.my/id/eprint/128495

Actions (login required)

View Item