Automated diagnosis of plus form and early stages of ROP using deep learning models

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Retinopathy of Prematurity (ROP) is a significant cause of vision loss in premature infants.
• Lower birth weight (BW) and gestational age (GA) are associated with increased ROP severity.
• Early detection and intervention are crucial for managing ROP and preventing irreversible vision impairment.

Purpose of the Study:

• To develop and evaluate an automated diagnostic system for classifying Plus disease and ROP stages.
• To assess the system's accuracy in identifying abnormal retinal vascularity and disease progression.
• To explore the potential of AI in supporting clinical screening for ROP.

Main Methods:

• Utilized a curated dataset of retinal fundus images.
• Developed a model for binary classification of Plus disease (Plus/Normal).
• Implemented a model for multi-class classification of ROP stages (Stage 0-3).

Main Results:

• Achieved high diagnostic accuracy: 0.996 for Plus disease detection.
• Demonstrated strong performance in ROP stage classification with 0.98 accuracy.
• The model shows significant potential for automated ROP screening.

Conclusions:

• The automated system shows high accuracy in classifying Plus disease and ROP stages.
• This AI-driven approach may enhance early diagnosis and timely intervention for ROP.
• Further multi-center validation is recommended to confirm clinical utility in reducing vision impairment in preterm infants.