AI model accurately detects retinopathy of prematurity needing specialist referral
Multisource Machine Learning Model for Detecting Referral-Warranted Retinopathy of Prematurity.
Xinwei Luo … Lifang He
Ophthalmology science · 2026
Purpose
To develop a multisource machine learning model for detecting referral-warranted retinopathy of prematurity (referral-warranted retinopathy of prematurity) using retinal images and demographics.
Methods
SUBJECTS: One thousand two hundred fifty-seven premature infants (mean birth weight 864 g; mean gestational age 27 weeks; 19.4% with referral-warranted retinopathy of prematurity) enrolled from 12 clinical centers in North America.
95.0%
Results
95.0%
Model correctly distinguished infants needing referral from those who didn't, 95% of the time (AUROC).
“
Conclusion · 1 of 2
The multisource ROPNet achieved high performance in referral-warranted retinopathy of prematurity classification by effectively integrating retinal images with demographics, demonstrating its potential for accurate risk stratification of referral-warranted retinopathy of prematurity.