AI analysis of routine eye photos spots lung and heart risks in premature infants
Mar 2nd 2026
A deep learning model analyzed retinal photos taken during standard retinopathy of prematurity screenings and identified risk for bronchopulmonary dysplasia and pulmonary hypertension in premature infants with high accuracy, but researchers say more validation is needed before clinical use.
- Researchers trained a deep learning model on retinal images from routine ROP screenings of 493 infants across seven NICUs.
- The combined imaging and clinical-data model predicted bronchopulmonary dysplasia with 82% accuracy and pulmonary hypertension with 91% accuracy.
- Accuracy remained when images showing clinical signs of ROP were excluded, suggesting the model detects signals beyond eye disease markers.
- Retinal imaging is already standard for retinopathy of prematurity screening, so the method may not require additional procedures.
- The study was published in JAMA Ophthalmology using data from the multi institution i-ROP consortium.
- Researchers say further validation is needed before the tool can be integrated into routine clinical care.