AI Detects Lung Disease in Premature Babies Through Eye Scans

Doctors may soon spot lung disease in premature infants weeks earlier than ever before, thanks to artificial intelligence that reads signs hidden in routine eye exams.

Researchers at the University of Colorado School of Medicine trained a deep learning model to predict bronchopulmonary dysplasia and pulmonary hypertension by analyzing retinal images already taken during standard screenings for premature babies. The AI achieved 82% accuracy for lung disease and an impressive 91% accuracy for heart-related complications.

The system works by examining photos taken during regular retinopathy of prematurity screenings, which nearly all premature infants already receive. Instead of requiring new tests or procedures, doctors could use images they're already capturing to identify babies at risk for serious breathing problems.

The study analyzed 493 infants at risk for eye complications related to premature birth. When researchers combined the AI's image analysis with standard demographic information like birth weight and gestational age, the predictions became significantly more accurate than using patient history alone.

What makes this discovery particularly exciting is that the AI detected disease markers even in eye images that showed no visible signs of retinopathy. The algorithm identified patterns invisible to the human eye, suggesting connections between eye development and lung health that doctors didn't know existed.

The Ripple Effect

This technology could transform care in neonatal intensive care units worldwide. Premature babies often face multiple health challenges simultaneously, and catching lung problems early gives doctors crucial time to intervene before conditions become life-threatening.

The approach also highlights a broader shift in medicine toward using existing data in smarter ways. Rather than subjecting fragile newborns to additional tests, doctors could extract more information from procedures already happening, reducing stress on babies and families while improving outcomes.

For parents of premature infants, the weeks and months in the NICU involve constant worry about their baby's developing organs. Early warning systems like this one could help medical teams stay ahead of complications, turning anxious waiting into proactive treatment.

The research team acknowledges these results need validation across larger, more diverse patient populations and different camera systems before widespread clinical use. However, the initial findings open an exciting new frontier in newborn care, where artificial intelligence serves as an extra set of eyes watching over the smallest patients.

The study, published in JAMA Ophthalmology, represents years of collaboration between ophthalmologists, pulmonologists, and computer scientists working together to give premature babies the best possible start.