Clinical Report: Retinal AI Predicts Neonatal Lung Disease

Overview

Deep learning models using retinal images, specifically trained on data from the i-ROP study, can predict bronchopulmonary dysplasia and pulmonary hypertension in premature infants. This study highlights the potential of integrating retinal imaging into routine neonatal care for early identification of at-risk infants.

Background

The ability to predict serious lung diseases in premature infants is crucial, as conditions like bronchopulmonary dysplasia and pulmonary hypertension significantly impact morbidity and mortality. Current screening practices for retinopathy of prematurity, which involve regular retinal imaging, provide an opportunity to leverage this existing data for broader diagnostic purposes. This study explores the feasibility of using retinal images to identify infants at risk for severe cardiopulmonary complications.

Data Highlights

Key Findings

• Deep learning models can predict bronchopulmonary dysplasia (BPD) and pulmonary hypertension (PH) using retinal images.
• The multimodal model for BPD achieved an AUC of 0.82, outperforming demographics-only and imaging-only models, indicating better predictive accuracy.
• For PH, the imaging-only model achieved an AUC of 0.91, indicating strong diagnostic capability.
• Secondary models without visible retinopathy signs yielded consistent predictive results.
• Findings suggest potential biological links between retinal changes and cardiopulmonary conditions.

Clinical Implications

Integrating retinal imaging into routine screening for retinopathy of prematurity may facilitate earlier identification of infants at risk for severe lung diseases. This approach could enhance clinical decision-making and improve outcomes for premature infants by enabling timely interventions, although challenges in implementation and training may arise.

Conclusion

This study provides proof of concept for using retinal imaging to predict neonatal lung disease, highlighting a promising avenue for improving risk stratification in premature infants. Further research is needed to validate these findings across diverse clinical settings and to address limitations such as the small pulmonary hypertension cohort and lack of external validation.

References

1. Ophthalmology Management, 2026 -- Study Shows Retinal Scans Predict Neonatal Lung Disease
2. Ophthalmology Management, 2023 -- AI Advances for Diabetic Retinopathy
3. Retinal Physician, 2025 -- Study: AI Delivers High Accuracy in IRD Diagnosis
4. Screening Examination of Premature Infants for Retinopathy of Prematurity, 2018
5. Bronchopulmonary dysplasia - PMC, 2020
6. Retinal Physician — Study: AI Delivers High Accuracy in IRD Diagnosis
7. Screening Examination of Premature Infants for Retinopathy of Prematurity
8. Bronchopulmonary dysplasia - PMC
9. AHA/ATS Guideline