Construction of a predictive model for retinopathy of prematurity using machine learning algorithms
1Department of Ophthalmology, The First Affiliated Hospital of Guangxi Medical University, Nanning, 530021, China.
BMC Ophthalmology
|October 1, 2025
Summary
A machine learning model using 46 risk factors accurately predicts retinopathy of prematurity (ROP). This tool aids in early clinical identification of infants at high risk for ROP, a leading cause of infant blindness.
Area of Science:
- Neonatal Medicine
- Ophthalmology
- Artificial Intelligence in Healthcare
Background:
- Retinopathy of prematurity (ROP) is a significant cause of vision impairment and blindness in newborns globally.
- Early detection and intervention are crucial for managing ROP and preventing vision loss.
Purpose of the Study:
- To develop and evaluate a predictive model for ROP using machine learning techniques.
- To identify key risk factors associated with ROP development in neonates.
Main Methods:
- A retrospective study analyzed 586 neonates undergoing ROP screening.
- Lasso regression identified 46 significant ROP risk factors from extensive data.
- Seven machine learning models were built and compared using performance metrics like AUC and accuracy.
Main Results:
- The Random Forest (RF) model achieved high performance, with an AUC of 0.981 on the testing set.
- The RF model demonstrated strong predictive capabilities with 95.7% accuracy and 0.751 Kappa coefficient on the testing set.
- The model identified 46 significant predictors from an initial 109 screened risk factors.
Conclusions:
- The developed RF predictive model shows significant potential for early ROP risk identification.
- This machine learning approach offers a valuable tool for clinicians to identify high-risk ROP populations proactively.
- The model's strong predictive performance supports its clinical utility in ROP management.
