Related Experiment Video
Updated: Jul 6, 2026
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Multisource Machine Learning Model for Detecting Referral-Warranted Retinopathy of Prematurity
Xinwei Luo1, Yong Chen2, Bowen Ying3
1Department of Computer Science and Engineering, Lehigh University, Bethlehem, Pennsylvania.
Insights
A new AI model, MS-ROPNet, accurately detects referral-warranted retinopathy of prematurity (RW-ROP) by combining retinal images and infant demographics. This advanced tool shows promise for improved risk stratification in premature infants.
Area of Science:
- Ophthalmology
- Medical Artificial Intelligence
- Neonatology
Background:
- Retinopathy of prematurity (ROP) is a leading cause of visual impairment in premature infants.
- Early detection of referral-warranted ROP (RW-ROP) is crucial for timely intervention and preventing vision loss.
- Current diagnostic methods can be resource-intensive and require specialized expertise.
Purpose of the Study:
- To develop and evaluate a multisource machine learning model (MS-ROPNet) for detecting RW-ROP.
- To integrate retinal imaging data with infant demographic information for enhanced detection accuracy.
Main Methods:
- A secondary analysis of data from the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity Study was performed.
- The MS-ROPNet model combined a VGG-Swin Transformer for image feature extraction and a random forest for demographic pattern analysis.
- The model was trained and validated using retinal images and demographic data from 1,257 premature infants across 12 clinical centers.
Main Results:
- The MS-ROPNet achieved high performance metrics, including an Area Under the Receiver Operating Characteristic Curve (AUROC) of 95.0 ± 0.7%.
- The model demonstrated excellent sensitivity and specificity, with adjustable cutoffs to prioritize either sensitivity or specificity.
- MS-ROPNet outperformed existing multisource and single-source models in detecting RW-ROP.
Conclusions:
- The MS-ROPNet model effectively integrates retinal images and demographic data for accurate RW-ROP classification.
- This AI-driven approach shows significant potential for precise risk stratification of RW-ROP in premature infants.
- The findings support the clinical utility of advanced machine learning in managing ROP.
Purpose:
To develop a multisource machine learning model for detecting referral-warranted retinopathy of prematurity (RW-ROP) using retinal images and demographics.
Design:
Secondary analysis of data from the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity Study.
Subjects:
One thousand two hundred fifty-seven premature infants (mean birth weight 864 g; mean gestational age 27 weeks; 19.4% with RW-ROP) enrolled from 12 clinical centers in North America.
Methods:
A multisource ROPNet (MS-ROPNet) model that combines a VGG-Swin Transformer model to extract features from retinal images and a random forest model that captures patterns of demographics was developed using central-view retinal images from 7741 eye visits with concurrent clinical eye examinations and demographic characteristics (birth weight, gestational age, sex, ethnicity, and age at retinal image). The MS-ROPnet model was compared to several existing machine learning models for detecting RW-ROP.
Main Outcome Measures:
Model performance metrics including the area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), sensitivity, specificity, and accuracy from fivefold cross-validation, with RW-ROP diagnosed by certified ophthalmologists as the reference standard.
Results:
The MS-ROPNet achieved an AUROC of 95.0 ± 0.7% (mean ± standard deviation), AUPRC of 78.0 ± 2.8%, sensitivity of 81.5 ± 4.0%, specificity of 94.3 ± 1.4%, and accuracy of 93.0 ± 0.9% under the default cutoff of 0.50 in predicted probability. After adjusting the cutoff to achieve higher sensitivity, MS-ROPNet had 90% sensitivity and specificity of 84.8 ± 2.5% (using cutoff 0.3504), and 95% sensitivity with specificity of 72.8 ± 6.9% (using cutoff 0.2074), which outperformed both existing multisource models and the best single-source model by ≥3.7% in specificity at the same sensitivity.
Conclusions:
The MS-ROPNet achieved high performance in RW-ROP classification by effectively integrating retinal images with demographics, demonstrating its potential for accurate risk stratification of RW-ROP.
Financial Disclosures:
The authors have no proprietary or commercial interest in any materials discussed in this article.
