Related Experiment Video
Updated: Dec 31, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
A practical model for the identification of congenital cataracts using machine learning
Duoru Lin1, Jingjing Chen1, Zhuoling Lin1
1State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Jinsui Road #7, Guangzhou, Guangdong 510060, People's Republic of China.
Ebiomedicine
|January 6, 2020
Summary
This study developed accurate models to identify infants at high risk for congenital cataracts (CCs), a leading cause of childhood blindness. These models can aid in early detection, especially in underserved regions.
Area of Science:
- Ophthalmology
- Pediatrics
- Public Health
Background:
- Congenital anomalies affect approximately 1 in 33 newborns globally.
- Congenital cataracts (CCs) are the primary cause of preventable childhood blindness.
- Effective identification models for infants at high risk of CCs are needed.
Purpose of the Study:
- To develop a practical and accurate model for identifying infants at high risk of congenital cataracts (CCs).
- To assess the model's performance in diverse clinical settings, including those with low disease prevalence.
Main Methods:
- A case-control study involving 2005 subjects (1274 CC cases, 731 controls).
- Development of CC identification models using random forest and adaptive boosting, based on birth conditions, family history, and environmental factors.
- Validation through internal cross-validation, external validation, and simulated clinical environments with varying CC prevalence.
Main Results:
- The developed CC identification models demonstrated high discrimination accuracy (AUC=0.91-0.96).
- Key risk factors identified include family history of CC, low parental education, and comorbidity.
- Models maintained stable performance across validation tests and simulated clinical scenarios.
Conclusions:
- The CC identification models accurately distinguish CC patients from healthy children.
- These models show potential as a complementary screening tool for congenital cataracts.
- The models are particularly valuable for screening in undeveloped and remote areas with limited healthcare access.
