Related Experiment Video
Updated: Jan 9, 2026
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
7.9K
Optimizing myopia prediction in children and adolescents using machine learning: a multi-factorial risk assessment
Yue Xi1, Wei Zhu2, Wenjing Yan3
1School of Physical Education, Shanghai University, Shanghai, China.
Frontiers in Medicine
|December 1, 2025
Summary
This study identified modifiable behavioral and sociodemographic factors contributing to childhood myopia. Machine learning models highlighted parental myopia and physical activity as key predictors for targeted prevention strategies.
Area of Science:
- Ophthalmology
- Public Health
- Data Science
Background:
- Childhood myopia is a growing concern, with most research focusing on genetics and environment.
- This study investigates modifiable behavioral, sociodemographic, and psychological factors.
- It also explores the utility of machine learning (ML) in identifying at-risk children.
Purpose of the Study:
- To identify modifiable risk factors for myopia in children and adolescents.
- To evaluate the effectiveness of machine learning models in predicting myopia.
- To inform targeted prevention and intervention strategies.
Main Methods:
- A cross-sectional survey of 2,086 students in Chinese schools.
- LASSO and logistic regression for risk factor identification.
- Ten machine learning algorithms (including LightGBM) were used for prediction, with SHAP for interpretation.
Main Results:
- Myopia prevalence was 25.12% in the study population.
- Key independent risk factors included parental myopia, only-child status, and physical activity.
- LightGBM demonstrated the best predictive performance (AUC = 0.738).
Conclusions:
- Machine learning models identified significant modifiable risk factors for childhood myopia.
- Findings suggest potential for targeted behavioral interventions and prevention strategies.
- ML models are valuable for identifying risk factors but not yet for clinical diagnosis.
