Clinical Report: EGS-Net: An Enhanced Machine Learning Approach for Forecasting High-Myopia Risk
Overview
This study presents the EGS predictive framework, which utilizes longitudinal school screening data to improve the prediction of high-myopia risk in children. The model demonstrated high recall and precision.
Background
The increasing prevalence of myopia, particularly high myopia, poses significant public health challenges. Early detection and intervention are critical to prevent the progression of myopia and associated complications. Traditional screening methods often fall short, necessitating innovative approaches like machine learning to enhance predictive accuracy.
Data Highlights
The EGS framework achieved a Recall of 0.9533 and Precision of 0.9211, with an AUC of 0.9992 for the AdaBoost model.
Key Findings
The EGS model integrates a multi-model ensemble architecture with a clinical risk-heuristic override module.
Longitudinal refractive trajectories were critical for improving model interpretability and prediction accuracy.
High Recall and Precision metrics indicate the model's potential for reliable future-risk identification.
5-fold cross-validation was employed for model tuning and evaluation.
SHAP analysis confirmed the importance of longitudinal features in the model.
Clinical Implications
The EGS framework offers a scalable solution for school-based myopia surveillance, prioritizing high-risk recall to facilitate timely clinical interventions.
Conclusion
The EGS predictive framework represents a significant advancement in the use of machine learning for myopia risk assessment.
