Related Experiment Video
Updated: Jun 17, 2025
07:20
Inducement and Evaluation of a Murine Model of Experimental Myopia
Published on: January 22, 2019
9.8K
A deep learning system for myopia onset prediction and intervention effectiveness evaluation in children
Ziyi Qi1,2, Tingyao Li3,4, Jun Chen1
1Department of Clinical Research, Shanghai Eye Disease Prevention and Treatment Center, Shanghai Eye Hospital, Shanghai Vision Health Center & Shanghai Children Myopia Institute, Shanghai, China.
NPJ Digital Medicine
|August 7, 2024
Summary
DeepMyopia, an AI system, accurately predicts myopia onset in children using retinal images. This AI tool aids early intervention, significantly reducing myopia progression and preventing vision loss.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Public Health
Background:
- Myopia prevalence is rising globally, posing a significant public health concern.
- Early detection and prediction of myopia in children are crucial for effective intervention.
- Retinal fundus images offer a readily accessible method for myopia examination.
Purpose of the Study:
- To introduce DeepMyopia, an AI-enabled decision support system for myopia detection and prediction in children.
- To facilitate targeted interventions for children at risk of myopia using routine retinal fundus images.
- To evaluate the performance and effectiveness of DeepMyopia in a large-scale cohort.
Main Methods:
- DeepMyopia utilizes a deep learning architecture trained on over 1.6 million retinal fundus images.
- Internal validation was performed on a large cohort, followed by external testing on datasets from seven sites in China.
- Performance was assessed using Area Under the Curve (AUC) for myopia onset prediction and risk stratification.
Main Results:
- DeepMyopia achieved high AUCs for 1-, 2-, and 3-year myopia onset prediction (0.908, 0.813, 0.810 internally; 0.796, 0.808, 0.767 externally).
- The system effectively stratified children into low- and high-risk groups (p < 0.001).
- In an emulated trial, DeepMyopia interventions showed a significant reduction in myopia progression (-17.8% ARR) and improved quality-adjusted life years.
Conclusions:
- DeepMyopia is a robust and efficient AI-based decision support system for guiding myopia interventions in children.
- The system demonstrates strong predictive capabilities and effectiveness in risk stratification.
- DeepMyopia holds potential for widespread application in public health strategies to combat the increasing prevalence of childhood myopia.
