Clinical Scorecard: Personalized Predictions and Interventions for Myopia Progression Using AI Technology
At a Glance
Category
Detail
Condition
Myopia and high myopia progression
Key Mechanisms
Axial length growth and spherical equivalent changes predicted via Transformer-based AI models
Target Population
Children and adolescents aged 8 to 18 years with myopia
Care Setting
Ophthalmology clinics and specialized eye hospitals
Key Highlights
High prevalence of myopia in Asia, with up to 80% of Chinese high school graduates affected and 10–20% having high myopia.
Current myopia control interventions include low-dose atropine, orthokeratology lenses, peripheral defocus spectacles, and repeated low-level red light therapy, each with specific limitations and risks.
Development of a Transformer-based Myopia Progression Predictive Model (MPPM) with Natural Progression Module (NPM) and Intervention Progression Module (IPM) enables individualized, long-term predictions of myopia progression and treatment effects.
Guideline-Based Recommendations
Diagnosis
Use longitudinal measurements of spherical equivalent (SE) and axial length (AL) to monitor myopia progression.
Apply machine-learning-based imputation to address missing AL data when necessary.
Management
Consider early intervention in high-risk pediatric patients using atropine 0.01%, orthokeratology lenses, peripheral defocus spectacles, or repeated low-level red light therapy.
Utilize AI-based predictive models to tailor intervention choice and timing based on individualized progression risk and expected treatment benefit.
Monitoring & Follow-up
Regular follow-up visits with refraction and axial length measurements to assess progression and treatment response.
Use AI model outputs to adjust management plans dynamically over the typical 10-year progression period.
Risks
Atropine may cause photophobia, transient near-vision impairment, or allergic reactions.
Orthokeratology lenses carry risks of corneal epithelial injury and infection.
Peripheral defocus spectacles involve high fitting costs limiting accessibility.
Repeated low-level red light therapy may pose potential retinal phototoxicity risks with long-term use.
Patient & Prescribing Data
Pediatric myopia patients aged 8–18 years receiving various myopia control interventions
AI models trained on large longitudinal cohorts enable prediction of individual treatment efficacy, optimizing intervention selection and potentially reducing unnecessary exposure to risks and costs.
Clinical Best Practices
Collect comprehensive longitudinal data including age, sex, SE, and AL for accurate prediction modeling.
Incorporate AI-based tools such as the MPPM to forecast natural progression and intervention outcomes for personalized care.
Balance benefits and risks of each intervention considering patient-specific factors and predicted treatment effects.
Ensure regular monitoring and adjust treatment plans based on ongoing progression and AI model feedback.
