Deep Learning Based Prediction of Myopia Control Effect in Children Treated With Overnight Orthokeratology

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Orthokeratology (Ortho-K) is a myopia control method using rigid gas permeable lenses.
• Predicting axial length (AL) elongation is crucial for managing myopia progression in children.
• Early identification of factors influencing AL elongation can optimize treatment outcomes.

Purpose of the Study:

• To develop and validate a deep learning model for predicting 12-month AL elongation in pediatric Ortho-K patients.
• To identify baseline factors and early corneal topographic changes associated with AL elongation.
• To explore the link between these predictive factors and the effectiveness of myopia control.

Main Methods:

• A cohort of 115 pediatric Ortho-K patients was analyzed.
• Baseline factors correlated with AL elongation were identified using Pearson correlation coefficients.
• A prediction model combining multiple linear regression and deep neural networks was developed and validated on independent patient groups.

Main Results:

• Age, spherical equivalent refractive error (SE), and sex were significant predictors of AL elongation (P <0.05).
• Corneal topographic variations, including height, area, and volume of the defocus region, were incorporated into the model.
• The validated deep learning model demonstrated high accuracy in predicting 12-month AL elongation.

Conclusions:

• A deep learning model effectively predicts axial length elongation in pediatric Ortho-K patients.
• The model integrates baseline patient data and corneal topographic changes for accurate predictions.
• This predictive tool can enhance myopia management strategies and personalize Ortho-K treatment.