Prediction of myopia onset and shift in premyopic school-aged children: a machine learning-based algorithm

Area of Science:

• Ophthalmology
• Pediatric Ophthalmology
• Computational Biology

Background:

• Myopia is a growing global health concern, particularly in school-aged children.
• Early identification of children at risk for myopia progression is crucial for timely intervention.
• Understanding longitudinal changes in ocular parameters is key to predicting myopia development.

Purpose of the Study:

• To investigate longitudinal changes in ocular parameters in premyopic children.
• To develop and validate a machine learning model for predicting myopia onset and shift within one year.
• To identify key predictive factors for myopia progression in this cohort.

Main Methods:

• Prospective cohort study of 320 premyopic children (aged 6-12 years).
• Regular measurements of visual acuity, spherical equivalent (SE), axial length (AL), corneal curvature (CC), and subfoveal choroidal thickness (SFCT).
• Machine learning algorithms employed for prediction, with SHAP analysis for feature interpretation.

Main Results:

• 49.3% of participants developed myopia within one year.
• Significant annual SE progression (-0.695 D) and AL elongation (0.356 mm).
• Machine learning model achieved high accuracy (AUC-ROC 0.963) for myopia onset prediction, with SE, parental myopia, SFCT, and age as key predictors.

Conclusions:

• Premyopic children demonstrate accelerated myopia progression.
• Machine learning models offer a promising approach for predicting myopia onset and progression.
• These predictive models can facilitate risk stratification and targeted prevention strategies.