Visual System Alterations for Identifying Teacher-Reported Academic Difficulties in Schoolchildren: A Machine...

Area of Science:

• Ophthalmology
• Educational Psychology
• Machine Learning

Background:

• Efficient visual processing is crucial for academic success in reading, writing, and attention.
• The specific visual domains most predictive of academic difficulties are not well-established.
• Teacher-reported academic difficulties are a significant concern in primary education.

Purpose of the Study:

• To evaluate five visual system domains for their ability to predict teacher-reported academic difficulties in schoolchildren.
• To compare the discriminative value of oculomotor function, accommodation, vergence, and axial length using machine learning.

Main Methods:

• An observational study included 506 primary schoolchildren.
• Academic functioning was teacher-rated and dichotomized to indicate academic difficulties.
• Machine learning classifiers were trained and validated using 5-fold cross-validation to assess five visual predictor groups.

Main Results:

• The accommodative system demonstrated the highest predictive performance (XGBoost: accuracy 0.952).
• Oculomotor function (clinical and instrumental assessment) also showed strong predictive capabilities.
• Vergence and axial length showed limited discriminative value for academic difficulties.

Conclusions:

• Functional visual domains, especially accommodation and oculomotor control, are strong predictors of academic difficulties.
• These findings highlight the potential of visual assessments in identifying children at risk for academic challenges.
• Further validation with larger cohorts and standardized outcomes is necessary for clinical application.