Investigating the influence of oculomotor functions on the TVPS-4 test of visual perceptual skills in school-age children using machine learning

Abstract

Introduction: This study aimed to assess the influence of oculomotor functions and binocularity on visual perception through a standard cognitive test of visual perception. Visual perception, essential for understanding and interacting with the environment, can be affected by impairments that often manifest during childhood and impact academic performance. Although there is evidence suggesting a link between visual perception and binocular vision, research on the exact nature of this relationship is limited. By focusing on school children, this research explored how binocularity and oculomotor functions contribute to visual perception, providing new insights into their potential impact on early cognitive development. Methods: A cross-sectional study was conducted with participants recruited from two school centres in Malaysia. Eligible subjects were registered school children aged 6-14 years, with no severe ophthalmic disorders that would prevent them from completing the assessments. Parental or guardian consent was obtained for all participants. Optometric assessments, performed by a trained investigator, included evaluations of binocular vision, visual acuity, vergence, fusional reserve, accommodation, and stereopsis. Visual perception was measured using the TVPS (4th edition). Data were analyzed descriptively using basic statistical methods and the Orange data mining and machine learning software. Associational analyses included Chi-Squared tests, fast correlation-based filter ranking, and Naive Bayes machine learning. Results: The study revealed strong and meaningful associations between several oculomotor functions and visual perception parameters measured by the Test of Visual Perceptual Skills (TVPS-4). Phoria at near and stereoacuity showed significant positive correlations with discrimination and sequential memory, while phoria at distance was linked to improvements in spatial relationships and figure-ground perception. Near point of convergence and stereoacuity positively impacted visual closure, and positive relative accommodation was related to enhanced visual memory and form constancy. Although principal components analysis (PCA) could not consolidate the data into single components, the use of machine learning provided valuable insights, identifying key oculomotor functions that are strong predictors of visual perception performance. These findings offer new perspectives for improving clinical decision-making and targeted interventions based on oculomotor function. Conclusion: The findings of this study introduce a novel approach to analyzing optometric data and demonstrate the feasibility of using machine learning to predict complex variable relationships in this field. The results provide strong evidence of a significant relationship between binocular vision impairments and visual perception abilities in school children. Notably, phoria and stereopsis were found to have the greatest influence on visual perception.