Artificial intelligence-assisted quantification of fundus tessellation in early-onset high myopia

To quantitatively evaluate fundus tessellation density (FTD) in children with early-onset high myopia (eo-HM) using artificial intelligence (AI)-assisted image analysis and to explore its association with axial length (AL).

This cross-sectional study included children aged ≤6 years with eo-HM, defined as spherical equivalent (SE) ≤ -6.00 D and/or AL > 26.00 mm. Cycloplegic refraction, AL measurement, and ultra-widefield fundus photography were performed. A deep learning-based algorithm quantified FTD across concentric macular zones (1 mm, 3 mm, 6 mm) and anatomical sectors. Correlation and multivariate regression analyses were used to assess associations between FTD and AL.

A total of 47 eyes from 31 children were analyzed. Mean SE was -9.35 ± 3.74 D and AL was 25.70 ± 1.50 mm. FTD declined from center to periphery (p < 0.001), with greater values in nasal and inferior sectors at 6 mm. Overall FTD correlated with AL (r = 0.46, p = 1.47 × 10^-3). In univariate regression, FTD significantly predicted AL (β = 18.16, p = 1.76 × 10^-5, R ² = 0.352). Multivariable analysis showed that FTD in the 6 mm nasal and superonasal sectors remained independently associated with AL after adjusting for age and sex.

AI-assisted quantification of FTD provides a sensitive and objective measure of early retinal changes in eo-HM. Regional FTD, especially in nasal and perifoveal regions, is strongly associated with axial elongation and may serve as a useful biomarker for early diagnosis and monitoring of pediatric high myopia.