Table 1.

Data-driven artificial intelligence (AI) models in high myopia and pathologic myopia.

Research	Year	Materials	Participants	AI Methods	Main Outcome	Evolutions and Performance
Lin, H. et al. [50]	2018	Refraction data	School-aged children	ML	Predicting the presence of high myopia	AUC: 0.802–0.976
Kaya, C. et al. [53]	2018	electrooculographic data	Adults (25–65 years old)	ML	Detecting hypermetropia and myopia refractive disorders	Sensitivity: 95.5%; specificity: 96%; classification accuracy: 90.91%
Ye, B. et al. [55]	2019	luminance, ultraviolet light levels, and step number data	Myopia patients	ML	Differentiating indoor and outdoor locations	Accuracy: 0.827–0.996; AUC: 0.90–0.99
Rampat, R. et al. [51]	2020	Wavefront aberrometry data	General population	ML	Predicting subjective refraction	mean absolute error: 0.094–0.301 diopters
Tang, T. et al. [54]	2020	Medical data	School-age myopic children	ML	Estimating physiological elongation of axial length	R square equals 0.87
Wei, L. et al. [52]	2020	Medical data	Myopia patients	ML	Improving the accuracy of IOL power predictions	mean absolute error: 0.25–0.29; median squared errors: 0.06–0.09
Yang, X. et al. [56]	2020	Medical data	Primary school children	ML	Studying influence of related factors on incidence of myopia in adolescents	Accuracy equals 0.92–0.93; Precision equals 0.95; Sensitivity equals 0.94; f1 equals 0.94; AUC equals 0.98; Specificity equals 0.94
Li, S.M. et al. [57]	2022	Medical data	Primary school children	ML	Detecting risk factors for myopia progression	Combined weight: 77%; Accuracy: over 80%

AUC, area under the receiver operating characteristic curves; ML, machine learning.