Table 2.

Performance of deep learning models for binary task (axial length ≥ 26.0 mm).

	Mean results of 10-fold cross validation				Test set (95% CI)
	AUC	Accuracy	Sensitivity	Specificity	AUC	Accuracy	Sensitivity	Specificity
ResNet 18	0.908 ± 0.048	0.898 ± 0.042	0.807 ± 0.107	0.997 ± 0.008	0.918 (0.886, 0.951)	0.872 (0.840, 0.899)	0.783 (0.704, 0.848)	0.902 (0.869, 0.929)
ResNet 50	0.932 ± 0.048	0.906 ± 0.033	0.920 ± 0.082	1.000	0.929 (0.899, 0.960)	0.872 (0.840, 0.899)	0.804 (0.728, 0.867)	0.895 (0.861, 0.923)
ViT	0.885 ± 0.075	0.884 ± 0.051	0.766 ± 0.151	1.000	0.924 (0.892, 0.955)	0.867 (0.836, 0.895)	0.693 (0.609, 0.769)	0.927 (0.897, 0.951)