TABLE 2.

Five-fold cross-validation of the performance of the algorithms in the total dataset.

		AUC (95% CI)	Accuracy (95% CI)	Specificity (95% CI)	Sensitivity (95% CI)
Algorithm I		0.995 (0.993, 0.996)	0.973 (0.969, 0.977)	0.981 (0.978, 0.985)	0.939 (0.933, 0.945)
		Macro-AUC	Accuracy (95% CI)	Quadratic-weighted kappa (95% CI)
Algorithm II		0.979 (0.972, 0.985)	0.967 (0.963, 0.971)	0.988 (0.986, 0.990)
		Image classification			ROI detection and lesion localization
	Classification	Accuracy (95% CI)	Specificity (95% CI)	Sensitivity (95% CI)	Recall	Precision	F1-score
Algorithm III	CNV	0.970 (0.966, 0.974)	0.970 (0.966, 0.974)	0.973 (0.969, 0.977)	0.916	0.789	0.848
	Fuchs	0.971 (0.967, 0.975)	0.971 (0.967, 0.975)	0.978 (0.975, 0.982)	0.915	0.864	0.889
	LC	0.994 (0.992, 0.995)	0.995 (0.993, 0.996)	0.684 (0.672, 0.695)	0.724	0.656	0.688
	Accuracy (95% CI)		Sensitivity (95% CI)	Specificity (95% CI)		Precision (95% CI)
Model-1	0.973 (0.969, 0.977)		0.939 (0.933, 0.945)	0.981 (0.978, 0.985)		0.926 (0.920, 0.933)
Model-2	0.984 (0.981, 0.987)		0.946 (0.941, 0.952)	0.992 (0.990, 0.995)		0.967 (0.963, 0.972)