Table 2 AUROC values and F1-scores for each fold and the mean over all folds of the five-fold cross validation for the different methods.

From: Glaucoma detection using non-perfused areas in OCTA

Plexus	Method	Fold1		Fold2		Fold3		Fold4		Fold5		Mean
Plexus	Method	AUROC	F1	AUROC	F1	AUROC	F1	AUROC	F1	AUROC	F1	AUROC	F1
Retina	CNN	0.89	0.86	0.93	0.86	0.78	0.68	0.95	0.86	0.85	0.72	0.88	0.79
	PDA	0.86	0.82	0.92	0.83	0.72	0.62	0.91	0.83	0.87	0.73	0.86	0.766
	GPDP	0.85	0.77	0.89	0.76	0.61	0.59	0.85	0.76	0.85	0.75	0.81	0.726
	PD	0.85	0.79	0.89	0.83	0.60	0.56	0.86	0.76	0.85	0.80	0.81	0.748
	VD	0.76	0.69	0.74	0.65	0.57	0.56	0.83	0.75	0.78	0.69	0.74	0.668
	IA	0.67	0.63	0.58	0.58	0.58	0.51	0.66	0.56	0.69	0.58	0.64	0.572
SVP	CNN	0.96	0.85	0.92	0.82	0.90	0.79	0.94	0.89	0.95	0.86	0.93	0.842
	PDA	0.93	0.89	0.90	0.81	0.86	0.78	0.94	0.87	0.92	0.82	0.91	0.834
	PD	0.91	0.86	0.90	0.86	0.81	0.73	0.93	0.83	0.89	0.75	0.89	0.806
	GPDP	0.88	0.79	0.91	0.80	0.75	0.65	0.92	0.79	0.89	0.79	0.87	0.764
	VD	0.83	0.70	0.85	0.72	0.76	0.72	0.92	0.87	0.85	0.76	0.84	0.754
	IA	0.76	0.77	0.88	0.79	0.80	0.73	0.82	0.77	0.84	0.70	0.82	0.752
ICP	PDA	0.83	0.75	0.90	0.75	0.85	0.70	0.92	0.83	0.89	0.79	0.88	0.764
	CNN	0.87	0.73	0.90	0.75	0.76	0.68	0.90	0.79	0.89	0.80	0.86	0.750
	PD	0.84	0.76	0.91	0.74	0.72	0.70	0.92	0.82	0.87	0.77	0.85	0.758
	GPDP	0.83	0.73	0.89	0.72	0.73	0.68	0.90	0.79	0.81	0.72	0.83	0.728
	VD	0.73	0.70	0.78	0.75	0.66	0.61	0.90	0.82	0.81	0.72	0.78	0.720
	IA	0.64	0.58	0.52	0.55	0.66	0.63	0.77	0.69	0.58	0.49	0.63	0.588
DCP	CNN	0.82	0.70	0.97	0.87	0.82	0.75	0.92	0.86	0.82	0.72	0.87	0.780
	PDA	0.80	0.75	0.94	0.87	0.71	0.68	0.88	0.82	0.75	0.70	0.82	0.764
	GPDP	0.73	0.63	0.93	0.79	0.64	0.59	0.90	0.83	0.71	0.62	0.78	0.692
	PD	0.80	0.73	0.86	0.79	0.65	0.61	0.84	0.73	0.71	0.69	0.77	0.717
	VD	0.60	0.59	0.82	0.66	0.64	0.61	0.80	0.72	0.61	0.61	0.70	0.638
	IA	0.52	0.48	0.62	0.59	0.57	0.48	0.57	0.51	0.53	0.48	0.562	0.508

For each plexus the methods are sorted according to their mean AUROC performance in descending order. The proposed features are highlighted in bold.
(CNN convolutional neural network, PDA perfusion deficite areas, GPDP geometric perfusion deficite percentage, PD perfusion distance, VD vessel density, IA intercapillary areas) and plexuses (Retina retina, SVP superficial vascular plexus, ICP intermediate capillary plexus, DCP deep capillary plexus) on the test set.

Back to article page

Table 2 AUROC values and F1-scores for each fold and the mean over all folds of the five-fold cross validation for the different methods.

Search

Quick links