Table 3.

Results for Scenario 2 Clinic-Wise Cross-Validation Experiments. The best results for each metric and label are highlighted in bold.

	Sensitivity
	MED-Net	DeepLab	SegNet	FCN	UNet
Background	0.89	0.94	0.95	0.90	0.77
Artifact	0.81	0.69	0.67	0.78	0.92
Meshwork	0.67	0.57	0.67	0.66	0.17
Nest	0.50	0.27	0.19	0.37	0.23
Ring	0.79	0.75	0.74	0.71	0.70
Aspecific	0.77	0.72	0.47	0.65	0.87
Average	0.77	0.71	0.69	0.72	0.68
	Specificity
	MED-Net	DeepLab	SegNet	FCN	UNet
Background	0.99	0.93	0.90	0.95	0.91
Artifact	0.92	0.93	0.95	0.91	0.98
Meshwork	0.91	0.90	0.85	0.89	0.95
Nest	0.99	1.00	1.00	1.00	1.00
Ring	0.94	0.91	0.94	0.93	0.94
Aspecific	0.95	0.95	0.98	0.98	0.79
Average	0.94	0.87	0.88	0.88	0.89
	Dice Coefficient
	MED-Net	DeepLab	SegNet	FCN	UNet
Background	0.92	0.85	0.80	0.86	0.92
Artifact	0.76	0.71	0.72	0.73	0.72
Meshwork	0.67	0.59	0.60	0.63	0.25
Nest	0.62	0.41	0.32	0.51	0.37
Ring	0.79	0.73	0.75	0.73	0.73
Aspecific	0.72	0.70	0.57	0.71	0.50
Average	0.77	0.70	0.69	0.72	0.64