Table 1.

The average performance metrics for segmenting retinal vessels in the DRIVE testing dataset (n=40).

Architecture	Dice Coefficient	Accuracy	PPV	Sensitivity
U-Net	0.678±0.026	0.872±0.000	0.659±0.052	0.808±0.045
SegNet	0.191±0.018	0.270±0.009	0.901±0.010	0.117±0.013
LinkNet	0.765±0.034	0.961±0.007	0.961±0.006	0.821±0.056
Inception U-Net	0.659±0.026	0.949±0.000	0.781±0.052	0.576±0.034
Attention U-Net	0.780±0.029	0.960±0.000	0.758±0.061	0.816±0.037
Res U-Net	0.798±0.026	0.965±0.000	0.815±0.048	0.790±0.043
R2 U-Net	0.798±0.025	0.965±0.000	0.808±0.052	0.794±0.036
Attn. Res U-Net	0.797±0.025	0.965±0.000	0.802±0.054	0.801±0.0364
U-Net++	0.770±0.035	0.962±0.000	0.810±0.069	0.742±0.047
Res U-Net++	0.755±0.030	0.956±0.000	0.742±0.059	0.780±0.042
Super U-Net	0.808±0.021	0.966±0.000	0.803±0.045	0.818±0.036