Automatic Retinal Vessel Segmentation Based on an Improved U-Net Approach

Table 6

Performance comparison of different methods for the HRF fundus database.

Method	ACC	TPR	TNR
AG-UNet [40]	0.9600	0.8297	0.9875
IterNet [42]	0.9623	0.7524	0.9743
M-GAN [41]	0.9700	0.6948	0.9931
NMF+3D U-Net [43]	0.9688	0.7451	0.9903
Our algorithm	0.9698	0.7478	0.9895