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

Table 7

Performance analysis of different modified U-Nets on DRIVE and STARE databases with respect to the measuring metrics.

Method	DRIVE			STARE
	ACC	TPR	TNR	ACC	TPR	TNR
MSFFU-Net [32]	0.9694	0.7762	0.9835	0.9537	0.7721	0.9885
Dense U-net [34]	0.9511	0.7986	0.9736	0.9538	0.7914	0.9722
AA-UNet [53]	0.9558	0.7941	0.9798	0.9640	0.7598	0.9878
DUNet [54]	0.9566	0.7963	0.9800	0.9641	0.7595	0.9878
EEA U-net [55]	0.9577	0.7918	0.9708	0.9445	0.8021	0.9561
Our algorithm	0.9701	0.8011	0.9849	0.9683	0.6032	0.9967