R2AU-Net: Attention Recurrent Residual Convolutional Neural Network for Multimodal Medical Image Segmentation

Table 2

Performance comparison of R2AU-Net and other networks on the DRIVE dataset.

Methods	F1-score	Sensitivity	Specificity	Accuracy	AUC
Hybrid features [22]	—	0.7252	0.9798	0.9474	0.9648
Trainable COSFIRE filters [23]	—	0.7655	0.9704	0.9442	0.9614
Three-stage filtering [24]	—	0.7250	0.9830	0.9520	0.9620
Deep model [25]	—	0.7763	0.9768	0.9495	0.9720
Cross-modality [26]	—	0.7569	0.9816	0.9527	0.9738
SegNet [19]	0.7992	0.7419	0.9833	0.9526	0.9752
U-Net [3]	0.8155	0.7908	0.9783	0.9544	0.9775
Attention U-Net [20]	0.8003	0.7272	0.9868	0.9538	0.9771
RU-Net [21]	0.8180	0.7999	0.9772	0.9547	0.9773
R2U-Net [21]	0.8187	0.7980	0.9779	0.9550	0.9775
R2AU-Net	0.8213	0.8036	0.9777	0.9555	0.9790