MHA-Net: A Multibranch Hybrid Attention Network for Medical Image Segmentation

Table 3

Eight standard performance measure values of different segmentation methods on the polyp dataset.

Method	Rec	Spec	Prec	DC	IoUp	IoUb	mIoU	Ac
U-Net [5]	0.6530	0.9940	0.8760	0.6481	0.5762	0.9564	0.7663	0.9588
UNet++ [31]	0.7705	0.9565	0.6375	0.6385	0.5401	0.9460	0.7431	0.9481
Ce-Net [16]	0.8129	0.9701	0.8167	0.7569	0.6873	0.9471	0.8172	0.9504
PraNet [24]	0.7820	0.9862	0.8110	0.7435	0.6644	0.9607	0.8126	0.9635
HarDNet-MSEG [32]	0.5886	0.9988	0.9255	0.6490	0.5736	0.9563	0.7649	0.9578
Our approach	0.8000	0.9905	0.8618	0.7692	0.6982	0.9640	0.8311	0.9656