Segmentation of Breast Tubules in H&E Images Based on a DKS-DoubleU-Net Model

Table 2

The quantitative comparison of tubule segmentation results obtained by different models on the test set of the public BRACS dataset.

Model	Dice	mIou	Recall	Precision	F1-Score
DKS-DoubleU-Net	0.9272	0.8655	0.9292	0.9305	0.9298
DoubleU-Net+DenseNet	0.9247	0.8617	0.9288	0.9259	0.9273
DoubleU-Net+KSM	0.9269	0.8655	0.9328	0.9264	0.9296
DoubleU-Net	0.9219	0.8570	0.9204	0.9319	0.9261
U-Net	0.8793	0.7887	0.9239	0.8537	0.8874
ResNet++	0.6547	0.5020	0.7450	0.6450	0.6914
DeepLabV3+	0.8413	0.7301	0.8547	0.8483	0.8515
HRNet	0.9089	0.8348	0.9368	0.8903	0.9130