Automatic Liver Segmentation in CT Images with Enhanced GAN and Mask Region-Based CNN Architectures

Table 3

A comparison of four liver segmentation algorithms based on six metrics.

No.	Algorithm type	Accuracy (%)	Recall (%)	Specificity (%)	Precision (%)	FOR (%)	FDR (%)
1	FCN-8s	73.2	75.3	75.0	74.4	20.1	25.6
2	U-Net	71.3	73.2	74.1	75.5	19.2	24.5
3	2D-FCN2	75.4	77.2	78.4	76.3	13.1	23.7
4	2D-FCN1	72.3	73.4	73.6	74.5	14.5	25.5
5	2D-dense-FCN	75.3	76.7	75.6	77.8	12.9	22.2
6	3D-FCN	82.3	81.7	82.9	80.1	10.1	19.9
7	H-DenseUNet	83.3	83.4	83.0	80.9	10.0	19.1
8	3D U-Net	86.3	87.7	88.3	87.3	11.5	12.7
9	IU-Net	88.3	89.3	90.2	92.9	16.3	7.1
10	GIU-Net	90.2	91.6	92.8	90.7	9.8	9.3
11	GAN Mask R-CNN	91.3	92.2	92.4	92.3	13.1	7.7