Skip to main content

Table 3 Quantitative evaluation results by different MR manufacturers.

From: Automated segmentation of liver segment on portal venous phase MR images using a 3D convolutional neural network

Metrics

S1

S2

S3

S4

S5

S6

S7

S8

AVG

GE

 DSC (%)

93.33

92.01

90.93

92.43

89.02

87.36

88.97

88.42

90.31

 MSD (mm)

1.18

3.33

3.01

2.38

3.71

4.62

4.11

3.93

3.28

 HD (mm)

1.34

3.86

3.33

2.79

4.02

4.76

4.32

4.22

3.58

 RV

1.02

1.04

1.02

0.98

1.11

0.93

0.95

0.95

1.00

Siemens

 DSC (%)

93.25

92.11

91.20

92.24

88.33

87.50

89.06

88.41

90.26

 MSD (mm)

1.17

3.29

2.97

2.38

4.17

4.53

4.10

3.97

3.32

 HD (mm)

1.35

3.81

3.20

2.87

4.28

4.69

4.42

4.22

3.61

 RV

1.05

1.07

0.97

0.92

1.14

0.90

0.90

1.05

1.00

Philips

 DSC (%)

92.28

93.35

91.26

91.06

83.79

88.28

90.83

86.47

89.67

 MSD (mm)

1.38

3.24

2.86

2.75

6.35

4.08

3.36

4.51

3.61

 HD (mm)

1.64

3.66

2.96

3.33

5.68

4.26

3.51

4.85

3.73

 RV

1.36

1.13

0.69

0.73

1.51

0.71

0.75

1.72

1.06

  1. S segment, AVG average, DSC dice similarity coefficient, HD hausdorff distance, MSD mean surface distance, RV volume ratio.