Skip to main content

Table 7 Different quantitative metrics for models trained on bi-cubically down-sampled data

From: Fully‑automated deep‑learning segmentation of pediatric cardiovascular magnetic resonance of patients with complex congenital heart diseases

 

Dice (%)

Jaccard (%)

Rel. volume difference (%)

\({\text{R}}_{{{\text{vol}}}}^{2} (\%)\)

t-test

p-value

LVED

 FCN

88.5 (18.4)

82.6 (20.3)

6.0 (10.3)

98.5

0.0147

 FCN-SAD

91.0 (14.9)

85.8 (17.0)

4.6 (6.1)

99.2

0.2739

 U-Net

85.5 (23.0)

79.5 (24.0)

10.7 (17.7)

89.8

4.755e−3

 U-Net-SAD

87.4 (21.1)

81.7 (22.3)

8.3 (15.9)

97.7

0.443e−3

LVES

 FCN

83.1 (22.6)

75.6 (23.8)

10.1 (9.3)

95.7

0.0786

 FCN-SAD

86.8 (16.5)

79.4 (18.9)

7.9 (5.1)

97.8

0.0945

 U-Net

81.6 (22.6)

73.4 (23.8)

23.4 (39.9)

79.1

9.913e−4

 U-Net-SAD

83.9 (20.7)

76.1 (22.0)

15.6 (27.0)

93.3

3.786e−4

RVED

 FCN

80.9 (22.9)

72.6 (24.3)

9.3 (14.2)

87.7

0.0159

 FCN-SAD

84.7 (18.8)

76.8 (20.8)

6.8 (8.6)

94.9

0.0858

 U-Net

76.5 (29.5)

69.0 (29.6)

12.4 (17.9)

81.8

0.0134

 U-Net-SAD

81.8 (22.8)

73.8 (24.4)

9.7 (12.4)

91.8

0.0251

RVES

 FCN

77.2 (22.8)

67.2 (23.6)

13.6 (15.6)

90.6

0.0226

 FCN-SAD

80.6 (19.7)

70.9 (21.2)

11.0 (13.8)

92.9

0.6585

 U-Net

70.2 (30.4)

60.9 (29.4)

18.5 (19.9)

82.6

0.151e−3

 U-Net-SAD

74.8 (25.5)

64.9 (25.6)

13.8 (15.4)

88.1

1.783e−3