Table 3 Segmentation performances for patients with residual tumor, for both architectures, all input configurations, and over the validation and test sets.

From: Segmentation of glioblastomas in early post-operative multi-modal MRI with deep neural networks

Input

Prot.

Arch.

Voxel-wise

Patient-wise

DSC-P

DSC-TP

Recall

Precision

HD95 (mm)

mAVE (ml)

A

Val

nnU-Net

46.94±24.03

49.51±21.70

94.42±6.69

62.96±6.60

37.55±32.40

0.97

AGU-Net

37.72±29.54

51.05±22.28

74.10±7.57

80.75±5.79

18.05±17.99

0.76

Test

nnU-Net

52.38±21.14

53.43±19.77

98.04

70.83

22.99±32.03

1.49

AGU-Net

38.06±27.45

46.21±22.80

82.35

84.31

21.55±33.40

1.37

B

Val

nnU-Net

52.97±22.66

55.62±19.63

95.08±5.73

66.82±6.06

29.02±31.02

0.51

AGU-Net

39.71±28.25

51.54±20.59

77.15±7.14

82.30±4.60

16.84±18.09

0.65

Test

nnU-Net

59.19±20.49

61.61±16.72

96.08

80.65

22.56±33.25

0.57

AGU-Net

43.76±27.61

53.14±20.23

82.35

87.76

21.28±35.39

0.89

C

Val

nnU-Net

52.43±22.45

54.72±19.77

95.55±7.10

63.70±6.68

35.82±35.02

0.60

AGU-Net

37.43±28.69

51.09±20.49

73.23±10.68

84.70±3.23

18.76±19.78

0.62

Test

nnU-Net

58.14±21.01

60.51±17.52

96.08

76.12

25.43±35.27

0.55

AGU-Net

42.33±27.87

53.97±18.51

78.43

95.24

20.73±36.04

0.55

D

Val

nnU-Net

52.80±22.59

55.26±19.73

95.35±6.40

66.21±5.72

25.08±28.91

0.48

AGU-Net

41.02±28.08

52.45±20.14

78.28±6.25

85.16±5.24

15.87±16.62

0.55

Test

nnU-Net

58.05±22.74

60.42±19.61

96.08

79.69

21.23±33.67

0.34

AGU-Net

40.84±28.62

52.07±20.96

78.43

93.02

19.97±26.64

0.69

E

Val

nnU-Net

53.61±22.57

55.81±19.97

95.86±6.38

63.86±6.93

29.47±31.84

0.64

AGU-Net

39.44±27.05

48.89±20.92

80.67±5.71

84.58±3.39

16.36±15.42

0.73

Test

nnU-Net

56.30±21.07

58.60±17.84

96.08

76.12

19.50±30.01

0.69

AGU-Net

41.23±25.72

47.78±20.93

86.27

89.80

21.38±35.04

1.11