Table 4 Results on the Camelyon+ dataset with VIT-S-extracted features.

From: Comprehensive Benchmark Dataset for Pathological Lymph Node Metastasis in Breast Cancer Sections

Methods

Acc (%)

AUC (%)

F1 (%)

Recall (%)

Precision (%)

Kappa

VIT-S12 (ImageNet pre-trained38)

Max-MIL

76.7 ± 5.30

77.2 ± 1.90

47.7 ± 1.17

45.4 ± 1.97

45.5 ± 6.72

0.58 ± 0.10

Mean-MIL

72.5 ± 4.64

74.0 ± 4.41

42.0 ± 3.99

41.8 ± 4.31

45.4 ± 7.88

0.42 ± 0.07

ABMIL19

81.4 ± 2.07

81.6 ± 3.62

51.0 ± 3.26

49.7 ± 4.06

52.9 ± 4.71

0.65 ± 0.04

Gate-ABMIL19

78.9 ± 2.05

82.3 ± 3.41

54.0 ± 3.10

53.0 ± 2.78

52.7 ± 2.55

0.61 ± 0.05

CLAM-SB25

81.9 ± 1.52

82.3 ± 3.57

54.0 ± 3.25

53.2 ± 3.30

57.0 ± 5.80

0.64 ± 0.06

CLAM-MB25

81.7 ± 3.08

84.8 ± 4.11

57.6 ± 2.34

56.9 ± 2.07

59.0 ± 3.74

0.61 ± 0.09

DSMIL34

77.1 ± 3.38

78.9 ± 3.28

49.6 ± 2.83

49.0 ± 2.75

51.2 ± 1.73

0.57 ± 0.07

TransMIL24

79.7 ± 1.78

83.9 ± 2.10

50.8 ± 4.61

47.1 ± 5.16

46.7 ± 7.02

0.63 ± 0.03

DTFD35

80.4 ± 1.10

78.5 ± 2.96

49.3 ± 2.60

46.4 ± 3.72

47.4 ± 7.18

0.63 ± 0.03

AMD-MIL33

81.7 ± 0.72

83.0 ± 2.83

53.1 ± 3.69

51.6 ± 3.35

55.1 ± 3.60

0.65 ± 0.04

WiKG36

80.8 ± 1.34

83.2 ± 2.79

51.6 ± 2.55

50.1 ± 5.04

51.0 ± 8.07

0.62 ± 0.03

FR-MIL37

79.2 ± 3.24

83.1 ± 2.03

54.5 ± 3.56

54.5 ± 1.93

57.4 ± 4.86

0.57 ± 0.06