Table 4 Cell detection and classification results.

From: Multi scale deep learning quantifies Ki67 index in breast cancer histopathology images

Model

\(Ki67^+\)

\(Ki67^-\)

Average

Prec.(%)

Rec.(%)

F1

Prec.(%)

Rec.(%)

F1

Prec.(%)

Rec.(%)

F1(%)

U-Net19

83.71

86.86

85.11

75.64

79.33

77.15

76.18

81.12

78.05

UNet++21

85.10

86.96

86.12

73.65

80.64

78.38

77.03

80.95

78.99

PathoNet17

86.77

88.32

87.68

77.87

82.33

80.15

80.32

83.43

82.92

TransUNet29

86.98

88.47

88.10

79.25

83.75

82.04

81.57

83.89

83.75

TransAttUnet30

87.65

90.11

88.33

80.30

83.87

83.07

82.04

83.97

84.16

Kpi-Net

88.78

90.57

89.56

80.86

84.35

82.94

82.82

86.46

84.25

  1. Experimental results of precision (Prec.), recall (Rec.), and F1-score (F1) for different models on the BCData test set, where \(Ki67^+\) represents immunohistochemically positive cells and \(Ki67^-\) denotes immunohistochemically negative cells. The best results in each column are highlighted in bold.
Back to article page