Table 14 Performance evaluation of the proposed MCAM model in three randomized experiments on testing datasets. The best-achieved values are in bold, while the second-highest values are underlined. The top values are individually highlighted in bold and underlined for normal and abnormal categories. [Values in %].

From: An interpretable framework for gastric cancer classification using multi-channel attention mechanisms and transfer learning approach on histopathology images

Channel

Class

1st Experiment

2nd Experiment

3rd Experiment

Avg. Acc

MGIC

SIC

MSIC

 

Sens

Spec

F1

Sens

Spec

F1

Sens

Spec

F1

 

SE

SimAM

ECA

N

98.96

98.14

98.36

98.24

98.57

98.69

98.65

98.67

98.78

 

A

98.52

98.27

98.11

98.67

98.42

98.57

98.35

98.24

98.77

98.84 Âħ 0.36

ECA

CBAM

SE

N

98.12

98.51

98.35

98.34

98.62

98.47

98.62

98.15

98.24

 

A

98.32

98.27

98.47

98.14

98.13

98.17

98.25

98.47

98.31

98.51 Âħ 0.11

SRM

CBAM

SRM

N

99.21

99.67

99.78

98.67

98.25

98.32

98.74

98.54

98.54

 

A

99.42

99.58

99.39

98.99

98.87

98.92

98.75

98.68

98.76

98.28 Âħ 0.31

SRM

CBAM

SE

N

99.21

99.67

99.78

98.67

98.25

98.32

98.62

98.54

98.69

 

A

99.08

99.27

99.67

98.80

98.32

98.58

98.79

98.99

98.90

98.10 Âħ 0.27

ECA

CBAM

SRM

N

99.42

99.58

99.39

98.87

98.72

99.02

98.75

98.68

98.76

 

A

99.75

99.28

99.63

98.80

98.82

98.58

98.79

98.99

98.90

98.39 Âħ 0.20