Table 1 Classification results of the multi-layer perceptron method.

From: Tumor cell identification and classification in esophageal adenocarcinoma specimens by hyperspectral imaging

Classifier

EAC and tumor stroma [%]

Squamous epithelium [%]

Background [%]

Used dataset

Sensitivity

77 ± 24

54 ± 25

90 ± 45

All 95 patients

Specificity

84 ± 46

81 ± 22

97 ± 8

Accuracy

78 ± 22

80 ± 20

97 ± 8

F1-score

82 ± 20

53 ± 23

75 ± 41

ROC AUC score

88 ± 46

71 ± 28

99 ± 49

Sensitivity

80 ± 25

49 ± 33

88 ± 45

Without neodjuvant chemotherapy

Specificity

74 ± 48

81 ± 23

97 ± 4

Accuracy

78 ± 22

77 ± 22

98 ± 4

F1-score

83 ± 21

42 ± 30

76 ± 40

ROC AUC score

90 ± 46

77 ± 39

99 ± 48

Sensitivity

86 ± 22

65 ± 29

71 ± 33

With neodjuvant chemotherapy

Specificity

91 ± 26

87 ± 21

98 ± 4

Accuracy

86 ± 21

85 ± 21

98 ± 4

F1-score

89 ± 18

58 ± 37

70 ± 33

ROC AUC score

94 ± 14

72 ± 27

100 ± 0

Sensitivity

60 ± 31

32 ± 20

82 ± 43

With synthetic RGB

Specificity

76 ± 44

69 ± 23

88 ± 19

Accuracy

60 ± 29

65 ± 24

88 ± 20

F1-score

66 ± 28

30 ± 18

59 ± 35

ROC AUC score

83 ± 44

47 ± 25

99 ± 48

Back to article page