Table 2 Side-by-side blinded comparison between histologically stained H&E and virtually generated H&E stain.

From: Dual contrastive learning based image-to-image translation of unstained skin tissue into virtually stained H&E images

Feature

Color

Resolution

Sharpness

Contrast

Brightness

Grader

G1

G2

G3

G1

G2

G3

G1

G2

G3

G1

G2

G3

G1

G2

G3

Agree

169

178

128

196

189

151

208

190

151

221

187

144

232

218

148

N %

71.6%

75.4%

54.2%

83.1%

80.1%

64%

88.1%

80.5%

64%

93.6%

79.2%

61%

98.3%

92.4%

62.7%

N.A. D.A

1

13

42

1

8

21

1

6

16

2

6

29

1

3

23

N %

0.4%

5.5%

17.8%

0.4%

3.4%

8.9%

0.4%

2.5%

6.8%

0.8%

2.5%

12.3%

0.4%

1.3%

9.7%

Disagree

66

45

66

39

39

64

27

40

69

13

43

63

3

15

65

N %

28%

19.1%

28%

16.5%

16.5%

27.1%

11.4%

16.9%

29.2%

5.5%

18.2%

26.7%

1.3%

6.4%

27.5%

All graders (3/3 agree)

104

125

133

131

145

N %

44.1%

53%

56.4%

55.5%

61.4%

Feature

Uniform illumination

Artifacts

Melanocytes

Keratinocytes

Inflammatory cells

Grader

G1

G2

G3

G1

G2

G3

G1

G2

G3

G1

G2

G3

G1

G2

G3

Agree

233

219

142

220

180

201

225

26

145

226

113

153

224

35

148

N %

98.7%

92.8%

60.2%

93.2%

76.3%

85.2%

95.3%

11%

61.4%

95.8%

47.9%

64.8%

94.9%

14.8%

62.7%

N.A. D.A

1

5

28

2

11

9

7

169

7

6

70

1

8

164

6

N %

0.4%

2.1%

11.9%

0.8%

4.7%

3.8%

3%

71.6%

3%

2.5%

29.7%

0.4%

3.4%

69.5%

2.5%

Disagree

2

12

66

14

45

26

4

41

84

4

53

82

4

37

82

N %

0.8%

5.1%

28%

5.9%

19.1%

11%

1.7%

17.4%

35.6%

1.7%

22.5%

34.7%

1.7%

15.7%

34.7%

All graders (3/3 agree)

137

154

20

94

29

N %

58.1%

65.3%

8.5%

39.8%

12.3%

Back to article page