Table 2 Confusion matrices and loss values of three examples, in terms of SEN and SPE, the best is Example 3 and the worst is Example 1.

From: AI outperformed every dermatologist in dermoscopic melanoma diagnosis, using an optimized deep-CNN architecture with custom mini-batch logic and loss function

Predicted class

True class

ACC

SEN

SPE

\(l_{MSE}\)

\(l_{CLF}\)

P

N

1

P’

12

2

90.0

60.0

97.5

0.10

0.2309

N’

8

78

2

P’

16

8

88.0

80.0

90.0

0.12

0.050

N’

4

72

3

P’

17

10

87.0

85.0

87.5

0.13

0.0384

N’

3

70

Back to article page