Table 6 Performance evaluated over the MClass-D dataset using different thresholds. In bold, three models achieve high efficiency, especially BLF that reaches SPE 93.8% with SEN 90%.

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

SEN % THRES

SPE %

SPE % THRES

SEN %

ORI

BON

BLF

SOTA

ORI

BON

BLF

SOTA

100.0

37.5

37.5

63.7

100.0

45.0

40.0

15.0

95.0

75.5

70.0

67.5

95.0

75.0

65.0

85.0

90.0

82.5

77.5

93.8

90.0

80.0

80.0

90.0

85.0

83.8

82.5

96.2

85.0

80.0

80.0

90.0

80.0

90.0

90.0

96.2

80.0

90.0

85.0

90.0

76.7

90.0

90.0

96.2

69.2

95.0

95.0

90.0

84.5

74.1

96.2

92.5

97.5

86.5

60.0

95.0

95.0

100.0

87.5