Skip to main content

Table 3 Performance of the AI models for the testing set

From: Artificial intelligence for non-mass breast lesions detection and classification on ultrasound images: a comparative study

AI Model

AUC

(95%CI)

Cut-off

Sensitivity(%)

Specificity(%)

Accuracy (%)

p value

Detection

MobileNet_448

0.999(0.997-1.000)

0.985

99.9

99.7

99.8

Na

DenseNet121_448

0.999(0.998-1.000)

0.676

99.7

99.6

99.6

>0.05

Classification

MobileNet_448

0.837(0.810–0.863)

0.489

70.5

80.3

74.6

Na

DenseNet121_448

0.738(0.705–0.771)

0.441

68.8

68.9

68.9

<0.05

  1. Note: p value: Comparison of MobileNet_448 with DenseNet121_448