Skip to main content

Table 2 Performance metrics for the DL model in the test sets

From: Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study

 

AUC (95% CI)

Sensitivity (95% CI)

Specificity (95% CI)

PPV (95% CI)

NPV (95% CI)

ACC

F1

MCC

Internal test set

0.908 (0.879–0.933)

83.23 (76.55–88.65)

83.61 (78.97–87.58)

72.83 (67.33–77.71)

90.43 (86.96–93.04)

83.48 (79.79–86.73)

0.777

0.650

External test sets

0.913 (0.881–0.939)

88.84 (83.72–92.79)

83.77 (77.76–88.70)

85.51 (81.00–89.10)

87.43 (82.48–91.13)

86.40 (82.63–89.61)

0.871

0.728

External test set A

0.908 (0.859–0.945)

88.00 (79.98–93.64)

85.57 (76.97–91.88)

86.28 (79.39–91.12)

87.37 (80.17–92.21)

86.80 (81.26–91.19)

0.871

0.736

External test set B

0.918 (0.871–0.952)

89.62 (82.19–94.71)

81.92 (72.63–89.10)

84.82 (78.34–89.62)

87.50 (79.87–92.51)

86.00 (80.41–90.49)

0.872

0.719

  1. DL deep learning, AUC area under the receiver operating characteristic curve, PPV positive predictive value, NPV negative predictive value, ACC accuracy, MCC Matthews correlation coefficient, CI confidence interval