Table 3 The diagnostic performance of different models on the external image dataset

From: Multimodal model for the diagnosis of biliary atresia based on sonographic images and clinical parameters

Model

AUC

AUPR

Accuracy (%)

Sensitivity (%)

Specificity (%)

P value*

Gallbladder model

0.893 (0.834, 0.937)

0.926 (0.890, 0.966)

85.8 (79.4, 90.9)

84.3 (74.7, 91.4)

87.6 (77.9, 94.2)

0.290

Triangular cord model

0.821 (0.752, 0.878)

0.854 (0.830, 0.884)

76.9 (69.5, 83.3)

80.7 (70.6, 88.6)

72.6 (60.9, 82.4)

<0.001

Conventional US model

0.936 (0.886, 0.969)

0.954 (0.932, 0.986)

83.3 (76.5, 88.8)

86.7 (77.5, 93.2)

79.5 (68.4, 88.0

1.0

Gallbladder-clinical model

0.922 (0.869, 0.959)

0.945 (0.907, 0.979)

84.6 (78.7, 90.4)

79.5 (74.7, 91.4)

90.4 (76.2, 93.2)

1.0

Triangular cord-clinical model

0.889 (0.828, 0.933)

0.913 (0.877, 0.949)

76.3 (68.8, 82.7)

79.5 (69.2, 87.6)

72.6 (60.9, 82.4)

0.035

Conventional US-clinical model

0.941 (0.891, 0.972)

0.956 (0.924, 0.981)

84.6 (78.0, 89.9)

85.5 (76.1, 92.3)

83.6 (73.0, 91.2)

–

  1. 95% confidence intervals are included in parentheses
  2. US ultrasound, AUC area under receiver operating characteristic curve, AUPR area under Precision-Recall curve
  3. *The P values were from the comparison between the AUC of the conventional US-clinical model and the AUCs of the other models. P-values were adjusted for multiple comparisons using the Bonferroni correction method. Differences between various AUCs were compared using a Delong test.
Back to article page