Skip to main content
. 2025 Dec 8;16:273. doi: 10.1186/s13244-025-02150-y

Table 4.

Diagnostic performance based on ADC

Malignant (n = 100) vs. Benign (n = 82)
Parameters With DLR (n = 182) Without DLR (n = 182) p-value
AUC 0.921 0.904 < 0.05
95% CI 0.885–0.957 0.863–0.945 /
Threshold ADC value (×10-3 mm2/s) 1.355 1.475 /
Sensitivity (%) 90.2 (74/82) 80.5 (66/82) < 0.001
Specificity (%) 78 (78/100) 83 (83/100) < 0.001
PPV (%) 77.1 (74/96) 79.5 (66/83) 0.002
NPV (%) 90.7 (78/86) 83.8 (83/99) 0.001
Accuracy (%) 83.5 (152/182) 81.9 (149/182) < 0.001
Youden index 0.68 0.63 /
Malignant (n = 100) vs. FNH (n = 18) + HH (n = 38)
Parameters With DLR (n = 156) Without DLR (n = 156) p-value
AUC 0.946 0.928 0.046
95% CI 0.914-0.977 0.891-0.966 /
Threshold ADC value (×10-3 mm2/s) 1.375 1.365 /
Sensitivity (%) 96.4 (54/56) 96.4 (54/56) /
Specificity (%) 79 (79/100) 72 (72/100) 0.121
PPV (%) 72.0 (54/75) 65.9 (54/82) 0.927
NPV (%) 97.5 (79/81) 97.3 (72/74) 0.249
Accuracy (%) 85.3(133/156) 80.8(126/156) 0.121
Youden index 0.75 0.68 /

Data are presented as AUC values, 95% confidence intervals, threshold values, and percentages (numerator/denominator)

ADC apparent diffusion coefficient, DLR deep learning reconstruction, FNH focal nodular hyperplasia, HH hepatic hemangioma, AUC area under the curve, CI confidence interval, PPV positive predictive values, NPV negative predictive values