Table 4 Results of unenhanced model and triphasic model predictive ability for distinguishing between iMAD and LPA
Model | AUC (95% CI) | Sensitivity (%) | Specificity (%) | Accuracy (%) | Cut-off | |
|---|---|---|---|---|---|---|
Clinical model | Training cohort | 0.764 (0.650–0.857) | 74.3 | 70.3 | 70.8 |  >  − 0.226 |
Testing cohort | 0.731 (0.557–0.864) | 57.1 | 86.4 | 72.2 | ||
External validation cohort | 0.680 (0.514–0.818) | 75.0 | 60.0 | 67.5 | ||
Unenhanced radiomics model | Training cohort | 0.916 (0.826–0.968) | 88.6 | 91.9 | 90.3 |  > 0.513 |
Testing cohort | 0.860 (0.704–0.953) | 78.6 | 72.7 | 72.2 | ||
External validation cohort | 0.838 (0.687–0.935) | 75.0 | 80.0 | 77.5 | ||
Radiomics nomogram | Training cohort | 0.951 (0.873–0.988) | 85.7 | 97.3 | 91.7 |  > 0.363 |
Testing cohort | 0.938 (0.805–0.980) | 78.6 | 95.5 | 88.5 | ||
External validation cohort | 0.893 (0.754–0.968) | 80.0 | 95.0 | 87.5 | ||