Table 2.
Model | Cutoff | Training cohort | Testing cohort | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
AUC | 95% CI | Accuracy | Specificity | Sensitivity | AUC | 95% CI | Accuracy | Specificity | Sensitivity | ||
SAD | 1.000 | 0.697 | 0.622–0.773 | 0. 659 | 0.719 | 0.461 | 0.729 | 0.619–0.839 | 0.513 | 0.750 | 0.500 |
SUVmax | 2.500 | 0.797 | 0.734–0.860 | 0.719 | 0.573 | 0.876 | 0.828 | 0.739–0.917 | 0.718 | 0.575 | 0.868 |
Rad-Score | − 0.044 | 0.849 | 0.796–0.903 | 0.768 | 0.802 | 0.730 | 0.828 | 0.782–0.919 | 0.769 | 0.725 | 0.816 |
Int-Score | 0.331 | 0.881 | 0.834–0.928 | 0.789 | 0.688 | 0.899 | 0.872 | 0.797–0.946 | 0.756 | 0.625 | 0.895 |
The values in italics indicate the cut-off values used in the diagnosis process
AUC, the area under receiver operating characteristic (ROC) curves; CI, confidence interval
*SUVmax ≥ 2.5 or SAD ≥ 1.0 cm to be malignant LNs, which are usually used as reference standards in our diagnosis process