Table 3 The performance of 5 SVM models for predicting NPC residue. AUC, area under the curve; P values are derived from the DeLong’s test of AUCs, where AUC of clinic is the reference standard for comparison.
Cohort | Model | AUC (95% CI) | Accuracy | Sensitivity | Specificity | P value |
|---|---|---|---|---|---|---|
Train | Clinic | 0.759 (0.727 ~ 0.791) | 0.728 | 0.698 | 0.828 | ref |
Clinic + FSCE-T1 Hab4 | 0.831 (0.785 ~ 0.876) | 0.808 | 0.793 | 0.837 | 0.030 | |
Clinic + CE-T1 Hab1 | 0.834 (0.783 ~ 0.884) | 0.784 | 0.771 | 0.810 | 0.075 | |
Clinic + T2 Hab2 | 0.861 (0.812 ~ 0.910) | 0.792 | 0.803 | 0.776 | 0.002 | |
Clinic + FSCE-T1 Hab4 + CE-T1 Hab1 + T2 Hab2 | 0.921 (0.867 ~ 0.975) | 0.824 | 0.821 | 0.830 | < 0.0001 | |
Test | Clinic | 0.659 (0.620 ~ 0.699) | 0.828 | 0.705 | 0.800 | ref |
Clinic + FSCE-T1 Hab4 | 0.781 (0.711 ~ 0.850) | 0.778 | 0.784 | 0.765 | 0.071 | |
Clinic + CE-T1 Hab1 | 0.714 (0.642 ~ 0.785) | 0.722 | 0.725 | 0.714 | 0.412 | |
Clinic + T2 Hab2 | 0.737 (0.664 ~ 0.810) | 0.741 | 0.788 | 0.667 | 0.120 | |
Clinic + FSCE-T1 Hab4 + CE-T1 Hab1 + T2 Hab2 | 0.811 (0.730 ~ 0.892) | 0.759 | 0.778 | 0.722 | 0.0047 |
