Table 3.
Performance of the radiomic model and integrated model for predicting large-volume LNM in the training and validation cohorts.
| AUROC (95% CI) | AUPR (95% CI) | ACC (95% CI) | SEN (95% CI) | SPE (95% CI) | PPV (95% CI) | NPV (95% CI) | |
|---|---|---|---|---|---|---|---|
| Integrated model | 0.910 (0.853–0.967)∗ | 0.463 (0.434–0.492) | 93 (88–96) | 68 (43–87) | 96 (91–98) | 65 (41–85) | 96 (92–99) |
| Radiomic model | 0.890 (0.837–0.942)∗ | 0.348 (0.326–0.370) | 86 (80–91) | 74 (49–91) | 87 (81–92) | 40 (24–58) | 97 (92–99) |
| Clinical model | 0.714 (0.590–0.838) | 0.255 (0.210–0.300) | 77 (70–83) | 47 (24–71) | 80 (74–86) | 22 (11–38) | 93 (87–97) |
| Integrated model | 0.883 (0.744–1)∗ | 0.494 (0.410–0.578) | 93 (88–97) | 64 (31–89) | 96 (91–99) | 64 (31–89) | 96 (91–99) |
| Radiomic model | 0.856 (0.753–0.958)∗ | 0.381 (0.333–0.428) | 83 (75–89) | 73 (39–94) | 84 (76–90) | 31 (14–52) | 97 (91–99) |
| Clinical model | 0.702 (0.529–0.876) | 0.226 (0.171–0.281) | 80 (71–86) | 45 (17–77) | 83 (75–89) | 21 (7–42) | 94 (87–98) |
∗ Significantly different (Delong test) p < 0.05 from the clinical model. AUPR, area under the precision-recall curve; AUROC, area under the receiver operator characteristic curve; ACC, accuracy; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value; SEN, sensitivity; SPE, specificity.