Table 3.
Results of clinical model and combined models for pCR prediction
| Training AUC | p value | Testing AUC | p value | Accuracy | Sensitivity | Specificity | |
|---|---|---|---|---|---|---|---|
| CRD model # |
0.769 (0.722–0.816) |
0.762 (0.679–0.845) |
0.736 | 0.578 | 0.812 | ||
| CD model |
0.754 (0.705–0.802) |
< 0.001* |
0.755 (0.672–0.839) |
0.112 | 0.660 | 0.734 | 0.624 |
| CR model |
0.716 (0.665–0.767) |
< 0.001* |
0.695 (0.656–0.714) |
0.005* | 0.695 | 0.656 | 0.714 |
| RD model |
0.709 (0.658–0.761) |
< 0.001* |
0.693 (0.602–0.784) |
< 0.001* | 0.629 | 0.594 | 0.647 |
| Clinical model |
0.642 (0.586–0.697) |
< 0.001* |
0.691 (0.600-0.782) |
< 0.001* | 0.619 | 0.828 | 0.519 |
P values were obtained by likelihood ratio test
* indicate statistically significant # indicate reference model for comparison
Accuracy, sensitivity, and specificity were obtained in testing set
CRD model: Clinical-Radiomic-Dynamic model; CD model: Clinical-Dynamic model; CR model: Clinical-Radiomic model; RD model: Radiomic-Dynamic model