Table 2.
Diagnostic performance of separate models.
| CEe | CEp | CEd | |
|---|---|---|---|
| Pre-radiomics model | |||
| No. of selected features | |||
| LASSO_ CV | 9 | 9 | 2 |
| Logistic_ CV | 3 | 5 | 1 |
| AUC (training/validation) | 0.759/0.617 | 0.827/0.694 | 0.649/0.539 |
| 95% CI of AUC | 0.647, 0.871/0.403, 0.830 | 0.723, 0.922/0.533, 0.856 | 0.527, 0.770/0.319, 0.759 |
| Sensitivity (training/validation) | 0.643/0.667 | 0.679/0.917 | 0.857/0.250 |
| Specificity (training/validation) | 0.800/0.633 | 0.857/0.500 | 0.386/0.967 |
| Accuracy (training/validation) | 0.755/0.643 | 0.806/0.619 | 0.520/0.762 |
| 1st-radiomics model | |||
| No. of selected features | |||
| LASSO_ CV | 10 | 9 | 13 |
| Logistic_ CV | 5 | 4 | 4 |
| AUC (training/validation) | 0.803/0.775 | 0.816/0.650 | 0.826/0.703 |
| 95% CI of AUC | 0.694, 0.913/0.627, 0.923 | 0.717, 0.915/0.432, 0.868 | 0.738, 0.914/0.514, 0.892 |
| Sensitivity (training/validation) | 0.756/0.667 | 0.786/0.667 | 0.821/0.417 |
| Specificity(training/validation) | 0.771/0.800 | 0.771/0.800 | 0.700/0.967 |
| Accuracy (training/validation) | 0.776/0.762 | 0.776/0.667 | 0.735/0.810 |
| Delta-radiomics model | |||
| No. of selected features | |||
| LASSO_ CV | 13 | 11 | 3 |
| Logistic_ CV | 9 | 7 | 1 |
| AUC (training/validation) | 0.917/0.842 | 0.803/0.764 | 0.708/0.697 |
| 95% CI of AUC | 0.861, 0.974/0.709, 0.974 | 0.64, 0.913/0.592, 0.936 | 0.594, 0.821/0.512, 0.883 |
| Sensitivity (training/validation) | 0.929/0.667 | 0.786/0.917 | 0.750/0.833 |
| Specificity (training/validation) | 0.829/0.900 | 0.771/0.667 | 0.629/0.700 |
| Accuracy (training/validation) | 0.857/0.833 | 0.776/0.738 | 0.663/0.738 |
Pre-radiomics features, the features from DCE-MRI before neoadjuvant chemotherapy (NAC); 1st-radiomics features, the features from DCE-MRI after the first cycle of NAC. CV, cross validation. AUC, area under receiver operating characteristic curve; CI, confidence intervals.