Table 3.
Sensitivity, specificity and AUC obtained from previously trained models inferring on the external validation cohort with corresponding data correction
| Dataset | Classifier | Sensitivity (%) | Specificity (%) | AUC |
|---|---|---|---|---|
| Images | ResNet50 | 92 | 24 | 0.64 |
| Radiomics | LR | 1 | 0 | 0.50 |
| SVM | 70 | 32 | 0.47 | |
| RF | 64 | 68 | 0.71 | |
| GB | 67 | 64 | 0.70 | |
| Radiomics with batch correction | LR | 1 | 0.07 | 0.54 |
| SVM | 47 | 57 | 0.48 | |
| RF | 53 | 86 | 0.75 | |
| GB | 97 | 61 | 0.80 |
AUC area under the curve, CNN convolutional neural networks, FE feature extraction, XGB Xgboost, GB gradient boosting, LR logistic regression, RF random forest, SVM support vector machine