Table 4.
Comparison of Adenocarcinoma aggressiveness estimation results using the heterogeneity and Definiens features.
| Feature type | Feature set | Feature selector | Classifier | AUROC | Acc. (%) |
|---|---|---|---|---|---|
| Staging | NA | NA | NA | 0.67 | 65 |
| Heterogeneity | hV 3 | mRMR 1* | J48 | 0.80 | 85 |
| Definiens | all 219 features | RfF 5 | J48 | 0.71 | 77.5 |
| Combined | RIDER +hV3 | RfF 5 | RFs | 0.90 | 85 |
*Weka v.3.8.1 provides the mRMR algorithm, which defines the optimal number of features for a particular dataset in terms of redundancy and relevance. As a result, the selected number of features varies.