Table 4 Comparison of classification performance by hot spot size in “CEUS”. All units except AUROC are in percent. Values in parentheses are 95 % confidence interval
Hot spot size (mm) | 2 | 3 | 4 | 5 | Entire tumor |
|---|---|---|---|---|---|
Number of selected features | 5 | 4 | 4 | 4 | 16 |
AUROC | 0.862 (0.792 - 0.924) | 0.855 (0.780 - 0.922) | 0.847 (0.776 - 0.912) | 0.851 (0.776 - 0.919) | 0.856 (0.778 - 0.921) |
ACC | 0.788 (0.717 - 0.850) | 0.795 (0.733 - 0.850) | 0.790 (0.733 - 0.850) | 0.796 (0.733 - 0.850) | 0.797 (0.733 - 0.850) |
Precision (PPV) | 0.825 (0.754 - 0.902) | 0.832 (0.759 - 0.895) | 0.824 (0.750 - 0.889) | 0.831 (0.760 - 0.889) | 0.825 (0.759 - 0.889) |
Recall (Sensitivity) | 0.898 (0.814 - 1.000) | 0.897 (0.791 - 1.000) | 0.903 (0.814 - 1.000) | 0.902 (0.814 - 1.000) | 0.913 (0.814 - 1.000) |
F1 | 0.858 (0.817 - 0.899) | 0.862 (0.817 - 0.901) | 0.860 (0.818 - 0.897) | 0.863 (0.815 - 0.903) | 0.865 (0.824 - 0.905) |
Specificity | 0.510 (0.176 - 0.765) | 0.535 (0.235 - 0.765) | 0.502 (0.176 - 0.765) | 0.527 (0.235 - 0.765) | 0.504 (0.235 - 0.706) |
NPV | 0.680 (0.500 - 1.000) | 0.692 (0.529 - 1.000) | 0.693 (0.526 - 1.000) | 0.703 (0.529 - 1.000) | 0.712 (0.545 - 1.000) |