Table 4.
Predictive model performance for Ki-67 in the internal and external test.
| Models | Features/classification | Test | Accuracy | 95% CI | AUC | 95% CI | Sensitivity | 95% CI | Specificity | 95% CI |
|---|---|---|---|---|---|---|---|---|---|---|
| Radiomics | Lasso+RandomForest | Internal test | 0.813 | 0.73 - 0.89 | 0.756 | 0.666 - 0.845 | 0.576 | 0.407 - 0.75 | 0.936 | 0.876 - 0.986 |
| External test | 0.751 | 0.669 - 0.83 | 0.677 | 0.567 - 0.778 | 0.516 | 0.322 - 0.708 | 0.837 | 0.753 - 0.917 | ||
| Lasso+LDA | Internal test | 0.813 | 0.74 - 0.88 | 0.756 | 0.669 - 0.841 | 0.576 | 0.405 - 0.741 | 0.937 | 0.873 - 0.985 | |
| External test | 0.753 | 0.66 - 0.83 | 0.679 | 0.582 - 0.779 | 0.518 | 0.344 - 0.708 | 0.839 | 0.753 -0.921 | ||
| LinearSVC+RandomForest | Internal test | 0.796 | 0.71 - 0.87 | 0.721 | 0.634 - 0.81 | 0.487 | 0.31 - 0.656 | 0.955 | 0.901 - 1 | |
| External test | 0.751 | 0.67 - 0.83 | 0.615 | 0.526 - 0.719 | 0.317 | 0.142 - 0.5 | 0.913 | 0.845 - 0.972 | ||
| LinearSVC+LDA | Internal test | 0.795 | 0.71 - 0.87 | 0.72 | 0.636 - 0.811 | 0.486 | 0.323 - 0.648 | 0.954 | 0.898 - 1 | |
| External test | 0.753 | 0.67 - 0.83 | 0.616 | 0.526 - 0.708 | 0.321 | 0.156 - 0.5 | 0.91 | 0.845 - 0.972 | ||
| ExtraTree+RandomForest | Internal test | 0.773 | 0.689 - 0.85 | 0.733 | 0.636 - 0.821 | 0.608 | 0.448 - 0.769 | 0.857 | 0.765 - 0.94 | |
| External test | 0.783 | 0.7 - 0.86 | 0.698 | 0.603 - 0.796 | 0.515 | 0.333 - 0.694 | 0.882 | 0.8 - 0.946 | ||
| ExtraTree+LDA | Internal test | 0.771 | 0.69 - 0.85 | 0.729 | 0.635 - 0.822 | 0.6 | 0.428 - 0.766 | 0.859 | 0.77 - 0.937 | |
| External test | 0.785 | 0.7 - 0.87 | 0.7 | 0.598 - 0.811 | 0.518 | 0.32 - 0.718 | 0.882 | 0.807 - 0.951 | ||
| Clinics+ Radiomics |
Lasso+RandomForest | Internal test | 0.867 | 0.81 - 0.93 | 0.82 | 0.741 - 0.897 | 0.671 | 0.514 - 0.821 | 0.969 | 0.923 - 1 |
| External test | 0.743 | 0.66 - 0.82 | 0.712 | 0.611 - 0.808 | 0.643 | 0.45 -0.809 | 0.78 | 0.684 - 0.875 | ||
| Lasso+LDA | Internal test | 0.866 | 0.79 - 0.93 | 0.817 | 0.727 - 0.896 | 0.665 | 0.499 - 0.818 | 0.968 | 0.919 - 1 | |
| External test | 0.777 | 0.69 - 0.85 | 0.67 | 0.569 - 0.771 | 0.441 | 0.25 - 0.631 | 0.899 | 0.823 - 0.96 | ||
| LinearSVC+RandomForest | Internal test | 0.846 | 0.77 - 0.91 | 0.796 | 0.705 - 0.88 | 0.639 | 0.47 - 0.8 | 0.953 | 0.893 - 1 | |
| External test | 0.732 | 0.64 - 0.82 | 0.628 | 0.534 - 0.74 | 0.403 | 0.217 - 0.613 | 0.854 | 0.774 - 0.925 | ||
| LinearSVC+LDA | Internal test | 0.847 | 0.77 - 0.91 | 0.796 | 0.705 - 0.881 | 0.639 | 0.468 - 0.794 | 0.953 | 0.892 - 1 | |
| External test | 0.731 | 0.64 - 0.81 | 0.626 | 0.516 - 0.725 | 0.398 | 0.206 - 0.577 | 0.854 | 0.767 - 0.931 | ||
| ExtraTree+RandomForest | Internal test | 0.815 | 0.74 - 0.89 | 0.757 | 0.67 - 0.843 | 0.576 | 0.407 - 0.743 | 0.938 | 0.876 - 0.985 | |
| External test | 0.773 | 0.68 - 0.85 | 0.666 | 0.562 - 0.767 | 0.435 | 0.25 - 0.629 | 0.896 | 0.823 - 0.96 | ||
| ExtraTree+LDA | Internal test | 0.815 | 0.74 - 0.89 | 0.758 | 0.663 - 0.838 | 0.578 | 0.4 - 0.736 | 0.937 | 0.873 - 0.985 | |
| External test | 0.744 | 0.66 - 0.83 | 0.713 | 0.614 - 0.82 | 0.644 | 0.444 - 0.826 | 0.781 | 0.684 - 0.873 |