Table 1.
The performance of radiomic models for predicting IDH mutation, 1p/19q codeletion, and TERT promoter mutation status.
| Genotype | WHO grade | Training cohort | Validation cohort | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AUC | ACC | SEN | SPE | PPV | NPV | AUC | ACC | SEN | SPE | PPV | NPV | ||
| IDH mutation | II–IV | 0.912 (0.880–0.944) | 0.870 (0.832–0.903) | 0.895 (0.842–0.942) | 0.850 (0.798–0.898) | 0.825 (0.764–0.879) | 0.911 (0.868–0.952) | 0.884 (0.830–0.934) | 0.824 (0.765–0.882) | 0.750 (0.656–0.841) | 0.898 (0.831–0.962) | 0.882 (0.804–0.956) | 0.779 (0.693–0.857) |
| II/III | 0.867 (0.814–0.914) | 0.842 (0.794–0.885) | 0.929 (0.885–0.969) | 0.712 (0.613–0.803) | 0.829 (0.767–0.888) | 0.870 (0.792–0.941) | 0.895 (0.831–0.948) | 0.827 (0.765–0.889) | 0.791 (0.683–0.886) | 0.868 (0.775–0.951) | 0.872 (0.784–0.952) | 0.786 (0.674–0.884) | |
| IV | 0.823 (0.743–0.894) | 0.932 (0.836–0.918) | 0.333 (0.182–0.500) | 0.985 (0.952–1.000) | 0.667 (0.500–1.000) | 0.943 (0.841–0.928) | 0.860 (0.778–0.929) | 0.816 (0.658–0.816) | 0.647 (0.556–0.741) | 0.952 (0.833–1.000) | 0.917 (0.889–1.000) | 0.769 (0.400–0.667) | |
| 1p/19q codeletiona | II–IV | 0.794 (0.732–0.842) | 0.706 (0.655–0.752) | 0.918 (0.855–0.970) | 0.633 (0.570–0.691) | 0.463 (0.389–0.537) | 0.957 (0.920–0.984) | 0.815 (0.751–0.878) | 0.723 (0.655–0.790) | 0.794 (0.686–0.906) | 0.694 (0.608–0.772) | 0.509 (0.409–0.625) | 0.894 (0.828–0.955) |
| II/III | 0.703 (0.630–0.772) | 0.612 (0.552–0.673) | 0.918 (0.859–0.970) | 0.433 (0.354–0.514) | 0.487 (0.413–0.564) | 0.900 (0.824–0.964) | 0.794 (0.713–0.865) | 0.679 (0.593–0.753) | 0.769 (0.632–0.893) | 0.636 (0.536–0.736) | 0.500 (0.375–0.622) | 0.854 (0.763–0.932) | |
| TERT promoter mutation | II–IV | 0.686 (0.627–0.741) | 0.676 (0.626–0.723) | 0.836 (0.777–0.887) | 0.509 (0.433–0.586) | 0.641 (0.578–0.706) | 0.747 (0.663–0.825) | 0.669 (0.580–0.748) | 0.655 (0.588–0.723) | 0.841 (0.766–0.915) | 0.446 (0.339–0.554) | 0.631 (0.549–0.718) | 0.714 (0.585–0.833) |
| II/III | 0.697 (0.628–0.763) | 0.673 (0.612–0.733) | 0.763 (0.684–0.837) | 0.588 (0.500–0.673) | 0.635 (0.554–0.713) | 0.725 (0.635–0.809) | 0.602 (0.497–0.710) | 0.654 (0.568–0.741) | 0.800 (0.698–0.896) | 0.472 (0.333–0.611) | 0.655 (0.542–0.750) | 0.654 (0.500–0.810) | |
| IV | 0.598 (0.470–0.722) | 0.685 (0.589–0.767) | 0.976 (0.930–1.000) | 0.290 (0.150–0.429) | 0.651 (0.551–0.746) | 0.900 (0.714–1.000) | 0.811 (0.689–0.918) | 0.658 (0.526–0.789) | 0.944 (0.842–1.000) | 0.400 (0.217–0.588) | 0.586 (0.433–0.733) | 0.889 (0.667–1.000) | |
Data within parentheses represent 95% confidence interval (CI).
AUC area under the curve, SEN sensitivity, SPE specificity, ACC accuracy, PPV positive predictive value, NPV negative predictive value.
aThe performance metrics of cT1WI-based radiomic model in prediction of 1p/19q codeletion status in grade IV gliomas could not be calculated because no patients had 1p/19q codeletion.