Table 3:
Sensitivity, specificity, PPV, NPV, and misclassification error rate in the differentiation of tumor grades and genotypes
| Sensitivity/Specificity, PPV/NPV | cMRIa | Multimodal MRIb,c |
|---|---|---|
| Gradeb | ||
| II | 73/65%, 67/72% | 84/82%, 82/84% |
| III | 65/73%, 72/67% | 82/84%, 84/82% |
| Misclassification error rate | 31% | 17% |
| Genotypec | ||
| 1p/19q Loss | 23/70%, 33/59% | 29/79%, 46/64% |
| 1p/19q Intact | 70/23%, 59/33% | 79/29%, 64/46% |
| Misclassification error rate | 48% | 40% |
Note:—CE indicates contrast enhancement; NPV, negative predictive value; PPV, positive predictive value.
The 6 most important predictors arising from the random forest analyses are:
a
CE+Location+Necrosis+Hemorrhage+Edema+Tumor borders
b
rCBF+rCBV+CE+ADC+Hemorrhage+ NAA/Cho at short TE
c
Location+ADC+rCBF+Cho/H2O+ NAA/Cr+lipids/H2O at short TE.