Table 2.
Summary of best multivariate models and results from model validation.
| FET PET | CE-MRI | Combined | ||
|---|---|---|---|---|
| Included features | PET_Volume | T1_stdValue | T1_LZE | |
| PET_GLNUr | T1_Volume | T1_GLNUz | ||
| PET_RLNU | T1_Compacity | T1_ZLNU | ||
| PET_LZHGE | T1_RLNU | T1_DWT3_GLNUz | ||
| PET_GLNUz | T1_LoG_ZLNU | PET_SRE | ||
| Accuracy | 83% | 81% | 89% | |
| Sensitivity | 88% | 67% | 85% | |
| Specificity | 75% | 90% | 96% | |
| AUC | 0.91 | 0.85 | 0.96 | |
| Model validation | ||||
| LOOCV | Accuracy | 72% | 71% | 83% |
| Sensitivity | 77% | 81% | 85% | |
| Specificity | 65% | 57% | 80% | |
| AUC | 0.75 | 0.74 | 0.86 | |
| 5-fold CV | Accuracy | 74% | 77% | 80% |
| Sensitivity | 81% | 84% | 85% | |
| Specificity | 65% | 67% | 75% | |
| AUC | 0.76 | 0.75 | 0.85 | |
| 10-fold CV | Accuracy | 76% | 74% | 83% |
| Sensitivity | 85% | 81% | 81% | |
| Specificity | 65% | 62% | 85% | |
| AUC | 0.79 | 0.77 | 0.84 | |
AUC: Area under the receiver-operating characteristics curve; CI: Confidence interval; CV: Cross-validation; DWT3: Discrete 3-dimensional wavelet transformation; GLNUr: Grey-level non-uniformity for run; GLNUz: Grey-level non-uniformity for zone; LoG: Laplacian-of-Gaussian filter; LOOCV: Leave-one-out cross-validation; LZE: Long-zone emphasis; LZHGE: Long-zone high grey-level emphasis; RLNU: Run length non-uniformity; ZLNU: Zone length non-uniformity.