Table 2.
List of significant multiparametric MR radiomic features to differentiate between pseudoprogression and early tumor progression using LASSO logistic regression
| Order | Wavelet-Transformation | Imaging Parameter | Radiomic Feature | Feature Type |
|---|---|---|---|---|
| 1 | Original | CE-T1WI | Covered image intensity range | First-order |
| 2 | LLL | CE-T1WI | Correlation (SD) | GLCM (2) |
| 3 | Original | FLAIR | SD (standard deviation) | First-order |
| 4 | LLL | ADC | Long run low gray-level emphasis (mean) | GLRLM |
| 5 | LLH | ADC | Correlation (mean) | GLCM (1) |
| 6 | HHL | CBV | Inverse difference moment normalized (mean) | GLCM (3) |
| 7 | HHL | CBV | Inverse difference normalized (mean) | GLCM (3) |
| 8 | Original | CBV | Sum of intensities | First-order |
| 9 | LLL | CBV | Haralick correlation (mean) | GLCM (2) |
| 10 | LLH | CBV | Haralick correlation (mean) | GLCM (2) |
| 11 | LHL | CBV | Difference average (mean) | GLCM (2) |
| 12 | LHL | CBV | Difference average (mean) | GLCM (2) |
Note: Numbers in parentheses represent 2 or 3 consecutive voxels in the texture analysis.
Abbreviations: H = high-pass filter; L = low-pass filter.