Table 2.
Nonredundant and Informative Features from Each Feature Extractor
| Feature Name | Univariate Analysis (AUC) |
|---|---|
| IBEX | |
| 135-1Correlation | 0.74 |
| LocalRangeStd | 0.72 |
| 1GaussAmplitude | 0.66 |
| VoxelSize | 0.62 |
| -333-4ClusterShade | 0.62 |
| Pyradiomics | |
| log-σ-2-0-mm-3D_firstorder_Minimum | 0.72 |
| log- σ2-0-mm-3D_glszm_SizeZoneNonUniformityNormalized | 0.70 |
| log-σ2-0-mm-3D_glcm_InverseVariance | 0.68 |
| wavelet-LHL_firstorder_Skewness | 0.67 |
| wavelet-LHH_firstorder_Skewness | 0.65 |
| wavelet-HHH_glszm_SmallAreaEmphasis | 0.65 |
| CIFE | |
| DWF_Z_H | 0.72 |
| Intensity_Minimum | 0.71 |
| Gabor_Max_Z | 0.68 |
| Intensity_Skewness | 0.65 |
Optimal features are listed for each individual extractor. These features are then used to build prediction models in the multivariate analysis. Each feature has a correlation coefficient <0.2 and an AUC > 0.6.