Description of Selected PET-Based Features That Were Included in Analysis
| Feature type | Feature | Description |
| Conventional | Metabolic tumor volume (MTV) | Volume of the tumor region with SUV > 2.5 |
| Maximum standardized uptake values (SUVmax) | SUVmax of the voxels within the tumor region | |
| Shape | Compactness | How closely and firmly united the tumor is |
| Histogram | Minimum standardized uptake value (SUVmin) | Minimum SUV of the voxels within the tumor region |
| Skewness | Measures the degree of asymmetry of the tumor SUV distribution | |
| Kurtosis | Measures the flatness of the tumor SUVs distribution (or how heavy-tailed or light-tailed the distribution is) relative to a normal distribution | |
| Gray level cooccurrence matrix (GLCM) | Maximum probability (MaxProb) | Determines how likely the most predominant SUV voxel pair occurs |
| Information measure of correlation (InfoCorr) | Measures the relationship between tumor voxels | |
| Normalized inverse difference (InvDiffnorm) | Measures tumor smoothness | |
| Normalized inverse difference moment (InvDiffmomnor) | Measures tumor homogeneity | |
| Inverse variance (InverseVar) | Inverse of the variability of the tumor voxel SUVs | |
| Gray level run length matrix (GLRLM) | Long run emphasis (LRE) | Measures coarse tumor textures |
| Run percentage (RunPct) | Measures homogeneity of the tumor and the distribution of runs in a specific direction | |
| Long run low gray emphasis (LRLGE) | Measures the joint distribution of low SUVs and long runs within the tumor | |
| Gray level size zone matrix (GLSZM) | Small zone emphasis (SZE) | Measures fine tumor textures |
| Large zone emphasis (LZE) | Measures coarse tumor textures | |
| Size zone variability (SZV) | Measures the similarity of the zones throughout the tumor | |
| Low gray small zone emphasis (LGSZE) | Measures the joint distribution of low SUVs and small zones within the tumor | |
| Low gray large zone emphasis (LGLZE) | Measures the joint distribution of low SUVs and large zones within the tumor | |
| Neighborhood gray tone difference matrix (NGTDM) | Busyness | Measures the change in SUV between multiple tumor voxels and their surroundings |
| Complexity | Measures the high level contents (e.g., edges and lines) |