Table 2.
Histogram descriptors | Definitiona |
---|---|
F1. Average region gray value [HU] | μ (Gray value in V) |
F2. Region contrast [HU] | F1–μ (Gray value outside of V) |
F3. Region gray value variation [HU] | σ (Gray value in V) |
F4. Margin gray value variation [HU] | σ (Gray value in M) |
Shape descriptors | |
F5. Irregularity | 2.2 × V1/3/M1/2 |
F6. Compactness | % of volume of V included in SP |
F7. Ellipsoid axes min‐to‐max ratio | Min‐to‐max ratio of semi‐axes of the ellipsoid fitted to V |
F8. Margin distance variation [mm] | σ (distances from the center of V to the margin of V) |
F9. Relative margin distance variation | F8/μ (distances from the center of V to the margin of V) |
F10. Average gradient direction | μ (gradient direction of each voxel in M) |
F11. Margin volume [mm3] | Σ (voxels in M) |
Margin descriptors | |
F12. Average radial gradient [HU] | μ (radial gradient of each voxel in M) |
F13. Radial gradient index (RGI) | F12/μ (magnitude of image gradient of each voxel in M) |
F14. Margin strength 1 | μ (magnitude of image gradient of each voxel in M)/F2 |
F15. Margin strength 2 | σ (magnitude of image gradient of each voxel in M)/F2 |
F16. Radial gradient variation | σ (radial gradient of each voxel in M) |
Texture descriptors | |
F17. GLCM|Energy | 3D version of 2D gray‐level co‐occurrence | Energy |
F18. GLCM|Contrast | 3D version of 2D gray‐level co‐occurrence | Contrast |
F19. GLCM|Correlation | 3D version of 2D gray‐level co‐occurrence | Correlation |
F20. GLCM|Homogeneity | 3D version of 2D gray‐level co‐occurrence | Homogeneity |
Surface curvature descriptors | |
F21. Total curvature | μ (|p1| + |p2| over S)/σ (|p1| + |p2| over S) |
F22. Mean curvature | μ (0.5(p1 + p2) over S)/σ (0.5(p1 + p2) over S) |
F23. Gaussian curvature | μ (p1 × p2 over S)/σ (p1 × p2 over S) |
V refers to the segmented lesion volume. M refers to the margin of the lesion volume. SP refers to the minimum sphere including V. S refers to the surface of V. p1 and p2 refer to the first and second principal component of S. μ and σ indicate mean and standard deviation.