Non-radiomics Morphology

Output

habitat_basic_features.csv

Definition

Morphological statistics derived from the habitat label map L(\mathbf{x}) (voxel label; 0 = background). No PyRadiomics intensity features are used.

Formula

\[\begin{split}N_h &= \text{count of distinct labels } k > 0 \quad (\text{num_habitats}) \\ V_{\mathrm{tumor}} &= \sum_{\mathbf{x}:\, L(\mathbf{x}) \neq 0} 1 \\ V_k &= \sum_{\mathbf{x}:\, L(\mathbf{x}) = k} 1 \\ \text{volume_ratio}_k &= V_k / V_{\mathrm{tumor}} \\ \text{num_regions}_k &= \text{face-connected components with label } k\end{split}\]

Connected components use SimpleITK ConnectedComponent with SetFullyConnected(False) (6-connectivity in 3D).

The exported CSV uses fixed column indices 1 … K where K = n_habitats from config (or auto-detected habitats.csv), not necessarily the distinct labels present in each subject’s map. Per-subject num_habitats counts distinct labels in that map and may differ from K.

Output columns

Column	Description
num_habitats	Number of distinct habitat labels (> 0) in this subject’s map
{k}_num_regions (k = 1 … K)	Face-connected component count for label k (0 if absent)
{k}_volume_ratio (k = 1 … K)	Fraction of tumor voxels with label k (0 if absent)

Implementation

habit/core/habitat_analysis/habitat_features/basic_features.py