ITH Score

Output

ith_scores.csv

Definition

Topological intra-tumor heterogeneity (ITH) score computed on the habitat label map. Each habitat label contributes connected-component statistics; the score increases when labels are fragmented into many small regions.

Formula

\[\mathrm{ITHscore} = 1 - \frac{1}{S_{\mathrm{total}}} \sum_i \frac{S_{i,\max}}{n_i}\]

\(S_{\mathrm{total}}\) = tumor voxel count; \(S_{i,\max}\) = largest connected component of label i; \(n_i\) = number of connected components of label i.

Connected components use SimpleITK ConnectedComponent (default 6-connectivity in 3D, same as non_radiomics).

Output columns

Column	Description
`ith_score`	ITH score (higher = more fragmented; not clamped to [0, 1])
`num_habitats`	Number of distinct habitat labels (> 0) in the map
`habitat_{i}_regions`	\(n_i\) (component count for label i; i = actual label value)
`habitat_{i}_largest_area`	\(S_{i,\max}\)
`habitat_{i}_area_ratio`	\(S_{i,\max} / n_i\)
`total_area`	\(S_{\mathrm{total}}\)

Notes

HABIT applies the ITHscore topology step to an existing habitat map (*_habitats.nrrd). It does not repeat the pixel-level radiomics clustering from the original ITHscore pipeline.

Implementation

habit/core/habitat_analysis/habitat_features/ith_features.py

References

Li J et al., European Radiology 2023 (PubMed · DOI).