Features computation and clinical reporting - raidionics/Raidionics GitHub Wiki

1. Across all tumor types

The following features are automatically computed and reported to the user for all tumor types. Their computation is performed in the MNI reference space, after SyN registration.

1.1. Volume

  • Variables (2): The total tumor volume in original patient space and MNI space, reported in ml.
  • Description: Simple counting of all nonzero voxels, multiplied by the MRI volume spacing information.

1.2. Multifocality

  • Variables (3): Whether the tumor is multifocal or not, the total number of foci, and the largest minimum distance between two foci (in mm).
  • Description: the largest minimum distance is computed using the 95th-percentile Hausdorff distance.

1.3. Laterality

  • Variables (3): the percentage of the total tumor volume within each hemisphere, and an assessment of midline crossing expressed as true/false.
  • Description: the laterality mask used for this computation has been generated from the MNI152 cortical structures atlas.

1.4. Cortical structures

Percentage of the tumor volume overlapping each cortical structure from the MNI atlas, the Harvard-Oxford atlas, and the Schaefer atlas (version 7 and 17).

1.5. Subcortical structures

Percentage of the tumor volume overlapping each structure from the BCB atlas. If no overlap, the minimum distance to the structure is provided (in mm).

2. Glioblastoma specific

2.1. Resectability

  • Variables (2): the expected residual tumor volume (in ml) and the expected resectability index.
  • Description: they are calculated using resection probability maps (left/right hemispheres) as introduced in this article.
⚠️ **GitHub.com Fallback** ⚠️