SpineData from Multi-Label

Summary

Converts a labeled segmentation volume (multi-label) into a structured SpineData object that downstream Spine algorithms can use. Use this when you have a spine label map from an external segmentation pipeline and need a SpineData with vertebrae, sacrum, and ilium for registration or analysis.

See also: SpineData.

Input

  • A single or multi-label 3D label map volume.

  • Each vertebra should be its own label value (instance segmentation): one integer id per vertebra region. Non-vertebra labels (e.g., sacrum/ilium) should not be present.

Output

  • A SpineData instance containing anatomical structures reconstructed from the label map.

  • Structures typically include vertebrae (with per-vertebra geometry), sacrum, and ilium where labeled.

Workflow

  1. Load or select your labeled volume.

  2. Run “SpineData from Multi-Label”.

  3. Inspect the generated SpineData in 3D; verify vertebra counts and positions.

  4. Use the SpineData as input to downstream altorithms such as registration or analysis (e.g., Poly Rigid Registration, Vertebra Dissection).

How vertebrae are derived

  • The label map is analyzed to estimate three canonical keypoints per vertebra (body, left pedicle, right pedicle).

  • One SpineData vertebra is created per detected keypoint triplet.

  • If per-vertebra masks are provided, they are attached to the corresponding vertebra and converted to meshes automatically.

Note

This conversion does not assign anatomical type names (C/T/L numbering).