Spine 2D Localization

Summary

Detects vertebrae and other key structures in X-Ray (2D) images using learned models, and can construct an ASCSet and SpineData from detections and optional segmentations—use for analyzing single-frame or multi-frame X-Ray images to obtain landmarks.

See also: SpineData.

Compatible data

  • Requires: SharedImageSet (single image or multi-frame set).

  • Implicit assumptions: sufficient image quality/contrast for the selected model set; correct frame orientation.

Output

  • Detections (keypoints), optional segmentations for sacrum/femurs/clavicles.

  • ASCSet and SpineData assembled from detected elements.

Key parameters

  • splitVertebrae (bool): If false, keypoints are added globally; if true, associated per-vertebra.

  • modelSet (string), modelSetVersion (string): Select the model set combination.

Workflow

  1. Load/select an X-Ray image or sequence.

  2. Choose a model set and version; adjust split behavior.

  3. Run localization; review detections and constructed SpineData/ASCSet.

Notes

  • Keypoint association handles some missing/mirrored keypoints and still forms vertebra entries in typical patterns.

  • Multi-frame inputs are supported; ensure correct frame selection/iteration in the UI.