Explore/add new compression metrics

[valosekj]

Originally posted as part of spinalcordtoolbox/spinalcordtoolbox#3733 (comment) -- moving here because the original SCT/#3733 issue eventually focused mainly on normalization to PAM50 and was closed by successfully merging SCT/#4003.

The project aims to add new relevant metrics to assess cervical spinal cord compression for degenerative cervical myelopathy.

Possible ref for triangular compression: 10.1093/brain/118.1.263

Other possible metrics

CSA for R-L

Compute CSA for right and left spinal cord by seperating from mid-point of R-L diameter.

TODO: Check in scikit-image how to extract the mid point (we use scikit-image to get the axes for AP/RL diam, so it’s a matter of fetching more outputs from that library)

Triangular compression

Compression can also have a triangular shape. However, metrics like symmetry/solidity would not be able to detect this compression.

Proposition:

1. With the spinal cord segmentation, scale compressed spinal cord to PAM50 using CSA.
2. Register to PAM50 template (center of mass or centermassrot)
3. Compute similarity metric (Dice, Hausdorff )

TODO: Make sure there isn’t a metric that can already assess this triangulation (another that is currently not implemented in process_seg)