Registration approaches on 7T MP2RAGE using `disc levels` and `rootlets_t2` segmentations

[Nilser3]

Description

The idea of this issue is to explore the registration approaches of one MS patient that was acquired twice on the same day, so that both registered images should be very similar in the PAM50 space.

Materials

• One patien
• 02 sessions (adquiered same day)
• Contrast: UNIT1 and T1map at 7T
• Version: SCTv7.0

Segmentations

1. spinalcord : r20250402 -> to segment on T1map
2. totalspineseg : r20250224 -> to obtain cervical levels on T1map
3. rootlets_t2 : r20250318 -> to segment on UNIT1 (because this model was trained on 7T UNI images)
4. MS lesion GT: 4 lesions on C1 to C7, after revision by raters. Manually corrected after automatic segmentation of lesion_ms_mp2rage r20240610 model.

Here is the QC for spinalcord, disc levels and MS lesion GT segmentations.

[Nilser3]

Issues with rootlets_t2 segmentations

Here are the rootlets where the rootlet_C6 is mixed with rootlet_C7 at the same level.

Session 02:

• Maximum Intensity Projection MIP visualization:
  

• 3D visualization:
  

[jcohenadad]

I'm sure @sandrinebedard has relevant insights

[sandrinebedard]

Issues with rootlets_t2 segmentations

Here are the rootlets where the rootlet_C6 is mixed with rootlet_C7 at the same level.

Session 02:

• Maximum Intensity Projection MIP visualization:
  
• 3D visualization:
  

I think I would just erase the overlapping levels

So your goal is to register both sessions to the PAM50 space?

You could also place a label at the center of the level and erase the entire rootlet level if we really don't see the rootlets, that would also work with the sct_register_to_template -lrootlet

[Nilser3]

Thanks @sandrinebedard

I think I would just erase the overlapping levels

Agree, done in the overlapped rootlets.

So your goal is to register both sessions to the PAM50 space?

Exact, ses-01 and ses-02 to PAM50 space

You could also place a label at the center of the level and erase the entire rootlet level if we really don't see the rootlets, that would also work with the sct_register_to_template -lrootlet

Yes, aberrant rootlets was replaced by midpoint levels:

Session 02: Maximum Intensity Projection MIP visualization:

[Nilser3]

Hi @sandrinebedard

I have a question about place a label at the center of the level, do you mean place in the center of vertebral levels or spinal levels?

because for example PAM50_rootlets_midpoint labels are different to center of vertebral levels:

Thanks you!

[sandrinebedard]

I mean the spinal level! So the center of the rootlets entry zone