Intervertebral foramen to caudal and rostral rootlets to update spinal levels: limitations

[sandrinebedard]

Description

We want to update the spinal levels in teh PAM50 using the measures from the Mendez 2021 paper. We discussed the potential limitations in #10 of placing direclty the segment length using linear scaling. Instead, we propose to use a surrogate measure for each level to find the center of each segment.

The following distance were provided from the Mendez 2021 paper:

• Intervertebral foramen -> caudal rootlet at dorsal entry
• Intervertebral foramen -> rostral rootlet at dorsal entry

The intervertevral foramen was measured the following way:

This parameter was obtained by measuring the distance between the most caudal point of each intervertebral foramen to the rostral rootlet of its corresponding dorsal root at the point of dorsal column entry using a slide caliper (Mendez, 2021)

Identifying the intervertebral foramen in the PAM50

Images from Mendez 2021

MRI Labels

How to label caudal point of the foramen (awaiting validation by a specialist):

• Get the level of the first frame (rostral) with spine segmentation (in the foramen region) for each spine level , take the most lateral point and go up 1 slice. (so the frame over the frame on the image)
  
  File with labeled foramen

Distances in the PAM50

Without standard deviation
With standard deviation

1. Compute the distance from the spinal cord centerline shifted to adjust in xy plane using the half width across dorsal
   columns

2. Find the closest distance corresponding to the measures from Mendez 2021 for rostral and caudal rootlet

RESULTS

We see that level 4, the caudal rootlet is at the same z than the intervertebral foramen.

The caudal distances are too small to be placed in the PAM50 space: the caudal entry is placed at the same slice as the intervertebral foramen.

Expected behaviour:

**Current behaviour: **

Bellow are some potential causes to explain this behaviour.

Potential limitations

1. Label in MRI differs than in ex vivo measurements

The ex vivo were measured on the dorsal rootlets, but the intervertebral foramen is more anterior. We might not be labeling the right location of the foramen in MRI images compared to the ex vivo measures.

2. Shrinking of ex vivo spinal cords (fixed)

• Fixed spinal cord experience some shrinking, resulting in smaller measures

3. Mean age is 90 y.o.

4. Height of subjects: 162 cm

5. Mean and std only on n=9

6. Straightening adds variablity and inacuracy in the PAM50 template

• Adds variability to the interverterbral foramen location in the PAM50 template.

Other

• Other approaches using fMRI (see Nawal Kinany at EPFL)

[sandrinebedard]

Next step: computing a ratio PAM50 Mendez et al. measures

Since the distances interverbral foramen - rostral/cadual rootlets were smaller from Mendex et al. paper then what we estimated in the PAM50, we tested computing a ratio for the cervical levels between the distance formen-rostral/caudal rootlets in the Mendez et al. paper vs the PAM50. Then, we applided the ratio to all levels.
Here are the steps applided:

1. We identified the rostral- caudal rootlets of levels C2 to C5 in PAM50 template.
2. We computed the distance rostral- caudal rootlet to the intervertebral foramen.
3. We computed a ratio for rostral and caudal distance in PAM50 vs Mendez paper.
4. Placed the distances of Mendez et al. paper using the ratio in the PAM50 template.

Results:

Ratio PAM50/Mendez:

Estimated distance foramen - rootlets with ratio

Here is the distance foramen - rostral/caudal rootlets in the PAM50 vs the estimated (--) ones from the Mendez et al. paper using a ratio of 1.5 for rostral distance and 1.3 for caudal.

Labels in the PAM50

Here are the results with the Mendez et al metrics with the ratio from C2 to T1. We see the spinal segment mid-point.
We can see that C8 (in orange) is estimated very close to C7.

We can see that C8 (in orange) is estimated very close to C7. We can see in the coronal view a bit the rootlets from each levels (here is a rough draft as an example).

Since we don't see the nerve rootlets in the PAM50 template for lower levels (e.g., lumbar), we have no way to validate the estimate of the spinal levels. We also have to consider that the foramen-rootlet distances for the lumbar cord in the Mendez et al. paper have more variability across subjects.

Data

spinal level labels:
foramen_label_without_std_center.nii.gz

manually labeled spinal levels in the PAM50:
label_caudal_rostral_roots.nii.gz

manually labeled intervertebral foramen:
test_foramen.nii.gz