Spinal cord morphometrics

MRI Processing & SCT
1

Hi everyone,
I am very new to Spinal Cord Toolbox and have just finished going through the tutorials, so apologies in advance if some of these questions are basic.
I am interested in calculating cervical spinal cord morphometrics in a cohort of patients with degenerative cervical myelopathy.
First question:
Our imaging is purely clinical: 1.5T, 2D MRI, not acquired according to the spine-generic protocol. As expected, advanced quantitative metrics (e.g., MTR) are not feasible, and my focus is on basic morphometric measurements (CSA, AP/RL diameters, eccentricity, etc.).
In this setting, is it reasonable to use SCT on 1.5T clinical MRI for morphometric analysis? Are there known limitations or caveats we should be particularly aware of when interpreting these measurements?

Second question:
After running the pipeline and reviewing the outputs, I noticed that some measurements at the maximum spinal cord compression (MSCC) level, particularly normalized AP diameter (e.g., PAM50 ratios), seem unexpectedly high or low.
I manually checked the images and, while segmentation and vertebral labeling appear accurate, the axial and coronal reformatted images generated by SCT appear noticeably low-resolution (almost “pixelated”). However, the original axial images in PACS do not appear nearly this degraded.
What is the underlying reason for this apparent loss of image quality in SCT visualizations?

P.S. I understand that some of the unexpected values may be due to focusing on the point of maximal compression rather than using level-based measurements, but before changing the analysis strategy I want to ensure that the measurements themselves are reliable and that I am not misunderstanding the image processing or normalization steps.

Any guidance would be greatly appreciated. Thank you for your time.

2

Hi @zegdalomer,

Welcome to the spinalcordmri.org forum, and thank you for your detailed questions. They’re not too basic at all, and we appreciate you reaching out. We would be happy to help!

Before responding fully, I’d just like to ask if you could please provide the following details:

  • The exact SCT commands used in your processing pipeline.
  • Any data or visualizations you can feasibly provide that illustrate the issues. We fully understand if this isn’t possible given the sensitive nature of the data, so even screenshots or a .zip of the generated QC report would be very helpful for diagnosing the issue. (The tutorial commands should include the -qc flag to generate QC reports by default.)

Thanks much!

Kindly,
Joshua

3

Hi @zegdalomer,

I’ll let you follow up on @joshuacwnewton’s comments, but I can already answer some of your first questions:

is it reasonable to use SCT on 1.5T clinical MRI for morphometric analysis?

Yes, definitely reasonable. 3T has higher SNR, but is also more prone to susceptibility artifacts, so for some sequences 1.5T might actually be more advantageous. That being said, it really depends on what sequence/parameters you’re using. Obviously, computing CSA on a sagittal scan with 5mm thickness is problematic as the cord is about 10mm diameter…

Are there known limitations or caveats we should be particularly aware of when interpreting these measurements?

What I mentioned above, and also various types of artifacts to watch out for (signal drop out, flow effects, Gibbs, etc.). It’s hard for me to advise without looking at your images-- as Joshua said, if you can share some QC reports that will be helpful

Cheers,
Julien

1 Like
4

Thanks, Joshua, for your reply.
Below are the commands I am currently using in my pipeline (SCT v7.2, macOS):

Reorient MRI to RPI

sct_image
-i Patient45data.nii
-setorient RPI
-o Patient45data_RPI.nii.gz

Spinal cord segmentation

sct_deepseg spinalcord
-i Patient45data_RPI.nii.gz

QC for spinal cord segmentation

sct_qc
-i Patient45data_RPI.nii.gz
-s Patient45data_RPI_seg.nii.gz
-p sct_deepseg_sc

Vertebral labeling

sct_deepseg spine
-i Patient45data_RPI.nii.gz
-qc ~/qc_singleSubj

CSA averaged across C3–C4

sct_process_segmentation
-i Patient45data_RPI_seg.nii.gz
-vert 3:4
-discfile Patient45data_RPI_totalspineseg_discs.nii.gz
-o Patient45_csa_C3C4.csv

PAM50-normalized, slice-wise shape metrics

sct_process_segmentation
-i Patient45data_RPI_seg.nii.gz
-discfile Patient45data_RPI_totalspineseg_discs.nii.gz
-perslice 1
-normalize-PAM50 1
-o Patient45_shape_PAM50.csv

I initially wanted to share a short screen recording to show the quality of the axial imaging, but I wasn’t able to upload it. Instead, I am uploading a fully de-identified NIfTI file from one representative MRI so you can directly inspect the results.
Thanks again for your time and help.

Thanks,
Hassan

Patient45data_RPI.nii.gz (4.9 MB)

5

as mentioned above Spinal cord morphometrics - #3 by jcohenadad these sagittal images are highly anisotropic: 3.3mm slice thickness-- CSA is quite unreliable at this resolution-- however you might want to try the new sct_compute_ascor or sct_compute_compression methods

6

Hi Julien,

Thank you for your helpful comments.
It is reassuring to know that using 1.5 T MRI data is acceptable. The slice thickness in our scans is approximately 3 mm.
The morphometric parameters we are interested in include cross-sectional area (CSA), anterior–posterior (AP) diameter, right–left (RL) diameter, eccentricity, solidity, and maximum spinal cord compression (MSCC), with comparisons performed between DCM patients and controls.
This project aims to expand the scope of data within our prospective myelopathy registry and to evaluate whether these SCT-derived morphometric variables have prognostic value in predicting response to surgical decompression.
Regarding your other comment, I hope the de-identified NIfTI file I shared is helpful for assessing the image quality and the robustness of the pipeline.

Thanks,
Hassan

7

Okay thanks!
I will try to learn those new commands…