Ponto-Medullary Junction not detected

SCT
1

Hi Community
I’m trying to detect PMJ in a T2 image from the spine generic dataset, however i had this message
my command
sct_detect_pmj -i C:\Users\MSI\Desktop\sci\segsaye\sub-milan07_T2w_sc_seg.nii.gz -c t2 -qc ctrl
I’m working in Windows 10

err
err1578×600 27.6 KB

err2
err3
err31461×981 26.7 KB

2

Hi,

I notice you input the segmentation instead of the actual image (the file after flag -i has the suffix _seg, so I assume it is a segmentation).

From the help:

sct_detect_pmj 

--
Spinal Cord Toolbox (git-jca/4298-flow-d396ad3a9090900cc62d4b6eb4ccbc6d63d4a6eb)

sct_detect_pmj 
--

usage: sct_detect_pmj -i <file> -c {t1,t2} [-h] [-s <file>] [-ofolder <folder>] [-o <file>]
                      [-qc <str>] [-qc-dataset <str>] [-qc-subject <str>] [-r {0,1}] [-v <int>]

Detection of the Ponto-Medullary Junction (PMJ). This method is based on a machine-learning
algorithm published in (Gros et al. 2018, Medical Image Analysis,
https://doi.org/10.1016/j.media.2017.12.001). Two models are available: one for T1w-like and another
for T2w-like images. If the PMJ is detected from the input image, a NIfTI mask is output
("*_pmj.nii.gz") with one voxel (value=50) located at the predicted PMJ position. If the PMJ is not
detected, nothing is output.

MANDATORY ARGUMENTS:
  -i <file>          Input image. Example: t2.nii.gz
  -c {t1,t2}         Type of image contrast, if your contrast is not in the available options (t1,
                     t2), use t1 (cord bright/ CSF dark) or t2 (cord dark / CSF bright)

Cheers,
Julien

3

Yes, I see. Thank you for the remark, but I still have the same output
sct_detect_pmj -i C:\Users\MSI\Desktop\sci\segsaye\sub-milan07_T2w.nii.gz -c t2 -qc ctrl

err4
err41527×649 27.4 KB

4

Can you send me the image privately so I can try to reproduce this issue and understand its cause?

5

yes sure, i sent it to your email

6

Hi @Tayssir_Boushila,

I can reproduce your error. Unfortunately this is due to a sub-performance of the PMJ detection method-- it sometimes fails. I tried it on milan06 and it worked fine.

However I agree that the output message is not super informative, and this is something we can improve on our end @joshuacwnewton

cheers

7

Thanks for the ping! I’ll open up an issue on GitHub to follow up with a more intuitive error message. :slight_smile:

Kind regards,
Joshua

1 Like