Ants Longitudinal CT Template-Build Stage fail

[dbrennan44]

Hi all,

ANTs Longitudinal Cortical Thickness is failing at the template-building stage with 2 scans. Oddly, this behavior does not happen each time antsCT is run (rerunning a subject can work). Here is an example output with error message:

output is attached for formatting
output.txt

I can provide the two images if it helps diagnose.

Thank you!

[cookpa]

What's your ANTs version?

[dbrennan44]

ANTs Version: v2.4.1.post7-g71c9890

from antsRegistration

[cookpa]

Might be data-dependent. I think this arises from scales estimation in the beginnings of registration

https://discourse.itk.org/t/zero-valued-spacing-is-not-supported-and-may-result-in-undefined-behavior/5382/6
#1348

Will try to reproduce with data

[dbrennan44]

here is a link

I will note, I manually truncated the images before processing because the FOV was very large and included the neck.

[cookpa]

Thanks, the images look like they should work. This works for me on Mac OS using 2.5.0.post12-g9516f51 (I ran it twice)

antsMultivariateTemplateConstruction.sh -d 3 -o template -b 0 -g 0.25 -i 4 -c 2 -j 4 -k 1 -w 1 -m 100x70x30x3 -n 1 -r 1 -s CC -t GR -y 1 -z scan_a.nii.gz scan_a.nii.gz scan_b.nii.gz

[dbrennan44]

great! My problem was happening within a call to antsLongitudinalCorticalThickness.sh. Can the templates be built outside of this script and used? Perhaps an initial template?

[cookpa]

Right, I used the template command from your original output.txt, which doesn't include the original call to the longitudinal script. Unfortunately it's not easy to use an initial template with that script.

We might debug further if you can try running the longitudinal script with -b 1 to keep temporary files. When it fails, does it leave behind a template T_template0.nii.gz? If the template is blank for some reason, the next iteration may fail with the error message you saw.

Also if you can add the original call to antsLongitudinalCorticalThickness.sh I'll try to reproduce the whole script call.

[dbrennan44]

heres the Template folder

there are a lot of intermediate templates

[cookpa]

Thanks! The template failed and is blank, which creates the error. Possibly related, I see that your input file names are identical - this will cause problems.

This is definitely something we should check for and fix at run time. Meanwhile, you might try changing the input files to MPRAGE_baseline.nii.gz and MPRAGE_followup.nii.gz

[dbrennan44]

I will try this and report, good to know! Previous successful runs with different data had different file names!

[dbrennan44]

I ran two subjects after changing the image basenames, and both ran thru antsLongitudinalCorticalThickness successfully. Think this was the problem, especially because a control cohort (1 scan each) ran fully successfully with no issue. Thank you!

[cookpa]

Great, thanks for sharing the info to help track this down! I think it is a simple fix to make the template scripts name their output uniquely. The longitudinal script already does this, so it should be fixed now.