5ttseg keep original voxel numbers and dimensions


After 5ttseg, the dimensions of the 5ttseg.mif output is different from that of the input, even when using -nocrop option. I’m using the WM output for subsequent steps, so it’s slightly problematic to have to do a rigid registration again to the original T1.
Could you please tell me how to keep all the original information in the segmentation output?

Many thanks

Hi Adriana,

I have some vague recollection of having fixed an issue with 5ttgen changing the output voxel grid despite having been provided with the -nocrop option. This shouldn’t be happening…

  • Are you using an up-to-date MRtrix3 version?
  • Which 5ttgen algorithm are you using?
  • Are you providing any other command-line options?

Assuming the voxel grid is just being cropped and not reoriented in any way, you should be able to use mrtransform with the -interp nearest and -template command-line options to get back to the original voxel grid with no loss of detail. But obviously my own interest is in solving the underlying issue.


Hi Rob,

Thanks for getting back to me.
Yes I am using MRtrix3 and the fsl algorithm.
I am currently using mrtransform following 5ttgen. But since I’m applying this to many images and have to QC after each step, I thought I might be missing something with the 5ttgen command to prevent it from changing the voxel grid.

Thanks again,

Yes I am using MRtrix3 …

Can you provide a more precise version (i.e. output of mrconvert -version)?

The following version, from June 22 2018
== mrconvert 3.0_RC3-51-g52a2540d ==

I know it was a while ago, but could you share how you solved this? I am experiencing the same problem and can’t seem to find any thread where this was answered.

I listed the issue on GitHub, but have not gotten around to investigating the source of the problem. If anybody can provide data that would allow me to reproduce the issue very quickly, rather than having to experiment with different modifications of the data to try to replicate independently, that would greatly increase its chances of surviving my time prioritisation… :pleading_face: