Generating 5tt image for HCP data

1

Hey,

I tried to generate a 5tt image for HCP 3T data. I tested it with the T1w_acpc_dc.nii.gz and with the T1w_acpc_dc_restore_brain.nii.gz file and the -premasked option.

Unfortunately the command ends with a warning: “[WARNING] Generated image does not perfectly conform to 5TT format:”

The _restore_brain run looks a lot better but I’m still wondering whether I should use the result or if I can tweak the generation somehow.

This is a slice of 5ttvis.mif from the _restore_brain input:

image
image.jpg1512×1576 202 KB

2

Hi Darius,

I’ve experienced this with HCP data. chances are that below (or above) that warning there will be a line saying how many voxels are affected (it is usually something tiny (between 1-5). You can also check which voxels are affected by running:

5ttcheck -mask prefix 5tt.mif

The voxels affected aren’t usually brain voxels.

I believe there was some other post about this at some point - if my memory serves the issue seems to stem from the way FSL operates…

3

Thanks,

that command resulted in a messasge telling me that the file is a correct 5tt image. Which is a bit odd because the generation produced the stated warning. Thanks for your help I will just use this image now.

Best,
Darius

4

This warning can arise either due to:

  • Use of an older version of MRtrix3, where some of the internal logic of 5ttgen fsl wasn’t quite accurate, leading to numerous non-unity-sum voxels;

  • Use of a non-tagged version of MRtrix3 with the right combination of updates and bugs, such that 5ttgen erroneously detects the “Image is OK!” message generated by 5ttcheck as an error message to be escalated;

  • Due to FSL fast sometimes providing one or two voxels where the sum of partial volume fractions is not 1.0.

In any case, as long as there aren’t any brain voxels where the sum of partial volume fractions is very different to 1.0, there shouldn’t actually be any consequences for tracking.

Rob