No streamlines in corpus callosum using ACT


Hello MRtrix experts,

I’m trying to use ACT, but for some reason streamlines in the corpus callosum seem to be rejected. I’m using single-shell data, b = 1000, voxel size = 2 mm isotropic, 64 directions, acquired in both AP and PA phase encoding directions. A Siemens Prisma machine was used to acquire the data.

Correction for EPI distortions, eddy currents and motion artefacts was done using FSL topup/eddy. The diffusion data was then aligned to a T1 scan using dtiInit function from the Vistasoft/MrDiffusion package.

The following commands were used MRtrix3 to track streamlines using ACT:

dwi2response tournier dwi.mif response.txt

dwi2fod csd dwi.mif response.txt fod.mif -mask dwi_mask.mif

5ttgen fsl T1.nii.gz 5tt.mif

tckgen fod.mif ACT.tck -act 5tt.mif -seed_image dwi_mask.mif -select 500000

This gave the following results:

As you can see, the body of the corpus callosum lacks streamlines. I tried using -backtrack and reducing the fod amplitude threshold to 0.05, and got slightly better results:

But still, the corpus callosum lacks streamlines. I checked the white matter mask from the 5tt file and tracked streamlines without ACT to make sure the problem is from ACT, and both were fine:

I think the problem is that fibers of the corpus callosum extend into the anatomical CSF compartment due to some volume averaging. Here’s an the fod map to show you what I mean:

Do you have any idea what could be causing this? Is my assumption about the volume averaging accurate?

Looking forward to hearing what everyone has to say. Thanks for the help and for the great software!



That would be a reasonable guess in general for a scenario like this; however, it’s not the explanation for what happened in your particular case.

Yep! :wink:

Nope! :stuck_out_tongue_winking_eye:

Your problem at hand is unrelated to ACT; but you’ll have to look slightly closer to spot it: the x-component (column) of your gradient table is inverted. So even your non-ACT tractogram is quite wrong/broken as well, it’s just not as obvious as in the ACT-constrained output. That is probably because not having ACT on, allowed for just slightly longer (but still broken) short pieces of tracks that can fill up the corpus callosum with a bunch of red.

To solve this, you’ll need to revisit the first few steps of your pipeline: something went wrong there with interpreting the gradient table.

I reckon this particular step is a possible candidate of where things went wrong, due to the conversion of formats / assumptions of the gradient table or image coordinate system:


You’re right, the x-cmponent of my bvecs file was inverted! I fixed the problem and everything is working as it should now! Thank you so much for the help :grinning: