Tractography parameters for tckgen

I started to work with the tractography on my DTI data. I have a multi shell DTI with 115 directions and 3 b0 (0, 1000, and 2000).
However, it is the first time that I use the tractography, and I have some doubts about the parameters that I have to use with tckgen.
I hope that someone can help me :slight_smile:
After the pre-processing, I resampled my data to 1.25 mm (originally, 2x2x2 mm)
After I used:
5ttgen on the T1-weighted images to generate the 5TT images
dwi2response with tournier
dwi2fod with csd
I created the white/grey matter border for seeds location
and I ran the tractography:

-algorithm iFOD2
-act 5TT
-nthreads 8
-minlength 15
-maxlength 250
-cutoff 0.06
n_seeds = 5 Milions

I am not sure if my parameters are totally correct. My big doubts are on -cutoff=0.06

Can someone that know better this part help me?


Hi Andrea,

After the pre-processing, I resampled my data to 1.25 mm (originally, 2x2x2 mm)

Firstly, it’s worth noting that this will change the default step size, which itself is based on voxel size. Is there a strong rationale behind this resampling? It’s used in the FBA pipeline for a specific purpose, which doesn’t necessarily translate to tractography; it can be beneficial for tractography as well, but it’s not something to just throw into a pipeline without knowledge of this consequence.

dwi2fod with csd

The dwi2fod csd algorithm will use only b=2000 data. Given you have multi-shell data, you can use the dwi2fod msmt_csd algorithm, which will make use of all of your DWI data, based on multi-shell response functions estimated using e.g. dwi2response dhollander.

My big doubts are on -cutoff=0.06

There is no correct answer to this question (or indeed any other here). I investigated cutoff a little when we revised the default parameters following software changes. The default in your circumstance is 0.05, so you could re-frame the question as “can I justify changing this parameter from the default?” (That way if it’s wrong you can blame me :expressionless: ) Same goes for minimum length: yours is considerably longer than the default in this case (2.5mm given your upsampling), and depending on what you’re doing this may or may not be consequential.


Hi Robert,
thank you so much for your answers.
I have just started to work in this part of the diffusion and I am trying to learn better everything.