DTIprep before analysis in MRtrix

Dear experts,

I apologise in advance for the question that is not directly related to MRtrix3.

I’m quite new to diffusion MRI and before starting the analysis in MRtrix3 I want to perform diffusion data quality check in DTIprep. In our project we have multishell data with b values:
b0 = 10​
b500 = 18​
b1250 = 36​
b2500 = 53

DTIprep seems to have an option to remove the gradients that it considers as “bad”. So I was wondering if removing gradients in a multishell data may somehow affect the analysis and results? Especially that different gradient are removed from each person?

Thank you in advance,
Paulina

Welcome Paulina!

Consequences of removal of diffusion volumes in the context of CSD is best understood in the context of maximum spherical harmonic degree. Assuming that the omission of volumes does not cross one of these numerical thresholds, the primary consequences are:

  • Decrease in “SNR” (i.e. you simply have less data);
  • Increase in condition number of the transformation between DWI intensities and SH coefficients (i.e. the calculations are “less stable”).

For small numbers, this is fine, even if the particular volumes removed are different for different participants. Though I would advise keeping track of the number of volumes omitted from each subject, and using that as a nuisance regressor in downstream analysis.

If however the omission of volumes does cross one of these thresholds, then that might be slightly more consequential. It no longer affects response function calculation as it once did (see the method cited in the amp2response command), but within CSD (both original and MSMT implementations) it will change the lmax of the initial transformation prior to introduction of the non-negativity constraints. The extent to which the final FOD estimation may be influenced by a change here I’m not sure; @jdtournier may have some insight, my suspicion is that the final results should not change drastically due to the lmax of the initial transform changing, but it’s a naïve sense.

Cheers
Rob