Any tips for working with Q-Ball data?

tractography
preprocessing

#1

I am working on comparing tractography from different acquisitions for surgical planning. One of the acquisitions given to me is a q-ball sequence (see gradients table below). A prior post indicated that q-ball support was removed from MRtrix, but before I go off looking at other packages, I thought I’d ask if anyone has used q-ball data for tractography in MRtrix. Currently I get this error (which makes sense):

145: tckglobal dwi.mif wm_response.txt -riso csf_response.txt -riso gm_response.txt -mask mask.mif -niter 1e9 -fod fod.mif -fiso fiso.mif tracks.tck
tckglobal: [ERROR] DWI volumes could not be classified into b-value shells; gradient encoding may not represent a HARDI sequence

Thanks!
-dan

144: mrinfo -dwgrad dwi.mif
          0           0           0           0
         -0   -0.999838  -0.0180093         100
         -0          -0          -1         100
  -0.999838          -0  -0.0180093         100
   0.707049   -0.707049  -0.0127376         200
   0.702398          -0   -0.711784         250
         -0   -0.702398   -0.711784         250
         -0    0.702398   -0.711784         250
  -0.707049   -0.707049  -0.0127376         200
  -0.702398          -0   -0.711784         250
   0.574484   -0.574484   -0.583041         350
   0.574484    0.574484   -0.583041         350
<snip>
   0.707099   -0.707099 -0.00477759         900
   0.704339          -0   -0.709863         900
         -0   -0.704339   -0.709863         900
         -0    0.704339   -0.709863         900
  -0.707099   -0.707099 -0.00477759         900
  -0.704339          -0   -0.709863         900
   0.665663   -0.665663   -0.337322        1000
   0.664308   -0.332154   -0.669603        1000
   0.664308    0.332154   -0.669603        1000
   0.665663    0.665663   -0.337322        1000
   0.332154   -0.664308   -0.669604        1000
   0.332154    0.664308   -0.669604        1000
         -0   -0.999988 -0.00499981        1000
         -0          -0          -1        1000
  -0.332154   -0.664308   -0.669604        1000
  -0.332154    0.664308   -0.669604        1000
  -0.665663   -0.665663   -0.337322        1000
  -0.664308   -0.332154   -0.669603        1000
  -0.664308    0.332154   -0.669603        1000
  -0.665663    0.665663   -0.337322        1000
  -0.999988          -0 -0.00499981        1000

#2

Short answer is no.

Long answer will need more info. What do you mean by ‘Q-ball data’ exactly? You’re passing what looks like a raw DWI image series (at least it’s called dwi.mif), is that the case? If not, what format are the images stored in? Are we talking about SH coefficients for the QBI data? Or samples along certain directions…?

Also, the DW scheme contains different b-values, which is unexpected for QBI data – the original method only works with single-shell data. The number of directions is also very small (only ~15 directions for the b=1000s/mm² shell?), and the max b-value is very low for QBI: the original method used b=4000s/mm², and the central approximation for the dODF provided by the Funk-Radon transform is valid in the limit of high b-values. In short, this looks nothing like any acquisition I’d use for a QBI reconstruction…

Where did you get these data? Personally, I’d go back to your source and verify what it was they had in mind with this. From what I’m looking at, I reckon this might have been intended for Diffusion Kurtosis Imaging, not Q-ball Imaging – and even then, the max b-value is very low compared to what would normally be recommend for DKI (in the region of b=2,500s/mm² would be more typical)…


#3

@jdtournier, thank you for the helpful response. I’m not familiar with QBI data (nor DKI), and was working from the series description. I converted the images using mrconvert, tried the usual pipeline and was a bit stumped by the error message.

The data was collected on a volunteer to see if different vendor sequences could resolve crossing fibers in the brain stem. Fitting in volunteers is a bit tricky in our setting and happens on the spur of the moment. My hunch is that they tried a bunch of DTI/DWI sequences. Next go-around will require a bit more thinking and planning.

Thanks again for the helpful response!