fixel-based analysis in stroke patients

Dear MRtrix experts,

I’m currently analyzing a dataset of stroke patients with 18 strokes and 17 controls. There are 5 patients with right-hemispheric and 13 with left-hemispheric stroke lesions. The main focus is on comparing the affected to the unaffected hemisphere (and the correlation with different behavioral measures). I am wondering if fixel-based analysis with statistical analysis of FD, FC, and FDC is possible in this cohort. How do I take the different sides of stroke lesions into account? Is it possible to flip the dwi images beforehand?

Thank you very much for your help,

Sabeth

Dear Stephanie,

Just my humble opinion, and also not a real expert ;-).
The biggest problem in stroke patients is the huge anatomical variations you will have in your images.
If you only want to compare healthy sides this is less of a problem, and mirroring your images (after FOD calculation seems handiest to me, as you don’t have to adjust your bvecs) will be quite feasible). However, If you want to make a comparison of all affected patients, the anatomical variation in your dataset might dominate your result.
Another strategy might be tracking some fibres and calculating FD, FC and FDC of these fibres and compare these between patients and hemispheres

For taking into account the effect of hemisphere, you might add this as a covariate in your analysis.

I hope this was of any help to you.

Cheers,
Jeroen

Is it possible to flip the dwi images beforehand?

mrtransform -flip

mirroring your images (after FOD calculation seems handiest to me, as you don’t have to adjust your bvecs) will be quite feasible).

Well, you still need to flip the FODs in that case. If your diffusion gradient table is embedded within the image header, then mrtransform will deal with the gradient table flipping automatically.

If you want to make a comparison of all affected patients, the anatomical variation in your dataset might dominate your result.

One would expect in this instance that the variation in stroke location will simply result in a diffuse effect of small magnitude that won’t reach statistical significance. So I maybe wouldn’t say “dominate” the result, given FWE control will still be guaranteed.


What I would however suggest @Stephanie is carefully considering exactly what it is that you are testing. I.e. Does one expect to see an effect in hemispheres affected by stroke that is independent of the infarct location? Or further: Are you looking for a group comparison effect that cannot be explained be the locations of the infarcts (i.e. essentially “regressing out” infarct loading)? For the latter there’s the potential to utilise some of the upcoming statistical analysis goodies, if ti would make the numerical testing more faithful to your actual scientific hypothesis.