I might be starting a project analyzing nonhuman primate MRI images, which will be my first time working with nonhuman neuroimaging. I was wondering if anyone had any suggestions for DWI processing pipelines for nonhuman primates. Would existing ones such as QSIPrep or Tractoflow work with some modification?
Hi, I would start by looking at the PRIME-RE (Primate Ressource Exchange) wiki:
In general, preprocessing steps for DWI images are identical to those for human data. Main differences are:
- orientation issues: At the beginning it is very important to look at your images with tools like FSLeyes and check if the labeled are displayed correctly. If you have a way to distinguish between left and right, it is very important to check that those label are well positioned. Then check if the bevy are well oriented by displaying a quick DTI estimation o your data (FSL drift tool or others will help you do that).
- distorsion issue: usually primate images require higher resolution due to the smaller brain size which implies bigger distorsions than in human and lower SNR if the images are not acquired at higher magnetic field.
Tools used for human such as topup and eddy should do the job anyway. Denoising and degibbs tools (from MRTRIX) should be use also (before topup and eddy).
- bias: Often primate images are acquired with surface coils with strong receive bias. Bias correction tools such as the one implanted in mrtrix (using AntS N4 or FSL FAST) should be used.
I hope this helps!
QSIPrep works for macaques, no modification necessary! Give it a try with --dwi-only and --infant and you should see some usable results. I haven’t tried any of the reconstruction workflows other than dsi_studio_gqi yet though
Thank you both for your responses!
@mattcieslak Thanks for the suggestion, the preprocessed image looks good! However, based on the error I got from QSIRecon, it looks like the reconstruction workflow requires a T1 image. Is there anyway to get around that for working with primate data?
The reconstruction workflows are all geared toward human data and might not work as intended on primate data. I’d suggest taking the preprocessed data into specialized software for reconstructing primate data. I believe DSI Studio has been working on this and MRTrix should have some documentation on it also.
One warning, if you’re using very recent dsi studio you will want to use the
--flip_by=1 argument. If using mrtrix, be sure to use the .b file, not the .bval and .bvec files