[Mrtrix-discussion] eddy correction and ideal number of streamlines for TDI

[Robert Smith]

Hi Todd

1. For high b-value data on Siemens scanners, we recommend avoiding
eddy_correct, for a couple of reasons. Firstly, because Siemens scanners
come with the twice-refocused spin-echo sequence, eddy current distortions
are typically sub-voxel, so not worth worrying about. Secondly,
eddy-correct is based on a simple affine registration between images; it
works for low b-value data because the SNR is high and the image contrast
is (relatively) homogeneous between DWIs, so it's just a matter of aligning
the images. But for high b-value data where the SNR is low and the image
contrast changes drastically between DWIs, eddy_correct can cause more
problems than it solves.
We hope to have a tool for motion and eddy current distortion correction of
high b-value data in the next major release of MRtrix.

2. This is an unfortunate glitch that made its way into MRtrix version
0.2.10. However the only error is the printing of that message; it does not
influence the data, processing or visualization in any way. You can find a
more detailed description of the cause (and solution) of this problem
here<http://www.nitrc.org/pipermail/mrtrix-discussion/2012-December/000597.html>;
if you implement the code modification proposed there, don't forget to run
the MRtrix build script afterwards!

3. As with most things tractography related, it's all subjective. I'd say
that 1 million is the bare minimum for generating a TDI from whole-brain
fibre-tracking; we tend to use 5 or 10 million. Curiously, the TDIs can
actually look 'worse' if you generate many more streamlines, because the
texture in the image (which helps to see the directionality of the
streamlines) gets smoothed out. You can play around with different
streamline counts and voxel resolution to see what you think is best for
your particular application.
If you generate a large number of streamlines, you can extract only the
first N streamlines from the file using the truncate_tracks command. This
will allow you to generate TDIs of the same subject & at the same
resolution for different numbers of streamlines, without having to
re-generate streamlines for every desired number.

Best regards
Rob


--

Robert Smith
PhD Candidate

The Florey Institute of Neuroscience and Mental Health
Melbourne Brain Centre - Austin Campus
245 Burgundy Street
Heidelberg Vic 3084
Ph: +61 3 9035 7128
Fax: +61 3 9035 7301
www.florey.edu.au



On Thu, Dec 13, 2012 at 10:18 AM, Todd Jolly <todd.jolly at uon.edu.au> wrote:

>  Hi all,
> I have been looking into using streamtrack SD_PROB for whole brain
> tracking to create TDI. My dMRI sequence was acquired on a 3T Siemens
> Verio, voxel size 2.2mm isotropic, using diffusion gradient b=3000 with 64
> directions. I was wondering if anyone would have any advice or suggestions
> in regard to the following:
>
> 1. Should perform eddy_current corrections on my dMRI images as the mrtrix
> wiki does not show a step for eddy current correction? I didn't want to use
> FSL just in case there were errors with the bvec gradient scheme when using
> it in mrtrix.
>
> 2. I also get the following error:
> disp_profile: unexpected number of SH coefficients - assuming response
> function
>
> My input is as follows
> erode subjectIDmask.mif -npass 3 - | mrmult subjectIDfa.mif - - |
> threshold - -abs 0.7 subjectIDsf.mif
> estimate_response subjectIDdwi.mif subjectIDsf.mif -lmax 8
> subjectIDresponse.txt
> disp_profile -response subjectIDresponse.txt
>
> I think I saw a post where Donald said that this has been fixed in the
> latest version so should I upgrade my version/ redo all analyses?
>
> 3. What is the optimum number of streamlines needed to achieve good
> quality TDI, I have seen some people use 4 million but others have used 2
> million?
>
> Thanks,
> Todd
>
> _______________________________________________
> Mrtrix-discussion mailing list
> Mrtrix-discussion at www.nitrc.org
> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20121213/4ac080a5/attachment.html