Print

Print
Hi Julia

Have you looked at 3way crossings? There should be clear differences there, have a look at bedpostx results after fitting 3way fibres.

Without knowing much about the data, and if you still find that there are not major differences between the single and multi-shell analysis, the safest would be to stick with the b=3000 shell.

Cheers
Stam



On 5 Jun 2014, at 22:25, Owen, Julia <[log in to unmask]> wrote:

Hi Stam,

Thanks so much for the help.

I rotated the bvecs and I am now getting reasonable tractography.  I guess my next question is one that you may not be able to answer, should I proceed with the two shell fitting (with the suboptimal difference in scan parameters) or just use one shell (b=1000 or b=3000)?  Empirically all three results look about the same from the test seed I tried, so I was hoping that you could tell me which one you would go with.  Again, this is likely not a question you can answer without looking at the data, but I thought I'd ask anyways.

Thanks again.

Best, Julia
From: FSL - FMRIB's Software Library [[log in to unmask]] on behalf of Stamatios Sotiropoulos [[log in to unmask]]
Sent: Thursday, June 05, 2014 12:26 AM
To: [log in to unmask]
Subject: Re: [FSL] Two shell bedpostx

Hi Julia

Yes, you need the sequence parameters to be matched across shells (otherwise you need to non-trivially account for T1 and T2 effects and also differences in baseline SNRs). Also you need to make sure that preprocessing is done jointly on the two shells. But even if your sequence parameters are not matched, you should be able to estimate orientations that make some sense. How did you concatenate, have you registered the two shells together and rotated the bvecs of the transformed shell?  

Cheers
Stam



On 5 Jun 2014, at 00:44, Owen, Julia <[log in to unmask]> wrote:

Hi,

I've read old posts about using bedpostx with two shells that say to use the --model=2 option.  I have data with two shells, one at b=1000 and one at b=3000.  I tried bedpostx with the --model=2 option (FSL 5.0.4) on the two shells concatenated together and got bad results (a test tractography run just looked like isotropic diffusion from the seed).  The tractography looks great with just the shells run separately, so I don't think it's a data/bvec issue.  A potential culprit is that the b=1000 and b=3000 data were collected with different scan parameters (different TE/TR), the study was not designed to do any modeling that needed multiple shells.  Does this fact preclude me from running bedpostx with both shells?  Any other suggestions?

Thanks!