You could do that but my opinion is that not doing rotation of the bvecs is a
tiny effect, whereas not doing any eddy current correction is a much more serious
effect. This is why we have gone for eddy current correction and not bvec rotation.
All the best,
On 20 May 2011, at 16:31, Dr T.M. Rolheiser wrote:
> This sounds pretty good so far. I will probably try to change the script by adding -dof 6, because then I could use the rotbvecs script (which depends on the .ecclog file) to alter the bvecs based on head motion only.
> This is a really helpful discussion board... I ought to do this more often.
> Thanks a bunch... Tyler
> On May 20 2011, Stamatios Sotiropoulos wrote:
>> Eddy correct just calls flirt. You can run flirt instead of eddy_correct through the command line and specify the settings you want to apply.
>> On 20 May 2011, at 16:01, Dr T.M. Rolheiser wrote:
>>> Thanks for the quick reply,
>>> I suppose the underlying issue is that I would like to correct bvectors for motion, but not for eddy currents, and if eddy_correct does both in a way that can't be separated, then I'm not sure whether it is prudent to employ a bvector rotation.
>>> The raw data is of very good quality, and the preprocessed images always look very good so perhaps this is less of a problem, and more of a curiosity.
>>> Is it possible to have eddy_correct simply use the 6-dof procedure? Would this involve changing the code?
>>> Thanks again... Tyler On May 20 2011, Stamatios Sotiropoulos wrote:
>>>> Hi Tyler,
>>>> 1) Eddy_correct is based on a simple model for eddy current artifacts that include scaling, translation and shear of the images. All these artifacts, including head motion can be captured by an affine transformation. That's why eddy_correct uses 12 DOFs rather than 6, which would only correct for head motion.
>>>> 2) The link "FDT theory" on the FDT webpage describes how bedpostx and probtrackx work. For dtifit there is no much theory, it is a simple least squares fit of the linear DTI model. Peter Basser's first DTI papers (any many other review papers since then) have all the information you need.
>>>> Hope this helps,
>>>> On 20 May 2011, at 13:02, Tyler wrote:
>>>>> Hello FSL Kingdom, hopefully someone can help me out with two specific questions. 1) Regarding the naming and function of the command eddy_correct: If all the command does is register the diffusion images to the reference image (12-degree of freedom affine), can this be considered an eddy-current correction? Granted it is important to align the diffusion images to a reference image before fitting a tensor, but how would one directly address the issue of correcting for eddy distortions independent of head motion? Does FSL have this capability? 2) As a secondary question, where can I find more detailed information about the FDT pipeline? The general information on the fsl webpage doesn't really provide enough information about the tools (e.g., how does the dtifit command calculate the FA/MD images). PS: I'm sorry if these posts have already been covered... I searched the archives but didn't come up with a direct answer.