Hi,
You should use 12 DOF as the extra DOF (scaling and skews/shears) account for the eddy current distortions. This is the whole point of having them there. So I strongly advise you to redo the registration with 12 DOF if you want to correct for the effects of eddy currents.
The -nosearch option just limits the initial search range for the rotations, making it more robust to artefacts generally, as long as the initial alignment is fairly close. If the initial alignment might be out by say 30 degrees or more in any axis then this option should not be used. It is roughly equivalent to -searchrx 0 0 -searchry 0 0 -searchrz 0 0.
The paddingsize option affects how extrapolation is done outside of the valid FOV. When rotating/shifting images it can be useful to extend the end slice, so that voxels are not lost from the analysis just because they've moved part of a voxel outside of the valid FOV. So we allow up to 1 voxel at the end slices (effectively just copying the end slices when necessary) so that small rotations and translations can still maintain useful signal in these slices. If your brain does not touch the end slices then this option won't really make any difference.
I hope this helps.
All the best,
Mark
On 15 Mar 2013, at 16:19, Vincent Koppelmans <[log in to unmask]> wrote:
> Dear FSL experts,
>
> I have an 'older' DWI dataset. It comprises 16 3D images: 1 B0 image and 15 diffusion weighted images.
>
> Instead of using the eddy_current script, I realigned the 15 images to the B0 image using FLIRT with 6 dof.
>
> I noticed eddy_current does not specify the number of degrees of freedom, and thus uses 12 dof.
>
> (line from eddy current script= ${FSLDIR}/bin/flirt -in $i -ref ${output}_ref -nosearch -o $i -paddingsize 1 >> ${output}.ecclog)
>
> I also noticed that it is using two options hat I am not sure of what they are doing, or why they are desired:
>
> -nosearch
> -paddingsize 1
>
> Could somebody explain:
>
> a) what the two options exactly do (the explanation that comes with flirt -help is not very extensive)
>
> b) if it is better to use the default 12 dof over the 6 dof I am using.
>
> Thanks,
>
> Vincent
|