Hi,
The -2d option in mcflirt is for when the input images are only one
slice
thick (hence 2d) rather than forcing the corrections to be only 2d.
If you
have only 2d slices like this then the third row of the matrix should be
0 0 1 0 or very close. If you have 3d images and you want to force a
2d transformation then you cannot do this directly with mcflirt but will
need to script calls to flirt instead.
If you have the mats directory from mcflirt then you can apply them
with applyxfm4D and specifying the input volume and directory for
the matrices. This is the easiest way. Alternatively you can also
script it with flirt, as mcflirt and flirt use exactly the same
conventions
for the matrices.
I'm afraid I can't think of a reference that discusses the exact
issues that you are raising here.
All the best,
Mark
On 29 Sep 2009, at 15:11, Daniel Kalthoff wrote:
> Hi folks,
>
> I have two questions regarding mcflirt and flirt:
>
> 1) When I used mcflirt with the -2d option I was expecting a MAT
> output close to identity
> (except for rotation around z and translation in x and y). However,
> the estimated motion
> parameters still include variations of the 3 other DOF. Is this how
> it's supposed to be?
>
> 2) I want to apply the same motion parameters (or their inverse) to
> another (artificial)
> dataset. Think I had to split up this dataset into its volumes and
> apply the MATs separately
> using flirt. How do I assure (in terms of command line parameters)
> that flirt does exactly
> the same with this dataset as mcflirt did with the original?
>
> 3) Does anyone know a good reference, in which the effect of motion
> correction on
> timecourse noise and autocorrelation is investigated?
>
> Thanks a lot,
>
> Daniel
>
|