Hi - indeed we definitely do not recommend running BET before motion  
correction as that in general reduces the accuracy of the motion  
estimation.

I would concentrate on the registration options themselves rather than  
BET here - I would revert to the default FEAT BET and thresholding  
options whilst trying the following:

  - First, use fslswapdim to reorient your functional AND structural  
to have the same orientation as the standard space images provided  
with FSL. Make sure that you get the options correct so that there is  
no warning about left-right swapping.

- Then, see if registrations are now happy. If not, try changing the  
registration settings in FEAT - for example for func-Struc, change  
"Normal search" o "No search", if that doesn't work try setting DOF to  
6 then 7 then 9. Etc.

Cheers.


On 6 Jun 2009, at 12:19, Jonathan Ipser wrote:

> Hello,
>
> I am new to fMRI analysis and using FSL, so excuse me if there is  
> something
> obvious I am overlooking in the following query.
>
> I am busy preprocessing fMRI data from a Go/NoGo task. Each person  
> gets
> scanned twice (once on placebo, once on medication), with 12 runs of  
> the
> task in each scan. I have noticed that the functional data is  
> inverted with
> respect to the standard space and T1 structural image after  
> registration for
> some of the  runs (but not for others).
>
> It was suggested to me on inspection of these images that this  
> inversion
> might be occuring due to the parameters for brain extraction of the  
> fMRI
> data being too lenient, with the result that there is a lot of noise  
> outside
> of the brain which is throwing the registration algorithm off.  
> Indeed, if I
> use the BET GUI to process the fMRI data, and then use the resulting  
> brain
> image as input to the preprocessing step in FEAT (with the BET brain
> extraction option under the Pre-stats tab disabled), the problem with
> registration disappears (please see the images at
> http://projects.pry.uct.ac.za/uploads/fMRI for examples).
>
> My first question is whether running BET on the fMRI data prior to
> processing motion correction using FEAT is a legitimate procedure?  
> It has
> been mentioned in another post to this list that running BET prior  
> to the
> other steps might confuse mcflirt, due to the creation of "artificial
> intensity boundaries" (message: 017723). However, this does seem to  
> solve
> the registration problem, and the differences in the average  
> displacement
> picked up by mcflirt are slight when doing it this way and enabling  
> bet in
> FEAT (though see below).
>
> I have subsequently noticed that enabling the BET function in FEAT  
> does not
> actually appear to make any difference to the registration (as seen  
> from the
> images produced). This is confirmed by the identical results  
> produced when
> switching this option of and off, and the fact that no difference is
> observed when changing the -f parameter from 0.3 to 0.5 in  
> featlib.tcl. My
> second question therefore is why this should be the case (it is a  
> whole
> brain analysis so there are no FOV issues that would switch bet  
> off)? Is
> there some way of forcing bet to run in FEAT?
>
> Regards,
> Jonathan Ipser
> Department of Psychiatry and Mental Health
> University of Cape Town
>


---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK
+44 (0) 1865 222726  (fax 222717)
[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------