Dear Mark,
Thanks for the advice.
Best regards,
Annouchka
__________________
Annouchka Van Impe
Motor Control Laboratory
Department of Biomedical Kinesiology
K.U.Leuven
Tervuursevest 101
3001 Heverlee, Belgium
Tel : +32 (0)16 32 90 66
E-mail : [log in to unmask]
________________________________________
Van: FSL - FMRIB's Software Library [[log in to unmask]] namens Mark Jenkinson [[log in to unmask]]
Verzonden: dinsdag 16 augustus 2011 12:36
Aan: [log in to unmask]
Onderwerp: Re: [FSL] FEAT initial_highres2highres
Dear Annouchka,
It is not easy to change the FEAT GUI for your own analysis.
You can change it for *everyone* quite easily by editing the
$FSLDIR/tcl/featlib.tcl file.
Alternatively, simply run FLIRT after you've run FEAT,
like you have done, then replace the example_func2highres.mat
file (or initial_highres2highres.mat if you are using an EPI
initial highres image) and then use "updatefeatreg" to update
your FEAT analysis using the new registrations.
I would advise the second option.
All the best,
Mark
On 16 Aug 2011, at 11:17, Annouchka Van Impe wrote:
> Hi,
>
> I didn't do the conversion to nifti myself. I got all the nifti images (epi scans as well as T1's) directly from the Philips scan computer. The pixdim values seem to match with the original settings on the scan computer.
>
> Meanwhile I have tried to register the full coverage EPI to the T1 structural just using FLIRT instead of the FEAT GUI. I've changed the cost function from 'correlation ratio' to 'mutual information' and this seems to improve the registration. Is it possible to change the code underlying the FEAT GUI so it uses 'mutual information' as the default setting for registration when opening the FEAT GUI?
>
> Thanks for your help.
>
> Annouchka
> __________________
> Annouchka Van Impe
> Motor Control Laboratory
> Department of Biomedical Kinesiology
> K.U.Leuven
> Tervuursevest 101
> 3001 Heverlee, Belgium
> Tel : +32 (0)16 32 90 66
> E-mail : [log in to unmask]
>
> ________________________________________
> Van: FSL - FMRIB's Software Library [[log in to unmask]] namens Jesper Andersson [[log in to unmask]]
> Verzonden: dinsdag 16 augustus 2011 10:25
> Aan: [log in to unmask]
> Onderwerp: Re: [FSL] FEAT initial_highres2highres
>
> Hi again,
>
> it is a little hard to see on those overlays, but I get a feeling that the scaling in the z-direction is different. It appears the full coverage EPI is slightly taller than the T1 structural. This could be caused by a "faulty" conversion from your native format (e.g. dicom) to nifti. Can you use fslhd to have a look at the value of pixdim3 and compare that to what you prescribed the interslice distance to be when you performed the scans?
>
> Jesper
>
> On 16 Aug 2011, at 09:02, Annouchka Van Impe wrote:
>
>>
>> Dear Jesper,
>>
>> I'm replying to your personal email address, because I could not upload documents on the listserver.
>>
>> Indeed, I meant the registration of the full coverage EPI image to the T1 structural. I've added an example of this registration step in an elderly subject.
>> In most subjects it's only around the ventricles that the registration is off, in some the registration of the cerebellum is also not good.
>> I've used the fslreorient2std command to make sure that the T1 image is in the same orientation as the epi images.
>>
>> Kind regards,
>>
>> Annouchka
>> ________________________________________
>> Van: FSL - FMRIB's Software Library [[log in to unmask]] namens Jesper Andersson [[log in to unmask]]
>> Verzonden: dinsdag 16 augustus 2011 9:36
>> Aan: [log in to unmask]
>> Onderwerp: Re: [FSL] FEAT initial_highres2highres
>>
>> Dear Annouchka,
>>
>> in your initial mail you indicate that you think the main problem was in the "the initial highres to the main structural image". Do you by that mean the registration of the full coverage EPI image to the T1 structural image? What does it look like? Are the registrations completely off, or are they mainly fine but with some parts looking worse?
>>
>> If it is the former, are the images in the same orientation to begin with? If not it may be useful to use fslswapdim to make sure they are.
>>
>> If it is the latter it might be caused by distortions in the EPI images. We strongly recommend correcting these using phase maps.
>>
>> Jesper
>>
>> On 16 Aug 2011, at 08:23, Annouchka Van Impe wrote:
>>
>>> Hi Steve,
>>>
>>> Thanks for replying.
>>> I tried lowering the warp resolution to 8mm, but the registration did not really improve.
>>> How far can I go in lowering the warp resolution? Are they any other processing steps that I should try in order to improve the registration.
>>> The preprocessing already includes a fieldmap unwarping.
>>>
>>>
>>> Cheers,
>>>
>>> Annouchka
>>>
>>>
>>> ________________________________________
>>> Van: FSL - FMRIB's Software Library [[log in to unmask]] namens Stephen Smith [[log in to unmask]]
>>> Verzonden: vrijdag 12 augustus 2011 8:31
>>> Aan: [log in to unmask]
>>> Onderwerp: Re: [FSL] FEAT initial_highres2highres
>>>
>>> Hi - if the problems are just around the ventricles, the issue is likely to be with the nonlinear registration, not the initial affine? Is the overall registration (globally) between functional and structural reasonable? If you have huge atrophy in some subjects then maybe you need to increase the detail of the nonlinear registration - so maybe reduce the warp resolution spacing?
>>>
>>> Cheers.
>>>
>>>
>>> On 11 Aug 2011, at 11:10, Annouchka Van Impe wrote:
>>>
>>> Dear all,
>>>
>>> I'm using Feat to analyze 2 sets of fMRI data: an elderly group and a young control group. I'm having problems with registering the data to MNI space, especially in the elderly group. The alignment around the ventricles is not so good. I'm using the following settings in the FEAT GUI:
>>>
>>> -Initial structural image: a whole brain-extracted epi scan (Normal search, 3DOF)
>>> -Main structural image: brain-extracted T1 (Normal search, 6DOF)
>>> -Standard space: MNI152_T1_1mm_brain (Normal search, 12 DOF, non-linear, warp resolution 10 mm)
>>>
>>> The problem seems to arise when registering the initial highres to the main structural image (registration of the main structural image to standard space is fine). I' ve tried changing the DOFs and changing the settings from 'normal search' to 'full search' but nothing seems to help much.
>>>
>>> Any suggestions on how to improve this registration would be greatly appreciated.
>>>
>>> Kind regards,
>>>
>>> Annouchka
>>>
>>>
>>>
>>> ---------------------------------------------------------------------------
>>> 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]<mailto:[log in to unmask]> http://www.fmrib.ox.ac.uk/~steve
>>> ---------------------------------------------------------------------------
>>>
>> <initial_highres2highres.png>
>
|
