Hi,
Sounds like you have tried all the standard options.
So please upload the relevant images (tar file is best) to:
http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi
and send us the reference ID number and we'll have a
look at it and see if a better registration can be found.
All the best,
Mark
On 27 Sep 2006, at 00:30, Rutvik Desai wrote:
> Thanks. I don't have a fieldmap, so I tried using various types of
> masks as weighting matrices. Several ventral slices were masked out to
> prevent any effects of signal loss in the anterior/ventral temporal
> lobe.
> Also tried a larger mask that removes roughly ventral half of
> the brain (including orbitofrontal areas), and it still wants to push
> the
> epi down. (I also tried physically removing slices, just in case the
> weighting
> option didn't work as expected, but that did not help either). Another
> mask was created by thresholding out low signal areas in the bottom
> half
> (but keeping eveything in the top half). This mask removes some
> inferior
> frontal and some temporal areas. It is a little better, but the problem
> persists.
>
> thanks,
> Rutvik
>
> On Sep 26, 2006, at 4:36 PM, Mark Jenkinson wrote:
>
>> Hi,
>>
>> I've seen similar things before. The best solution is usually to
>> apply
>> unwarping and signal loss masking, as often the incorrect translation
>> is an attempt to match up the inferior frontal area, which has little
>> signal in the epi, with the background.
>>
>> Do you have a fieldmap?
>> If so, I recommend using fugue for the registration - which is
>> easiest through the Feat GUI. If not, then I suggest making a
>> mask by hand which is zero in the areas that have signal loss
>> and one everywhere else (including the background areas
>> adjacent to parts of the brain that have good signal). Then
>> try using this mask as a cost function weighting in flirt.
>>
>> Hopefully one of these will fix the problem.
>> Let us know how you get on.
>>
>> All the best,
>> Mark
>>
>>
>> On 26 Sep 2006, at 20:15, Rutvik Desai wrote:
>>
>>> I am trying to register an epi to an anatomy using flirt to get
>>> better alignment. The anatomy has
>>> been skull-stripped. The anatomy is 256X256X134, with 0.81x0.81x1.0
>>> mm voxels. The epi is
>>> 64x64x36, with 3.25x3.25x3.30 mm voxels. Seems that flirt always
>>> wants to push the epi down (in
>>> z direction, ventrally). I have tried various things, inlcluding
>>> chopping off the bottom parts of
>>> anatomy and/or the epi. This happens with a number of subjects.
>>> Typically the epi is 2-3mm too
>>> high, and is pushed down by 5-6 mm, making it too low. I have also
>>> tried all cost functions, and
>>> tried reducing the 'finesearch' value to 6 and 8 from 18. Any help
>>> would be appriciated. I can send
>>> an anat and an epi if necessary. Thanks!
>>>
|