Hi,
Sometimes it can help to take an ROI, cutting down the image FOV
rather than masking. Plus, it can be very difficult to get good bias
correction working on an EPI. Also, if you make the structural data
available then we can try a couple of other techniques here and
see if they work.
All the best,
Mark
On 12 Nov 2010, at 17:19, Marta Moraschi wrote:
> Hi Wolf,
> first of all thank for replying me.
> Yes, I've first tried to register the functional data (from surface coil) to
> the whole brain and it didn't work, but I've also tried to zero parts of the
> brain that are not present in the functional data, but it didn't work too.
> As you have suggested, running flirt, I've used masks selecting the part of
> functional data that did not show an intensity loss (in both cases, thus
> using the whole brain and the 'reduced whole brain').
> The data are oriented the same so I've also tried to use small search
> angles.
> I've not mentioned in my first mail but I'm working on the mean image
> obtained when correcting data for motion (spm).
>
> Thanks!!!!
> Marta
>
>
>
>
>
> Dr. Marta Moraschi -MARBILab -"Enrico Fermi" Center @ Fondazione Santa Lucia
> -
> Via Ardeatina 306 - Rome, Italy - Tel +390651501527
> -----Messaggio originale-----
> Da: FSL - FMRIB's Software Library [mailto:[log in to unmask]] Per conto di
> wolf zinke
> Inviato: venerdì 12 novembre 2010 16.34
> A: [log in to unmask]
> Oggetto: Re: [FSL] surface coil data registration
>
> Hi,
>
> Do you try to register the functional data to the whole brain? Maybe it
> helps, if you zero parts of the brain that for sure are not present in
> the functional data (fslmaths -roi option could help). Also, it might be
> helpful, if you use masks selecting the part of your functional data
> that does not show a strong intensity loss (flirt -inweight and
> -refweight options). With a rigid body registration, this might work out
> good enough. Make sure that the data is already oriented the same, so
> you can use small search angles. Instead of using -nosearch, I set
> -searchrx (/y and /z) to a small range.
> Is the functional data already corrected for motion? If so, you might
> try the average image (fslmaths -Tmean) instead of only the first
> volume. corratio seems to give reasonable results for me.
> This are just a few suggestions I can make based on my experience with
> monkey surface coil data, so maybe this could help you. I guess zeroing
> parts of the reference volume might help quiet a bit.
>
> Good luck,
> wolf
>
>
>
>
> Marta Moraschi wrote:
>> Dear all,
>>
>> I’m trying to register epi images of human visual cortex acquired with a
> surface coil ( ftp://lab-g1.phys.uniroma1.it/pub/surfcoil.nii.gz), thus
> suffering from large intensity variation and partial brain coverage, to T2
> weighted images acquired with a volume coil. I've run FLIRT, but it didn't
> work very well.
>> I’ve tried to use FAST before, in order to get the restored epi. I’ve
> tried to register the restored epi, using its bias field (previously max
> intensity normalised) as input weighting volume; I’ve mainly used 6 dof and
> normalised mutual information as cost function, but in point of fact I’ve
> tried every combination of DOF<6 and cost function. In any case the result
> was very far away from being good.
>> As advised in the forum, I’ve also tried to use the Nudge Gui, thus i’ve
> found my initial alignment and then i’ve run FLIRT with the search set
> "Already virtually aligned (no search)" and also use a cost-function
> weighting volume (i've tried both the intensity normalised bias field and a
> mask set to 1 everywhere except in the areas where distortion and signal
> loss are evident, both were previously reported in the partial epi space
> with nudge). In particular I’ve used the suggested command line:
>>
>> flirt -in partial_brain -ref anatomical -dof 6 -nosearch -init
> nudge_matrix.mat -inweight partial_brain_weighting_volume -out partial2anat
> -omat partial2anat.mat
>>
>> However, running flirt I’ve noticed a worsening of the alignment i've
> reached before with the Nudge gui.
>>
>> I would appreciate any advice. Many thanks in advance.
>>
>> Marta
>>
>
|