Dear Theodor,
These are good questions - especially brain extraction.
In your case I would not do brain extraction at all.
It will be more helpful to have the non-brain structures on
the side where you don't have brain in order to constrain the
registration more naturally.
Your pipeline looks good and I would use the b0 image for
registration, mainly because it will show sensible non-brain
structures. It should work, hopefully, with or without using
a mask. But you can always try both, just in case.
All the best,
Mark
On 22 Nov 2011, at 23:40, Theodor R?ber wrote:
> Dear Mark,
>
> Thank you so much for your answer! I have tried to quickly sum up the procedure that you have described just to be sure that I got it right. Furthermore, I have some questions. Would you mind to take a quick look at both?
>
> 1. Use flirt to register the DTI to the individual T1.
> 2. Use flirt AND fnirt to register the individual T1 to the MNI T1-template.
> 3. Take the resulting transformation matrices to register the DTI (the one that has been registered to the individual T1) to the MNI T1-template.
>
> Where would you do the brain extraction? At the end, after the DTI has been registered to the MNI T1-template?
>
> When registering the DTI to the individual T1, would you recommend to use the b0 as input volume?
>
> Is it necessary to use masking, when registering the DTI to the individual T1 with flirt? I thought that it wouldn’t, since the DTI and the T1 have “the same” lesion?
>
> Thanks again - your help is greatly appreciated!
>
> Best,
>
> Theodor
>
> On Tue, 22 Nov 2011 09:49:53 +0000
> Mark Jenkinson <[log in to unmask]> wrote:
>> Dear Theodor,
>> You definitely need to use masking to get this to work.
>> We have had success previously in dealing with hemispherectomy
>> patients so I know it is possible. The trick is in getting the mask
>> right, so that it is zero in the missing hemisphere, but ones everywhere
>> else, including the background and the skull/scalp areas. If you do
>> this, and try various options for cost functions then you should be
>> able to get the registrations to work. Note that you are likely to be
>> best off trying to get the DTI-T1 registration to work first with FLIRT
>> only (trying various cost functions potentially and making sure that
>> the orientations start similarly - ideally use fslreorient2std on each
>> first). Then try to get that individual's T1 registered to the MNI
>> T1 template, initially with FLIRT and then with FNIRT. The latter
>> may be trickier and you may need to be very careful with the
>> mask and possibly change the control point spacing and/or the
>> regularisation parameters, but start with the standard config file
>> and a good mask and see how you get on.
>> All the best,
>> Mark
>> On 21 Nov 2011, at 17:07, Theodor R?ber wrote:
>>> Dear FSL-Experts,
>>> I have a bunch of DTI-datasets of patients who have undergone functional hemispherectomy. Since I want to run voxelwise analyses, I need to normalize these highly deformed brains.
>>> I have tried to normalise the FA-maps to the MNI-template using flirt/fnirt and masking out the lesions but the results were not really convincing. Do you have any advice on how the normalisation could be done best?
>>> Any input on this is highly appreciated.
>>> Thanks in advance!
>>> Best,
>>> Theodor
>
> ===================
> Theodor Rüber
> Remigiusplatz 1
> 53111 Bonn
> Tel.: 0228/24005219
>
|