Hi,

Yes - that's right - you need to crop the structural to make invwarp  
work
faster.  If you remove all the space from around the scalp in all  
directions
(using fslroi) then you should be able to get invwarp to run in less  
than
an hour.  But make sure that you do the cropping *before* either
FLIRT or FNIRT is run.  Everything needs to be run using the cropped  
image.

As for FLIRT - use the MNI152_T1_2mm with 12 DOF.

See the documentation at:
   http://www.fmrib.ox.ac.uk/fsl/fnirt/index.html#fnirt_examples
for examples on how to run it (after you've cropped the image
to start with).

All the best,
	Mark


On 27 Oct 2009, at 09:25, Alessio Moscato wrote:

> Hi Jesper..
>
> thanks a lot for your precious indications.. as you easily  
> understand I'm not very expert in fnirt..
>
> I'll try what you told me.. only other 2 little questions:
>
> 1. all of these indications don't change anything about the invwarp  
> problem we discussed at the beginning? to reach short invwarp  
> process time the cool move is to crop the structural volume..is it  
> correct?
>
> 2. when you say me to use
>
> " fnirt --ref=MNI152_T1_2mm --config=T1_2_MNI152_2mm.cnf -- 
> in=my_scan --cout=my_warps --aff= ... "
>
> the affine transform what could be? I flirted my_T13D to MNI_05mm  
> with 12 DOF.. I suppose that in order to use your suggested command  
> I need to flirt my_T13D to MNI_2mm..is it correct? how many DOF are  
> suitable?
>
> I really thank you for everything
>
> All the best
>
> Alessio
>
> On Tue, Oct 27, 2009 at 10:11 AM, Jesper Andersson <[log in to unmask] 
> > wrote:
> Hi again,
>
>> Hi Jesper.. yes, it took 30 hours
>>
>> I report you the command I used:
>>
>> fnirt --ref=MNI152_T1_05mm.nii.gz --in=my_T13D.nii -- 
>> aff=my_affine_transform_fromT13DtoMNI05mm.mat
>>
>> Could be this the problem?
>
> yes! First of all you are using the "default defaults", and they are  
> not particularly well suited for anything (I should get around to  
> changing them). We recommend always using a config-file for  
> specifying how fnirt is to be run. When registering to the MIN152 T1- 
> weighted scan you should typically use the T1_2_MNI152_2mm.cnf file.
>
> Secondly, you are registering to the 0.5mm template. Since you are  
> only estimating the warps to a 10mm resolution (which is _not_ the  
> same as a 10mm accuracy of alignment of structure) it is  
> computationally wasteful and unnecessary to use a template with such  
> high resolution (it will not give you any better results).
>
> I recommend instead to register to the MNI152 2mm template. If, for  
> some reason, you want to have your warped images in the space of the  
> 0.5mm template you can still use the warps estimated from the 2mm  
> template to accomplish that. Simply do
>
> fnirt --ref=MNI152_T1_2mm --config=T1_2_MNI152_2mm.cnf --in=my_scan  
> --cout=my_warps --aff= ...
> applywarp --ref=MNI152_T1_0.5mm --in=my_scan --warp=my_warps -- 
> interp=spline --out=my_scan_in_0.5mm_standard_space
>
> Good luck Jesper
>
>
>