Hi,

Yes, CBF images should be registered much as is done for FMRI.
You should choose an image that shows lots of anatomical
structure for the registration, and then apply the transformation
to the other images that are in the original space.  That is, after
motion correction, you should have a set of images that are
aligned and then used to calculate CBF etc.  These images are
in original space - like the example_func space for our FMRI
processing, or equivalently, the same as the b=0 image for diffusion
imaging.  Any other images derived from this set of aligned images
will also be an image in the original space, although they normally
do not look like anatomical images (e.g. CBF maps, statistical images,
diffusion maps).

If you try to register anything like a CBF map or statistical image
to an anatomical image directly it will not work.  This is why you
need to take an image in the original space that still has some  
anatomical
details (e.g. the b=0 image, or one image from an EPI time-series
which we select as our example_func) and register this to the
structural image, then apply this transformation to all the other
images.

See our various documents such as the FSL course, FEAT/FLIRT
web pages and FAQ for more information on how we register EPI
to structural images for FMRI.  This should help make the process
clear.

All the best,
	Mark




On 18 Dec 2007, at 20:42, WangPing wrote:

> Hi Mark:
>
> One question just come up to me based on your email as below, I  
> think my situation is a little similar to fMRI registration.  For  
> example, we still consider the within-subject registration: in  
> fMRI, we need to register EPI images to the structure images.  In  
> my case, I need to register CBF (the resolution looks like EPI  
> images) maps to structure images.  I think I can use the  
> registration scheme as fMRI does.  So I am wondering how to  
> register EPI to structure images in fMRI within-subject?
>
> You mentioned "It is more usual to use an image that is already   
> aligned (an example functional or the b=0 weighting diffusion  
> image) for the  registrations and then transform these other images  
> using the spatial transformation found  by the previous  
> registration."  Could you explain this a little more with an  
> example of fMRI registration?  (You mean need an aligned EPI image?  
> How to get this)
>
> Best Regards,
> Ping
>
> ----------------------------------------
>> Date: Mon, 17 Dec 2007 10:07:03 +0000
>> From: [log in to unmask]
>> Subject: Re: [FSL] non-linear registration
>> To: [log in to unmask]
>>
>> Just to add to this:
>> If you are trying to get registrations within-subject then you
>> probably don't want to use
>> non-linear registration unless you are expecting substantial
>> distortions.  For the
>> within-subject case, 6 dof should be the *correct* degrees-of-freedom
>> to use, which
>> can be done with flirt.
>>
>> Can you describe more fully in what way that are "not good".
>> Note that you can only successfully register images that show the
>> same structures
>> (albeit with different contrasts).  So registering a CBF map (or ADC)
>> could be
>> problematic.  It is more usual to use an image that is already
>> aligned (an example
>> functional or the b=0 weighting diffusion image) for the
>> registrations and then
>> transform these other images using the spatial transformation found
>> by the
>> previous registration.
>>
>> I hope this is helpful.
>> All the best,
>> 	Mark
>>
>>
>>
>> On 17 Dec 2007, at 09:56, Steve Smith wrote:
>>
>>> Hi - FSL 4.0 includes the IRTK nonlinear registration tool; see the
>>> TBSS scripts for example usage. However you're on your own wrt
>>> getting this to work on multimodal data, as we've not had much
>>> experience on this with IRTK.
>>>
>>> The next release of FSL will include FNIRT, a generic nonlinear
>>> registration tool which seems to work really well.
>>>
>>> Cheers.
>>>
>>>
>>>
>>>
>>> On 14 Dec 2007, at 19:05, Ping Wang wrote:
>>>
>>>> Dear FSL users:
>>>>
>>>> I have data in this situation: each subject has images from
>>>> different modalities:  T1,T2, ADC, CBF.  Now I only consider one
>>>> subject, I want to register all other modilitiy images to T2
>>>> image.  I know Flirt only do linear registration, I tested the
>>>> results by Flirt, some are not good.  Should I use non-linear
>>>> registration?  I am wondering which tool in FSL can do non-linear
>>>> registration?  tbss?
>>>>
>>>> Thanks, Ping
>>>>
>>>> Windows Live Writer，支持离线撰写博客内容，随时随地想写就写。 立即
>>>> 使用！
>>>
>>>
>>> -------------------------------------------------------------------- 
>>> --
>>> -----
>>> 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]    http://www.fmrib.ox.ac.uk/~steve
>>> -------------------------------------------------------------------- 
>>> --
>>> -----
>>>
>
> _________________________________________________________________
> MSN 中文网，最新时尚生活资讯，白领聚集门户。
> http://cn.msn.com
>