Print

---

I don't have any easy solutions for this - other than to modify the
Normalise to MNI option.  In principle though, the deformations could be
composed with one that maps between the Dartel template and MNI space.


There is still much more to be done in respect to longitudinal
registration.  Dartel is based on aligning tissue class images, which
results in more accurate inter-subject alignment than does working with
the original intensities (Heckemann et al, NeuroImage, in press).

However, the best model for longitudinal registration is likely to be
different.  The current segmentation model assumes that brains consist
of only grey and white matter, but more tissue types (ie mixtures of
grey and white) would probably need to be modelled to achieve very
accurate longitudinal alignment of tissue class images.  The cortex is
only a few voxels thick, so identifying thickness differences of only a
few percent would need accuracies that are much less than the dimensions
of a voxel.  At this spatial scale, little implementational details can
make a big difference.

Best regards,
-John


On Tue, 2010-02-16 at 02:30 -0800, kambiz rakhshan wrote:
> Dear John and other users,
>  
> Thank you for your reply. Although I have carefully read the sections
> that you pointed to, I still have a problem since the result of that
> section is deformation not the flow field. I am doing Longitudinal
> dartel and I want to combine two warps.The result of the combination
> is going to be used in Normalize to MNI function in SPM8 as a flow
> field but if I understood correctly,the result of the compose function
> is deformation which can not be used as a flow field in Normalize to
> MNI function.right?
>  
> In short, I want to compose the flow field from within- subject DARTEL
> with flow field from between-subject DARTEL.The result should be in a
> format which could be used as a flow field for the final Normalize to
> MNI function!
> Any suggestion would be highly appreciated!
>  
> Thanks in advance!
> Kami
> --- On Fri, 2/12/10, John Ashburner <[log in to unmask]> wrote:
> 
>         
>         From: John Ashburner <[log in to unmask]>
>         Subject: Re: Composition of two Dartel's flow fiels
>         To: [log in to unmask]
>         Date: Friday, February 12, 2010, 11:29 AM
>         
>         Two deformation fields from Dartel can be composed, which
>         allows the
>         image of one subject to be warped to match the image of
>         another.  The
>         "Using DARTEL" chapter of the manual gives an example.
>         
>         The flow fields themselves are not be composed - only the
>         deformations
>         that are generated from them.  With a slightly different
>         formulation of
>         the registration, it may be possible to appropriately combine
>         initial
>         velocity fields using the Baker–Campbell–Hausdorff formula.  I
>         could be
>         wrong here though.
>         
>         Best regards,
>         -John
>         
>         On Fri, 2010-02-12 at 07:40 -0800, kambiz rakhshan wrote:
>         > Dear all,
>         >  
>         > Is there any function in SPM which only compose two flow
>         fields
>         > extracted from Dartel? I don't want to apply it to an image
>         in this
>         > stage, but instead later on in normalize to MNI as a flow
>         field?
>         >  
>         >  
>         >  
>         > Best
>         > Kami
>         > 
>         
>         
>         -- 
>         John Ashburner <[log in to unmask]>
>         
> 


-- 
John Ashburner <[log in to unmask]>