Hi - I am surprised by this. We do this regularly and, in general, it
works well.
Some notes -
1) We register FROM the T2-weighted b=0 images TO the T1-weighted scan.
2) We extract the brain from both images before we start (for exactly the
reason you suggest)
NB - bet does not always work well with the defaults with dwi reference
images, but you can pretty much always tune it by hand to make things
good - I start off with "-f 0.4" which tends to work better than default.
3) In general, mutual information cost function in flirt tends to work
best ( but not hugely better than default)
4) Remember that this is a affine registration, and there will be some
warping between the EPI and the structural image. The worse the warping,
the worse you can do with affine registration. This will not only affect
the registration in areas of bad distrotion, but throughout the brain, as
the global cost function will try to match the local distortion. Hence the
worse you distortion, the worse your registration will be generally. If
this is bad, you should think of collecting field maps to help
registration.
Hope this is useful,
Tim
-------------------------------------------------------------------------------
Tim Behrens
Centre for Functional MRI of the Brain
The John Radcliffe Hospital
Headley Way Oxford OX3 9DU
Oxford University
Work 01865 222782
Mobile 07980 884537
-------------------------------------------------------------------------------
On Fri, 13 Aug 2004, Hoang Tran wrote:
> Does anyone have any experience (good or bad) with registering a DTI image
> to a T1 image? The images have different spatial resolutions, but are
> oriented the same way.
>
> I've been trying to do so, but the results are terrible. I've tried the
> following methods so far:
>
> 1) Registered full head DTI (b=0) with full head T1 => poorly registered
> because DTI image does not have enough contrast on the skull and skin, i.e.
> those areas are relatively dark in the DTI relative to the brain. In the
> T1, that's not the case.
>
> 2) Registered brain DTI (b=0) with brain T1 (brains were extracted using
> BET with some post BET editing) => poorly registered but have no idea why
> it should be
>
> Any suggestions or insights would be much appreciated.
>
> Hoang Tran
> Neurobehavioral Research, Inc.
> 201 Tamal Vista Blvd.
> Corte Madera, California 94925
> 415.927.6627 (Office)
> 415-924-2903 (Fax)
> [log in to unmask]
>
|