Try it using the segmentation.  It generally works a bit better.  See
the evaluations by Klein et al for a comparison of which normalisation
approaches seem to work best.

Best regards,
-John

On Fri, 2010-06-18 at 09:06 -0400, Xu, Ben (NIH/NINDS) [E] wrote:
> Hi,
>
> I'd appreciate it if anyone can give me some suggestions for the
> normalization problem I have. I'm trying to normalize subjects' T1
> images to the MNI template brain with SPM8. The T1 images were
> acquired with a high resolution (1x1x1 voxel size) MPRAGE sequence
> and are good quality. They have been coregistered to the subject's own
> EPI images. Although the normalized images looked generally ok
> toward more lateral side of the brain, they are clearly over
> stretched (about 5 millimeters or more) in the middle part of the
> brain, particularly in the low half and toward the posterior part of
> the brain (see attached images). As a result, part of the visual
> cortex, for example, in the normalized subject's T1 image is now
> mapped to the cerebellum in the template brain. I have tried to
> increase "non-linear regularization" to 10, 20, 30, or even higher. It
> didn't make much difference in correcting the over stretch. And this
> appeared to be true for other subjects' normalized T1 images, also.
> With lower resolution (1x1x4 voxel size) T1 images, the problem is
> even more pronounced. Are there any other default parameters that I
> should try to correct this problem?
>
> Thank you for your help.
>
> Ben
> --------------------------------
> Benjamin Xu, Ph.D.
> 7D51 Bldg 10 Center Dr.
> Human Cortical Physiology Section/NINDS
> The National Institutes of Health
> Bethesda, MD 20892
> Tel: 301-402-7968
> Fax: 301-480-2286
> Email: [log in to unmask]

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