Dear Johannes,
If you have lesions in the image, then you may consider trying some sort of
lesion masking. There is a NeuroImage paper by Matthew Brett et al that
could help. Other things to try involve changing the defaults for estimating
spatial normalisation parameters - increasing the regularisation or reducing
the number of basis functions.
Best regards,
-John
On Wednesday 02 November 2005 18:16, Johannes Rüter wrote:
> Dear SPM Users,
>
> I'm pretty new to SPM and have encountered a strange error while trying to
> normalize my data.
>
> I have data from a single subject with a very large bilateral medial
> occipito-temporal lesion. Unfortunately this whole brain scan consist of
> only twenty transversal T2-weighted slices. After converting my images
> into analyze images using MRIcro, I designate the large regions of
> interest, export them into analyze format as well and feeds everything
> into SPM2.
> Doing the normalization step by step, applying a basically default
> normalization (linear transformation only, 2nd degree B-spline
> interpolation, a slightly larger bounding box), the result is quite good
> at first glance. Taking a closer look it becomes obvious, that on the one
> hand the head has been clockwisely tilted by nearly 45 degree, and one the
> other hand that the ROI distribution has changed significant (e.g. into
> the cerebellum), probably a result of the bad fit. Orienting the source
> image onto the bicommissural line does not help, nor does tilting the
> source image counterclockwise from there.
> Which parameters should I check (and/or change) to get rid of the tilt in
> the normalization process and thereby getting a better fit to the
> template? Are there any methods or parameters to deal with these extended
> lesions?
>
> Thanks in advance for the advice
>
> Johannes
|