On Monday 06 May 2002 17:17, Evelyn Eger wrote:
> Dear SPMers,
>
> I did a fMRI study where I acquired 18  slices (3 mm thickness) parallel
> to the line between occipital and temporal pole. When I try to normalise
> these volumes (to the EPI-template), the nonlinear transformations do
> not work (I get the message: The field of view is too small to attempt
> nonlinear registration).

SPM99 requires at least a 6cm field of view.  This is coded around line 759
of spm_sn3d.m:

        if any(fov<60),
                warning('The field of view is too small to attempt nonlinear registration\n');
                params(1:4)=0;
        end;

>
> Now I want to coregister the EPI-images to the stucturals instead
> (target - T1, object - EPI), then normalise the structurals to the T1
> template and apply normalisation parameters to the EPIs. When I reslice
> the images during coregistration as is default, they become very large
> (21 MB) because of the higher resolution of the structural images. Do I
> have to write the coregistered images at this stage at all? I thought it
> is maybe enough to have the transformation parameters in the .mat file,
> but I am not sure.

No.  It is preferable to use the "coregister only" option.  The actual resliced
images are not really useful for estimating warps from, because they may contain
regions of missing data (filled with zeros).

I would generally suggest having the T1 as the "object"/"source" image, and
one of the EPIs as the target.  This way, only the .mat file of a single image
needs to be changed.

>
> Is there maybe another (better) way in my case to get more accurately
> normalised images for a group analysis?

There are a lot of problems with distortions and dropout of EPI images
(see spm archives).  If warps are to be estimated from a T1 image and
applied to a series of EPI images, then some kind of distortion correction
should ideally be applied first.

The approach you would take depends on the severity of the dropout and distortion.
If the EPI images are highly distorted, with relatively little dropout, then
matching an EPI to an EPI template is preferable.

If the dropout is severe, then you would ideally need to spatially normalise via
an extra image.  If you can acquire an image with the same distortion as the EPI,
a good field of view, but without the droupout, then this would probably be the
ideal extra image.

A general comment about spatial normalisation of T1 images is that it generally
works best for skull stripped (scalp edited) images, matching them to a skull
stripped template.  The objective of most functional imaging studies is to match
grey matter, so I would suggest that for spatially normalising good quality T1
weighted images, one could probably get best results by first segmenting the
images and estimating the warps by matching _seg1.img to a grey matter template
(first disabling "brain masking" via the <Defaults> button.).


Best regards,
-John

--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491  or  +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420  http://www.fil.ion.ucl.ac.uk/~john