Thanks Alex.
I'll just elaborate a little more on the first part of the query though, and hope
that someone else will answer the third.
| 1. When I coregistrate our high-resolution anatomical data to the
| functional images, the realigned anatomicals are written with the
| resolution of the functionals. Is this inevitable? Can the resolution of
| the anatomical images be preserved within the realignment to the
| functional? How do I do this? I have not found the opportunity at the
| defaults bottom, so maybe it would need some editing in the code. I
| guess, another option would be to write the functionals with a higher
| resolution but I have no clue how to do this before normalization and I
| donīt know whether this is accurate or may inflate the statistics.
|
| | 1) If you coregister your structural to your functional image then I think that
| | it is inevitable that the voxel size of the coregistered structural image will
| | be the same as the functional image. I can think of two possible ways around the
| | problem, both assume that you have normalized your individual subjects'
| | functional data:
| |
| | i) Coregister the structural image to the (mean) functional image but coregister
| | only (do not reslice- which produces an r* image). Then you normalize the
| | structural image using the *sn3d.mat file generated from normalizing your mean
| | functional image to the (presumably) EPI template. The structural image is
| | resliced once only, giving a n*structural image rather than an nr*structural
| | image. As you have not resliced the structural image, I believe that it should
| | have the resolution of the MNI template (2x2x2mm).
| |
| | ii) Another option would be to not coregister the structural image to the mean
| | functional image, but to normalize it separately to the (presumably) T1
| | template. If both normalizations work well, then both you functional and your
| | structural data will be in MNI space.
| |
| | I hope that others will comment, especially if I have misled you here.
If your objective is ultimately to superimpose your functional activations on to
your high resolution structural image, then the route I would take would be:
1) Without spatial normalisation:
Coregister the structural image to the mean functional.
Do the stats on the realigned functional images.
Simply display the activations on the structural image (not the resliced
version).
This works because the .mat file that is written for the structural image encodes
the relative positioning of the structural image relative to the mean functional image.
You can check that this has worked with the <Check Reg> button, and selecting the
unresliced high res structural and the mean (or any of the individual functional images).
2) With spatial normalisation:
Coregister the structural image to the mean functional.
Estimate spatial normalisation parameters either from the mean functional or the
structural image.
Apply the spatial normalisation parameters to the functional images to get spatially
normalised functional images with whatever resolution you like.
Hit the <Defaults> button, select spatial normalisation, then the option for
changing the defaults on how the images are written, and specify something like
1x1x1mm resolution.
Write the spatially normalised structural image using the new default voxel sizes
(selecting the original structural image, rather than the resliced one).
The .mat files created by the coregistration (or realignment for that matter) are
incorporated into the affine part of the spatial normalisation procedure. Processing
the structural image in this way means that it is not resampled down to the resolution
of the functional images at any stage.
Best regards,
-John
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|