> Using the option to write spatially normalized grey matter files as part of
> the new Segment procedure, I notice that the voxel size and bounding box
> will always be derived from the template. I think it would be useful to
> have these as user-selectable. For example, most VBM studies use 1x1x1 mm
> voxel sizes but the template is 2x2x2 mm.
The reason for the higher resolution for most VBM studies is to reduce the
effects of partial volume. For the new code, there is no need, as the images
themselves are not warped. It is the tissue probability maps that are warped
to match the images, so there is no interpolation of the data. You will also
note that the wc*.img and mwc*.img images are slightly smooth. This is
because the native space tissue class images are smoothed slightly in order
to reduce any potential aliasing effects.
>
> At line 151 of spm_preproc_write.m, the structure "so" is created with the
> vox and bb fields set to NaN; so changes would need to be made there.
> Perhaps fields relating to writing spatial normalization files could be
> added to the "opts" field in the "jobs{.}.spatial{.}.preproc" structure.
It would have been a possibility, but I chose not to implement these options.
If you want spatially normalised images at a different resolution, voxel sizes
etc, then you could use the Normalise button to apply the normalisation
parameters to the native space tissue classes. I didn't include these
various options, as I figured that having the same bounding box and voxel
sizes for all spatially normalised tissue classes would reduce the chance of
people mixing them up in the same study, and wondering why the stats didn't
work.
Best regards,
-John
|