> > * spm doesn't seem to want to do analysis over two sessions if I only
> > preprocess the data up to the sliceTiming stage (no normalization nor
> > smoothing). I guess that's understandable as my images don't have the exact
> > same orientation, but is there a way to do it?
>
> You can use the Realign->ResliceOnly option to reslice all images to have the
> same orientations etc.
I ran dicom import, realignment (motion correction) and slice timing
correction for each session separately already. Do you mean I now run
the realignment routine with the reslice only option, with as input
the slice timing corrected images from each session? or should I skip
the first insession realignment altogether and do the realignment with
images from all sessions?
> > * I coregistered the functional images from both sessions to an anatomical
> > from one of the two sessions. When I run CheckReg, it looks very well
> > coregistered, so I'm happy... until I run normalization, after which my
> > images don't match very well any more. Is this expected, if not what am I
> > doing wrong?
>
> I'm guessing that you may have coregistered one of the functional images,
> without telling SPM that the rest of the images in the series should be moved
> in the same way (specify these files as "Other"). For cases like this, it is
> helpful to use the CheckReg button with a selection of the images, both
> before and after spatial normalisation. This should indicate which images
> are in alignment with which other images, and should give pointers about what
> may have gone wrong.
No, I coregistered the mean functional to the anatomical in each case,
by wiggling the mean functional around (source image), and I asked spm
to keep my slice timing corrected images coregistered with the mean
functional image. However, your first point (across sessions
realignment) might do the trick, by better matching the orientations
compared to my coregistration of both sessions to the same anatomical
image. Although, my images seemed pretty well coregistered before
normalization, and the coregistration algorithm and the realignment
algorithm are pretty similar, aren't they?
Thanks for your advice.
- Julien
|