Dear Agnieszka,
Your strategy looks fine for ascending/descending images. Interleaved
images are slightly trickier, and nobody really agrees on a good
solution for processing these. Karl says it should be done in the
statistical analysis by including derivatives of the HRF in the model,
whereas others give other recommendations.
If slice-timing correction has to be used for your interleaved data,
then it would only really make sense to do it before the realignment. I
am reluctant to say anything else, given that I have not seen the
required empirical evidence that would be necessary to make an informed
choice.
My own view is that a combined timing/motion/GLM generative model is the
only really correct answer, but this would require a serious amount of
work to develop and is unlikely to appear in SPM any time soon.
Segmentation is generally preferred for estimating spatial normalization
parameters in SPM5, because it gives better results than the spatial
normalization itself. The reason that it was not recommended for SPM2
is that the SPM2 version of Segmentation wasn't combined with any
warping model.
Best regards,
-John
-----Original Message-----
From: A.A.Reid [mailto:[log in to unmask]]
Sent: Monday, June 04, 2007 1:42 PM
To: Ashburner John (PSYCHOLOGY)
Subject: the order of the preprocessing steps - couple of questions
Hi John,
I just have a quick couple of question about the order of the
pre-processing fMRI data with SPM5. I have been following the example of
Face Data from SPM5 manual. The order of the pre-processing steps is as
follows:
1) Realignment
2) Slice Timing Correction
3) Co-registration
4) Segmentation
5) Normalise
6) Smoothing
This order is used for a continuous EPI of 24 descending slices; I am
trying to analyse a sequence of 44 slices which were acquired in
interleaved fashion. Would it be better in my situation to do the Slice
Timing Correction in step 1 and Realignment in step 2, to avoid the
danger
of shifting voxels to adjacent slices (and hence to different time
points)? If yes, would I need to change anything else in this
order/processing?
What advantages does one gain by doing Segmentation? In the same example
for Face Data, but processed using SPM2 no Segmentation was applied?
I would be grateful for a fast answer. Thank you for your help.
Agnieszka
|