Hi,
For ASL images, I always realign them before subtraction, and
for realignment, I just follow the general procedure, although Geoffrey
Aguirre recommends the separate realignment. But you should choose an
image in the middle rather than the first image as the reference volume
if you find heavy motion in the realignment time courses. According to
JJ Wang's paper, spatial smoothing is very important for ASL images. I'd
like to use 3x3x3 or 3x3x4 times of the voxel size for the normalized
images, for the non-normalized images, 3x3x2 is ok.
Regards!
Ze Wang
Center for functional neuroimaging,
School of Medicine,
Upenn
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
On Behalf Of Michael Hermes
Sent: Wednesday, July 13, 2005 6:00 AM
To: [log in to unmask]
Subject: [SPM] Perfusion MRI processing
Dear SPM community,
in an arterial spin labelling (ASL) perfusion MRI study, a sequence of
40 interleaved labelled and non-labelled images (label, non-label,
label, ...) is acquired and pairwise subtracted to get the perfusion
images. Averaging over the 40 images is done in order to increase SNR.
My question concerns the order of (pre-) processing steps, since there
doesn't seem to be consent in the ASL community.
Would it make more sense...
1) to first realign all of the 40 label and 40 non-label images
separately and then substract the two mean realigned images? Or...
2) to first subtract the label and non-label images and then realign the
40 difference images (incl. averaging)?
Furthermore, if you are doing a ROI analysis (extracting the mean
perfusion in a large ROI like the vascular region of the posterior
cerebral artery) instead of a voxel-based analysis (no random field
theory!), is it necessary to smooth the image? If so, what would be a
criterion for the size of the smoothing kernel, e.g. for a acquired
voxel size of [2 2 8], normalized to a voxel size of [2 2 2]?
Thanks in advance,
Michael Hermes
|