Steve, Jim,
Thanks for the replies!
Jim, I did verify the slice order with GE. They assured me that interleaved is the
default with the new Signa TwinSpeed system. But I ran a phantom anyway with
contiguous slices as well as slices with 0.5 and 1.5mm inter-slice separation. I did
not see any significant differences between them in the ICA maps (no STC). So my
guess is that the acquisition is in fact interleaved and there's probably no cross talk
due to excitation of adjacent slices. Moreover, the problem is not really intensity
scaling of the whole slice, but rather, the intensity and shape of the ICA maps when I
use STC. I was also told that the pulses are evenly distributed across the TR.
Steve, with regard to motion artifacts, I don't know if motion is a serious problem
here. I used a bite bar in these experiments and both the absolute and relative
mean displacement is around .1 and .05 mm, respectively. Our voxel size is 1.5 x
1.5 x 4.5 mm.
I ran my finger tapping data with
(a) no STC
(b) STC=UP
(c) STC=DOWN
(d) STC=interleaved up 1,3,5,...2,4,6,...
(f) several other interleaved ST order files.
My results were:
(a) 19 ICA components. Mild checkerboards.
(b) 49 ICA components. Mild checkerboards.
(c) 45 ICA components. Mild checkerboards.
(d) 59 ICA components. Strong checkerboards.
(e) over 50 ICA components. Strong checkerboards.
A few other relevant facts: the Philips scanner next door shows the same behavior,
however it shows serious checkerboarding even without STC. It also does
interleaved acquisition.
It seems to me that if there is some temporally correlated noise which is only present
for the first half of the TR (e.g. a head movement that lasts for 2 sec but TR=4 sec),
then you should get a checkerboard pattern in some of your IC components. On the
other hand, shouldn't STC reduce the number of ICA components such that there is
only one or two components that account for that noise? In other words, why do I get
59 components when I do STC but only 19 without? I ran another subject and found
the same type of result, 14 ICA components without STC and 52 ICA components
with STC.
I put the data for the finger tapping (~40 Megs), slice order file, and design.fsf on our
website. It's all in www.fmri.org/jgdata/finger.htm. I would really appreciate if you
could take a look. Thanks a lot.
cheers,
jack
--
Jack Grinband, PhD
Mind & Brain Institute
Columbia University
518 PI Annex W. 168th St.
NY, NY 10032
212-543-6931 ext 306
[log in to unmask]
|