Hi Guys,
When I add slice timing correction, Melodic gives me a bunch of components that
look like checker boards. That is, I have a 25 slice volume and slices 2,4,6,8 have a
lot of signal but 1,3,5,7 have very little. The opposite is true on the other side, i.e.
25,23,21,19 have lots of signal and 24, 22,20,18 have very little. The middle slices
12,13,14,15 have about the same amount.
My initial thought was that my slice order file was wrong. I verified with GE that the
scanner acquired bottom up/interleaved (i.e. 1,3,5,...,2,4,6,...). Nevertheless, I also
tried (25, 23, 21,...24,22,20), (2,4,6,...1,3,5), and a few other combinations. In all
cases, I got about 52 components. Very few had any structure to them; they looked
speckled. None had timecourses that looked like my design. Nearly all had this
checkerboard pattern to them.
I also tried running Melodic with s.t. correction turned off. I got only 14 components.
Much less of the checkerboard, but you can still see it if you look hard. None of the
maps had much structure or the right timecourse.
The experiment was a simple finger tapping task, 20s on, 20s off, for 5min. When I
ran FEAT, w/o s.t. correction, I got beautiful activation in motor cortex and
cerebellum, and virtually nothing else.
So here are the questions:
1. Is the slice timing correction being implemented incorrectly (i.e is my order file
way off or something like that)? Is this type of ICA map typical of slice timing
problems (i.e should I submit it to the little fMRI shop of horror)?
2. What does it mean that FEAT shows really nice activation in one area, but Melodic
doesn't find a map that matches this activation? I haven't seen any components that
look like the Melodic example data.
thanks,
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]
|