Hi Jack, you're pretty much right in your conclusions, and many thanks for
pointing this out! I have been doing some similar analyses over the
weekend using MELODIC to quantitate whether STC changes intrinsic
dimensionality, and yes ICA turns out to be a very useful tool for picking
up these problems. In fact, FSL in general assumes that slices were taken
bottom up; however, yes, all the slicetimer documentation/bubble-help is
misleading about the order it is doing things; we will make this much
clearer in the next release of FSL (3.1, _hopefully_ within a couple of
months). For the moment, yes, people need to select the opposite slice
timing ordering than is actually the case - sorry!
BTW, I should point out that slice-timing-correction is not in general
necessarily a helpful thing to do; the slight loss of temporal information
(because of the necessary interpolation used to achieve the time shift)
may outweigh the advantage of having perfect "slice timing". Furthermore,
it is not very satisfying to have to do the timing correction as a
separate stage from the motion correction; integrating these is something
which we are working on for the future.
Many thanks, Steve.
ps - given all the above, your suggested reverse order is not quite right
- it should be 24,22,20....25,23,21......when I used this on your data the
dimensionality didn't then increase.
On Fri, 28 Feb 2003, Jack Grinband wrote:
> Hi Steve,
> I figured out the checkerboard problem. FSL assumes that the slices are acquired
> from top to bottom. All three of the scanners that I looked at acquired bottom up.
> Consequently, 1,3,5,..2,4,6,.. should really have been 25,23,21,...24,22,20,....
>
> I couldn't find any info in the manual regarding which slice, top or bottom, is the "first"
> slice. It might be a good idea to add this to the next version of the manual (or even
> make the user specify directly within FEAT) so others don't make the same mistake I
> did.
>
> Anyway, one of the things I did was to compare the STC in FSL with STC in SPM.
> When I ran Melodic, I got the following results:
> No STC: 10 components
> STC SPM: 10 components
> STC FSL: 19 components
>
> When I compared the maps whose timecourse matched my model, I found that both
> STC in FSL and in SPM gave me almost identical maps and both maps had
> smoother timecourses and more localized activation. This component was the 7th
> map in the No STC and STC SPM conditions but the 9th map in STC FSL.
>
> So, my question is why are there so many more maps using the FSL STC? My
> feeling is that there should be the same number or fewer components after STC.
> Thanks,
>
> jack
>
Stephen M. Smith MA DPhil CEng MIEE
Associate Director, FMRIB and Analysis Research Coordinator
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|