Hi Jack,
> mcflirt -in func_BET_del4 -out prefiltered_func_data_mcf -plots -refvol 61 -rmsrel -rmsabs
> What are the -rmsrel and -rmsabs options? I couldn't find any reference to them in
> the manual.
these are simply telling mcflirt to output the summary statistics about
overall estimated motion, so yes they don't affect the output image data.
> I did the following experiment on 6 different data sets:
> 1. take func_data and run bet
> 2. delete first 4 volumes using avwsplit and avwmerge
a couple of comments here - you should NOT be running bet BEFORE motion
correction - the act of brain extraction will probably bias the motion
estimation to underestimate the motion. secondly, you can more easily
delete the first 4 volumes by using avwroi instead.
> then one of three operations:
> 3a. command line mcflirt (default options) and Melodic (MCFLIRT, BET & STC = "off",
> delete 0 vols).
> 3b. Melodic (MCFLIRT="on", BET & STC = "off", delete 0 vols).
> 3c. FEAT (prestats only, MCFLIRT="on", BET & STC = "off", FWHM=0, highpass="off",
> delete 0 vols) and Melodic (MCFLIRT, BET, & STC = "off", delete 0 vols).
>
> 3a and 3b gave me the same answer every time. 3c gave approximately the same
> number of ICs, sometimes 1 more or 1 less or equal. However the components
> themselves were different. For example, in one data set, a component which I
> believe was visual cortex activation was component 4 in 3a/3b but component 18 in
> 3c. The timecourses of all the components were different (although the amount of
> difference varied by component). Also, some components were present in one
> analysis but not in the other.
>
> The only other thing I can think of is that the "High pass filter cutoff" = 100s in my
> FEAT analysis. But it shouldn't be applied since I'm only doing prestats, and in any
> case, I turned temporal filtering off. What am I doing wrong?
probably the small difference is due to the fact that the FEAT pre-stats
will also be carrying out grand-mean intensity normalisation (which
happens even if you turn off "intensity normalisation", which also
includes a simple thresholding step as well. you won't be getting this
step in 3a and 3b, so that probably explains things.
Note that Melodic does not necessarily give the same answer every time
anyway, as the unmixing matrix is initialised randomly.
ttfn :)
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
|