selxavg-sess vs selxavg3-sess - Freesurfer

9 Nov 2010


      Hi all,
I'm still trying to figure out what's going on with the differences I'm
getting between selxavg-sess and selxavg3-sess.  I"m using freesurfer4.5. I
was using selxavg3-sess (as suggested on the wiki) but I was getting weird
activation maps when I compared any condition vs. baseline, where most of
the brain was more activity for baseline (which was fixation) - see the
attached picture for act-fix-gamma-selxavg3-rh.png.   So I ran selxavg-sess
which was what I used to run with an older version of freesurfer, and when I
did this I got a more appropriate picture - see act-fix-gamma-rh.png
The experiment starts with fixation (baseline) so I tried running the
analysis with and without skipping the first 2 TRs to account for the BOLD
spike, but it makes no difference in the activation map.
These are essentially the commands I ran:
mkanalysis-sess.new -analysis loc_loc -TR 2 -paradigm loc.dat -designtype
blocked -funcstem fmc -runlistfile loc_runs.txt -inorm -nconditions 2
-timewindow 20 -gammafit 2.25 1.25
mkcontrast-sess -analysis loc_loc -contrast shape_vs_fix -a 2 -c 0
selxavg3-sess -s 101006SP_loc_loc -df loc_loc.dir -analysis loc_loc
Thoughts?  Or should I just not use selxavg3-sess?
Katie
---------- Forwarded message ----------
From: Katie Bettencourt kcrum@bu.edu
Date: Tue, Nov 2, 2010 at 2:31 PM
Subject: Questions about event related processing and selxavg-sess and
selxavg3-sess
To: freesurfer maillist freesurfer@nmr.mgh.harvard.edu
Hi all,
I've been trying to analyses both an event related and block design
experiments and have noticed a couple things that are confusing me and I
would appreciate any help anyone could give me.
1.  In trying do the event-related analysis, I'm am a bit confused by the
FIR vs Gamma analysis and the use of the time window and tprestim in the FIR
analysis.  I have an experiment with 6 set size conditions and 1 fixation
condition.  each trial is 6s long with a TR of 1.5
I have run both of these commands:
mkanalysis-sess.new -analysis supIPS_loc -TR 1.5 -paradigm supIPS.dat
-designtype event-related -funcstem fmc -motioncor -runlistfile
supIPSruns.txt -inorm -nskip 2 -nconditions 6 -gammafit 2.25 1.25
mkanalysis-sess.new -analysis supIPS_loc -TR 1.5 -paradigm supIPS.dat
-designtype event-related -funcstem fmc -motioncor -runlistfile
supIPSruns.txt -inorm -nskip 2 -nconditions 6 -tprestim 2 -timewindow 20
The FIR gives me a time course window, which makes me think that perhaps
that is the one I want to use, but it shows very little activation.  This
data has been previously analyzed in Brain Voyager, so I know it's not a
case of the conditions not actually causing activation, but for some reason,
the FIR analysis doesn't show any.  I'm not sure if this is due to a bad
time window/tprestim settings, or something else.  However the gamma fit
analysis shows a bunch of activity where I expect to see it, but no time
course window.  (attached are pngs of the differences for the
act_vs_fixation comparison).
2.  In addition, in both the event related (gamma) and a normal block design
(different experiment) I get very different results (at least for certain
comparisons) depending on whether I use selxavg-sess (with stxgrinder-sess
for each contrast) or selxavg3-sess.  The differences are shown in the
attached pngs (act_vs_fix-gamma.png (which is the selxavg-sess and
stxgrinder-sess analysis and act_vs_fix-gamma-selxavg3).  In the block
design, both selxavg3 and selxavg (+stxgrinder) give me very similar
activation when comparing two non-null conditions (ie. shape_vs_noise) but
very different (and similar to the differences in the attached pictures)
activations when comparing against the null (ie shape_vs_fix or
noise_vs_fix).  What am I doing wrong here?
Thanks for your help!
Katie