Does anyone have any thoughts on analyzing PET data with FSL? The only postings I could
find regarding PET analysis are a year old. Here is what I am trying to do...
I have mutiple subjects, each with 12 PET scans. Of the 12, there are 2 replications of 6
tasks, with 1 of the 6 being baseline/rest. I modeled it as 6 EVs. The problem that I am
having is a rank deficiency of 2.98 10^-8, presumably because I am modeling rest. I am
interested in looking at many contrasts, task-rest and task-task, and then averaging across
subjects (how to best register the data being a separate issue). What is the most appropriate
way to model this kind of study?
Regarding the other FSL parameters, here is what I have chosen as being appropriate for PET
analysis:
Misc: threshold set at about 20-30%
Data: TR=1 (meaningless), Delete volumes=0, High pass filter=1 (meaningless)
Pres-stats: Slice timing=off, Mcflirt=off (I did motion correction outside of FEAT), BET=on,
Spatial smoothing=8mm, Intensity nomalization=on, Temporal filtering=off
Stats: FILM=off, Convolution=off, Apply temporal filtering=off, Othogonalise=off, Temporal
derivative=off
I chose Full Model Setup and used a separate design.txt file for each EV which contained 12
points (for each EV design.txt file there were two 1s and ten 0s).
Is this an appropriate design? Has anyone analyzed a PET study with a similar design before
in FSL?
Any comments would be appreciated. Thanks. --michael
|