Dear community,
I would be happy to get some comments and advice concerning our analysis of
the following fMRI paradigm.
Subjects processed a mental rotation task. Each trial consisted of the
following three steps/“conditions”:
A) Presentation of a geometrical object (duration 3 seconds)
B) Mental Rotation of the stimulus (duration 5 seconds), no visual input
during this period
C) Presentation of either a mirror-reversed or a matching version of the
geometrical object presented in A) (duration 2 seconds)
The task of the subjects was to decide whether the object presented in A)
was identical to the object presented in C). This had to be done within the
last 2 seconds. Each trial was preceded and followed by the presentation of
a fixation cross (so to say “the baseline period”).
There was no temporal gap between the three steps/conditions of a trial.
The intertrial interval was randomly varied having a mean around 10.5 s. TR
was 1.007 s, 18 EPI slices were acquired, 488 volumes/session, 24
trials/session.
The main question we are interested in is whether brain activation during
step/condition B is higher than during A. To this end, we chose the
following approach (after standard preprocessing) using the fMRI models
setup option in SPM99:
The model consists of 3 predictors (the three conditions, present-rotate-
match/respond), each with variable stimulus onset times and durations
(because of the random ITI).
For each condition, we specified the exact onset and duration (in TR units)
The three conditions were modeled as events convolved with the canonical
hrf.
Low-pass and high-pass filtering were also used/incorporated.
After model estimation, we used a t-contrast (–1 1 0) to identify voxels
with higher activation during the rotation condition.
Basically, I’d like to know whether you think this is – in general – a
correct approach, or whether anybody comes up with a better or more
elaborated idea. More specifically, the main thing I am concerned about is
the rapid succession of different “events” (i.e. conditions) and
potential “carry over effects” this might cause. For example, I’d like to
know whether it would make sense (and why) to add the temporal derivative
to our model, whether I should use (and why) Volterra interactions (up to
now, I did not) etc.
Any help, comments, criticism…. will be gratefully appreciated,
-Claus
|