Next: Results
Up: FMRI data
Previous: FMRI data
We use two different datasets. The first is a boxcar visual
experiment with a reversing checkerboard boxcar stimulus (30
seconds on, 30 seconds off). The second is a single-event pain
experiment, for which the stimulus was a thermal noxious stimuli
of 3 seconds duration administered to the dorsum of the
volunteer's left hand using an electrical resistor to generate
heat with varying ISI (between seconds and seconds).
For both experiments echo planar images (EPI) were acquired using
a 3 Tesla system with TR=3 seconds, time to echo (TE) = 30ms,
in-plane resolution 4mm and slice thickness 7mm. The first 4 scans
were removed and the data was motion corrected using
MCFLIRT (Jenkinson et al., 2002) and high-pass filtered as described
in Woolrich et al. (2001). The data is not spatially smoothed.
We use an HRF resolution of (i.e. 0.5 seconds). We then
fit the model with an autoregressive order of and use a
basis set of the top 3 eigenHRFs shown in figure 3. We
use two different models, one with no HRF constraints ( and
) and one with HRF constraints (
and
as given in equation 17). We
use an f-contrast to pick out the linear combinations of the basis
functions. The resulting f-statistics are f-to-z converted to
produce spatial maps of pseudo-z-statistics. The fully adaptive
mixture modelling described in section 5 is then used
to provide probabilities of a voxel being activated.
Next: Results
Up: FMRI data
Previous: FMRI data