Hi Markus
>
> 1) I set up a model with only one condition sound that gets an
> 1^st order parametric modulation intensity. Further I include 6
> movement-regressors. The Stimulus duration is 5 seconds, the TR is
> 7.7. seconds (Block design). Despite the gap of 2.7 seconds where no
> stimuli is presented the entire regressor Sn(1) Stimulus*bf(1) (see
> attached Design Matrix.jpg) is continuously positive without any gap. Why?
>
if sounds are presented every 2.7 sec, after convolution there will be
no gap -
>
> 2) Next I want to take a look at the results. I defined a F-Contrast
> with ones at the position of the linear regressor (Contrast vector is:
> repmat([0 1 0 0 0 0 0 0],[1 4]), No. 1: offset of linear regression
> No. 2: Gradient off linear regression No. 3-8: 6 movement
> parameters). The attached brain-glass-view (eigenvariate.jpg) should
> show those areas that respond linear to the intensity of the stimulus.
> From this point of view the brain glass view looks very fine because
> typical auditory areas show high F-Values. Now I plot the response of
> the global maximum at -40,22,6 (the cluster F value is 144) with an
> eigenvariate analysis of that region (VOI-sphere with 0 mm radius).
> The response shows values between -2 and 2 as can be seen at the lower
> right corner in (eigenvariate.jpg) . 100% of variance is explained. My
> understanding is that the voxel response shows an 100%
> correlation/agreement with the intensity of the stimulus - high pass
> filtered and adjusted for movement related activation. Is this right?
>
funny I never noticed the % of variance stuff .. would it be the %
explained by the 1st eigen variate (since it is computed across voxels
and you have only 1 voxel = 100%)
>
> 3) But (!), if I set up a parameter plot of the response in that
> region (session 1) - I get a very strange result. This can be seen in
> the figure Parametric response.jpg in the lower right corner. The
> values decrease linear from 0.0687 to 0.0686. So there isnt any
> change at all at that voxel a voxel with the global maximum in the
> F-Contrast (and its decreasing, wrong direction) not at 16,4s and
> not at other times.
>
> a. Why
>
> b. How can plot the response at other times?
>
> c. How is the parametric response exactly calculated? from the
> corresponding beta-image?
>
not you used an F test so the map is for increases as well as decreases
- I'm not sure why this is like that, but you must have some effects in
other sessions otherwise the F test would not have give this area as
activated - you use plot / scan or time / - for the parametric session
per session ; data for the mesh are in the workspace - the beta images
are the reg coef of each regressor
cyril
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|