Dear Michael,
you can treat CBF perfusion estimates as independent measurements,
using an ANOVA to model each session you took.
In the literature, several authors have shown that the CBF perfusion
estimates are not temporally correlated, at least to a reasonable
approximation (you can have a look at a paper by Aguirre et al.,
Neuroimage 15:488-500, and at one by Mumford et al., Neuroimage
33:103-114). In my data, I have a very small negative correlation
between consecutive CBF acquisitions. It is really small however, and
it appears to tally well with the data presented in Mumford et al.
If you use SPM to model your data at first level, you should be aware
that SPM does a lot more than just estimating an autoregressive
covariance (which can be swicthed off when specifying the model). On
top of that, data are temporally filtered, for example, when using the
fMRI module.
Best wishes,
Roberto Viviani
Quoting Michael Froelich <[log in to unmask]>:
> Dear SPMers:
>
> We are working on continuous spin labeling data where we are
> acquiring sessions during which a task is applied.
> There is a session of a baseline resting state and three sessions
> during which a task is applied.
> Each sessions consists of 30 volumes (since in cont. spin labelling
> the TR is very long).
>
> Is it appropriate to concatenate the four session into one and then
> make up the contrasts by assigning onset and duration times that
> correspond to the appropriate position in the concatenated series of
> volumes, as if we had one very long session.
>
> This approach would of course be problematic if SPM used a special
> covariance structure like an autoregressive covariance but it
> probably would not matter if SPM was using a compound symmetry
> covariance structure. I am not sure what's "under the hood".
>
> Any thoughts on this?
>
> Regards,
>
> Michael Froelich
> University of Alabama
>
|