Dear SPM-ers and DCM-ers,
For a stochastic DCM analysis, I extract BOLD time courses from a resting state fMRI dataset using the VOI tool, adjusting for the effects of interest (a set of physiological regressors), thereby removing effects of motion and other nuisance regressors present in my design matrix.
My question is whether there is any low-pass filtering / temporal smoothing being performed on those time courses, either when extracting the time course (with the VOI tool) or afterwards in stochastic DCM.
If I understand it correctly, SPM implements a form of temporal smoothing at the first level using autoregressive modeling; AR(1). If I extract a time course while adjusting for the effects of interest, is this AR(1) also performed on that time course? Or later in stochastic DCM?
The reason for my question is that I noticed some high-frequency variability in the BOLD time courses extracted from my VOI, which appears (partly) non-neuronal. When plotting the observed vs predicted BOLD time courses from a DCM, it appears that these higher-frequent components are not modelled (ie not visible in the predicted time courses).
Thanks for your input!
Yours,
Rick Helmich
|