I would like to use deconvolved time series from selected ROIs as inputs for functional or effective connectivity analysis. I know 3dDeconvolve can be used to calculate the deconvolution of a 3D+time dataset with a specified input stimulus time series. I also know the option “-fitts” can be used to output a dataset which contains the full-model time series fit to the input data. But my task paradigm contains two stimuli, and I want to separate them. Should I use 3dSynthesize to synthesize a fit dataset based on the two stimuli separately? Do you think that it makes sense to synthesize a fit dataset using one stimulus response for the later connectivity analysis?
Thanks for your advice in advance.
my task paradigm contains two stimuli, and I want to separate them.
You could, for example, use 3dSynthesize to combine the average response based on your presumed HDR for each stimulus type plus confounds such as drifting, head motion effects, etc, and subtract it from the original time series using 3dcalc. However, there are some big caveats. Basically there is no perfect/effective way to truly separate the BOLD signal between two stimuli for a couple of reasons: 1) the typically presumed HDR function is a crude approximation, and 2) the estimated response is an average across all the trials (i.e., the real BOLD response varies across trials).
Thanks for your advice. Owing to the big caveats, would you more recommend to use the output from “-fitts” as the input for the later connectivity analysis? I will extract the time series of the “fitts” dataset from the selected ROIs for the connectivity analysis. If that’s the case, is it not possible to see the differences between the two stimuli in the connectivity results?
Thanks again for your help.
I will extract the time series of the “fitts” dataset from the selected ROIs for the connectivity analysis.
The output from -fitts is just the combined effects from all the regressors in the model.
If that’s the case, is it not possible to see the differences between the two stimuli in the connectivity results?
It’s not clear to me what kind of “connectivity analysis” you’re performing. Ideally it would be better to do everything with the original data with as little data manipulation as possible.
We are going to use Granger causality to do the effective connectivity analysis. Some papers suggested that hemodynamic deconvolution is used as a preprocessing step before Granger causality analysis to remove the effect of hemodynamic response. The papers said that hemodynamic deconvolution removes the inter-subject and inter-regional variability of the HRF, as well as its smoothing effect, thus increasing the effective resolution of the signal. The underlying neuronal variables are obtained by applying hemodynamic deconvolution, which are then input into a dynamic MVAR model to obtain condition-specific connectivity values. So, which output from 3dDeconvolve should be used as an input for Granger causality analysis?
which output from 3dDeconvolve should be used as an input for Granger causality analysis?
The closest thing that is related to your description is the approach I mentioned in my first message in this thread: use 3dSynthesize followed by 3dcalc.