Getting Relevant fits from afni_proc/3dREMLfit

I’m trying to sort out some interesting behavior in my fMRI data and I’d like to look at just the “fit of interest”: what AfNI thinks the Hemodynamic response should look like, without any of the polynomials to remove motion, noise that changes over time, etc. In other words, I’d like to get the model that is used to select the beta value in the “#RUN_Coeff” BRIK/HEAD pair. Is there a way to get that out of the fitts file or with additional options to 3dREMLfit? Sorry if this is an unclear or poorly-founded question.
Thanks so much in advance,

Hi Josh,

Note that if you are using GAM, BLOCK, SPMG*, or similar, then you are actually telling 3dREMLfit what the HRF is supposed to look like. The cases where the program is figuring that out are with multiple basis functions, an particularly in the case of TENT.

And in order for the program to get a useful estimate of what the HRF looks like, it is very important to model drift and motion effects, or they will corrupt the very estimates that you seek. It does not seem reasonable to remove those model parameters.

Maybe there is an aspect I am missing. Please feel free to provide more details and possibly even a model (e.g. 3dDeconvolve command).

  • rick

Thank you for your response. Sorry, I misphrased my question. I don’t really want just the HRF-- I would like the anticipated BOLD response (essentially, an HRF convolved with event delta-functions and scaled by a beta-weight) and a “cleaned” time series for each brain voxel. Presumably I can generate the BOLD response with an AfNI utility, but I had hoped it might be stored somewhere in afni_proc’s output (the model is wrapped in afni_proc rather than run separately). More importantly, though, I would like to view the “cleaned” data. One idea proposed in the lab was to model the nuisance regressors only, and then take the residuals as “cleaned” data-- is this a good idea, or no? Is there perhaps a better way to do it?

Oh, then you probably want to use 3dSynthesize. But in order to use it, the prior 3dDeconvolve command would need the -cbucket option included.

Add “-regress_make_cbucket yes” to the command. It would then include an appropriate -cbucket option in 3dDeconvolve. So after the proc script is finished running, you would be able to use 3dSynthesize to extract whatever you want.

Pleas see 3dSynthesize -help for details.

  • rick