I am interested in looking at the resting-state component of the time series during task-based fMRI data. So for task-based fMRI, we are trying to find what parts of the time-series match the task design, and we are ignoring all of the residual error which in this case actually is the resting-state component of the time series. I understand how can we sort out task-related activation using proc.py’s regress function, but I was wondering if it is possible to investigate the residual error (non task-related time-series) portion of the time series using AFNI. If it is possible, could you please let me know how can we do that?
I’m looking forward to learning how to regress out the task-related time-series and to get the non-task related time-series using the task-related fMRI data. Thank you in advance for your help!
Hmmm, how much of a gap is there in your task data between events? I don’t know that one can treat the gaps between events (and this meaning the event itself, convolved with a reasonable HRF, which greatly extends the duration of what is meant by “event”) as rest. While resting state acquisitions are just somewhat unconstrained periods of brain activity (and many people complain about this), I think that is still quite different than being a tiny gap between stimulated events. Many people make the case that resting state scans should be >10 mins, say, to “settle down” into stability—this is very different than what might be at most a few seconds between stimuli, and most task presentations do not have large gaps between event blocks (even before considering the HRF, which again would further lengthen the duration of what one would term part of the “event” measurement).
It is true that there is a residual output by afni_proc.py, but I also don’t think this could be considered “the resting state component” of a task stimulus. Given the imperfection of GLM modeling, there would probably be too many relics of the task left in there. (And again, I can hear people asking how this would be different than a normal resting state scan, where there is audio stimulation from the scanner, visual stimulus from lights in the room, etc.—that still strikes me as a separate consideration).
So, you might be able to look at the residual and see how it looks across the brain, but that would probably not be comparable to a resting state scan. If anything, it would be closer to what people look at for measuring the smoothness of noise, for things like clustering.
A gap between events was 3 sec. I agree that the gap between the task can be treated as resting-state. I was searching for if the between-task time-series data could be meaningful and if it is possible to investigate this using AFNI. Thank you so much for your thoughtful answer. It really helped!
OK, 3 seconds between events would not be enough time to let the BOLD response settle back to some “baseline” after an event, anyways. I still don’t think it would be feasible to pull out rest moments between the tasks very easily (or in a way to mimic actual resting state very closely) without suuuuper long gaps.
You mention this: “I agree that the gap between the task can be treated as resting-state.”
… but I don’t actually think the gaps between the task could be treated as resting state.
Thank you for your thoughts. Also, thank you for the correction. Sorry, it was a typo - I meant to say “the gap between the task can’t be treated as resting-state”. Have a good weekend!
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.