If I have a participant with 2 scan sessions (a few weeks apart). One of the anatomicals is unusable. What is the best way to run it with fmriprep?
Is there good way of utilizing anat from one session to save the funcs from another?
How would fmriprep behave if one of the sessions is just missing the anatomical because I deleted it?
In general, is there a recommended way of dealing with rejected data?
As you hinted, removing it.
fMRIPrep will merge all the T1w images it finds across sessions. At least one T1w must be present for fMRIPrep to work. Depending on co-registration, fMRIPrep may drop further scans. But, if you know that scan should be excluded, please do so before running fMRIPrep.
No that I know of. However, there’ve been discussions under the BIDS specification umbrella about how to flag images that should be excluded, but that did not fly.
That’s good to know. Can that interact with the longitudinal flag/analysis if that is being used?
I wonder if there can be a future version of BIDS where we can flag/organize rejected data without deleting it altogether.
Potentially yes, but do you want to consider your T1w data longitudinal?
That will need someone to propose it and submit a PR to the bids-specification repo. If you were to push on that, this conversation will show you what are the potential argument in favor/against such new semantics.
The reason for longitudinal is that in some cases I have 2 sessions close together in time and then 2 more a couple of years later. one of the 4 sessions has a bad anat.
In other cases I have only 2 sessions (close in time) and one anat is bad. In which case, this can be analyzed without the longitudinal flag.
I meant “longitudinal” in an anatomical sense. In other words, do you expect substantial anatomical changes between sessions? By substantial I mean, e.g., a surgery or images were acquired 20 years apart.
Got it. Some participants are a developmental sample. So in principal a couple of years can make an anatomical difference.
I can probably run different subs with different flags in a way that makes most sense for the individual in question.
as little as 7 during first scan sessions.
I guess then you should treat sessions as different subjects in that case.
That’s interesting. Howcome?
Because for fMRIPrep, the anatomy is relevant to place things around. When there are no big anatomical changes, we average the T1w images across sessions to remove noise. Treating sessions as individual participants will help registration steps to be more accurate. If anatomy changes a lot between sessions, the averaging is not going to do any good.