If you do resting-state fMRI and you do any sort of functional connectivity analysis, you should probably read this new paper from Blaise Frederick:
https://www.nature.com/articles/s41562-024-01908-6
I've been banging the drum on systemic LFOs for some time. Here's another example of how not properly thinking through the physiology of the entire human can produce misleading changes in so-called FC in the fMRI data. That said, I don't think Blaise has the full story here, either. For one thing, the big dips in his Fig 1b suggest that something is being partially offset with the on-resonance adjustment that is conducted automatically at the start of each EPI time series, so I have a residual concern that there are magnetic susceptibility effects contributing here somewhere. (Perhaps the magnetic susceptibility effects are what's left to drift higher after RIPTiDe correction, as in Fig 6b, for example.) The point is that not having independent measures of things like arousal, or proper models of physiologic noise components like sLFOs, or a full understanding of what's happening in the scanner hardware (including head support) during the experiment can lead to an assumption that things are neural when there are better explanations available.
---
Link added on 6/23/2024: Blaise Frederick discussing systemic LFOs on "Coffee Break!"
Thanks for engaging with our paper! We attributed the dips in Fig 1b. to the brief resurgences in subject arousal that occur between task runs, when the experimenter verbally checks in with the subject over the scanner intercom. And the residual drift in Fig 6b. to some combination of genuine arousal-related neural reconfiguration and noise that's not taken care of by FIX (see pink line in Fig 7., where combining RIPTiDe with non-FIX denoising options gets rid of that residual drift). Also certainly possible that magnetic susceptibility effects are at play, but wanted to highlight these other interpretations as well!
ReplyDeleteCertainly possible explanations. But we don't need to guess! More experiments will show what it is! Simply concatenating runs together without waking the subject, or turning off the on-resonance adjustment, things like that. The FC crowd have a lot to understand yet!
ReplyDelete