global signal regression (GSR)

[Jeriffer Smith]

Hi,

  I have a question about bold monitor. I wonder whether the result we get from bold monitor directly needs a GSR?  Is this operation necessary for fitness?

 I know my question maybe a bit stupid, thanks for your time!

[John Griffiths]

Hi Jerrifer. 

It is a very reasonable question, and doesn't have a straightforward answer. 

I would say the short answer is 'no' , and less-short answer is 'no and yes'.

Like many preprocessing steps, GSR is intended to remove artefactual contributions from fMRI data. These can and likely do have multiple sources, including movement, pulsation, etc. Those may well be the major contributions to the average signal that is removed by GSR, but (like all preprocessing) there is likely also some non-artefactual, i.e. genuine brain signal being removed when you do GSR. So the 'No and Yes' answer would take the view of doing both to your simulated activity, and with the rationale that applying GSR to the sims is replicating the effect of the GSR on the 'true' brain activity signals in the fmri data. 

All this said, unless you're interested in exploring this as a methodological question, and are familiar with these nuances, don't do GSR on your sims.

Final thought for the groupq: one might consider the argument that we should actually be adding an actual scanner artefact model signal, and analyzing that with exactly the same pipeline as the data. Maybe we should...

HTH,

John

[daniele marinazzo]

Thanks a lot for these thoughts John

the thing is that the "global" is just the ultimate ROI, in this sense no wonder that it contains "actual signal". And even more so given that BOLD is literally made by physiological signals.

If we were sure that with TVB we are simulating exactly the real brain activity, then we would have solved the physiology correction forever, since we should just subtract the empirical data to the real one and obtain the "bathwater".

Or, we should simulate physiology and add it to the framework. Some are trying (two cents from my group as well), but surprise surprise, it's super difficult :)

[Jeriffer Smith]

Thanks for your nice thoughts! It really helps me a lot.

[Daniele Marinazzo]

This relevant paper, on a related topic, just came out

Aquino KM, Fulcher B, Oldham S, Parkes L, Gollo L, Deco G, Fornito A. On the intersection between data quality and dynamical modelling of large-scale fMRI signals. Neuroimage. 2022 Mar 8:119051. doi: 10.1016/j.neuroimage.2022.119051. Epub ahead of print. PMID: 35276367.

--
You received this message because you are subscribed to the Google Groups "TVB Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tvb-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tvb-users/b4f9c10b-dc4d-45f0-801d-2129bc9a05ffn%40googlegroups.com.