Aligning EEG and MRI in simultaneous EEG-fMRI

AFNI Message Board
1

AFNI version info (afni -ver):
24

code text  # or delete if not needed

I am just getting started with simultaneous EEG-MRI. When I collect the data, EEG and fMRI will be aligned by receiving the TR pulse from the scanner, which will distribute it to the EEG machine. My question is, can and how do you align data such as these? Can AFNI be used? If so, would there be a pipeline for these things?

Thanks

2

AFNI doesn't have tools that are intended to analyze EEG data, although they can be. Most folks will use their preferred EEG-fMRI software (e.g. FieldTrip, MNE-Python, BrainStorm, EGI Net Station, BrainVision Analyzer2, Compumedics Curry)

Once your EEG data has been processed then it can be combined with the fMRI data in a variety of ways depending on your experimental hypotheses. And if you decide to perform source analysis on your EEG data, then it can be visualized in AFNI/SUMA and group analysis can be performed in the source/surface space using the variety of tools in AFNI.

What is your experiment and hypotheses that you wish to answer?

3

Hi Peter,

Thanks for the input. My experiment will try to put people to sleep inside the scanner to see if some fMRI-based fluid movement metrics in sleep and will be different from their counterparts in wakefulness. The MRI's TR pulse will be used as the sync clock. I will be using EGI Net Station. I did not know that this software could be used to align EEG with fMRI. Could you talk more about how I can do this?

Thanks again!

4

That makes sense! Net Station can do the heavy lifting on the EEG side of cleaning things, but you'll still need AFNI to handle the fMRI data once you have cleaned and classified your EEG data for sleep stage / wakefulness.

I would start by pre-processing your EEG data in Net Station, something like:

  • Remove MR Gradient Artifacts
  • Detect BCG with either EEG or QRS
  • Regress/mitigate the BCG with OBS
  • Filter (either broad band like 0.1-30 or frequency specific like just alpha)
  • Segment based on TR markers
  • Do some artifact detection / bad channel replacement
  • Re-reference to Average Reference
  • Baseline Correct

At this point you'll have EEG semi-aligned with your fMRI data as each epoch will correspond to the fMRI TR. You have a number of choice points on how to proceed. My first inclination would be to proceed if you did broadband filtering, you could do wavelet calculations on the epochs. Then you could export the wavelet calculations using the statistical export in Net Station so that you'll have one set of numbers per-TR. These could then be used to classify fMRI TRs as wake/sleep and processed in AFNI.

There are more complicates and/or intricate ways to get at this, so let me know if you have already made decisions on how to classify the sleep/wake data.