Estimate envelope correlation per band

💬 Support & Discussions
1

Environment details
MNE version: 0.23.4
OS: Windows 10

Hi,

My goal is to estimate the all-to-all envelope correlation in sensor space. I want to perform such task for the theta, alpha, and beta bands.

I know with mne_connectivity.spectral_connectivity you can estimate the spectral connectivity in multiple frequency bands using fmin and fmax parameters. However, for mne_connectivity.envelope_correlation, there are no such parameters.

My question is, knowing this, do I need to filter my Epochs in the band of interest and then apply mne_connectivity.envelope_correlation to estimate the envelope correlation in that band or am I missing something?

Thanks in advance.

Eduardo

2

Hi @eduardo,
Yes, that’s what I’d do. Take a look at this example - it does what you want to achieve but performs this in source space (connectivity in sensor space is generally a risky buisness).

1 Like
3

Hey @mmagnuski, thanks for the comment.

May I ask you what do you mean when saying connectivity in sensor space is a risky business? I thought pair-wise orthogonalization prevented from undesired effects.

In fact, most of the papers I have read on connectivity (in my case for Alzheimer’s detection), rely on sensor space connectivity. For instance, this one.

4

May I ask you what do you mean when saying connectivity in sensor space is a risky business? I thought pair-wise orthogonalization prevented from undesired effects.

Sorry, that comment was not clear, and not needed. :slight_smile: You are right that here are ways to reduce the leakage/volume conduction (which are frequently used in source space connectivity as well), but they don’t correct for variability in source projections (differences in source orientations across subjects) for example. As a consequence I’d expect connectivity in channel space to be more difficult to interpret, less consistent and less sensitive to effects. Similar concerns are voiced for example in OHBM COBIDAS standards. However, I’m aware source reconstruction can also be problematic and should be interpreted cautiously; also this is a very broad topic and I don’t want to derail this thread with my private reservations. :slight_smile: