Connectivity calculation using PLI

✉️ Mailing List Archive (read-only)
1

External Email - Use Caution

Hi,

I want to compute connectivity between sources using the PLI metric.
As far as I understand it, PLI is computed at the electrode level and
not at the source level. Does anyone know how to transform the
connectivity matrix between channels into a connectivity matrix
between sources? (I've looked at the inverse operator but I have no
idea how to apply it to the connectivity matrix).

Thanks,

Fleur
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20200304/ae6b14dc/attachment.html

2

External Email - Use Caution

hi Fleur,

you cannot really transfer PLI from sensor to source. You need to estimate
PLI directly in source space eg using parcels.

here an example

https://mne.tools/stable/auto_examples/connectivity/plot_mne_inverse_label_connectivity.html#sphx-glr-auto-examples-connectivity-plot-mne-inverse-label-connectivity-py

or

https://mne.tools/stable/auto_examples/connectivity/plot_mne_inverse_psi_visual.html#sphx-glr-auto-examples-connectivity-plot-mne-inverse-psi-visual-py

note that the choice of the parcellation and metric has been the topic
of many papers
with various recommendations.

HTH
Alex

3

External Email - Use Caution

Hello,

Thanks a lot for your help!

I actually want to z score connectivities for each subject. Is there a way to compute connectivity for each epoch in the source space? (I mean: a more efficient way than iterating over all epochs)

4

External Email - Use Caution

Hi Fleur,

our spectral connectivity code does not implement single-trial
connectivity, it implements phase statistics across trials.
If you want source-level connectivity, you need epochs in source space.
To do that in a memory-efficient manner, you can use
`return_generator=True` in
https://mne.tools/dev/generated/mne.minimum_norm.apply_inverse_epochs.html
and then for loop by source localizing one epoch at a time.

Best,
Denis

5

External Email - Use Caution

Hello,

I tried source localizing one epoch at a time, and then computing connectivity for each epoch.

However, when I average the connectivity matrices over epochs, I do not obtain the same matrix as the one obtained by computing directly connectivity over all epochs in source space. Connectivity matrices for one epoch are made only of 0 and 1.

Is there something I'm missing ?

6

External Email - Use Caution

Hi Fleur,

The code in spectral connectivity does not yield any meaningful
single-epoch results.
At this point, the only valid usage of the function is to compute
connectivity across trials.

Denis