PLV Connectivity

Discussions
1

Hello everyone,

I have a question regarding PLV connectivity with brainstorm. I have 25 trials of -200 to 800 ms for each subject(20 subjects) of EEG recordings, I wanna perform PLV but I have a doubt. I wanna know how does brainstorm calculate PLV, it is possible or is it correct to calculate it with these amount of trials and length window? Would it be possible to have a edge problem in the trials if PLV filters in specific frequency bands?

Which is the best way to perform PLV with this kind of data??

Thanks in advance!

2

Hola Sandra!

It will depend on the range of frequencies you’d like to look at. Bear in mind that is advisable to have a window length of at least 2/3 cycles of the slowest frequency.
The data is bandpassed using a FFT-FIR based filter and by default, signals are mirrored, before applying it, to avoid edge effects.
Regarding the trials, Brainstorm allows you to save the connectivity results individually or to average them all together.
It also has the option of computing PLV across trials, keeping time information. This requires the average of many trials though, which in your case is not very high.

I hope that help. Cheers,

3

Hi Sandra,
Response from Peter (PhD in our lab):

For sure there can be edge artifacts, that depends on the frequency band you’re interested in. There would be more problems with lower frequencies of course. To get a sense of this, you could compute a time-frequency map and click on ‘Hide edge effects’ to see what’s the extent of this in different frequencies. This way you’d also know better what you’re dealing with before doing something more complicated like PLV.

Also you could import epochs with some extra time before and after (e.g. -400 to 1000ms), filter the data and then compute PLV on the original epochs (e.g. -200 to 800ms).

4

Thanks Guiomar and Francois for your replies.

I was interested in analyzing all frequency bands including delta as well… I was wondering if a better approach could be to concatenate all the trials in time so I could have more time so then I could see slow frequencies. I saw that Brainstorm has a function to concatenate trials… Could I use it in this context?

Thanks again!

5

Hi Sandra,

Concatenating trials can allow you to estimate better lower frequencies without the edge effects.
I added for PLV the same option as in the Coherence process, that allows you to concatenate the input files instead of averaging the output files.

However, I’m not sure that is going to do any good to the PLV estimation accuracy. If you concatenate two trials that were not contiguous in the recordings, you add jumps/discontinuities in the recordings.
These jumps translate in frequency domain to bursts of high frequencies, and also a phase reset…
I will let Guiomar and Peter commenting on this part…

Francois