Functional connectivity- magnitude vs. imaginary coherence

Discussions
,
1

Hi,

I am working on resting state EEG data, and I have two groups controls and patients, for which I would like to calculate the coherence between scouts of interest for each subject (1xN), as we would like to know which group has an increased or decreased connectivity for a particular frequency band. So I tested using both the magnitude square coherence and imaginary coherence method.

I had calculated the source per subject and per trial using the sLORETA method for unconstrained sources and the head model was calculated using the BEM (volume) option.

I calculated the coherence (A x B) by adding the trials for each subject into process 2 and selecting the source option, followed by selecting 1 scout each of interest from A and B (the volume atlas was same for both). The scout function was: mean, apply the scout option was: before. The window length for PSD estimation was 0.5 (I have 1 second trials), overlap was 50% and highest frequency of interest was 30 Hz. I had selected the output as average cross spectra.

I am seeing for example for a subject, with the magnitude squared coherence method I am getting a coherence peak at 4.8 Hz, while if I use the imaginary coherence method then I am getting a peak at 8.7 and 13.6 Hz.

Am I doing something wrong? Is there any preference as to which method should be used? since this would effect our results and hypothesis in a big way.

Any help regarding this is greatly appreciated.

Thank you so much.

Best,
Ami

2

Hi,

Is there any feedback I could get regarding my previous post.
Thank you so much for your help.

Best,
Ami

3

Coherence and Imaginary Coherence are two different metrics for assessing connectivity. If they provide different outcomes, this means that some of the effects detected with coherence might be due to sources that are strictly synchronized either physiologically or because of signal leakage from source reconstruction. Regardless, I recommend you apply the Scout Function after, instead of before, as otherwise the source time series are averaged within the scout before being subjected to connectivity analysis. By doing so, sensitivity, especially at higher frequencies is degraded. It is more computationally intensity to run the "after" option but again, effect sensitivity is greater.

4

it is believed that Imaginary Coherence is more accurate since it avoids the volume conduction effect,

1 Like
5

Thank you so much for your help and feedback.

Best,
Ami

1 Like