Most of the available literatures on functional connectivity (FC) usually employed big number of EEG channels. This can be from 32 up to the range of hundreds.
However in [1], they mentioned 8 channels are possible for FC. Yet, I wonder, if there is minimal number of EEG channels for FC. Also, is it possible to use only 5 channels?
[1] Monitoring Effective Connectivity in the Preterm Brain: A Graph Approach to Study Maturation
Please, take into accaunt one manuscript in Attachment. And then you will make a conclusion.
The minimum number of electrodes to calculate connectivity between signal X(t) and Y(t) is two: the two electrodes measuring signals X(t) and Y(t).
What you have to consider is whether the connectivity that you calculate is physiologically meaningful? E.g., two electrodes close to one another will (almost) always show high connectivity, because of spread in the electric field. Different signal strength between conditions can also affect connectivity measures. More channels do not change these issues.
As mentioned above, with a clear hypothesis about what you want to accomplish is a good start. I think that looking at connections between two well-defined signals is as good (or even better) as having loads of electrodes and calculating all-to-all connectivity. Another advice is to consider how potential trivial differences in the signals you are comparing affects the connectivity measure that you use. E.g., do power differences between conditions/groups/etc. make a difference?
If you want to work with graph theory metrics you should use at least 20 channels.
https://sph.umd.edu/sites/default/files/files/Rubinov_Sporns_2009.pdf
If you just use simple channel-based EEG coherence for functional connectivity analysis, the minimise number of EEG channels is 2. The coherence function in EEGLAB can do it. If you prefer to use component-based approaches for EEG functional connectivity, I suggest you can use SIFT toolbox in EEGLAB.
Some of my channels have gone bad and should be excluded.
I am using a 32 channel cap.
Do you suggest me deal with those excluded channels by interpolation?
Or can I use "areas" instead of single electrodes, that are actually averages of a couple of neighboring channels (each area being the average of say 3 channels)? This way, each area will be computed from a 3 electrodes, and if one electrode is missing, the "area" is still there (now calculated from 2 remaining electrodes).
Which strategy is better?
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA