Hi everyone! i was wondering what would be the best way to analyze functional connectivity during an event-related brain potential (such as N200, P3b or N400). First of all, I wanted to ask if there was any reason why this should be avoided, and if this should be done with averaged data or individual trials. In addition, we are aiming to study functional connectivity in the brain, by applying source localization algorithms to the ERP data. Finally, I would like to know if there's any open source software that you would reccommend for this kind of analysis.
Thank you so much for your attention!
Phase amplitude correlation over several electrode pairs (front/back. left right) using individual EEG frequency bands , especially low gamma. The analysis should be on individual trials .PAC softtware I believe is available through MAtlab.
Going from source analysis to connectivity would require many untested assumptions
There are lots of very good open source toolboxes for doing exactly what you want. I recommend checking out FieldTrip (http://www.fieldtriptoolbox.org/) and MNE (http://martinos.org/mne/stable/index.html). Both support basic preprocessing, source analysis, and connectivity. There are also good tutorial documentation for getting started and doing this type of analysis.
The best way is a whole different answer. I do not think there is a 'golden' best way. There a lots of reasons why this should be avoided, but also lots of reasons it can work. It depends on what you are looking for.
If you are interested in all-to-all connectivity EEG might not be the optimal tool. The spread of the electric field on the scalp can make sensor space connectivity meaningless since the same underlying signal is picked up by several sensors independently. For a discussion of applications and potential danger see: http://onlinelibrary.wiley.com/doi/10.1002/hbm.20745/full and http://journal.frontiersin.org/article/10.3389/fnsys.2015.00175/full.
Close-range connectivity is difficult and you should at least use a non-amplitude based connectivity measure like imaginary-coherence or phase lag index, etc. With high density EEG you could do source reconstruction which would attenuate (but NOT remove) the volume conduction problem. A seed based approach might reveal long range connections.
That's a good question which also interests me; and is without a clear answer, afaik :(
In case it helps, there is some coverage of the basics in our recent review, http://www.nowpublishers.com/article/Details/HCI-065 (fulltext also available on my RG page).
For tools, I tend to prefer EEGLAB because of the large body of plugins and free code available for many esoteric methods. However if licence costs are an issue then the free Octave version of EEGLAB should provide much of the same functionality. Some people don't like using GUIs - so we've built a batch scripting engine to work with EEGLAB: https://peerj.com/articles/cs-108/
Thank you so much to everyone, you've been really helpful. I realize there's not much consensus on the subject still, but i do wonder if everyone agrees that connectivity should be analyzed on single trial ERPs only. In this case, I would like to ask how do you summarize the data and how you choose the trials for analysis.
Thanks again!
- Badges
- Science topic
- Similar topics
- Biological Science
- Neuroscience