[Eeglablist] Preprocessing steps before Time Frecuency Analysis

[Tarik S Bel-Bahar]

Hello thanks Ruben. Some of the titles below from google scholar might
be of use to you before you move forward, mostly from the last few
years. Cheers!




****************************************************************

Please also search on google for "eeg referencing eeglablist" or other
combinations of keywords if you haven't yet. There are several earlier
notes about biosemi, mastoids, and average referencing.

About the rereferencing, I recommend you take a step back, review the
matlab code in eeglab, contact the manufacturer for their suggestions,
check out chapters and articles on the topics. Some groups rereference
just to mastoids, some to average reference, some to both. Y Overall,
good, you should be double-checking the results of different
processing pipelines yourself. Try only average rerefencing after you
have removed bad channels. Regarding mastoids, your best bet is just
find the last 5 to 10 articles using your eeg system in top journals,
and mimic their rereferencing steps and sequence. Try also contacting
the researrchers directly for their rationale. If you get more clarity
please share with the list.

The jury is still out regarding using blind-source separation methods
and average referencing in relation to connectivity metrics. There's
not enough published data on these issues.
Major connectivity groups and researchers have used ICA, at least for
cleaning up the polluted EEG data, sometimes in top journals like
Neuroimage.
It's preferable to show yourself the results with and without ICA
cleaning rather than relying on preferences based on few published
reports.
It's also an opportunity to clarify some of these issues in the field
with your work :)


***Recent articles for Ruben: (including examples/comparisons with
average reference analyses)

Dynamics of functional and effective connectivity within human
cortical motor control networks

Reliability of event‐related EEG functional connectivity during visual
entrainment: Magnitude squared coherence and phase synchrony estimates

Disturbed phase relations in white matter hyperintensity based
vascular dementia: An EEGdirected connectivity study

Impact of the reference choice on scalp EEG connectivity estimation

Being Conscious of Methodological Pitfalls in Functional Brain Network Analysis

Opportunities and methodological challenges in EEG and MEG resting
state functional brain network research

Measuring large-scale synchronization with human MEG and EEG:
challenges and solutions

How reliable are MEG resting-state connectivity metrics?

The effect of epoch length on estimated EEG functional connectivity
and brain network organisation

A tutorial review of functional connectivity analysis methods and
their interpretational pitfalls

Measuring electrophysiological connectivity by power envelope
correlation: a technicalreview on MEG methods

Unifying blind separation and clustering for resting-state EEG/MEG
functional connectivityanalysis

Identification of brain networks with high time/space resolution using dense EEG


















































On Fri, Jul 22, 2016 at 5:29 AM, Ruben Perellón Alfonso
<ruben.palfonso at gmail.com> wrote:
> Hello Tarik,
>
> thank you very much for your helpful reply. I would like to elaborate some
> more:
>
> We are planing to do coherence analysis, for which ICA and average reference
> seem not to be recommended (e.g. Uhlaas 2006 & 2009 for reference in TF
> analysis, Thatcher 2010 for reference and ICA in coherence analysis)
>
> The the most important at this point is that I am still unsure about the
> re-referencing procedure in EEGLAB. I tried re-referencing to mastoids (TP9,
> TP10) directly after data import (after import reference is unknown in
> eeglab), and compared with the re-referencing procedure described in EEGLAB
> wiki (first get common reference (FCz) back to data, then do average
> reference, and finally do mastoids re-referencing), the resulting signals
> are quite different, so this two are obviously quite different procedures in
> their results. Given that, it would be quite important to be sure that both
> procedures are actually correct, namely is direct re-referencing to mastoids
> just a wrong thing to do in EEGLAB for data recorded with common reference?
> If to do first average reference is a requirement to do mastoids
> re-referencing in EEGLAB, then average reference should be done without
> considering the bad channels, as you say, this would mean one should first
> reject the bad channels before average reference, and only after,
> interpolate them, right?
>
> Best,
> Ruben
>
> 2016-07-22 6:10 GMT+02:00 Tarik S Bel-Bahar <tarikbelbahar at gmail.com>:
>>
>> Hello Ruben, some notes below that may be of use. best wishes.
>>
>>
>> *************************************
>> Regarding your epoch rejection, order seems generally correct. I would
>> suggest also visually inspecting and making changes in thresholds if
>> necessary, and using other sub-tools in the channel-based rejection GUI.
>> With 128 channels, the data is well suited to ICA based artifact-detection
>> and ICA-based analyses. If you haven't had a chance to, examine
>> time-frequency dynamics for particular ICs. See also several tools for
>> cleaning data that come with eeglab, including "trim outlier" and the
>> ASR-based continuous data cleaning.
>>
>> One can drop the mastoid channels after rereference to them.
>> No need to put back in the Fcz. Consider just dropping bad channels,
>> average reference with good channels (see e.g, prep pipeline), and then
>> interpolate bad channels.
>> However, plenty of researchers "rereference" to all channels at the start
>> of their pipelines rather than only using good channels.
>> One could interpolate Fcz as a last step if one wants to report on it.
>>
>> If you haven't had a chance to, see Makoto's pipeline suggestions, eeglab
>> online tutorials for time-frequency analysis and all eeglab aspects, as well
>> google eeglab list for past mentions regarding your topics.
>>
>>
>>
>> On Sat, Jul 16, 2016 at 2:26 AM, Ruben Perellón Alfonso
>> <ruben.palfonso at gmail.com> wrote:
>>>
>>> Hello!
>>>
>>> I am about to pre-proces data before doing Morlet Wavelet Transform.
>>> Some trials where already rejected based on visual rejection criteria and
>>> bad channels where also identified via visual inspection..
>>> These are the next planned steps:
>>>
>>> 1. Re-reference data to linked Mastoids (TP9, TP10)
>>> 2..Interpolate bad channels.
>>> 3. Reject epochs based on threshold -+500microV
>>> 4. Reject epochs based on improbable events, 6SD per channel 2SD for all
>>> channels.
>>> 5. Do wavelet transform
>>>
>>> I would really appreciate the input regarding how correct the order of
>>> this steps looks to you, and whether you consider some fundamental step is
>>> missing.
>>>
>>> I also have a concrete question regarding step 1:
>>> We recorded using a 128 channels Brain Producs system, with a common
>>> reference at FCz. This reference does not appear on import of data into
>>> EEGLAB of course. Now I am wondering how to properly re-reference the data.
>>> In the EEGLAB wiki it says that for linked mastoids, one should first create
>>> a dummy channel of zeros for the common reference, then do average reference
>>> keeping the common reference, and only then re-reference again to the
>>> mastoids. But why can't I just directly re-reference to mastoids? Once
>>> re-reference is done, is there any reasons why I should keep the reference
>>> channels in the data?
>>>
>>> Thanks
>>> Ruben
>>>
>>> _______________________________________________
>>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>>> To unsubscribe, send an empty email to
>>> eeglablist-unsubscribe at sccn.ucsd.edu
>>> For digest mode, send an email with the subject "set digest mime" to
>>> eeglablist-request at sccn.ucsd.edu
>>
>>
>