[Eeglablist] Re-referencing & channel names

[Tarik S Bel-Bahar]

Hi Curtis, brief notes below that may be useful. Cheers.


*********************************************************
Not necessary (but often suggested) to reref before ICA. There are several
camps of preferences about rereferencing before/after ICA.
Recommendation would be to go with recent pubs in high-quality journals
using ICA-19-EEG, Then at least your choice is defendable by published
precedent.

Consider also CSD/Laplacian (though not often used in conjunction with
ICA), and yes, keep remembering the limitations of average reference and/or
ICA for low-density-EEG (< ~40-60 channels). Things may get better in the
future for low-density-EEG with better methods and tech. Yes, definitely,
19-chan data certainly give much less valid and fewer neural ICs. However
19-chan data, in practice, does give up at least a few interpretable good
ICs, especially when the data is high-quality. Some researchers only use
ICA for low-density-EEG to remove clearly artifactual ICs (e..g, eye,
muscle, and zombie ICs).

You should be okay with linked-ears only as recorded.  Check if previous
teams published linked-ears 19-chan ICA without further referencing in good
or high-quality journals.
There's certainly a wide range of 19-chan ICA papers out there, including
in higher-quality journals (easy to googlescholar).

One could do test with a few files to see the differences if any, or report
on two or more referencing schemes (and their effect on study effects).
So in brief, a few in-house tests comparing a few approaches might be the
best thing here.

Would advise caution if creating new channels and then using them for some
referencing procedure.

Some channel renaming suggestions below.
In brief the labels are addressable and modifiable via matlab script in a
loop that acts on each .set file
that you would like to modify (rather than modify the EDFs themselves), or
do the manual editing of info.

way 1?
import a new channel .txt file that
*save out the current channel locations via eeglab gui Channel Editor) -
from a loaded EDF file with the messed up channel labels
*edit/save the channel locations file in a text editor
*import the new channel locaitons file in a loop that opens, changes, and
saves each .set file

way 2?
in a loop through that opens each .set file....
*loop through the EEG.chanlocs.labels
*find labels that contain (contains matlab function) strings you don't want
and replace/delete (strrep matlab function) those substrings
*update/save the EEG structure and .set file

way 3?
*Go into one file's EEG.chanlocs structure. Get the labels column. Save as
new cell array of strings.
*Edit the cell array of channel labels
*Then apply it, via code, into the first column of EEG.chanlocs, and
update/save EEG structure.


* if you haven't had a chance to, check eeglab scripting wiki, and makoto
suggestions page, for a range of useful info that may be of use
*********************************************************




















On Thu, Apr 30, 2020 at 3:56 AM Curtis Bingham <cnbingham at gmail.com> wrote:

> Folks,
> Must you re-reference to perform ICA?
> And if so, how do you re-reference to linked-ears?
> How do you change the channel labels from a dataset?
>
>
> 1.) We’re getting up to speed using eeglab. We’re recording in 19 channels
> with a linked-ears reference. The tutorials and Arnaud’s video overviews
> talk about re-referencing regularly. Must you re-reference to perform ICA?
> If the data were collected using linked-ears is there any reason to
> re-reference the data, assuming we’re not examining a very localized
> trouble-spot?  Or can the data be used as-is, without re-referencing?
>
>
>
> 2.) If we must re-reference, how do you do so with linked-ears?  Are A1 &
> A2 equivalent to T9 & T10? The literature shows an average reference is
> inaccurate unless you have very high-density coverage (i.e. more than 19
> channel cap). We don’t yet have the understanding to do REST. Is it
> sufficient to manually create two new channels in our data via the Channel
> Locs window to represent mastoid placement, and modify the channel locs
> (.ced) file to append to it the dimensions for T9 & T10? Or is there an
> equivalent polar dimension for A1 & A2?
>
>
>
> 3.) How do you change the channel labels from a dataset? We’re recording
> with a NeuroField Q20 19-channel device. The output EDF file names channels
> with prefix & suffix that need to be modified before channel locations are
> set. Names are of the format EEG-ZZ-LE, where ZZ=traditional 10-20 sites.
> Presently we’re manually deleting both EEG- prefix and -LE suffix and it is
> time-consuming. Is there an easy way to pre-process the EDF file to strip
> these extra characters?
>
>
>
>
>
>
>
>
>
> _______________________________________________
> 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