[Eeglablist] rejecting/interpolating bad channels?

Thu May 12 11:56:58 PDT 2011

Dear Sirs,

I found the same issue to be a bit troubling, in as much as the opposite
problem arises too: a Kurtosis threshold 5 can sometimes reject channels
which are in fact quite acceptable once averaged in ERPs, at least according
to my research purposes. I think this has to do with the nature of the
Kurtosis measure of dispersion's probability. That is the reason why I ended
up using a joint approach Kurtosis threshold 15 and ERPLAB absolute
threshold and step-like AR functions performed on each of the neural
channels. I place a marker when channels are exceeding a defined threshold
in microVs, and I then consider the number of those markers along with the
Kurtosis value, for each channel. You will need to play around with these
thresholds too, according to your purposes and to your data, but this works
fine for me.

Cheers,

Mahesh

Mahesh M. Casiraghi
PhD candidate - Cognitive Sciences
Roberto Dell'Acqua Lab, University of Padova
Pierre Jolicoeur Lab, Univesité de Montréal
mahesh.casiraghi at umontreal.ca

I have the conviction that when Physiology will be far enough advanced, the
poet, the philosopher, and the physiologist will all understand each other.
Claude Bernard

On Wed, May 11, 2011 at 5:14 PM, Tarik S Bel-Bahar
<tarikbelbahar at gmail.com>wrote:

> Greetings, I am wondering why as well.
> In particular, does anyone have an EEGLAB-based algorithm
> that catches these extremely bad channels, or very noisy ones without big
> amplitude changes, in either continuous or epoched data ?
>
> Andrew, the bad channel rejection probably works differently if you are
> using continuous or epoched data.
> I bet that the epoched (or windowed) data bad channel detection would catch
> the big bad channel.
>
>
>
> On Mon, May 9, 2011 at 4:34 PM, Andrew Hill <andrewhill at ucla.edu> wrote:
>
>> Thanks James - Exactly what I needed.
>>
>> Another question for the list about automagic channel rejection.
>> This picture shows some EEG and the channels that are marked as bad with
>> kurtosis / probability thresholds of 5:
>> http://salamander.net/stage/pastebin/autorejects.jpg
>>
>> You will notice it doesn't remove channels that are wildly out of range -
>> even dropping the thresholds on the methods to 2 or under, those extremely
>> bad channels aren't "caught"...
>>
>> Anyone have an idea why?
>>
>> Best,
>> Andrew
>>
>> On May 9, 2011, at 6:10 AM, James Desjardins wrote:
>>
>> Hi Andrew,
>>
>> The outputs listed in the pop_rejchan help text can be added following
>> the EEG output at the command line as follows:
>>
>> [EEG,indelec] = pop_rejchan(EEG,'elec',1:64], ...
>> 'threshold',5,'norm','on','measure','kurt');
>>
>> Then used in eeg_interp as follows:
>> EEG = eeg_interp(EEG,indelec)
>>
>>
>> Also, it may be good practice to store that vector in the EEG
>> structure so you can keep the information when you subsequently save
>> the dataset. Other rejection variables are stored in the EEG.reject
>> field.
>>
>> The following should work nicely and the marked channels will be
>> stored when you save the dataset:
>>
>> [EEG,EEG.reject.indelec] = pop_rejchan(EEG,'elec',1:64], ...
>> 'threshold',5,'norm','on','measure','kurt');
>>
>> EEG = eeg_interp(EEG,EEG.reject.indelec)
>>
>>
>> I have been using a procedure similar to this in my data processing
>> stream where I interpolate bad channels then include them in my
>> average reference (the interpolation allows for a consistent and
>> spatially balanced average reference channel across participants
>> regardless of what channels are marked bad). I keep track of which
>> channels have been interpolated (in the case above using the
>> EEG.reject.indelect field) and then I do not include them in
>> procedures where interpolation is inappropriate (eg. ICA).
>>
>> I hope that this is helpful.
>>
>> ps. I am going to use the term "automagically" regularly from now on.
>>
>> James Desjardins
>> Technician, MA Student
>> Department of Psychology, Behavioural Neuroscience
>> Cognitive and Affective Neuroscience Lab
>> Brock University
>> 500 Glenridge Ave.
>> St. Catharines, ON, Canada
>> L2S 3A1
>> 905-688-5550 x4676
>>
>>
>> Quoting Andrew Hill <andrewhill at ucla.edu>:
>>
>> Hi folks,
>>
>>
>> I'm trying to figure out how to first automagically detect bad
>>
>> channels and then interpolate them on the fly, before saving my EEG
>>
>> data set.
>>
>> Using pop_rejchan seems to do what it should:
>>
>> ...
>>
>> EEG = pop_rejchan(EEG ,'threshold',5,'norm','on','measure','kurt');
>>
>> Computing kurtosis for channels...
>>
>> 4 electrodes labeled for rejection
>>
>> Removing 4 channel(s)...
>>
>> ...
>>
>>
>> But these are of course not removed - just marked as red in the plot
>>
>> and I'm assuming labeled in the EEG object somewhere.
>>
>>
>> I see an output parameter "indelec" in the pop_rejchan help, but
>>
>> after running pop_rejchan I don't see an "indelec" variable created,
>>
>> nor a EEG.indelec one.
>>
>>
>> Ideally I'd like to take the indices of any marked-bad channels and
>>
>> then immediately interpolate, maybe something like this?
>>
>>
>> EEG = pop_rejchan(EEG, 'elec',[1:64]
>>
>> ,'threshold',5,'norm','on','measure','kurt');
>>
>> EEG = eeg_interp(EEG, EEG.indelec)
>>
>>
>> But that of course isn't syntatically correct - how/where is this
>>
>> information stored?
>>
>> Is anyone else doing something like this?
>>
>>
>> Also, does the pop_reref function know enough to exclude channels
>>
>> that have been marked bad?
>>
>> I couldn't see anything relevant in the function help.
>>
>>
>> Thanks,
>>
>> Andrew
>>
>>
>>
>> _______________________________________________
>>
>> 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
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> 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
>>
>>
>>
>> _______________________________________________
>> 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
>>
>
>
> _______________________________________________
> 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110512/70c9a953/attachment.html>