Re: [Erppcatoolkit-support] Using PARE correction for EEG data

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

Okay, got a sense of it now!  Your question has a lot of interesting angles to it:

1) You really have another electrode so you might not want to discard it.  CMS (in the Biosemi system) is your recording reference electrode.  As I understand it from the online documentation, Biosemi terms this initial data “unreferenced” which is really misleading.  The data necessarily has a reference (you can’t have voltage data that doesn’t have a reference) and the reference site is CMS.  This site is, by definition, a flat line with zero voltage.  To save disk space Biosemi is not including this flat line with the rest of the data but it’s there.  This is relevant since it stops being flat and boring if you rereference.  You should put it back into your dataset one way or another prior to rereferencing.

2) It looks like you used the .ced file that comes with the EP Toolkit, which in turn is a modified version of the one that Biosemi provides.  An issue you are running into is that the electrode positions for EXG1-8 are not available in this CED file and neither is CMS.  This is a problem for PARE since it needs the electrode locations in order to interpolate the head between the electrode positions.  You should see if you can estimate the electrode positions and add them into the ced file.

3) If you want to contrast nose-reference and average reference, you need to be using the same set of electrodes to obtain comparable results.  So when you compute the average reference you need to include the nose reference: EEG = pop_reref(EEG, [1:64 69], 'keepref','on’).  Note also that the EP Toolkit ced file has EXG1-8 marked as BAD, which is an EP Toolkit notation (not meaningful for EEGlab) which tells EP Toolkit to drop those channels.  When you rereferenced to nose, the nose channel was changed from BAD to REF, which meant it was no longer dropped (as indicated in the command window output when the .set file is imported).  So the nose reference file ends up with a nose channel whereas the average reference file does not.  So again a problem since the two files are no longer comparable.

4) The EEGlab rereferencing directions (http://sccn.ucsd.edu/wiki/Chapter_04:_Preprocessing_Tools) has an odd comment about leaving out the reference channels when calculating average reference.  No idea what they are talking about.  

5) I fixed a couple bugs in the Topos view that was causing it to crash when working with your data.

6) I’m not seeing what you said about the nose-reference-PARE not making much difference compared to the average-reference-PARE and that would indeed be very odd.  

7) However, I do see what you mean about getting different results depending on which you started with.  Thinking about it, it makes sense that the PARE corrected average reference would be dependent on the initial reference scheme.  With regular rereferencing, rereferencing between different electrodes is arithmetically interchangeable.  With PARE correction you’re now including all the interpolated space in between the electrodes and there is no guarantee that they will yield equivalent solutions.  One way of putting it is that whereas in an average reference the electrodes all have the same weight, in PARE some electrodes (notably the ones on the periphery) have much more weight since they have the greatest influence on the interpolated underside of the head.  Junghofer et al did not address this issue.  You can play around with this by using the rereference option (right-click menu) in the Topos View pane.

8) So I’m going to change my implementation of the procedure so that average reference is always computed prior to PARE correction so at least it’ll provide the same result regardless of the initial reference scheme.  Also, your dataset reminded me that a PARE-corrected dataset will necessarily need to drop all the electrodes for which coordinates are not available.

I’ll send you a beta version with these changes implemented for testing and will include them in the next release.

Cheers!

Joe

--------------------------------------------------------------------------------

Joseph Dien, PhD
Research Associate
Cognitive Neurology
The Johns Hopkins University School of Medicine

On Jul 4, 2015, at 7:32 AM, Maria Teresa Wijaya <mtwijaya@u.nus.edu> wrote:

> Hi Joe,
> 
> Yes I used EXG5 as the nose electrode. In addition to the 64 scalp electrodes, I used 8 external ones as follows: 
> 
> 65 - EXG1 - above left eye
> 66 - EXG2 - below left eye
> 67 - EXG3 - left mastoid
> 68 - EXG4 - beside right eye
> 69 - EXG5 - nose
> 70 - EXG6 - right mastoid
> 71 - EXG7 - left arm 
> 72 - EXG8 - right arm
> 
> Here's my processing pipeline up to the point where I generate the files I sent to you. Everything is done in EEGLAB 10 (Matlab 2010b).
> 
> Nose-referenced files:
> 1. Import from bdf, referenced to EXG5, high pass filtered at 0.1 Hz, low pass filtered at 20 Hz. The command that I used for the referencing is EEG = pop_reref(EEG, [69],'keepref','on')
> 2. Manual rejection of non-typical artifacts
> 3. Removal of bad channels
> 4. ICA (runica) and removal of eye blinks and heart rate components
> 5. Interpolation of bad channels
> 
> Average-referenced files:
> 1-3. same as above
> 4. Interpolation of bad channels
> 5. Re-reference to the average of 64 electrodes. The command that I used is EEG = pop_reref(EEG, [1:64], 'keepref','on')
> 6. ICA (runica) and removal of eye blinks and heart rate components
> 
> Thank you. 
> 
> Regards,
> Maria
> From: Joseph Dien <jd...@ma...>
> Sent: Saturday, July 4, 2015 1:09 PM
> To: Maria Teresa Wijaya
> Cc: erp...@li...
> Subject: Re: [Erppcatoolkit-support] Using PARE correction for EEG data
>  
> I think I see.  You used one of your external channels (EXG5) as a nose channel and then rereferenced to it?  I need to know what EXG1-8 were used as in your dataset and what steps you took after data collection with respect to the reference to generate the two files (as in what software and what settings did you use).
> 
> Thanks!
> 
> Joe
> --------------------------------------------------------------------------------
> 
> Joseph Dien, PhD
> Research Associate
> Cognitive Neurology
> The Johns Hopkins University School of Medicine
> 
> On Jul 3, 2015, at 9:31 PM, Joseph Dien <jd...@ma...> wrote:
> 
>> Before I can address the PARE question, I’m having some trouble with your .set files.  There is an aspect of them which is crashing my code.  Could you tell me how the rereferencing was accomplished?  To both nose and average reference?  In particular, EXG5 is identified as being the reference electrode for the nose referenced dataset rather than CMS.
>> 
>> Joe
>> --------------------------------------------------------------------------------
>> 
>> Joseph Dien, PhD
>> Research Associate
>> Cognitive Neurology
>> The Johns Hopkins University School of Medicine
>> 
>> 
>> On Jun 28, 2015, at 10:16 PM, Maria Teresa Wijaya <mtwijaya@u.nus.edu> wrote:
>> 
>>> Yes, I did, sorry, I forgot about that. I've uploaded the files with the correct names in the folder: https://www.dropbox.com/sh/k2fpxx99kt6u0vg/AACpdOaLICsFdDBpLjollVWma?dl=0
>>> 
>>> Thank you.
>>> 
>>> Maria
>>> 
>>> 
>>> From: Joseph Dien <jd...@ma...>
>>> Sent: Monday, June 29, 2015 3:31 AM
>>> To: Maria Teresa Wijaya
>>> Cc: erp...@li...
>>> Subject: Re: [Erppcatoolkit-support] Using PARE correction for EEG data
>>>  
>>> I’m sorry, I can’t read the .set files you uploaded.  It looks like you renamed the files?  .set files don't let you do that.  Could you send me the files with the original names?
>>> 
>>> Joe
>>> 
>>>> On Jun 26, 2015, at 05:58, Maria Teresa Wijaya <mtwijaya@u.nus.edu> wrote:
>>>> 
>>>> Hi Joe,
>>>> 
>>>> I have uploaded the files to dropbox: https://www.dropbox.com/sh/k2fpxx99kt6u0vg/AACpdOaLICsFdDBpLjollVWma?dl=0
>>>> 
>>>> In my computer I could not load the ones with the event information so I removed the event information and stored them in the .csv file; these can be found in the "no event" folder. The files in the "with event" folder are from the same data set only with the event information intact.
>>>> 
>>>> Thank you!
>>>> 
>>>> Maria
>>>> 
>>>> 
>>>> From: Joseph Dien <jd...@ma...>
>>>> Sent: Thursday, June 25, 2015 8:55 AM
>>>> To: Maria Teresa Wijaya
>>>> Cc: erp...@li...
>>>> Subject: Re: [Erppcatoolkit-support] Using PARE correction for EEG data
>>>>  
>>>> How odd!  Could you send me the nose-referenced and averaged-referenced data files so I can look into this?
>>>> 
>>>> Joe
>>>> 
>>>>> On Jun 23, 2015, at 08:08, Maria Teresa Wijaya <mtwijaya@u.nus.edu> wrote:
>>>>> 
>>>>> Dear Joe, 
>>>>> 
>>>>> Thank you for your quick reply. I have actually tried applying the PARE referencing to both nose-referenced and average-referenced data. In the former, it barely made any noticeable difference while in the latter it did change the grand average waveform. I am not sure why it did not work on the nose-referenced data. Is it actually okay to apply PARE referencing to a data set that has already been referenced to the average of the scalp electrodes?
>>>>> 
>>>>> Thank you.
>>>>> 
>>>>> Regards,
>>>>> Maria
>>>>> 
>>>>> 
>>>>> From: Joseph Dien <jd...@ma...>
>>>>> Sent: Friday, June 19, 2015 9:27 PM
>>>>> To: Maria Teresa Wijaya
>>>>> Cc: erp...@li...
>>>>> Subject: Re: [Erppcatoolkit-support] Using PARE correction for EEG data
>>>>>  
>>>>> Thanks!  You do not need to re-reference first.  The PARE option is just a modified average-reference procedure (referenced to the average of the full interpolated head rather than just the average of the electrode locations).  Use the “re-reference” option in the Topomaps view to get a sense of what it is doing.
>>>>> 
>>>>> Cheers!
>>>>> 
>>>>> Joe
>>>>> --------------------------------------------------------------------------------
>>>>> 
>>>>> Joseph Dien, PhD
>>>>> Research Associate
>>>>> Cognitive Neurology/Neuropsychology
>>>>> The Johns Hopkins University School of Medicine
>>>>> 
>>>>> 
>>>>>> On Jun 19, 2015, at 01:18, Maria Teresa Wijaya <mtwijaya@u.nus.edu> wrote:
>>>>>> 
>>>>>> Hello,
>>>>>> 
>>>>>> First of all, thank you for the amazing EP toolkit. I am currently trying to make use of the PARE correction feature. My data were recorded in Biosemi (64 channels) and have been filtered and re-referenced to nose in EEGLAB.
>>>>>> 
>>>>>> My question is, do I have to first re-reference my data to average before using the PARE correction? 
>>>>>> 
>>>>>> Thank you and apologize if this question is too basic.
>>>>>> 
>>>>>> Regards,
>>>>>> 
>>>>>> Maria
>>>>>> ------------------------------------------------------------------------------
>>>>>> _______________________________________________
>>>>>> Erppcatoolkit-support mailing list
>>>>>> Erp...@li...
>>>>>> https://lists.sourceforge.net/lists/listinfo/erppcatoolkit-support
>>>>> 
>>>> 
>>>> --------------------------------------------------------------------------------
>>>> 
>>>> Joseph Dien, PhD
>>>> Research Associate
>>>> Cognitive Neurology/Neuropsychology
>>>> The Johns Hopkins University School of Medicine
>>>> 
>>>> 
>>> 
>>> --------------------------------------------------------------------------------
>>> 
>>> Joseph Dien, PhD
>>> Research Associate
>>> Cognitive Neurology/Neuropsychology
>>> The Johns Hopkins University School of Medicine
>>> 
>> 
>> ------------------------------------------------------------------------------
>> Don't Limit Your Business. Reach for the Cloud.
>> GigeNET's Cloud Solutions provide you with the tools and support that
>> you need to offload your IT needs and focus on growing your business.
>> Configured For All Businesses. Start Your Cloud Today.
>> https://www.gigenetcloud.com/_______________________________________________
>> Erppcatoolkit-support mailing list
>> Erp...@li...
>> https://lists.sourceforge.net/lists/listinfo/erppcatoolkit-support