[FieldTrip] rereference to repaired channel

Ingrid Nieuwenhuis inieuwenhuis at berkeley.edu
Mon Dec 12 18:00:06 CET 2011


Hi Jorn,

Thanks a lot! And fortunately, the mastoid has plenty neighbours ('E95'  
'E96'  'E99'  'E101'  'E107'  'E108'), since I'm using high density EEG 
(128 EGI). The spherical spline method sounds pretty cool as well. Would 
be cool to implement one day indeed!

Cheers,
Ingrid

On 12/12/2011 7:18 AM, "Jörn M. Horschig" wrote:
> Hi Ingrid,
>
> ft_channelrepair is simply taking the average of the neighbouring 
> channels. So, you would need to specify your neighbours and based on 
> this, and it just gets the average of the neighbours of your mastoid. 
> However, imho, you cannot really define neighbours for a mastoid 
> channel. There is a method to interpolate not using nearest neighbours 
> but spherical splines, however this is not (yet) part of FieldTrip, 
> see also here:
> http://bugzilla.fcdonders.nl/show_bug.cgi?id=634
> But, I am not sure if this would help in this case either. If you 
> really want to use linked mastoid as a reference, I would probably 
> reject that subject due to bad signal quality. I don't know if anyone 
> on the list has a better idea (or more experience).
>
> Second, if you first rereference, all noise from this one channel will 
> leak into all other channels, as you said. As long as your mastoid is 
> not a bad channel, there should be no difference whether you first 
> rereference or then interpolate bad channels or the other way around 
> (assuming that I got no mathematical blackout right now).
> Should be easy to verify, e.g. if you have channels A, B, and C (all 
> neighbours) and want to reference to R, then your referenced channels 
> would be A-R, B-R and C-R.
> If B is a bad channel, and you interpolate first, you get B=(A+C)/2 
> (i.e. just the average of the neighbouring sensors). If you then 
> rereference you will obtain (A+C)/2 - R, which is the same as (A+C) /2 
> - (R+R)/2 = (A-R + C-R) /2.
> If you first rereference and then interpolate you obtain B-R = (A-R + 
> C-R) /2 immediately (because A-R and C-R are the only neighbours of B-R).
>
> Hope it helps! Best,
> Jörn
>
> On 12/11/2011 9:06 PM, Ingrid Nieuwenhuis wrote:
>> Hi all,
>>
>> I've measured EEG data with Cz as the reference. I now want to 
>> rereference to linked mastoids. However, in one participant one of 
>> the mastoid electrodes is really noisy. Is it valid to throw out the 
>> noisy channel, then fix it (with ft_channelrepair), and then 
>> rereference to this fixed mastoid channel? In other words, is 
>> channelrepair a linear operation? Or would I be mixing (a little) Cz 
>> into all channels now...? Any other options?
>>
>> Or another way of asking: do I get the same result if I first 
>> rereference and then fix bad channels, compared to first fixing bad 
>> channels and then rereference? In the participant with the bad 
>> mastoid chan, I don't have a choice, I have to first fix the channel 
>> first, or I'd be mixing noise into all channels. But, in other 
>> participants, where I also have bad channels (but not the mastoid 
>> ones), does the order matter, which order is better? I'd think first 
>> rereference, then fix... Correct?
>>
>> Thanks!!
>> Ingrid
>>
>
>

-- 
Ingrid Nieuwenhuis PhD
Postdoctoral Fellow
Sleep and Neuroimaging Laboratory
Department of Psychology
University of California, Berkeley
California 94720-1650
Tolman Hall, room 5305




More information about the fieldtrip mailing list