Induced activity

Sat Mar 27 11:41:48 CET 2010

hi bobby,

i agree with michael that there is no method -at least that i'm aware of- to get rid of evoked parts, just due to the issues he mentions.

the solution of reporting the averaged single-trial TF response along with ITC would be ok if one could assume a strict sequence of evoked and induced responses (e.g. <300 ms evoked; >300 ms induced). in practice this however is not the case and induced and evoked processes overlap considerably. This will make problems when you are specifically interested in what is induced: the evoked parts will dominate the entire picture. overlaps may also be created due to the analysis method. e.g. a practical problem my group sometimes encounters is that pre-stimulus alpha desynchronizations are sometimes rendered almost "invisible" in time-frequency analysis due to post-stimulus increases at low frequencies (that largely compose the ERP), that "bleed" into the pre-stimulus period (also an issue regarding baseline correction of post-stimulus responses).
these issues will not be solved by reporting inter-trial coherence.

so depending on what you are interested in, i think that sometimes the best (not conceptually ideal) you can do is to remove the time-frequency representation of the ERP (of course before baseline correction; this would not increase computation time a lot and you could report both if needed). of course you need to be aware of the limitations michael mentioned and decide whether you want to forgo investigating something altogether due to a lack of 100% valid methods. in general however, one could pose a similar question for many things we do e.g. regarding source localization: even though we can not be 100% sure whether the "blobs" or dipole fits we see are valid we do it anyway (or at least many of us).

if anybody has a better idea rather than primitive ERP removal, for separating evoked and incuced activity when they overlap, then it would be great to share it with the community.

cheers,
nathan

On 26.03.2010, at 19:24, Michael Wibral wrote:

> Hi Bobby,
> 
>  
> for fundamental reasons there is (of yet?) no valid approach to get at "pure" induced activity. To see this, imagine a process that produces induced responses, but in a particular participant and day you get by chance mostly responses with relatively similar phases - such that you get a nonnegligible average (some would call it an ERP). Is it still induced? Did it convert to evoked (without the underlying process having changed) ?
> 
>  
> As at least three different mechanisms (additive evoked activity, phase rest of ongoing oscillations, asymmetric amplitude modulation of ongoing oscillations) have been shown to produce non-vanishing averages / ERPs I would also strictly advise against regressing out "evoked" activity from each trial to get induced activity.
> 
> I also do not see a need for analysis of "pure" induced activity. You can calculate power and inter-trial phase-coherence for each time-frequency bin. For the various peaks in your time-frequency representation you quote these two values and most people should be fine with this.
> 
>  
> Michael
> 
>  
>  
>  
>  
>  
> 
> 
> 
> 
> Von: Adrian Bartlett <adrian.m.bartlett at GMAIL.COM>
> Gesendet: Mar 25, 2010 6:51:51 PM
> An: FIELDTRIP at NIC.SURFNET.NL
> Betreff: Re: [FIELDTRIP] Induced activity
> 
> Hi Bobby,
> This sounds like you may have a case of line noise contamination, which is usually seen at 50/60Hz (depends on which continent you are on). While I am not familiar with it's implementation in FIELDTRIP, there are notch filtering solutions to eliminate this contamination. Using multitaper FFT will usually result in the 0 power (post-notch-filtering) at ~60 Hz being smoothed over by the narrowband frequency bias.
> - A
> 
> On Thu, Mar 25, 2010 at 1:33 PM, Bobby Stojanoski <stojanoski at utsc.utoronto.ca> wrote:
> Dear Fieldtrippers,  
>              
> I am a relatively new user of fieldtrip and am very impressed!
>  
> I am interested in comparing differences at certain frequencies – induced 40-100 Hz – between 2 experimental conditions.  My understanding was that I can calculate induced activity in the gamma range by calculating the power for each trial (subject average -- freqanalysis) and then averaging across subjects (grand average -- freqdescriptives/freqgrandaverage).
>  
> To my dismay, when I plotted the results of my grandaverage I found a band of power at 55 - 65 Hz for the entire duration of my epoch. I should add this was not the case when I plotted power across trials for each participant.   
>  
> Earlier discussions mention computing induced+evoked (using freqanalysis and freqgrandaverage) and subtracting that from evoked (using timelockanalysis+freqanalysis) to get extract induced only activity. However, later posts suggest that this is not a valid approach.
>  
>  
> 1. Where have I made my mistake?
>  
> 2. If ((induced+evoked)-evoked)) is not valid, what is the correct approach to calculating induced activity at 40 - 100 Hz?
>  
>  
> Any help would be greatly appreciated!
>  
>  
> Thank you
> Bobby Stojanoski
> 
> 
> 
> ----------------------------------
> 
> The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis.
> 
> http://listserv.surfnet.nl/archives/fieldtrip.html
> 
> http://www.ru.nl/fcdonders/fieldtrip/
> 
> ----------------------------------
> 
> The aim of this list is to facilitate the discussion between users of the FieldTrip toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis.
> 
> http://listserv.surfnet.nl/archives/fieldtrip.html
> 
> http://www.ru.nl/fcdonders/fieldtrip/
> 
> ----------------------------------
> 
> The aim of this list is to facilitate the discussion between users of the FieldTrip toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis.
> 
> http://listserv.surfnet.nl/archives/fieldtrip.html
> 
> http://www.ru.nl/fcdonders/fieldtrip/
> 
> <Michael Wibral.vcf>

----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20100327/ff6ccedd/attachment-0002.html>