[Eeglablist] ICA decomposition of possibly 2 different conditions....

Arnaud Delorme arno at ucsd.edu
Sat Jan 16 19:28:09 PST 2010


Dear Klados,

yes, you should keep the two conditions together. Otherwise, it is  
going to be hard to compare ICA component activities in the two  
conditions - you would have to find matching components and you would  
never know if the difference you observe between the two conditions  
for a given component pair is due to the difference in component  
carateristics (scalp projection, etc...) or if it due to the  
difference of brin activity in the two conditions. If you have the  
same components for both conditions, the first problem does not arise.  
So keep both conditions together and process the continuous file.

Hope this helps,

Arno

On Jan 12, 2010, at 11:42 PM, Marco Congedo wrote:

> Hello Klados,
>
>    you can use the fact that you have two conditions explicitly to  
> recover sources characteristic of each condition. The appropriate  
> framework is described in length in
> On the blind source separation of human electroencephalogram by  
> approximate joint diagonalization of second order statistics.
>
> Congedo M, Gouy-Pailler C, Jutten C.
>
> Clin Neurophysiol. 2008 Dec;119(12):2677-86.
>
> Cheers,
> Marco Congedo
> Senior Scientist, cnrs
> GIPSA-lab Grenoble
>
> On Tue, Jan 12, 2010 at 11:42 PM, Klados Manousos  
> <mklados at gmail.com> wrote:
> Dear IIya
>
> According to my opinion i suppose that you have to separate them  
> because other pools of neuron (different sources) are activated with  
> eyes opened and other with eyes closed...And if you see the  
> mathematical background because you have a two condition experiment  
> the underlying sources will be more than in one codition....so some  
> sources will be merged in one component...i think that it is not  
> preferable...it is better to reduce "normally" your data  
> dimensionality by separating the recording session....and a 2 mins  
> record is not a short epoch for ICA...There is an heuristic rule  
> which says that ICA can be succesfully applied in epochs with sample  
> points grater than channel^2 => Sample Points>= Channels^2.
>
> I hope to helped you...
>
> 2010/1/12 Ilya Adamchic <dr.ilya at yahoo.com>
> Dear All.
>
>
> We have an experiment set up, when a patient should sit 2 min with  
> eyes closed and 2 min with eyes open and so on several times. This  
> is a spontaneous recording: no actions are performed in both  
> conditions. When we perform ICA decomposition, is it reasonable to  
> leave these 2 conditions mixed as a continuous file and let ICA  
> decompose the continuous file. Or we need to separate eyes closed  
> from eyes open conditions and only then perform ICA on the short  
> epochs, which might be to short (2 min) for the 128 channel  
> decomposition.
>
> Thank you all for your replies in advance.
>
>
>
> _______________________________________________
> 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
>
>
>
> -- 
> Klados A. Manousos
>  Research Assistant
> Group of Applied Neurosciences
> Lab of Medical Informatics, Medical School
> Aristotle University of Thessaloniki
> Thessaloniki, Greece
> _________________________________________________
> Tel: +30-2310-999332
> Website: http://lomiweb.med.auth.gr/gan/mklados
>
>
>
>
> _______________________________________________
> 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
>
>
>
> -- 
> Marco Congedo
> http://www.lis.inpg.fr/pages_perso/congedo/MC_Home.html
> <ATT00001..txt>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20100116/89da7c26/attachment.html 


More information about the eeglablist mailing list