[Eeglablist] Different subjects, ICA compare

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Mon May 18 11:11:14 PDT 2015

Dear Mostafa,

Now you are standing on the same frontier as we do; we don't have an
established solution for that issue.
Here are the solutions I've seen so far

1. Use Bayesian Hierarchical modeling to compensate the missing values.
Tim Mullen and Wes Thompson have been suggesting this method for years. Tim
told me that he was in the final phase of checking it. But I haven't seen
it yet. If it works, it will be the best solution.

2. Use Network Projection developed by Nima Bigdely-Shamlo after SIFT
I kind of tried it. I made a mistake in preprocessing and failed, but
theoretically I should success in the second shot. This is a variation of
Measure Projection, but instead of using a single dipole location all ICs
are treated as pairs. I don't think code for Network Projection is publicly
available, so contact Nima.

3. Use LORETA to backproject the scalp activity into cortical grids and
take ROIs to run SIFT
I have never tried this but some of my colleagues in SCCN have tried it
with success. Personally, I'm not attracted to this approach since ICA is
not used in solving the inverse problem in the process of backprojection
from the scalp to cortex.

4. Use SIFT and Measure Projection 'project connectivity'
I have never used this, but I guess what it does is the 'total information
in/outflow' projection. When you sum across column or row of the
connectivity matrix and you can obtain the 'total information in/outflow'
which could be treated just like ERSP. I know Tim and Nima implemented it
together some time ago and should be functional and available from GUI.

5. Hand-pick more or less corresponding ICs across subjects
I tried it and the result was ok. Of course the problem is that you'll lose
a lot of subjects due to 'missing ICs' (and also 'too many ICs' due to
subspace issue could be a problem)


On Sun, May 17, 2015 at 6:55 AM, Mostafa IR <mostafa.rouzbahani at gmail.com>

> I want to use SIFT plugin for finding causal connection between brain
> sources. SIFT visualization option (time frequency grid) use brain sources
> information for calculating connection. Of course ICA decomposition gives
> these information to SIFT.
> I want to compare two different subjects time frequency grid and their
> coupling. But ICA for each one of them gives different information so I
> have a different brain sources. I mean that for example source 3 in subject
> 1 is not the same in subject 2 so I can't compare coupling between sources
> of these two subject.
>  Do you have any idea about comparing different subject with different
> brain sources?
> Thank YOU
> _______________________________________________
> 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

Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20150518/ad80f286/attachment.html>

More information about the eeglablist mailing list