[FieldTrip] Cross-Spectral Input to ft_eloreta

Nakhnikian, Alexander Alexander_Nakhnikian at hms.harvard.edu
Wed Oct 29 18:19:51 CET 2014


I'm trying to perform source analysis on an EEG data set using ft_eloreta.  I've set up the analysis using the dipole and volume conduction models included in fieldtrip.  I want to perform the analysis in frequency space, which means I need to enter a cross-spectral matrix for the variable Cf.  The cross-spectral matrix returned by ft_freqanalysis doesn't work.  ft_eloreta runs into trouble on line 163 "csd     = dip.filter{i}*Cf*dip.filter{i}' " where  Cf is the cross-spectral density matrix returned by ft_freqanalysis.  This matrix has many rows and its number of columns is equal to the number of frequencies analyzed.  The row number is (numChannels*(numChannels-1))^2/2, so it's assigning a row index to each pair-wise combination of channels, ignoring complex conjugate pairs and autospectra.  The problem is that dip.filter{i} has dimensionality 3 x numChannels, which means Cf must be a numChannels x numChannels matrix to make the linear algebra work out.  Is there a program in fieldtrip that returns a square cross-spectral density matrix for the frequencies of interest?   Are there methods papers by Pascual-Marqui or others that deal with how to construct the cross-spectral matrix?

I am new both to fieldtrip and to eLORETA so any advice would be greatly appreciated.  

Thanks in advance,

Alexander Nakhnikian

More information about the fieldtrip mailing list