ValueError: The eigenspectrum of the leadfield at this voxel is complex. Consider reducing the rank of the leadfield by using reduce_rank=True.

Mailing List Archive (read-only)
1

External Email - Use Caution

Hai group,
I encountered the problem while executing the function make_lcmv. The error
is ,
ValueError: The eigenspectrum of the leadfield at this voxel is complex.
Consider reducing the rank of the leadfield by using reduce_rank=True.

As recommended i changed included the parameter reduce_rank=True in the
function 'make_lcmv', even after that the code gives the same error.
I am not sure what is the problem.
This work uses 19 channel EEG signals.

2

External Email - Use Caution

There might be a voxel at the center of the source space that is nearly
zero in all three directions. Does it help if you pass exclude=10 or so in
the setup_volume_source_space call?

Eric

3

External Email - Use Caution

Hai Eric,
I tried exclude=10 but no luck.
What else i can change to solve this error?

4

External Email - Use Caution

hi Pooja,

can you share a code snippet to replicate this problem?
just a few lines of python with the reading of the data covariance, the forward
and the computation of the make_lcmv so we can investigate.

Alex

5

External Email - Use Caution

Hai Alex,
Thanks for help.
Here is a snippet :
##creating volume source space
src= setup_volume_source_space(subj, pos=5.0,mri=aseg_fname,
bem=fname_model,mindist=7.0,
subjects_dir=subjects_dir)
##forward model
fwd = mne.make_forward_solution(info=raw.info, trans=None, src=src,
bem=fname_bem,
mindist=2.0, meg=False, eeg=True, n_jobs=1)
##computing noise covariance
noise_cov = mne.compute_covariance(epochs, tmin=0, tmax=0.2,
method='empirical', rank='full')
##computing data covariance
data_cov = mne.compute_covariance(epochs, tmin=0.2, tmax=epochs.tmax,
method='empirical',
rank='full')

6

External Email - Use Caution

Hi Pooja,

I used your code with the MNE sample data (EEG) as input, however, I could
not recreate the problem, so it seems to be connected to your data ... Or
were you able to recreate the problem with any of the example data sets?
Some ideas: Did you inspect your head model for the case where this happens
in your data? Did you check whether the leadfield is [0., 0., 0.] anywhere?

Best,
Britta