LCMV with common spatial filter?

✉️ Mailing List Archive (read-only)
1

External Email - Use Caution

Dear MNE experts,

We use LCMV for localization of visual gamma oscillations in several
experimental conditions. We expect that the power will differ between
conditions, but that in all these conditions the source will be
approximately in the same area. Therefore, we would like to use a common
spatial filter for all conditions.

We calculated noise_cov and data_cov on the full data and then applied the
resulting common filter to separate conditions, as described in ?examples?:
https://mne.tools/0.15/auto_examples/inverse/plot_lcmv_beamformer_volume.html

This approach visibly reduced signal power in comparison with using
specific filters for each condition.

We have earlier used the common spatial filter with LCMV beamformers in
Fieldtrip and it worked well. However, I guess there is different approach
to assessment of covariance in the Fieldtrip.

What approach to calculation of the LCMV spatial filter would you
recommend us to use in the MNE?

Best,

Elena

2

External Email - Use Caution

hi Elena,

make_lcmv compute the filters and apply_lcmv allow you to apply them
to any condition. it's the recommended way when you compare conditions.

are you observing different results on the same data between Fieldtrip and MNE?

we have just published
https://www.sciencedirect.com/science/article/pii/S1053811920302846

Amit Jaiswal, Jukka Nenonen, Matti Stenroos, Alexandre Gramfort,
Sarang S. Dalal, Britta U. Westner, Vladimir Litvak, John C. Mosher,
Jan-Mathijs Schoffelen, Caroline Witton, Robert Oostenveld, Lauri
Parkkonen,
Comparison of beamformer implementations for MEG source localization,
NeuroImage, Volume 216, 2020,

and our results show very similar results between packages.

HTH

Alex