[Mne_analysis] question about LCMV

[Andrea Brovelli]

Hi all,

in fact, I did not look into the details of the lcvm code. For what 
concerns dics, I think the output is the ratio of powers (data/noise), 
that is what is normally called "neural activity index" NAI. In don't 
think the 1997 Van Veen paper has defined the NAI 
(https://www.lsv.uni-saarland.de/fileadmin/teaching/dsp/ss15/BF_VeenBuckley.pdf). 
For a ref about NAI, you can look at this chapter:

http://www.ccs.fau.edu/~fuchs/pub/Beamer.pdf

Or the webpage of Fieldtrip:

http://www.fieldtriptoolbox.org/tutorial/beamformer#neural_activity_index

More generally, it would be good to have the possibility in MNE to have 
different normalization techniques (in addition to "None", in some case 
people may prefer not to normalise wrt noise). Keeping in mind that 
power values are not normally distributed (they are chi-squared), some 
ways to normalise are:

1) dB = 10 * log10(data/noise)

2) z-score = [data - mean(noise)] / std(noise)

3) z-score of logs = [log(data) - mean(log(noise))] / std(log(noise))  
(the log-transform make power values approx. gaussian, as sqrt-transform)

4) ERD/S = [data - mean(noise)] / mean(noise)

... just to cite a few

Some inspiration here (lines 198-207 ):

https://github.com/fieldtrip/fieldtrip/blob/master/ft_freqbaseline.m

bye

Andrea




Le 14/12/2016 à 21:43, Alexandre Gramfort a écrit :
> Hi,
>
> cc Andrea would recently looked a lot at this code.
>
> stc.data contains the output of the filter and taking the magnitude if non
> fixed orientation is used. I am not sure how to call this. Maybe 
> Andrea can tell
> you how he refers to this.
>
> if you apply LCMV to 2 conditions you should use the same noise cov 
> and data cov
> in which case the filters are the same and the stc.data are definitely 
> comparable.
>
> my 2c
> Alex
>
> On Wed, Dec 14, 2016 at 6:21 PM, daisy <smartcandies at 163.com 
> <mailto:smartcandies at 163.com>> wrote:
>
>     Hi, experts
>
>     I have some questions  when I computed LCMV beamformer on evoked
>     data. Could anyone help me?
>
>     (1)I use this
>     command：stc=lcmv(evoked_con1,forward,noise_cov,data_cov,reg=0.01).
>     Then I can plot the source time courses(stc). The xlabel is
>     time(ms). But what is the ylabel? Because the beamformer weight
>     was normalized by noise, according to the paper of Van Veen in
>     1997, the ylabel should be neural activity index. But for one
>     example in the
>     gallery(http://martinos.org/mne/stable/auto_examples/inverse/plot_lcmv_beamformer_volume.html?highlight=lcmv
>     <http://martinos.org/mne/stable/auto_examples/inverse/plot_lcmv_beamformer_volume.html?highlight=lcmv>),
>     the ylabel is LCMV value. Is the ‘lcmv value’ same with ’neural
>     acitvity index’ or ‘pseudo-z value’(Robinson and Vrba 1999)?
>     (2) If I apply LCMV beamformer on evoked data of two conditions,
>     then I can get two stc files. Can I compare them directly? For
>     example, calculate the difference of time course of these two
>     conditions in given time and location? Is it meaningful?
>
>     Thank you!
>
>     -Best
>      Xiaoxu
>
>
>     _______________________________________________
>     Mne_analysis mailing list
>     Mne_analysis at nmr.mgh.harvard.edu
>     <mailto:Mne_analysis at nmr.mgh.harvard.edu>
>     https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>     <https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis>
>
>
>     The information in this e-mail is intended only for the person to
>     whom it is
>     addressed. If you believe this e-mail was sent to you in error and
>     the e-mail
>     contains patient information, please contact the Partners
>     Compliance HelpLine at
>     http://www.partners.org/complianceline
>     <http://www.partners.org/complianceline> . If the e-mail was sent
>     to you in error
>     but does not contain patient information, please contact the
>     sender and properly
>     dispose of the e-mail.
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20161215/25f7c741/attachment.html