depth weighting

✉️ Mailing List Archive (read-only)
1

Hi MNE users,

Just a quick question regarding depth weighting. So far I have only
applied the default settings (weightexp = 0.8 and weightlimit = 10) but I
would like to experiment a bit with these parameters.

I tried to run 'mne_inverse_operator' identically to
'mne_do_inverse_operator' from the command line but it always gives me the
usage options (i.e. those that would be returned with the --help option)
irrespective of what command line options I specify.

What am I missing? I am using version 2.6.1.

Also, any comments on why the defaults were chosen as 0.8 and 10 would be
helpful.

Thanks and best regards,

Pavan Ramkumar
Brain Research Unit
MEG Core
Low Temperature Laboratory
Aalto University School of Science
Espoo, Finland

2

Hi Pavan,
   Not sure why mne_inverse_operator is scooting out in your case..In
general, it has fewer defaults than mne_do_inverse_operator and
requires you to specify a larger set of options for it to run at all.

  As far as the default parameters go: "Our results show that using p =
0.7?0.8 in general yields most accurate localization accuracy, as
quantified by our shift metrics" is a sentence from

http://www.ncbi.nlm.nih.gov/pubmed/16520063

The limit of 10 just prevents the deep sources from being blown up too much.

Regards,
Hari

3

Hi Pavan,

If you want to try using mne_inverse_operator, your best bet is to read the script for mne_do_inverse_operator and feed it all of the details, which it normally would. That script adds details to the call of mne_inverse_operator, which are required inputs.

Good Luck!
D

4

Thanks a lot Dan and Hari!

I did try looking into the script of mne_do_inverse_operator but I haven't
managed to figure it out yet. I'm not sure what command line args to
specify for input parameters I don't use such as $eeg and $fixed

I read through Lin et al. 2006 and it appears from their discussion that
MNE without dSPM is better for deep sources. This was useful.

I'm currently playing with simulated sources in odd places like medial
prefrontal cx and cingulate cx; but a depth weighting exponent of 0.8
estimates the cingulate source in the posterior STG/auditory cx, even for
a surface decimation of ~3-mm separation (ico-5). I wonder whether
increasing the exponent would help in this regard.

Thanks for any advice.

Best regards,
Pavan