[FieldTrip] LCMV and SAM

Luo, Jessie (NIH/NIMH) [V] luoj at mail.nih.gov
Sat Jul 23 18:34:01 CEST 2011

Thanks Jed and Johanna for your input. 

From: Johanna Zumer [johanna.zumer at donders.ru.nl]
Sent: Friday, July 22, 2011 2:26 AM
To: Email discussion list for the FieldTrip project
Subject: Re: [FieldTrip] LCMV and SAM

Hi all,

As of June 6, 2011 'beamformer_sam' was put back into the Fieldtrip toolbox (after it had gone missing for some time), although, a warning that it has been awhile since the code was tested.

I think there are other differences than just the orientation selection (between the original-CTF SAM and LCMV in FT), although that is a large part of it.
Note the 'fixedori' option in both FT functions (beamformer_sam and beamformer_lcmv).   If set to 'yes' in beamformer_lcmv, it will give a scalar output (similar, maybe equivalent, to fixedori='robert' output in beamformer_sam), both which use SVD to find the orientation. (Default for 'fixedori' in beamformer_lcmv is 'no' so will give a vector output by default).  The default option for 'fixedori' in beamformer_sam is 'spinning' which is the method described by Jed, and I think is what matches the CTF implementation.


On 21 July 2011 23:21, Jed Meltzer <jedmeltzer at yahoo.com<mailto:jedmeltzer at yahoo.com>> wrote:
Without getting into software implementation questions (basically fieldtrip has LCMV, and CTF software has SAM) I think the main difference is that LCMV is a "vector" beamformer.  At any given location, it estimates a virtual signal in all three spatial dimensions (or only two if you constrain it to tangential sources, which makes sense for MEG but not for EEG). So you have three virtual signals at each point, and how you combine them is up to you - take the biggest one, or the magnitude of the whole vector, or any other combination.  I'm not sure what the default for power mapping is in fieldtrip (I have mainly used SAM for beamforming so far, but I use fieldtrip for other purposes).

SAM has an extra step involved where it "optimizes" a dipole orientation at each location to maximize the signal, so you only get one signal at each location.  In this sense it's nonlinear.  The calculation is more complex due to the optimization, but the result is simpler to deal with.  This is called a "scalar" beamformer.  For pros and cons, you might look up papers on vector vs. scalar beamformers in general.  Here's one recent one that I saw that compared them and has further references:

Quraan, M. A., S. N. Moses, et al. (2011). "Detection and localization of hippocampal activity using beamformers with MEG: a detailed investigation using simulations and empirical data." Hum Brain Mapp 32(5): 812-827.

fieldtrip mailing list
fieldtrip at donders.ru.nl<mailto:fieldtrip at donders.ru.nl>

More information about the fieldtrip mailing list