[FieldTrip] SAM beamforming on Neuromag data

[Elena Orekhova]

Hi Michael,

> If you run into rank-deficiency issue with the covariance matrix a tiny amount of regularization should fix this.

What lambda you would recommend?

Elena


________________________________
From: fieldtrip-bounces at donders.ru.nl [fieldtrip-bounces at donders.ru.nl] on behalf of Michael Wibral [michael.wibral at web.de]
Sent: Monday, May 30, 2011 5:58 PM
To: Email discussion list for the FieldTrip project
Subject: Re: [FieldTrip] SAM beamforming on Neuromag data


Hi Elena,

as far as I can see from the neuromeg discussion list and the maxfilter papers the properties of the components removed by the maxfilter do not require a leadfield update.
If you run into rank-deficiency issue with the covariance matrix a tiny amount of regularization should fix this.
(Note: If someone who reads this is of a different opinion, please let me know!)


Michael

________________________________
Von: "Elena Orekhova" <Elena.Orekhova at neuro.gu.se>
Gesendet: May 30, 2011 4:30:34 PM
An: "Email discussion list for the FieldTrip project" <fieldtrip at donders.ru.nl>
Betreff: Re: [FieldTrip] SAM beamforming on Neuromag data

Thank you for this.

I have more concerns.  I applied MaxFilter to the data. Since MaxFilter reduces the rank of the covariance matrix by removing noisy components, it may influence the beamformer results.
Is it safe to do beamforming with MaxFiltered data?

Elena

________________________________
From: fieldtrip-bounces at donders.ru.nl [fieldtrip-bounces at donders.ru.nl] on behalf of Michael Wibral [michael.wibral at web.de]
Sent: Monday, May 30, 2011 2:08 PM
To: Email discussion list for the FieldTrip project
Subject: Re: [FieldTrip] SAM beamformeing on Neuromag data


Hi Elena,

as far as I know, the leadfield computation should be aware of the different UNITS (not only scales) of gradiometers and magnetometers. There was a problem with the sign of the leadfields but that should have been fixed.

There is one more fundamental problem however, that you should be aware of (doesn't invalidate your source analysis but bears potential for fine-tuning), which is the projection of noise:
In beamforming the unit gain constraint guarantees that you get your source signal back with unit gain. Added on top however is neurophysiological crosstalk (minimized) and sensor noise of the sensors with the largest weights in your Beamformer (not reducible). So different sensor types willhave different (inverse) leadfield strengths, theerfore also fiofefrent source noise levels. the relative benefits of each sensor type changes from location to location, so a location (and data) dependend weighting would in principle be best.
I am not sure if and how this is implemented if FT (Bayesian weighting would be optimal here..)

What you could do as a first step is to beam separately and compare the results.

Michael


________________________________
Von: "Elena Orekhova" <Elena.Orekhova at neuro.gu.se>
Gesendet: May 30, 2011 1:08:29 PM
An: "fieldtrip at donders.ru.nl" <fieldtrip at donders.ru.nl>
Betreff: [FieldTrip] SAM beamformeing on Neuromag data

@font-face { font-family: "\FF2D \FF33 \660E \671D "; }@font-face { font-family: "\FF2D \FF33 \660E \671D "; }@font-face { font-family: "Cambria"; }p.MsoNormal, li.MsoNormal, div.MsoNormal { margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: Cambria; }pre { margin: 0cm 0cm 0.0001pt; font-size: 10pt; font-family: Courier; }span.HTMLPreformattedChar { font-family: Courier; }.MsoChpDefault { font-family: Cambria; }div.WordSection1 { page: WordSection1; }@font-face { font-family: "\FF2D \FF33 \660E \671D "; }@font-face { font-family: "Cambria Math"; }@font-face { font-family: "Cambria"; }p.MsoNormal, li.MsoNormal, div.MsoNormal { margin: 0cm 0cm 0.0001pt; font-size: 12pt; font-family: Cambria; }pre { margin: 0cm 0cm 0.0001pt; font-size: 10pt; font-family: Courier; }span.HTMLPreformattedChar { font-family: Courier; }.MsoChpDefault { font-family: Cambria; }div.WordSection1 { page: WordSection1; }
Dear All,
I try to run beamformer analysis on the auditory MEG data (Neuromag) and have basic questions.

1.Magnetometers and gradiometers Neuromag sensors have different scales. Does the Fieldtrip take care of this difference or should I normalize the data?   It yes, how to  normalize?


2. I would like to do SAM analysis  of evoked field and look at the time courses at ROIs  (virtual channels).  The only tutorial example I have found was for the lcmv-beamformer
(cfg.method = 'lcmv'; http://fieldtrip.fcdonders.nl/example/lcmv-beamformer). I am not sure
which parameters should I specify in ft_sourceanalysis if cfg.method = 'sam'.





I would be most grateful for any example script of this type analysis!



Regards,

Elena

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20110531/96932714/attachment-0001.html>