Hi Vladimir,

Thanks! I will try it out and will let you know!

best,

Stephan

________________________________________________________
Stephan Moratti, PhD

see also: http://web.me.com/smoratti/

Universidad Complutense de Madrid
Facultad de Psicología
Departamento de Psicología Básica I
Campus de Somosaguas
28223 Pozuelo de Alarcón (Madrid)
Spain

and

Center for Biomedical Technology
Laboratory for Cognitive and Computational Neuroscience
Parque Científico y Tecnológico de la Universidad Politecnica de Madrid
Campus Montegancedo
28223 Pozuelo de Alarcón (Madrid)
Spain


email: [log in to unmask]
Tel.:    +34 679219982

El 08/05/2012, a las 23:51, Vladimir Litvak escribió:

Dear Stephan and Erick,

Please try the attached fix for the problems with lead field scaling.
It does seem to improve things on our example. It'll be in the next
SPM8 update.

Vladimir

On Thu, May 3, 2012 at 11:48 AM, Erick Britis Ortiz
<[log in to unmask]> wrote:
Dear Stephan, Vladimir, Rik,

we have a CTF system, but we had a problem that might be related to
yours. I am taking the opportunity to inform Vladimir too.

The leadfield calculation changed from r4010 to r4290. When the
preparation steps (forward model, etc) for the inversion are done with
one version of SPM, but the leadfield calculation is done with another,
the leadfields might end up with a different scale than if calculated
with just one version. The values are the same, just multiplied by 10^-10.

This is only noticeable in the group inversion, due to the svd in the
leadfield realignment. On r4010, the leadfield values are between 1e-2
and 1, all fine. On r4290, the values drop to around 1e-12, and the
calculation bumps into floating point precision, resulting in 0 spatial
modes.

To reproduce it, (1) do a complete inversion with SPM r4010, (2) rename
the SPMgainmatrix.mat file, (3) re-start MATLAB with SPM r4290, (4)
re-calculate the leadfields and compare the two.

A quick solution is to erase any stale SPMgainmatrix.mat files and start
the source localization from scratch again ("clear" it, in the GUI).
Probably re-doing just the forward model would be enough, but better
safe than sorry. If even this does not work, I suggest trying with r4010.

Good luck,
Erick


On 2012-05-03 08:14, [log in to unmask] wrote:
Hi Vladimir and Rik,

Thanks for your fast reply. I will try to use less subjects and see what happens.

Best,

Stephan

________________________________________________________
Stephan Moratti, PhD

see also: http://web.me.com/smoratti/

Universidad Complutense de Madrid
Facultad de Psicología
Departamento de Psicología Básica I
Campus de Somosaguas
28223 Pozuelo de Alarcón (Madrid)
Spain

and

Center for Biomedical Technology
Laboratory for Cognitive and Computational Neuroscience
Parque Científico y Tecnológico de la Universidad Politecnica de Madrid
Campus Montegancedo
28223 Pozuelo de Alarcón (Madrid)
Spain


email: [log in to unmask]
Tel.:    +34 679219982

El 02/05/2012, a las 22:17, Rik Henson escribió:


I can confirm that I also sometimes get 0 spatial modes left when using group inversion for MSP, whether Mags only, Grads only or both (ie it appears to be the group inversion that is problematic). The same data gives reasonable inversions when I use MSP without group inversion (ie individual MSP), or MNM or COH. So it must be a problem with the realignment of leadfields; not the leadfields themselves (and I get the same problem whether I use single-shell or single-sphere forward models).

Unfortunately I haven't yet tracked down why this happens, because on other Neuromag datasets in the past, the group inversion worked fine. I have confirmed that the spm_eeg_invert function (in which the leadfield realignment is done) has not changed during that time, so I can only assume that something else "deep down" in the forward modelling (eg FieldTrip) functions has changed since then?

Note that I don't think it can be anything to do with the way the data is preprocessed (eg via SSS or SSP), because the MEG data are not used when realigning the leadfields. (Though note that, even if group inversion did work, any spatial adjustment done to the data, eg via SSS or SSP, should, in principle, also be applied to the forward modelling. I'm not sure what is done with any SSP vectors when SSS is applied, but assume that any such SSP vectors present are applied when the data are converted (via the FieldTrip-MNE FIF-reading functions) - and would hope they are also therefore stored somewhere so FieldTrip can adjust leadfields accordinly, but I think we would need advice from a FieldTrip expert....?)

Sorry not to be more help - I am still puzzling!

BW,R

-------------------------------------------------------
DR RICHARD HENSON
Assistant Director for Neuroimaging
MRC Cognition & Brain Sciences Unit
15 Chaucer Road
Cambridge, CB2 7EF
England
EMAIL:  [log in to unmask]
URL:    http://www.mrc-cbu.cam.ac.uk/people/rik.henson/personal
TEL     +44 (0)1223 355 294 x501
FAX     +44 (0)1223 359 062
MOB     +44 (0)794 1377 345
-------------------------------------------------------
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of [log in to unmask] [[log in to unmask]]
Sent: 02 May 2012 19:49
To: [log in to unmask]
Subject: [SPM] MEG group comparison neuromag

Dear SPM users,

I tried to do a group inversion with neuromag MEG data (2 conditions, 18 subjects). When I include MEGPLANAR I get 0 spatial modes and NaNs. For the MEG I get 1 spatial mode (if I invert individually I get 66 modes for MEG and 104 modes for MEGPLANAR). So, if I only use MEG to get a solution, I get for all subjects, conditions the same solutions (and a very unlikely one, beam former, MNE and LORETA get visual cortex activity, GS group inversion gets auditory activity that is equal for all conditions and subjects).

I saw similar postings in the SPM list in 2010 and it was advised to wait for a newer SPM version. I have used the newest 2012 SPM version. I did tSSS with the MEG data, but I do not know if SPM applies the SSP vectors to the data (it shouldn't when using tSSS). Could this be the problem?

Best,

Stephan

________________________________________________________
Stephan Moratti, PhD

see also: http://web.me.com/smoratti/

Universidad Complutense de Madrid
Facultad de Psicología
Departamento de Psicología Básica I
Campus de Somosaguas
28223 Pozuelo de Alarcón (Madrid)
Spain

and

Center for Biomedical Technology
Laboratory for Cognitive and Computational Neuroscience
Parque Científico y Tecnológico de la Universidad Politecnica de Madrid
Campus Montegancedo
28223 Pozuelo de Alarcón (Madrid)
Spain


email: [log in to unmask]
Tel.:    +34 679219982

El 02/05/2012, a las 20:03, Liam Mason escribió:

Hi Carlos,

Did you get any answer of if it is possible to extract %SC from a group of subjects? I'm afraid the only way I know is to do them iteratively as you say. If marsbar is able to do it from a second-level (group) SPM file, then would be keen to know!


Liam

---
School of Psychological Sciences
University of Manchester
0161 275 2692

________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Ming-Tsung Tseng [[log in to unmask]]
Sent: 24 April 2012 18:14
To: [log in to unmask]
Subject: Re: [SPM] Compare 2 methods to extract %signal change in MarsBaR

Hi Matthew,

Excellent to get your clear explanation!!

So I can use Method 2 to extract Beta values in a group of subjects, right??

One more question: Since %SC and Beta value can both represent neural response in an ROI, which one is better?......I mean, if I want to extract %SC, I have to use Method 1 which will be performed subject by subject and time-consuming (or any time-saving way to get %SC within many ROIs in a group of subjects)?

Thanks indeed!

Carlos


<spm_cond_units.m><spm_eeg_invert.m><spm_eeg_lgainmat.m>