Paul/Lisa -
There was a change in group inversion between older and current versions of SPM8, concerning the nature of the "realignment" of gain matrices across subjects. The older version (as described in Litvak & Friston, 2008, Neuroimage) worked reasonably well, but was sensitive to the choice of initial (referent) subject. The newer version developed by Karl (and as described in Henson, Wakeman, Litvak & Friston, 2011, Frontiers in Human Neuroscience) re-references to an "average" gain matrix across subjects, so should be largely independent of referent subject.
However, we too have noticed that the newer version sometimes fails when gain matrices are too dissimilar across subjects (aborting with 0 common modes, as you describe). Nonetheless, we have found that first projecting the sensor data to a common space (as is possible in the MaxFilter software for Elekta Neuromag systems, eg to "trans" to space of one subject, or to the default helmet space) solves this problem (at least in datasets we have tried so far), with minimal loss of spatial modes (eg maintaining 60-80 for Neuromag magnetometers or gradiometers).
I appreciate this may not help you (if you don't use MaxFilter), and the group optimisation of M/EEG spatial priors may require some further work, in which case you may want to stick with individual inversions.
Best wishes
Rik
---------------------------------------------------------------
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 x522
FAX +44 (0)1223 359 062
MOB +44 (0)794 1377 345
---------------------------------------------------------------
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Paul Sowman
Sent: 27 September 2011 08:08
To: [log in to unmask]
Subject: Re: [SPM] MEG Group Inversion
Dear Vladamir, I too have had the same experience as posted by Lisa. You suggested that this was because the initial head positions across the subjects were too disparate. However, following Lisa's post I tried the same data set on an older SPM version and the group inversion seems to work whereas it doesn't on the latest version (0 spatial modes etc...). Is it possible that the latest version has tighter constraints on head position?
Regards, Paul
|