[FieldTrip] Head movement correction for source time-courses
anne.urai at gmail.com
Thu Sep 24 21:35:31 CEST 2015
If you want to regress out single-trial variance that's explained by head motion see this tutorial: http://www.fieldtriptoolbox.org/example/how_to_incorporate_head_movements_in_meg_analysis, I think it's equally valid to do this at the sensor vs source level.
Another method that is described in Stolk et al http://www.ncbi.nlm.nih.gov/pubmed/23246857 and originally Uutela et al http://www.ncbi.nlm.nih.gov/pubmed/11707098 is to incorporate changes in the grad structure into you source reconstruction procedure.
This is implemented in ft_headmovement, but I've personally not found it very intuitive to use - perhaps one of the authors could elaborate? Perhaps it's worth adding a short section on the wiki on this..
Anne E. Urai, MSc
PhD student | Institut für Neurophysiologie und Pathophysiologie
Universitätsklinikum Hamburg-Eppendorf | Martinistrasse 52, 20246 | Hamburg, Germany
On 24 Sep 2015 at 14:46:34, Fritsche, M. (Matthias) (m.fritsche at student.ru.nl) wrote:
I started looking into offline head movement correction of MEG data wondered about the following:
Is it possible/sensible to correct time-courses of source reconstructed data for head movement (that is timepoint-by-timepoint for individual sources)?
The GLM approach with ft_regressconfound only computes trial-by-trial power estimates, whose variance due to head movement is removed, right? Is there a theoretical consideration that renders the timepoint-by-timepoint correction for individual sources pointless? Or is there maybe a technical limitation, e.g. in terms limited computation resources?
I couldn’t find any information on this on the wiki, or the mailing list archive. Please excuse if I should have overlooked something or if this is a silly question.
fieldtrip mailing list
fieldtrip at donders.ru.nl
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the fieldtrip