AW: [FIELDTRIP] TFRs on virtual source data, Neuromag 122

Nina Nina.Kahlbrock at UNI-DUESSELDORF.DE
Wed Apr 28 15:05:40 CEST 2010


Hi Nathan,



thank you very much for your response!

Performing source analysis on data not normalized to the template brain now
gives me reasonable results. I will however, try to figure out, what I was
doing wrong with the normalized data (even though plotting of my virtual
grid looked fine there, too).



Thanks again!



Nina





  _____

Von: FieldTrip discussion list [mailto:FIELDTRIP at NIC.SURFNET.NL] Im Auftrag
von Nathan Weisz
Gesendet: Dienstag, 27. April 2010 15:41
An: FIELDTRIP at NIC.SURFNET.NL
Betreff: Re: [FIELDTRIP] TFRs on virtual source data, Neuromag 122



hi nina,



did you validate the position by plotting them on the individuals MRI?

I mean the ones you define in xgrid, ygrid and zgrid.



when i project my data into source space to do e.g. time-frequencystuff i
do:

cfg=[];

cfg.vol=vol;

cfg.grid.pos=rois; %%here is a difference to your code

cfg.grad=data.grad;

cfg.channel={'MEG'};



grid=prepare_leadfield(cfg);



then i continue with the sourceanalysis stuff.



good luck,

n



On 27.04.2010, at 14:23, Nina wrote:





Hi all,



I am working on constructing a virtual sensor for my 122 Neuromag MEG data
in order to compute time frequency analysis on this virtual sensor.

The results I am getting so far are not in line with the sensor level data.
On sensor level I see a very clear gamma response. It is also present on
source level in the frequency domain, using dics, however, the time course
of the source is not as strong and frequency confined on source level. Does
anybody have any ideas on what I might be doing wrong?

In the following, I would like to explain what I have done so far and what
my problems are. I am attaching the script I used in text format.



About the data set:

It contains trials of three different conditions. These trials are of
variable lengths (spread evenly over the conditions) and there are not
always the same number of trials in each condition.



What I have done so far:

1.	based on TFRs on sensor level I chose each subject's strongest gamma
frequency
2.	for each subject, I took this frequency (+/- 5 Hz) and calculated
spatial filters for stimulation and baseline periods, averaged over all
three conditions using a DICS beamformer
3.	for each voxel, the ratio of poststimulus power to prestimulus power
was computed
4.	from that I took the voxel with maximum power increase and used it
as my voxel of interest,
5.	for this voxel of interest I calculated a new dipole grid with only
one voxel. It is in the same location as the strongest voxel from step 3.
6.	then I went back to my functional data and used the FT function
'timelockanalysis' to compute the covariance matrices for all my sensors and
trials (keeping single trials), trying different time windows for covariance
computation, but always calculating power for the whole time period:

a.	pre stimulus [-2 0] (but using the  whole trial for
timelockanalysis; time = [-2 3],
b.	post stimulus [0 3] (but using the  whole trial for
timelockanalysis; time = [-2 3],
c.	the whole time period [-2 3],
d.	pre stimulus [-2 0] (using only that time window for
timelockanalysis; time = [-2 0],
e.	post stimulus [0 2] (using only that time window for
timelockanalysis; time = [0 2]

7.	the covariance matrices were put into source analysis, again
computing spatial filters for the voxel of interest (using rawtrial = 'yes')
8.	NaNs, that were due to different lengths of trials, in dipole
moments resulting from source analysis were removed
9.	then I put the resulting dipole moments of the three directions
(x,y,z) into a structure that resembles that of preprocessed data
10.	time frequency representations of power were calculated using a
multitaper approach
11.	When looking at the three directions (x,y,z,) separately in a time
frequency plot, this gives me

a.	somehow meaningful results for the covariance window being pre
stimulus (5.a), however, they are a lot weaker than on sensor level.
b.	no meaningful results for the covariance window being post stimulus
(5.b) or the whole time period (5.c)

12.	for 5. d/e relative changes to baseline were calculated for each of
the trials

a.	this gives me somehow meaningful results, but very weak and not
constrained to the before found frequency ranges



Does anybody have experience with this kind of analysis? Do you have any
suggestions about which step might be causing these troubles?



Thank you all in advance for any help!



Nina



----------------------------------

The aim of this list is to facilitate the discussion between users of the
FieldTrip toolbox, to share experiences and to discuss new ideas for MEG and
EEG analysis.

http://listserv.surfnet.nl/archives/fieldtrip.html

http://www.ru.nl/fcdonders/fieldtrip/

<virtualsensor.txt>



----------------------------------

The aim of this list is to facilitate the discussion between users of the
FieldTrip toolbox, to share experiences and to discuss new ideas for MEG and
EEG analysis.

http://listserv.surfnet.nl/archives/fieldtrip.html

http://www.ru.nl/fcdonders/fieldtrip/


----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20100428/7e87b7f8/attachment.html>


More information about the fieldtrip mailing list