[FieldTrip] ROI selection of beamformer grid points
"Jörn M. Horschig"
jm.horschig at donders.ru.nl
Wed Jun 26 15:26:41 CEST 2013
I basically use two approaches (in the end, both failed, so any other
hints are appreciated):
(a) Select voxels purely based on anatomical labels, as found in an
atlas or in literature.
(b) Select voxels based on some local maxima or minima, e.g. power
maximum or maximum difference of log-ratio
(a) should be pretty straight forward. In essence it involves getting
MNI coordinates, inversely warping your grids to MNI space, getting
closest voxel. If you have your region of interest not in MNI
coordinates you need to transform them. I found some tal2mni functions
on the web for this, but note that this is just an estimate. Of course,
(a) is also applicable if you have a localizer task using fMRI and want
to focus on some localized voxels.
(b) is a bit more tricky, because you might be faced with huge
inter-subject variability. Best of course would be to have the
subject-specific, fMRI localized voxel. What I done in the past is to
define a rough region of interest, e.g. posterior neocortex (based on
some quick&dirty coordinate thresholding), using ft_volumesmooth to
apply a gaussian blur on single subject-activity and then select the
voxel that suits me best (i.e. the one of maximum activity). Of course
your ROI could also be based on the grand-average or what have you. I
had the feeling that especially this latter approach (base ROI on GA +/-
3 cm, smooth individual subject data, select most sensitive voxel)
worked quite well, but I cannot tell for sure, because in the end my
results were not reliable enough.
Oh and btw, if the question just aims on 'how' to select
programming-wise: Match the coordinate with your template, store the
index based on the template-grid and use this index on your
subject-specific grid to get voxel of interest in subject-specific
On 6/26/2013 3:07 PM, Marieke van de Nieuwenhuijzen wrote:
> Dear Fieldtrippers,
> I am running my analyses on time courses reconstructed in source space. Basically, that means that my working dataset is a matrix of grid point x time. What I want to do now is do some analyses on a subset of that dataset, a bit analogous to selecting some sensors to restrict analyses to. Therefore, what I would ideally want to do, is select a subset of grid points corresponding to a specific location (for example only the occipital grid points, or only the grid points corresponding to a specific atlas label).
> Does anyone have any suggestions about how I should go about selecting specific grid points? Is there perhaps some grid based atlas, or is it possible to select grid points based on their corresponding mni coordinates which you get after running ft_sourceinterpolate and ft_volumenormalise (in other words, is it possible to reverse ft_volumenormalise and ft_sourceinterpolate to map the mni coordinates to the grid points instead of the grid points to mni representation).
> Any pointers would be much appreciated.
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
Jörn M. Horschig
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognitive Neuroimaging
Radboud University Nijmegen
Neuronal Oscillations Group
FieldTrip Development Team
P.O. Box 9101
NL-6500 HB Nijmegen
E-Mail: jm.horschig at donders.ru.nl
Trigon, room 2.30
NL-6525 EN Nijmegen
More information about the fieldtrip