volumewrite, volumenormalize and mri orientations

[Jan Mathijs Schoffelen]

Dear Marie,

I hope that Tom's comments have been helpful. On the other hand, you mention
that you have problems with volumenormalise, and since the normalization is
one of your goals (and a prerequisite for labeling the cortical areas) I
think it would be nice to debug your problem and enable you to stay within
fieldtrip.

> I've tried both 'analyze' and 'spm' filetype options with quite
>varying results (see attached pdf, method spm is first). But as you
>can see neither matches the default spm orientation found in T1.mnc
>(third).


Regarding the results you sent along, it makes a big difference whether the
visualization tool just stupidly plots the block of data on the screen, or
takes the coordinate system into account. The coordinate system is defined
by an affine transformation matrix, and tells you how to toggle between
voxel-coordinates and head-coordinates. An original analyze-file does not
have an affine matrix attached, so any visualization tool will stupidly plot
the block of data on the screen. Spm-based analyze files have the matrix
hidden somewhere in the header.
Anyway, in order to be able to compare the different outputs, it is crucial
that the volumes are defined in the same coordinate-system, thus being spm,
rather than ctf based (because this will prohibit comparison with the
template).
The plotting tool you used apparently takes the affine-matrix into account
(if present, so the second one is just the dumb plotting of the block of
data), but in addition assumes the analyze xyz-convention which I mentioned
last time. Since the ctf affine matrix allows you (or the plotting software)
to interpret the voxels in terms of head coordinates according to ctf's
convention, this gives you the volume to be rotated by 90 degrees along the
z-axis.

>I also had problems using volumenormalise inside fieldtrip and got
>some very strange results again I think due to a mis-match in
>orientation between my mri + functional and the T1 template. As I
>said yesterday, when I use sourceplot the 3 slices seem to match the
>fieldtrip norm (as depicted in the beamformer tutorial), but checking
>the .anatomy field of my sourinterpolated data has x running from
>right to left, y from front to back and z from top to bottom which
>does not match the ctf system co-ordinates. I am not doing anything
>to the mri file except loading it in during sourceinterpolate and I
>am pretty sure it was created correctly in MRIViewer.

Importantly, there is a difference between the 'x', 'y' and 'z' axes in the
sourceplots, since this function 'stupidly' plot the block of data, and is
thus voxel-coordinate based, rather than head-coordinate based. So if the
anatomies look like those in the fieldtrip tutorial I would be curious to
see where your normalization fails.
Did you try volumenormalise with cfg.nonlinear = 'no'? This will cause the
normalisation to be only a linear one, and makes it run faster, and works
typically good enough.

Yours,

Jan-Mathijs


----------------------------------
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/fcdonders/fieldtrip.