When images are displayed in SPM, there is usually a negative voxel size
shown. The voxels of proper Analyze images should be stored so that x
increases from right to left, y increases from back to front, and z increases
from bottom to top. The T&T system has the x going from left to right.
Therefore a flip is needed to map from the way the data are stored, to the
way that they are displayed. In order to do this, SPM needs to consider that
one of the voxel sizes is reversed. For reasons that are more to do with
psychology than maths or physics, I chose to show the left-right direction as
reversed.
Whether the co-ordinates should be flipped really depends on how the Analyze
images were created. Some labs store their voxels with x increasing from
left to right, in which case a flip would be the wrong thing to do. Labs
with their data stored in this order should have their defaults changed
(permenantly) in order to tell SPM the order of the voxels, and whether
flipping is necessary.
I'm not sure how you got your data into Analyze format, so can't comment on
the order that the voxels are stored and whether flipping should be done.
There is a little more info about this at
http://www.fil.ion.ucl.ac.uk/spm/software/spm2/#Compatability
and also on pages 10 and 11 of
http://www.fil.ion.ucl.ac.uk/spm/doc/books/hbf2/pdfs/Ch2.pdf
Best regards,
-John
> I was hoping that you might be able to give me some assistance on a
> technical problem I am currently having during the spatial pre-processing
> of fMRI data using SPM2.
>
> I have 2 data sets. The first of which (data set 1) appears to be being
> processed as expected. The dimensions of the 3mm anatomical slice used for
> co-registration and for determining parameters for normalisation are
> 64x64x64 (with a voxel size of 3x3x3). The dimensions of the fMRI data set
> are 64x30x64 (with a voxel size of 3x6x3). This produces a final
> realigned, normalised, smoothed image with a voxel size of -2x2x2.
>
> Images from the second data set (data set 2) however, are being initially
> displayed with dimensions of 64x64x64 (with a voxel size of -3x3x3) for the
> 3mm anatomical data and dimensions of 64x30x64 (with a voxel size of -
> 3x6x3) for the fMRI data set. The final realigned, normalised, smoothed
> images have a voxel size of -2x2x2.
>
> However as you can see from the images attached, the final images produced
> from data set 2 are smaller in comparison to those produced from data set
> 1, even though the voxel size apparently has remained the same.
>
> Initially, with data set 2, I changed the ‘resize {x}’ value to -1 to flip
> all of the original images to cancel the negative voxel size. However,
> this didn’t make any difference to the final realigned, normalised,
> smoothed images, which still have a voxel size of -2x2x2 and were still
> smaller on the ‘display’ than those from data set 1.
>
> As I am new to using SPM2, I cannot think of what the problem might be and
> why the final images are being produced with a negative voxel size, which
> seems to be independent of whether the original images (those before pre-
> processing) have a negative voxel value or not and why the images appear to
> be smaller on display.
|