> If MNI coordinates are in fact in units of mm, does SPM account for voxel
> sizes and shapes when displaying voxel coordinates?
> More specifically, does a common MNI coordinate (5 -2 3) correspond to the
> same geometric position for a common reference original point when different
> voxel sizes are used, e.g. [2 2 2] and [3 3 3.75]. If not, how do we deal
> with this issue.
There is a 4x4 affine transtorm matrix in the headers, that allows you
to go from i,j,k indices to x,y,z mm coordinates. Obtaining the x,y,z
coordinates is simply a matrix multiplication:
xyz = M(1:3,:)*[i j k 1]'
>
> This has puzzled me for quite some time...
>
> For Bas,
> 1) We might have diverged understanding of the concepts of "voxel space"
> and "mm space ( which I assume is some sort of stereotactic spaces such as
> MNI) ".
mm space may refer to anything you like. For spatially normalised
images it is the coordinate in MNI space. For NIfTI images just
converted from DICOM, it is the position within the scanner. For
coregistered images, it is something slightly different, and is used
for mapping between voxel indices in two images.
ijk2 = inv(M2)*M1*[i j k 1]';
ijk2 = ijk2(1:3)
>
> For example, let's say that a particular image has dimension of [79 95 68].
>
> To what I know, its "voxel space" ranges from 1 to 79*95*68
> or alternatively, from (0, 0, 0) to (79, 95, 68). In other words, the
> coordinates of a certain voxel in "voxel space" are in fact the matrix
> indices of that voxel.
For your example, i would go from 1 to 79, j from 1 95 and k from 1 to 68.
>
> If I am correct, then coordinates in "voxel space" must be integers, since
> it is impossible to have floating point numbers as matrix indices. But I
> found floating point numbers, e.g. 29.6 19.1 33.3 when displaying such
> images. So where did I go wrong?
The centres of voxels are integers, but in Display, Check Reg etc, it
is possible to click between the voxel centres.
>
> 2) One more thing, I assume that V.mat is the affine transformation matrix
> used in normalising the raw images to MNI space. So I don't see why and how
> it can be applied for the purpose of voxel to mm conversion...
V.mat is the affine transform matrix referred to previously (ie M).
Best regards,
-John
> ----- 原始邮件 -----
> 发件人:Bas Neggers <[log in to unmask]>
> 收件人:<[log in to unmask]>
> 抄送人:<[log in to unmask]>
>
> 主题:Re: [SPM] Correspondence between mms and voxel coordinates
> 日期:2012年06月27日 16点53分
>
|