Dear Mark,
thanks for your reply and explanation.
On Friday 09 September 2005 14:24, Mark Jenkinson wrote:
> Hi Martin,
>
> That matrix is guaranteed to map the voxels outside of your FOV.
> The translation values in the fourth column always need to
> be the opposite sign to the main diagonal value, since at the
> moment your new x coordinate (for instance) is given by:
> xnew = -1*xold - 99.618134
> As the coordinate conventions in flirt always use scaled voxel
> coordinates (voxel coord * voxel dimension) they are always
> positive, hence your xnew will be negative here and so outside
> the FOV.
>
> I don't know how you got this matrix, but it seems that you are
The matrix transforms the acquired image from the scanner image space to the
physical patient space. That is, the scanner knows exactly, where the FOV
starts in the physical space of the operating room (I am dealing with
interventional MRI here). I have set of four matrices, which should transform
transform images from space A to space D and I thought to use applyXFM to
test the correctness of the transform and convert_xfm to concatenate the
matrices A...D.
> using the origin in some way, however, the coordinates that
> flirt deals with internally (and relate to the matrix) do not care
> about the origin value, and actually ignore it. The mm coordinates
I think this is the crucial information for me that internally the image
origin is not taken into account in FSL. I'm afraid that in the context of my
work it is essential, isn't it? But I should still be fine convert_xfm for
concatenation of the matrices, shouldn't I?
> shown by fslview do use the origin, but not the flirt matrices.
>
Thanks again and have a nice day.
Martin
|