Yes I understand your points perfectly and agree with you completely.
The reason I asked the question is that the fsl web page claims that the
fslview displays images in radiological convention. As soon as you state
this, some things have to change on display as the directional vectors
change. I did not observe this behavior and was forced to ask the
Now, I understand that the statement of "FSLView always shows images in
radiological convention." is a design goal which is not fully supported
yet. I just had to know that. I was worried that I misunderstood
something in NIFTI specifications.
Thanks for your help
On Thu, 21 Oct 2004, Graham Wideman wrote:
> Couple of points, neither specifically related to NIFTI, but perhaps helpful.
> 1. What do you mean by the orientation of your data "corresponding to
> the neurological one."
> Strictly speaking, this is a little off-target -- neurological vs radiological conventions are *viewing* issues, not data-order issues.
> It is up to the viewer software to present the data to you in
> one or other of those preferred views regardless of order of data in the file.
> That said, it does appear to be the case that due to intractable viewers in the past, some data users have developed customs where they rearranged the order of data in their data files to get it to show in the viewer the way they want --
> but that's "fixing" the problem by introducing another problem.
> In NIFTI, you can store the data in any of 48 different orders, which might
> be described as (L-R, S-I, A-P) or (I-S, P-A, L-R) and so on (48 different permutations).
> But so long as the NIFTI header info describes the image data order
> correctly, you as the user would be unaware of the data order because the
> viewer should simply present it correctly. And if that viewer is set to show
> data in radiological view, it will do that regardless of NIFTI data order.
> 2. In order to troubleshoot these puzzles it can be useful to have a
> viewer that knows nothing about header info, and simply shows you the raw
> data in its raw order. One such example is available at my site:
> 3. And obviously the above story is further modified for viewers that can rotate and interpolate in 3D, and are not merely showing slices along one of the three "data axes".
> Hope that helps,
> At 10/20/2004 09:56 PM, you wrote:
> >I have created an img/hdr NIFTI pair with orientation corresponding to
> >the neurological one. Because such files look like analyze pair, the old
> >and the new FSLview can display the images. I would, however, expect
> >different image orientations because analyze coordinate system is fixed
> >where as NIFTI supports directional cosines.
> >What bothers me is that old FSLview (1.0 beta 7) and the new one (2.3.5)
> >produce exactly the same displays.
> >Are not I correct in my expectations? I do not understand what I am
> >Thank you for your kind help,