what kind of coordinate system ImageVis3D uses?

ImageVis3D simply maps the local grid coordinate axes to world space axes. So far we ignore any transformation stored in DICOM (or any other format such as NRRD, except for scaling). As long as we were only supporting a single volume this did not matter much but with multiple datasets it does matter I agree. For geometry you can already specify any affine transformation (simply use the transform button in the document explorer).

The way most software packages address this issue is by parsing the vendor and machine strings from the DICOM file and then look into a table where that particular scanner stores the correct information. As this results in very ugly and hard tom maintain code we decided a heuristic to figure out which value has the most appropriate spacing (for instance we prefer un-isotropic over isotropic) but I'm sure some vendors mess that up so hard that the heuristic fails :-). 

But nonetheless you may also simply be hitting some bug that I'm not aware of in out DICOM reader and that is the reason for seeing the wrong spacing, I'll take a look as soon as I find a moment. Can you tell me what you think the z-Spacing for those datasets are supposed to be?

Cheers

   Jens