Hi Andriy,
Since you worked on MultiVolume modules, I am assigning this issue to you.
Let me know what you think.
Jc

Cross-reference to forum exchange.

http://massmail.spl.harvard.edu/public-archives/slicer-users/2013/006470.html

I think this distinction is handled somewhere in the guts of ArchetypeVolumeReader already. It does not rely on metadata, I think RGB, DF and multivolume are already handled differently. We can add metadata, but these files will be generated by tools other than Slicer, so we will need some heuristic anyway.

I think someone who worked with that part of Slicer would be able to provide a more authoritative comment (e.g., Steve or Jim).

The nrrd spec does a nice job of supplying the metainformation needed to interpret the data. Some other formats also have enough information (nifti, for example). Sometimes though there may be multiple valid interpretations and the user will have to pick.

We probably want to define a nhdr fields for our canonical representation for each use case so they can be read back unambiguously. Other formats will just need to be handled with importers or dialogs.

The translation from file representation to Slicer interpretation is pretty complicated. We cleaned it up a good bit last summer which allowed us to introduce a MultiVolumeReader that is really just a subclass of one of the NRRD readers.

1. We do use nrrd attributes to represent MultiVolumes. And of course NRRD attributes for DWI, DTI, etc.

2. The issue I usually come against is that Slicer likes to interpret 3 component vector images as RGB. For the DCE MRI case, our introduction of the MultiVolumeReader addressed this.