I can't quite figure out how to properly create DICOM NM SPECT (TOMO)
files with multiple detectors and the standard appears to be quite
fuzzy in that regard.

Suppose I had a gantry with two opposing detectors which had different
physical properties and radii and I was required to tell for each
frame if it came from detector one or two and which radius the
detector had in that position.

Since the detectors are logically different I set "Number of
Detectors" to 2 and create two Items in the "Detector Information
Sequence".

After this it gets fuzzy. The radii (according to the current
standard) are to be encoded in the "Rotation Information Sequence"
which is supposed to contain a single Item. The sequence contains
several attributes ("Start Angle", "Radial Position" etc.) which refer
to "the detector" (which one? the gantry as a whole?).

Radial Position can contain one average value or one value per angular
view.
How is angular view defined though?

Is there one angular view per gantry step, or one angular view for
every angle any of the detectors did take?
What if two detectors took the same angle (with different apertures)
but had different radii?

I've seen a couple of NM datasets from different vendors which don't
quite seem to know how either;
putting the radii into the "Detector Information Sequence" (where they
don't belong (according to the standard)),
putting only the first detector's radii into the "Rotation Information
Sequence" or putting both detectors' radii into
the "Rotation Information Sequence" (yet "Number of Frames in
Rotation" was only half the value of "Number of Frames").

I would be very grateful if anyone could shed some light on this
matter.

Thanks in advance,
Nils Gladitz