+http://groups.google.com/group/comp.protocols.dicom/browse_thread/thread/50bd859e0232a735/6040b545240c8fa4
+Quote:
+For completeness, there are three ways an application can
+correlate CT and PET (or other sets) using DICOM only:
+
+1) Code all the images with the same Frame of Reference UID.
+At least one major vendor does this now.
+
+2) Make assumptions based on specific implementation behavior
+based on data elements that match (e.g. perhaps Study Time, or
+matching values in the General Equipment Module).
+Without actual trial and experience verifying the correlation,
+I agree, that an assumption is not wise
+
+3) Use the recent Supplement 73 Spatial Registration Storage SOP Classes.
+This capability allows explicitly specifying the spatial transformation
+from one Frame of Reference to another. However, this supplement
+is probably too recent to be in use.
+
+[Is there a SOP class for CT-PET images?]
+http://groups.google.com/group/comp.protocols.dicom/browse_thread/thread/bb96abb730624a8a/5a78f1ff0199a727