Page 606 |
DICOM PS3.17 2020a - Explanatory Information |
The scope of referential integrity required is defined to be the Patient. Instances in one Study may be referenced from another (e.g., as prior images).
KKK.7 Persistence and Determinism
The rules for conversion specify that the SOP Instance and Series Instance UIDs of converted images be changed, and that the same UIDsbeusedeachtimethataqueryorretrievalisperformed.Thestrictseparationofthetwo"views"ofthesameinformation,coupled with the "determinism" that results in the same identification and organization of each view every time, are required for stability across successive operations.
Were this not to be the case, for example, the results of a query (C-FIND) might be different from the results of a subsequent retrieval (C-MOVEorC-GET),orforthatmatter,successivequeries.Further,referencestospecificinstanceUIDineitherviewmayberecorded in external systems (e.g., in an EMR), hence it is important that these remain stable and accessible.
This places a burden on the Q/R SCP to either retain a record of the mapping of UIDs from one view to the other, or to use some deterministicprocessthatresultsinthesameUIDs(onecouldenvisagesomehashingscheme,forinstance).Howthisisimplemented is beyond the scope of the Standard to define. The determinism requirement does not remove the uniqueness requirement; in partic- ular it is not appropriate to attempt to derive new UIDs by adding a suffix to a UID generated by a different application, for example.
There is no time limit placed on the determinism; it is expected to be indefinite, at least within the control of the system. This is a factor that should be taken into account both in the design of federated Q/R SCPs that may integrate subsidiary SCPs that support this mechanism. It should also be considered during migration to a new Q/R SCP, which ideally should support the mechanism, and should support the same mapping from one view to another as was provided by the Q/R SCP being migrated. This may be non-trivial, sincethealgorithmforconversionmaybedifferentbetweenthetwosystems.Itmaybenecessarytodefinesomepersistent,standard, serialized mapping of one set of UIDs to the other.
KKK.8 Source References
It is also useful to save references in converted SOP instances to their source. Accordingly, converted instances are required to contain such references, both for image conversions as well as for ancillary instances that may be updated, such as Presentation States and Structured Reports.
Obviously, the references to the source instances for the conversion are excluded from conversion themselves. If the instances have been converted on different systems, however, there is a possibility that the source references will be "replaced" and a record of the "chain" of multiple conversions will not be persisted.
There is no mechanism to define forward references in the source to the converted instances, since that would imply changing the source instances from their original form, and while this is acceptable within the scope of the normal "coercion" that a Storage SCP is permitted to perform, it is probably not sufficiently useful to justify the effort. This does imply some asymmetry however, depending on the direction of conversion (classic to enhanced or vice versa); only one set will contain the references.
In performing round trip conversion, without access to the source instances, the referenced source UIDs can be used as the UIDs for the newly created converted instances.
KKK.9 Uncertainty Principle
When does a converted view come into existence? By definition, when it is "observed". However, a practical question is when to start conversion. A Study is never, theoretically, complete, yet the semantics for conversion and consistency are defined at the Study level.
Another practical question is whether or not to make the received instances available, even though the converted ones may not yet have been created.
In the absence of the concept of "study completion" in DICOM, no firm rules can be defined. However, in practice, most systems have aninternal"completion"concept,whichmayormaynotberelatedtothecompletionofthePerformedProcedureStepsthatarerelated to the sets of instances in question, or may be established through some other mechanism, such as operator intervention, possibly via a RIS message (e.g., after QC checks are signed off as complete, or after a Study has been declared as "ready to read").
Asystemmayelectto"dynamically"beginconversionasinstancesarriveandupdatetheinformationintheconversionasnewinstances are encountered, or it may wait until some state is established that allows it to perform the conversion "statically". In either case, the informationintheconvertedviewviathequery/retrievalmechanismsshouldbeimmutableoncemadeavailable.I.e.,onceaconversion