Develop methods for validating the DICOM SNOMED mappings

[gaurav]

David Clunie identified some mappings that are incorrect (quoting from his comment: the matches between "abdominal structure" and "radiating chest pain in abdomen", or "abdominal artery" and "abdominal aorta artery", or "instrument, device", versus "patient-reported survey instrument").)

We should find some ways of validating the mappings we've generated, hopefully without having to manually check all the mappings. One broad validation method could be by identifying the classes of things that the subject and object of the mappings are -- if it subject turns out to be a substance while the object turns out to be a method, for instance, it's a good bet that the mapping was incorrect (we might be able to do this using the NCImt hierarchy in cancerDHC/tools#20). We could also compare our mappings to other mappings (#21, cancerDHC/tools#23).

I'm creating this issue to track all the different approaches to doing this. We can close it once we have a solid approach for validating these mappings.