Stream: implementers
Topic: FHIR representation of dermoscopic examination
Joakim Antus (Mar 10 2022 at 09:51):
Hi, what would be a preferred way to represent a dermoscopic examination consisting of multiple images using FHIR R4?
We already use Observation to capture laboratory and vitals data. And I have noted that Observation.category has a code for imaging, however I can't seem to find a satisfying solution for our dermoscopic examinations.
One solution I'm looking at is to use Observation.derivedFrom with references to images using either DocumentReference or Media.
Another solution would be to use a valueAttachment extension and place all images in Observation.component.
Or am I looking at this the wrong way and should use Procedure or ImageStudy instead? Or something else?
Also, I have very limited experience using DICOM and would prefer to avoid it.
Any suggestions are very much appreciated!
Lloyd McKenzie (Mar 10 2022 at 15:46):
In R4, images are typically captured as Media as it has the descriptive elements around the image. In R5, that moves to DocumentReference.
Joakim Antus (Mar 10 2022 at 16:10):
OK, would Observation.derivedFrom be appropriate for representing the collection of all Media resources created during the examination? Or is there a better solution for representing this collection of Media resources?
Elliot Silver (Mar 10 2022 at 16:11):
What format are the original images in? If they are not in DICOM, @Lloyd McKenzie's suggestion is good, if they are in DICOM, it's a little more complicated.
Elliot Silver (Mar 10 2022 at 16:12):
derivedFrom documents which content the observation is derived from, not content that is derived from the observation.
Elliot Silver (Mar 10 2022 at 16:16):
Generally, you'd find all images from a procedure by looking for all Media.basedOn or DocumentReference.basedOn the ServiceRequest that Procedure is also basedOn.
Kai Kewley (Snowstorm) (Mar 10 2022 at 20:24):
To represent the "dermoscopic" aspect there is a concept "427690002 |Diagnostic dermatoscopy of skin (procedure)|" in SNOMED ( https://snomed.info/id/427690002 ). Not sure if that can be used or a good idea, just commenting.
Joakim Antus (Mar 10 2022 at 21:51):
The images will be in a non-DICOM format, most likely JPEG.
Using Media.basedOn might work in most situations, however, we are targeting primary care and there might be situations where an dermoscopic examination has to be performed without an underlying ServiceRequest to base it on. So, how would one represent a collection of Media resources in that situation.
Only referencing the Encounter won't be viable either, since some patients have multiple skin moles that needs examining and we must be able to identify what images belongs to what mole.
Elliot Silver (Mar 10 2022 at 21:55):
You can tie the Media or the Observation to the mole with the standard extension body-site, which lets you specify a BodyStructure associated with the image/finding.
Joakim Antus (Mar 11 2022 at 10:31):
Thank you for all the help and different suggestions.
After evaluating the different options, we've decided to go by creating a new ServiceRequest on the fly, set the ServiceRequest.intent to proposal and ServiceRequest.requester to the Practitioner and then reference that ServiceRequest in all Media resources using Media.basedOn.
John Moehrke (Mar 11 2022 at 12:53):
note if you are just getting started with Media... you might want to use DocumentReference to start with... this because DocumentReference supports the Media usecases, and in R5 Media goes away.
Joakim Antus (Mar 11 2022 at 13:00):
Great point! Yes, we are just getting started so will use DocumentReference instead of Media :thumbs_up:
Last updated: Apr 12 2022 at 19:14 UTC
