Stream: implementers
Topic: Is Media long for this world?
Elliot Silver (Dec 09 2016 at 21:45):
As part of the Media and Imaging thread (https://chat.fhir.org/#narrow/stream/implementers/subject/Media.20.26.20Imaging.20Study.20Resource ) I've been looking at Media. The scope includes "Photos of patients and staff for identification purposes". Cool, lets see how it's used in Patient...Oh, it isn't. Patient uses an attachment for the photo. OK, well, what about another use? The scope also includes "Photos and videos of diagnostic or care provision procedures for recording purposes." That suggests photos and videos as an Observation, which is supported by another thread (https://chat.fhir.org/#narrow/stream/implementers/subject/Where.20to.20best.20place.20clinical.20photos.3F). So, let's go look at Observation... Nope, that uses an attachment as well. Well, then surely the Media resource should tell me all the many places it is used...yes, its used in DiagnosticReport, and...well, DiagnosticReport.
Is there a reason for the poor adoption of Media? Is there a long term vision for the resource?
Grahame Grieve (Dec 09 2016 at 21:46):
I think that patient.photo and Observation.attachment predate Media, and at least Patient.photo should have been changed to use it.
Grahame Grieve (Dec 09 2016 at 21:47):
what to do about observation.value is less obvious - maybe Media should be a choice in Observation.related.target
Grahame Grieve (Dec 09 2016 at 21:47):
and/or we could make it clear how to reference a Media resource in Observation.valueAttachment
Elliot Silver (Dec 09 2016 at 21:50):
Observation.valueAttachment.url contains the URL for a Media resource, which contains an attachment which contains the photo?
Grahame Grieve (Dec 09 2016 at 21:53):
y. one way to do it.
Elliot Silver (Dec 09 2016 at 21:54):
Is there any reason not to add Media as an Observation value?
Elliot Silver (Dec 09 2016 at 22:00):
OK, added GF#12454 to convert Patient.photo to Media.
Grahame Grieve (Dec 09 2016 at 22:00):
yes, there's technical reasons not to add Media as an observation.value. It's a reference, not a datatype.
Grahame Grieve (Dec 09 2016 at 22:01):
something to think about
Elliot Silver (Dec 09 2016 at 22:20):
Aren't there other resource that have choices between datatypes and references? Provenance.who[x] is uri or Reference(...).
Lloyd McKenzie (Dec 09 2016 at 23:34):
Yes, but having a resource be the result of an Observation is a slippery slope we might not want to tread on
Brian Postlethwaite (Dec 10 2016 at 05:49):
I was planning to use Media for things like wound management, or photo's of client's homes etc for community maintanance purposes.
Brian Postlethwaite (Dec 10 2016 at 05:49):
Don't know that I'd want the current Client Photo to be Media resource, an extra thing to go get. And also this would then have both resources pointing at each other, not ideal either.
And do want to know which one is current.
Eric Haas (Dec 10 2016 at 20:49):
Not crazy about adding media to observation. already have it in DR
Elliot Silver (Dec 12 2016 at 17:38):
Media referenced from the DR seems appropriate for an image included in the report, but what is the correct way to record clinical imaging that doesn't end up in the report? I imagine many more images are taken then end up being referenced directly in a report.
Using media for wound management is not great because Media is missing most of the context needed, e.g. what is the encounter or episode of care? order number? body part? Is Media an Event? If so, where are the other event-like elements (e.g. basedOn)?
As it is, Media either needs more clinical information, so that it can stand on its own, or needs to be referenced from something that can convey that clinical information.
Elliot Silver (Dec 12 2016 at 17:40):
I'd argue the case is at least as strong for a large object like a photo not to be included in the Patient resource, as for it to be included directly, so I don't buy that argument. I do agree that the Patient to Media circular reference is an issue.
Eric Haas (Dec 12 2016 at 20:00):
Observation + valueAttachment gives you all the stuff you mentioned. Observation is an Event resource and still need to do that reconciliation wiht the logical model for things like basedOn. This bring us full circle to your original question. What does Media give that Observation + valueAttachment doesn't. more metadata...
Elliot Silver (Dec 13 2016 at 19:39):
Or, poking at this from the other end: What does Media give that Binary doesn't? Is knowing the dimensions/playing length useful, or does that just duplicate information in the actual binary content? Media has a Patient reference, but is there a reason why Binary can't/doesn't? Is Media better achieved through core extensions on Binary?
Grahame Grieve (Dec 13 2016 at 20:07):
you can search for the metadata things in media easily.
John Moehrke (Dec 13 2016 at 21:15):
Media has a subject , thus allowing for patient specific data. Binary might have patient specific data, but that fact must be learned by the relationships the Binary has with a metadata resource like DocumentRefererence, or other.
John Moehrke (Dec 13 2016 at 21:17):
There are other useful meta elements in Media, including those found in Attachment datatype. But they are less compelling to me than the subject element.
Elliot Silver (Dec 13 2016 at 21:32):
"Media has a subject, thus allowing for patient-specific data" -- is there a reason Binary doesn't have patient? Can the two be collapsed into one resource? Or, do we add more to Media making it "ClinicalMediaObservation" and put space between the Binary/Media resources that way?
John Moehrke (Dec 13 2016 at 21:55):
Binary is intended for purely binary content. "This resource is generally used as the target of a Document Reference or an Attachment, when a FHIR server finds it convenient to manage the content within the same overall REST framework as the other resources." Binary is designed to be as thin as possible, holding only the content and the mime-type of that content. Binary will be returned to you as the binary content (not base64Binary encoded) if you can accept that binary content (MIME-TYPE). Hence if it is of mime-type pdf, and you GET it with accept headers allowing pdf, you will get only the pdf. This does not happen with Media.
John Moehrke (Dec 13 2016 at 21:55):
Media can point at Binary
John Moehrke (Dec 13 2016 at 21:56):
Media.content.url can be a Binary
John Moehrke (Dec 13 2016 at 21:57):
Of course under the turtles-all-the-way-down meme... Binary can also point at Media, but that is just silly.
Elliot Silver (Dec 13 2016 at 21:58):
Right -- forgot the weird Binary dual personality thing. OK, forget I ever suggested unifying Media and Binary.
Eric Haas (Dec 14 2016 at 16:10):
I guess my point was Obs+ valueAttachment gives you patient but leaves out dimension. So is Media really needed as a ref from Observation. ( see much earlier in thread )
John Moehrke (Dec 14 2016 at 17:06):
Yes Eric. Media has a reasonable minimal set of metadata for common photo quality media.
Brian Postlethwaite (Dec 15 2016 at 03:05):
I think that metadata on the media is appropriate, and beneficial over binary.
Last updated: Apr 12 2022 at 19:14 UTC
