Stream: implementers
Topic: example of imagingStudies with images?
Jose Costa Teixeira (Sep 21 2017 at 20:09):
Hi. Is there an instance of an imaging study with images? anyone know an example instance resource out there?
John Moehrke (Sep 21 2017 at 21:06):
Jose, The only way to do that is to have a PACS system available on the Internet. As the images are accessed by using DICOM web access methods. @Elliot Silver @Brad Genereaux right?
Jose Costa Teixeira (Sep 21 2017 at 21:16):
Can't a study have contained media?
Brad Genereaux (Sep 21 2017 at 22:35):
Can't a study have contained media?
Looking at the resource, http://hl7.org/fhir/imagingstudy.html, there is no attachment for an embedded binary.
Nate Walker (Sep 21 2017 at 22:39):
Is this not outlined by this?
ImagingStudy is used for DICOM imaging and associated information. Use Media to track non-DICOM images, video, or audio. Binary can be used to store arbitrary content. DocumentReference allow indexing and retrieval of clinical “documents” with relevant metadata.
Jose Costa Teixeira (Sep 22 2017 at 05:25):
Ok, perhaps stepping back: If i have a procedure and no PACS, perhaps I don't use DICOM, what resource should I use? I want to capture the date, patient, evt any relevant observations, and a bunch of key images or videos.
Patrick Werner (Sep 22 2017 at 08:37):
For photos and video you can use the Media resource, the photo/video itself would be contained as binary data or reference in the content of the Media resource.
Jose Costa Teixeira (Sep 22 2017 at 10:13):
Thanks. My question is what is the resource for the study which allows me to include media
John Moehrke (Sep 22 2017 at 14:44):
@Jose Costa Teixeira Use DocumentReference. DocumentReference is available for ANY kind of content, as long as you have a MIME-TYPE. It is, as XDS is, very liberal in the definition of 'document'. DocumentReference is NOT limited to CDA documents, nor clinical documents. It is there to provide metadata to ANY document.
Jose Costa Teixeira (Sep 22 2017 at 14:48):
Thanks John. OK, and what is the umbrella that bundles it all? like "while performing this procedure, we got the following outcome, the following text notes, and the following images". Perhaps a question to OO?
John Moehrke (Sep 22 2017 at 14:56):
Isn't that an Observation? I am not the expert there... I think you are asking a question that @Hans Buitendijk is working to resolve in OO... Umbrella in XDS equivalent is the DocumentManifest.. but I don;t think that is your question.
Eric Haas (Sep 22 2017 at 17:13):
Yes OO along with II is trying to sort out the Observation vs Media vs Binary vs DocRef and I think should add ImagingStudy. On another note I think that "document" needs to be defined in this use case I think we should be more liberal in Media like in DocRef too and not limit to audio, video and images those boundaries always seemed arbitrary to me
Jose Costa Teixeira (Sep 22 2017 at 17:22):
ok, but regardless of the media type, I need to have the token for the procedure and all the things hanging from it.
AFAIK, procedure only contains outcome, which seems kind of a final report.
so:
Eric Haas (Sep 22 2017 at 17:23):
I don't know what you mean
Jose Costa Teixeira (Sep 22 2017 at 17:25):
1. Shouldn't procedure allow to document things that happen during the procedure? Catheter labs, those images when people are putting pacemakers, surgeries... Limiting that to procedure.outcome seems limiting to the final result, judging from the definition.
2. Outcome (DiagnosticReport) (or whatever attribute would be) should support images, but also other stuff like media, no?
Eric Haas (Sep 22 2017 at 17:55):
It is currently framed in Procedure to use DiagnosticReport to tie all those things together. (OO has a bit of work to to closely review DiagnosticReport in the context of being a sxreport or dentalrep) @Michelle (Moseman) Miller will need to rustle up some use cases and examples for this...
Jose Costa Teixeira (Sep 22 2017 at 17:57):
Ok, then how can i submit my examples? simple case: capturing images during surgery
Eric Haas (Sep 22 2017 at 19:14):
DR --> Dr.image.link type= reference(Media)
Elliot Silver (Sep 22 2017 at 22:35):
@Jose Costa Teixeira, another option currently is to use Observation.valueAttachment and either include the content directly or by URL reference.
Your use case hits close to the heart of the OO/II discussion about the future of Media, Observation, ImagingStudy. I hope there is a better answer in the next few months.
Jose Costa Teixeira (Sep 22 2017 at 22:53):
I think Diagnostic Report seems closer to reality than Observation.
Jose Costa Teixeira (Sep 22 2017 at 22:53):
I will bring this up to OO
Eric Haas (Sep 22 2017 at 23:31):
@Jose Costa Teixeira is already a hot topic did you get the doodle poll from Hans?
Jose Costa Teixeira (Sep 23 2017 at 05:09):
@Eric Haas no, i did not receive. can you please fwd?
Jose Costa Teixeira (Sep 25 2017 at 19:18):
Just to wrap up: Seems that the best approach for now would not be imagingStudy, but ProcedureRequest.Report.
Report is a DiagnosticReport to which I can easily add images.
Jose Costa Teixeira (Sep 25 2017 at 19:23):
this left me with some open questions:
1. diagnosticReport.image called Image but it can also point to videos. Wondering if that is misleading.
2. ImagingStudy seems to be the RIS part of a study, and it seems that the link to the images would be via procedureReference, which then has a report which points to a diagnosticReport which contains images??
3. Is imagingStudy aligned with the workflow patterns and separate Request from Event? And considering the different request statuses that are basedOn each other? It seems to be a bit compressed, e.g. when I see reason = reason why it was requested. Should it not be Reason it was performed?
Elliot Silver (Sep 25 2017 at 19:40):
1. Yes, this should probably be renamed to media or similar, but that element might go away in light of the Media/Observation merge discussion. So I wouldn't do anything now, but please file a tracker item so it doesn't get lost.
2. ImagingStudy is the RIS part of a DICOM study. The link to the DICOM images is via the information (endpoints) in the ImagingStudy resource. For non-DICOM imaging, the images would be described by Media, or Observation, which would both link to the ProcedureRequest. (Hmm, neither Observation nor Media link to the Procedure, just the PR, that is a gap--either in my understanding or the resources--but a separate issue.) No DiagnosticReport is needed to associate the images with a procedure, for either DICOM or non-DICOM imaging.
3. ImagingStudy should be aligned with the Event pattern. If you notice discrepancies, please file a tracker item. I see we didn't align the definition of reason. But wouldn't the reason performed (in most cases) be that it was requested? I suspect that reason requested might be more useful. However, if you submit an issue, we'll look at aligning.
Jose Costa Teixeira (Sep 25 2017 at 19:41):
thanks!
Daniel Rutz (Sep 26 2017 at 11:46):
FWIW, in the ongoing discussion working through OO/II/etc. I think that we need to ensure that the Observation's value is not a media (attachment, etc.) and that the Observation.value is always a demonstrative statement about the patient (etc.). If that statement is made based on assessment of an image that should be in the .related. But yes, all of these points about the lack of clear boundary conditions and gaps for certain cases are things we need to deal with.
One other key point I want to keep is that I think we do need to keep Media, to ensure that we can accurately represent a DMS and/or BLOB server's content as that's a potentially relevant use case I don't want to lose.
Jose Costa Teixeira (Sep 26 2017 at 20:41):
@Daniel Rutz , if i interpret this as "an image is not an observation. observation is a statement about something the patient has/is..., and not a general container for something that is known about something" then i agree.
Jose Costa Teixeira (Sep 26 2017 at 20:43):
I think the images during surgery is a good case: Images are not the end goal of the surgery. If during a procedure, we need to capture images , and for liability and QC, we need to store those images, these images _could_ be called "report" but report seems more "final outcome". Images during the procedure are closer to "intermediary results" but there should be another term for that. I'd ask the native English speakers if further precision in terms is convenient.
Jose Costa Teixeira (Sep 26 2017 at 20:45):
i think that procedure should have some sort of "annotations" directly...
Elliot Silver (Sep 26 2017 at 21:36):
@Daniel Rutz I think we have different views.
Why is a blood pressure cuff reading an observation, but an image isn't? Neither are directly observed, neither involve an interpretation, and both are measurements provided by a device. I could see separating "readings" from basic interpretations, but the current design uses Observation for both, and that wouldn't resolve the issue, since I'd classify both the BP and image as "readings".
For the attachment question, I think it is a difference of viewing an "attachment" as the thing that is attached, rather than the thing that does the attaching. For things kept in a DMS or BLOB server, that's been the domain of DocumentReference. You might feel that the use of the word "Document" is misleading, as not everything referenced is equivalent to a paper document, but it is no more misleading than DMS.
Jose Costa Teixeira (Sep 26 2017 at 22:24):
@Elliot Silver The way i read your text, would this mean that observation could have an image, besides of a bunch of values?
Elliot Silver (Sep 27 2017 at 19:35):
@Jose Costa Teixeira I'm talking about future direction, not current state. I'm saying is that one approach is that an image could be an observation (currently via Observation.valueAttachment; potentially in the future, via Observation.valueMedia or some other mechanism). Separately, you could have an observation (potentially with "a bunch of values") related to an image (recorded in another Observation, or potentially in the future in Media or ImagingStudy).
Abbie Watson (Oct 10 2017 at 22:05):
Hi. Is there an instance of an imaging study with images? anyone know an example instance resource out there?
Last year, the Society for Imaging Informatics in Medicine had a hackathon where they implemented a number of Radiology specific resources in FHIR, including the ImagingStudy resources, and cross referencing with PACS and DICOM fields. A fork of the FHIR resources is below in the Clinical Meteor GitHub org:
https://github.com/clinical-meteor/siim-dataset
You could take any XR, CT, PT, or MA modality images, and cross reference with the resources listed to create some mocked up data.
Also, if anybody would be willing to toss me $1000 or $2000 (i.e. funds to go to FHIR Dev Days), I'd be more than happy to clean the SIIM data up, cross reference with images from The Cancer Imaging Archive (TCIA), and make an authoritative resource for the FHIR community.
Last updated: Apr 12 2022 at 19:14 UTC
