FHIR Chat · AI Algorithm within Device · Dicom-sr

Stream: Dicom-sr

Topic: AI Algorithm within Device


view this post on Zulip Diana_Ovelgoenne (Dec 09 2021 at 07:54):

Hello,
I went through device to see how to fit the Algorithm and I am wondering if we could use specialization, systemType for the Algorithm name and version for the algorithm version:
image.png
https://www.hl7.org/fhir/device.html
@Jonathan Whitby , @Elliot Silver
@Christopher Lindop
Do you see any other alternatives?
@Hans Buitendijk we are struggling using device for AI Algorithms, is my proposal ok? or specialization must be really used only for standards compliance?

view this post on Zulip Elliot Silver (Dec 09 2021 at 17:11):

I have not looked at Device much, but my inclination would to put this information in Device.version, with type = software version, component = AI. I also note that Device can model a heirarchy of devices; would having an AI component be useful, or just push the problem down one level?

view this post on Zulip Diana_Ovelgoenne (Dec 10 2021 at 07:29):

@Elliot Silver thanks for the tip about the hierarchies, not my first choice as it can become cumbersome to handle, on the other hand it can be clean. Parent is the software itself, and then having one child per algorithm. Still you have the ambiguity that it has to be specified as other on http://hl7.org/fhir/device-nametype as there is nothing for AI. I'll give it a try.

view this post on Zulip Elliot Silver (Dec 10 2021 at 08:05):

@Diana_Ovelgoenne did you see the other part of my comment about Device.version? That's the route I'd pursue first.

view this post on Zulip Diana_Ovelgoenne (Dec 10 2021 at 08:11):

@Elliot Silver Yes i saw it, but I need the AI code, name, version and I feel they don't really fit inside the version only. Using the hierarchy the fields just fit in quite well with the device structure. On Tuesday we will discuss this at the II call, so we'll come with a resolution (hopefully).

view this post on Zulip Elliot Silver (Dec 10 2021 at 08:12):

Not Device.specialization.version, but Device.version.

view this post on Zulip Elliot Silver (Dec 10 2021 at 08:22):

By the way, I filed jira issue F#32812 about the fact that deviceName.type has a required binding. Looks like it accepted, but made less useful (to you).

view this post on Zulip Diana_Ovelgoenne (Dec 10 2021 at 10:01):

Yes I meant Device.version


Last updated: Apr 12 2022 at 19:14 UTC