Difference between revisions of "AI Interoperability in Imaging White Paper"

From IHE Wiki
Jump to: navigation, search
(Public Comment: bolded links to doc and comment page)
 
(13 intermediate revisions by 3 users not shown)
Line 1: Line 1:
This page is draft.
+
This white paper is intended to document:
 +
* tasks related to medical imaging that might be performed or supported by AI models
 +
* "use cases" that describe the steps required to create and use AI Models
  
== Public Comment and Navigating the AI Interoperability in Imaging Document ==
+
The paper’s use of “imaging” as a shorthand is intended to cover all forms of medical imaging, not just radiology.
  
=== [https://www.youtube.com/watch?v=Sj6qB8VFBy8 Watch the Introductory Video on YouTube!] ===
+
The IHE Planning and Technical Committees will use this white paper to identify logical groups of needs that would constitute functional Profiles. The white paper content will help ensure the Profiles are properly scoped and do not overlook key issues.
  
The AI Interoperability in Imaging White Paper is a long document. Reviewers are encouraged to provide feedback, even if only on portions of the document. To optimize their efforts, reviewers may consider:
+
==Navigating the White Paper==
* Paying particular attention to the Questions in sectionSection 1.4.  
+
<youtube>pOjldQBiytg</youtube>
* Focusing on specific areas that align with their interests and expertise:
+
 
** 2.1 Applications of AI in Imaging focuses on activities that AI can be applied to. The use cases in chapter 3 will be driven by these applications. This section may appeal to clinical reviewers.
+
Although the white paper is a long document, readers may use the document structure to focus on sections of interest.
*** Note: this white paper specifically strives to broaden the map of applications beyond radiology image analysis. Reviewers are encouraged to consider:
+
 
** Imaging specialties beyond radiology, such as cardiology, pathology, dentistry, ophthalmology, dermatology, etc.
+
* Section 2.1 - lists examples of '''tasks to which AI can be applied'''. They are collected into 8 "stages" in the imaging pipeline starting with placing orders.
*** Tasks beyond image analysis, such as ordering, acquisition, reporting, and departmental operations analysis.
+
* Section 2.2 - provides brief definitions of the '''personas and systems''' that participate in the subsequent use case sections.
** 2.2 Personas and Systems provides brief definitions that may clarify the subsequent use case sections. This section may appeal to both clinical and technical reviewers.
+
* Section 3.1 & 3.2 - describe '''managing repositories and assembling datasets''' for training, testing and validation
** 3 Use Cases focuses on the steps that create, update, and use AI Applications. At the beginning of each Use Case Group and Use Case includes a statement of the task covered in the block. This section may appeal to data scientists and technical reviewers.
+
* Section 3.3 & 3.4 - describe '''creating and distributing AI models'''
** 4 Entities focuses on common data structures and metadata in AI workflows. This section may appeal to technical reviewers.
+
* Section 3.5 & 3.6 & 3.7 - describe '''external validation, clinical use, and feedback''' for AI models
 +
* Section 3.8 - describes common mechanics like '''de-identification, annotation, and provenance'''
 +
* Section 4 - starts sketching '''common data structures and metadata''' in AI workflows.
 +
 
 +
==Public Comment==
 +
 
 +
'''[https://youtu.be/pOjldQBiytg Watch the Introductory Video on YouTube!]''' (also embedded above) for an overview of the purpose and content of the white paper.
 +
 
 +
Reviewers are encouraged to provide feedback, even if only on portions of the document.
 +
* consider focusing on specific sections (see above) that align with your interests and expertise
 +
* review the Questions in Section 1.4. on which the committee is particularly seeking feedback
 +
** Have imaging specialties beyond radiology, such as cardiology, pathology, dentistry, ophthalmology, dermatology, etc. been adequately covered?
 +
** Have tasks beyond image analysis, such as ordering, acquisition, reporting, and departmental operations analysis been adequately covered?
 +
 
 +
'''The document is available for download [https://www.ihe.net/resources/public_comment/#radiology here.] Comments can be submitted [https://www.ihe.net/Radiology_Public_Comments/ here.]'''

Latest revision as of 15:41, 12 March 2021

This white paper is intended to document:

  • tasks related to medical imaging that might be performed or supported by AI models
  • "use cases" that describe the steps required to create and use AI Models

The paper’s use of “imaging” as a shorthand is intended to cover all forms of medical imaging, not just radiology.

The IHE Planning and Technical Committees will use this white paper to identify logical groups of needs that would constitute functional Profiles. The white paper content will help ensure the Profiles are properly scoped and do not overlook key issues.

Navigating the White Paper

Although the white paper is a long document, readers may use the document structure to focus on sections of interest.

  • Section 2.1 - lists examples of tasks to which AI can be applied. They are collected into 8 "stages" in the imaging pipeline starting with placing orders.
  • Section 2.2 - provides brief definitions of the personas and systems that participate in the subsequent use case sections.
  • Section 3.1 & 3.2 - describe managing repositories and assembling datasets for training, testing and validation
  • Section 3.3 & 3.4 - describe creating and distributing AI models
  • Section 3.5 & 3.6 & 3.7 - describe external validation, clinical use, and feedback for AI models
  • Section 3.8 - describes common mechanics like de-identification, annotation, and provenance
  • Section 4 - starts sketching common data structures and metadata in AI workflows.

Public Comment

Watch the Introductory Video on YouTube! (also embedded above) for an overview of the purpose and content of the white paper.

Reviewers are encouraged to provide feedback, even if only on portions of the document.

  • consider focusing on specific sections (see above) that align with your interests and expertise
  • review the Questions in Section 1.4. on which the committee is particularly seeking feedback
    • Have imaging specialties beyond radiology, such as cardiology, pathology, dentistry, ophthalmology, dermatology, etc. been adequately covered?
    • Have tasks beyond image analysis, such as ordering, acquisition, reporting, and departmental operations analysis been adequately covered?

The document is available for download here. Comments can be submitted here.