Skip to content

Evaluation, Reproducibility, Benchmarks Meeting 30

Nicholas Heller edited this page Nov 27, 2024 · 1 revision

Minutes of Meeting 30

Date: 27th November, 2024

Present

  • Olivier
  • Annika
  • Carole
  • Nicola
  • Michela
  • Nick
  • Lena

Mission Statement Update

  • Discussed whether we'd like to update our working group's name
    • It might be nice to make the name shorter and more understandable
    • Is evaluation + benchmarking redundant? Not necessarily. Benchmarking is more about the rigorous comparison between competing methods, whereas evaluation is more generally about assessing the usefulness of a system in a clinical setting.
    • The consensus is "Evaluation and Benchmarking Working Group"
  • 3 to 4 sentences describing the working group --
    • Carole will add this blurb to Google Drive and invite comments/suggestions
      • If you have a different email you'd like to use for sharing folders/files via Google, please reach out to Nick at [email protected]

Opening to Community

  • MONAI would like these working groups to be more accessible to and engaged with the public
  • Ideas for how we might go about this:
    • Surveys to the public? We've had some experience with this, but didn't get much engagement
    • MONAI days + related initiatives where we have a specific need have been very effective
    • In general, we do want new, highly motivated members, but we don't want the group to get so big that it is unwieldy and inefficient
    • New opportunities via the website
      • Make sure community is aware of current projects
      • Open up a "suggestion box" or other avenue for the community to indicate what they'd like us to work on
      • Need to talk to the team behind the website changes to see how much flexibility we have here

Task Forces

  • First one for the confidence intervals project
    • To provide guidelines for reporting
    • And practical implementation of tasks in Metrics Reloaded
    • Olivier to take the lead
  • Second one about data
    • What makes data suitable for benchmarking? -- related to dataset quality work in progress for MICCAI lighthouse challenges
    • Working with the development team to integrate data downloads/loaders into MONAI itself
    • Michela to take the lead
  • Overall structure
    • Each task force to prepare some brief slides for WG meetings
    • Alternate focus for each meeting
    • Next meeting (January 22nd), we will play it by ear for which task force will be the focus
Clone this wiki locally