This is a discussion/paper-trail ticket documenting one specific line of outreach under #[main ticket number]'s general call — not a feature request in its own right.
Context:
HAIG scoped which existing biomedical segmentation challenges to approach about interactive-segmentation collaboration, as part of standardising interactive-segmentation evaluation practice.
Method:
~5-year scan (2021–2026), ranked primarily on feasibility of actually hopping onto an existing setup (recurring + currently-active organising team as a hard filter, plus existing team ties, and a real-clinician-annotation angle where relevant), and secondarily on alignment with the under-tested axes named in Parhom's eval pipeline talk. Radiological imaging as primary scope on top of both.
Current shortlist (top → bottom): CVPR SEGFM3D, TopCoW/TopBrain, ToothFairy3, autoPET, PENGWIN, HECKTOR, ISLES, SegRap, FLARE. Full reasoning: see the shared doc: https://docs.google.com/document/d/1QhO0eJiVtWGpYzWAZioRmnvb0V1xyXBieBDggL7RyTY/edit?usp=sharing
Why post this here: creating a public paper trail within the MONAI project, given the overlap between HAIG's goals and MONAI's interactive-segmentation tooling (MONAI Label).
This is a discussion/paper-trail ticket documenting one specific line of outreach under #[main ticket number]'s general call — not a feature request in its own right.
Context:
HAIG scoped which existing biomedical segmentation challenges to approach about interactive-segmentation collaboration, as part of standardising interactive-segmentation evaluation practice.
Method:
~5-year scan (2021–2026), ranked primarily on feasibility of actually hopping onto an existing setup (recurring + currently-active organising team as a hard filter, plus existing team ties, and a real-clinician-annotation angle where relevant), and secondarily on alignment with the under-tested axes named in Parhom's eval pipeline talk. Radiological imaging as primary scope on top of both.
Current shortlist (top → bottom): CVPR SEGFM3D, TopCoW/TopBrain, ToothFairy3, autoPET, PENGWIN, HECKTOR, ISLES, SegRap, FLARE. Full reasoning: see the shared doc: https://docs.google.com/document/d/1QhO0eJiVtWGpYzWAZioRmnvb0V1xyXBieBDggL7RyTY/edit?usp=sharing
Why post this here: creating a public paper trail within the MONAI project, given the overlap between HAIG's goals and MONAI's interactive-segmentation tooling (MONAI Label).