Mitosis Domain Generalization Challenge 2025 (MIDOG)¶
Welcome to the MICCAI 2025 edition of the Mitosis Domain Generalization Challenge (MIDOG).
NEWS (2026-06-08): 📃 The preprint of our challenge paper is out. Find all about it in our blog post.
- 🔗 Full details and participation info: https://midog2025.deepmicroscopy.org
- 💬 Join our community on Discord: https://discord.gg/xEuqXjMqTk
Post-Challenge Questionnaire¶
If you have participated in this challenge and want to participate in the challenge paper, it is important that you fill out the post-challenge questionnaire:
https://forms.gle/M41zs7qZ9hhkYHb67
Why this challenge matters¶
Mitotic figures are a key biomarker in tumor grading — yet detecting and classifying them reliably remains one of the most difficult tasks in digital pathology. Current AI models often fail when faced with the variability found in clinical samples: staining artifacts, inflammation, necrosis, or unseen scanners.
MIDOG 2025 addresses this by testing algorithms not just in ideal conditions, but across diverse, challenging, and unpredictable domains — bringing us closer to clinical robustness.
What’s new in 2025?¶
This year's challenge introduces two tracks, each addressing critical gaps in current AI capabilities:
Track 1 – Robust Mitosis Detection¶
Detect mitotic figures not only in curated regions of interest (ROIs) — as in previous challenges — but also in:
- Adverse tissue regions (necrosis, inflammation, staining artifacts)
- Randomly selected tissue crops from whole slide images (WSIs)
This simulates the real-world deployment scenario where no region is guaranteed to be "clean."
👉 Learn more
Track 2 – Atypical Mitosis Classification¶
Detected mitoses must now be subclassified into:
- Normal mitotic figures
- Atypical mitotic figures — which are independent markers of tumor malignancy and patient prognosis
This is a clinically important, yet understudied, task.
👉 Learn more
Datasets¶
We provide a large, diverse training dataset from multiple domains with expert-verified annotations, following the high standards of previous MIDOG editions.
📄 Details on dataset access
How To Submit¶
Please find more details on how to submit on this page.
We provide two docker templates, including baseline algorithms: - Track 1: github link - Track 2: github link
Timeline¶
| Date | Milestone |
|---|---|
| ✅ March 25 | Launch of challenge website |
| ✅ April 8 | Training data & dataset description released |
| ✅ July 30 | Baseline methods & docker templates released |
| ✅ August 15 | Preliminary evaluation set submission opens |
| ✅ August 30 | Final set submission opens (00:01 AM) |
| ✅ August 31, 23:59 CEST | Preliminary set closes (track 2) |
| ✅ Sept 1, 23:59 CEST | Preliminary test set closes (track 1), Submission deadline for Docker container + 2-page abstract |
| ✅ Sept 23 | Announcement of results at MICCAI 2025 |
💡 Join us!¶
MIDOG 2025 is more than a challenge — it’s a step toward deployable AI in oncology. We invite researchers, developers, and clinicians to help push the boundaries of what AI can truly achieve in cancer diagnostics.
👥 Join the discussion and ask questions on our Discord:
https://discord.gg/xEuqXjMqTk
Together, let’s make mitosis detection reliable, explainable, and clinically useful.