Challenge Phases & Rules¶

The MIDOG 2025 Challenge is structured into three key phases to ensure fair, robust, and reproducible evaluation of mitosis detection and classification algorithms.

Phase 1 – Training Phase¶

• Access to training data is available. We provide diverse data sets, including multiple data domains across whole slide images and hotspot region of interests. See our dataset page.
• Participants develop and train their algorithms using the provided data and publicly and openly available datasets (private datasets are not allowed).
• The focus is on creating models that generalize well across staining, scanning, and tissue variability.

Phase 2 – Technical Validation Phase¶

• A preliminary evaluation set is made available for evaluation on grand-challenge.
• Participants are required to submit their algorithms as Docker containers.
• Only one submission per team and day is permitted. Tries to circumvent this limitation (e.g., setting up of new teams) result in exclusion from the challenge.
• This phase ensures:
  • Technical compatibility
  • Successful inference execution
  • Preliminary feedback (not used for ranking)
• ⚠️ Note: This is not the final test set and has no bearing on the final ranking.

Phase 3 – Final Submission Phase¶

• Participants submit final Docker containers for evaluation on the hidden test set.
• Submissions must be self-contained, fully automated, and reproducible.
• Participants are requested to publish a brief (approx. 2 pages) description of their method and results on a preprint site (such as arxiv.org, medrxiv.org) together with their submission. We do provide a template (double column, IEEE style) for this. There is no explicit page limit for that description, but it has to include a conclusive description of the approach of the participating team. Participants of both tracks can combine both approaches in a single preprint.
• Only one submission per team is allowed.
• Final evaluation is performed centrally on the platform under standardized conditions.

Results & Presentation¶

• Final results will be presented at the MICCAI 2025 MIDOG workshop (September 23, 2025).
• Top-performing methods will be highlighted, and participants will be invited to present their approaches.

Key Rules¶

• Participation: Participants can decide to participate in track 1, track 2, or both.
• Reproducibility: All submissions must be made as Docker containers and operate without internet access.
• Fair Play: Excessive use of GPU resources is prohibited.
• Transparency: Abstracts describing each method are required alongside the final submission.
• Conflicts of Interest: Researchers belonging to the institutes of the organizers are not allowed to participate to avoid potential conflict of interest.

Have a look at the complete list of rules here.

Publication policy¶

• Participants may publish papers including their official performance on the challenge data set, given proper reference of the challenge. There is no embargo time in that regard.
• We aim to publish a summary of the challenge in a peer-reviewed journal. Participating teams are free to publish their own results in a separate publication.

🔗 Full details: https://midog2025.deepmicroscopy.org
💬 Questions? Join our Discord community