Participation¶

Submission¶

The algorithms are submitted as Docker container images according to the Grand Challenge documentation. You can find a submission template in the TopAneu-26 Repository.

Expected outputs are:

Task1: A json file containing the detected aneurysm locations of the image, formatted as below.

NOTE: This is different from the json format the training data is provided as in location_jsons due to compatibility with existing Grand Challenge sockets.

NOTE: The order of the values in the .json file is not of importance.

Task2: A 3D MetaImage file (.mha) containing the multi-class aneurysm segmentation mask, similar to those provided in the location_masks of the training data. Must be in the range of our data labels (0-52) and in the datatype 8 bit unsigned integer.

TIP: Take a look at our submission templates to base your docker image on, a more detailed description of the template is in the README.

Task 2 Submissions Also Count for Task 1¶

Any submission to Task 2 will automatically be considered for the Task 1 leaderboard!

For every Task 2 (segmentation) submission, we will:

Obtain the Task 1's location-list from the predicted segmentation masks based on class presence.
Automatically evaluate the location-list using the Task 1 metrics.
Include the resulting Task 1 metrics on the Task 1 leaderboard.

Therefore, you do not need to submit your Task 2 segmentation algorithm separately to Task 1.

Unless you have a separate, tailored classification algorithm that performs differently from our native mask-to-list conversion, please do not double-submit your Task 2 algorithm to Task 1. Your Task 2 submission will already be evaluated for both leaderboards.

Inference Runtime Environment¶

There are limited resources available for your solution, please keep these in mind when designing your approach (see Tutorials page):

The maximum processing time per case for each phase is set to 12 minutes, after this time your container will be terminated and considered failed.
Your container can request a maximum of 32 GB RAM (1 GB of which is reserved for operation purposes)
Your container can request an NVIDIA T4 GPU with 16 GB of VRAM

Evaluation¶

The evaluation methods, together with testing and documentation, are provided for each task in our Repository folder. For more details on the methodology, see the README in the respective task folder.

Leaderboards¶

Submissions to the final-test phases will be reviewed by the organizers before they are made public on the leaderboard. Note that only submissions that we can verify will be included in the leaderboards.

Rules¶

Exclusion Policy¶

We reserve the right to exclude from TopAneu-26 challenge those participants/teams breaking any of these rules.

Account and Team¶

A single account per participant. You cannot submit from multiple grand-challenge accounts. Each participant is allowed to solely submit from a single account.
Each participant can join at most one team.
Participants in a team are not allowed to make individual submissions outside of their team.
Normally, each team can only submit one algorithm to each final test phase.
If you or your team have sufficiently distinct algorithms, you may submit more than one algorithm to a final test phase, subject to approval from us.
Please reach out to us if you want to split into sub-teams or submit different algorithms.
We will exclude anonymous submissions unless we can verify your profile.
Every team can submit twice per day to the sanity check phases and only once over the whole submission period to the final submission phases. The sanity check phase is meant to validate and debug your submission (Only successful runs contribute to the submission limit).

Participants are only allowed to share scripts among their team members. Code sharing with other participants should be made publicly available through our forum.

Members of the Organizing Institutes¶

Members of the organizers' direct research groups can participate and their results can be included in the publications and the leaderboard. However, they are not eligible for awards. This policy also applies to teams that include such members.

Data Usage¶

We provide training data from multiple centers for this challenge. Usage of external data is permitted, however in this case you also need to provide your models trained without external data for a more fair comparison in our challenge manuscript. The usage of external data must be transparently communicated in the submission. If you used external training data and are invited to contribute to the challenge manuscript we will reach out to request your method trained on TopAneu-26 data only.

Publication Policy¶

The challenge results will be summarized and published in a journal manuscript. The top-5 ranked teams for each task (excluding the organizers) will be invited to contribute to the challenge publication.

To be eligible for consideration, teams must:

Disclose their methods used for the submission via a questionnaire from us.
Share the submitted Docker containers with the organizers for verification and reproducibility.
Provide valid contact and identity information by completing the Contact Form when making their submission.

We will contact the eligible top-ranked teams by email with further instructions and requests for the required materials before Sep 27th.

The top teams will also be invited to present and share their methods at our MICCAI event. Please look out for further information via email.

Each winning team can have maximum three co-authorships for the challenge paper. Additional authors from the top submissions can be included upon request with justification according to the ICMJE authorship guidelines.

Participating teams may submit their results separately without any publication embargo.

Last Updated: 11 Sep 2026