TopBrain Segmentation Challenge Version 2¶

TopBrain will have a special joint track with the collaborating TopAneu Challenge for the upcoming MICCAI event. The tasks for veins and stenosis will take place from October through the end of December 2026. Here are the details for the activities leading up to MICCAI.

Call for Submissions: The TA36 Track at MICCAI¶

🧠 TopAneu 36-Class (TA36) Vessel Ground Truth Released¶

The 36-class vessel anatomy used by TopAneu consists of two infraclinoid vessel classes introduced in the new edition, together with the 34 common artery labels used for both CTA and MRA in the 2025 edition. The ground-truth annotations for these 36 classes have now been added to the TopBrain data release at https://zenodo.org/records/21972006. You can find them under labelsTr_topbrain_v2_topaneu36class, which contains the 36-class vessel annotations corresponding to the vessel anatomy used in TopAneu. Note that TA36 is essentially a subset of the artery labels in TopBrain V2, excluding the extracarotid artery (ECA) classes.

🐋 Official TopAneu Vessel Docker Released¶

The Docker used to generate the "silver-standard" vessel masks in the TopAneu data release is now publicly available on Zenodo at https://zenodo.org/records/21959166. This Zenodo record includes the container image along with a detailed guide for running it locally. The model was trained on the TA36 annotations in labelsTr_topbrain_v2_topaneu36class, and will serve as the organizer baseline method for the TA36 track.

🛤 TA36 Track¶

Building on these releases, we invite both the TopBrain and TopAneu communities to benchmark their vessel segmentation algorithms in our special TA36 Track at MICCAI:

• Task: Predict TopAneu 36-class vessel segmentation (TA36) on the hidden annotated TopBrain test set.

• Evaluation: Submissions will be evaluated against gold-standard vessel annotations on the hidden test set and compared with the official TopAneu Docker, which will serve as the baseline benchmark.

💪 Why participate? For TopAneu participants: If you already have a vessel segmentation module for TopAneu, submit it to the TA36 track and let us do the evaluation for you. The hidden test set provides an independent and detailed evaluation of your vessel segmentation component against gold-standard annotations, together with a comparison to the official Docker baseline.

💪 Why participate? For TopBrain participants: You might win both challenges! Internally, we found that models trained on the TA36 annotations provided useful vessel anatomy for aneurysm localization, despite being trained on a non-aneurysm dataset. We have used this capability to help flag and resolve several cases of misclassified aneurysms in the TopAneu data. In other words, accurate TA36 vessel segmentation can be useful for downstream aneurysm localization.

One vessel segmentation model. Two challenges. Two chances to win. 🏆🏆

To accommodate both challenges, we will align the TA36 submission phase with the TopAneu final submission period. We aim to close the TA36 submission approximately one or two days after the TopAneu deadline, providing participants with some additional buffer time to submit their methods.

We look forward to seeing your methods and results!

More information about the TA36 track and submission procedure will be available soon.