TopBrain Segmentation Challenge for Whole Brain Vessel Anatomy (TopBrain2025)

The proposed TopBrain challenge is an extension to the previously held TopCoW challenge [1], as we now extend the vessel annotations from the base of the brain to the whole brain vessel anatomy. For the past two MICCAI conferences from 2023 to 2024, the TopCoW challenge was successfully organized with positive responses. It was the first large annotated dataset and benchmark for multi-class segmentation of the Circle of Willis (CoW), an important vessel network at the base of the brain. We released to the public 250 angiography scans on two common angiographic modalities, magnetic resonance angiography (MRA) and computed tomography angiography (CTA). Each brain image was annotated voxel-wise with 13 vessel classes of the CoW anatomy. Our two iterations of the TopCoW challenge received more than 360 registrations from participants from six continents. Over 25 teams made high-quality submissions to the benchmark. The best solutions achieved nearly 90% class-average Dice for the CoW segmentation, above 85% intersection-over-union (IoU) for CoW detection, and above 85% class-balanced accuracy for CoW variant classification. The TopCoW challenge was one of the officially featured challenges on the grand-challenge platform, and was at some point with the most submissions. TopCoW arxiv preprint has since been cited by around 20 publications, many of which were accepted into impactful conferences like MICCAI and ECCV.

The positive responses from the TopCoW challenge indicate the interest from the MICCAI community for detailed, accurate, and multi-class segmentation on the complex brain vessel anatomy. The good performances of the submitted algorithms on CoW show potential of the technology to be extended to all vessel anatomy of the brain. Through the TopCoW challenge, our organizers have also established the workflow and expertise to prepare, annotate, and release large datasets with clinically verified and accurate vessel anatomy for brain images. We can build upon the TopCoW challenge to extend the annotation and dataset, from CoW to all the multi-class anatomical vessels in the angiography modalities. Thus we propose to organize such an extension of the TopCoW challenge that we call the TopBrain challenge.

The TopBrain challenge will continue the benchmark on the CoW, like in the TopCoW challenge, since the CoW is part of the whole brain vessel anatomy. However, the TopBrain challenge will cover much more than the CoW. It includes all the other major artery and vein anatomies of the brain, many of which are equally clinically relevant as the CoW and are of research interest themselves. To name a few: The ophthalmic artery (OA) is relevant to blindness and eye diseases; External carotid arteries (ECA) are important for cerebral bypass surgery; Superior cerebellar artery (SCA) supplies blood to the cerebellum, along with branches along the vertebral artery (VA) and basilar artery (BA) such as the PICA and AICA; Vessels downstream of CoW are also clinically relevant such as the detailed segments of M2, M3 and M4 of the middle cerebral artery (MCA) as they often need to be located down to their section region for stroke treatment; A branch of the ECA called the middle meningeal artery (MMA) is linked to epidural hematoma, but can be often mistaken in other annotations for peripheral vessels of MCA like M4 due to the proximity. The TopBrain challenge will also cover veins of the brain, such as the superior sagittal sinus, basal veins, and the transverse sinus, which are related to venous strokes. The whole brain vessel anatomy, including all major arteries and veins that can be identified in angiography, are annotated and segmented in the TopBrain challenge.

The proposed TopBrain challenge is beneficial to both clinicians and medical imaging researchers. A whole brain vessel anatomy segmentation with anatomical labelings and localizations can be combined with existing models for clinical tasks and applications, such as vessel stenosis or occlusion localization and aneurysm localization. Such a complete anatomy model can also be used in medical report generation and disease management and personalized treatment planning. Based on the clinical impact, data value, and research potential, we look forward to organizing the new TopBrain challenge and are sincerely welcoming submissions.

[1] Yang, K., Musio, F., Ma, Y., Juchler, N., Paetzold, J. C., Al-Maskari, R., ... & Menze, B. (2023). Benchmarking the CoW with the TopCoW Challenge: Topology-Aware Anatomical Segmentation of the Circle of Willis for CTA and MRA. arXiv preprint arXiv:2312.17670.