TopBrain: Segmentation Challenge for Whole Brain Vessel Anatomy: Version 2 with More Labels and Clinical Focus
Rattachement africain : ch, cn, de, es, fr, gb, us, nl. Niveau de preuve : code pays fourni par la source.
Le résumé fourni par la source
The second edition of the TopBrain challenge seeks to build on the momentum of its inaugural edition at MICCAI 2025, which introduced the first benchmark for multiclass segmentation of 48 landmark brain vessels for both arterial and venous systems in CT and MR angiography. The inaugural edition established a vibrant community of over 200 global registrants and 11 independent teams. Notably, two teams exceeded the performance of organizer-developed strong baselines, demonstrating both the community’s capacity for innovation and the challenge’s role as a catalyst for algorithmic development. Beyond the challenge itself, the original TopBrain dataset has also shown substantial research utility, surpassing 800 downloads on Zenodo and contributing to a podium-finish solution in the RSNA Aneurysm Detection Challenge. In the updated TopBrain 2.0 version, we aim to improve the challenge in several key areas identified by community feedback and detailed post-challenge analysis. TopBrain 2.0 is specifically designed to bridge the gap between a testing-the-water baseline segmentation model and a clinically robust, diagnostically useful brain vessel analysis tool. The updated TopBrain 2.0 challenge aims to improve on the initial results of TopBrain 1.0 with more labels and clinical focus. We propose to expand the anatomical and pathological labels, add tailored metrics and a clinically-oriented task, and incorporate diverse multi-modal data. TopBrain 2.0 hopes to establish a robust technical and translational benchmark for whole brain vessel segmentation and relevant downstream clinical tasks. We are highly enthusiastic about hosting TopBrain 2.0 at MICCAI and look forward to advancing this field with the research community.
Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.
Le contrôle bibliographique ouvert
Les institutions déclarées
Une affiliation ne permet pas de déduire la nationalité d’un auteur.