The circle of Willis connects major arteries supplying the brain, but its highly variable anatomy makes manual assessment challenging. DQBM’s Menze Lab, together with collaborators at University Hospital Zurich, ZHAW and international institutions, developed the TopCoW challenge to benchmark AI methods for mapping this network in CT and MR angiography.
The researchers used virtual reality to annotate 13 vessel components and evaluated algorithms on 226 scans from multiple centres. Leading methods accurately mapped vessels and supported anatomical variant classification and aneurysm localisation relative to surrounding arteries.
The publicly released datasets and top-performing algorithms provide a foundation for further research and clinical tool development. The study did not assess effects on patient outcomes.
Read the publication in NEJM AI