Results of the CROWN challenge on automated assessment of circle of Willis morphology

Iris Vos, Ynte Ruigrok, Hugo Kuijf

Keywords: circle of Willis, intracranial arteries, biomedical image analysis challenge, classification, quantification, magnetic resonance angiography

Abstract:

Automated assessment of circle of Willis (CoW) morphology may aid the identification of imaging risk factors associated with intracranial aneurysm (IA) development. However, comparative studies that explore optimal methodological approaches are currently lacking.
To systematically compare the performance of automated methods to a clinical reference standard, we initiated a scientific challenge associated with MICCAI 2023. This challenge comprised two tasks: automated classification of CoW anatomical variants and automated
prediction of CoW artery diameters and bifurcation angles. The challenge dataset comprised 300 TOF-MRA images for training and 300 for testing, all manually annotated. Method were evaluated using balanced accuracy, mean absolute error, and Pearson correlation coefficient metrics. This short paper will present the first results and evaluation of the challenge. The challenge remains open for future submissions, serving as a benchmark for evaluating methods aimed at assessing CoW morphology.

Submission Number: 69