Automated Segmentation of Coronary Arteries


Please sign up on the grand-challenges website, then fill out the data confidentiality form and email a scanned copy of the signed form to .

The data will be released for unrestricted research purposes at the conclusion of the challenge, however you will not be able to download the data now and participate in the challenge without signing the data confidentiality form.


You are provided with a training set of 40 Cardiac Computed Tomography Angiography (CCTA) with contrast agent showing the coronary arteries, comprising of 20 healthy patients and 20 patients with confirmed coronary artery disease. Annotations produced by three expert annotators are provided for this training set.

Further 20 CCTA images (10 healthy and 10 patients with disease) will be released as the test set used for evaluation of the segmentation algorithms. Ground truth for the test set will not be provided while the challenge is in progress.


The task of this challenge is to segment the coronary artery lumen; not including calcified regions or other disease present. The developed methods must be fully automatic. Dice score and 95% Hausdorff distance will be used to evaluate the produced results.


Refer to the detailed challenge description.


The top 10 participants will presesnt their results at MICCAI 2020, 8th October 2020: (09:00-13:00 UTC )



  • Ramtin Gharleghi
  • Gihan Samarasinghe
  • Arcot Sowmya
  • Susann Beier