On this page you can find our implementation from the Kubiac team of the aneurysm segmentation algorithm we submitted for the ADAM challenge in MICCAI 2020.
Our submission ranked 4th for the aneurysm detection challenge, with a sensitivity of 0.60 and a false positive rate of 0.36, and 3rd for the aneurysm segmentation challenge (Dice 0.28, Mean Hausdorff distance 18.13, Volume similarity 0.39). See the full ranking for all teams and the docker containers.
Our approach is based on an ensamble of 18 networks, trained using 3 different architectures/loss functions using 6 different validation sets. More information and aknowledgements will later be added to this page.