CAMUS article
Accurate segmentation of 2D echocardiographic images has been an ongoing problem for more than 30 years. The reasons for that are threefold: i) the very nature of echocardiographic images (poor contrast, brightness inhomogeneities, variation of speckle pattern along the myocardium, significant tissue echogenicity variability within the population, etc) makes it difficult to accurately localize cardiac regions; ii) the lack of publicly-available large scale 2D echocardiographic datasets; iii) the lack of multi-expert annotations on large datasets to assess the minimum error margin beyond which segmentation methods would be considered as accurate as human experts.

While these reasons are well accepted by the community, a large number of echocardiographic image segmentation methods have been proposed over the past decades. Unfortunately, most of it have been validated on small private datasets (usually with a few dozen patients) thus making their comparison with other methods almost impossible. As a consequence, semi-automatic or manual annotation is still used in clinic due to the lack of accuracy and reproducibility of fully-automatic cardiac segmentation methods.

Based on this observation, we decided to create the CAMUS project. The purpose of our project is to provide the community with the necessary materials to solve the problem of 2D echocardiographic image segmentation.

This work has published to IEEE TMI journal. You must cite this paper for any use of the CAMUS database

  • S. Leclerc, E. Smistad, J. Pedrosa, A. Ostvik, et al.
    "Deep Learning for Segmentation using an Open Large-Scale Dataset in 2D Echocardiography" in IEEE Transactions on Medical Imaging, vol. 38, no. 9, pp. 2198-2210, Sept. 2019.

    doi: 10.1109/TMI.2019.2900516