Overview

MRI cardiac segmentation remains a medical imaging challenge with no satisfying current solution available for the clinic neither in commercial or open-source software. In addition to the intrinsic challenge of analysing cardiac images, the main difficulties encountered in clinical routine are:

• the limited time to perform the acquisitions which can lead to decreased quality of the acquired images and the presence of artefacts;


• the use of different MRI equipment which leads to an intrinsic variability in the properties of the images to be analysed;


• the evolution over time of the patho-physiology of the heart following treatment (over several weeks, months, years).

Through this challenge, we will provide the community with a public database of 439 MR examinations with fully annotated cardiac LGE MRI obtained at the acute and late phases of myocardial infarction after reperfusion therapy, with a variety of image properties encountered in clinical routine.

We hypothesize that such a database will allow us to identify the most efficient segmentation methods that may be used in daily clinical routine.