The anatomy of the left ventricle (LV) along with a detection of scar tissue in the myocardial muscle provides important information for the treatment of heart failure. Heart failure often arises from ventricular dyssynchrony, when left and right ventricles are not contracting at the same time. This dyssynchrony can be treated with the implantation of a bi-ventricular pacemaker in cardiac resynchronization therapy (CRT). However, the presence of scar tissue hinders the electrical signal coming from the CRT device, which negatively affects the success of the treatment. Integrating the knowledge of the location of scar tissue to treatment planning could improve the efficacy of CRT.

Commonly, Late-Gadolinium-Enhanced (LGE) MR imaging is performed to visualize both scar tissue and the anatomy of the heart. The use of such images for therapeutic procedures currently lacks automatic segmentation approaches. Manual segmentation methods are available, but their use is time consuming and tedious. Therefore, it would be desirable if the segmentation of the myocardium and the scar classification were completely automatic and independent of any manual interaction.

The current research focus is on fully automatic LV anatomy and scar segmentation from LGE-MRI.