Using cardiac cine magnetic resonance, we introduce a new structured output model for the region of interest (ROI) detection of the left ventricle of the heart using a deep belief network (DBN).
We also propose a new structured output model for the delineation of the endocardial and epicardial borders using another DBN.
Finally, we extend a common level set method that takes: a) the ROI detection above to initialise the optimisation process, and b) the delineation of the endocardial and epicardial borders above to constrain the level set evolution.