Graph-based knowledge-driven discrete segmentation of the left ventricle

In this paper, we propose a novel similarity-invariant approach to model-based segmentation of the left ventricle. The method assumes a control point representation of the model and an arbitrary interpolation strategy. First, we construct the prior manifold using the distributions of the relative normalized distances between pairs of control points within the training set. Then, we introduce a geometric partition of the space using a Voronoi decomposition that aims to determine relationships between the control points and the image domain. Knowledge-based segmentation can then be expressed using a Markov Random Field, where the pairwise potentials encode the variation of the shape, while the singleton potentials refer to the data term through the Voronoi decomposition of the space. State-of-the art techniques from linear programming are considered to optimize the designed function.