Reconstruction of three-dimensional cardiac shapes in biplane angiography: a fuzzy and evolutionary approach

A 3D reconstruction method for the left ventricle is proposed. The reconstruction is performed from true preprocessed images acquired in biplane angiography, by working directly in the 3D space and by considering the real acquisition geometry. The reconstruction method includes two stages. The first stage provides an approximate reconstruction for the ventricle shape. Several attributes from the ventriculographic images are used to define the fuzzy sets describing the ventricle, its contour and the background region. Combination of each pair of sets according to the fuzzy relation cylindrical closure operation and the Dempster-Shafer theory, allows an approximate binary shape to be found. The second stage improves the reconstruction by progressively deforming the initial shape by using a simple genetic algorithm (SGA). The SGA optimizes an objective function that measures the projection error and includes a connexity constraint between voxels of the object.