A knowledge-based method for 3-D coronary imaging

A knowledge-based method for the 3-D reconstruction of coronary arteries using biplane angiography is described. The reconstruction problem was divided into two subproblems: (1) reconstruction of the tree skeleton and (2) reconstruction of the arterial cross section. Only the solution of the first subproblem is discussed here. Basically, the reconstruction of the tree skeleton was achieved by matching the projections of each vessel segment and subsequently backprojecting the coupled projections. As regards the matching phase, both imaging geometry and anatomical knowledge were used. In particular, owing to the allocation of epicardial vessels on a well-behaved ellipsoidlike surface, the projections of a given vessel segment must satisfy a set of geometrical conditions. These depend on the position of the vessel on the heart surface with respect to the X-ray sources. As for reconstruction of the coronary tree skeleton, an algorithm based on the imaging geometry, the bounding of the backprojection error, and a contiguity criterion was developed