Abstract:
We describe a system for estimating the epicardial surface of the heart using data obtained from biplane coronary cinéangiograms. The 3-dimensional (3D) geometry of the left coronary arterial tree at an instant in time is interactively reconstructed as an ensemble of 3D Bézier cubics. This provides a compact representation of the tree structure, incorporating the 3D location of the bifurcation points and their connectivity as well as the locii of the connecting vessels. A finite element model for the surface is defined using bicubic Hermite basis functions to interpolate prolate spheroidal geometric parameters. As the arteries are not uniformly distributed around the left ventricular epicardium, weighted spline-type smoothness constraints are incorporated into the error function along with the least squares error estimate. A trial fit to data describing the superficial arteries of an isolated dog heart was compared to a uniform and dense data set covering the entire epicardial surface obtained from the same heart. Good agreement was found in elements containing coronary data, with the error increasing in a controlled manner in the remaining elements. Results with angiographic data are also given. The model is readily extended to fit time-varying surfaces with the inclusion of a time basis function and is intended for use in subsequent vessel tracking algorithms.