Towards the automatic reconstruction of a 3-D animated model for the coronary tree from biplane angiograms

The problem is discussed of building a 3-D temporal geometric model of the coronary tree from a succession of biplane (dual views) cardioangiogram images. The model represents geometric topologic information, which can serve as a basis for further research of the organ, and enhance medical diagnosis of angiograms in the optimal planning of bypass surgery. The cardioangiogram image data is a sequence of about 70 image pairs taken during about four cardiac cycles. The major problems dealt with are the recognition of arterial nodes and vessels in successive images, and the registration of matching objects in the two views, to convert them to 3-D coordinates, through all 70 pairs of images in a single case. In reaching towards a fully automatic system, the following points are emphasized: automatic recognition of arterial nodes and vessels in a series of 2-D images along the time axis; and semiautomatic solution of the two-views registration problem, needed to construct a 3-D model