Three dimensional structure recognition in digital angiograms using Gauss-Markov methods

Existing methods for automatically finding arteries in coronary angiograms rely on preprocessing (digital subtraction or edge enhancement). Structure recognition in unprocessed images will enable the analysis of a wider range clinical images (of varying quality). The authors have previously reported on a prototype which works on such unsubtracted and unprocessed digital angiograms. They now present a system designed to process image pairs and thereby perform recognition in three dimensions. This approach, the “Deformable Template Matcher” (DTM), combines a-priori knowledge of the arterial tree (encoded as mathematical “templates”) with a stochastic deformation process described by a hidden Markov model. An introduction so the technique is presented along with examples of its application to bi-plane images and a discussion of the computational implications