Knowledge-based system for the three-dimensional reconstruction of blood vessels from two angiographic projections.

A knowledge-based system for the three-dimensional reconstruction of blood vessels from wide-angle coronary and stereoscopic cerebral angiographic projections is developed. For the reconstruction of the coronary vessels, the left coronary artery (LCA) is automatically labelled on standard RAO and LAO projections, using anatomical models of the LCA. The labelling system succeeds in giving the most important coronary arteries a correct anatomical label. These labelling results enable us to find corresponding segments in both images. In the case of the reconstruction of the cerebral vessels however, such an anatomical model is clearly unavailable. To find corresponding segments, small-angle projections must be relied on, resulting in very similar images. Owing to the small angular separation between both projections, the three-dimensional reconstruction will be less accurate. Once the corresponding segments in both projections are obtained, the three-dimensional artery trajectory is reconstructed with dynamic programming techniques. The three-dimensional reconstructed coronary vessels are also used for an automatic quantification of stenotic lesions.