RECONSTRUCTING THE 3-D MEDIAL AXES OF CORONARY-ARTERIES IN SINGLE-VIEW CINEANGIOGRAMS

We describe a technique for reconstructing the skeletal structure of c oronary arteries from a succession of frames of a single-view cineangi ogram. We use local features in each frame to determine correspondence s of arterial segments in successive frames. We define a similarity me asure in 2-D image space as the change in angular coordinates of corre sponding pairs. We use a form of gradient descent to find those depth coordinates that minimize the average deviation of the 3-D angular coo rdinates of all points on the skeleton from the coordinates produced b y a 3-D scaling transformation. In experiments with software models th e reconstruction error was approximately two pixels when the initial g uessed reconstruction was as large as 30 pixels.