Coronary artery angiography is still the most commonly used and most reliable medical imaging system. In angiograms, the three-dimensional structure of the heart arteries can be observed in two-dimensional images. This situation complicates the visual examination of the heart vessels and may lead to incorrect measurements. The precision of the measurements is important in terms of the roles they play at the diagnostic and therapeutic applications, such as stent placement. Three-dimensional modeling systems have been developed in order to obtain the correct numerical values from angiogram images . It became possible to obtain accurate quantitative information about vessels, in particular the area of the lesion, by using 3D models created from images which are taken from different angles. The studies in this area are proceeding in the direction of creating a more precise three-dimensional model. In this thesis, three-dimensional modeling is carried out by using edge information from vessels. In order to obtain a more precise edge information, a parametric edge estimation algorithm is used. Also, an approach was developed according to the imaging geometry to calculate the surface of the vessel and vessel cross sectional areas were modelled via ellipsoids. The performance of the new edge estimation method and the ellipsoidal estimation were presented comparatively. It was observed that the proposed methods outperformed the current ones in literature.
Key Words: Angiogram, calibration, edge estimation, 3D modelling, ellipsoid fitting, vessel tracking, centerline extraction.