Robust segmentation of coronary arteries in cine angiography for 3d modeling

X-ray angiography is state of the art in diagnosing cardiovascular diseases. Angiographic imaging is based on contrast enhanced biplanar projection images of the left ventricle and the coronary vessels. A novel automated segmentation and modeling method for the coronary vessel tree is developed. Segmented vessels are the basis for three-dimensional reconstruction of the coronary arteries. Vessel areas are enhanced by Gabor filtering and resulting regions are merged employing similarity criteria. A hierarchical graph model of the entire vessel tree is determined. Vessel walls are autonomously assessed. For proper analysis of the segmented vessel shape a smoothing algorithm preserving the vessel course is developed. The vessel graph is iteratively adapted according to adjacent frames in the angiographic series. A time series of the reconstructed vessel tree is generated. Results show accurate segmentation of even small vessels. The method proved to be robust even with noisy angiographic image data. Vectorization of results builds entire hierarchical trees enabling model based investigation of obstructed vessels.