Feature extraction of convex surfaces on blood vessels using cone-beam CT images

Presents a method for feature extraction from convex blood vessel surfaces to detect abnormalities such as aneurysms. The convex surface of an aneurysm is considered to be peak surface. Feature extraction is based on the surface representation formed by differential geometries and is performed in two steps: the segmentation into different surface types and calculation of the protuberance degree of the peak surface. During the segmentation procedure, the peak surface regions are extracted from the 3-D blood vessel image based on the signs of Gaussian and mean curvatures. To calculate the protuberance degree, the area of each extracted peak surface is calculated using its image under the Gauss map. Applying the proposed method to a patient´s blood vessel, the authors demonstrate that it could be a promising technique for feature extraction from convex blood vessel surfaces