Automatic identification of internal carotid artery from 3DRA images.

Geometric characteristics and arrangement of the cerebral vessels are assumed to be related to the development of vascular diseases. Identifying anatomical segments and bifurcations of the cerebral vasculature allows the comparison of these characteristics across and within subjects. In this paper, we focus on the automatic identification of internal carotid artery (ICA) from 3D rotational angiographic images. The steps of the proposed method are the following: Arterial vascular tree is first segmented and centerlines are computed. From a set of centerlines, vascular tree topology is constructed and its bifurcations geometrically characterized. Finally, ICA terminal bifurcation is detected, which enables ICA identification. To detect ICA terminal bifurcation, a support vector machine classifier is trained. We processed 82 images to obtain 274 feature vectors of bifurcations around the ICA. 10×5-fold cross-validation showed an average accuracy of 99.6%, with 99.5% specificity and 100% sensitivity. The two most discriminating bifurcation features were: radius ratio between the smaller branch and its parent vessel, and the long-axis component of the smaller branch vector.