Cerebral vessel enhancement using rigid registration in three-dimensional CT angiography

Abstract

In this paper, we propose a robust 3D rigid registration technique for detecting cerebral aneurysms, arterial stenosis, and other vascular anomalies in a brain CT angiography. Our method is composed of the following four steps. First, a set of feature points are selected using a 3D edge detection technique within skull base. Second, a locally weighted 3D distance map is constructed for leading our similarity measure to robust convergence on the maximum value. Third, the similarity measure between feature points is evaluated repeatedly by selective cross-correlation. Fourth, bone masking is performed for effectively removing bones. Experimental results show that the performance of our method is very promising compared to conventional methods in the aspects of its visual inspection and robustness. In particular, our method is well applied to vasculature anatomy of patients with an aneurysm in the region of the skull base. © Springer-Verlag 2004.