Vascular Roadmap Generation by Registration and Blending of Multiple Enhanced X-ray Angiograms

Abstract

In catheterization for hepatoma, contrast areas in X-ray angiograms are enhanced to visualize the blood vessel structures. However, it is difficult to observe the entire blood vessel in several cases due to an incomplete contrast filling during imaging. The present paper generates a vascular roadmap that enables visualization of an entire blood vessel by combining multiple enhanced images. Since images of different temporal phases are distorted by body motion, corrections by non-rigid registration are needed. We first estimated the feature-based deformations between temporally adjacent images. Then, we registered temporally distant images by accumulating the images. Finally, the vascular roadmap was generated by blending the motion-corrected images. The proposed method was applied to 13 angiograms and was compared with other registration methods. As a result, the proposed method achieved efficient deformation with a shorter processing time and generated a vascular roadmap with improved vascular structure visibility.

Keywords

• Image registration
• interventional radiology (IVR)
• moving least squares (MLS)
• robust principal component analysis (RPCA)
• X-ray angiogram