北京理工大学 增强现实智能医疗实验室

Abstract—This paper presents a novel energy back-projective composition model (EBPCM) for 3-D reconstruction of the coronary arteries from two mono-plane angiographic images. A major problem with the commonly used parameter deformable model is that the predefined correspondences may become non-strict matching after the curve evolution, which generally leads to large extra calculation errors. In this study, the energy field in the image is back-projected to 3-D space and decomposed into three independent components in the world coordinates centered at the iso-center of the C-arm. Then, the components from different views are composited together according to the rotation and scaling relationship of the imaging angles. The composited energy field hence is utilized as the external force to control the evolution of the vascular structure in 3-D space. As the driving force is iteratively updated according to energy in the two projection images, the non-strict matching can be effectively avoided. Also, the proposed method is very flexible, which can be composited with any energy fields such as Generalized Gradient Vector Flow (GGVF) and Potential Energy (PE) etc. Experiments demonstrate that the proposed method is very effective and robust, when using GGVF as the external force, the reconstruction RMS error can be reduced to about 0.595 mm in the 3-D space.