作者: David Rivest-Henault , Hari Sundar , Mohamed Cheriet
关键词: Image segmentation 、 Feature extraction 、 Fluoroscopy 、 Image registration 、 Cardiac interventions 、 Iterative reconstruction 、 Artificial intelligence 、 Computer vision 、 Affine transformation 、 Medicine 、 Solid modeling
摘要: A 2D/3D nonrigid registration method is proposed that brings a 3D centerline model of the coronary arteries into correspondence with bi-plane fluoroscopic angiograms. The registered overlaid on top interventional angiograms to provide surgical assistance during image-guided chronic total occlusion procedures, thereby reducing uncertainty inherent in 2D images. methodology divided two parts: global structural alignment and local registration. In both cases, vessel centerlines are automatically extracted from images, serve as basis for algorithms. first part, an energy minimization used estimate affine transformation aligns performance nine general purpose optimizers has been assessed this problem, detailed results presented. second fully compensate any shape discrepancy. This based variational framework, uses simultaneous matching reconstruction process compute With typical run time less than 3 s, algorithms fast enough interactive applications. Experiments five different subjects presented show promising results.