基于血管匹配的三维超声与CT图像的配准方法

提出一种基于血管匹配的三维超声与CT图像配准的新方法.首先,基于水平集方法自动分割出CT图像中的血管；其次,由于超声图像中的声影与血管均属于低回声区域,我们结合声影形成的物理原理及图像纹理特性,自动检测出声影区域,以提高配准的鲁棒性；最后,采用进化算法,将CT图像中分割出的血管与超声图像中低回声区域进行匹配.在肝脏体模和临床脾脏数据上进行了实验验证,自动配准的成功率在95％以上,平均目标配准误差在2 mm以内,实验结果验证了本方法的可行性.

Three- dimensional Registration of Ultrasound and CT based on Vessels

In this paper,we presented a framework to rigidly register a pre-operative computed tomography(CT) scan to the intra-operative ultrasound(US) volume.Our main focus was on removing the US shadow artifacts and deriving an appropriate similarity measure based on vessels.The proposed shadow removal method was based on the physical principle of US shadow formation and the observation that a number of shadows happen in larger areas and contain more noises than the vessel parts.Due to the fact that the vessels in the US images commonly occur in regions with relatively low intensity,the proposed registration method performed the correlation between the segmented vessels from CT images and low intensity regions in US images.We applied this method to a phantom liver and a real spleen.The success rate of registration is more than 95%,and the average target registration error is less than 2 mm,which demonstrates the feasibility of our method.