基于点云的膝关节胫骨三维CT与MRI图像配准  

Point Cloud Based Registration of 3D CT and MRI Tibia Image

作　　者：何巍[1] 魏国栋[1] 师为礼[1] 苗语[1] 何飞[1] 杨华民[1] 

机构地区：[1]长春理工大学计算机科学与技术学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2015年第5期131-135,共5页Journal of Changchun University of Science and Technology(Natural Science Edition)

摘　　要：同时利用CT和MRI图像对膝关节及胫骨疾病进行诊断和治疗,是目前临床上的常规方法。精确配准膝关节胫骨部分三维CT和MRI图像可以更有效地利用这两种图像中的组织信息。目前,基于图像特征或者图像灰度值的方法可以实现图像间的配准,但是都存在一些配准精度低或者时间复杂度高的缺陷。针对膝关节胫骨组织的刚性特征,采用一种基于骨组织表面重建的点云配准方法。首先对数据集进行骨表面重建,通过重建后的结果提取点云数据,对点云数据进行下采样并利用重心进行初始变换,最后利用ICP算法对点云数据进行配准。对该方法进行配准结果分析及配准精度实验,表明这种方法简单可靠并且配准精度较高。In order to improve the accuracy of diagnosis or operation on knee and tibia,the registration between CT and MRI 3D images is required in clinical application. Accurate registration can help doctors understand anatomical structures correctly. In order to improve the algorithm for registering CT and MRI images of tibia,this paper proposes a registration method based on surface point cloud of bone. First,surface reconstruction of bone is performed by using the Marching Cubes method,and point clouds are extracted as the reconstructed result. Second,a down sample method is utilized for the point clouds to reduce computational consumption,and the transformation of displacement is estimated by using center of gravity. Finally,ICP method is used for point clouds registration. The experimental results show that the proposed method has high registration accuracy and low computational consumption.

关 键 词：配准 三维重建 ICP 点云 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]