基于仿射CPD的股骨医学图像配准方法  

Registration method based on affine CPD femoral medical image

作　　者：刘露[1] 刘春宇[1] 周洋[2] 张斌[1] 王培东[1] 

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080 [2]哈尔滨医科大学附属肿瘤医院,哈尔滨150080

出　　处：《计算机应用研究》2015年第8期2543-2546,共4页Application Research of Computers

基　　金：黑龙江省自然科学基金资助项目(F201208);哈尔滨市科技创新人才研究专项基金资助项目(2013RFQXJ101)

摘　　要：股骨医学图像配准是股骨三维重建技术的主要研究内容,通过机器辅助获得配准的参数决定了三维重建的准确性。针对传统ICP算法普遍存在的准确度和鲁棒性较低的问题,提出一种基于仿射CPD的股骨医学图像配准方法,使配准自动化及精准化。该方法通过边缘提取获得浮动、参考点集;在配准后获取结果参数;通过结果参数将股骨通用模型摆成股骨个体化骨折姿态。实验结果表明:平移、旋转的平均误差分别小于2.1 mm、1.6°;配准失败率仅在10%左右;三维重建后的3D图像真实度较高。该方法对于图像配准和三维重建是可行的和高效的。Femoral medical image registration is the main research content of the femur 3D reconstruction technology, registration parameters obtained through the auxiliary machine determines the accuracy of the 3D reconstruction. In view of the problem of traditional ICP algorithm has the low accuracy and robustness, this paper put forward a kind of method based on affine CPD femoral medical image registration, to make registration automation and precision. The method obtained floating and reference point set through edge detection ; then got result parameters after registration; through the result parameters, femoral universal model was put into individual femoral fracture posture. The result shows that the average error of translation and rotation is less than 2.1 mm, 1.6 ° respectively; registration failure rate is only around 10% ; the validity of 3D image after 3D reconstruction is higher. The method described by this paper for image registration and 3D reconstruction is feasible and efficient.

关 键 词：股骨医学图像 自动化 精准化 传统ICP算法 仿射CPD算法 配准方法 股骨三维重建 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]