北京生物医学工程

基于DRR及相似性测度的2D-3D医学图像配准算法

2D/3D medical image registration algorithm based on DRR and the similarity measure

作者：麦永锋孙启昌贾鹏飞陈晓军

单位：上海交通大学机械与动力工程学院（上海 200240）上海交通大学数字医学临床转化教育部工程研究中心（上海 200240）上海交通大学医疗机器人研究院（上海 200240）

关键词： 2D/3D配准；数字重建放射影像；相似性测度；Powell算法；归一化互相关；梯度差分

分类号：R318.04

出版年·卷·期（页码）：2021·40·3（263-272）

摘要：

目的针对术前三维电子计算机断层扫描(computed tomography,CT)图像和术中二维X线图像的配准问题提出了精确高效的全自动配准算法，使得该算法能够适应不同X线图像的风格。方法首先通过数字重建放射影像(digitally reconstructured radiograph， DRR)技术，把CT图像转换成DRR图像，从而将CT和X线图像的配准转换成DRR图像和X线图像的配准。然后在传统的归一化互相关 (normalized cross correlation，NCC)和梯度差分(gradiant difference，GD)相似性测度指标基础上提出融合NCC和GD的归一化互相关-梯度差分(normalized cross correlation - gradient difference，NG)指标，并基于NG指标计算DRR图像和X线图像的相似性。通过Powell算法迭代求取相似性测度的极值，从而获得配准矩阵。最后在人体头颅CT数据上采用“黄金标准”判断法量化配准系统的精度，并通过脊柱CT和X线配准案例验证系统在实际场景的性能。结果基于DRR及NG相似性测度的2D/3D图像配准系统的配准距离误差为0.51 mm，角度误差为0.40°。结论基于DRR及NG相似性测度的2D/3D图像配准算法具有较好的配准精度，能适应不同风格的二维输入图像，基于该算法的配准系统具有较好的鲁棒性。

Objective To accurately, efficiently, and automatically register the preoperative computed tomography（CT） image and the inoperative X-ray image in order to improve the adaptability to different X-ray image styles. Methods CT was firstly converted into digital reconstruction radiography （DRR) images through the digital reconstruction radiography technology. The registration of CT and X-ray was converted into the registration of DRR and X-ray images. Then, based on the traditional similarity measures, normalized cross correlation（NCC） and gradient difference（GD）, normalized cross correlation - gradient difference（NG） was proposed to calculate the similarity between DRR and X-ray. Powell algorithm was used to get the extremum value of the similarity metric. The registration matrix was obtained after the optimization process. At last，using the CT image of the skull, the “Gold standard” judgment method was used to quantitatively evaluate the registration algorithm. The spine CT and X-ray registration case was also applied to qualitatively evaluate the performance of the registration system in practical application. Results The registration distance errors of the angle and position of the 2D/3D medical image registration algorithm based on DRR and the NG similarity measure were 0.40°and 0.51mm. Conclusions The 2D / 3D image registration algorithm based on DRR and the NG similarity measure can improve the accuracy and adapt to different image styles. The registration system based on the algorithm is robust.

参考文献：

[1]?张薇, 黄毓瑜, 栾胜, 等. 基于灰度的二维/三维图像配准方法及其在骨科导航手术中的实现[J]. 中国医学影像技术, 2007, 23(7): 1080-1084.?

Zhang W, Huang SY, Luan S, et al.?Implementation of intensity-based 2D/3D image registration in orthopaedic navigation system[J]. Chinese journal of medical imaging technology, 2007,23(7):1080-1084.

[2]?Qin C , Cao Z , Fan S , et al. An oral and maxillofacial navigation system for implant placement with automatic identification of fiducial points[J]. International Journal of Computer Assisted Radiology and Surgery, 2019.

[3]?张冉,王雷,夏威,高欣.2D/3D图像配准中的相似性测度和优化算法[J].激光与红外,2014,44(01):98-102.?

Zhang R,Wang L,Xia W,et al.Comparison of similarity measurement and optimization methods in 2D/3D image registration[J]. Laser&Infrared,2014,44(1):98-102.

[4]?李言, 武王将, 杨博鑫, 等. 基于2D-3D配准的术中腓骨旋转不良检测方法[J]. 北京生物医学工程, 2019, 38(1):56-62.

Li Y, Wu WJ, Yang BX, et al. Method of detectingintraoperative fibular malrotation based on 2D-3D registration technology?[J].?Beijing Biomedical Engineering,?2019,?38(1):52－58．

[5]?Wang M,Song Z. Automatic localization of the center of fiducial markers in 3D CT/MRI images for image-guided neurosurgery[J]. Pattern Recognition Letters, 2009, 30 (4): 414-420.

[6]?张建法,张峰峰,孙立宁, 等.基于C形臂的手术导航相机标定方法[J].计算机应用,2016,36(8):2327-2331.

Zhang?JF,?Zhang?FF,?Sun?LN,?et al. Camera calibration method of surgical navigation based on C-arm[J]. ?Journal of Computer Applications,?2016,36(8):2327-2331.

[7]?Yu W , Tannast M , Zheng G . Non-rigid free-form 2D-3D registration using A B-spline-based statistical deformation model[J]. Pattern Recognition, 2016:S003132031630293X.

[8]?刘坤, 吕晓琪, 谷宇, 等. 快速数字影像重建的2维/3维医学图像配准[J]. 中国图象图形学报, 2016, 21(1): 69-77.

Liu?K,?Lyu XQ,?Gu Y,?et al． The 2D/3D medical image registration algorithm based on rapid digital image reconstruction[J].?Journal of image and graphics,?2016,?21(1): 69-77.

[9]?刘达, 谢永召, 倪自强, 等. 基于局部方差采样的2D/3D医学图像配准技术[J]. 高技术通讯, 2013, 23(8): 827-832.

Liu D,?Xie YZ,?Ni?ZQ,?et al.?2D/3D medical image registration based on local variance sampling[J]. Chinese high technology letters,?2013,?23(8): 827-832.

[10]?陈龙, 张峰峰, 于凌涛, 等. 归一化互信息与多分辨率融合的2D-3D配准方法[J]. 哈尔滨工程大学学报, 2020, 41(2):243-249.

Chen?L,?Zhang?FF,?Yu LT,?et al. A 2D-3D registration method based on normalized mutual information and multi-resolution fusion[J]. Journal of Harbin Engineering University,?2019,?41(2): 243-249．

[11]?Pengfei Jia, Lu Xu, Xiaojun Chen, et al. An intensity-based rigid registration of 2D C-Arm and 3D CT images for image-assisted navigation system. Proceedings of the 31st International Conference on Computer Assisted Radiology and Surgery (CARS 2017), June, 2017.

[12]?Tornai?GJ, Cserey G , Pappas I . Fast DRR generation for 2D to 3D registration on GPUs[J]. Medical Physics, 2012, 39(8).

[13]?Toth D , Panayiotou M , Brost A , et al. 3D/2D Registration with Superabundant Vessel Reconstruction for Cardiac Resynchronization Therapy[J]. Medical Image Analysis, 2017, 42:160.

[14]?Mu Z. A Fast DRR Generation Scheme for 3D-2D Image Registration Based on the Block Projection Method[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. 2016: 169-177.

[15]?Folkerts M , Jia X , Choi D , et al. SU-E-I-35: A GPU Optimized DRR Algorithm[J]. Medical Physics, 2011, 38(6):3403.

[16]?Cong W, Liang X, Dong J, et al. Accurate CT-ultrasound image registration using simulated transformation optimization[C] //Eleventh International Conference on Digital Image Processing (ICDIP 2019). International Society for Optics and Photonics, 2019, 11179: 111790V.

[17]?ébastien Clippe S, Sarrut D, Malet C, et al. Patient setup error measurement using 3D intensity-based image registration techniques[J]. International Journal of Radiation Oncology* Biology* Physics, 2003, 56(1): 259-265.

[18]?Kim J, Fessler J A, Lam K L, et al. A feasibility study of mutual information based setup error estimation for radiotherapy[J]. Medical Physics, 2001, 28(12): 2507-2517.

[19]?Alam F, Rahman S U, Ullah S, et al. Medical image registration in image guided surgery: Issues, challenges and research opportunities[J]. Biocybernetics and Biomedical Engineering, 2018, 38(1): 71-89.

[20]?苏长慧, 夏桂梅, 屈向红. 基于Powell算法的改进微粒群算法[J]. 宁夏师范学院学报, 2013(06):70-73.

Su CH, Xia GM, Qu XH. Improved Particle Swarm Optimization Algorithm Based on the Powell Algorithm. Journal of Ningxia Normal University, 2013(06):70-73.

[21]?杨帆,张汗灵. 蚁群算法和Powell法结合的多分辨率三维图像配准[J]. 电子与信息学报, 2007, 29 (3): 622-625.

Yang F, Zhang HL. Multiresolution 3D Image Registration Using Hybrid Ant Colony Algorithm and Powell's Method[J]. Journal of Electronics & Information Technology, 2007, 29 (3): 622-625.

[22]?Flood PDL , Banks SA . Automated Registration of 3-D Knee Implant Models to Fluoroscopic Images Using Lipschitzian Optimization[J]. IEEE Transactions on Medical Imaging, 2018, 37(1):326-335.

[23]?Li?GH, Gao?JY, Kafka OL, et al. Implementation and application of the multiresolution continuum theory[J]. Computational mechanics, 2019, 63(4): 631-647.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录