基于2D-3D配准的术中腓骨旋转不良检测方法  

Method of detecting intraoperative fibular malrotation based on 2D-3D registration technology

作　　者：李言 武王将 杨博鑫 丁云鹤 王巨鹏 孙振辉[4] 杨智[1,2] LI Yan;WU Wangjiang;YANG Boxin;DING Yunhe;WANG Jupeng;SUN Zhenhui;YANG Zhi(College of Biomedical Engineering,Capital Medical University,Beijing 100069;Beijing Key Laboratory of Fundamental Research on Biomechanics in Clinical Application,Beijing 100069;Orthopedics Department,Xiqing Hospital,Tianjin 300380;Orthopedics Department,Tianjin Hospital,Tianjin 300211)

机构地区：[1]首都医科大学生物医学工程学院,北京100069 [2]临床生物力学应用基础研究北京市重点实验室,北京100069 [3]天津市西青医院骨科,天津300380 [4]天津市天津医院骨科,天津300211

出　　处：《北京生物医学工程》2019年第1期52-58,共7页Beijing Biomedical Engineering

基　　金：国家自然科学基金(61227802);北京市教育委员会科技发展计划一般项目(KM201610025009);首都医科大学基础-临床科研合作基金项目(2016-JL-PT-X02)资助

摘　　要：目的通常检测腓骨旋转不良的方法是对术中踝关节2D X射线影像做目视评估,但主观判断往往导致结果误差较大;目前公认的准确检测腓骨旋转不良的方法需借助术后3D CT数据,但会为患者引入较大的辐射量。为此,本文提出了一种基于2D-3D配准技术准确识别术中腓骨旋转不良的低剂量、低成本的精准复位腓骨的可行方法,以此验证所提方法的可行性。方法研究对象为单一腓骨CT数据,初始位置定义为参考位,然后借助Mimics软件对其做不同程度的旋转变换,生成12组腓骨旋转畸形测试位的CT数据,模拟术中腓骨的不同姿态,包括6组腓骨内旋和6组腓骨外旋。测试位腓骨术中C臂图像由投影仿真程序生成。通过将术中C臂图像和参考位CT数据做2D-3D配准来识别这12组测试位相对于参考位的旋转不良程度。得到的结果和金标准比对,从而评估2D-3D配准的准确性;其中,金标准为参考位和12组测试位的3D-3D配准结果。另外,因为与投影轴平行方向检测位移不敏感,故本文用两幅正交位投影的配准结果做补偿。结果 10次测试12组数据配准结果在绕x轴、y轴和z轴旋转的平均角度(及沿3个方向的平均位移)误差分别为1. 19°(0. 56 mm)、0. 72°(0. 84 mm)和0. 81°(0. 65 mm);标准差依次为0. 43°(0. 38 mm)、0. 51°(0. 47 mm)和0. 58°(0. 5 mm);最大误差分别是2. 13°(1. 76 mm)、2. 74°(1. 90 mm)和2. 10°(2. 16 mm)。结论 2D-3D配准方法可为临床腓骨复位提供精度更高的监控工具,其误差远小于目前10°旋转的目测误差。相比于术前CT做手术规划、术后CT做手术评估,本文方法借助术前CT和术中C臂图像不仅可达到准确评估的目的,而且可实现术中动态评估,故其辐射剂量更低,患者医疗成本更低,治疗更及时、更有效。Objectives Distal fibular malrotation is one of the main reasons leading to poor functional outcomes in ankles.The general method of detecting fibular malrotation is to evaluate intraoperative 2D X-ray images based on surgeons’experiences.This subjective judgment tends to bring bigger errors.Whereas,accurate estimation of fibular malrotation requires postoperative 3D CT scans.It will bring in extra radiational doses to patients.This paper proposes to better detect the intraoperative distal fibular malrotation using 2D-3D registration technology with low-doses and low-costs,which registers intraoperative fluoroscopies in 2D to the preoperative CT volume.Methods In order to verify the proposed method’s feasibility,we studied the procedure of identifying the fibular malrotation in intra-operation scenario using simulation data.We took a cadaver fibula bone CT data as a research subject.Its initial position was defined as the reference position.From the CT data,12 intraoperative volumetric datasets were simulated with internal rotation(IR)and external rotation(ER)gestures in Mimics-software.2D C-arm images at 12 gestures were also simulated by forward projection.Fibular rotations at 12 gestures were estimated using the suggested 2D-3D registration method,the registration results’accuracy was compared with gold standards.The gold standards were from the 3D-3D registration between the 3D simulation data and the reference 3D data.To minimize the large errors in insensitive axis,two orthogonal fluoroscopies were used in 2D-3D registration.Results The average registration errors of 10 tests in rotation angles(and translations)at 12 gestures in x-axis,y-axis and z-axis were 1.19°(0.56 mm),0.72°(0.84 mm)and 0.81°(0.65 mm),respectively.The corresponding standard deviations were 0.43°(0.38 mm),0.51°(0.47 mm)and 0.58°(0.50 mm).Overall,the maximum errors in cartesian coordinate system were 2.13°(1.76 mm),2.74°(1.90 mm)and 2.10°(2.16 mm)respectively.Conclusions The proposed 2D-3D registration method opens a possib

关 键 词：踝关节骨折 腓骨旋转不良 2D-3D 配准 优化算法 医学图像手术导航 

分 类 号：R318.04[医药卫生—生物医学工程]