Objective: To propose a new method of 3D medical image registration based on fiducial markers and to verify its feasibility. Methods: The positioning tools with 6 fiducial markers were fixed with 25 experimental objects, which consisted of lumbar and cervical model bones, animal (pig) vertebrae and cadaveric vertebrae, respectively. Three-dimensional X-ray scans were performed to obtain medical images. Markers in the collected 25 three-dimensional medical images were identified and positioned. Finally, the rigid registration matrices and registration errors that obtained from above results were calculated. Results: In 25 images with 150 fiducial markers, 170 points were obtained through the preliminary screening of marching cubes(MC) combined with geometrical characteristic of markers. After the second fine screening, 131 correct fiducial markers were obtained. Because the number of fiducial markers has redundance, all the datasets of 25 groups were successfully registered and all of the errors were less than 1mm. Conclusion: This method can automatically identify and locate the marker points, and it has good identification effect and high image registration accuracy. And the registration error can meet the clinical requirements.%目的:提出一种新的基于标记点的三维医学图像配准方法,并对其可行性进行验证.方法:将带有6个标记点的定位工具分别与由腰椎、颈椎模型骨、动物(猪)脊椎骨和尸体脊椎骨等构成的25例试验对象固定在一起,进行三维X射线扫描,从而获得医学图像.识别采集的25幅三维医学图像中的标记点,计算得到刚体配准矩阵和配准误差.结果:将带有150个标记点的25幅图像,经过移动立方体(MC)算法结合标记点几何特征初步粗筛得到170个标记点,经过二次精细筛选之后得到131个正确标记点.因标记点数目有冗余,25组数据全部配准成功,误差均<1 mm.结论:基于标记点的三维医学图像配准方法可以自动识别标记点,具有较好的识别效果和图像配准精度,配准误差可满足临床要求.
展开▼
