摘要
目的提出基于表面特征点的CT-MRI 3D/3D刚性配准(方法一)和基于表面轮廓点云的CT-MRI 3D/3D刚性配准方法(方法二),为头部肿瘤靶区勾画等临床所需的图像融合场景提供必要基础。方法利用开源图形应用函数库Visualization Toolkit(VTK)对CT和MRI图像序列进行三维面绘制重建后,拾取人脸特征点,并基于迭代最近点算法对三维特征点集进行配准。为减少拾取特征点的人为干扰,进一步提出提取2D图像人脸轮廓线后再重建得到人脸特征点云,并对配准区间进行限定,比较两种配准方法的效果。结果经多个角度进行评估,方法一由于人工拾取特征点的误差,Dice评分低于方法二;与其他配准方法相比,两种配准方法均方根误差值均较低,且方法二表现更突出。两种方法在配准精度、鲁棒性、逆一致性及配准速度等方面均有很好的表现。结论方法一的优势主要体现在配准速度上,方法二的优势主要体现在配准精度上,在鲁棒性、逆一致性等方面两种方法均有较好的表现,均能得到较好的配准结果,对头部肿瘤准确的诊断或制定出合适的治疗方案有很大的积极价值。
Objective To propose a rigid registration method of CT-MRI 3D/3D based on surface feature points(method 1)and a rigid registration method of CT-MRI 3D/3D based on surface contour point cloud(method 2),to provide the necessary basis for image fusion scenes required by clinic,such as the sketch of head tumor target area.Methods After 3D surface rendering and reconstruction of CT and MRI image sequences using the open-source graphic application library Visualization Toolkit(VTK),facial feature points were picked up and registered with the 3D feature point set based on the iterative nearest point algorithm.In order to reduce the human interference in picking up the feature points,a method was further proposed which extracted the face contour of 2D image and reconstructed it to face feature point cloud,limited the registration interval,the effectiveness of two registration methods were compared.Results After evaluation from multiple angles,method 1 had a lower Dice score compared to method 2 due to the error in manually picking feature points.Compared with other registration methods,the root mean square error values of the two registration methods were lower,and the performance of the method 2 was more prominent.The two methods had good performance in registration accuracy,robustness,inverse consistency and registration speed.Conclusion The advantages of method 1 were mainly reflected in the registration speed,and the advantages of method 2 were mainly reflected in the registration accuracy.Both methods have good performance in terms of robustness and inverse consistency.Both methods can get good registration results,which is of great positive value for the accurate diagnosis of head tumors or the formulation of appropriate treatment plans.
作者
和陆兴
和树仁
何炳春
和秋遇
贺建林
HE Luxing;HE Shuren;HE Bingchun;HE Qiuyu;HE Jianlin(Department of Medical Engineering,The 920th Hospital of Joint Logistics Support Force,Kunming Yunnan 774775,China)
出处
《中国医疗设备》
2023年第11期73-80,共8页
China Medical Devices
