目的 探讨基于解剖标记的点配准在人头部CT及MRI双模态配准精度。 方法 4例患者行CT及MRI扫描，将扫描数据导入3D Slicer 4. 8.1以体素渲染方式进行三维重建。在10个备选的体表解剖特征点中，基于各点的5次定位误差在CT和MRI图像上选出相对应8个的解剖部位标记点坐标，按标记点数分为4、5、6、7、8个标记点组。利用奇异值分解（SVD）算法对两个模型进行对应点配准，利用标记点配准误差（FRE）对配准精度进行评估。 结果 双模态融合误差的FRE为2.68～6.54 mm。4、5、6、7、8个标记点组分别为（5.14±0.97）mm、（4.71±0.64）mm、（4.45±0.59）mm、（4.13±0.55）mm、（3.54±0.72）mm，相邻组间差异无统计学意义，8个标记点的FRE与4个标记点的FRE相比差异有统计学意义（P<0.05）。 结论 面部表面的解剖学特征可用于CT-MRI的双模态融合；CT-MRI的双模态融合FRE随着标记点数的增加而减小。

Objective To explore the accuracy of double modality fusion between CT and MRI for human head, based on pair anatomical points registration. Methods CT and MRI were approached in 4 patients, digital imaging and communications in medicine (DICOM) files were loaded on 3D Slicer (version 4. 8.1), the surface and brain were reconstructed automatically with volume rendering as reconstructed models, 10 surface characters were tested with localization error and would be used as fiducial points, 8 of the points would be chosen as pairpoints for registration according the localizing error after 5 localizations, 5 registration groups based on 4, 5, 6, 7, 8 fiducial points were divided. Singular value decomposition (SVD) was employed for registration as a basic algorithm, and fiducial registration error (FRE) was used to evaluate the registration error. Results FRE values of double modality fusion were 2.68-6.54 mm. FRE values of 4, 5, 6, 7, 8 fiducial points were (5.14±0.97)mm, (4.71±0.64)mm, (4.45±0.59)mm, (4.13±0.55)mm, (3.54±0.72)mm respectively. There was no statistical difference between neighbor groups, while in 4 and 8 points groups, the FRE values were different statistically. Conclusion Surface anatomic characters can be used for double modality fusion between CT and MRI; FRE values after CT-MRI double modality fusion are reduced with the increase of points.