Clifford代数空间上的三维颅部感兴趣区配准

3D Cranium Registration Based on Clifford Algebra and ROI

针对3D颅部医学图像配准中存在的配准精度不高、运算复杂、配准效率低等问题,在创新性地圈定了感兴趣配准区域的基础上,提出了一种基于Clifford代数的全新的几何特征轴构造方法。起初从参考模态与浮动模态中依次提取特征点,通过该特征点实现配准感兴趣区(ROI)圈定;其次利用感兴趣区的点云数据集到其质心的距离测度构造几何特征轴,并计算相应的旋转算子完成浮动模态相对于参考模态的高效、高精度配准。这样的配准方式有效地避免了多模态图像成像时配准区域非完全匹配导致的误差,并减少待处理的数据量,同时消除了无效配准区域产生的局部最优点的影响,进而降低了配准的误差。实验表明,感兴趣区处理后的待配准图像,经新算法仿真配准,能够精确地定位组织器官的三维位置,执行效率高且配准误差较小,是一种有效的3D颅部医学图像配准方法。

Aimed at the disadvantages of the traditional method for 3D skull medical images registration, such as low precision, complex operation and inefficiency, the region of interest for registration innovatively is delineated and a new method for constructing geometric feature axis is proposed based on Clifford algebra. Firstly, the feature points, which realize the delineation of region of interest （ ROI） , are extracted in turn from the reference mode and floating mode. Then the geometric feature axes are structured by using the distance measure from the point cloud data sets of ROI to their barycenter. And the corresponding rotation operators are calculated to complete the high- efficient and high- accurate registration of the floating mode respect to the reference mode. This method can effectively avoid the error caused by the non-matching registration area when the multi-modality is imaging. And it reduces the amount of data to be processed. At the same time,the effect of local optimal point in the invalid registration region is eliminated, and the registration error is reduced. Experiment shows that the simulation registration by new algorithm can accurately locate the three-dimensional position of tissue organ aiming at the images to be registered dealt with ROI. It,s an effective registration method for 3D skull medical images with higher execution efficiency and minor registration error.