CN113450395B - Surgical navigation space registration method and device based on optical tracking probe point drawing - Google Patents

Abstract

Translated fromChinese

基于光学跟踪探针划取点的手术导航空间配准方法及装置，更加易于操作且成本更低，配准精度与基于基准点配准得到的结果相近。这种基于光学跟踪探针划取点的手术导航空间配准方法包括以下步骤：(1)利用一个通用的人面部模板，通过模板匹配方法以剔除图像空间中冗余的表面数据；(2)使用光学跟踪探针在真实人面部取点，并提出一个约束指标，以重建均匀的人面部轮廓；(3)通过粗到精的增量点云配准方法以及配准约束指标，约束所采集点云的完整性和配准结果的正确性，最终将图像空间与真实空间精确对齐。

The surgical navigation space registration method and device based on the points drawn by the optical tracking probe are easier to operate and lower in cost, and the registration accuracy is similar to the result obtained based on the reference point registration. This space registration method for surgical navigation based on the points drawn by the optical tracking probe includes the following steps: (1) using a common human face template to eliminate redundant surface data in the image space through template matching; (2) Use the optical tracking probe to take points on the real human face, and propose a constraint index to reconstruct a uniform human facial contour; (3) Constrain the collected points through the coarse-to-fine incremental point cloud registration method and the registration constraint index The integrity of the point cloud and the correctness of the registration results finally align the image space with the real space precisely.

Description

Translated fromChinese

基于光学跟踪探针划取点的手术导航空间配准方法及装置Surgical navigation space registration method and device based on optical tracking probe drawing points

技术领域technical field

本发明涉及医学图像处理的技术领域，尤其涉及一种基于光学跟踪探针划取点的手术导航空间配准方法，以及基于光学跟踪探针划取点的手术导航空间配准装置。The present invention relates to the technical field of medical image processing, in particular to a surgical navigation space registration method based on optical tracking probe drawing points and a surgical navigation space registration device based on optical tracking probe drawing points.

背景技术Background technique

作为手术导航系统的关键步骤，不准确的图像配准将会导致时间损失、不良事件或手术失败，例如切除位置错误或不完整的治疗。往往需要选择一个坐标系统作为世界坐标系，然后将系统中的所有坐标系统匹配到我们所选择的世界坐标系统中。As a critical step in surgical navigation systems, inaccurate image registration will lead to lost time, adverse events, or surgical failures, such as wrongly positioned resections or incomplete treatments. It is often necessary to select a coordinate system as the world coordinate system, and then match all coordinate systems in the system to the world coordinate system we choose.

传统的图像配准都是通过基于基准点的配准方法实现的。在基于基准点的配准方法中，需要在图像空间和人物理空间中分别找到至少3个不共线的点对。图像空间中的点通过在屏幕上用鼠标点击来获取，人物理空间中的点通过使用被跟踪的探针接触它们来获取。这样，就可以通过最小化所有对应点之间距离的平均值来计算两个空间之间的转换关系。然而，由于基于基准点的配准方法存在操作过程复杂、成本高、花费时间长、有创伤等固有问题，已经不再是这个方向的研究热点，研究者们致力于寻找一种快速、无标、无创的手术导航空间配准方法。近年来，陆续出现了一系列基于表面的图像配准方法，这种方法因其无标、无创、节约时间等优点成为研究的热点，逐渐被应用于颅底外科、神经外科、肝胆外科、口腔科及骨关节科等多个导航领域。以往的多项研究表明，当前手术导航中基于表面配准方法大多采用粗到精的配准策略，在粗配准过程中通过手动选择等方式为随后的精配准提供良好的初始姿态。已经提出了用于图像空间中点云自动配准的各种方法，并进行了临床测试。Fan等人提出了一种提取标识点的算法，可以自动将扫描的真实空间中的点云配准到图像空间中重建的表面，但需要扫描较为完整的头部点云，这无疑增加了临床使用难度。Hakje等人提出一种基于探针取点的自动图像配准算法，利用最小二乘投影和ICP算法来筛选探针所拾取的点，但忽略了初始配准姿态和ICP易于陷入局部最优的问题。Liu等人提出了一种基于3D曲面特征描述的自动配准算法，并使用ICP算法得到最优化的结果。Traditional image registration is realized by the registration method based on fiducial points. In the fiducial-based registration method, at least 3 non-collinear point pairs need to be found in the image space and the human physical space respectively. Points in image space are acquired by clicking on the screen with the mouse, and points in human physical space are acquired by touching them with the tracked probe. In this way, the transformation relationship between two spaces can be calculated by minimizing the average of the distances between all corresponding points. However, due to the inherent problems of complex operation process, high cost, long time-consuming, and trauma, the registration method based on fiducial points is no longer a research hotspot in this direction. Researchers are committed to finding a fast, standard-free , Non-invasive surgical navigation spatial registration method. In recent years, a series of surface-based image registration methods have emerged one after another. This method has become a research hotspot due to its advantages of being non-standard, non-invasive, and time-saving, and has been gradually applied in skull base surgery, neurosurgery, hepatobiliary surgery, and oral surgery. Department of Orthopedics and Orthopedics and other navigation fields. A number of previous studies have shown that most of the current surface-based registration methods in surgical navigation use a coarse-to-fine registration strategy, and provide a good initial pose for the subsequent fine registration through manual selection during the coarse registration process. Various methods for automatic registration of point clouds in image space have been proposed and clinically tested. Fan et al. proposed an algorithm for extracting marker points, which can automatically register the scanned point cloud in the real space to the reconstructed surface in the image space, but it needs to scan a relatively complete head point cloud, which undoubtedly increases the clinical Difficulty of use. Hakje et al. proposed an automatic image registration algorithm based on probe point selection, using least square projection and ICP algorithm to screen the points picked up by the probe, but ignored the initial registration pose and ICP that tend to fall into local optimum question. Liu et al. proposed an automatic registration algorithm based on 3D surface feature description, and used the ICP algorithm to obtain the optimal result.

目前常见的点云获取设备主要有被跟踪的激光测距仪、被跟踪的探针以及被跟踪的三维扫描仪三种。被跟踪的激光测距仪和被跟踪的三维扫描仪这两种方法的基本原理是：将跟踪标志物附着在激光测距仪(或三维扫描仪)上，然后利用特定的方法(如手眼标定)对其进行标定，获取其成像数据与跟踪标志物之间的固定转换关系，这样即可对其进行实时跟踪，并将扫描的点云数据变换到跟踪设备坐标系下，完成配准过程。然而，由于临床环境复杂，3D扫描仪所扫描的数据往往包含大量无效点云需要手动去除，这无疑大大增加了配准任务的难度。且价格高昂的3D扫描仪也对其广泛应用于临床环境产生了阻碍。At present, the common point cloud acquisition equipment mainly includes three kinds of tracked laser range finders, tracked probes and tracked 3D scanners. The basic principle of the two methods of the tracked laser rangefinder and the tracked 3D scanner is: attach the tracking marker to the laser rangefinder (or 3D scanner), and then use a specific method (such as hand-eye calibration ) to calibrate it to obtain the fixed conversion relationship between its imaging data and tracking markers, so that it can be tracked in real time, and the scanned point cloud data is transformed into the coordinate system of the tracking device to complete the registration process. However, due to the complex clinical environment, the data scanned by 3D scanners often contain a large number of invalid point clouds that need to be manually removed, which undoubtedly greatly increases the difficulty of the registration task. And the high price of 3D scanners has also hindered its widespread use in clinical settings.

发明内容Contents of the invention

为克服现有技术的缺陷，本发明要解决的技术问题是提供了一种基于光学跟踪探针划取点的手术导航空间配准方法，其更加易于操作且成本更低，配准精度与基于基准点配准得到的结果相近。In order to overcome the defects of the prior art, the technical problem to be solved by the present invention is to provide a surgical navigation space registration method based on optical tracking probe drawing points, which is easier to operate and lower in cost, and the registration accuracy is the same as that based on Similar results were obtained for fiducial registration.

本发明的技术方案是：这种基于光学跟踪探针划取点的手术导航空间配准方法，其包括以下步骤：The technical solution of the present invention is: this kind of surgical navigation space registration method based on optical tracking probe drawing point, it comprises the following steps:

(1)利用一个通用的人面部模板，通过模板匹配方法以剔除图像空间中冗余的表面数据；(1) Utilize a common human face template to remove redundant surface data in the image space through template matching method;

(2)使用光学跟踪探针在真实人面部取点，并提出一个约束指标，以重建均匀的人面部轮廓；(2) Use the optical tracking probe to take points on the real human face and propose a constraint index to reconstruct a uniform human facial contour;

(3)通过粗到精的增量点云配准方法以及配准约束指标，约束所采集点云的完整性和配准结果的正确性，最终将图像空间与真实空间精确对齐。(3) Through the coarse-to-fine incremental point cloud registration method and the registration constraint index, the integrity of the collected point cloud and the correctness of the registration results are constrained, and finally the image space is accurately aligned with the real space.

本发明利用一个通用的患者面部模板，提出一种模板匹配方法以剔除图像空间中冗余的表面数据，使用光学跟踪探针在真实患者面部取点，并提出一个约束指标，以重建均匀的患者面部轮廓，提出一种粗到精的增量点云配准方法以及配准约束指标，以约束所采集点云的完整性和配准结果的正确性，最终将图像空间与真实患者空间精确对齐，所提出的配准方法集成在了导航系统中，并进行了仿体实验和临床试验验证，证实更加易于操作且成本更低，配准精度与基于基准点配准得到的结果相近。The present invention uses a general patient face template, proposes a template matching method to eliminate redundant surface data in the image space, uses optical tracking probes to take points on the real patient's face, and proposes a constraint index to reconstruct a uniform patient Facial contours, a coarse-to-fine incremental point cloud registration method and a registration constraint index are proposed to constrain the integrity of the collected point cloud and the correctness of the registration results, and finally accurately align the image space with the real patient space , the proposed registration method is integrated in the navigation system, and has been verified by phantom experiments and clinical trials, which proves that it is easier to operate and lower cost, and the registration accuracy is similar to the results obtained based on fiducial point registration.

还提供了基于光学跟踪探针划取点的手术导航空间配准装置，其包括：A spatial registration device for surgical navigation based on optical tracking probe drawing points is also provided, which includes:

数据剔除模块，其配置来利用一个通用的人面部模板，通过模板匹配方法以剔除图像空间中冗余的表面数据；A data removal module configured to use a common human face template to remove redundant surface data in the image space through a template matching method;

人脸重建模块，其配置来使用光学跟踪探针在真实人面部取点，并提出一个约束指标，以重建均匀的人面部轮廓；A face reconstruction module configured to use an optical tracking probe to take points on a real human face and propose a constraint index to reconstruct a uniform human facial contour;

配准对齐模块，其配置来通过粗到精的增量点云配准方法以及配准约束指标，约束所采集点云的完整性和配准结果的正确性，最终将图像空间与真实空间精确对齐。The registration and alignment module is configured to constrain the integrity of the collected point cloud and the correctness of the registration result through the coarse-to-fine incremental point cloud registration method and the registration constraint index, and finally accurately align the image space with the real space align.

附图说明Description of drawings

图1是根据本发明的基于光学跟踪探针划取点的手术导航空间配准方法的流程图。FIG. 1 is a flow chart of a method for surgical navigation spatial registration based on optical tracking probe points drawn according to the present invention.

具体实施方式Detailed ways

如图1所示，这种基于光学跟踪探针划取点的手术导航空间配准方法，其包括以下步骤：As shown in Figure 1, this surgical navigation spatial registration method based on optical tracking probe drawing points includes the following steps:

(1)利用一个通用的人面部模板，通过模板匹配方法以剔除图像空间中冗余的表面数据；(1) Utilize a common human face template to remove redundant surface data in the image space through template matching method;

(2)使用光学跟踪探针在真实人面部取点，并提出一个约束指标，以重建均匀的人面部轮廓；(2) Use the optical tracking probe to take points on the real human face and propose a constraint index to reconstruct a uniform human facial contour;

(3)通过粗到精的增量点云配准方法以及配准约束指标，约束所采集点云的完整性和配准结果的正确性，最终将图像空间与真实空间精确对齐。(3) Through the coarse-to-fine incremental point cloud registration method and the registration constraint index, the integrity of the collected point cloud and the correctness of the registration results are constrained, and finally the image space is accurately aligned with the real space.

本发明利用一个通用的患者面部模板，提出一种模板匹配方法以剔除图像空间中冗余的表面数据，使用光学跟踪探针在真实患者面部取点，并提出一个约束指标，以重建均匀的患者面部轮廓，提出一种粗到精的增量点云配准方法以及配准约束指标，以约束所采集点云的完整性和配准结果的正确性，最终将图像空间与真实患者空间精确对齐，所提出的配准方法集成在了导航系统中，并进行了仿体实验和临床试验验证，证实更加易于操作且成本更低，配准精度与基于基准点配准得到的结果相近。The present invention uses a general patient face template, proposes a template matching method to eliminate redundant surface data in the image space, uses optical tracking probes to take points on the real patient's face, and proposes a constraint index to reconstruct a uniform patient Facial contours, a coarse-to-fine incremental point cloud registration method and a registration constraint index are proposed to constrain the integrity of the collected point cloud and the correctness of the registration results, and finally accurately align the image space with the real patient space , the proposed registration method is integrated in the navigation system, and has been verified by phantom experiments and clinical trials, which proves that it is easier to operate and lower cost, and the registration accuracy is similar to the results obtained based on fiducial point registration.

优选地，所述步骤(2)中，人的面部点云是由光学跟踪探针在人面部划动收集得到的，并且人的皮肤表面是从术前CT影像中分割并重建得到的；利用一个标准化模板分割出人面部相对较硬的区域用于配准，同时在拾取点云和精配准步骤分别加入约束条件，以提高算法的鲁棒性；4PCS算法将两个点云配准，以获得良好的初始姿态；然后执行ICP算法，以得到最佳的姿态矩阵，这通过最小化公式(1)来实现：Preferably, in the step (2), the face point cloud of the person is collected by the optical tracking probe on the face of the person, and the skin surface of the person is segmented and reconstructed from the preoperative CT image; A standardized template is used to segment the relatively hard area of the face for registration, and at the same time, constraints are added to the picking point cloud and fine registration steps to improve the robustness of the algorithm; the 4PCS algorithm registers the two point clouds, to obtain a good initial attitude; then perform the ICP algorithm to obtain the best attitude matrix, which is achieved by minimizing formula (1):

其中，R和t为能够最小化两个点云之间欧氏距离的旋转矩阵和平移向量；p_i和q_i分别表示两个点云中第i个相对应的点。Among them, R and t are the rotation matrix and translation vector that can minimize the Euclidean distance between the two point clouds; p_i and q_i represent the i-th corresponding points in the two point clouds, respectively.

优选地，所述步骤(1)中，在配准过程中加入中间模板，以剔除冗余的面部区域；所使用的中间模板是一个覆盖人额头和鼻子区域的T字形表面数据，该模板由高斯人脸平均模型中手动截取；定义由医学影像中获取的人表面数据的原始点集为P；首先，使用4PCS算法将该模板对齐到所提取的人面部数据中，对齐后的模板表示为M；然后，利用kd-tree算法在P中快速搜索与M中的点距离小于给定阈值d_δ(本文取d_δ＝10mm)的所有点；最终，得到P中被模板以d_δ距离覆盖的点集P^*用于后续配准，表示为公式(2)：Preferably, in the step (1), an intermediate template is added in the registration process to remove redundant facial regions; the intermediate template used is a T-shaped surface data covering the forehead and nose area of a person, and the template is composed of Manual interception in the Gaussian face average model; define the original point set of human surface data obtained from medical images as P; first, use the 4PCS algorithm to align the template to the extracted human face data, and the aligned template is expressed as M; then, use the kd-tree algorithm to quickly search for all points in P that are less than a given threshold d_δ (this paper takes d_δ = 10mm) in P; finally, it is obtained that P is covered by the template with d_δ distance The point set P^* for subsequent registration is expressed as formula (2):

P^*＝{P_i|d(P_i,M)<d_δ},(i＝1,2,…,N) (2)P^* ＝{P_i |d(P_i ,M)<d_δ },(i=1,2,…,N) (2)

其中，d(P_i,M)表示点P_i到点集M中所有点的最近距离；N为点集P中的点的个数。P^*为医学影像空间中的人面部数据，被用作后续点云配准算法的输入。Among them, d(P_i , M) represents the shortest distance from point P_i to all points in point set M; N is the number of points in point set P. P^* is the face data in the medical image space, which is used as the input of the subsequent point cloud registration algorithm.

优选地，所述步骤(2)中，点拾取与点云配准同步进行，随着所拾取点的增加，真实人的面部轮廓逐渐被重建出来；在拾取过程中，用最近点距离(Nearest Point Distance，NPD)来约束点拾取和添加过程，以保证所重建点云的均匀性；NPD约束与点云配准中的3个约束指标共同构成完整的约束条件，只有当这4个约束全部被满足时，才认为当前点拾取和配准结果是最佳的；Preferably, in the step (2), the point picking is carried out synchronously with the point cloud registration, and with the increase of the picked points, the facial contour of the real person is gradually reconstructed; in the picking process, the nearest point distance (Nearest Point Distance, NPD) to constrain the point picking and adding process to ensure the uniformity of the reconstructed point cloud; the NPD constraint and the three constraint indicators in the point cloud registration together constitute a complete constraint condition, only when the four constraints are all When is satisfied, the current point picking and registration results are considered to be the best;

对划取点云过程添加了NPD约束，定义当前拾取的点集为Q，含有N个点，对于即将添加的第N+1个点q_N+1，利用公式(8)检测其是否满足NPD约束：NPD constraints are added to the point cloud drawing process, and the currently picked point set is defined as Q, which contains N points. For the N+1th point q_N+1 to be added, use formula (8) to check whether it satisfies NPD constraint:

其中，d(q_N+1,Q)表示点q_N+1到Q中所有点的最近距离；

(本文取

)为预先设定的距离阈值；若该式成立，则将新拾取的点q_N+1添加到点集Q中，否则丢弃。Among them, d(q_N+1 , Q) represents the shortest distance from point q_N+1 to all points in Q;

(This article takes

) is the preset distance threshold; if this formula is true, add the newly picked point q_N+1 to the point set Q, otherwise discard it.

优选地，所述步骤(3)中，与点云配准中的3个约束指标分别是覆盖率(CoverageRatio，CR)、奇异点占比(Outliers Ratio，OR)和表面配准误差(Surface RegistrationError，SRE)；当这3个约束指标全部被满足时，则当前点集拾取和配准过程结束，得到最佳的空间配准矩阵；4PCS算法被用来执行粗配准，以获得两组点集之间良好的初始姿态；ICP算法被用来执行精配准，以进一步提高配准精度；Preferably, in the step (3), the three constraint indicators in the point cloud registration are coverage ratio (CoverageRatio, CR), singular point ratio (Outliers Ratio, OR) and surface registration error (Surface RegistrationError , SRE); when all three constraints are satisfied, the current point set picking and registration process ends, and the best spatial registration matrix is obtained; 4PCS algorithm is used to perform rough registration to obtain two sets of points Good initial pose between sets; ICP algorithm is used to perform fine registration to further improve registration accuracy;

CR反映了所拾取点集覆盖目标区域的比例，用于评价划取范围是否可靠，目标区域指T模板所截取的点集P^*，给定点到点集的距离阈值

将所划取的点集划分为局内点和局外点，局内点集定义为：CR reflects the proportion of the picked point set covering the target area, and is used to evaluate whether the drawn range is reliable. The target area refers to the point set P^* intercepted by the T template, and the distance threshold from a given point to point set

The drawn point set is divided into intra-office point and out-of-office point, and the intra-office point set is defined as:

其中，T′表示当前配准结果；

表示利用T′将P^*和Q对齐后

与点集Q的最近距离；N_P为点集P^*中点的个数；则CR的定义表示为：Among them, T' represents the current registration result;

Indicates that after aligning P^* and Q using T′

The shortest distance to the point set Q; N_P is the number of points in the point set P^* ; then the definition of CR is expressed as:

与CR不同，OR反映了所拾取点集中局外点的占比，用于判断当前配准结果是否陷入局部最优，给定点到点集的距离阈值

则所拾取点集中的局外点Q_outlier和OR被定义如下：Different from CR, OR reflects the proportion of outliers in the picked point set, and is used to judge whether the current registration result falls into a local optimum, and the distance threshold from a given point to a point set

Then the outlier points Q_outlier and OR in the picked point set are defined as follows:

SRE被用来评估导航中表面配准的精确度，表示所拾取点集Q中每一个点到点集P^*的最小距离的平均值，SRE定义为：SRE is used to evaluate the accuracy of surface registration in navigation, which represents the average value of the minimum distance from each point in the picked point set Q to the point set P^* , and SRE is defined as:

本领域普通技术人员可以理解，实现上述实施例方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成，所述的程序可以存储于一计算机可读取存储介质中，该程序在执行时，包括上述实施例方法的各步骤，而所述的存储介质可以是：ROM/RAM、磁碟、光盘、存储卡等。因此，与本发明的方法相对应的，本发明还同时包括一种基于光学跟踪探针划取点的手术导航空间配准装置，该装置通常以与方法各步骤相对应的功能模块的形式表示。该装置包括：Those of ordinary skill in the art can understand that all or part of the steps in the method of the above-mentioned embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When executed, it includes the steps of the method in the above embodiments, and the storage medium may be: ROM/RAM, magnetic disk, optical disk, memory card, etc. Therefore, corresponding to the method of the present invention, the present invention also includes a space registration device for surgical navigation based on optical tracking probe drawing points, which is usually expressed in the form of functional modules corresponding to each step of the method . The unit includes:

数据剔除模块，其配置来利用一个通用的人面部模板，通过模板匹配方法以剔除图像空间中冗余的表面数据；A data removal module configured to use a common human face template to remove redundant surface data in the image space through a template matching method;

人脸重建模块，其配置来使用光学跟踪探针在真实人面部取点，并提出一个约束指标，以重建均匀的人面部轮廓；A face reconstruction module configured to use an optical tracking probe to take points on a real human face and propose a constraint index to reconstruct a uniform human facial contour;

配准对齐模块，其配置来通过粗到精的增量点云配准方法以及配准约束指标，约束所采集点云的完整性和配准结果的正确性，最终将图像空间与真实空间精确对齐。The registration and alignment module is configured to constrain the integrity of the collected point cloud and the correctness of the registration result through the coarse-to-fine incremental point cloud registration method and the registration constraint index, and finally accurately align the image space with the real space align.

在利用光学跟踪探针拾取点集的过程中，利用所提出的方法同步执行粗到精的点云配准。结合所提出的约束指标，实时检测当前配准质量。CR指标将点拾取区域限制在了“T”模板匹配得到的人脸表面数据周围，以保证利用探针所重建表面的完整性。OR指标直接反映了当前拾取点集的质量和配准质量，确保点云配准结果不会陷入局部最优。SRE作为一个比较常规的指标，则计算了两个点集之间的匹配程度。在本文所提出的4个指标NPD、CR、OR和SRE的约束下，引导操作者完成人面部点集拾取和重建过程，最终得到最佳的图像到真实配准的空间变换。Coarse-to-fine point cloud registration is performed synchronously with the proposed method during the process of picking a point set with an optical tracking probe. Combined with the proposed constraint metrics, the current registration quality is detected in real time. The CR index limits the point picking area around the face surface data obtained by "T" template matching to ensure the integrity of the surface reconstructed by the probe. The OR index directly reflects the quality and registration quality of the currently picked point set, ensuring that the point cloud registration results will not fall into local optimum. As a more conventional indicator, SRE calculates the degree of matching between two point sets. Under the constraints of the four indicators NPD, CR, OR and SRE proposed in this paper, the operator is guided to complete the process of face point set picking and reconstruction, and finally the best image-to-real registration space transformation is obtained.

以上所述，仅是本发明的较佳实施例，并非对本发明作任何形式上的限制，凡是依据本发明的技术实质对以上实施例所作的任何简单修改、等同变化与修饰，均仍属本发明技术方案的保护范围。The above are only preferred embodiments of the present invention, and are not intended to limit the present invention in any form. Any simple modifications, equivalent changes and modifications made to the above embodiments according to the technical essence of the present invention are still within the scope of this invention. The protection scope of the technical solution of the invention.

Claims (3)

1. The surgical navigation space registration method based on the optical tracking probe marking point is characterized in that: which comprises the following steps:

(1) Utilizing a universal human face template to remove redundant surface data in an image space by a template matching method;

(2) Using an optical tracking probe to take points on the real human face, and providing a constraint index to reconstruct a uniform human face contour;

(3) The integrity of the acquired point cloud and the correctness of a registration result are restricted through a coarse-to-fine incremental point cloud registration method and a registration constraint index, and finally, an image space is accurately aligned with a real space;

in the step (2), the point cloud of the human face is obtained by the optical tracking probe through paddling and collecting on the human face, and the skin surface of the human is obtained by segmentation and reconstruction from the preoperative CT image; segmenting a relatively hard region of the human face by utilizing a standardized template for registration, and simultaneously adding constraint conditions in the steps of point cloud picking and precise registration respectively to improve the robustness of the algorithm; registering the two point clouds by a 4PCS algorithm to obtain a good initial posture;

the ICP algorithm is then performed to obtain the optimal attitude matrix, which is achieved by minimizing equation (1):

wherein, R and t are a rotation matrix and a translation vector which can minimize the Euclidean distance between two point clouds; p is a radical of_i And q is_i Respectively representing ith corresponding points in the two point clouds;

in the step (2), point picking and point cloud registration are synchronously carried out, and the facial contour of a real person is gradually reconstructed along with the increase of the picked points; in the picking process, the NPD point picking and adding process is restricted by the nearest point distance so as to ensure the uniformity of the reconstructed point cloud; the NPD constraint and 3 constraint indexes in the point cloud registration jointly form a complete constraint condition, and only when the 4 constraints are all met, the current point picking and registration result is considered to be optimal;

NPD constraint is added in the process of drawing point cloud, a currently picked point set is defined as Q and contains N points, and for the (N + 1) th point Q to be added, the point is added_N+1 Using equation (8) to detect whether it satisfies the NPD constraint:

wherein d (q)_N+1 Q) represents a point Q_N+1 Closest distance to all points in Q;

is a preset distance threshold; if this equation holds true, the newly picked point q is pointed out_N+1 Adding the point set Q, otherwise, discarding;

in the step (3), the 3 constraint indexes in point cloud registration are respectively coverage ratio CR, singular point ratio OR and surface registration error SRE; when all the 3 constraint indexes are met, finishing the current point set picking and registering process to obtain an optimal spatial registering matrix; the 4PCS algorithm is used to perform a coarse registration to obtain a good initial pose between the two sets of points; the ICP algorithm is used to perform fine registration to further improve registration accuracy;

CR reflects the proportion of the picked point set covering the target area for evaluating whether the drawing range is reliable or not, and the target area refers to the point set P intercepted by the T template^* Given a threshold of distance from point to point set

Dividing the drawn point set into local points and local points, wherein the local point set is defined as:

wherein T' represents a current registration result;

represents P by T^* After alignment with Q

The closest distance to the set of points Q; n is a radical of_P As a set of points P^* The number of midpoints; the definition of CR is then expressed as:

unlike CR, OR reflects the percentage of picked points to localized outliers, andgiving a distance threshold value from a point to a point set when judging whether the current registration result falls into the local optimum

The outer point Q of the picked point set_outlier And OR is defined as follows:

SRE is used to assess the accuracy of surface registration in navigation, representing each point in the picked point set Q to the point set P^* SRE is defined as:

2. the surgical navigation spatial registration method based on optical tracking probe scribing points according to claim 1, characterized in that: in the step (1), an intermediate template is added in the registration process to remove redundant face regions; the used middle template is T-shaped surface data covering the forehead and nose areas of the human body, and the template is manually intercepted from a Gaussian face average model; defining an original point set of human surface data acquired from a medical image as P; firstly, aligning the template into the extracted human face data by using a 4PCS algorithm, wherein the aligned template is represented as M; then, quickly searching for points in P with the distance less than a given threshold value d from the point in M by using a kd-tree algorithm_δ All points of (a); finally, the template in P is obtained as d_δ Set of points P covered by distance^* For subsequent registration, expressed as equation (2):

P^* ＝{P_i |d(P_i ,M)<d_δ },(i＝1,2,…,N) (2)

wherein d (P)_i M) represents the point P_i The closest distance to all points in the point set M; n is the number of points in the point set P, P^* Is human face data in the medical image space and is used as an input of a subsequent point cloud registration algorithm.

3. The device for realizing the surgical navigation space registration method based on the optical tracking probe scribing point as the claim 1 is characterized in that: it includes:

the data removing module is configured to remove redundant surface data in the image space by a template matching method by utilizing a universal human face template;

a face reconstruction module configured to take points on a real face using an optical tracking probe and to propose a constraint index to reconstruct a uniform human face contour;

and the registration alignment module is configured to constrain the integrity of the acquired point cloud and the correctness of a registration result through a coarse-to-fine incremental point cloud registration method and a registration constraint index, and finally, accurately align the image space with the real space.

Images