An oral cbct automatic registration method

Abstract

The application provides an oral CBCT automatic registration algorithm, which comprises the following steps: a tooth centroid heat map label is made based on an oral CBCT image; a tooth centroid detection model is trained; tooth centroid detection and post-processing are respectively performed on preoperative and postoperative CBCT images to obtain three-dimensional coordinates of each tooth centroid point; a clustering algorithm is used to cluster the tooth centroid coordinates to separate the upper and lower jaws; the tooth and jaw centroid points are sorted to obtain an ordered centroid point set of two CBCT data; a point registration algorithm is used to register the tooth and jaw centroid points; CBCT point cloud registration is performed to obtain a preoperative and postoperative oral CBCT registration conversion relationship. The method combines centroid extraction based on deep learning, a point registration algorithm and an ICP algorithm, has the advantages of simple operation, fast registration speed and high precision, provides accurate data basis for medical staff to perform postoperative evaluation of dental implant surgery, and has wide application prospect and value.

Description

Technical Field

[0001] This invention relates to the field of medical image processing, and more particularly to an automatic registration method for oral CBCT. Background Technology

[0002] In recent years, dynamic navigation technology has been widely used and developed in the field of dental implant surgery. The main process of dynamic navigation implant surgery is to plan the relative position of the implant in the patient's oral cavity CBCT before the operation, and to navigate and track the relative position of the CBCT model, the planned implant, and the actual implant in real time during the operation, so as to assist the doctor in placing the actual implant into the planned implant position, that is, to perform the operation more precisely.

[0003] Assessing the implantation accuracy after dental implantation is an important evaluation criterion for the effectiveness of dynamic navigation implant surgery. However, the registration accuracy of the CBCT model before and after surgery directly affects the accuracy and reliability of the assessment results. Therefore, before assessing implantation accuracy, it is necessary to first register the postoperative oral CBCT to the position of the preoperative oral CBCT before calculating the accuracy error between the actual implant and the planned implant.

[0004] Currently, commonly used CBCT registration methods in oral surgery include point registration and Iterative Closest Point (ICP) algorithms. The point registration algorithm involves manually selecting at least three pairs of registration points at the same location on both pre- and post-operative CBCT models, such as the cusps or indentations of a tooth. The transformation relationship between these two sets of points is then calculated and applied to CBCT registration. The advantage of this method is its high success rate due to manual point selection. However, the manual selection process is cumbersome, and the subjective nature of point selection can lead to lower registration accuracy. The ICP algorithm first extracts isosurfaces from two CBCT models, then performs point cloud registration on these isosurfaces to obtain their transformation relationship. Finally, this transformation relationship is applied to the CBCT model registration. This method is an automatic registration algorithm with advantages such as high registration accuracy and ease of operation when successful. However, the drawback is that the two CBCT models may have different acquisition times and differ in the presence or absence of implants, leading to potentially large differences in the extracted isosurfaces and thus a lower registration success rate. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art described in the background section and provide an automatic registration method for oral CBCT. This method combines centroid extraction, point registration algorithms, and ICP algorithms based on deep learning. It has the advantages of simple operation, fast registration speed, and high accuracy, providing accurate data for postoperative evaluation of dental implant surgery by medical staff and has broad application prospects and value.

[0006] This invention is achieved using the following technical solution: A method for automatic registration in oral CBCT is provided, comprising the following steps:

[0007] Thermal labels for tooth centroids based on oral CBCT images;

[0008] Training a tooth centroid detection model;

[0009] The centroids of teeth were detected and post-processed on the CBCT images before and after the operation to obtain the three-dimensional coordinates of the centroid of each tooth.

[0010] Clustering algorithms are used to cluster the centroid coordinates of the teeth to separate the upper and lower jaws;

[0011] The centroids of the teeth are sorted to obtain an ordered set of centroids from the two CBCT datasets;

[0012] The centroids of the dentition are registered using a point registration algorithm;

[0013] CBCT point cloud registration was performed to obtain the preoperative and postoperative oral CBCT registration conversion relationship.

[0014] Furthermore, the process of creating thermal mapping labels for tooth centroids based on oral CBCT images includes:

[0015] Based on the oral CBCT images, three orthogonal slices are displayed, including transverse, sagittal, and coronal planes;

[0016] For each tooth, delineate the foreground and background areas on three orthogonal slices;

[0017] Based on the initially outlined foreground and background regions, the segmentation results of the teeth are obtained using a graph cut algorithm.

[0018] The center point of each tooth is calculated based on its segmentation results, thus obtaining the centroid of the tooth.

[0019] Gaussian smoothing is applied to the centroid image to obtain the heat map label of the tooth centroid.

[0020] Furthermore, the smoothing radius of the Gaussian smoothing applied to the centroid image is set to 2 mm.

[0021] Furthermore, the thermal image label of the tooth centroid is a three-dimensional image with a numerical range of [0, 1], where the value at the centroid point is 1, and the value of voxel points farther away from the centroid is smaller.

[0022] Furthermore, the training model for detecting the tooth centroid includes:

[0023] A deep learning model structure is used to train the input CBCT image and the corresponding centroid heatmap label until the loss function converges, thus obtaining the parameters of the neural network model used for centroid detection.

[0024] Furthermore, the deep learning model adopts the UNet model, and the loss function adopts the mean squared error loss function MSELoss.

[0025] Furthermore, the step of performing tooth centroid detection and post-processing on pre- and post-operative CBCT images to obtain the three-dimensional coordinates of the centroid point of each tooth includes:

[0026] Using the tooth centroid detection model, centroid detection was performed on CBCT images before and after the operation, and the results of the centroid detection were obtained as heat maps.

[0027] The heatmap is binarized with a threshold of 0.5;

[0028] Connectivity analysis is performed on the binarized image, where each connected component represents a tooth;

[0029] Calculate the center point of each connected domain to obtain the centroid coordinates of each tooth in the LPS coordinate system.

[0030] Furthermore, the step of using a clustering algorithm to cluster the centroid coordinates of the teeth to separate the upper and lower jaws includes:

[0031] Using the two points with the smallest and largest Z-axis coordinates among the centroids as the initial cluster centers, calculate the distance in the Z direction from each centroid to the two cluster centers;

[0032] Each centroid is assigned a cluster based on the distance, with the number of clusters set to 2, namely the maxilla and mandible. The cluster to which the cluster center with the smaller distance belongs is the cluster to which the centroid belongs.

[0033] After each centroid is assigned a cluster, the center points of the two clusters are calculated, and the two center points are used as the new cluster centers. The distance from each centroid to the cluster center is calculated again and the cluster is assigned.

[0034] Repeat the above process until the cluster center no longer changes. At this point, the centroids are clustered into two clusters, representing the centroid set of the maxillary teeth and the centroid set of the mandibular teeth, respectively.

[0035] Furthermore, the clustering algorithm employs the K-Means algorithm.

[0036] Furthermore, the sorting of the centroids of the dentition yields an ordered set of centroids from the two CBCT datasets, including:

[0037] Based on the planned implant location, select the corresponding set of centroids. If the planned implant is in the maxilla, sort the set of centroids in the mandible; if the planned implant is in the mandible, sort the set of centroids in the maxilla.

[0038] The coordinate range of the oral CBCT image in the LPS coordinate system is denoted as [xmin, ymin, zmin] to [xmax, ymax, zmax].

[0039] A corner point is determined based on the image range. There are four points that can be used as corner points: [xmin, ymax, zmin], [xmin, ymax, zmax], [xmax, ymax, zmin], and [xmax, ymax, zmax].

[0040] Select one of the corner points as the target corner point, find the centroid point closest to the target corner point as the first point, search the set of centroid points for the centroid point closest to the target corner point as the next point, and remove the found point from the set;

[0041] Repeat the above process until the set is empty, resulting in two ordered centroid sets of CBCT data, denoted as set A and set B, respectively. P A and set P B At this point, the centroids before and after the operation correspond one-to-one.

[0042] Furthermore, the registration of the centroid points of the dentition using the point registration algorithm includes:

[0043] Using the point registration algorithm, calculate the set of two centroids. P B arrive P A Transformation matrix T 1 As shown in the following formula:

[0044]

[0045] Furthermore, the CBCT point cloud registration is performed to obtain the preoperative and postoperative oral CBCT registration conversion relationship, including:

[0046] The preoperative CBCT model is denoted as A, and the postoperative CBCT model is denoted as B;

[0047] Through the relation matrix Move the postoperative CBCT model B to a new location. This serves as the result of coarse registration;

[0048] For A and Two sets of point cloud data were obtained by isosurface extraction. and The following relationship is satisfied:

[0049]

[0050]

[0051] Point cloud calculation using the ICP algorithm To Point Cloud Conversion relationship It satisfies the following transformation relationship:

[0052]

[0053] Finally, the transformation matrix from postoperative CBCT model B to preoperative CBCT model A can be obtained. The following relationship is satisfied:

[0054]

[0055] This invention proposes an automatic registration method for oral CBCT images. This method combines deep learning-based centroid extraction, point registration algorithms, and the ICP algorithm to automatically register oral CBCT images of the same patient. It boasts advantages such as simple operation, fast registration speed, and high accuracy, providing accurate data for postoperative evaluation of dental implant surgery and demonstrating broad application prospects and value. Compared with existing manual point selection registration algorithms, this invention offers both ease of operation and higher registration accuracy; compared with the ICP algorithm, the proposed method has a higher registration success rate. Attached Figure Description

[0056] The features, advantages and technical effects of exemplary embodiments of the present invention will now be described with reference to the accompanying drawings.

[0057] Figure 1 This is a flowchart illustrating the automatic registration method for oral CBCT provided in an embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram of the process for creating thermal labels for the centroid of teeth according to an embodiment of the present invention;

[0059] Figure 3 This is a schematic diagram of the centroid detection results provided in an embodiment of the present invention;

[0060] Figure 4 This is a schematic diagram of the model prediction and post-processing results provided in the embodiments of the present invention;

[0061] Figure 5This is a schematic diagram of the four corner points of the CBCT image range provided in the embodiment of the present invention;

[0062] Figure 6 This is a schematic diagram of the centroid sorting process provided in an embodiment of the present invention. Detailed Implementation

[0063] The features and exemplary embodiments of various aspects of this disclosure will now be described in detail. To make the objectives, technical solutions, and advantages of this disclosure clearer, the disclosure will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are intended to explain this disclosure only and not to limit it. For those skilled in the art, this disclosure can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this disclosure by illustrating examples.

[0064] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0065] To better understand the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0066] Figure 1 This is a flowchart illustrating the automatic registration method for oral CBCT provided in an embodiment of the present invention.

[0067] like Figure 1 As shown, the present invention provides an automatic registration method for oral CBCT, which includes the following steps:

[0068] S101, Generating thermal labels for tooth centroids based on oral CBCT images;

[0069] S102, training a tooth centroid detection model;

[0070] S103 performs tooth centroid detection and post-processing on CBCT images before and after the operation to obtain the three-dimensional coordinates of the centroid point of each tooth.

[0071] S104, Clustering algorithm is used to cluster the centroid coordinates of the teeth to separate the upper and lower jaws;

[0072] S105, sort the centroids of the teeth and jaws to obtain an ordered set of centroids from the two CBCT data;

[0073] S106, the centroid of the tooth jaw is registered using a point registration algorithm;

[0074] S107, perform CBCT point cloud registration to obtain the preoperative and postoperative oral CBCT registration conversion relationship.

[0075] Figure 2 This is a schematic diagram of the process for creating thermal labels for the centroid of teeth according to an embodiment of the present invention.

[0076] like Figure 2 As shown, the thermal mapping of tooth centroids based on oral CBCT images described in S101 includes:

[0077] Based on the oral CBCT images, three orthogonal slices are displayed, including transverse, sagittal, and coronal planes;

[0078] For each tooth, delineate the foreground and background areas on three orthogonal slices;

[0079] Based on the initially outlined foreground and background regions, the segmentation results of the teeth are obtained using a graph cut algorithm.

[0080] The center point of each tooth is calculated based on its segmentation results, thus obtaining the centroid of the tooth.

[0081] Gaussian smoothing is applied to the centroid image to obtain the heat map label of the tooth centroid.

[0082] Alternatively, other segmentation algorithms that achieve the same or similar results can be used when segmenting teeth, such as level sets, seed point growth, etc.

[0083] Optionally, the smoothing radius of the Gaussian smoothing applied to the centroid image is set to 2 mm.

[0084] Optionally, the thermal image label of the tooth centroid is a three-dimensional image with a numerical range of [0, 1], where the value at the centroid point is 1, and the value of voxel points farther away from the centroid is smaller.

[0085] Optionally, the training of the tooth centroid detection model described in S102 includes:

[0086] A deep learning model structure is used to train the input CBCT image and the corresponding centroid heatmap label until the loss function converges, thus obtaining the parameters of the neural network model used for centroid detection.

[0087] Optionally, the deep learning model adopts the UNet model, and the loss function adopts the mean squared error loss function MSELoss.

[0088] Optionally, the deep learning model may also adopt other deep learning model structures that can achieve the same or similar effects, and the loss function shown may also adopt other loss functions that can achieve the same or similar effects.

[0089] Optionally, S103 describes performing tooth centroid detection and post-processing on pre- and post-operative CBCT images to obtain the three-dimensional coordinates of the centroid point of each tooth, including:

[0090] Using the aforementioned tooth centroid detection model, centroid detection is performed on CBCT images, such as... Figure 3 As shown, the results of the centroid detection are presented as a heatmap;

[0091] The heatmap is binarized with a threshold of 0.5;

[0092] Perform connected component analysis on the binarized image, such as Figure 4 As shown, each connected component represents a tooth;

[0093] Calculate the center point of each connected domain to obtain the centroid coordinates of each tooth in the LPS coordinate system.

[0094] Optionally, the clustering algorithm used in S104 to cluster the centroid coordinates of the teeth to separate the upper and lower jaws includes:

[0095] Using the two points with the smallest and largest Z-axis coordinates among the centroids as the initial cluster centers, calculate the distance in the Z direction from each centroid to the two cluster centers;

[0096] Each centroid is assigned a cluster based on the distance, with the number of clusters set to 2, namely the maxilla and mandible. The cluster to which the cluster center with the smaller distance belongs is the cluster to which the centroid belongs.

[0097] After each centroid is assigned a cluster, the center points of the two clusters are calculated, and the two center points are used as the new cluster centers. The distance from each centroid to the cluster center is calculated again and the cluster is assigned.

[0098] Repeat the above process until the cluster center no longer changes. At this point, the centroids are clustered into two clusters, representing the centroid set of the maxillary teeth and the centroid set of the mandibular teeth, respectively.

[0099] Optionally, the clustering algorithm uses the K-Means algorithm.

[0100] Alternatively, the clustering algorithm may also employ other clustering algorithms that can achieve the same or similar results, such as the Learning Vectorization (LVQ) algorithm, the KNN algorithm, etc.

[0101] Optionally, the sorting of the centroids of the dentition described in S105 yields an ordered set of centroids from the two CBCT datasets, including:

[0102] Based on the planned implant location, select the corresponding set of centroids. If the planned implant is in the maxilla, sort the set of centroids in the mandible; if the planned implant is in the mandible, sort the set of centroids in the maxilla.

[0103] The coordinate range of the oral CBCT image in the LPS coordinate system is denoted as [xmin, ymin, zmin] to [xmax, ymax, zmax].

[0104] A corner point is determined based on the image range, such as... Figure 5 As shown, there are four points that can be used as corner points, namely [xmin, ymax, zmin], [xmin, ymax, zmax], [xmax, ymax, zmin], and [xmax, ymax, zmax].

[0105] Select point 3 as the target corner point, such as Figure 6 As shown, find the centroid 1 that is closest to the corner point 3 as the first point, search for the centroid 2 that is closest to the centroid 1 in the set of centroid points as the next point, and remove the centroid 1 from the set;

[0106] Repeat the above process until the set is empty, resulting in two ordered centroid sets of CBCT data, denoted as set A and set B, respectively. P A and set P B At this point, the centroids before and after the operation correspond one-to-one.

[0107] Optionally, the point registration algorithm described in S106 for registering the centroids of the dentition includes:

[0108] Using the point registration algorithm, calculate the set of two centroids. P B arrive P A Transformation matrix T 1 As shown in the following formula:

[0109]

[0110] Optionally, the CBCT point cloud registration described in S107, to obtain the preoperative and postoperative oral CBCT registration conversion relationship, includes:

[0111] The preoperative CBCT model is denoted as A, and the postoperative CBCT model is denoted as B;

[0112] Through the relation matrix Move the postoperative CBCT model B to a new location. This serves as the result of coarse registration;

[0113] For A and Two sets of point cloud data were obtained by isosurface extraction. and The following relationship is satisfied:

[0114]

[0115]

[0116] Point cloud calculation using the ICP algorithm To Point Cloud Conversion relationship It satisfies the following transformation relationship:

[0117]

[0118] Finally, the transformation matrix from postoperative CBCT model B to preoperative CBCT model A can be obtained. The following relationship is satisfied:

[0119]

[0120] Alternatively, other registration algorithms that can achieve the same or similar results can be used when performing CBCT point cloud registration, such as RANSAC point cloud registration algorithm, 4PCS registration algorithm, etc.

[0121] This invention proposes an automatic registration method for oral CBCT images. This method combines deep learning-based centroid extraction, point registration algorithms, and the ICP algorithm to automatically register oral CBCT images of the same patient. It boasts advantages such as simple operation, fast registration speed, and high accuracy, providing accurate data for postoperative evaluation of dental implant surgery and demonstrating broad application prospects and value. Compared with existing manual point selection registration algorithms, this invention offers both ease of operation and higher registration accuracy; compared with the ICP algorithm, the proposed method has a higher registration success rate.

[0122] Although the invention has been described with reference to preferred embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, the technical features mentioned in the various embodiments can be combined in any manner, provided there is no structural conflict. The invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. An automatic registration method for oral CBCT, characterized in that, Includes the following steps: S101, Generating thermal labels for tooth centroids based on oral CBCT images; S102, training a tooth centroid detection model; S103 performs tooth centroid detection and post-processing on CBCT images before and after the operation to obtain the three-dimensional coordinates of the centroid point of each tooth. S104, Clustering algorithm is used to cluster the centroid coordinates of the teeth to separate the upper and lower jaws; The centroids are clustered into two clusters, representing the set of centroids of the maxillary teeth and the set of centroids of the mandibular teeth, respectively. S105. Based on the planned implant location, select the corresponding centroid set. If the planned implant is in the maxilla, sort the centroid set of the mandible; if the planned implant is in the mandible, sort the centroid set of the maxilla, thus obtaining two ordered centroid sets from CBCT data. S106, using the point registration algorithm, calculate the two centroid sets P. B To P A The transformation matrix T1; S107, perform CBCT point cloud registration to obtain the preoperative and postoperative oral CBCT registration conversion relationship.

2. The method for automatic registration of oral CBCT according to claim 1, characterized in that, The thermal mapping of tooth centroids based on oral CBCT images described in S101 includes: Based on the oral CBCT images, three orthogonal slices are displayed, including transverse, sagittal, and coronal planes; For each tooth, delineate the foreground and background areas on three orthogonal slices; Based on the initially outlined foreground and background regions, the teeth segmentation result is obtained using a graph cut algorithm. The center point of each tooth is calculated based on its segmentation results, thus obtaining the centroid of the tooth. Gaussian smoothing is applied to the centroid image to obtain the heat map label of the tooth centroid.

3. The method for automatic registration of oral CBCT according to claim 2, characterized in that, The smoothing radius of the Gaussian smoothing applied to the centroid image is set to 2 mm.

4. The method for automatic registration of oral CBCT according to claim 2, characterized in that, The thermal image label of the tooth centroid is a three-dimensional image with a numerical range of [0, 1], where the value at the centroid point is 1, and the value of the voxel points farther away from the centroid is smaller.

5. The method for automatic registration of oral CBCT according to claim 1, characterized in that, The training of the tooth centroid detection model described in S102 includes: A deep learning model structure is used to train the input CBCT image and the corresponding centroid heatmap label until the loss function converges, thus obtaining the parameters of the neural network model used for centroid detection.

6. The method for automatic registration of oral CBCT according to claim 5, characterized in that, The deep learning model used is the UNet model, and the loss function is the mean squared error loss (MSELoss) function.

7. The method for automatic registration of oral CBCT according to claim 1, characterized in that, S103 describes the process of detecting and post-processing the centroids of teeth in pre- and post-operative CBCT images to obtain the three-dimensional coordinates of the centroid of each tooth, including: Using the tooth centroid detection model, centroid detection was performed on CBCT images before and after the operation, and the results of the centroid detection were obtained as heat maps. The heatmap is binarized with a threshold of 0.5; Connectivity analysis is performed on the binarized image, where each connected component represents a tooth; Calculate the center point of each connected domain to obtain the centroid coordinates of each tooth in the LPS coordinate system.

8. The method for automatic registration of oral CBCT according to claim 1, characterized in that, S104 describes using a clustering algorithm to cluster the centroid coordinates of the teeth to separate the upper and lower jaws, including: Using the two points with the smallest and largest Z-axis coordinates among the centroids as the initial cluster centers, calculate the distance in the Z direction from each centroid to the two cluster centers; Each centroid is assigned a cluster based on the distance, with the number of clusters set to 2, namely the maxilla and mandible. The cluster to which the cluster center with the smaller distance belongs is the cluster to which the centroid belongs. After each centroid is assigned a cluster, the center points of the two clusters are calculated, and the two center points are used as the new cluster centers. The distance from each centroid to the cluster center is calculated again and the cluster is assigned. Repeat the above process until the cluster center no longer changes. At this point, the centroids are clustered into two clusters, representing the centroid set of the maxillary teeth and the centroid set of the mandibular teeth, respectively.

9. The method for automatic registration of oral CBCT according to claim 8, characterized in that, The clustering algorithm used is the K-Means algorithm.

10. The method for automatic registration of oral CBCT according to claim 1, characterized in that, As described in S105, the corresponding set of centroids is selected based on the planned implant location for the surgery. If the planned implant is in the maxilla, the set of centroids in the mandible is sorted. If the planned implant is in the mandible, the centroid set of the maxilla is sorted to obtain an ordered set of centroids from two CBCT data sets, including: Based on the planned implant location, select the corresponding set of centroids. If the planned implant is in the maxilla, sort the set of centroids in the mandible; if the planned implant is in the mandible, sort the set of centroids in the maxilla. The coordinate range of the oral CBCT image in the LPS coordinate system is denoted as [xmin, ymin, zmin] to [xmax, ymax, zmax]. A corner point is determined based on the image range. There are four points that can be used as corner points: [xmin, ymax, zmin], [xmin, ymax, zmax], [xmax, ymax, zmin], and [xmax, ymax, zmax]. Select one of the corner points as the target corner point, find the centroid point closest to the target corner point as the first point, search the set of centroid points for the centroid point closest to the target corner point as the next point, and remove the found point from the set; Repeat the above process until the set is empty, resulting in two ordered centroid sets of CBCT data, denoted as set P. A and set P B At this point, the centroids before and after the operation correspond one-to-one.

11. The method for automatic registration of oral CBCT according to claim 10, characterized in that, The point registration algorithm described in S106 is used to calculate two centroid sets P. B To P A The transformation matrix T1 includes: Using the point registration algorithm, calculate the two centroid sets P. B To P A The transformation matrix T1 is shown in the following equation: 。 12. The method for automatic registration of oral CBCT according to claim 11, characterized in that, S107 describes the CBCT point cloud registration process to obtain the preoperative and postoperative oral CBCT registration conversion relationship, including: The preoperative CBCT model is denoted as A, and the postoperative CBCT model is denoted as B; The postoperative CBCT model B is moved to a new position using transformation matrix T1. This serves as the result of coarse registration; For A and Two sets of point cloud data were obtained by isosurface extraction. and The following relationship is satisfied: ； ； Point cloud calculation using the ICP algorithm To Point Cloud Conversion relationship It satisfies the following transformation relationship: ； Finally, the transformation matrix from postoperative CBCT model B to preoperative CBCT model A can be obtained. The following relationship is satisfied: 。

Images