Automatic registration method and device based on CBCT data and point cloud data, surgical equipment and storage medium

Abstract

The invention relates to the technical field of image processing, and discloses an automatic registration method and device based on CBCT data and point cloud data, surgical equipment and a storage medium, and the method comprises the steps: obtaining to-be-registered CBCT data and to-be-registered point cloud data; determining a first centroid and a first principal axis of the to-be-registered CBCT data, and determining a second centroid and a second principal axis of the to-be-registered point cloud data; and performing translation alignment registration on the to-be-registered CBCT data and the to-be-registered point cloud data based on the first centroid and the second centroid, and performing rotation alignment registration on the to-be-registered CBCT data and the to-be-registered point cloud data based on the first main axis and the second main axis to obtain a target registration result. Compared with the prior art that registration needs to be completed manually by the user, registration can be automatically achieved, and the registration efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a method, apparatus, surgical equipment and storage medium for automatic registration of CBCT data and point cloud data. Background Technology

[0002] In modern medicine, especially in neurosurgery, precise surgical navigation is crucial. With the development of technology, cone-beam computed tomography (CBCT) has been widely used for surgical navigation due to its ability to provide high-resolution three-dimensional images.

[0003] To achieve surgical navigation, it is necessary to register the patient's CBCT data with point cloud data obtained from scanning the patient's body surface, and then perform surgical navigation based on the registration results. However, existing registration methods generally require users to manually set marker points to register the CBCT data and point cloud data, which is inefficient. Summary of the Invention

[0004] The main objective of this application is to provide an automatic registration method, device, surgical equipment, and storage medium based on CBCT data and point cloud data, aiming to solve the technical problem that the registration of CBCT data and point cloud data requires manual operation by the user, resulting in low registration efficiency.

[0005] To achieve the above objectives, this application provides an automatic registration method based on CBCT data and point cloud data, the method comprising: Acquire the CBCT data to be registered and the point cloud data to be registered; Determine the first centroid and first principal axis of the CBCT data to be registered, and determine the second centroid and second principal axis of the point cloud data to be registered; Based on the first centroid and the second centroid, the CBCT data to be registered and the point cloud data to be registered are translated and aligned, and based on the first principal axis and the second principal axis, the CBCT data to be registered and the point cloud data to be registered are rotated and aligned to obtain the target registration result.

[0006] In one embodiment, the step of obtaining the target registration result includes: The results obtained after translational alignment and registration and rotational alignment and registration are used as the initial registration results, and the initial transformation matrix is ​​obtained based on the initial registration results; Based on the initial registration result, determine the nearest point of the point cloud data to be registered in the CBCT data to be registered, and construct the nearest point pair based on the nearest point; The sum of distances is determined based on the nearest point pairs, the sum of distances is minimized, and the initial transformation matrix is ​​updated iteratively based on the minimization result; When the update iteration is completed, the initial registration result is adjusted using the updated initial transformation matrix to obtain the target registration result.

[0007] In one embodiment, the step of determining the total distance based on the nearest point pair includes: Determine the current curvature value of the nearest point pair; Based on the current curvature value, query the preset weight setting mapping relationship table to determine the target weight corresponding to the nearest point pair; The total distance is determined based on the target weight and the nearest point pair.

[0008] In one embodiment, before the step of determining the nearest point in the CBCT data to be registered based on the initial registration result, the method further includes: The point cloud data to be registered is downsampled to obtain downsampled point cloud data at different scales, and the downsampled point cloud data are sorted. The step of determining the nearest point of the point cloud data to be registered in the CBCT data to be registered based on the initial registration result includes: The target downsampled point cloud data is determined according to the sorting results, and the nearest point of the target downsampled point cloud data in the CBCT data to be registered is determined according to the initial registration results; The step of adjusting the initial registration result using the updated initial transformation matrix to obtain the target registration result upon completion of the update iteration includes: When the update iteration is completed, it is determined whether the target downsampled point cloud data has reached the highest scale; If the target downsampled point cloud data reaches the highest scale, the initial registration result is adjusted using the updated and iterated initial transformation matrix to obtain the target registration result.

[0009] In one embodiment, after the step of determining whether the target downsampled point cloud data has reached the highest scale, the method further includes: If the target downsampled point cloud data does not reach the highest scale, the target downsampled point cloud data is updated according to the sorting result; Return to the step of determining the nearest point of the target downsampled point cloud data in the CBCT data to be registered based on the initial registration result.

[0010] In one embodiment, the step of minimizing the sum of distances and updating the initial transformation matrix based on the minimization result includes: Obtain the deformation model corresponding to the point cloud data to be registered, and construct a joint objective function based on the total distance and the deformation model; The joint objective function is solved to obtain the optimal deformation field parameters; The deformation of the point cloud data to be registered is adjusted according to the optimal deformation field parameters, and the current registration error of the point cloud data to be registered after deformation adjustment is determined. The optimal deformation field parameters are updated and iterated using the deformation-adjusted point cloud data to be registered. The step of adjusting the initial registration result using the updated initial transformation matrix to obtain the target registration result upon completion of the update iteration includes: If the current registration error meets the preset error condition, the initial registration result is adjusted using the updated and iterated optimal deformation field parameters to obtain the target registration result.

[0011] In one embodiment, after the step of acquiring the CBCT data to be registered and the point cloud data to be registered, the method further includes: Thin-plate spline interpolation is performed on the point cloud data to be registered to obtain the deformation model corresponding to the point cloud data to be registered.

[0012] In one embodiment, the step of adjusting the initial registration result using the updated and iterated optimal deformation field to obtain the target registration result includes: The initial registration result is adjusted using the updated optimal deformation field to obtain an accurate registration result; Based on the accurate registration results, determine the registration error of the marker points and the target registration error; If the difference between the registration error of the marker point and the target registration error meets the preset difference error, it is determined whether the target registration error is higher than the preset error threshold. If not, the precise registration result shall be used as the target registration result.

[0013] In one embodiment, the step of determining the first centroid and the first principal axis of the CBCT data to be registered includes: Anisotropic diffusion filtering is performed on the CBCT data to be registered to obtain filtered CBCT data. Determine the first centroid and the first principal axis of the filtered CBCT data.

[0014] In one embodiment, the step of determining the first centroid and the first principal axis of the filtered CBCT data includes: Determine whether there is a region containing a metal implant in the filtered CBCT data; If so, then morphological closing operation is performed on the filtered CBCT data of the metal implant region to repair it, and repaired CBCT data is obtained. The first centroid and the first principal axis of the repaired CBCT data are determined.

[0015] In one embodiment, the step of determining the second centroid and the second principal axis of the point cloud data to be registered includes: Extract target feature points from the point cloud data to be registered, the target feature points including SIFT corner points and / or curvature extrema points; The second centroid and second principal axis of the point cloud data to be registered are determined based on the target feature points.

[0016] In one embodiment, the step of extracting target feature points from the point cloud data to be registered includes: Extract the initial feature points from the point cloud data to be registered; The contour line is determined based on the registration point data, and the deformation value corresponding to the initial feature point is determined based on the contour line. The initial feature points are filtered based on the deformation values ​​to obtain the target feature points.

[0017] Furthermore, to achieve the above objectives, this application also proposes an automatic registration device based on CBCT data and point cloud data, the device comprising: The data acquisition module is used to acquire the CBCT data to be registered and the point cloud data to be registered; The parameter determination module is used to determine the first centroid and the first principal axis of the CBCT data to be registered, and to determine the second centroid and the second principal axis of the point cloud data to be registered. The alignment and registration module is used to perform translational alignment and registration of the CBCT data to be registered and the point cloud data to be registered based on the first centroid and the second centroid, and to perform rotational alignment and registration of the CBCT data to be registered and the point cloud data to be registered based on the first principal axis and the second principal axis, so as to obtain the target registration result.

[0018] In one embodiment, the alignment and registration module is further configured to: use the results obtained after translational alignment and registration and rotational alignment and registration as the initial registration result, and obtain an initial transformation matrix based on the initial registration result; determine the nearest point of the point cloud data to be registered in the CBCT data to be registered according to the initial registration result, and construct a nearest point pair according to the nearest point; determine the total distance based on the nearest point pair, minimize the total distance, and update and iterate the initial transformation matrix based on the minimization result; when the update and iteration are completed, adjust the initial registration result using the updated and iterated initial transformation matrix to obtain the target registration result.

[0019] In one embodiment, the alignment and registration module is further configured to determine the current curvature value of the nearest point pair; query a preset weight setting mapping relationship table according to the current curvature value to determine the target weight corresponding to the nearest point pair; and determine the total distance based on the target weight and the nearest point pair.

[0020] In one embodiment, the alignment and registration module is further configured to downsample the point cloud data to be registered to obtain downsampled point cloud data at different scales, and sort the downsampled point cloud data. The alignment and registration module is further configured to determine the target downsampled point cloud data according to the sorting result, and to determine the nearest point of the target downsampled point cloud data in the CBCT data to be registered according to the initial registration result; The alignment and registration module is further configured to determine whether the target downsampled point cloud data has reached the highest scale when the update iteration is completed; if the target downsampled point cloud data has reached the highest scale, the initial registration result is adjusted using the updated initial transformation matrix to obtain the target registration result.

[0021] In one embodiment, the alignment and registration module is further configured to update the target downsampled point cloud data according to the sorting result if the target downsampled point cloud data does not reach the highest scale; and return to perform the operation of determining the nearest point of the target downsampled point cloud data in the CBCT data to be registered based on the initial registration result.

[0022] In one embodiment, the alignment and registration module is further configured to: acquire the deformation model corresponding to the point cloud data to be registered; construct a joint objective function based on the total distance and the deformation model; solve the joint objective function to obtain optimal deformation field parameters; adjust the deformation of the point cloud data to be registered according to the optimal deformation field parameters; determine the current registration error of the adjusted point cloud data; and update and iterate the optimal deformation field parameters using the adjusted point cloud data. The alignment and registration module is further configured to adjust the initial registration result using the updated and iterated optimal deformation field parameters, when the current registration error meets the preset error conditions, to obtain the target registration result.

[0023] In one embodiment, the data acquisition module is further configured to perform thin-plate spline interpolation on the point cloud data to be registered to obtain the deformation model corresponding to the point cloud data to be registered.

[0024] In one embodiment, the alignment and registration module is further configured to adjust the initial registration result using the updated and iterated optimal deformation field to obtain an accurate registration result; determine the marker point registration error and the target registration error based on the accurate registration result; if the difference between the marker point registration error and the target registration error meets a preset difference error, determine whether the target registration error is higher than a preset error threshold; if not, then use the accurate registration result as the target registration result.

[0025] In one embodiment, the parameter determination module is further configured to perform anisotropic diffusion filtering on the CBCT data to be registered to obtain filtered CBCT data; and determine the first centroid and the first principal axis of the filtered CBCT data.

[0026] In one embodiment, the parameter determination module is further configured to determine whether there is a metal implant region in the filtered CBCT data; if so, to perform morphological closing operation repair on the filtered CBCT data of the metal implant region to obtain repaired CBCT data; and to determine the first centroid and the first principal axis of the repaired CBCT data.

[0027] In one embodiment, the parameter determination module is further configured to extract target feature points in the point cloud data to be registered, the target feature points including SIFT corner points and / or curvature extrema points; and determine the second centroid and the second principal axis of the point cloud data to be registered based on the target feature points.

[0028] In one embodiment, the parameter determination module is further configured to extract initial feature points from the point cloud data to be registered; determine a contour line based on the point cloud data to be registered, and determine the deformation value corresponding to the initial feature points based on the contour line; and filter the initial feature points based on the deformation value to obtain target feature points.

[0029] In addition, to achieve the above objectives, this application also proposes a storage medium storing an automatic registration program based on CBCT data and point cloud data, wherein when the automatic registration program based on CBCT data and point cloud data is executed by a processor, it implements the steps of the automatic registration method based on CBCT data and point cloud data as described above.

[0030] Furthermore, to achieve the above objectives, this application also proposes a surgical device, which includes: a memory, a processor, and an automatic registration program based on CBCT data and point cloud data stored in the memory and executable on the processor. When the automatic registration program based on CBCT data and point cloud data is executed by the processor, it implements the steps of the automatic registration method based on CBCT data and point cloud data as described above.

[0031] This application provides an automatic registration method, apparatus, surgical device, and storage medium based on CBCT data and point cloud data. The method includes: acquiring CBCT data to be registered and point cloud data to be registered; determining a first centroid and a first principal axis of the CBCT data to be registered, and determining a second centroid and a second principal axis of the point cloud data to be registered; performing translational alignment registration on the CBCT data to be registered and the point cloud data to be registered based on the first centroid and the second centroid, and performing rotational alignment registration on the CBCT data to be registered and the point cloud data to be registered based on the first principal axis and the second principal axis to obtain a target registration result.

[0032] This application can first determine the first centroid and first principal axis of the CBCT data to be registered, and determine the second centroid and second principal axis of the point cloud data to be registered. Since the centroid represents the geometric center and the principal axis represents the main direction, the CBCT data to be registered and the point cloud data to be registered can be translated and aligned using the first and second centroids, and rotated and aligned using the first and second principal axes, thus obtaining the target registration result. Therefore, compared to existing methods that require manual registration by the user, this application can automatically perform registration, improving registration efficiency. Attached Figure Description

[0033] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram of the surgical equipment structure of the hardware operating environment involved in the embodiments of this application; Figure 2This is a flowchart illustrating the first embodiment of the automatic registration method based on CBCT data and point cloud data in this application; Figure 3 This is a flowchart illustrating the second embodiment of the automatic registration method based on CBCT data and point cloud data in this application; Figure 4 This is a flowchart illustrating the third embodiment of the automatic registration method based on CBCT data and point cloud data in this application; Figure 5 This is a structural block diagram of the first embodiment of the automatic registration device based on CBCT data and point cloud data in this application.

[0036] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0037] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0038] Reference Figure 1 , Figure 1 This is a schematic diagram of the surgical equipment structure of the hardware operating environment involved in the embodiments of this application.

[0039] like Figure 1 As shown, the surgical device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include an interface for connecting to a display screen; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. In this application, the wired interface of the user interface 1003 may be a USB interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be high-speed random access memory (RAM) or non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0040] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the surgical device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0041] like Figure 1 As shown, the memory 1005, which is identified as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an automatic registration program based on CBCT data and point cloud data.

[0042] exist Figure 1 In the surgical device shown, the network interface 1004 is mainly used to connect to the backend server and communicate with the backend server; the user interface 1003 is mainly used to connect to the user device; the surgical device calls the automatic registration program based on CBCT data and point cloud data stored in the memory 1005 through the processor 1001, and executes the automatic registration method based on CBCT data and point cloud data provided in the embodiments of this application.

[0043] It is important to note that precise surgical navigation is crucial in modern medicine, especially in neurosurgery. With the advancement of technology, cone-beam computed tomography (CBCT) has been widely used for surgical navigation due to its ability to provide high-resolution three-dimensional images.

[0044] To achieve surgical navigation, it is necessary to register the patient's CBCT data with point cloud data obtained from scanning the patient's body surface, and then perform surgical navigation based on the registration results. However, existing registration methods generally require users to manually set marker points to register the CBCT data and point cloud data, which is inefficient.

[0045] Therefore, to address the aforementioned shortcomings, this embodiment provides an automatic registration method based on CBCT data and point cloud data. This embodiment first determines the first centroid and first principal axis of the CBCT data to be registered, and then determines the second centroid and second principal axis of the point cloud data to be registered. The centroid represents the geometric center, and the principal axis represents the main direction. Therefore, translational alignment registration can be performed using the first and second centroids, and rotational alignment registration can be performed using the first and second principal axes to obtain the target registration result. Thus, compared to existing methods that require manual registration by the user, this embodiment can automatically achieve registration, improving registration efficiency.

[0046] For ease of understanding, the following is combined with Figures 2 to 5 The automatic registration method based on CBCT data and point cloud data provided in the embodiments of this application will be described in detail.

[0047] Reference Figure 2 , Figure 2This is a flowchart illustrating the first embodiment of the automatic registration method based on CBCT data and point cloud data according to this application. The first embodiment of the automatic registration method based on CBCT data and point cloud data according to this application is presented as follows: Figure 2 As shown, in this embodiment, the specific method includes: Step S10: Obtain the CBCT data to be registered and the point cloud data to be registered.

[0048] It is understood that the method of this embodiment can be applied to any device with data processing, program execution, and automatic registration capabilities, such as surgical equipment, and this embodiment does not impose any limitations on this. However, for ease of subsequent understanding, this embodiment uses surgical equipment (hereinafter referred to as the equipment) as the execution subject to describe this embodiment and the following embodiments.

[0049] It should be understood that the aforementioned CBCT data to be registered can be data obtained by scanning the object in three dimensions using a CBCT scanner, and the aforementioned CBCT data to be registered can include a three-dimensional volume image of the object. The aforementioned point cloud data to be registered can be data obtained by scanning the surface of the object using a three-dimensional scanner to obtain a set of surface points, and the aforementioned point cloud data to be registered can be used to describe the shape of the object's surface.

[0050] In this embodiment, the user can scan the object in advance using a CBCT scanner to obtain CBCT data to be registered, and scan the object using a 3D scanner to obtain point cloud data to be registered. Both data are then transmitted to the aforementioned device, so that the aforementioned device can acquire the CBCT data to be registered and the point cloud data to be registered.

[0051] Step S20: Determine the first centroid and the first principal axis of the CBCT data to be registered, and determine the second centroid and the second principal axis of the point cloud data to be registered.

[0052] It should be noted that the first centroid can be the geometric center of the CBCT data to be registered, and the second centroid can be the geometric center of the point cloud data to be registered. In practical use, the first centroid can be obtained by calculating the average coordinates of all feature points in the CBCT data to be registered, and the second centroid can be obtained by calculating the average coordinates of all feature points in the point cloud data to be registered.

[0053] It should also be noted that the first principal axis can be the longest, second longest, and shortest directions in the CBCT data to be registered. Similarly, the second principal axis can be the longest, second longest, and shortest directions in the point cloud data to be registered. In practical applications, the first principal axis can be obtained by performing principal component analysis on the CBCT data to be registered, and the second principal axis can be obtained by performing principal component analysis on the point cloud data to be registered.

[0054] Step S30: Perform translational alignment registration on the CBCT data to be registered and the point cloud data to be registered based on the first centroid and the second centroid, and perform rotational alignment registration on the CBCT data to be registered and the point cloud data to be registered based on the first principal axis and the second principal axis to obtain the target registration result.

[0055] In this embodiment, since the centroid is the geometric center, after obtaining the first and second centroids, alignment and registration can be performed by translating the first centroid of the CBCT data to be registered with the second centroid of the point cloud data to be registered. This ensures that the first and second centroids coincide, eliminating translational differences and placing them in the same position in space. Furthermore, since the principal axis represents direction, after obtaining the first and second principal axes, alignment and registration can be performed by rotating the first principal axis of the CBCT data to be registered with the second principal axis of the point cloud data to be registered. This ensures that the first and second principal axes are aligned in the same direction, eliminating rotational differences and placing them in the same direction. After the above translational and rotational alignment and registration, automatic registration of the CBCT data and point cloud data to be registered can be achieved, thereby obtaining the target registration result.

[0056] Furthermore, considering that the CBCT data to be registered typically has a high noise level, in order to improve the accuracy of subsequent registration, in this embodiment, the steps of determining the first centroid and the first principal axis of the CBCT data to be registered include: Step S21: Perform anisotropic diffusion filtering on the CBCT data to be registered to obtain filtered CBCT data; Step S22: Determine the first centroid and the first principal axis of the filtered CBCT data.

[0057] It should be noted that the anisotropic diffusion filtering described above can be a type of filtering that reduces noise while preserving edge features. Anisotropic diffusion filtering can dynamically adjust the filtering strength based on the local gradient information of the data, reducing the filtering strength in regions with large gradients (such as edges) to preserve edges, and increasing the filtering strength in regions with small gradients (such as smooth regions) to reduce noise.

[0058] In this embodiment, after the device obtains the CBCT data to be registered, it can first select a suitable diffusion coefficient function, such as the Perona-Malik function, etc., which is not limited in this embodiment. Next, it determines the gradient of each feature point in the CBCT data to be registered, obtaining gradient data. Then, it calculates the diffusion coefficient using the aforementioned diffusion coefficient function based on the obtained gradient data, obtaining diffusion coefficient data. Finally, it updates the initial CBCT data to be registered using the diffusion coefficient data, and repeats the above process iteratively until the iteration requirements are met. These iteration requirements can be a preset number of iterations, etc., and the filtered CBCT data is output. Then, the first centroid and the first principal axis can be obtained from the filtered CBCT data, removing noise while preserving edge features.

[0059] Furthermore, considering that metal implants may be present in the patient's body during surgery, these implants can be metal objects used in the surgery and installed within the patient's body, such as metal plates or screws; this embodiment does not impose any limitations on this. Since metal implants easily produce artifacts in CBCT data, affecting subsequent calculations, in this embodiment, the step of determining the first centroid and the first principal axis of the filtered CBCT data includes: Step S221: Determine whether there is a region containing a metal implant in the filtered CBCT data.

[0060] It should be noted that the aforementioned metal implant region can be any area within the subject's body where a metal implant exists. After obtaining the filtered CBCT data, the device can first determine whether a metal implant region exists. Specifically, a threshold segmentation method can be used. Due to the high density characteristics of metal, a segmentation threshold is set, and regions in the filtered CBCT data that exceed this threshold are identified as the aforementioned metal implant region.

[0061] Step S222: If so, perform morphological closing operation on the filtered CBCT data of the metal implant region to obtain the repaired CBCT data; Step S223: Determine the first centroid and the first principal axis of the repaired CBCT data.

[0062] Understandably, when there is no metal implant area in the filtered CBCT data, the device can directly determine the first centroid and the first principal axis based on the filtered CBCT data.

[0063] If there is a metal implant region in the filtered CBCT data, morphological closing operation can be performed on the filtered CBCT data to repair the surface fracture or discontinuity caused by artifacts, and the repaired CBCT data can be obtained.

[0064] Specifically, the aforementioned device can first perform connectivity analysis on the metal implant area to obtain connected regions, where each connected region can represent an independent metal implant part; then, a unique label is assigned to each connected region to obtain a corresponding labeled image, in which the pixel value of each metal implant area can correspond one-to-one with the label value of the corresponding unique label; then, the aforementioned device can select a suitable structural element shape, such as a sphere or a cube, preferably a spherical structure, because a spherical structure can better adapt to artifacts in different directions.

[0065] After determining the shape of the structuring element, an appropriate structuring element size can be selected based on the size of the metal implant region and the characteristics of artifacts. Finally, based on the shape and size of the structuring element, a dilation operation is performed on each metal implant region in the marked image. Specifically, the structuring element can be placed around the metal region, expanding its boundary outwards to fill small holes and breaks. After dilation, an erosion operation is performed on the dilated marked image based on the shape and size of the structuring element, shrinking the structuring element inwards from the boundary of the metal region. This removes any excess material introduced by the dilation operation, making the shape of the metal region smoother and more natural. This ultimately yields the repaired CBCT data. Finally, the first centroid and first principal axis can be obtained from the repaired CBCT data.

[0066] Furthermore, to improve the robustness of the point cloud data to be registered, the step of determining the second centroid and the second principal axis of the point cloud data to be registered includes: Step S23: Extract target feature points from the point cloud data to be registered, the target feature points including SIFT corner points and / or curvature extrema points; Step S24: Determine the second centroid and second principal axis of the point cloud data to be registered based on the target feature points.

[0067] It should be noted that the aforementioned target feature points can be feature points reflecting the geometric structure in the point cloud data to be registered. Since SIFT corner points and / or curvature extrema points have strong robustness, SIFT corner points and / or curvature extrema points can be used as the aforementioned target feature points in this embodiment.

[0068] It should also be noted that the aforementioned SIFT corner points can be points in the point cloud data to be registered that have undergone Scale-Invariant Feature Transform (SIFT), and are generally located at edges or corners. The aforementioned curvature extrema points can be points in the point cloud data to be registered where the curvature value reaches a local maximum or local minimum, and are also generally located at edges or corners.

[0069] It should be emphasized that in this embodiment, only SIFT corner points can be used as target feature points for subsequent use, or only curvature extrema points can be used as target feature points for subsequent use, or both can be used.

[0070] Specifically, when SIFT corner points are used, the above-mentioned device can calculate the gradient magnitude and gradient direction of each feature point in the point cloud data to be registered, and determine the local extreme points based on the gradient direction. Then, based on the gradient magnitude, the local extreme points are filtered by setting a threshold or non-maximum suppression, thereby using the obtained SIFT corner points as the target feature points.

[0071] When using curvature extrema, the above-mentioned device can first calculate the curvature of each feature point in the point cloud data to be registered. This curvature can reflect the degree of curvature of the point cloud surface. Then, it can find local extrema in each curvature value. Finally, it can filter the local extrema by setting a threshold or non-maximum suppression, so that the obtained curvature extrema are used as the target feature points.

[0072] Finally, the aforementioned device can determine the second centroid and second principal axis of the point cloud data to be registered based on these target feature points.

[0073] Furthermore, to avoid excessive distortion, in this embodiment, the step of extracting target feature points from the point cloud data to be registered includes: Step S231: Extract the initial feature points from the point cloud data to be registered.

[0074] It should be noted that the initial feature points mentioned above can be SIFT corner points and / or curvature extrema points in the point cloud data to be registered.

[0075] In practical use, the above-mentioned device can obtain the SIFT corner points and / or curvature extreme points of the point cloud data to be registered as the above-mentioned initial feature points through the above-mentioned process.

[0076] Step S232: Determine the contour line based on the registration point data, and determine the deformation value corresponding to the initial feature point based on the contour line; Step S233: Filter the initial feature points based on the deformation values ​​to obtain target feature points.

[0077] Understandably, the aforementioned contour line can be the surface contour line of the scanned area of ​​the object. The aforementioned deformation value can be the distance variation value of the feature.

[0078] In practical use, the aforementioned device can calculate the corresponding normal vector for each initial feature point, which reflects the local orientation of the point cloud surface. Then, it can analyze each normal vector and select points with significant variations as boundary points. Connecting these boundary points yields the aforementioned contour line. After obtaining the contour line, the distance from each initial feature point to the nearest contour line is calculated, serving as the aforementioned deformation value. Once the deformation value is obtained, a preset deformation threshold (e.g., 2mm) can be set. The device can then retain initial feature points with deformation values ​​less than or equal to 2mm as target feature points, thereby filtering out feature points with excessive deformation and preventing excessive distortion.

[0079] This embodiment first determines the first centroid and first principal axis of the CBCT data to be registered, and then determines the second centroid and second principal axis of the point cloud data to be registered. The centroid represents the geometric center, and the principal axis represents the main direction. Therefore, translational alignment registration can be performed between the CBCT data to be registered and the point cloud data to be registered using the first and second centroids, and rotational alignment registration can be performed using the first and second principal axes to obtain the target registration result. Therefore, compared to existing methods that require manual registration by the user, this embodiment can automatically perform registration, improving registration efficiency.

[0080] Reference Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the automatic registration method based on CBCT data and point cloud data in this application. Based on the first embodiment described above, the second embodiment of the automatic registration method based on CBCT data and point cloud data in this application is proposed.

[0081] Considering that registration via the centroid and spindle may not be accurate enough (e.g., error tolerance of ±0.5mm), therefore in this embodiment, as... Figure 3 As shown, the steps for obtaining the target registration result include: Step S31: Take the result obtained after translational alignment and registration and rotational alignment and registration as the initial registration result, and obtain the initial transformation matrix based on the initial registration result.

[0082] It should be noted that the initial registration result mentioned above can be the preliminary registration result obtained after translational alignment registration and rotational alignment registration. The initial transformation matrix mentioned above can be a transformation matrix used to register the point cloud data to be registered to be consistent with the CBCT data to be registered. For ease of subsequent explanation, this embodiment denotes the initial transformation matrix as follows: .

[0083] In practical use, after obtaining the initial registration result, this initial registration result can be the matrix required to register the point cloud data to the CBCT data. Therefore, the matrix corresponding to this initial registration result can be used as the aforementioned initial transformation matrix. .

[0084] Step S32: Determine the nearest point of the point cloud data to be registered in the CBCT data to be registered based on the initial registration result, and construct the nearest point pair based on the nearest point.

[0085] Understandably, the aforementioned nearest point can be the closest point between the target feature point in the point cloud data to be registered and the point in the CBCT data to be registered after registration. In practical use, the aforementioned device will consider each target feature point in the registered point cloud data. The corresponding nearest point can be found in the CBCT data to be registered. This can be achieved by calculating the Euclidean distance. This is done after obtaining the nearest point. Then, for each target feature point Construct the corresponding nearest point pair ( , ).

[0086] Step S33: Determine the total distance based on the nearest point pair, minimize the total distance, and update the initial transformation matrix based on the minimization result.

[0087] It should be understood that the sum of the aforementioned distances can be the sum of the Euclidean distances between all nearest point pairs. After obtaining all nearest point pairs, the device can sum the distances between all nearest point pairs to obtain the total distance. After obtaining the total distance, the device can adjust the initial transformation matrix using optimization methods (such as least squares). To minimize the total distance obtained, and when the total distance is minimized, the adjusted initial transformation matrix can be... As the initial transformation matrix after the update iteration.

[0088] Step S34: When the update iteration is completed, the initial registration result is adjusted using the updated initial transformation matrix to obtain the target registration result.

[0089] It should be noted that during the update iteration, the aforementioned device can determine whether the preset iteration conditions are met, thereby determining whether the update iteration is complete. In this embodiment, the preset iteration conditions may be that the total distance is less than a certain preset distance threshold, or that the number of iterations reaches a certain preset number threshold, etc., and this embodiment does not impose any restrictions on these conditions.

[0090] It should also be noted that if the aforementioned preset iteration conditions are not met, the device can update the initial transformation matrix after iteration. The data is then fed into the point cloud data to be registered, resulting in a new initial registration result. The process is then returned to step S32 above.

[0091] When the aforementioned preset iteration conditions are met, the device can determine that the update iteration is complete, and then utilize the final obtained updated initial transformation matrix. The initial registration result is adjusted and then applied to the point cloud data to be registered to obtain the final target registration result.

[0092] Furthermore, considering that different locations may require different registration accuracies, to enhance robustness, in this embodiment, the step of determining the total distance based on the nearest point pair includes: Step S331: Determine the current curvature value of the nearest point pair.

[0093] It should be noted that the aforementioned current curvature value can be the degree of curvature in geometry between the nearest point pair. In practical use, after the device determines the nearest point pair, it can calculate the curvature to obtain the current curvature value corresponding to the nearest point pair.

[0094] Step S332: Query the preset weight setting mapping relationship table according to the current curvature value to determine the target weight corresponding to the nearest point pair; Step S333: Determine the total distance based on the target weight and the nearest point pair.

[0095] It is understood that the aforementioned preset weight setting mapping table can be a table storing the mapping relationship between the current curvature value and the corresponding target weight. The specific mapping relationship can be set according to the actual situation; in this embodiment, a larger target weight can be set for a larger current curvature value.

[0096] In practical use, after the aforementioned device obtains the current curvature value, it can query the target weight corresponding to each nearest point pair based on the current curvature value. After obtaining each target weight, a weighted summation can be performed when summing the distances. That is, after determining the distance between each nearest point pair, it is multiplied by the corresponding target weight and then summed to obtain the total distance. This introduces a weighting mechanism, which has a greater impact on the registration process with high curvature values, thereby improving the registration accuracy and robustness.

[0097] Furthermore, considering the large amount of data in the point cloud data to be registered, direct registration of the point cloud data may lead to slow convergence. Therefore, to increase the convergence speed, in this embodiment, before the step of determining the nearest point of the point cloud data to be registered in the CBCT data to be registered based on the initial registration result, the method further includes: Step S302: Downsample the point cloud data to be registered to obtain downsampled point cloud data at different scales, and sort the downsampled point cloud data.

[0098] It should be noted that before determining the nearest point, the aforementioned device can first downsample the point cloud data to be registered, thereby obtaining point cloud data of different scales (different resolutions) as the downsampled point cloud data. For example, downsampled point cloud data of three scales can be obtained, corresponding to low resolution, medium resolution, and high resolution, respectively. Of course, it can also correspond to more scales, and this embodiment does not limit this.

[0099] After obtaining downsampled point cloud data at various scales, the downsampled point cloud data can be sorted in order of scale from low to high, which facilitates subsequent iterations from low to high.

[0100] The step of determining the nearest point of the point cloud data to be registered in the CBCT data to be registered based on the initial registration result includes: Step S321: Determine the target downsampled point cloud data according to the sorting result, and determine the nearest point of the target downsampled point cloud data in the CBCT data to be registered according to the initial registration result.

[0101] It is understood that the target downsampled point cloud data mentioned above can be the downsampled point cloud data used in this iteration. In this embodiment, the iteration can be performed in a manner from low to high scale.

[0102] In practical use, after the above devices are sorted, the downsampled point cloud data with the lowest scale can be selected as the target downsampled point cloud data. Based on the initial registration result, the nearest point in the CBCT data to be registered is found for each point in the current target downsampled point cloud data, forming the nearest point pair. The total distance is determined based on the nearest point pair, and the total distance is minimized. The initial transformation matrix is ​​then updated and iterated based on the minimization result.

[0103] The step of adjusting the initial registration result using the updated initial transformation matrix to obtain the target registration result upon completion of the update iteration includes: Step S341: When the update iteration is completed, determine whether the target downsampled point cloud data has reached the highest scale; Step S342: If the target downsampled point cloud data reaches the highest scale, the initial registration result is adjusted using the updated and iterated initial transformation matrix to obtain the target registration result.

[0104] It should be understood that the aforementioned highest scale can be the highest scale obtained during downsampling. When the device determines that the update iteration is complete, it can determine whether the target downsampled point cloud data has reached the highest scale. If the highest scale has been reached, the initial transformation distance obtained after the update iteration can be used to adjust the initial registration result to obtain the target registration result.

[0105] Furthermore, when it is not the highest scale, the step of determining whether the target downsampled point cloud data has reached the highest scale, as described above, further includes: Step S343: If the target downsampled point cloud data does not reach the highest scale, then update the target downsampled point cloud data according to the sorting result; Step S344: Return to the step of determining the nearest point of the target downsampled point cloud data in the CBCT data to be registered based on the initial registration result.

[0106] In practical use, when the aforementioned device determines that the target downsampled point cloud data is not at the highest scale, it can select the next higher-scale downsampled point cloud data as the new target downsampled point cloud data according to the sorting results. It then returns to the step of determining the nearest point of the target downsampled point cloud data in the CBCT data to be registered based on the initial registration results. That is, it redetermines the new nearest point based on the new target downsampled point cloud data and the initial registration results, and continues iterating until the iteration is complete and the highest scale is reached.

[0107] Reference Figure 4 , Figure 4 This is a flowchart illustrating the third embodiment of the automatic registration method based on CBCT data and point cloud data in this application. Based on the above embodiments, the third embodiment of the automatic registration method based on CBCT data and point cloud data in this application is proposed.

[0108] Furthermore, considering that non-rigid tissues (such as brain tissue) may exist in the human body, the deformation of non-rigid tissues differs somewhat from the registration of deformations in rigid tissues. To improve registration accuracy, such as... Figure 4 As shown, in this embodiment, the steps of minimizing the sum of distances and updating the initial transformation matrix based on the minimization result include: Step S334: Obtain the deformation model corresponding to the point cloud data to be registered, and construct a joint objective function based on the total distance and the deformation model.

[0109] It should be noted that the aforementioned deformation model can be a model describing the deformation of non-rigid organization in the point cloud data to be registered. The specific deformation model can be set according to the actual situation, such as a Bayesian deformation model, etc. This embodiment does not impose any restrictions on this. The aforementioned joint objective function can be an objective function composed of the sum of distances and the regularization term of the deformation model, which can be used to optimize the registration process.

[0110] In practical use, after obtaining the total distance, the above-mentioned device can first obtain the deformation model involved in the point cloud data to be registered, and then define the objective function, which may include data items (the total distance between the nearest point pairs) and regularization items (regularization of the deformation model).

[0111] Specifically, in this embodiment, the joint objective function can be... .in Here, N represents the deformation field parameters, and N represents the number of target feature points. Let be the distance between the i-th nearest point pair. The sum of the above distances, This is the regularization term for the deformation field parameters in the deformation model. This is a regularization parameter, which can be set according to the actual situation. This embodiment does not impose any restrictions on this.

[0112] Step S335: Solve the joint objective function to obtain the optimal deformation field parameters.

[0113] Understandably, the aforementioned optimal deformation field parameters can be parameters indicating the deformation required for the point cloud data to be registered to transform into the CBCT data to be registered. In practical use, after the aforementioned device constructs and obtains the joint objective function, a numerical optimization method (e.g., gradient descent) can be used to solve the joint objective function to obtain the optimal deformation field parameters. In this embodiment, the aforementioned optimal deformation field parameters may include weighting coefficients. and control points .

[0114] Step S336: Adjust the deformation of the point cloud data to be registered according to the optimal deformation field parameters, and determine the current registration error of the point cloud data to be registered after deformation adjustment.

[0115] It should be understood that the aforementioned current registration error can be the registration error between the point cloud data to be registered and the CBCT data to be registered after the current deformation adjustment.

[0116] In practical use, once the aforementioned equipment obtains the optimal deformation field parameters, these parameters can be applied to the current point cloud data to be registered, thereby adjusting the deformation of the data. The optimal deformation field parameters can be applied using a deformation field expression, specifically: .in Let x be the target feature point in the point cloud data to be registered after adjustment, and let x be the target feature point in the point cloud data to be registered before adjustment. For the i-th control point, This is the weighting coefficient corresponding to the i-th control point. Used to determine the degree of influence of control points on deformation. These are radial basis functions, which can be used to measure points. With control points The contribution of the relative distance between them to the deformation. For point With control points The relative distance between them.

[0117] Furthermore, after obtaining the optimal deformation field parameters, for each target feature point in the point cloud data to be registered, the optimal deformation field parameters can be substituted into the above deformation field expression to calculate the deformation contribution of all control points. Next, the deformation contribution is used to adjust the target feature points before adjustment, thereby obtaining the deformed point cloud data to be registered. Finally, the current registration error between the deformed point cloud data to be registered and the unadjusted point cloud data is determined.

[0118] Step S337: Update and iterate the optimal deformation field parameters using the deformation-adjusted point cloud data to be registered.

[0119] After deformation adjustment, the optimal deformation field parameters can be updated and iterated using the deformed point cloud data to be registered. Specifically, this can involve determining whether the current registration error of this iteration meets a preset error condition. This preset error condition can be set according to the actual situation, and this embodiment does not impose any restrictions on it.

[0120] If the preset error condition is not met, it indicates that the iteration is not completed, and then the process returns to the above steps of determining the nearest point of the point cloud data to be registered in the CBCT data to be registered based on the initial registration result, that is, determining the nearest point of the new point cloud data to be registered in the CBCT data to be registered, reconstructing the nearest point pair, and recalculating the optimal deformation field parameters.

[0121] The step of adjusting the initial registration result using the updated initial transformation matrix to obtain the target registration result upon completion of the update iteration includes: Step S345: If the current registration error meets the preset error condition, adjust the initial registration result using the updated and iterated optimal deformation field parameters to obtain the target registration result.

[0122] If the current registration error meets the preset registration error, the update iteration can be determined to be complete. Then, the optimal deformation field parameters obtained in this iteration can be used to adjust the point cloud data to be registered in the initial registration result, thereby obtaining the target registration result.

[0123] It should be emphasized that, since the optimal deformation field parameters are replaced with the initial transformation matrix used in the above iteration in this embodiment, it is more suitable for the transformation of non-rigid structures and improves the accuracy of registration.

[0124] Furthermore, in order to obtain the above-mentioned deformation model, in this embodiment, after the step of acquiring the CBCT data to be registered and the point cloud data to be registered, the method further includes: Thin-plate spline interpolation is performed on the point cloud data to be registered to obtain the deformation model corresponding to the point cloud data to be registered.

[0125] It should be noted that the above-mentioned thin plate spline interpolation can be a method for surface fitting and interpolation, which is applicable to handling non-rigid deformation and can generate smooth deformation fields.

[0126] In this embodiment, after the device obtains the point cloud data to be registered, it can select a set of control points in the point cloud data to be registered to define the deformation field, and define a target position for each control point. The target position can be the position that the control point is expected to move to during the registration process. Finally, the thin plate spline interpolation method is used to construct the deformation model based on the original position and target position of the control point, that is, to obtain the above deformation field expression.

[0127] Furthermore, to improve the accuracy of the registration results, in this embodiment, the step of adjusting the initial registration result using the updated and iterated optimal deformation field to obtain the target registration result includes: Step S3451: Adjust the initial registration result using the updated and iterated optimal deformation field to obtain an accurate registration result; Step S3452: Determine the registration error of the marker points and the registration error of the target based on the accurate registration results.

[0128] It should be noted that the above-mentioned precise registration result can be the registration result obtained under the condition that the current registration error meets the preset error conditions. The above-mentioned fiducial registration error (FRE) can be the error between the actual position and the expected position of the fiducial markers used in the registration process. The above-mentioned target registration error (TRE) can be the error between the actual position and the true position of the registered target point (such as key points of an anatomical structure).

[0129] In practical use, if the current registration error meets the preset error conditions, the initial registration result can be adjusted using the latest updated optimal deformation field to obtain an accurate registration result. Then, based on the accurate registration result, the deviation between the registered position of the marker point in the point cloud data to be registered and its corresponding position in the CBCT data to be registered is measured as the aforementioned marker point registration error. Based on the accurate registration result, the deviation between the registered key anatomical structure points in the point cloud data to be registered and the actual positions of these points in the CBCT data to be registered is measured as the aforementioned target registration error.

[0130] Step S3453: If the difference between the registration error of the marker point and the target registration error meets the preset difference error, determine whether the target registration error is higher than the preset error threshold. Step S3454: If not, then the precise registration result is taken as the target registration result.

[0131] It is understood that the aforementioned preset difference error can be used to determine whether the difference between the registration error of the marker point and the registration error of the target point is large, and can be set according to the actual situation. The aforementioned preset error threshold can be used to determine whether the target registration error is too large, and can also be set according to the actual situation. In this embodiment, 0.5mm can be used for illustration.

[0132] In practical use, after the above-mentioned device obtains the FRE and TRE, it can first determine whether the difference between the FRE and TRE meets the preset difference error. If it does not meet the preset difference error, it can be said that the difference between the two is large. Then, it can return to the step of using the result obtained after translational alignment registration and rotational alignment registration as the initial registration result for re-iteration.

[0133] If the condition is met, it indicates that the difference between the two is small, and then it can be determined whether the TRE is greater than 0.5mm. If it is greater than 0.5mm, it indicates that the difference is still large, and then the step of using the result obtained after translational alignment and rotational alignment as the initial registration result can be executed again for re-iteration. If it is less than or equal to 0.5mm, it indicates that the difference is small, and then the obtained accurate registration result can be used as the target registration result.

[0134] In addition, refer to Figure 5 , Figure 5 This is a structural block diagram of the first embodiment of the automatic registration device based on CBCT data and point cloud data in this application; as shown... Figure 5 As shown in the embodiments of this application, an automatic registration device based on CBCT data and point cloud data is also proposed. The device includes: Data acquisition module 501 is used to acquire CBCT data to be registered and point cloud data to be registered; The parameter determination module 502 is used to determine the first centroid and the first principal axis of the CBCT data to be registered, and to determine the second centroid and the second principal axis of the point cloud data to be registered. The alignment and registration module 503 is used to perform translational alignment and registration of the CBCT data to be registered and the point cloud data to be registered based on the first centroid and the second centroid, and to perform rotational alignment and registration of the CBCT data to be registered and the point cloud data to be registered based on the first principal axis and the second principal axis, so as to obtain the target registration result.

[0135] This embodiment first determines the first centroid and first principal axis of the CBCT data to be registered, and then determines the second centroid and second principal axis of the point cloud data to be registered. The centroid represents the geometric center, and the principal axis represents the main direction. Therefore, translational alignment registration can be performed between the CBCT data to be registered and the point cloud data to be registered using the first and second centroids, and rotational alignment registration can be performed using the first and second principal axes to obtain the target registration result. Therefore, compared to existing methods that require manual registration by the user, this embodiment can automatically perform registration, improving registration efficiency.

[0136] In one implementation, the parameter determination module 502 is further configured to perform anisotropic diffusion filtering on the CBCT data to be registered to obtain filtered CBCT data; and determine the first centroid and the first principal axis of the filtered CBCT data.

[0137] In one implementation, the parameter determination module 502 is further used to determine whether there is a metal implant region in the filtered CBCT data; if so, morphological closing operation is performed on the filtered CBCT data of the metal implant region to repair it, thereby obtaining repaired CBCT data; and the first centroid and the first principal axis of the repaired CBCT data are determined.

[0138] In one implementation, the parameter determination module 502 is further configured to extract target feature points from the point cloud data to be registered, the target feature points including SIFT corner points and / or curvature extrema points; and determine the second centroid and the second principal axis of the point cloud data to be registered based on the target feature points.

[0139] In one implementation, the parameter determination module 502 is further configured to extract initial feature points from the point cloud data to be registered; determine a contour line based on the point cloud data to be registered, and determine the deformation value corresponding to the initial feature points based on the contour line; and filter the initial feature points based on the deformation value to obtain target feature points.

[0140] Based on the first embodiment of the automatic registration device based on CBCT data and point cloud data described in this application, a second embodiment of the automatic registration device based on CBCT data and point cloud data is proposed.

[0141] In this embodiment, the alignment and registration module 503 is further configured to take the results obtained after translational alignment and registration and rotational alignment and registration as the initial registration result, and obtain an initial transformation matrix based on the initial registration result; determine the nearest point of the point cloud data to be registered in the CBCT data to be registered according to the initial registration result, and construct a nearest point pair according to the nearest point; determine the total distance based on the nearest point pair, minimize the total distance, and update and iterate the initial transformation matrix based on the minimization result; when the update and iteration are completed, adjust the initial registration result using the updated and iterated initial transformation matrix to obtain the target registration result.

[0142] As one implementation, the alignment and registration module 503 is further configured to determine the current curvature value of the nearest point pair; query a preset weight setting mapping relationship table according to the current curvature value to determine the target weight corresponding to the nearest point pair; and determine the total distance based on the target weight and the nearest point pair.

[0143] As one implementation, the alignment and registration module 503 is further used to downsample the point cloud data to be registered, obtain downsampled point cloud data at different scales, and sort the downsampled point cloud data. The alignment and registration module 503 is further configured to determine the target downsampled point cloud data according to the sorting result, and determine the nearest point of the target downsampled point cloud data in the CBCT data to be registered according to the initial registration result; The alignment and registration module 503 is further configured to determine whether the target downsampled point cloud data has reached the highest scale when the update iteration is completed; if the target downsampled point cloud data has reached the highest scale, the initial registration result is adjusted using the updated initial transformation matrix to obtain the target registration result.

[0144] As one implementation, the alignment and registration module 503 is further configured to update the target downsampled point cloud data according to the sorting result if the target downsampled point cloud data does not reach the highest scale; and return to perform the operation of determining the nearest point of the target downsampled point cloud data in the CBCT data to be registered based on the initial registration result.

[0145] Based on the above embodiments of the automatic registration device based on CBCT data and point cloud data of this application, a third embodiment of the automatic registration device based on CBCT data and point cloud data of this application is proposed.

[0146] In this embodiment, the alignment and registration module 503 is further configured to acquire the deformation model corresponding to the point cloud data to be registered, and construct a joint objective function based on the total distance and the deformation model; solve the joint objective function to obtain the optimal deformation field parameters; adjust the deformation of the point cloud data to be registered according to the optimal deformation field parameters, and determine the current registration error of the adjusted point cloud data; and update and iterate the optimal deformation field parameters using the adjusted point cloud data. The alignment and registration module 503 is further configured to adjust the initial registration result using the updated and iterated optimal deformation field parameters to obtain the target registration result when the current registration error meets the preset error conditions.

[0147] As one implementation, the data acquisition module 501 is also used to perform thin plate spline interpolation on the point cloud data to be registered to obtain the deformation model corresponding to the point cloud data to be registered.

[0148] As one implementation, the alignment and registration module 503 is further configured to adjust the initial registration result using the updated and iterated optimal deformation field to obtain an accurate registration result; determine the marker point registration error and the target registration error based on the accurate registration result; if the difference between the marker point registration error and the target registration error meets a preset difference error, determine whether the target registration error is higher than a preset error threshold; if not, then use the accurate registration result as the target registration result.

[0149] Other embodiments or specific implementations of the automatic registration device based on CBCT data and point cloud data described in this application can be referred to the above-mentioned method embodiments, and will not be repeated here.

[0150] In addition, to achieve the above objectives, this application embodiment also provides a storage medium storing an automatic registration program based on CBCT data and point cloud data. When the automatic registration program based on CBCT data and point cloud data is executed by a processor, it implements the steps of the automatic registration method based on CBCT data and point cloud data as described above.

[0151] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system 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 system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0152] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0153] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as a read-only memory image (ROM) / random access memory (RAM), magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0154] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. An automatic registration method based on CBCT data and point cloud data, characterized in that, The method comprises: obtaining CBCT data to be registered and point cloud data to be registered; determining a first centroid and a first principal axis of the CBCT data to be registered, and determining a second centroid and a second principal axis of the point cloud data to be registered; performing translational alignment registration on the CBCT data to be registered and the point cloud data to be registered based on the first centroid and the second centroid, and performing rotational alignment registration on the CBCT data to be registered and the point cloud data to be registered based on the first principal axis and the second principal axis, to obtain a target registration result.

2. The method of claim 1, wherein, The step of obtaining the target registration result comprises: taking the result obtained after translational alignment registration and rotational alignment registration as an initial registration result, and obtaining an initial transformation matrix based on the initial registration result; determining a nearest point of the point cloud data to be registered in the CBCT data to be registered according to the initial registration result, and constructing a nearest point pair according to the nearest point; determining a distance sum based on the nearest point pair, minimizing the distance sum, and updating the initial transformation matrix based on the minimization result; when the updating iteration is completed, adjusting the initial registration result by using the initial transformation matrix after the updating iteration to obtain the target registration result.

3. The method of claim 2, wherein, The step of determining the distance sum based on the nearest point pair comprises: determining a current curvature value of the nearest point pair; querying a preset weight setting mapping relationship table according to the current curvature value to determine a target weight corresponding to the nearest point pair; determining the distance sum based on the target weight and the nearest point pair.

4. The method of claim 2, wherein, Before the step of determining the nearest point of the point cloud data to be registered in the CBCT data to be registered according to the initial registration result, the method further comprises: down-sampling the point cloud data to be registered to obtain down-sampled point cloud data of different scales, and sorting each down-sampled point cloud data; The step of determining the nearest point of the point cloud data to be registered in the CBCT data to be registered according to the initial registration result comprises: determining a target down-sampled point cloud data according to the sorting result, and determining a nearest point of the target down-sampled point cloud data in the CBCT data to be registered according to the initial registration result; The step of adjusting the initial registration result by using the initial transformation matrix after the updating iteration to obtain the target registration result when the updating iteration is completed comprises: when the updating iteration is completed, judging whether the target down-sampled point cloud data reaches a highest scale; if the target down-sampled point cloud data reaches the highest scale, adjusting the initial registration result by using the initial transformation matrix after the updating iteration to obtain the target registration result.

5. The method of claim 4, wherein, After the step of judging whether the target down-sampled point cloud data reaches the highest scale, the method further comprises: if the target down-sampled point cloud data does not reach the highest scale, updating the target down-sampled point cloud data according to the sorting result; returning to execute the step of determining the nearest point of the target down-sampled point cloud data in the CBCT data to be registered according to the initial registration result.

6. The method of claim 2, wherein, The step of minimizing the distance sum and updating the initial transformation matrix based on the minimization result iteratively comprises: obtaining a morphable model corresponding to the point cloud data to be registered, and constructing a joint objective function based on the distance sum and the morphable model; solving the joint objective function to obtain optimal morphing field parameters; deforming and adjusting the point cloud data to be registered according to the optimal morphing field parameters, and determining a current registration error of the deformed and adjusted point cloud data to be registered; updating the optimal morphing field parameters using the deformed and adjusted point cloud data to be registered; The step of adjusting the initial registration result using the initial transformation matrix updated after the updating iteration to obtain a target registration result when the updating iteration is completed comprises: if the current registration error meets a preset error condition, adjusting the initial registration result using the optimal morphing field parameters updated after the updating iteration to obtain a target registration result.

7. The method of claim 6, wherein, The step of adjusting the initial registration result using the optimal morphing field updated after the updating iteration to obtain a target registration result comprises: adjusting the initial registration result using the optimal morphing field updated after the updating iteration to obtain an accurate registration result; determining a marker point registration error and a target registration error based on the accurate registration result; if the difference between the marker point registration error and the target registration error meets a preset difference error, determining whether the target registration error is higher than a preset error threshold; if not, taking the accurate registration result as a target registration result.

8. The method of claim 1, wherein, The step of determining the first centroid and the first principal axis of the CBCT data to be registered comprises: performing anisotropic diffusion filtering on the CBCT data to be registered to obtain filtered CBCT data; determining the first centroid and the first principal axis of the filtered CBCT data.

9. The method of claim 8, wherein, The step of determining the first centroid and the first principal axis of the filtered CBCT data comprises: determining whether there is a metal implant region in the filtered CBCT data; if yes, performing morphological closing operation repair on the filtered CBCT data of the metal implant region to obtain repaired CBCT data; determining the first centroid and the first principal axis of the repaired CBCT data.

10. The method of claim 1, wherein, The step of determining the second centroid and the second principal axis of the point cloud data to be registered comprises: extracting target feature points in the point cloud data to be registered, the target feature points comprising SIFT corner points and / or curvature extreme points; determining the second centroid and the second principal axis of the point cloud data to be registered based on the target feature points.

11. An apparatus for automatically registering CBCT data with point cloud data, comprising: The device comprises: a data acquisition module configured to acquire CBCT data to be registered and point cloud data to be registered; a parameter determination module configured to determine a first centroid and a first principal axis of the CBCT data to be registered, and determine a second centroid and a second principal axis of the point cloud data to be registered; An alignment registration module is configured to perform translational alignment registration of the CBCT data to be registered and the point cloud data to be registered based on the first centroid and the second centroid, and perform rotational alignment registration of the CBCT data to be registered and the point cloud data to be registered based on the first principal axis and the second principal axis, to obtain a target registration result.

12. A surgical apparatus, characterized by The surgical device comprises a memory, a processor, and a CBCT data and point cloud data automatic registration program stored on the memory and executable on the processor, and the CBCT data and point cloud data automatic registration program, when executed by the processor, implements the steps of the CBCT data and point cloud data automatic registration method according to any one of claims 1 to 10.

13. A storage medium, characterized by The storage medium stores a CBCT data and point cloud data automatic registration program, and the CBCT data and point cloud data automatic registration program, when executed by a processor, implements the steps of the CBCT data and point cloud data automatic registration method according to any one of claims 1 to 10.

Images