Guide plate three-dimensional model generation method, device, terminal equipment and storage medium

Abstract

The present invention application provides a method, device, terminal equipment and storage medium for generating a three-dimensional model of a guide plate, including: acquiring three-dimensional image data of a target area of ​​the maxillary sinus; generating point cloud data of a segmented area, and performing registration processing using an iterative closest point algorithm to obtain three-dimensional point data; fusing the three-dimensional point data of each time point, and adjusting the fusion result through a grid optimization algorithm to obtain a three-dimensional reconstruction model of the segmented area; identifying anatomical landmarks; constructing a local coordinate system; and then generating a three-dimensional model of the guide plate. The present invention application accurately identifies bone, soft tissue and blood vessels; separately registers the point cloud data of each time point to ensure the temporal consistency of the data; adjusts the fusion result through a grid optimization algorithm to make the grid both uniform and consistent with the actual biological tissue structure, thereby ensuring the grid quality of the three-dimensional reconstruction model; the refinement of the three-dimensional model of the guide plate can be improved to match the actual anatomical structure.

Description

Technical Field

[0001] The present invention relates to the field of 3D printing, and in particular to a method, device, terminal equipment and storage medium for generating a three-dimensional model of a guide plate. Background Art

[0002] Problems such as insufficient bone volume in the maxillary posterior teeth area are difficult points in dental implant technology. This situation can be solved by maxillary sinus floor lift surgery, which mainly includes two surgical procedures: the alveolar crest approach (internal lift) and the lateral wall window approach (external lift). There are many surgical methods for external maxillary sinus lift through lateral wall windowing, such as using an ultrasonic bone knife or a ball drill to grind off the lateral wall bone for windowing. These methods cannot preserve the lateral wall bone. At the same time, in actual operation, the position and range of the grinding, as well as the avoidance of important blood vessels in the adjacent area are roughly judged based on the CBCT before windowing, which is not very accurate. In addition, if the windowing range is too large or too small, and the position is too high or too low, it will affect the surgical operation and lifting effect, and greatly affect the treatment prognosis.

[0003] Digital 3D printed guide technology is a new method used in maxillary sinus lift surgery in recent years. It designs the shape and positioning points of the digital guide to effectively avoid important anatomical structures and reduce surgical risks. At present, the intraoral positioning and fixation methods of the guide are divided into mucosa-supported and tooth-supported, as well as detachable guides. These methods have the problem of unstable fixation. These problems are often caused by the poor matching degree between the guide data and the actual anatomical structure of the patient, and the low accuracy of the generated guide data. Summary of the invention

[0004] The present invention application provides a guide plate three-dimensional model generation method, device, terminal equipment and storage medium, which improves the accuracy of the guide plate three-dimensional model.

[0005] In order to solve the above technical problems, the present invention provides a method for generating a three-dimensional model of a guide plate, comprising:

[0006] Acquire three-dimensional image data of a target area of ​​the maxillary sinus, and segment the three-dimensional image data using a deep learning segmentation algorithm to identify regional image information of the target area to obtain a segmented area; the regional image information includes bone area information, soft tissue area information, and blood vessel direction information;

[0007] Generate point cloud data of the segmented area, and use an iterative closest point algorithm to perform registration processing on the point cloud data at each time point to obtain three-dimensional point data;

[0008] Fusing the three-dimensional point data at each time point, adjusting the fusion result by a grid optimization algorithm, and thereby obtaining a three-dimensional reconstruction model of the segmented area;

[0009] Based on the three-dimensional reconstruction model, a plurality of anatomical landmarks are identified by a geometric feature extraction algorithm to obtain information of each anatomical landmark; design reference points of a target guide are selected from the identified anatomical landmarks, and a local coordinate system is constructed using the anatomical landmark information;

[0010] Based on the screened reference points and in combination with the local coordinate system, a parametric modeling method is used to generate an initial contour of the guide plate; based on the initial contour of the guide plate, a surface fitting algorithm is used to refine the contact area between the target guide plate and the bone surface to obtain the target contour of the guide plate, and then the three-dimensional model of the guide plate is generated based on the target contour of the guide plate.

[0011] As a preferred solution, the point cloud data at each time point is registered using an iterative closest point algorithm to obtain three-dimensional point data, including:

[0012] Generate a corresponding time series according to the acquisition timestamp of each point in the point cloud data;

[0013] According to the time sequence, the point cloud data is loaded into a preset coordinate system to obtain resolution and density information of the point cloud data;

[0014] Performing denoising processing on the point cloud data according to the resolution and density information;

[0015] The overlapping area information of the denoised point cloud data is obtained, and the number of iterations and the convergence threshold of the iterative closest point algorithm are determined; a preset initial transformation matrix is ​​used, and according to the number of iterations and the convergence threshold, the point cloud data at each time point are respectively registered by the iterative closest point algorithm to obtain the three-dimensional point data.

[0016] As a preferred solution, the fusion result is adjusted by a grid optimization algorithm to obtain a three-dimensional reconstruction model of the segmented area, including:

[0017] Generate an initial grid according to the three-dimensional coordinate information of the fusion result;

[0018] Determining the edge topological relationship of the initial grid based on a preset topological data structure and the three-dimensional coordinate information;

[0019] Determining the connectivity of the initial mesh by using the edge topological relationship, thereby obtaining adjacency information of each facet in the initial mesh;

[0020] Based on the adjacency information, a geometric algorithm is used to optimize and adjust the angle and side length of each facet to obtain an optimized mesh;

[0021] Calculating the patch density and patch distribution in the optimized grid; determining the energy function in the Markov random field under the energy minimization condition constraint according to the patch density and patch distribution;

[0022] The optimized grid is iteratively optimized through the energy function to obtain a three-dimensional reconstruction model of the segmented area.

[0023] As a preferred solution, based on the three-dimensional reconstruction model, a plurality of anatomical landmarks are identified by a geometric feature extraction algorithm, specifically:

[0024] Based on the three-dimensional reconstruction model, initial landmarks are obtained by using a geometric feature extraction algorithm;

[0025] Determine the position information of each initial landmark point through edge detection and curvature analysis;

[0026] According to the position information of the initial landmark points, the morphological features of each initial landmark point are identified in combination with the relative positions of each initial landmark point and the corresponding adjacent anatomical structure;

[0027] The initial landmark points that meet the preset characteristics are used as the anatomical landmark points.

[0028] As a preferred solution, the initial contour of the guide plate is generated by using a parametric modeling method based on the selected reference points and in combination with the local coordinate system, including:

[0029] Acquire preset size parameters, shape parameters and inclination angle of the target guide plate, wherein the size parameters include the length, width, thickness and curvature radius of the guide plate;

[0030] The end point coordinates and the center point coordinates of the target guide plate are determined according to the size parameters and the reference point, and then the initial contour of the guide plate is generated in the local coordinate system according to the shape parameters and the inclination angle.

[0031] As a preferred solution, generating the initial contour of the guide plate according to the shape parameters and the inclination angle includes:

[0032] Acquire the initial bone surface model of the target area by three-dimensional scanning technology; use the bone area information to refine the initial bone surface model to obtain a three-dimensional bone surface model;

[0033] Obtaining bone surface curvature data based on the bone surface three-dimensional model, and fitting the bone surface curvature data using a least squares method to obtain a fitting result;

[0034] Based on the fitting results and in combination with preset channel modeling data, the starting point, end point and trajectory of the target instrument channel are located;

[0035] A plurality of marking points are generated according to the starting point, the end point and the trajectory of the target instrument channel; the initial contour of the guide plate is generated by combining the marking points with the shape parameters and the tilt angle.

[0036] As a preferred solution, the generating the guide plate three-dimensional model based on the guide plate target contour includes:

[0037] Modeling the geometric structure parameters and material properties of the guide plate target profile by finite element method to create an initial three-dimensional simulation model;

[0038] Applying loads and boundary conditions on the basis of the initial three-dimensional simulation model, and obtaining a stress distribution cloud map through finite element simulation;

[0039] According to the stress distribution cloud map combined with the starting point, end point and trajectory of the target instrument channel, a guide plate area with a stress greater than a preset threshold is identified;

[0040] Based on the guide plate area having a stress greater than a preset stress threshold, locally strengthening the initial three-dimensional simulation model to update geometric structure parameters and material properties of the initial three-dimensional simulation model;

[0041] The unit division is re-performed according to the updated initial three-dimensional simulation model to obtain the three-dimensional model of the guide plate.

[0042] Correspondingly, the present invention also provides a guide plate three-dimensional model generation device, including a segmentation module, a registration module, a reconstruction module, a construction module and a generation module; wherein,

[0043] The segmentation module is used to obtain three-dimensional image data of the target area of ​​the maxillary sinus, and use a deep learning segmentation algorithm to segment the three-dimensional image data to identify regional image information of the target area to obtain a segmented area; the regional image information includes bone area information, soft tissue area information and blood vessel direction information;

[0044] The registration module is used to generate point cloud data of the segmented area, and to perform registration processing on the point cloud data at each time point using an iterative closest point algorithm to obtain three-dimensional point data;

[0045] The reconstruction module is used to fuse the three-dimensional point data at each time point, adjust the fusion result through a grid optimization algorithm, and thus obtain a three-dimensional reconstruction model of the segmented area;

[0046] The construction module is used to identify a number of anatomical landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm to obtain information of each anatomical landmark; select design reference points of the target guide plate from the identified anatomical landmarks, and construct a local coordinate system using the anatomical landmark information;

[0047] The generation module is used to generate the initial contour of the guide plate based on the screened reference points and in combination with the local coordinate system using a parametric modeling method; based on the initial contour of the guide plate, the contact area between the target guide plate and the bone surface is refined through a surface fitting algorithm to obtain the target contour of the guide plate, and then the three-dimensional model of the guide plate is generated based on the target contour of the guide plate.

[0048] As a preferred solution, the registration module uses an iterative closest point algorithm to perform registration processing on the point cloud data at each time point to obtain three-dimensional point data, including:

[0049] The registration module generates a corresponding time series according to the acquisition timestamp of each point in the point cloud data;

[0050] According to the time sequence, the point cloud data is loaded into a preset coordinate system to obtain resolution and density information of the point cloud data;

[0051] Performing denoising processing on the point cloud data according to the resolution and density information;

[0052] The overlapping area information of the denoised point cloud data is obtained, and the number of iterations and the convergence threshold of the iterative closest point algorithm are determined; a preset initial transformation matrix is ​​used, and according to the number of iterations and the convergence threshold, the point cloud data at each time point are respectively registered by the iterative closest point algorithm to obtain the three-dimensional point data.

[0053] As a preferred solution, the reconstruction module adjusts the fusion result through a grid optimization algorithm to obtain a three-dimensional reconstruction model of the segmented area, including:

[0054] The reconstruction module generates an initial grid according to the three-dimensional coordinate information of the fusion result;

[0055] Determining the edge topological relationship of the initial grid based on a preset topological data structure and the three-dimensional coordinate information;

[0056] Determining the connectivity of the initial mesh by using the edge topological relationship, thereby obtaining adjacency information of each facet in the initial mesh;

[0057] Based on the adjacency information, a geometric algorithm is used to optimize and adjust the angle and side length of each facet to obtain an optimized mesh;

[0058] Calculating the patch density and patch distribution in the optimized grid; determining the energy function in the Markov random field under the energy minimization condition constraint according to the patch density and patch distribution;

[0059] The optimized grid is iteratively optimized through the energy function to obtain a three-dimensional reconstruction model of the segmented area.

[0060] As a preferred solution, the construction module identifies several anatomical landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm, specifically:

[0061] The construction module obtains initial landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm;

[0062] Determine the position information of each initial landmark point through edge detection and curvature analysis;

[0063] According to the position information of the initial landmark points, the morphological features of each initial landmark point are identified in combination with the relative positions of each initial landmark point and the corresponding adjacent anatomical structure;

[0064] The initial landmark points that meet the preset characteristics are used as the anatomical landmark points.

[0065] As a preferred solution, the generation module generates the initial contour of the guide plate using a parametric modeling method based on the selected reference points and in combination with the local coordinate system, including:

[0066] The generating module obtains preset size parameters, shape parameters and inclination angle of the target guide plate, wherein the size parameters include the length, width, thickness and curvature radius of the guide plate;

[0067] The end point coordinates and the center point coordinates of the target guide plate are determined according to the size parameters and the reference point, and then the initial contour of the guide plate is generated in the local coordinate system according to the shape parameters and the inclination angle.

[0068] As a preferred solution, the generating module generates the initial contour of the guide plate according to the shape parameters and the inclination angle, including:

[0069] The generating module obtains the initial bone surface model of the target area by three-dimensional scanning technology; uses the bone area information to perform fine processing on the initial bone surface model to obtain a three-dimensional bone surface model;

[0070] Obtaining bone surface curvature data based on the bone surface three-dimensional model, and fitting the bone surface curvature data using a least squares method to obtain a fitting result;

[0071] Based on the fitting results and in combination with preset channel modeling data, the starting point, end point and trajectory of the target instrument channel are located;

[0072] A plurality of marking points are generated according to the starting point, the end point and the trajectory of the target instrument channel; the initial contour of the guide plate is generated by combining the marking points with the shape parameters and the tilt angle.

[0073] As a preferred solution, the generating module generates the guide plate three-dimensional model based on the guide plate target contour, including:

[0074] The generation module models the geometric structure parameters and material properties of the guide plate target profile by a finite element method to create an initial three-dimensional simulation model;

[0075] Applying loads and boundary conditions on the basis of the initial three-dimensional simulation model, and obtaining a stress distribution cloud map through finite element simulation;

[0076] According to the stress distribution cloud map combined with the starting point, end point and trajectory of the target instrument channel, a guide plate area with a stress greater than a preset threshold is identified;

[0077] Based on the guide plate area having a stress greater than a preset stress threshold, locally strengthening the initial three-dimensional simulation model to update geometric structure parameters and material properties of the initial three-dimensional simulation model;

[0078] The unit division is re-performed according to the updated initial three-dimensional simulation model to obtain the three-dimensional model of the guide plate.

[0079] Correspondingly, the present invention application also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the guide plate three-dimensional model generation method when executing the computer program.

[0080] Correspondingly, the present invention application also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the guide plate three-dimensional model generation method.

[0081] Compared with the prior art, the present invention has at least the following beneficial effects:

[0082] The present invention application provides a method, device, terminal device and storage medium for generating a three-dimensional model of a guide plate. The generating method comprises: acquiring three-dimensional image data of a target area of ​​a maxillary sinus, segmenting the three-dimensional image data using a deep learning segmentation algorithm to identify regional image information of the target area and obtain a segmented area; the regional image information includes bone regional information, soft tissue regional information and blood vessel direction information; generating point cloud data of the segmented area, registering the point cloud data at each time point using an iterative closest point algorithm, and obtaining three-dimensional point data; fusing the three-dimensional point data at each time point, and optimizing the fusion result using a grid optimization algorithm. Adjustments are made to obtain a three-dimensional reconstruction model of the segmented area; based on the three-dimensional reconstruction model, a number of anatomical landmarks are identified through a geometric feature extraction algorithm to obtain information on each anatomical landmark point; design reference points of a target guide are screened out from the identified anatomical landmark points, and a local coordinate system is constructed using the anatomical landmark point information; based on each screened reference point, in combination with the local coordinate system, a parametric modeling method is used to generate an initial contour of the guide; based on the initial contour of the guide, a surface fitting algorithm is used to refine the contact area between the target guide and the bone surface to obtain a target contour of the guide, and then a three-dimensional model of the guide is generated based on the target contour of the guide. The present invention applies a segmentation algorithm based on deep learning to segment the three-dimensional image of the target area and accurately identify the bone, soft tissue and blood vessels; the point cloud data at each time point are respectively registered to ensure the temporal consistency of the data; the fusion result is adjusted by a grid optimization algorithm to make the grid both uniform and consistent with the actual biological tissue structure, thereby ensuring the grid quality of the three-dimensional reconstructed model; a number of anatomical landmarks are identified by a geometric feature extraction algorithm, and a local coordinate system is constructed, on which basis a parametric modeling method is used to generate the guide plate contour, so that the generation of the guide plate three-dimensional model is more refined than the existing technology, and can match the anatomical structure of the target area of ​​the maxillary sinus, and the guide plate generated by subsequent 3D printing can be more stable during surgery. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 : A flow chart of an embodiment of a method for generating a three-dimensional model of a guide plate provided by the present invention.

[0084] Figure 2 : A schematic diagram of the effect of an application example of a guide plate three-dimensional model generated by the guide plate three-dimensional model generation method provided by the present invention.

[0085] Figure 3 : A structural schematic diagram of an embodiment of a guide plate three-dimensional model generating device provided by the present invention. DETAILED DESCRIPTION

[0086] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0087] Embodiment 1

[0088] Please refer to Figure 1 , Figure 1 The present invention provides a method for generating a three-dimensional model of a guide plate, comprising steps S1 to S5; wherein:

[0089] Step S1, acquiring three-dimensional image data of a target area of ​​the maxillary sinus, and using a deep learning segmentation algorithm to segment the three-dimensional image data to identify regional image information of the target area to obtain a segmented area.

[0090] The regional image information includes bone regional information, soft tissue regional information and blood vessel direction information. The present application can obtain the patient's dentition and gingival 3D surface STL data through an intraoral 3D scanner, and simultaneously shoot a large-field CBCT (oral and maxillofacial cone beam CT) containing the maxillary sinus, obtain DICOM (Digital Imaging and Communications in Medicine) format data, etc., to obtain the 3D image data of the maxillary sinus target area.

[0091] Furthermore, when processing three-dimensional medical imaging data, deep learning-based segmentation algorithms, such as convolutional neural networks (CNNs), can be used to accurately identify the bone, soft tissue, and blood vessel orientation in the target area of ​​the maxillary sinus.

[0092] The image segmentation model used can be pre-trained. In the training phase, a large amount of labeled or annotated medical image data can be used, and the Dice coefficient can be used as a loss function to optimize the tissue boundary accuracy of the model output. For example, the Dice coefficient target value can be set to 9 or more to ensure high accuracy.

[0093] Furthermore, an image segmentation model that meets a preset convergence condition is used to segment the three-dimensional image data of the target area of ​​the maxillary sinus to identify the regional image information of the target area and obtain a segmented area. The preset convergence condition can be set according to the loss function.

[0094] Step S2, generating point cloud data of the segmented area, and using an iterative closest point algorithm to perform registration processing on the point cloud data at each time point to obtain three-dimensional point data.

[0095] In this step, the segmented image data can be converted into three-dimensional point cloud data using voxel point conversion technology. This implementation usually involves converting the position and type of each voxel (bone, soft tissue, etc.) into a point in the point cloud. Then, the iterative closest point (ICP) algorithm is applied to these point cloud data to align the point clouds from different time points to improve the temporal consistency of the data.

[0096] As a preferred implementation, the point cloud data at each time point is registered using an iterative closest point algorithm to obtain three-dimensional point data, including:

[0097] Generate a corresponding time series according to the acquisition timestamp of each point in the point cloud data; load the point cloud data into a preset coordinate system according to the time series to obtain the resolution and density information of the point cloud data; perform denoising on the point cloud data according to the resolution and density information; obtain the overlapping area information of the denoised point cloud data, and determine the number of iterations and convergence threshold of the iterative closest point algorithm; use a preset initial transformation matrix, and according to the number of iterations and the convergence threshold, perform registration processing on the point cloud data at each time point through the iterative closest point algorithm to obtain the three-dimensional point data.

[0098] In this preferred embodiment, the order of data can be determined by timestamps to obtain an ordered point cloud data set; for the ordered point cloud data set, a unified coordinate system is used for data processing to ensure that all point cloud data are in the same coordinate system, thereby providing an accurate benchmark for data registration; the resolution and density information of the point cloud data are used to characterize the data characteristics of the point cloud, thereby analyzing the noise level therein and implementing noise reduction processing, which can reduce the impact of noise on the point cloud data; based on the denoised data, the size of the overlapping area of ​​the data is obtained. The overlapping area is a key factor in the stability of the ICP algorithm, which can help determine whether the overlapping area is sufficient to support high-quality data registration. At the same time, the overlapping area information is used to determine the number of iterations and convergence threshold of the ICP algorithm to adapt to the data characteristics, so that the ICP algorithm can obtain higher registration quality with fewer iterations, reduce the number of invalid iterations, and achieve a better balance between convergence speed and registration quality.

[0099] Step S3, fusing the three-dimensional point data at each time point, adjusting the fusion result by a grid optimization algorithm, and thereby obtaining a three-dimensional reconstruction model of the segmented area.

[0100] In this step, the 3D point data that has been registered is fused, and a grid is generated based on the 3D coordinate information of the fusion result. After a number of processing and optimization, a 3D reconstruction model of the segmented area is obtained. This step can provide data support by generating a 3D reconstruction model of the segmented area, helping to accurately generate a 3D model of the target guide.

[0101] In some preferred embodiments, the fusion result is adjusted by a mesh optimization algorithm to obtain a three-dimensional reconstruction model of the segmented area, including: generating an initial mesh according to the three-dimensional coordinate information of the fusion result; determining the edge topological relationship of the initial mesh based on a preset topological data structure and the three-dimensional coordinate information; using the edge topological relationship to judge the connectivity of the initial mesh, thereby obtaining the adjacency information of each facet in the initial mesh; based on the adjacency information, using a geometric algorithm to optimize and adjust the angle and side length of each facet to obtain an optimized mesh; calculating the facet density and facet distribution in the optimized mesh; based on the facet density and facet distribution, determining the energy function in the Markov random field under the energy minimization condition constraint; iteratively optimizing the optimized mesh through the energy function to obtain the three-dimensional reconstruction model of the segmented area.

[0102] By implementing the implementation method of the present application, by judging the connectivity of the mesh, the smoothness and consistency of the mesh can be obtained to ensure the reasonable structure of the mesh; in addition, combined with the relationship between adjacent facets, the topological association between each facet can be effectively identified to facilitate subsequent geometric processing and optimization, and obtain a better quality optimized mesh. On this basis, the energy function in the Markov random field is determined according to the facet density and distribution of the optimized mesh. At this time, the energy function combines factors such as mesh density to minimize the overall energy to achieve the optimal configuration of the mesh, and obtain a more accurate three-dimensional reconstruction model of the segmented area.

[0103] Step S4, based on the three-dimensional reconstructed model, a number of anatomical landmarks are identified through a geometric feature extraction algorithm to obtain information of each anatomical landmark; the design reference points of the target guide are selected from the identified anatomical landmarks, and a local coordinate system is constructed using the anatomical landmark information.

[0104] In a preferred implementation of this step, the three-dimensional reconstructed model is used to identify several anatomical landmark points through a geometric feature extraction algorithm, specifically: based on the three-dimensional reconstructed model, initial landmark points are obtained through a geometric feature extraction algorithm; the position information of each initial landmark point is determined through edge detection and curvature analysis; based on the position information of the initial landmark points, the morphological characteristics of each initial landmark point are identified in combination with the relative position of each initial landmark point and the corresponding adjacent anatomical structure; the initial landmark points that meet the preset characteristics are used as the anatomical landmark points.

[0105] This implementation method can obtain a digital representation of the user's anatomical structure based on a three-dimensional reconstruction model. The three-dimensional reconstruction model is processed using a geometric feature extraction algorithm to identify key anatomical landmarks. The precise location of the landmarks is determined by edge detection, curvature analysis and other methods. Furthermore, by using position information to identify the morphological features of the landmarks, such as abnormal protrusions or depressions, a clinical correlation with pathological phenomena can be established, and then combined with preset features for identification. The anatomical landmarks obtained can adapt to different types of anatomical structures and complex situations, while improving the recognition accuracy and extraction efficiency of the landmarks.

[0106] Step S5, based on the screened reference points, combined with the local coordinate system, a parametric modeling method is used to generate an initial contour of the guide plate; based on the initial contour of the guide plate, a surface fitting algorithm is used to refine the contact area between the target guide plate and the bone surface to obtain the target contour of the guide plate, and then the three-dimensional model of the guide plate is generated based on the target contour of the guide plate.

[0107] In this step, the preset size parameters, shape parameters and inclination angle of the target guide plate are obtained, and the size parameters include the length, width, thickness and curvature radius of the guide plate; the endpoint coordinates and center point coordinates of the target guide plate are determined according to the size parameters and the reference point, and then in the local coordinate system, the initial contour of the guide plate is generated according to the shape parameters and the inclination angle.

[0108] Among them, the preset size parameters, shape parameters and inclination angle of the target guide plate can be obtained according to the design requirements of the guide plate. Through these basic parameters, the preliminary outer dimensions of the guide plate can be obtained, laying the foundation for the subsequent detailed design. According to these basic parameters, combined with the above-mentioned reference points, the coordinate points at both ends and the middle position of the guide plate can be preliminarily determined. These coordinate points can be used to accurately describe the spatial position of the guide plate in the local coordinate system, ensuring that the guide plate can be correctly installed and used according to the design intent.

[0109] Furthermore, considering that the generated three-dimensional model of the guide plate can be combined with factors such as the channel through which the instrument enters to further improve the accuracy, the generating the initial contour of the guide plate according to the shape parameters and the inclination angle includes: obtaining the initial bone surface model of the target area through three-dimensional scanning technology; using the bone area information, refining the initial bone surface model to obtain the three-dimensional model of the bone surface; obtaining bone surface curvature data based on the three-dimensional model of the bone surface, fitting the bone surface curvature data using the least squares method to obtain a fitting result (Gaussian curvature and average curvature can be used as two indicators); based on the fitting result, combined with preset channel modeling data, locating the starting point, end point and trajectory of the target instrument channel; generating a number of marking points according to the starting point, end point and trajectory of the target instrument channel (for example, according to the curvature distribution obtained by fitting, a spatial geometric calculation method can be used, with the minimum bone surface curvature point as the starting point and the maximum curvature point as the end point to help locate the instrument channel); using the marking points in combination with the shape parameters and the inclination angle to generate the initial contour of the guide plate.

[0110] Furthermore, the generation of the three-dimensional model of the guide plate based on the target contour of the guide plate includes: modeling the geometric structure parameters and material properties of the target contour of the guide plate by the finite element method, creating an initial three-dimensional simulation model (the initial three-dimensional simulation model may include parameters such as the size, shape, elastic modulus, Poisson's ratio and yield strength of the guide plate); applying loads and boundary conditions on the basis of the initial three-dimensional simulation model, and obtaining a stress distribution cloud map by finite element simulation; identifying the guide plate area greater than the preset stress threshold, i.e., the guide plate weak area, where the stress is concentrated or exceeds the material yield strength, according to the stress distribution cloud map combined with the starting point, end point and trajectory of the target instrument channel; based on the guide plate area greater than the preset stress threshold, locally strengthening the initial three-dimensional simulation model to update the geometric structure parameters and material properties of the initial three-dimensional simulation model; re-dividing the unit according to the updated initial three-dimensional simulation model to obtain the three-dimensional model of the guide plate. The update may be considered to be in the form of iterative update, and each iteration may record the number of simulation analysis and structural adjustment to ensure the efficiency and effect of the entire optimization process.

[0111] like Figure 2The figure shows an application example of the invention application using a guide plate three-dimensional model generation method to generate a guide plate three-dimensional model. The target guide plate is a U-shaped low-window maxillary sinus floor external lifting guide plate. The generated guide plate is a U-shaped design with an open upper boundary. The inner frame of the guide plate is used as the indicator edge when bone removal and window opening. The inner boundary line extends 2mm outward as the outer contour line of the U-shaped window to ensure the strength of the guide plate. The lower boundary is flush with the maxillary sinus floor, the front boundary is flush with the maxillary sinus anterior wall, and the rear boundary is positioned corresponding to the farthest planned implant. The maxillary sinus floor and the outer edge position are accurately positioned using a four-point square as the positioning point, and the design is a concave shape to facilitate intraoperative positioning and prevent slippage. The height of the window does not exceed the bottom of the vestibule groove, and the specific height can be adjusted in place according to the soft tissue conditions during the operation. For the target guide plate, a 3D printer can be used to complete the processing of the integrated model of the base plate and the maxillary sinus external lifting guide plate.

[0112] Accordingly, the present invention also provides a guide plate three-dimensional model generation device, including a segmentation module 101, a registration module 102, a reconstruction module 103, a construction module 104 and a generation module 105; wherein,

[0113] The segmentation module 101 is used to obtain three-dimensional image data of the target area of ​​the maxillary sinus, and use a deep learning segmentation algorithm to segment the three-dimensional image data to identify regional image information of the target area to obtain a segmented area; the regional image information includes bone area information, soft tissue area information and blood vessel direction information;

[0114] The registration module 102 is used to generate point cloud data of the segmented area, and use an iterative closest point algorithm to perform registration processing on the point cloud data at each time point to obtain three-dimensional point data;

[0115] The reconstruction module 103 is used to fuse the three-dimensional point data at each time point, and adjust the fusion result through a grid optimization algorithm, so as to obtain a three-dimensional reconstruction model of the segmented area;

[0116] The construction module 104 is used to identify a number of anatomical landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm to obtain information of each anatomical landmark; select design reference points of the target guide from the identified anatomical landmarks, and construct a local coordinate system using the anatomical landmark information;

[0117] The generation module 105 is used to generate the initial contour of the guide plate based on the screened reference points in combination with the local coordinate system using a parametric modeling method; based on the initial contour of the guide plate, the contact area between the target guide plate and the bone surface is refined through a surface fitting algorithm to obtain the target contour of the guide plate, and then the three-dimensional model of the guide plate is generated based on the target contour of the guide plate.

[0118] As a preferred solution, the registration module 102 uses an iterative closest point algorithm to perform registration processing on the point cloud data at each time point to obtain three-dimensional point data.

[0119] The registration module 102 generates a corresponding time series according to the acquisition timestamp of each point in the point cloud data;

[0120] According to the time sequence, the point cloud data is loaded into a preset coordinate system to obtain resolution and density information of the point cloud data;

[0121] Performing denoising processing on the point cloud data according to the resolution and density information;

[0122] The overlapping area information of the denoised point cloud data is obtained, and the number of iterations and the convergence threshold of the iterative closest point algorithm are determined; a preset initial transformation matrix is ​​used, and according to the number of iterations and the convergence threshold, the point cloud data at each time point are respectively registered by the iterative closest point algorithm to obtain the three-dimensional point data.

[0123] As a preferred solution, the reconstruction module 103 adjusts the fusion result through a grid optimization algorithm to obtain a three-dimensional reconstruction model of the segmented area, including:

[0124] The reconstruction module 103 generates an initial grid according to the three-dimensional coordinate information of the fusion result;

[0125] Determining the edge topological relationship of the initial grid based on a preset topological data structure and the three-dimensional coordinate information;

[0126] Determining the connectivity of the initial mesh by using the edge topological relationship, thereby obtaining adjacency information of each facet in the initial mesh;

[0127] Based on the adjacency information, a geometric algorithm is used to optimize and adjust the angle and side length of each facet to obtain an optimized mesh;

[0128] Calculating the patch density and patch distribution in the optimized grid; determining the energy function in the Markov random field under the energy minimization condition constraint according to the patch density and patch distribution;

[0129] The optimized grid is iteratively optimized through the energy function to obtain a three-dimensional reconstruction model of the segmented area.

[0130] As a preferred solution, the construction module 104 identifies several anatomical landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm, specifically:

[0131] The construction module 104 obtains initial landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm;

[0132] Determine the position information of each initial landmark point through edge detection and curvature analysis;

[0133] According to the position information of the initial landmark points, the morphological features of each initial landmark point are identified in combination with the relative positions of each initial landmark point and the corresponding adjacent anatomical structure;

[0134] The initial landmark points that meet the preset characteristics are used as the anatomical landmark points.

[0135] As a preferred solution, the generating module 105 generates the initial contour of the guide plate by using a parametric modeling method based on the selected reference points and in combination with the local coordinate system, including:

[0136] The generating module 105 obtains preset size parameters, shape parameters and tilt angle of the target guide plate, wherein the size parameters include the length, width, thickness and curvature radius of the guide plate;

[0137] The end point coordinates and the center point coordinates of the target guide plate are determined according to the size parameters and the reference point, and then the initial contour of the guide plate is generated in the local coordinate system according to the shape parameters and the inclination angle.

[0138] As a preferred solution, the generating module 105 generates the initial contour of the guide plate according to the shape parameters and the inclination angle, including:

[0139] The generating module 105 obtains the initial bone surface model of the target area through the three-dimensional scanning technology; uses the bone area information to perform a fine processing on the initial bone surface model to obtain a three-dimensional bone surface model;

[0140] Obtaining bone surface curvature data based on the bone surface three-dimensional model, and fitting the bone surface curvature data using a least squares method to obtain a fitting result;

[0141] Based on the fitting results and in combination with preset channel modeling data, the starting point, end point and trajectory of the target instrument channel are located;

[0142] A plurality of marking points are generated according to the starting point, the end point and the trajectory of the target instrument channel; the initial contour of the guide plate is generated by combining the marking points with the shape parameters and the tilt angle.

[0143] As a preferred solution, the generating module 105 generates the guide plate three-dimensional model based on the guide plate target contour, including:

[0144] The generating module 105 models the geometric structure parameters and material properties of the target contour of the guide plate by using the finite element method to create an initial three-dimensional simulation model;

[0145] Applying loads and boundary conditions on the basis of the initial three-dimensional simulation model, and obtaining a stress distribution cloud map through finite element simulation;

[0146] According to the stress distribution cloud map combined with the starting point, end point and trajectory of the target instrument channel, a guide plate area with a stress greater than a preset threshold is identified;

[0147] Based on the guide plate area having a stress greater than a preset stress threshold, locally strengthening the initial three-dimensional simulation model to update geometric structure parameters and material properties of the initial three-dimensional simulation model;

[0148] The unit division is re-performed according to the updated initial three-dimensional simulation model to obtain the three-dimensional model of the guide plate.

[0149] Correspondingly, the present invention application also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the guide plate three-dimensional model generation method when executing the computer program.

[0150] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal, and various interfaces and lines are used to connect various parts of the entire terminal.

[0151] The memory can be used to store the computer program, and the processor realizes various functions of the terminal by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0152] Correspondingly, the present invention application also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the guide plate three-dimensional model generation method.

[0153] Wherein, if the module integrated in the guide plate three-dimensional model generating device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0154] Compared with the prior art, the present invention has at least the following beneficial effects:

[0155] The present invention application provides a method, device, terminal device and storage medium for generating a three-dimensional model of a guide plate. The generating method comprises: acquiring three-dimensional image data of a target area of ​​a maxillary sinus, segmenting the three-dimensional image data using a deep learning segmentation algorithm to identify regional image information of the target area and obtain a segmented area; the regional image information includes bone regional information, soft tissue regional information and blood vessel direction information; generating point cloud data of the segmented area, registering the point cloud data at each time point using an iterative closest point algorithm, and obtaining three-dimensional point data; fusing the three-dimensional point data at each time point, and optimizing the fusion result using a grid optimization algorithm. Adjustments are made to obtain a three-dimensional reconstruction model of the segmented area; based on the three-dimensional reconstruction model, a number of anatomical landmarks are identified through a geometric feature extraction algorithm to obtain information on each anatomical landmark point; design reference points of a target guide are screened out from the identified anatomical landmark points, and a local coordinate system is constructed using the anatomical landmark point information; based on each screened reference point, in combination with the local coordinate system, a parametric modeling method is used to generate an initial contour of the guide; based on the initial contour of the guide, a surface fitting algorithm is used to refine the contact area between the target guide and the bone surface to obtain a target contour of the guide, and then a three-dimensional model of the guide is generated based on the target contour of the guide. The present invention applies a segmentation algorithm based on deep learning to segment the three-dimensional image of the target area and accurately identify the bone, soft tissue and blood vessels; the point cloud data at each time point are respectively registered to ensure the temporal consistency of the data; the fusion result is adjusted by a grid optimization algorithm to make the grid both uniform and consistent with the actual biological tissue structure, thereby ensuring the grid quality of the three-dimensional reconstructed model; a number of anatomical landmarks are identified by a geometric feature extraction algorithm, and a local coordinate system is constructed, on which basis a parametric modeling method is used to generate the guide plate contour, so that the generation of the guide plate three-dimensional model is more refined than the existing technology, and can match the anatomical structure of the target area of ​​the maxillary sinus, and the guide plate generated by subsequent 3D printing can be more stable during surgery.

[0156] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for generating a three-dimensional model of a guide plate, characterized in that: include: Acquire three-dimensional image data of a target area of ​​the maxillary sinus, and segment the three-dimensional image data using a deep learning segmentation algorithm to identify regional image information of the target area to obtain a segmented area; the regional image information includes bone area information, soft tissue area information, and blood vessel direction information; Generate point cloud data of the segmented area, and use an iterative closest point algorithm to perform registration processing on the point cloud data at each time point to obtain three-dimensional point data; Fusing the three-dimensional point data at each time point, adjusting the fusion result by a grid optimization algorithm, and thereby obtaining a three-dimensional reconstruction model of the segmented area; Based on the three-dimensional reconstruction model, a plurality of anatomical landmarks are identified by a geometric feature extraction algorithm to obtain information of each anatomical landmark; design reference points of a target guide are selected from the identified anatomical landmarks, and a local coordinate system is constructed using the anatomical landmark information; Based on the selected reference points, in combination with the local coordinate system, a parametric modeling method is used to generate an initial outline of the guide plate; based on the initial outline of the guide plate, a surface fitting algorithm is used to refine the contact area between the target guide plate and the bone surface to obtain a target outline of the guide plate, and then a three-dimensional model of the guide plate is generated based on the target outline of the guide plate; The iterative closest point algorithm is used to perform registration processing on the point cloud data at each time point to obtain three-dimensional point data, including: Generate a corresponding time series according to the acquisition timestamp of each point in the point cloud data; According to the time sequence, the point cloud data is loaded into a preset coordinate system to obtain resolution and density information of the point cloud data; Performing denoising processing on the point cloud data according to the resolution and density information; Obtaining the overlapping area information of the denoised point cloud data, and determining the number of iterations and the convergence threshold of the iterative closest point algorithm; using a preset initial transformation matrix, and according to the number of iterations and the convergence threshold, performing registration processing on the point cloud data at each time point by the iterative closest point algorithm, to obtain the three-dimensional point data; The step of adjusting the fusion result by a grid optimization algorithm to obtain a three-dimensional reconstruction model of the segmented area includes: Generate an initial grid according to the three-dimensional coordinate information of the fusion result; Determining the edge topological relationship of the initial grid based on a preset topological data structure and the three-dimensional coordinate information; Determining the connectivity of the initial mesh by using the edge topological relationship, thereby obtaining adjacency information of each facet in the initial mesh; Based on the adjacency information, a geometric algorithm is used to optimize and adjust the angle and side length of each facet to obtain an optimized mesh; Calculating the patch density and patch distribution in the optimized grid; determining the energy function in the Markov random field under the energy minimization condition constraint according to the patch density and patch distribution; Iteratively optimizing the optimized grid by using the energy function to obtain a three-dimensional reconstruction model of the segmented area; Based on the three-dimensional reconstruction model, a plurality of anatomical landmarks are identified by a geometric feature extraction algorithm, specifically: Based on the three-dimensional reconstruction model, initial landmarks are obtained by using a geometric feature extraction algorithm; Determine the position information of each initial landmark point through edge detection and curvature analysis; According to the position information of the initial landmark points, the morphological features of each initial landmark point are identified in combination with the relative positions of each initial landmark point and the corresponding adjacent anatomical structure; Using the initial landmark point that meets the preset characteristics as the anatomical landmark point; The method of generating the initial contour of the guide plate by using a parametric modeling method based on the selected reference points and in combination with the local coordinate system includes: Acquire preset size parameters, shape parameters and inclination angle of the target guide plate, wherein the size parameters include the length, width, thickness and curvature radius of the guide plate; Determine the end point coordinates and the center point coordinates of the target guide plate according to the size parameters and the reference point, and then generate the initial contour of the guide plate in the local coordinate system according to the shape parameters and the inclination angle; Generating the initial contour of the guide plate according to the shape parameter and the inclination angle comprises: Acquire the initial bone surface model of the target area by three-dimensional scanning technology; use the bone area information to refine the initial bone surface model to obtain a three-dimensional bone surface model; Obtaining bone surface curvature data based on the bone surface three-dimensional model, and fitting the bone surface curvature data using a least squares method to obtain a fitting result; Based on the fitting results and in combination with preset channel modeling data, the starting point, end point and trajectory of the target instrument channel are located; A plurality of marking points are generated according to the starting point, the end point and the trajectory of the target instrument channel; the initial contour of the guide plate is generated by combining the marking points with the shape parameters and the tilt angle.

2. A guide plate three-dimensional model generating device, characterized in that: It includes segmentation module, registration module, reconstruction module, construction module and generation module; among them, The segmentation module is used to obtain three-dimensional image data of the target area of ​​the maxillary sinus, and use a deep learning segmentation algorithm to segment the three-dimensional image data to identify regional image information of the target area to obtain a segmented area; the regional image information includes bone area information, soft tissue area information and blood vessel direction information; The registration module is used to generate point cloud data of the segmented area, and to perform registration processing on the point cloud data at each time point using an iterative closest point algorithm to obtain three-dimensional point data; The reconstruction module is used to fuse the three-dimensional point data at each time point, adjust the fusion result through a grid optimization algorithm, and thus obtain a three-dimensional reconstruction model of the segmented area; The construction module is used to identify a number of anatomical landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm to obtain information of each anatomical landmark; select design reference points of the target guide plate from the identified anatomical landmarks, and construct a local coordinate system using the anatomical landmark information; The generation module is used to generate an initial guide plate contour by a parametric modeling method based on the selected reference points and in combination with the local coordinate system; based on the initial guide plate contour, the contact area between the target guide plate and the bone surface is refined by a surface fitting algorithm to obtain a guide plate target contour, and then generate the guide plate three-dimensional model based on the guide plate target contour; The registration module uses an iterative closest point algorithm to perform registration processing on the point cloud data at each time point to obtain three-dimensional point data. The registration module generates a corresponding time series according to the acquisition timestamp of each point in the point cloud data; According to the time sequence, the point cloud data is loaded into a preset coordinate system to obtain resolution and density information of the point cloud data; Performing denoising processing on the point cloud data according to the resolution and density information; Obtaining the overlapping area information of the denoised point cloud data, and determining the number of iterations and the convergence threshold of the iterative closest point algorithm; using a preset initial transformation matrix, and according to the number of iterations and the convergence threshold, performing registration processing on the point cloud data at each time point by the iterative closest point algorithm, to obtain the three-dimensional point data; The reconstruction module adjusts the fusion result by a grid optimization algorithm, thereby obtaining a three-dimensional reconstruction model of the segmented area, including: The reconstruction module generates an initial grid according to the three-dimensional coordinate information of the fusion result; Determining the edge topological relationship of the initial grid based on a preset topological data structure and the three-dimensional coordinate information; Determining the connectivity of the initial mesh by using the edge topological relationship, thereby obtaining adjacency information of each facet in the initial mesh; Based on the adjacency information, a geometric algorithm is used to optimize and adjust the angle and side length of each facet to obtain an optimized mesh; Calculating the patch density and patch distribution in the optimized grid; determining the energy function in the Markov random field under the energy minimization condition constraint according to the patch density and patch distribution; Iteratively optimizing the optimized grid by using the energy function to obtain a three-dimensional reconstruction model of the segmented area; The construction module identifies several anatomical landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm, specifically: The construction module obtains initial landmarks based on the three-dimensional reconstruction model through a geometric feature extraction algorithm; Determine the position information of each initial landmark point through edge detection and curvature analysis; According to the position information of the initial landmark points, the morphological features of each initial landmark point are identified in combination with the relative positions of each initial landmark point and the corresponding adjacent anatomical structure; Using the initial landmark point that meets the preset characteristics as the anatomical landmark point; The generation module generates the initial contour of the guide plate by using a parametric modeling method based on the selected reference points and in combination with the local coordinate system, including: The generating module obtains preset size parameters, shape parameters and inclination angle of the target guide plate, wherein the size parameters include the length, width, thickness and curvature radius of the guide plate; Determine the end point coordinates and the center point coordinates of the target guide plate according to the size parameters and the reference point, and then generate the initial contour of the guide plate in the local coordinate system according to the shape parameters and the inclination angle; The generating module generates the initial contour of the guide plate according to the shape parameter and the inclination angle, including: The generating module obtains the initial bone surface model of the target area by three-dimensional scanning technology; uses the bone area information to perform fine processing on the initial bone surface model to obtain a three-dimensional bone surface model; Obtaining bone surface curvature data based on the bone surface three-dimensional model, and fitting the bone surface curvature data using a least squares method to obtain a fitting result; Based on the fitting results and in combination with preset channel modeling data, the starting point, end point and trajectory of the target instrument channel are located; A plurality of marking points are generated according to the starting point, the end point and the trajectory of the target instrument channel; the initial contour of the guide plate is generated by combining the marking points with the shape parameters and the tilt angle.

3. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the method for generating a three-dimensional model of a guide plate according to claim 1 is implemented when the processor executes the computer program.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the guide plate three-dimensional model generating method according to claim 1.

Images