Curvature enhancement based tooth model reconstruction method, apparatus, device and medium

Abstract

The application provides a tooth model reconstruction method and device based on curvature enhancement, equipment and medium, which comprises the following steps: acquiring CBCT images and intraoral scanning data of a patient in the same period; marking tooth models of the CBCT images and crown models of the intraoral scanning data respectively; registering the crown models of the intraoral scanning data to the tooth models of the CBCT images through a rigid ICP algorithm; and based on a curvature enhancement implicit function network, finely registering the three-dimensional shape of the crown in the tooth model, so as to obtain a high-quality tooth model. The application generates a high-quality tooth model from CBCT images. Only one kind of data, i.e. CBCT images, needs to be input into a trained network, and then a high-quality tooth model of each tooth can be obtained. This not only improves the diagnosis and treatment efficiency of tooth orthodontics, but also can be applied to a digital orthodontic system to quickly design a tooth arrangement scheme for each patient, and then arrange a specific tooth correction plan.

Description

Technical Field

[0001] This invention relates to the field of digital orthodontics, and more particularly to a method, apparatus, device, and medium for reconstructing tooth models based on curvature enhancement. Background Technology

[0002] With the development of computer-aided technology, digital dental systems have been widely used in dental hospitals and clinics for oral disease diagnosis, clinical surgical restoration, treatment planning, and postoperative evaluation. In digital orthodontic treatment, CBCT images are often used to reconstruct patient tooth models, simultaneously acquiring information on the crown and root. However, due to the low resolution of CBCT images, crowns generated from CBCT images often lose important geometric surface details, failing to meet the needs of precise clinical treatment planning. Therefore, in current clinical practice, CBCT images are usually combined with high-resolution intraoral scans captured by an intraoral scanner. However, the process of scanning and combining these two types of data is very time-consuming and is affected and limited by patient conditions and acquisition equipment, often making it impossible to acquire both pairs of data for the same patient. Summary of the Invention

[0003] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide a method, apparatus, device and medium for tooth model reconstruction based on curvature enhancement, which can solve the deficiencies in the acquisition and fusion of CBCT images and intraoral scan models in the prior art.

[0004] To achieve the above and other related objectives, this invention provides a method for reconstructing a tooth model based on curvature enhancement. The method includes: acquiring CBCT images and intraoral scan data of a patient at the same time; labeling the tooth model in the CBCT images and the crown model in the intraoral scan data respectively; registering the crown model in the intraoral scan data to the tooth model in the CBCT images using a rigid ICP algorithm; and refining the three-dimensional shape of the crown in the tooth model after fine registration based on a curvature enhancement hidden function network to obtain a high-quality tooth model.

[0005] In one embodiment of the present invention, the registration of the crown model of intraoral scan data to the tooth model of CBCT image using the rigid ICP algorithm includes: estimating the corresponding points in the crown model and the tooth model, and performing iterative calculation based on the least squares method; calculating the distance between the vertices of the crown model and the tooth model, removing the crown portion of the tooth model by removing points with a distance less than a preset value, and replacing it with the crown model of the intraoral scan data; and using the Poisson surface reconstruction algorithm to fill the lateral cracks of the registered tooth model, so as to fill the missing part of the surface of the registered tooth model and form a closed surface.

[0006] In one embodiment of the present invention, the curvature-enhanced implicit function network includes: an instance segmentation module for detecting and segmenting tooth models of each tooth from CBCT images; a surface reconstruction module for reconstructing tooth models based on the implicit function representation; and a curvature enhancement module for enhancing the learning of local geometric features of the crown portion.

[0007] In one embodiment of the present invention, the method of reconstructing a tooth model based on a latent function includes: stacking the tooth model of each tooth obtained by the instance segmentation module and the tooth model of the labeled CBCT image together, and inputting them into an encoder to obtain a multi-scale feature map; acquiring multiple access points and sampling them on the feature map to obtain multi-scale features of each access point; inputting the multi-scale features of each access point into a multilayer perceptron as a classifier to obtain a binary classification of each access point on and off the surface, which is used as a latent function representation of the three-dimensional shape.

[0008] In one embodiment of the present invention, the method further includes: sampling access points near the surface of the tooth with a Gaussian distribution during the training phase and / or uniformly acquiring multiple dense points in space as access points during the testing phase; dividing the space into discrete three-dimensional voxels using the Marching Cubes algorithm, and extracting isosurfaces in each three-dimensional voxel using triangular facets to obtain a complete explicit surface representation.

[0009] In one embodiment of the present invention, the learning of the local geometric features of the enhanced crown portion includes: using KNN to find the nearest access points for each vertex of the registered tooth model; performing a weighted average based on the multi-scale features of each access point to obtain the features of each vertex; and inputting the features of each vertex into a multilayer perceptron as a regressor to predict the curvature value of each vertex.

[0010] To achieve the above and other related objectives, the present invention provides a tooth model reconstruction device based on curvature enhancement. The device includes: an acquisition module for acquiring CBCT images and intraoral scan data of a patient at the same time; a registration module for annotating the tooth model in the CBCT image and the crown model in the intraoral scan data respectively, and registering the crown model in the intraoral scan data to the tooth model in the CBCT image using a rigid ICP algorithm; and an optimization module for optimizing the three-dimensional shape of the crown in the finely registered tooth model based on a curvature enhancement hidden function network to obtain a high-quality tooth model.

[0011] To achieve the above and other related objectives, the present invention provides a computer device, the device comprising: a memory and a processor; the memory storing a computer program, and the processor executing the computer program to implement the method described above.

[0012] To achieve the above and other related objectives, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0013] As described above, the present invention provides a method, apparatus, device, and medium for reconstructing a tooth model based on curvature enhancement. This involves acquiring CBCT images and intraoral scan data of a patient at the same time; labeling the tooth model in the CBCT images and the crown model in the intraoral scan data respectively; registering the crown model from the intraoral scan data to the tooth model in the CBCT images using a rigid ICP algorithm; and obtaining a high-quality tooth model based on the three-dimensional shape of the crown in the finely registered tooth model using a curvature enhancement implicit function network.

[0014] It has the following beneficial effects:

[0015] This application generates high-quality tooth models from CBCT images. By simply inputting CBCT images into a trained network, a high-quality tooth model for each tooth can be obtained. This not only improves the efficiency of orthodontic treatment but can also be applied to digital orthodontic systems to quickly design tooth alignment plans for each patient, thereby arranging specific orthodontic treatment plans. Attached Figure Description

[0016] Figure 1 The diagram shows a flowchart of a tooth model reconstruction method based on curvature enhancement according to one embodiment of the present invention.

[0017] Figure 2 The diagram shows a scenario of tooth model reconstruction based on curvature enhancement in one embodiment of the present invention.

[0018] Figure 3 The diagram shows a scenario of registering a crown model and a tooth model in one embodiment of the present invention.

[0019] Figure 4 The diagram shown is a schematic representation of a model based on a curvature-enhanced implicit function network in one embodiment of the present invention.

[0020] Figure 5 This diagram illustrates a comparison of different reconstruction results in one embodiment of the present invention.

[0021] Figure 6 The diagram shows a module schematic of a tooth model reconstruction device based on curvature enhancement in one embodiment of the present invention.

[0022] Figure 7 The diagram shown is a structural schematic of a computer device according to one embodiment of the present invention. Detailed Implementation

[0023] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0024] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0025] This application is primarily for digital orthodontics. Orthodontics involves applying external force to correct teeth, and different treatment plans need to be developed based on each patient's individual circumstances. Clinically, the following factors need to be considered: First, the dental arch curve of the oral cavity to observe whether the teeth are aligned properly; second, the positional relationship between the tooth root and the alveolar bone to ensure that the external force applied to the tooth does not exceed its growth limit; and third, the occlusion between the upper and lower rows of teeth, mainly observing the crown portion of the teeth to ensure that the upper and lower rows of teeth can occlude well after correction.

[0026] For orthodontic treatment, data from both CBCT images and intraoral scan models are required. However, scanning and combining these two types of data is cumbersome and time-consuming, and due to limitations in equipment and patient conditions, it is often impossible to obtain paired data from both modalities for the same patient.

[0027] To address this issue, this application focuses on this problem and proposes a curvature-enhanced hidden function network based on artificial intelligence methods. This network generates high-quality tooth models that contain both root information and fine-grained crown details from widely used CBCT images. This is crucial for improving the efficiency of orthodontic treatment.

[0028] like Figure 1 The diagram illustrates a flowchart of a tooth model reconstruction method based on curvature enhancement according to an embodiment of the present invention. As shown, the method includes:

[0029] Step S101: Acquire CBCT images and intraoral scan data of a patient at the same time period.

[0030] In the field of digital dentistry, CBCT images are an important type of 3D imaging data with the most widespread clinical applications. Digital dental systems require the reconstruction of 3D models of patients' teeth and bones to enable rapid design of orthodontic treatment plans directly within the system. Therefore, tooth recognition and instance segmentation based on CBCT images has always been a fundamental step and a hot research topic in the diagnosis and treatment of digital orthodontics. Many traditional algorithms have been explored by designing handcrafted features for tooth segmentation from background tissue. Currently, researchers have designed various emerging network models based on deep learning methods in artificial intelligence in recent years, and have achieved stable and excellent experimental results. In computer-aided technology, based on tooth segmentation from CBCT, 3D models of teeth can be easily generated using traditional computer graphics methods. However, due to limitations in imaging technology and radiation exposure, the spatial resolution of CBCT images is relatively low, with common voxel sizes between 0.2-1.0 mm. While this is sufficient to observe the complete shape and root features of teeth, it is still insufficient to observe the fine geometric details of the crown surface, making it difficult to use for judging occlusion. Therefore, this application requires the design of new methods to obtain high-quality tooth models.

[0031] Because CBCT data has a low resolution, the segmented crown portion cannot achieve the precision of an intraoral scanning model, making it difficult to determine the occlusion of the upper and lower teeth. Therefore, in orthodontics, doctors typically combine CBCT images with another modality of data—an intraoral scanning model—to observe and analyze the occlusal relationship of the patient's upper and lower jaws.

[0032] Intraoral scanners are a new type of invasive optical scanning head that scans objects using projected laser or structured light sources. They can directly acquire high-precision 3D shape and texture information of soft and hard tissues such as teeth, gums, and mucosa within the patient's oral cavity, and then generate 3D models through software processing. Compared to traditional plaster models and impression techniques, this direct digital scanning eliminates many cumbersome traditional steps, reduces labor costs and material consumption, significantly simplifies clinical procedures, and improves efficiency. It also enables computer-aided model analysis, restoration fabrication, and treatment design. In recent years, intraoral scanners have gained wider adoption due to their efficiency and convenience, and significant progress has been made in the segmentation of dental crowns based on intraoral scan models.

[0033] CBCT images have been widely used for both types of data over the years. They can acquire information on both the tooth root and the crown at the same time. However, their resolution is low, making it difficult to observe the details of the crown to determine the occlusal relationship. Therefore, in clinical practice, they can be combined with the detailed crown obtained from the intraoral scanning model as a supplement.

[0034] It should be noted that, because the two types of data were collected from devices from different manufacturers, spatial misalignment may occur when combining them. Therefore, this application requires that these two pairs of data were collected from the same patient within the same time period.

[0035] like Figure 2 The figure illustrates a scenario of tooth model reconstruction based on curvature enhancement in one embodiment of this application. The figure shows the objective of this application: generating high-quality tooth models solely from CBCT images. The solution of this application is to use a high-precision crown model obtained from an intraoral scan to guide the network in generating a high-quality tooth model from CBCT images. This allows for the direct design of tooth alignment schemes for each patient, leading to specific orthodontic plans and appliance design.

[0036] Step S102: Label the tooth model in the CBCT image and the crown model in the intraoral scan data respectively, and register the crown model in the intraoral scan data to the tooth model in the CBCT image using the rigid ICP algorithm.

[0037] Since the CBCT images and intraoral scan data were collected from the same patient at the same time, the teeth maintained a rigid state. Therefore, this application uses the rigid ICP algorithm for registration, registering the crown model from the intraoral scan data onto the tooth model from the CBCT.

[0038] like Figure 3 The figure shows a schematic diagram illustrating the registration of a crown model and a tooth model in one embodiment of this application. As shown, this application combines CBCT images and intraoral scan data to construct a high-precision tooth model to supervise network training. The entire construction process consists of three steps.

[0039] First, for paired CBCT images and intraoral scan models, the tooth model is labeled on the CBCT image, and the crown model is labeled on the intraoral scan model.

[0040] Then, the rigid ICP algorithm is applied to register the crown model onto the tooth model, thereby aligning the two models.

[0041] Finally, the crown portion of the tooth model obtained from the CBCT images is removed, and the remaining root portion is merged with the crown model extracted from the intraoral scan using the Poisson surface reconstruction algorithm. In this way, the final tooth model contains both root information and high-quality crown details, which can be used to train a supervised network.

[0042] In one embodiment of this application, the registration of the crown model from intraoral scan data to the tooth model from CBCT images using the rigid ICP algorithm includes:

[0043] A. Estimate the corresponding points in the crown model and the tooth model, and perform iterative calculations based on the least squares method;

[0044] B. Calculate the distance between the vertices of the crown model and the tooth model, remove the crown portion of the tooth model by removing points with a distance less than a preset value, and replace it with the crown model of the intraoral scan data;

[0045] C. The Poisson surface reconstruction algorithm is used to fill the lateral cracks of the registered tooth model, so as to fill the missing parts of the surface of the registered tooth model and form a closed surface.

[0046] In this embodiment, ICP registration first estimates corresponding points in the two models, and then performs iterative calculations based on the least squares method. After registration, the crown portion of the CBCT tooth model needs to be replaced with the crown model of the intraoral scan model. Specifically, this is done by calculating the distance between the vertices of the two models and removing points that are closer together, thus removing the crown portion of the CBCT tooth model. This results in a crack on the side of the tooth, so this application finally uses the Poisson surface reconstruction algorithm to fill this crack. Poisson surface reconstruction is a reconstruction method based on implicit functions, which transforms discrete sample points of a surface into a continuously integrable surface function, thereby filling in the missing parts of the surface and forming a closed surface. In this way, the high-quality tooth model required by this application is finally synthesized.

[0047] Step S103: Based on the curvature-enhanced hidden function network, the three-dimensional shape of the crown in the tooth model after fine registration is obtained to obtain a high-quality tooth model.

[0048] To effectively represent the intricate 3D shapes in high-quality dental models, this application employs implicit functions as the representation method. Defined in a continuous space, it can represent 3D shapes at arbitrary resolutions and can be used to learn intricate surfaces with different topologies. It primarily determines whether an arbitrary 3D coordinate point in space is inside or outside the surface by performing binary classification of the coordinates, implicitly representing the surface. By taking enough points in the continuous space and performing this binary classification, boundaries can be gradually generated in space, thus implicitly forming the surface. Therefore, implicit function networks have achieved excellent performance in shape recovery, shape completion, and super-resolution reconstruction. One pioneering classic work involves learning a signed distance field (SDF) through a deep neural network, which constrains the generated geometry by measuring the distance from points to the shape surface. Researchers subsequently proposed an improved implicit function network to predict continuous outputs of multi-scale features. Although these methods have demonstrated excellent performance on synthetic 3D datasets, reconstructing intricate and complex geometries in real-world data remains very challenging, especially for the crown portion of our high-quality dental model.

[0049] In one embodiment of this application, the curvature-enhanced implicit function network includes:

[0050] The instance segmentation module is used to detect and segment tooth models for each tooth from CBCT images;

[0051] The surface reconstruction module is used to reconstruct tooth models based on implicit function representations.

[0052] The curvature enhancement module is used to enhance the learning of local geometric features of the crown portion.

[0053] like Figure 4 The figure shows a schematic diagram of a model based on a curvature-enhanced implicit function network in one embodiment of this application. As shown, the network structure mainly consists of three parts: 1) an instance segmentation module, used to detect and segment each tooth from CBCT images; 2) a surface reconstruction module, which generates a high-quality tooth model based on the implicit function representation; and 3) a curvature enhancement module, which enhances the network's learning of local geometric details of the crown. Solid arrows in the figure represent the training and testing processes, while dashed arrows are only used for training. After training the network with sufficient paired data, in the testing or usage phase, we only need to input CBCT images to reconstruct a high-quality tooth model.

[0054] Specifically, the goal of this application is to generate high-quality tooth models solely from CBCT images, and the network structure designed for this purpose is mainly divided into three modules.

[0055] The first part is the instance segmentation module. First, the CBCT image is segmented into instances of teeth. Specifically, this application first detects the position of each tooth, and then segments each tooth.

[0056] The second part is the surface reconstruction module, which is inspired by implicit function networks.

[0057] In one embodiment of this application, the implicit function-based representation for reconstructing a tooth model includes:

[0058] A. Stack the tooth model of each tooth obtained by the instance segmentation module and the tooth model of the labeled CBCT image together, and input them into the encoder to obtain a multi-scale feature map.

[0059] B. Obtain multiple access points and sample them on the feature map to obtain multi-scale features for each access point;

[0060] C. Input the multi-scale features of each access point into a multilayer perceptron, which serves as a classifier, to obtain a binary classification of each access point on and off the surface, which is then used as a latent function representation of the three-dimensional shape.

[0061] Specifically, this application stacks the images of each tooth obtained by the instance segmentation module with the tooth models from the labeled CBCT images, and inputs them into the encoder to obtain multi-scale feature maps. Then, some access points are selected in continuous space, and these points are sampled on the obtained feature maps to obtain multi-scale features for each point. Finally, the multi-scale features of each point are input into a multilayer perceptron, which acts as a classifier, to obtain a binary classification of each point as inside and outside the surface, which serves as a latent function representation of the three-dimensional shape.

[0062] In one embodiment of this application, the method further includes:

[0063] A. During the training phase, access points are sampled near the surface of the teeth using a Gaussian distribution, and / or during the testing phase, multiple dense points are uniformly acquired in space as access points.

[0064] B. The space is divided into discrete three-dimensional voxels by using the Marching Cubes algorithm. Isosurfaces are extracted from each three-dimensional voxel using triangular facets to obtain a complete explicit surface representation.

[0065] In terms of sampling access points, this application samples access points near the surface of the tooth using a Gaussian distribution during the training phase, aiming to distribute the points on the tooth surface, especially near the crown surface. During the testing phase, a number of dense points are uniformly selected in space, and then the Marching Cubes algorithm is used to obtain an explicit surface, which is the high-quality tooth model that this application aims to obtain.

[0066] The Marching Cubes algorithm described in this application is a 3D surface rendering algorithm that divides space into discrete 3D voxels. Within each voxel, isosurfaces are extracted using triangular facets, ultimately yielding a complete surface representation. Using these two modules, we can obtain preliminary reconstruction results of the tooth model.

[0067] In experimental studies, this application found that the surface of the tooth crown has very rich geometric details. If only a binary classification method of surface inside and outside is used for learning, the resulting tooth model still lacks some geometric features on the curved surface of the crown, such as local shape variations. Therefore, this application designed a third part of the network, the curvature enhancement module.

[0068] The third part is the curvature enhancement module. Curvature is a very important geometric feature on a three-dimensional surface, which can represent the degree of bending and shape changes in a local area. The crown, which is of most interest in this application, has a very rich variety of such shape changes. Therefore, we want to design a network specifically for this purpose, so that it can effectively learn the curvature feature during training.

[0069] In one embodiment of this application, the learning of the local geometric features of the enhanced crown portion includes:

[0070] A. For each vertex of the registered tooth model, use KNN to find its nearest access points;

[0071] B. Perform a weighted average based on the multi-scale features of each access point to obtain the features of each vertex;

[0072] C. Input the features of each vertex into a multilayer perceptron, which acts as a regressor, to predict the curvature value of each vertex.

[0073] Specifically, for each vertex in the real model, this application uses KNN to find its nearest access points and performs a weighted average based on their multi-scale features to obtain the vertex's features. This feature is then input into a multilayer perceptron, which acts as a regressor, to predict the curvature value of each vertex. This application aims to enhance the training of surface reconstruction through this module, enabling the network to learn finer local changes and geometric details on the crown, thus resulting in a reconstructed tooth model with more refined crown details.

[0074] like Figure 5 The figure shows a comparison of different reconstruction results. Taking three teeth as an example, the reconstruction results are illustrated. The first column shows the manual annotations based on CBCT images, the second column shows the tooth segmentation results based on CBCT images, the third and fourth columns show the surface reconstruction results of other hidden function networks, the fifth column shows the reconstruction results of the method designed in this application, and the last column shows a high-quality tooth model combining CBCT images and intraoral scan models. It can be seen from the figure that CBCT-based tooth annotation can generate basic tooth shapes, but due to the limited resolution of CBCT images, many details on the crown cannot be observed or accurately annotated. For the implicit function-based surface reconstruction method, it can be observed that the tooth surface generated by OccNet based on coordinate learning is too smooth, and can only reconstruct the overall approximate shape; IFNet uses multi-scale feature extraction, which can generate a more fine-grained surface; however, due to the large domain gap between the CBCT image and the tooth model, there is a lot of noise in the reconstruction results.

[0075] It is worth noting that the reconstruction results of the method designed in this application are more consistent with high-quality tooth models. Both the overall shape and the local geometric details of the crown surface can be reconstructed well, demonstrating the effectiveness of the reconstruction method designed in this application.

[0076] Overall, this application proposes a curvature-enhanced hidden function network for generating high-quality tooth models from CBCT images. Only CBCT images need to be input into the trained network to obtain high-quality tooth models for each tooth. This not only improves the efficiency of orthodontic treatment but also can be applied to digital orthodontic systems to quickly design tooth alignment schemes for each patient and subsequently arrange specific orthodontic treatment plans.

[0077] like Figure 6 The figure shows a schematic diagram of a tooth model reconstruction device based on curvature enhancement according to an embodiment of the present invention. As shown, the tooth model reconstruction device 600 based on curvature enhancement includes:

[0078] The acquisition module 601 is used to acquire CBCT images and intraoral scan data of a patient at the same time period;

[0079] The registration module 602 is used to annotate the tooth model of the CBCT image and the crown model of the intraoral scan data respectively, and to register the crown model of the intraoral scan data to the tooth model of the CBCT image through the rigid ICP algorithm.

[0080] The optimization module 603 is used to determine the three-dimensional shape of the crown in the tooth model after fine registration based on the curvature-enhanced implicit function network, so as to obtain a high-quality tooth model.

[0081] It is understandable that the device 600, through the operation of its various modules, can achieve the following: Figure 1 The tooth model reconstruction method based on curvature enhancement is described above.

[0082] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented by processing element calls to software, while others are implemented in hardware. For example, optimization module 603 can be a separate processing element, or it can be integrated into a chip in the above device. Alternatively, it can be stored as program code in the memory of the above device, and its function can be called and executed by a processing element of the device. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element mentioned here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed through integrated logic circuits in the hardware of the processor element or through software instructions.

[0083] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more digital signal processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together to form a system-on-a-chip (SOC).

[0084] like Figure 7 The figure shows a schematic diagram of a computer device according to an embodiment of the present invention. As shown, the computer device 700 includes a memory 701 and a processor 702. The memory 701 stores a computer program, and the processor 702 executes the computer program to implement the following... Figure 1 The tooth model reconstruction method based on curvature enhancement is described above.

[0085] The memory 701 may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0086] The processor 702 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0087] In one embodiment of the present invention, a computer-readable storage medium stores a data transmission program thereon, which, when executed by a processor, implements as follows: Figure 1 The tooth model reconstruction method based on curvature enhancement is described above.

[0088] As will be understood by those skilled in the art, all or part of the steps of the above-described method embodiments can be implemented using computer program-related hardware. The aforementioned image processing program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0089] These computer programs can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0090] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0091] In summary, the present invention provides a method, apparatus, device, and medium for reconstructing a tooth model based on curvature enhancement, which can effectively overcome the various shortcomings of the prior art and has high industrial application value.

[0092] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for reconstructing a tooth model based on curvature enhancement, characterized in that, The method includes: Acquire CBCT images and intraoral scan data of a patient at the same time period; The tooth model in the CBCT image and the crown model in the intraoral scan data are labeled respectively. The crown model in the intraoral scan data is registered to the tooth model in the CBCT image using the rigid ICP algorithm. The three-dimensional shape of the crown in the tooth model after fine registration based on the curvature-enhanced hidden function network is used to obtain a high-quality tooth model. The curvature-enhanced implicit function-based network includes: The instance segmentation module is used to detect and segment tooth models for each tooth from CBCT images; The surface reconstruction module is used to reconstruct tooth models based on implicit function representations. The curvature enhancement module is used to enhance the learning of local geometric features of the crown portion; The implicit function-based representation for reconstructing the tooth model includes: The tooth model of each tooth obtained by the instance segmentation module and the tooth model of the labeled CBCT image are stacked together and input into the encoder to obtain a multi-scale feature map. Multiple access points are acquired and sampled on the feature map to obtain multi-scale features for each access point; The multi-scale features of each access point are input into a multilayer perceptron, which acts as a classifier, to obtain a binary classification of each access point on and off the surface, which is then used as a latent function representation of the three-dimensional shape.

2. The tooth model reconstruction method based on curvature enhancement according to claim 1, characterized in that, The process of registering the crown model from intraoral scan data to the tooth model from CBCT images using the rigid ICP algorithm includes: The corresponding points in the crown model and the tooth model are estimated, and iterative calculations are performed based on the least squares method; Calculate the distance between the vertices of the crown model and the tooth model, remove the crown portion of the tooth model by removing points whose distance is less than a preset value, and replace it with the crown model of the intraoral scan data; The Poisson surface reconstruction algorithm is used to fill the lateral cracks of the registered tooth model, so as to fill the missing parts of the surface of the registered tooth model and form a closed surface.

3. The tooth model reconstruction method based on curvature enhancement according to claim 1, characterized in that, The method further includes: During the training phase, access points are sampled near the surface of the teeth using a Gaussian distribution, and / or during the testing phase, multiple dense points are uniformly acquired in space as access points. The Marching Cubes algorithm is used to divide the space into discrete three-dimensional voxels. In each three-dimensional voxel, isosurfaces are extracted using triangular patches to obtain a complete explicit surface representation.

4. The tooth model reconstruction method based on curvature enhancement according to claim 1, characterized in that, The learning of the local geometric features of the enhanced crown portion includes: For each vertex of the registered tooth model, use KNN to find the nearest access points; A weighted average is performed based on the multi-scale features of each access point to obtain the features of each vertex; The features of each vertex are input into a multilayer perceptron, which acts as a regressor, to predict the curvature value of each vertex.

5. A tooth model reconstruction device based on curvature enhancement, characterized in that, The device includes: The acquisition module is used to acquire CBCT images and intraoral scan data of a patient at the same time period; The registration module is used to annotate the tooth model in the CBCT image and the crown model in the intraoral scan data respectively. The crown model in the intraoral scan data is registered to the tooth model in the CBCT image using the rigid ICP algorithm. The optimization module is used to determine the 3D shape of the crown in the tooth model after fine registration based on the curvature-enhanced hidden function network, so as to obtain a high-quality tooth model. The curvature-enhanced implicit function-based network includes: The instance segmentation module is used to detect and segment tooth models for each tooth from CBCT images; The surface reconstruction module is used to reconstruct tooth models based on implicit function representations. The curvature enhancement module is used to enhance the learning of local geometric features of the crown portion; The implicit function-based representation for reconstructing the tooth model includes: The tooth model of each tooth obtained by the instance segmentation module and the tooth model of the labeled CBCT image are stacked together and input into the encoder to obtain a multi-scale feature map. Multiple access points are acquired and sampled on the feature map to obtain multi-scale features for each access point; The multi-scale features of each access point are input into a multilayer perceptron, which acts as a classifier, to obtain a binary classification of each access point on and off the surface, which is then used as a latent function representation of the three-dimensional shape.

6. A computer device, characterized in that, The device includes: a memory and a processor; the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1-4.

Images