Oral cavity posterior mold tooth loss repairing method and system based on metal framework and deep learning

Abstract

The invention relates to the technical field of tooth restoration, in particular to an oral cavity posterior mold tooth loss restoration method and system based on a metal framework and deep learning, and the method comprises the following steps: firstly, designing a metal framework for fixing a restoration body; acquiring three-dimensional point cloud data of the oral cavity of the patient by using oral cavity scanning equipment; carrying out three-dimensional reconstruction on the tooth at the missing part by utilizing a self-attention implicit function network; and finally, customizing a dental restoration according to the reconstructed three-dimensional tooth model, and mounting the dental restoration on a metal frame. The metal frame is used for fixing the prosthesis, damage to mold teeth on the two sides of the missing tooth position is avoided, and damage to gingiva of the missing tooth position is avoided. The missing teeth are accurately complemented through the self-attention implicit function network, and the aesthetic property and comfort of the prosthesis are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tooth restoration, and in particular to a method and system for restoring missing teeth in the posterior oral cavity based on a metal frame and deep learning. Background Art

[0002] In life, we often encounter the situation where teeth in the mouth are pulled out due to disease, while the teeth on both sides of the missing tooth are intact. After the tooth is missing, if the missing part is not filled, the roots of the healthy teeth on both sides will collapse to the missing part, causing the roots of the intact teeth on both sides to loosen and tilt, affecting the lifespan of the healthy teeth on both sides.

[0003] For the restoration of the second and third model teeth in the oral arrangement, the settings are adjusted through manual experience to fill the missing teeth. If the design error is large, not only will the adjacent teeth be damaged, but the gums in the missing tooth position will also be damaged; and the restoration effect often fails to meet user satisfaction. Summary of the invention

[0004] The present invention aims to solve the problem of repairing the second and third mold teeth of the upper and lower jaws, and proposes a method and system for repairing missing posterior mold teeth in the oral cavity based on a metal frame and deep learning, which avoids damage to adjacent teeth and damage to the gums at the current missing tooth position, and improves the aesthetics and comfort of the restoration.

[0005] To achieve the above purpose, the technical solution adopted is:

[0006] The present invention provides a method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning, comprising:

[0007] Design a metal framework to hold the restoration;

[0008] Use oral scanning equipment to obtain three-dimensional point cloud data of the patient's oral cavity;

[0009] The self-attention implicit function network is used to reconstruct the missing teeth in three dimensions;

[0010] A dental restoration is custom made based on the reconstructed 3D tooth model and mounted on a metal framework.

[0011] According to the oral posterior model missing tooth repair method based on metal frame and deep learning of the present invention, further, according to the location of the missing teeth, an adaptive metal frame is designed, which fits tightly to the healthy teeth on both sides and is connected at the line part of the center points of the teeth to form a clasp-like structure.

[0012] According to the method for restoring missing teeth in the posterior oral cavity based on a metal frame and deep learning of the present invention, further, the process of three-dimensionally reconstructing the teeth at the missing part using a self-attention implicit function network is as follows:

[0013] A deep learning model is used for feature extraction, feature fusion and prediction, and the self-attention mechanism is applied in the deep learning model. The input of the model is the three-dimensional point cloud data of the patient's mouth, and the output is the three-dimensional point cloud data of the missing teeth. A loss function is constructed to train the deep learning model.

[0014] According to the method for restoring missing teeth in the posterior oral mold based on a metal frame and deep learning of the present invention, further, by calculating the weight value of each point in the three-dimensional point cloud of the missing teeth after sampling, the weight training model is added to the loss function, specifically:

[0015] First, the normal vectors of all points in the three-dimensional point cloud of the missing teeth are calculated, and the distribution of the normal vectors in the point cloud is analyzed using statistical methods. Then, the deviation between the normal vector of the target point and the overall distribution is calculated to obtain the weight value Z1 of the normal vector of the current point.

[0016] Use the LOF algorithm to calculate the outlier value of each point, and select the point with the largest outlier value as the sampling point of the point cloud;

[0017] The density value Z2 of the current point is calculated using a neighborhood-based method, and the weight factor is constructed using the weight value Z1 of the normal vector of the current point and the density value Z2;

[0018] The eigenvalues ​​of the three-dimensional covariance matrix of the point cloud are calculated using the weight factor of each point after sampling, and the eigenvalues ​​λ1, λ2, and λ3 in three directions are obtained. The average of the eigenvalues ​​in the three directions is calculated to obtain λ, which is the complexity factor of the point cloud object.

[0019] The complexity factor λ is introduced into the loss function to train the model.

[0020] According to the method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning of the present invention, further, the process of calculating the outlier value of each point using the LOF algorithm is as follows:

[0021] For each point, determine its k nearest neighbors and calculate its minimum distance to the k nearest neighbors;

[0022] Calculate the local reachability distance of each point;

[0023] Calculate the local reachability density of each point. The local reachability density is the inverse of the local reachability distance.

[0024] The LOF score is then calculated based on the local reachable density, and then all points are sorted, and points with a score greater than the set LOF score threshold are selected as the points for this point cloud sampling.

[0025] According to the method for restoring missing teeth in the posterior oral cavity based on a metal frame and deep learning of the present invention, further, the local reachable distance calculation formula of a point p in the point cloud is:

[0026] reach-dist' k (p, o)=B·max{dist(p,o)·Z1,k-dist(o)}

[0027] Among them, reach-dist' k (p, o) is the local reachable distance of point p; Z1 is the weight value of the normal vector of point p; B is the adjustment coefficient; max{·} is the maximum value function; dist(p, o) is the Euclidean distance between point p and point o; k-dist(o) is the minimum distance from point o to its k nearest neighbors.

[0028] According to the oral posterior mold missing tooth restoration method based on metal framework and deep learning of the present invention, further, the weight factor Z0 of each point is expressed as follows:

[0029]

[0030] in, is the weight coefficient of Z1, is the weight parameter of Z2.

[0031] According to the method for restoring missing teeth in the posterior oral mold based on a metal framework and deep learning of the present invention, the complexity factor λ is further introduced into the loss function, and the loss function calculation formula is:

[0032]

[0033] Among them, L is the loss function, N is the number of samples, and y i is the true label, is the predicted probability; A is the adjustment coefficient.

[0034] According to the method for restoring missing teeth in the posterior oral cavity based on a metal frame and deep learning of the present invention, the method further includes: using a three-dimensional geometric matching algorithm to evaluate the matching degree between the restoration and the surrounding teeth, and fine-tuning the restoration according to the matching results.

[0035] Furthermore, the present invention also provides a system for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning, comprising:

[0036] The Metal Framework Fabrication module is used to design a metal framework for fixing the restoration;

[0037] A scanning module, used to obtain three-dimensional point cloud data of the patient's oral cavity using an oral scanning device;

[0038] Missing teeth reconstruction module, which is used to reconstruct the missing teeth in three dimensions using a self-attention implicit function network;

[0039] The mounting module is used to customize the dental restoration according to the reconstructed 3D tooth model and mount it on the metal framework.

[0040] The beneficial effects achieved by adopting the above technical solution are:

[0041] 1. The present invention uses a self-attention implicit function network to accurately complete missing teeth, which can capture the non-local features of the tooth model while retaining the geometric details of the shape. The restored teeth are difficult to distinguish from natural teeth in appearance, which improves the aesthetics and comfort of the restoration and ensures that the restoration effect meets the needs of patients.

[0042] 2. The present invention uses metal to embrace the mold teeth on both sides of the missing tooth position to achieve fixation during tooth restoration of the missing tooth position, thereby avoiding damage to the mold teeth on both sides of the missing tooth position and the gums of the missing tooth position, thereby avoiding the occurrence of gum diseases to a certain extent; and the metal frame itself has long-term stability, thereby extending the service life of the restoration.

[0043] 3. Combined with the weight value of the normal vector of the point, the LOF score of the point is further calculated, and sampling points are selected according to the score, so that the sampled points are points with higher eigenvalues, which can reduce the amount of calculation.

[0044] 4. The feature information of the object point cloud is extracted through calculation, and the complexity factor of the point cloud object is calculated. The complexity factor is used to adjust the loss function. The complexity factor can help the model adjust the learning strategy when facing objects of different complexities and improve the generalization ability of the model. In addition, the introduction of the complexity factor enables the loss function to distinguish the importance of different samples, so that the model can focus more computing resources on samples with more complex geometric features and richer information, rather than distributing them evenly to all samples, thereby improving training efficiency and speeding up model training. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings of the embodiments of the present invention, wherein the drawings are only used to illustrate some embodiments of the present invention, but not to limit all embodiments of the present invention thereto.

[0046] Figure 1 It is a flow chart of a method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning according to an embodiment of the present invention;

[0047] Figure 2 is a schematic structural diagram of a metal frame according to an embodiment of the present invention;

[0048] Figure 3 Schematic diagram of missing tooth completion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings of specific embodiments of the present invention to clearly and completely describe the exemplary scheme of the embodiment of the present invention. Unless otherwise defined, the technical terms or scientific terms used in the present invention should be the common meanings understood by people with ordinary skills in the field.

[0050] This embodiment discloses a method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning. Figure 1 As shown, the following steps are included:

[0051] Step S1, designing a metal frame for fixing the restoration.

[0052] According to the location of the missing teeth, an adaptive metal frame is manually designed. The frame can fit closely to the healthy teeth on both sides and connect at the line connecting the center points of the teeth to form a clasp-like structure, such as Figure 2 to secure the restoration.

[0053] First, the stress distribution calculation is performed: the finite element analysis software is used to simulate the stress distribution of the metal frame when it is subjected to force. Then, the deformation calculation is performed, and vectors are introduced into the triangular unit to calculate the deformation of the metal frame. The metal frame design is optimized through stress distribution and deformation calculation to ensure the stability and durability of the restoration.

[0054] Step S2: using an oral scanning device to obtain three-dimensional point cloud data of the patient's oral cavity.

[0055] Use high-precision oral scanning equipment to obtain the three-dimensional point cloud data of the patient's oral cavity. Import the scanned dental model data into the restoration system to provide accurate tooth and gum structure information for subsequent restoration design, ensuring the accuracy of the restoration design.

[0056] High-precision oral scanning equipment: Use oral scanning equipment with sub-millimeter accuracy to obtain three-dimensional point cloud data of the patient's oral cavity.

[0057] Data import and preprocessing: Import the scanned dental model data into the restoration system, and use CAD / CAM software to preprocess the data, including denoising, smoothing and enhancement, to ensure the accuracy and usability of the data.

[0058] Step S3: Use the self-attention implicit function network to perform three-dimensional reconstruction of the missing teeth.

[0059] Use deep learning models for feature extraction, feature fusion, and prediction, such as Resnet networks or 2D / 3D Transformers, to improve the performance of voxel 3D reconstruction methods when there are many input images. And apply the self-attention mechanism in the deep learning model. This application is based on the observation that learning how to predict implicit values ​​is similar to calculating the global response of an image pixel position when a query point is given. The input of the deep learning model is the 3D point cloud data of the patient's mouth, and the output is the 3D point cloud data of the missing teeth. Finally, a loss function is constructed to train the deep learning model.

[0060] After obtaining the 3D point cloud data of the missing teeth, the point cloud data is filtered (to remove outliers and noise anomalies), and sampling is performed during the model training process to reduce the amount of calculation; the weight value of each point is calculated after sampling, and the weight is added to the loss function to improve the generalization ability of the model and improve the training efficiency. The calculation process is as follows:

[0061] Step S301, first calculate the normal vectors of all points in the three-dimensional point cloud of the missing teeth, use statistical methods (such as covariance matrix) to analyze the distribution of normal vectors in the point cloud, and then calculate the deviation between the normal vector of the target point and the overall distribution to obtain the weight value Z1 of the normal vector of the current point.

[0062] Step S302, use the LOF algorithm to calculate the outlier value of each point. Since the point cloud data has been filtered and the outliers and noise outliers have been removed, when the LOF algorithm is used to calculate the outlier value of each point, the point with a larger outlier value is the point with a larger feature in the point cloud, and can further better reflect the feature information of the point cloud.

[0063] A. First, select an appropriate number of neighbors k and calculate the three-dimensional space distance between each point and all other points; for each point, determine its k nearest neighbors and calculate its minimum distance to the k nearest neighbors, where the value of k can be adjusted and set according to the actual situation.

[0064] B. Calculate the local reachable distance of each point. The calculation formula for the local reachable distance of point p in the point cloud is:

[0065] reach-dist' k (p, o)=B·max{dist(p,o)·Z1,k-dist(o)}

[0066] Among them, reach-dist' k(p, o) is the local reachable distance of point p; Z1 is the weight value of the normal vector of point p; B is the adjustment coefficient; max{·} is the maximum value function; dist(p, o) is the Euclidean distance between point p and point o; k-dist(o) is the minimum distance from point o to its k nearest neighbors.

[0067] Here, in the process of calculating the local reachable distance, the weight value Z1 of the point's normal vector is added. By combining the normal vector weight Z1, the LOF algorithm can more accurately capture the geometric and structural features in the point cloud, thereby improving the accuracy of anomaly detection; enhancing the ability to distinguish different indicators, assigning different weights to variables of different dimensions in the LOF algorithm, and strengthening the ability of different indicators to distinguish anomalies, so that the algorithm can better identify truly important anomalies; and can more accurately identify abnormal feature points.

[0068] C. Calculate the local reachability density of each point. The local reachability density is the inverse of the local reachability distance.

[0069] D. Calculate the LOF score based on the local reachable density, then sort all the points, set the LOF score threshold, and select the points greater than the set LOF score threshold as the points for this point cloud sampling. The threshold setting can be adjusted according to the actual situation on site.

[0070] Step S303: Calculate the density value Z2 of the current point using a neighborhood-based method, and use a KD tree or other search structure to perform a neighborhood search to find all neighboring points within a certain radius around each point. The number of points in the neighborhood of the current point is the density value Z2 of the current point, and the expression of the weight factor of each point is:

[0071]

[0072] in, is the weight parameter of Z1, is the weight parameter of Z2.

[0073] Step S304, using the geometric features of the point cloud to quantify the complexity, using the weight factor of each point after sampling to calculate the eigenvalue of the "-dimensional covariance matrix of the point cloud, obtaining the eigenvalues ​​λ1, λ2, λ3 in three directions, and calculating the average of the eigenvalues ​​in the three directions to obtain λ, λ is the complexity factor of the point cloud object, and the complexity factor represents the complexity of the current point cloud object.

[0074] The self-attention implicit function network performs 3D reconstruction of the missing teeth. The network can effectively capture the non-local features of the tooth model and maintain the geometric details of the shape, thereby achieving accurate completion of the missing teeth. Figure 3 shown.

[0075] Among them, in the loss function in the calculation logic, the binary cross entropy loss function is used to train the network and optimize the parameters of the implicit function network. The calculation formula of the loss function is as follows:

[0076]

[0077] Among them, L is the loss function, N is the number of samples, yi is the true label, is the predicted probability; A is the adjustment coefficient.

[0078] The complexity factor is added to the loss function as a function of the eigenvalue to guide the model training process. It can help the model adjust its learning strategy and improve the generalization ability of the model when facing objects of different complexity.

[0079] Step S4: Customize a tooth restoration according to the reconstructed three-dimensional tooth model and install it on the metal framework.

[0080] According to the reconstructed 3D tooth model, the appropriate tooth material and model are selected to customize the tooth restoration. The customized tooth restoration is tightly combined with the metal framework (which can be fixed by adhesive) to ensure the stability and functionality of the restoration. The restoration is tested for mechanical properties, including flexural strength and wear resistance, to ensure its long-term stability.

[0081] Step S5: Use a three-dimensional geometric matching algorithm to evaluate the matching degree between the restoration and the surrounding teeth, and fine-tune the restoration according to the matching results to ensure the precise fit of the restoration.

[0082] First, by automatically identifying the characteristic points of teeth, key information can be provided for subsequent geometric matching. This can be achieved by combining the DBSCAN and K-Means hybrid clustering algorithms, which can accurately detect the characteristic points of teeth. Then, the deep learning framework is used to segment the three-dimensional point cloud of teeth from the three-dimensional point cloud data of the oral cavity and three-dimensionally model them. The three-dimensional geometric matching algorithm is used to evaluate the matching degree between the restoration and the surrounding teeth. According to the matching results, the restoration is fine-tuned to improve the comfort and aesthetics of the restoration.

[0083] Corresponding to the above method, this embodiment also discloses a system for restoring missing teeth in the posterior oral cavity based on a metal frame and deep learning, comprising:

[0084] The Metal Framework Fabrication module is used to design a metal framework for fixing a restoration.

[0085] The scanning module is used to obtain three-dimensional point cloud data of the patient's oral cavity using an oral scanning device.

[0086] The missing teeth reconstruction module is used to reconstruct the missing teeth in three dimensions using a self-attention implicit function network.

[0087] The mounting module is used to customize the dental restoration according to the reconstructed 3D tooth model and mount it on the metal framework.

[0088] Unless otherwise specifically stated, the components, steps, numerical expressions and values ​​set forth in these embodiments do not limit the scope of the present invention.

[0089] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0090] The units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person of ordinary skill in the art may use different methods to implement the described functions for each specific application, but such implementation is not considered to be beyond the scope of the present invention.

[0091] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits, and accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present invention is not limited to any specific form of combination of hardware and software.

[0092] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning, characterized in that: include: Design a metal framework to hold the restoration; Use oral scanning equipment to obtain three-dimensional point cloud data of the patient's oral cavity; The self-attention implicit function network is used to reconstruct the missing teeth in three dimensions; A dental restoration is custom made based on the reconstructed 3D tooth model and mounted on a metal framework.

2. The method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning according to claim 1, characterized in that: According to the location of the missing teeth, an adaptive metal frame is designed, which fits tightly to the healthy teeth on both sides and is connected at the line connecting the center points of the teeth to form a clasp-like structure.

3. The method for restoring missing teeth in the posterior oral cavity based on metal framework and deep learning according to claim 1, characterized in that: The process of three-dimensional reconstruction of missing teeth using the self-attention implicit function network is as follows: A deep learning model is used for feature extraction, feature fusion and prediction, and the self-attention mechanism is applied in the deep learning model. The input of the model is the three-dimensional point cloud data of the patient's mouth, and the output is the three-dimensional point cloud data of the missing teeth. A loss function is constructed to train the deep learning model.

4. The method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning according to claim 3, characterized in that: By calculating the weight value of each point in the three-dimensional point cloud of the missing teeth after sampling, the weight training model is added to the loss function, specifically: First, the normal vectors of all points in the three-dimensional point cloud of the missing teeth are calculated, and the distribution of the normal vectors in the point cloud is analyzed using statistical methods. Then, the deviation between the normal vector of the target point and the overall distribution is calculated to obtain the weight value Z1 of the normal vector of the current point. Use the LOF algorithm to calculate the outlier value of each point, and select the point with the largest outlier value as the sampling point of the point cloud; The density value Z2 of the current point is calculated using a neighborhood-based method, and the weight factor is constructed using the weight value Z1 of the normal vector of the current point and the density value Z2; The eigenvalues ​​of the three-dimensional covariance matrix of the point cloud are calculated using the weight factor of each point after sampling, and the eigenvalues ​​λ1, λ2, and λ3 in three directions are obtained. The average of the eigenvalues ​​in the three directions is calculated to obtain λ, which is the complexity factor of the point cloud object. The complexity factor λ is introduced into the loss function to train the model.

5. The method for restoring missing teeth in the posterior oral cavity based on metal framework and deep learning according to claim 4, characterized in that: The process of calculating the outlier value of each point using the LOF algorithm is: For each point, determine its k nearest neighbors and calculate its minimum distance to the k nearest neighbors; Calculate the local reachability distance of each point; Calculate the local reachability density of each point. The local reachability density is the inverse of the local reachability distance. The LOF score is then calculated based on the local reachable density, and then all points are sorted, and points with a score greater than the set LOF score threshold are selected as the points for this point cloud sampling.

6. The method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning according to claim 5, characterized in that: The calculation formula for the local reachable distance of point p in the point cloud is: reach-dist’ k (p,o)=B·max{dist(p,o)·Z1,k-dist(o)} Among them, reach-dist' k (p,o) is the local reachable distance of point p; Z i is the weight value of the normal vector of point p; B is the adjustment coefficient; max{·} is the maximum value function; dist(p,o) is the Euclidean distance between point p and point o; k-dist(o) is the minimum distance from point o to its k nearest neighbors.

7. The method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning according to claim 4, characterized in that: The weight factor Z0 of each point is expressed as follows: in, is the weight coefficient of Z1, is the weight parameter of Z2.

8. The method for restoring missing teeth in the posterior oral cavity based on a metal framework and deep learning according to claim 4, characterized in that: The complexity factor λ is introduced into the loss function, and the loss function calculation formula is: Among them, L is the loss function, N is the number of samples, and y i is the true label, is the predicted probability; A is the adjustment coefficient.

9. The method for restoring missing teeth in the posterior oral cavity based on metal framework and deep learning according to claim 1, characterized in that: The method also includes: using a three-dimensional geometric matching algorithm to evaluate the matching degree between the restoration and surrounding teeth, and fine-tuning the restoration according to the matching result.

10. An oral posterior model tooth loss restoration system based on metal framework and deep learning, characterized in that: include: The Metal Framework Fabrication module is used to design a metal framework for fixing the restoration; A scanning module, used to obtain three-dimensional point cloud data of the patient's oral cavity using an oral scanning device; Missing teeth reconstruction module, which is used to reconstruct the missing teeth in three dimensions using a self-attention implicit function network; The mounting module is used to customize the dental restoration according to the reconstructed 3D tooth model and mount it on the metal framework.

Images