Artificial intelligence technology assisted digital design method for oral denture

By performing data annotation and intelligent neural network training on the three-dimensional tooth jaw model, the key points and regular coordinate systems are automatically identified, and the problems of confusion in the initial direction and complex soft tissue shape are solved, the automated design of oral dentures is realized, and the design intelligence is improved. Efficiency reduces medical costs and time consumption.

CN118981946APending Publication Date: 2024-11-19NANJING WEIYIN TECHNOLOGY CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411013081.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-19

Smart Images

  • Figure CN118981946A_ABST
    Figure CN118981946A_ABST
Patent Text Reader

Abstract

The invention discloses an artificial intelligence technology-assisted digital design method for oral dentures, which comprises the following steps of: performing data annotation on a three-dimensional dental model, exporting matched data, and constructing a network learning database; an automatic position correcting module is constructed, key points of the three-dimensional dental model are automatically identified, a coordinate system of the three-dimensional dental model is calculated and corrected, the problem that directions of dental models of different data sources are disordered is avoided, and new coordinates of all point clouds in the original three-dimensional dental model are calculated based on the corrected coordinate system; performing curve fitting based on the key point coordinates output by the automatic position correcting module, and calculating key features of each point cloud of the three-dimensional dental model; the key features are input into an intelligent neural network for network learning training, classification prediction of each point cloud is output, segmentation boundary prediction is formed, segmentation boundaries are refined by using a segmentation refining algorithm, three-dimensional intelligent modeling of the oral denture is performed, the intelligent degree of the design process is improved, and the overall working efficiency is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of oral medicine, and in particular to an artificial intelligence technology-assisted oral denture digital design method. Background Art

[0002] With the advancement of computer technology and the rise of the big data era, artificial intelligence technology has been widely used in many industries. Especially in the field of stomatology, digital technology based on intelligent neural networks has shown significant growth potential. Artificial intelligence has shown great advantages in providing accurate, efficient and practical medical services. It can reduce the subjectivity of doctors in clinical operations, reduce fatigue and visual errors, and achieve standardization and accuracy of operations. These technologies are not only minimally invasive, time-saving and efficient, but also can significantly reduce medical costs and time consumption, and effectively respond to the increasing demand for oral medical care due to the aging population. However, artificial intelligence technology in the field of oral medicine is still in its infancy, and there are few reports on the digital design methods of oral dentures based on automatic alignment and intelligent feature extraction of three-dimensional dental models. The dental model has obvious directionality, and technicians only design denture restorations in the occlusal direction.

[0003] However, edentulous digital models from different data sources have different initial directions. The chaotic initial coordinate system input can only enable the neural network to extract key anatomical features, which is prone to produce erroneous results. Furthermore, the soft tissue shape of the three-dimensional dental model is irregular, with no obvious boundaries or curvature changes. The local coordinate changes of the three-dimensional points are difficult to capture and learn, which further increases the difficulty of extracting key anatomical features and makes it difficult to complete the automated digital design of oral dentures. A large amount of manual interaction is still required, which is time-consuming and has a low level of intelligence. Summary of the invention

[0004] The purpose of the present invention is to provide an artificial intelligence technology-assisted oral denture digital design method that can effectively solve the problems existing in the above-mentioned prior art.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solution: an artificial intelligence technology-assisted oral denture digital design method, comprising:

[0006] Annotate the 3D dental model, export the supporting data, and build a network learning database;

[0007] Build an automatic alignment module to automatically identify the key points of the 3D dental model, calculate and align the coordinate system of the 3D dental model, and calculate the new coordinates of all point clouds in the original 3D dental model based on the aligning coordinate system;

[0008] Perform curve fitting based on the key point coordinates output by the automatic alignment module to calculate the key features of each point cloud of the 3D dental model;

[0009] The key features are input into the intelligent neural network for network learning and training, and the classification prediction of each point cloud is output to form a segmentation boundary prediction. The segmentation refinement algorithm is used to refine the segmentation boundary and perform three-dimensional intelligent modeling of oral dentures.

[0010] Preferably, the network learning database is preprocessed, including:

[0011] Convert the three-dimensional dental model into manifold mesh data;

[0012] The manifold 3D mesh of the 3D dental model is flattened into an isomorphic 2D mesh; and the transformation relationship between UV coordinates and XYZ coordinates is obtained at the same time, realizing a bidirectional reversible mapping between space and plane;

[0013] Among them, the 3D grid space With 2D grid plane The parameter formula between the points p(u,v) on is as follows:

[0014] p(u,v)=s(x(u,v),y(u,v),z(u,v)).

[0015] Preferably, the data annotations of the key points in the 2D grid are input into the automatic alignment module, the key points are automatically identified and mapped back to the three-dimensional dental model, and the predicted key points are output; the loss function of the points is calculated and back-propagated and fed back, the weights of the front-layer neurons are adjusted, and the forward and backward propagation processes are repeated until convergence.

[0016] Preferably, the loss function is: Among them, (Ii,si) is the i-th training sample pair, that is, Ii is the i-th dental model, and si is the data annotation of the coordinate vector of the positioning key point corresponding to the i-th dental model.

[0017] Preferably, the coordinate system of the three-dimensional dental model is calculated and rectified based on the output predicted key points to form a new coordinate system; and the new coordinates of all point clouds of the entire original edentulous jaw model are recalculated in the new coordinate system, and the calculation formula is as follows:

[0018]

[0019] Preferably, the key features include: the coordinates of the point cloud (x, y, z), the normal vector of the point cloud (Nx, Ny, Nz), the angle θ between the line connecting the point to the origin and the x-axis in the new coordinate system, the distance sgn(d) from the point to the curve, and the local angle θ-local between the point and the curve.

[0020] Preferably, the distance sgn(d) is calculated as follows:

[0021] d=distance(Arch curve,PointE)

[0022]

[0023] PointE is the key point, and PointN is the point on the curve closest to PointE.

[0024] Preferably, all points in the original jaw mesh are represented as The final feature map of B Among them, F l -F9 is expressed as each key feature.

[0025] Beneficial effects: The present invention automatically identifies the key positions of the three-dimensional dental model through the automatic alignment module, and uses the key points to perform the alignment coordinate system, thereby calculating the new coordinates of all point clouds, realizing the automatic alignment of the model, and avoiding the problem of confusion in the directions of dental models from different data sources; reducing manual operations and improving the degree of intelligence:

[0026] Among them, the present invention utilizes a database in combination with an intelligent neural network to realize the automated design of three-dimensional intelligent modeling of oral dentures, which reduces a large amount of manual interaction and improves the intelligence level of the design process; this not only reduces human errors and fatigue, but also shortens the design time and improves overall work efficiency.

[0027] Curve fitting and 3D point cloud feature extraction technology based on key point coordinates can effectively capture and learn the soft tissue shape and local coordinate changes of the 3D dental model, significantly improving the model adaptability and the accuracy of feature extraction, thereby ensuring the quality of automated design.

[0028] In addition, the method of the present invention can standardize and automate the digital design process, significantly reduce medical costs and time consumption, effectively address the problem of increased oral medical needs caused by an aging population, optimize the allocation of medical resources, and improve the overall level of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0030] In the attached picture:

[0031] Figure 1 5 key points of the jaw arch curve of the present invention are marked as an example diagram;

[0032] Figure 2 is a labeled example diagram of the oral denture automatic intelligent design network of the present invention;

[0033] Figure 3 It is a bidirectional reversible mapping diagram of the space-plane of the dental model of the present invention;

[0034] Figure 4 It is a three-dimensional automatic orthographic image of edentulous jaw based on orthographic key points of the present invention;

[0035] Figure 5 It is a personalized dental arch curve fitting diagram based on automatic identification of key points of the present invention;

[0036] Figure 6 It is a key characteristic diagram of the present invention for calculating the three-dimensional dental and jaw model using the jaw arch curve;

[0037] Figure 7 This is an example diagram of the automatic design based on the intelligent neural network of the present invention. DETAILED DESCRIPTION

[0038] The embodiments of the present invention are described below in conjunction with the drawings in the embodiments of the present invention. The terms used in the implementation mode of the present invention are only used to explain the specific embodiments of the present invention, and are not intended to limit the present invention. The embodiments of the present application are described below in conjunction with the drawings.

[0039] Embodiment: An artificial intelligence-assisted oral denture digital design method, comprising:

[0040] Annotate the 3D dental model, export the supporting data, and build a network learning database;

[0041] Data annotation includes key point annotation, physiological and anatomical feature area annotation and target boundary annotation, specifically: In a specific embodiment, refer to Figure 1 As shown, the three-dimensional jaw model (M) of the maxilla and mandible is obtained by three-dimensional scanning, and key point data is annotated on the three-dimensional jaw model through CAD software, generally by dental experts, and 5 key points of the three-dimensional jaw model are annotated. The personalized jaw arch curve of the three-dimensional jaw model (M) can be fitted through these five key points P1-P5; specifically, Figure 1 The right maxillary node P1, the right canine positioning point P2, the mesial point of the central incisor P3, the left canine positioning point P4, and the left maxillary node P5; export supporting data, including the three-dimensional dental models of the maxilla and mandible, and 5 key points (P1-P5), and incorporate them into the jaw arch curve key point recognition network learning database for training and learning.

[0042] refer to Figure 2As shown in the figure, the physiological and anatomical feature areas and the target boundaries are annotated to build a learning database for the automatic intelligent design network of oral dentures. Dental experts use CAD software to annotate on the three-dimensional tooth-jaw model, outline the boundaries of the physiological and anatomical feature areas (including clinical crown margin lines, maxillary and mandibular dental arch curves, maxillary and mandibular vestibule groove turning lines, maxillary and mandibular alveolar ridge top areas, maxillary hard areas, palatal wrinkles, zygomatic processes, maxillary tuberosity areas, mandibular buccal shelf areas, mandibular molar retro-pad areas, etc.); outline the target boundaries of digital design (including the boundaries of the three-dimensional printed brackets of full dentures and removable partial dentures, the cervical margin lines of fixed crowns and the occlusal contact areas, etc.); export supporting data, including the three-dimensional tooth-jaw models of the maxilla and mandible and their expert-annotated edge information, as the learning database for the automatic intelligent design network of oral dentures, as shown in the figure. Figure 2 As shown in the figure, (a) an example of annotation of physiological and anatomical feature areas; (b) an example of annotation of target boundaries for digital design of full denture metal base.

[0043] like Figure 3 As shown, firstly, the three-dimensional dental model is converted into manifold mesh data and uniformly converted into manifold mesh data; through the bidirectional reversible mapping of space-plane, the three-dimensional coordinates can be converted into two-dimensional coordinates, and the two-dimensional key points can be reversely mapped back to the three-dimensional edentulous jaw model; by flattening the manifold 3D dental mesh, it is flattened into an isomorphic 2D mesh, and the conversion relationship between UV coordinates (2D mesh) and XYZ coordinates (3D mesh) is obtained at the same time, realizing the bidirectional reversible mapping of space-plane;

[0044] 3D mesh space 2D mesh plane The parameter formula between the points p(u,v) on is as follows:

[0045] p(u,v)=s(x(u,v),y(u,v),z(u,v));

[0046] Here, S represents a specific three-dimensional grid space; Mathematical symbol that means "subset"; Denotes that S is a three-dimensional real space A subset of ; P: Here P represents a specific two-dimensional grid plane; Denotes that P is the two-dimensional real number plane A subset of .

[0047] Construct an automatic alignment module, input the data annotations of the key points in the 2D grid into the automatic alignment module, automatically identify the key points, and map them back to the three-dimensional dental model, and output the predicted key points. For example, in the above-mentioned specific embodiment, the trained predicted points are P1', P2', P3', P4', and P5'; calculate the loss function of the point and back-propagate feedback, adjust the weights of the front-layer neurons, repeat the forward and back-propagation processes until convergence, that is, the error signal is minimized, and complete the training of the positioning module; specifically, the input is the three-dimensional scanned dental model in the learning database, and the output (i.e., the fitting target) is a vector s=[x1, y1, x2, y2, ..., xN, yN] composed of the dental positioning key point labels. T , where N is the number of positioning key points, (xi, yi) is the coordinate of the key point of the anatomical feature of the i-th edentulous jaw; the goal of the dentition positioning key point detection model is to find a nonlinear mapping relationship: φ: I→s;

[0048] During training, the loss function is: Among them, (Ii,si) is the i-th training sample pair, that is, Ii is the i-th dental model, si is the data annotation of the coordinate vector of the positioning key point corresponding to the i-th dental model; the training process is to minimize the objective function as much as possible through repeated iterations of the optimization algorithm. At this time, the nonlinear mapping This is the optimal model corresponding to the neural network; using this model, key points can be automatically identified and mapped back to the three-dimensional dental model.

[0049] Based on the above embodiment, the posterior coordinate system Axis of the three-dimensional dental model can be determined by using the prediction points P1', P2', P3', P4', and P5'; Figure 4 As shown, the specific steps are as follows: First, calculate the new x-axis, y-axis, z-axis and coordinate origin: Get from the first point P1' and the fifth point P5' Through point P3' The perpendicular line of the new coordinate system is set to O, with its foot as the origin of the new coordinate system. After normalization, it is used as the new x-axis unit vector; the vector After normalization, it is used as the unit vector of the y-axis; the new unit vector of the z-axis is obtained by multiplying the unit vector of the x-axis by the unit vector of the y-axis; finally, the new coordinates of all point clouds of the entire original edentulous jaw model are recalculated in the new coordinate system using the new x-axis, y-axis, z-axis and coordinate origin. The calculation formula is as follows:

[0050]

[0051] Where x': the x coordinate in the new coordinate system; y': the y coordinate in the new coordinate system; z': the z coordinate in the new coordinate system; (x, y, z): the coordinates of the original point; (xO ,y O ,z O ): coordinates of the origin of the new coordinate system; The new x-axis unit vector; The new unit vector of the y axis; The new unit vector of the z-axis; “·”: represents the dot product (inner product) operation.

[0052] Perform curve fitting based on the key point coordinates output by the automatic alignment module to calculate the key features of each point cloud of the 3D dental model;

[0053] Based on the above embodiments, reference Figure 5 As shown, through P1'-P5', the personalized dental arch curve of the three-dimensional dental model can be fitted; the personalized dental arch curve fitting is performed using parameter curves including Bezier curves, B-spline curves, NURBS curves and polynomial curves, among which, Figure 5 According to the new and old coordinate conversion formula, the x' and y' coordinates of the predicted points P1'-P5' in the new coordinate system are obtained, and the z' coordinate is not shown.

[0054] The key features include: the coordinates of the point cloud (x, y, z), the normal vector of the point cloud (Nx, Ny, Nz), the angle θ between the line connecting the point to the origin and the x-axis in the new coordinate system, the distance sgn(d) from the point to the curve, and the local angle θ-local between the point and the curve;

[0055] refer to Figure 6 As shown in the figure, the feature extraction method is as follows: for any point (PointE) on the three-dimensional dental model, its coordinate values ​​(x, y, z) and normalized normal vector (Nx, Ny, Nz) are used as the 1st to 6th features; the coordinates (x, y) of the point are converted into polar coordinates (R, θ), where θ is used as the 7th feature of PointE:

[0056]

[0057] Furthermore, find the point (PointN) closest to PointE on the jaw arch curve, and calculate the signed Euclidean distance sgn(d) between PointE and the jaw arch curve as the eighth feature; if the Euclidean distance from PointE to the origin (PointO) is greater than the distance from PointN to PointO, sgn(d) is positive (such as example point 1); otherwise, sgn(d) is negative (such as example point 2); the calculation formula is as follows:

[0058] d=distance(Arch curve,PointE)

[0059]

[0060] Finally, calculate the x-axis to vector The rotation angle, that is, the local angle θ between point E and the jaw arch curve local , as the 9th feature;

[0061] After extracting 9 features for each point, the state of the input data is re-described in mathematical language. All points in the original jaw mesh are represented as The final feature map of B 1 = 1-9 is divided into the above 9 characteristics;

[0062] F1=x; F2=y; F3=z; F4=Nx; F5=Ny; F6=Nz; F7=θ; F8=sgn(d); F9=θ local ;

[0063] The key features are input into the intelligent neural network for network learning and training. After steps including convolution, deconvolution, activation function, pooling, full connection, etc., the classification prediction of each point is output to form a segmentation boundary prediction. The segmentation refinement algorithm is used to refine the segmentation boundary calculation error and back-propagate feedback to adjust the weights of the front-layer neurons. The forward and back-propagation processes are repeated until the network converges and the network training is ended. Figure 7 As shown in the figure, the physiological and anatomical feature area boundary or the target boundary of the digital design output by the network can replace the manual boundary drawing of the technician in the digital design, and perform three-dimensional intelligent modeling of oral dentures. The output file can be directly used for three-dimensional printing intelligent manufacturing. Figure 7 As shown, the purple area is the intelligently identified full denture metal base retention mesh area, the green area is the intelligently identified large connector area, and the red arrows are the intelligently identified tissue termination points and inner and outer termination lines.

[0064] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. For ordinary technicians in this technical field, after knowing the contents recorded in the present invention, they can make several equivalent changes and substitutions without departing from the principle of the present invention. These equivalent changes and substitutions should also be regarded as belonging to the protection scope of the present invention.

Claims

1. An artificial intelligence-assisted oral denture digital design method, characterized in that: include: Annotate the 3D dental model, export the supporting data, and build a network learning database; Build an automatic alignment module to automatically identify the key points of the 3D dental model, calculate and align the coordinate system of the 3D dental model, and calculate the new coordinates of all point clouds in the original 3D dental model based on the aligning coordinate system; Perform curve fitting based on the key point coordinates output by the automatic alignment module to calculate the key features of each point cloud of the 3D dental model; The key features are input into the intelligent neural network for network learning and training, and the classification prediction of each point cloud is output to form a segmentation boundary prediction. The segmentation refinement algorithm is used to refine the segmentation boundary and perform three-dimensional intelligent modeling of oral dentures.

2. The method for digital design of oral dentures assisted by artificial intelligence technology according to claim 1, characterized in that: Preprocess the network learning database, including: Convert the three-dimensional dental model into manifold mesh data; The manifold 3D mesh of the 3D dental model is flattened into an isomorphic 2D mesh; and the transformation relationship between UV coordinates and XYZ coordinates is obtained at the same time, realizing a bidirectional reversible mapping between space and plane; Among them, the 3D grid space With 2D grid plane The parameter formula between the points p(u,v) on is as follows: p(u,v)=s(x(u,v),y(u,v),z(u,v)).

3. The method for digital design of oral dentures assisted by artificial intelligence technology according to claim 2, characterized in that: The data annotation of the key points in the 2D grid is input into the automatic alignment module, which automatically identifies the key points and maps them back to the 3D dental model, and outputs the predicted key points; The loss function of the calculated point is back-propagated to adjust the weights of the neurons in the previous layer, and the forward and backward propagation processes are repeated until convergence.

4. The method for digital design of oral dentures assisted by artificial intelligence technology according to claim 3, characterized in that: The loss function is: Among them, (Ii,si) is the i-th training sample pair, that is, Ii is the i-th dental model, and si is the data annotation of the coordinate vector of the positioning key point corresponding to the i-th dental model.

5. The method for digital design of oral dentures assisted by artificial intelligence technology according to claim 3, characterized in that: The coordinate system of the three-dimensional dental model is calculated and rectified based on the output predicted key points to form a new coordinate system; and the new coordinates of all point clouds of the entire original edentulous jaw model are recalculated in the new coordinate system. The calculation formula is as follows:

6. The method for digital design of oral dentures assisted by artificial intelligence technology according to claim 1, characterized in that: The key features include: the coordinates of the point cloud (x, y, z), the normal vector of the point cloud (Nx, Ny, Nz), the angle θ between the line connecting the point to the origin and the x-axis in the new coordinate system, the distance sgn(d) from the point to the curve, and the local angle θ-local between the point and the curve.

7. The method for digital design of oral dentures assisted by artificial intelligence technology according to claim 6, characterized in that: The distance sgn(d) is calculated as: d=distance(Arch curve,PointE) PointE is the key point, and PointN is the point on the curve closest to PointE.

8. The artificial intelligence-assisted oral denture digital design method according to claim 6, characterized in that: All points in the original jaw mesh are represented as The final feature map of B Among them, F1-F9 are respectively expressed as key features.

Citation Information

Cited By

  • Three-dimensional modeling method, system and equipment for false tooth support and medium

    CN120726263A

  • A method, system, device and medium for three-dimensional modeling of a denture base

    CN120726263B