A method for automatically generating a prosthesis from three-dimensional scan data, a generator for automatically generating a prosthesis from three-dimensional scan data, and a computer-readable recording medium having a program recorded thereon for causing a computer to execute the same

Abstract

To provide an automated method for generating a prosthesis from three-dimensional scan data, the method reducing time for generating the prosthesis by using a geometric deep learning.SOLUTION: The automated method for generating a prosthesis from three-dimensional scan data includes: extracting prep-information of a prepared tooth from the three-dimensional scan data; generating two-dimensional projection images by projecting the three-dimensional scan data on the basis of the prep-information; and generating a three-dimensional prosthesis on the basis of the two-dimensional projection images using a generative adversarial network including a two-dimensional encoder and a three-dimensional decoder.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for automatically generating a prosthesis from three-dimensional scan data, a generator for automatically generating a prosthesis from three-dimensional scan data, and a computer-readable recording medium having a program recorded thereon for causing a computer to execute the same. More specifically, the present invention relates to a method for automatically generating a prosthesis from three-dimensional scan data, a generator for automatically generating a prosthesis from three-dimensional scan data, and a computer-readable recording medium having a program recorded thereon for causing a computer to execute the same, which uses geometric deep learning to shorten the production time of the prosthesis.

Background Art

[0002] Three-dimensional oral scan data refers to data obtained by scanning teeth, the oral cavity, or an object imitating or reconstructing the same with a three-dimensional scanner. Dental treatments such as inlays, onlays, crowns, implants, and orthodontics acquire a patient's oral data and are used for prosthesis or implant design, orthodontic appliance production, and the like.

[0003] Conventionally, a method of directly imitating the oral cavity using alginate or the like and then manually producing a prosthesis has been mainly used. In order to create an anatomically correct prosthesis, a dentist or dental technician grasps the degree of wear of the surrounding teeth, comprehensively understands the tooth number of the tooth and the meshing information of the opposing tooth, and then generates a result. The conventional prosthesis generation method can be manually modified according to the oral condition of each patient based on a general tooth shape in consideration of such information.

[0004] In addition, conventionally, since the process of generating a prosthesis is performed manually, there is a problem that the work fatigue of a dentist or dental technician increases, and the accuracy and productivity of the result decrease. In addition, there is a problem that the deviation in the quality and required time of the prosthesis is large depending on the skill level of the operator.

Summary of the Invention

Problems to be Solved by the Invention

[0005] An object of the present invention is to provide a method for automatically generating a prosthesis from three-dimensional scan data that shortens the production time of the prosthesis using geometric deep learning.

[0006] Another object of the present invention is to provide a generator for automatically generating a prosthesis from three-dimensional scan data.

[0007] Still another object of the present invention is to provide a computer-readable recording medium on which a program for causing a computer to execute a method for automatically generating a prosthesis from the three-dimensional scan data is recorded.

Means for Solving the Problems

[0008] A method for automatically generating a prosthesis from three-dimensional scan data according to an embodiment for realizing the object of the present invention includes a step of extracting prepared information of a prepared tooth from the three-dimensional scan data, a step of generating a two-dimensional projection image obtained by projecting the three-dimensional scan data based on the prepared information, and a step of generating a three-dimensional prosthesis based on the two-dimensional projection image using an adversarial generation network including a two-dimensional encoder and a three-dimensional decoder.

[0009] Furthermore, the method includes a step of extracting a margin line of the prepared tooth, and the step of extracting the prepared information extracts the prepared information using prepared mesh data extracted using the margin line.

[0010] The prepared information includes a position of the prepared tooth, and the position of the prepared tooth is a center of gravity of the prepared mesh data.

[0011] The prepared information includes the position of the prepared tooth, and the position of the prepared tooth is the center of the margin line.

[0012] The prepared information includes the direction of the prepared tooth, and the direction of the prepared tooth is determined using the normal vector of the surface of the prepared mesh data.

[0013] When the direction of the prepared tooth is d, the number of surfaces of the prepared mesh data is N, and the normal vector is JPEG0007699849000001.jpg6170 where xopt is the direction in which the normal vector of the point of the prepared mesh data is not hidden, and T is the swapping function that exchanges the rows and columns of the matrix, JPEG0007699849000002.jpg8170 is satisfied.

[0014] The prepared information includes the position and the direction of the prepared tooth, the two-dimensional projection image is generated using a projection plane, the projection plane is arranged at a predetermined distance from the position of the prepared tooth, and is defined such that the mating tooth of the prepared tooth or the adjacent tooth of the prepared tooth can be seen.

[0015] The pixel value of the two-dimensional projection image is defined as the distance to the closest point that hits the three-dimensional scan data when a ray is emitted from the projection plane in the direction of the normal vector of the projection plane.

[0016] The two-dimensional encoder receives the two-dimensional projection image and outputs a latent vector.

[0017] The three-dimensional decoder receives the latent vector and generates the coordinates of the points forming the three-dimensional prosthesis.

[0018] Furthermore, it includes the step of generating the prosthesis correct data used for the learning of the adversarial generation network, and the step of generating the prosthesis correct data converts the first correct data corresponding to the prepared tooth into the second correct data with a fixed connection relationship by using deformable registration.

[0019] The step of generating the prosthesis correct data divides the hexahedron-shaped initial model into eight equal parts and deforms the initial model so as to be close to the shape of the first correct data to generate the second correct data.

[0020] Furthermore, it includes the step of learning the adversarial generation network, and the step of learning the adversarial generation network includes a first learning step of inputting the prosthesis correct data into a 3D encoder to generate a latent vector, and inputting the latent vector into the 3D decoder to restore it to the prosthesis correct data.

[0021] The step of learning the adversarial generation network further includes a second learning step of using a generator to generate a learned 3D prosthesis and using a discriminator to determine whether the learned 3D prosthesis is true.

[0022] The generator includes the 2D encoder and the 3D decoder learned in the first learning step, and the discriminator includes the 3D encoder learned in the first learning step.

[0023] In the step of learning the adversarial generation network, the learning objective function uses a loss that compares the distance difference between the points of the correct mesh data and the points of the predicted mesh data. The number of points of the correct mesh data is the same as the number of points of the predicted mesh data. where the loss is L, the number of points of the correct mesh data is X, and the points of the correct mesh data are JPEG0007699849000003.jpg8170 、When the points of the predicted mesh data are JPEG0007699849000004.jpg8170 set to, JPEG0007699849000005.jpg9170 it satisfies.

[0024] The first correct data corresponding to the prepared tooth is converted into second correct data with a fixed connection relationship using deformable registration, and the prosthesis correct data is the second correct data.

[0025] A generator for automatically generating a prosthesis from three-dimensional scan data according to an embodiment for realizing the object of the present invention includes a two-dimensional encoder and a three-dimensional decoder. The two-dimensional encoder receives a two-dimensional projection image of a prepared tooth of three-dimensional scan data and outputs a latent vector. The three-dimensional decoder receives the latent vector and generates coordinates of points forming a three-dimensional prosthesis for the prepared tooth.

[0026] A program for causing a computer to execute a method for automatically generating a prosthesis from the three-dimensional scan data is recorded on a computer-readable recording medium.

Advantages of the Invention

[0027] According to the method for automatically generating a prosthesis from three-dimensional scan data according to the present invention, prepared information of a prepared tooth is automatically extracted from the three-dimensional scan data, and based on the prepared information, a two-dimensional projection image obtained by projecting the three-dimensional scan data is generated, and a three-dimensional prosthesis can be automatically generated using an adversarial generation network including a two-dimensional encoder and a three-dimensional decoder.

[0028] In the method for automatically generating a prosthesis of the present invention, the two-dimensional coordinates of the prosthesis are generated and not restored to three-dimensional coordinates. Instead, the three-dimensional coordinates of the prosthesis are directly generated using an adversarial generation network including a two-dimensional encoder and a three-dimensional decoder. Therefore, complex post-processing for restoring two-dimensional coordinates to three-dimensional coordinates is unnecessary.

[0029] Further, in the method for automatically generating a prosthesis of the present invention, since the three-dimensional coordinates of the prosthesis are directly generated, it can be applied to the anterior tooth region where there is insufficient information for the meshing surface and it is difficult to restore two-dimensional coordinates to three-dimensional coordinates.

[0030] In this way, the prosthesis is automatically generated from the three-dimensional scan data, the manufacturing time and process of the prosthesis can be shortened, and the quality of the prosthesis can be improved.

Brief Description of the Drawings

[0031]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Best Mode for Carrying Out the Invention

[0032] Regarding the embodiments of the present invention shown in the text, the specific structural or functional descriptions are merely exemplified for the purpose of explaining the embodiments of the present invention. The embodiments of the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described in the text.

[0033] The present invention can be subjected to various modifications and can have various forms. Specific embodiments will be illustrated in the drawings and described in detail in the text. However, this is not intended to limit the present invention to specific disclosed forms, and it should be understood to include all modifications, equivalents, and alternatives included within the spirit and technical scope of the present invention.

[0034] Terms such as first, second, etc. are used to describe various components, but the components should not be limited by the terms. The terms are used for the purpose of distinguishing one component from another. For example, without departing from the scope of the rights of the present invention, the first component can be referred to as the second component, and similarly, the second component can also be referred to as the first component.

[0035] When it is said that a certain component is "connected to" or "connected with" another component, it should be understood that it can be directly connected or connected to the other component, but there can also be other components in between. On the other hand, when it is said that a certain component is "directly connected to" or "directly connected with" another component, it should be understood that there are no other components in between. Other expressions for explaining the relationship between components, namely, "between ~" and "immediately between ~", or "adjacent to ~" and "directly adjacent to ~", etc. should be interpreted in the same way.

[0036] The terms used in this application are merely used to describe specific embodiments and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, terms such as "including" or "having" are intended to specify the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, and it should be understood that the presence or addition possibility of one or more other features, numbers, steps, operations, components, parts, or combinations thereof is not precluded in advance.

[0037] Unless otherwise defined, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or overly formal sense unless clearly defined in this application.

[0038] On the other hand, when a certain embodiment can be implemented differently, the functions or operations specified within a specific block can occur differently from the procedures specified in the flowchart. For example, two consecutive blocks can actually be performed substantially simultaneously, and depending on the related functions or operations, the said blocks can also be performed in reverse.

[0039] Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the accompanying drawings. The same reference numerals are assigned to the same components in the drawings, and duplicate descriptions of the same components are omitted.

[0040] FIG. 1 is a flowchart showing a method for automatically generating a prosthesis from three-dimensional scan data according to an embodiment of the present invention.

[0041] As shown in FIG. 1, the method for automatically generating a prosthesis from three-dimensional scan data according to this embodiment includes a step of extracting prepared information of a prepared tooth from the three-dimensional scan data (step S100), a step of generating a two-dimensional projection image obtained by projecting the three-dimensional scan data based on the prepared information (step S200), and a step of generating a three-dimensional prosthesis based on the two-dimensional projection image using a generative adversarial network including a two-dimensional encoder and a three-dimensional decoder (step S300).

[0042] The method for automatically generating a prosthesis from the three-dimensional scan data of this embodiment is performed by a computing device.

[0043] FIG. 2 is a detailed flowchart showing the method for automatically generating a prosthesis from the three-dimensional scan data of FIG. 1. FIG. 3 is a diagram showing the margin line of the prepared tooth of FIG. 2.

[0044] Referring to FIGS. 1 to 3, the method for automatically generating a prosthesis from the three-dimensional scan data further includes a step of extracting the margin line (Margin Line, ML) of the prepared tooth.

[0045] In order to utilize the present invention, three-dimensional scan data (three-dimensional mesh data) obtained by scanning a dental arch together with a prepared tooth from a three-dimensional scanner, and the margin line (ML) of the prepared tooth are required.

[0046] Here, the three-dimensional scan data refers to data obtained by scanning a tooth and oral cavity or an object imitating or reconstructing the same with a three-dimensional scanner. For example, the three-dimensional scan data is mesh data including three-dimensional vertices and triangle faces (Triangles) or quadrilateral faces (Rectangles) generated by connecting the points. There is no limitation on the file extension of the three-dimensional scan data, and for example, it is any one of ply, obj, and stl.

[0047] Here, the prepared tooth means a tooth prepared for a crown, and the prepared tooth means a tooth with a part of the tooth removed. Specifically, in order to generate a single crown, it is necessary to grind down the entire natural tooth so that the prosthesis can be easily fitted, and the natural tooth that has gone through this process is referred to as the prepared tooth. Also, the margin line (ML) means the edge of the prepared tooth. The margin line (ML) indicates the boundary between the prepared tooth and the tooth root.

[0048] Figure 3 shows the margin line (ML) of the prepared tooth. For example, the margin line (ML) is automatically extracted from the three-dimensional scan data. For example, the margin line (ML) is automatically extracted from the three-dimensional scan data using an artificial neural network.

[0049] For example, the step of extracting the margin line (ML) includes the step of extracting partial scan data corresponding to the prepared tooth from the three-dimensional scan data, the step of mapping the partial scan data into a predetermined two-dimensional space using a transformation matrix, the step of determining a curvature value from the data mapped in the two-dimensional space to obtain a two-dimensional margin line, and the step of converting the two-dimensional margin line into a three-dimensional margin line using the inverse matrix of the transformation matrix.

[0050] For example, the curvature value is any one of a maximum curvature value, a minimum curvature value, a Gaussian curvature value, and an average curvature value.

[0051] In the case of the upper surface of the tooth, the curvature value has a relatively constant value. On the other hand, at the boundary between teeth or at the part where the tooth and the tooth root are in contact, the curvature value changes greatly. Therefore, the margin line (ML) of the tooth can be determined using the curvature value.

[0052] Figure 4 is a diagram showing the step (step S100) of extracting the prepared information of the prepared tooth in Figure 1.

[0053] Referring to FIGS. 1 to 4, in order to input the three-dimensional scan data into the deep learning model, a process (Prep Info Extractor) for extracting prepared information for the prepared teeth of the three-dimensional scan data is required.

[0054] For example, in the step of extracting the prepared information (step S100), the prepared information is extracted using the prepared mesh data extracted using the margin line (ML).

[0055] For example, the prepared information includes the position (p) and direction (d) of the prepared tooth. For example, the position (p) of the prepared tooth is the center of gravity (Prep center) of the prepared mesh data.

[0056] For example, the position of the prepared tooth is p, the number of vertices of the prepared mesh data is K, and the vertices are JPEG0007699849000006.jpg6170 when JPEG0007699849000007.jpg9170 is satisfied.

[0057] Differently, the position (p) of the prepared tooth may also be the center of the margin line (ML).

[0058] The direction (d) of the prepared tooth is determined using the normal vector of the surface of the prepared mesh data. The direction (d) of the prepared tooth indicates the protruding direction of the prepared tooth. The direction (d) of the prepared tooth indicates the direction (Insertion Direction) in which the prosthesis model is inserted into the prepared tooth.

[0059] For example, the direction of the prepared tooth is d, the number of surfaces of the prepared mesh data is N, and the normal vector is JPEG0007699849000008.jpg6170 where x opt is the direction in which the normal vector of the points of the prepared mesh data is not hidden, and T is a swapping function that exchanges the rows and columns of a matrix, then JPEG0007699849000009.jpg9170 is satisfied.

[0060] In the direction x, expressing the fact that the normal vector n is not hidden by an equation, x T n > 0, which means that the angle between x and n is an acute angle. If the angle between x and n is an acute angle, then x T n > 0, and if the angle between x and n is a right angle, then x T n = 0, and if the angle between x and n is an obtuse angle, then x T n < 0. Therefore, xopt is the direction in which the average value of the angles with the normal vectors on the surfaces of each prepared mesh data is the lowest.

[0061] FIG. 5 is a diagram showing the step (step S200) of generating the two-dimensional projection image of FIG. 1.

[0062] Referring to FIGS. 1 to 5, based on the position (p) and direction (d) of the prepared tooth, two-dimensional projection images (Projected Images) obtained by projecting the three-dimensional scan data can be generated.

[0063] The method for automatically generating a prosthesis from the three-dimensional scan data further includes, prior to the step of generating a two-dimensional projection image (step S200), the step of aligning the three-dimensional scan data with the origin of a predetermined coordinate system in the direction of the predetermined coordinate system. The three-dimensional scan data thus aligned is referred to as processed 3D Models.

[0064] The two-dimensional projection image is generated using a projection plane (A d1 , A d2 , A d3 ). The projection plane (A d1 , A d2 , A d3) is defined to be arranged at a predetermined distance from the position (p) of the prepared tooth, such that the mating tooth of the prepared tooth, or the adjacent tooth of the prepared tooth, is visible.

[0065] For example, when the three-dimensional scan data includes only one of the upper jaw data or the lower jaw data, the two-dimensional projection image is defined such that the adjacent teeth of the prepared tooth are clearly visible.

[0066] For example, when the three-dimensional scan data includes a pair of upper jaw data and lower jaw data, the two-dimensional projection image is defined such that the mating tooth and the adjacent teeth of the prepared tooth are clearly visible.

[0067] The pixel value of the two-dimensional projection image is defined as the distance to the closest point that hits the three-dimensional scan data when a ray is emitted in the direction of the normal vectors (d1, d2, d3) of the projection plane (A d1 , A d2 , A d3 ) from the projection plane (A d1 , A d2 , A d3 ). Different from this, the pixel value of the two-dimensional projection image can also be defined as the perspective view of the three-dimensional scan data when a ray is emitted in the direction of the normal vectors (d1, d2, d3) of the projection plane (A d1 , A d2 , A d3 ) from the projection plane (A d1 , A d2 , A d3 ).

[0068] FIG. 5 shows a case where the number of the two-dimensional projection images is three, but the present invention is not limited thereto.

[0069] FIG. 6 is a diagram showing an adversarial generation network used in the step (step S300) of generating the three-dimensional prosthesis model of FIG. 1.

[0070] Referring to FIGS. 1 to 6, the adversarial generation network is also referred to as Geometric AI.

[0071] The input of the geometric AI is the 2D projection image, and the output of the geometric AI is a 3D prosthesis model. Here, the 3D prosthesis model is a 3D single crown model (3D Crown Model).

[0072] The adversarial generation network includes a 2D encoder (Image Encoder) and a 3D decoder (Mesh Decoder). The 2D encoder receives the 2D projection image and outputs a latent vector. The 3D decoder can receive the latent vector and generate the coordinates of the points forming the 3D prosthesis.

[0073] Specifically, the 2D encoder is a function that receives, on the input side, projected image data of size M and provides, on the output side, an encoded latent vector. JPEG0007699849000010.jpg7170 and provides, on the output side, an encoded latent vector.

[0074] In the 3D decoder, the latent vector created by the 2D encoder is received on the input side, and finally a 3D output corresponding to the crown model is produced.

[0075] In the present invention, in order to learn for data with a fixed connection relationship, the 3D output can only predict the position of each point of the mesh. Accordingly, instead of 2D convolution for stereotyped data, operations applicable to non-stereotyped data are used. For example, the 3D decoder can utilize graph convolution operations such as GCN, ChebConv, GraphConv, PointNetConv, DynamicEdgeConv, SpiralConv.

[0076] As shown in FIG. 5, a three-dimensional crown mesh is generated from a two-dimensional projection image. The geometric AI combines the two-dimensional projection images obtained in multiple directions to generate the three-dimensional crown mesh. The adversarial generation network includes an image decoder that understands two-dimensional image information and a mesh decoder that re-analyzes the information understood in two dimensions into three dimensions.

[0077] FIG. 7 is a diagram showing a method for generating correct data used for learning the adversarial generation network of FIG. 6.

[0078] Referring to FIGS. 1 to 7, the method for automatically generating a prosthesis from the three-dimensional scan data further includes a step of generating prosthesis correct data used for learning the adversarial generation network. The step of generating the prosthesis correct data converts first correct data (CR1) corresponding to the prepared tooth into second correct data (CR2) with a fixed connection relationship using deformable registration.

[0079] For example, the step of generating the prosthesis correct data can divide a hexahedron-shaped initial model (M1) into eight equal parts, deform the initial model (M1) so as to be close to the shape of the first correct data (CR1), and generate the second correct data (M2).

[0080] Specifically, in order to generate the three-dimensional crown model proposed in the present invention, learning of the geometric AI is required. For learning of the geometric AI, three-dimensional scan data including the prepared tooth and actual crown mesh data (CR1) made by a dental technician corresponding to the prepared tooth are required. However, since the tooth morphology and the positions of the feature points of each patient are different and the shape of the scan data is not constant, it is unnatural to directly use the crown mesh data (CR1) made by the dental technician for model learning.

[0081] As a solution to this, a method of performing deformable registration on the crown mesh data (CR1) created by a dental technician to a polygon mesh (tooth library) representing a general tooth shape, a hexahedron initial model, or a spherical initial model is used. Deformable registration refers to a method of aligning a source mesh having an indefinite property (connection relationship) with the connection relationship of a previously defined target mesh.

[0082] The crown mesh data (CR1) created by each dental technician has different properties (connection relationships) from each other, but when the deformable registration is used, the properties of the crown mesh data (CR1) can be made the same.

[0083] As shown in FIG. 6, as a method of deformable registration, a Mesh shrink wrapping technique is used. This is a method of dividing a hexahedron quad mesh through a number of steps and creating a mesh shape having a target shape (CR1). The crown mesh data (CR2) with the same properties in this way is used as the correct data for training the model.

[0084] That is, by the Mesh shrink wrapping technique of FIG. 6, an algorithm for each step is applied to the initial mesh (M1) having a hexahedron shape to generate intermediate data such as M2, M3, M4, M5, and finally, the second correct data (CR2) having a fixed connection relationship with a shape like the first correct data (CR1) can be generated.

[0085] FIG. 8 is a diagram showing a method of learning the adversarial generation network of FIG. 6.

[0086] Referring to FIGS. 1 to 8, the method of automatically generating a prosthesis from the three-dimensional scan data further includes a step of learning the adversarial generation network.

[0087] The step of training the adversarial generation network includes a first training step (Training Stage 1) of inputting the correct patch data (e.g., CR2 in FIG. 7) into a 3D encoder (Mesh Encoder) to generate a latent vector (Z), and inputting the latent vector into the 3D decoder (Mesh Decoder) to restore it to the correct patch data (e.g., CR2 in FIG. 7). Here, the correct patch data is the second correct data (CR2).

[0088] For example, the step of training the adversarial generation network further includes a second training step (Training Stage 2) of using a generator to generate a learned 3D patch and using a discriminator to determine whether the learned 3D patch is True or Fake.

[0089] For example, the generator includes the 2D encoder (Image Encoder) and the 3D decoder (Pretained Mesh Decoder) learned in the first training step (Training Stage 1). The discriminator includes the 3D encoder (Pretained Mesh Encoder) learned in the first training step (Training Stage 1).

[0090] In the step of training the adversarial generation network, the training objective function uses a loss that compares the distance difference between the points of the correct mesh data and the points of the predicted mesh data. The number of points of the correct mesh data is the same as the number of points of the predicted mesh data. The loss is L, the number of points of the correct mesh data is X, the points of the correct mesh data are JPEG0007699849000011.jpg8170 , and the points of the predicted mesh data are JPEG0007699849000012.jpg8170 Then, JPEG0007699849000013.jpg9170 is satisfied.

[0091] Specifically, the adversarial generation network includes two model generators and discriminators. The adversarial generation network uses the following optimization method to obtain better results.

[0092] Since the connection relationship is fixed, in terms of the correct data points JPEG0007699849000014.jpg8170 and the points predicted by the deep learning model JPEG0007699849000015.jpg8170 are in one-to-one correspondence. Therefore, it is not necessary to use a loss with high complexity such as chamfer loss. In order to train the deep learning model proposed in the present invention, it can be trained with a simple L1L2 loss.

[0093] The learning consists of two steps. In the first learning step (Training Stage 1), first, in order to represent 3D prosthesis data from the latent vector (Z), the mesh decoder is learned. This uses the training method of an autoencoder. The correct data (CR2) is input as the input of the mesh encoder, and the process of compressing and expanding the size of the correct data (CR2) with the vector (Z) in the latent space and restoring it to the original state is repeated. Thus, the fixed connection relationship between the mesh encoder and the mesh decoder can be understood.

[0094] As the second training step (Training Stage 2), a generative adversarial network (GAN) structure model is trained. The generator is composed of an image encoder that extracts feature vectors from two-dimensional images and a mesh decoder trained in the first training step (Training 1), and the discriminator is composed of the mesh encoder trained in the first training step (Training 1). When the generator uses the two-dimensional projection image to create three-dimensional prosthesis data, the discriminator determines how real it is and provides the determined information to the generator, thereby training the generator. The generator trained in this way can produce a usable three-dimensional prosthesis with only simple post-processing without another complex algorithm.

[0095] According to this embodiment, prepared information of prepared teeth is automatically extracted from the three-dimensional scan data, and based on the prepared information, a two-dimensional projection image obtained by projecting the three-dimensional scan data is generated, and a three-dimensional prosthesis can be automatically generated using a generative adversarial network including a two-dimensional encoder and a three-dimensional decoder.

[0096] In the prosthesis automatic generation method of the present invention, the two-dimensional coordinates of the prosthesis are generated, and the three-dimensional coordinates of the prosthesis are directly generated using a generative adversarial network including a two-dimensional encoder and a three-dimensional decoder without restoring them to three-dimensional coordinates. Therefore, complex post-processing for restoring two-dimensional coordinates to three-dimensional coordinates is unnecessary.

[0097] Further, in the prosthesis automatic generation method of the present invention, since the three-dimensional coordinates of the prosthesis are directly generated, it can also be applied to the anterior tooth region where information on the meshing surface is insufficient and it is difficult to restore two-dimensional coordinates to three-dimensional coordinates.

[0098] In this way, by automatically generating a prosthesis from the three-dimensional scan data, the production time and process of the prosthesis can be shortened, and the quality of the prosthesis can be improved.

[0099] According to an embodiment of the present invention, there is provided a computer-readable recording medium having a program recorded thereon for causing a computer to execute a method for automatically generating a prosthesis from the three-dimensional scan data according to the above-described embodiment. The above method can be created by a program executed by a computer, and can be implemented by a general-purpose digital computer that operates the above program using a computer-readable medium. Further, the data structure used in the above method is recorded on a computer-readable medium by a plurality of means. The computer-readable medium can include program instructions, data files, data structures, etc., alone or in combination. The program instructions recorded on the medium are those specifically designed and configured for the present invention, and those known to and usable by ordinary technicians in the field of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as optical disks, and hardware devices specifically configured to store and execute program instructions such as ROMs, RAMs, and flash memories. The program instructions include not only machine language codes created by compilers, but also high-level language codes executed by a computer using an interpreter or the like. The above-described hardware device is configured to operate as one or more software modules for performing the operations of the present invention.

[0100] In addition, the above-described method for automatically generating a prosthesis from three-dimensional scan data can also be implemented in the form of a computer program executed by a computer stored in a recording medium, or an application.

[0101] [Industrial Applicability] The present invention relates to a method for automatically generating a prosthesis from three-dimensional scan data and a computer-readable recording medium having a program recorded thereon for causing a computer to execute the same, and can shorten the labor for manufacturing the prosthesis and improve the accuracy and productivity of the prosthesis.

[0102] In the foregoing, the preferred embodiments of the present invention have been described with reference thereto. However, those skilled in the art in the relevant technical field will understand that the present invention can be variously modified and changed without departing from the spirit and scope of the present invention described in the following claims.

Claims

1. extracting prepared information of a prepared tooth from three-dimensional scan data; generating a two-dimensional projection image obtained by projecting the three-dimensional scan data based on the prepared information; generating three-dimensional prosthesis data based on the two-dimensional projection image using an adversarial generation network including a two-dimensional encoder and a three-dimensional decoder; comprising: the prepared tooth is a tooth obtained by cutting a part of a natural tooth; further comprising extracting a margin line of the prepared tooth; the step of extracting the prepared information extracts the prepared information using prepared mesh data extracted using the margin line; the prepared information includes the position and the direction of the prepared tooth; the two-dimensional projection image is generated using a projection plane; further comprising generating ground truth prosthesis data used for learning the adversarial generation network; the step of generating the ground truth prosthesis data converts first ground truth data corresponding to the prepared tooth into second ground truth data with a fixed connection relationship using deformable registration; further comprising a step of learning the adversarial generation network; the step of learning the adversarial generation network includes a first learning step of inputting the ground truth prosthesis data into a three-dimensional encoder to generate a latent vector, and inputting the latent vector into the three-dimensional decoder to restore the ground truth prosthesis data; the ground truth prosthesis data is the second ground truth data; the step of learning the adversarial generation network further comprises: a method for automatically generating prosthesis data from three-dimensional scan data, characterized in that the step of learning the adversarial generation network includes a second learning step of generating learned three-dimensional prosthesis data using a generator and determining whether the learned three-dimensional prosthesis data is true using a discriminator.

2. The position of the prepared tooth is the centroid of the prepared mesh data, characterized in that. The method for automatically generating prosthesis data from three-dimensional scan data according to Claim 1.

3. The position of the prepared tooth is the center of the margin line, characterized in that. The method for automatically generating prosthesis data from three-dimensional scan data according to Claim 1.

4. The direction of the prepared tooth is determined using the normal vector of the surface of the prepared mesh data, characterized in that A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 1.

5. The direction of the prepared tooth is d, the number of surfaces of the prepared mesh data is N, and the normal vector is where xopt is the direction in which the normal vector of the points of the prepared mesh data is not hidden, and T is a swapping function that exchanges the rows and columns of a matrix, characterized in that it satisfies A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 4.

6. The projection plane is arranged at a predetermined distance from the position of the prepared tooth and is defined so that the mating tooth of the prepared tooth or the adjacent tooth of the prepared tooth can be seen, characterized in that A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 1.

7. The pixel value of the two-dimensional projection image is defined as the distance to the closest point that hits the three-dimensional scan data when a ray is emitted from the projection plane in the direction of the normal vector of the projection plane, characterized in that A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 6.

8. The two-dimensional encoder receives the two-dimensional projection image and outputs a latent vector, characterized in that A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 1.

9. The three-dimensional decoder receives the latent vector and generates the coordinates of the points forming the three-dimensional prosthesis data, characterized in that A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 8.

10. The step of generating the prosthesis correct data divides the initial hexahedron-shaped model into eight equal parts and deforms the initial model so as to approximate the shape of the first correct data to generate the second correct data, characterized in that A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 1.

11. The generator includes the two-dimensional encoder and the three-dimensional decoder learned in the first learning step, The discriminator includes the three-dimensional encoder learned in the first learning step, characterized in that A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 1.

12. In the step of training the adversarial generation network, the training objective function uses a loss that compares the distance difference between the points of the correct mesh data and the points of the predicted mesh data. The number of points of the correct mesh data is the same as the number of points of the predicted mesh data. The loss is L, the number of points of the correct mesh data is X, the points of the correct mesh data are , and the points of the predicted mesh data are When, It is characterized by satisfying A method for automatically generating prosthesis data from the three-dimensional scan data according to claim 1.

13. A computer-readable recording medium on which a program for causing a computer to execute the method according to any one of claims 1 to 12 is recorded.

Images