Method for automatically separating teeth from 3D scan data using tooth boundary curves, and computer-readable recording medium having recorded thereon a program for executing the method on a computer
The use of AI neural networks for tooth detection and boundary generation in three-dimensional scan data addresses inaccuracies in conventional methods, improving accuracy and efficiency in tooth separation.
Patent Information
- Application Number
- JP2024529847
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2021-12-29
- Filing Date
- 2022-01-27
- Publication Date
- 2024-11-11
- Estimated Expiration
- 2042-01-27
AI Technical Summary
Conventional tooth separation methods in dentistry face challenges such as inaccurate masking of tooth areas, invasion of masks into adjacent teeth or gums, misidentification of multiple teeth as one, and difficulty in separating partial teeth, leading to reduced accuracy and increased computational resources.
An automatic tooth separation method using artificial intelligence neural networks for tooth detection and boundary generation, involving mesh parameterization and tooth boundary curve generation, to accurately separate teeth from three-dimensional scan data.
The method improves accuracy by preventing incorrect masking and misidentification, reduces computational effort, and enables separation of partial teeth using the same model, enhancing overall tooth separation efficiency.
Smart Images

Figure 2024541724000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an automatic tooth separation method from 3D scan data using tooth boundary curves, and a computer-readable recording medium having a program recorded thereon for executing the same on a computer, and more particularly to an automatic tooth separation method from 3D scan data, which detects teeth from scan data through deep learning, performs mesh parameterization for each tooth, and generates tooth boundary curves within mapping data of each tooth through deep learning, and a computer-readable recording medium having a program recorded thereon for executing the same on a computer. [Background technology]
[0002] In dentistry, there is a need for technology to separate teeth from a patient's 3D scan data for diagnosis, analysis, prosthetics production, etc. In particular, digital orthodontics using oral scanners is becoming more common in dentistry. In orthodontics, it is important to predict the tooth alignment and occlusion and establish an appropriate plan, and for this, tooth separation is essential.
[0003] A commonly used method for tooth separation is as follows: first, tooth scan data is obtained using an oral scanner. Then, an operator manually specifies the tooth boundaries, specifies the plane to be used for tooth separation using the axis information and the tooth boundaries, and finally checks the separated surfaces between the teeth and corrects any necessary parts to finish. Then, this process is repeated for all teeth to obtain the separated tooth data.
[0004] When workers perform this task manually, they must specify boundaries for 3D data with the naked eye on a 2D screen, which results in problems such as reduced accuracy and the need for high levels of skill and effort on the part of the worker.
[0005] Conventional tooth separation technology mainly uses a method of masking the plane corresponding to the teeth, which can lead to problems such as some areas of the teeth not being masked (Fig. 1, Fig. 2, Fig. 3), the mask not being formed within one tooth and affecting adjacent teeth or gums (Fig. 4, Fig. 5), and the mask judging multiple teeth as one tooth (Fig. 6).
[0006] For example, the arrow in Figure 1 indicates that the upper region of the tooth is not masked, the arrow in Figure 2 indicates that a hole has been formed in a portion of the upper region of the tooth, and the arrow in Figure 3 indicates that the tooth boundary region is not masked. The dotted oval in Figure 4 indicates that the mask has invaded other adjacent teeth, and the dotted oval in Figure 5 indicates that the mask has left the tooth and invaded the gums. The arrow in Figure 6 indicates that the mask has determined that two teeth are one tooth, and also indicates that the mask has left the tooth and invaded the gums.
[0007] In addition, the conventional tooth separation technology is limited to a technology for separating the entire tooth due to the limitations of the algorithm. In actual clinical practice, scan data of partial teeth, rather than the entire tooth, is often used, and the conventional tooth separation technology has a problem in that it is difficult to separate teeth from scan data of partial teeth. In addition, in order to separate teeth from scan data of partial teeth using deep learning, a model that is trained appropriately for partial tooth data is further required, which may result in excessive memory usage, excessive speed reduction, etc. Summary of the Invention [Problem to be solved by the invention]
[0008] It is an object of the present invention to provide a method for automatic tooth segmentation of 3D scan data, which is performed automatically by an artificial intelligence neural network for tooth detection, mesh parameterization, and an artificial intelligence neural network for tooth border generation.
[0009] Another object of the present invention is to provide a computer-readable recording medium having a program recorded thereon for causing a computer to execute the method for automatically separating teeth from 3D scan data. [Means for solving the problem]
[0010] In one embodiment of the present invention for achieving the above-mentioned object of the present invention, a method for automatically separating teeth from 3D scan data using a tooth boundary curve includes the steps of: detecting teeth in the scan data using a first artificial intelligence neural network; extracting tooth scan data from the scan data based on the tooth detection result; generating tooth mapping data corresponding to a predetermined space based on the tooth scan data; inputting the tooth mapping data into a second artificial intelligence neural network to generate a tooth boundary curve; and mapping the tooth boundary curve onto the scan data.
[0011] The method for automatic tooth segmentation of 3D scan data using tooth boundary curves further includes converting the 3D scan data into a 2D image, which is input to the first artificial intelligence neural network.
[0012] The step of detecting teeth in the scan data includes the steps of the first artificial intelligence neural network receiving the two-dimensional image and outputting an output two-dimensional image including two-dimensional tooth detection information, and converting the two-dimensional tooth detection information into three-dimensional tooth detection information. Detecting the teeth in the scan data includes generating a tooth bounding box that encloses one tooth.
[0013] The step of extracting tooth scan data includes extracting the tooth scan data corresponding to one tooth using a tooth cutting box corresponding to the tooth bounding box.
[0014] The tooth cutting box includes the tooth bounding box and is larger than the tooth bounding box. Detecting the teeth in the scan data includes mapping center points of the tooth bounding boxes onto the tooth surface to generate tooth feature points.
[0015] The step of extracting tooth scan data includes extracting the tooth scan data corresponding to one tooth using a tooth cut area formed based on the tooth feature points. The method for automatically separating teeth from 3D scan data using tooth boundary curves further includes converting the tooth scan data into curvature data representing a curvature value for each point in the tooth scan data.
[0016] The curvature data represents the minimum curvature value for each of the points. The method for automatically separating teeth from 3D scan data using the tooth boundary curve further includes a step of inverting the gradation of the curvature data so that when the minimum curvature value of each point is large, it is represented in white, and when the minimum curvature value of each point is small, it is represented in black.
[0017] The step of generating the tooth mapping data includes mapping the curvature data of the tooth scan data into the predetermined space to generate the tooth mapping data. The inputs of the second artificial intelligence neural network are the mapping data and the tooth bounding box, and the output of the second artificial intelligence neural network is the tooth boundary curve.
[0018] Generating the tooth boundary curve includes rotating the tooth bounding box to generate a diamond contour. Further, generating the tooth boundary curve includes shifting a location of a vertex of the diamond contour within the mapping data by the curvature value to generate a pole point.
[0019] Furthermore, the step of generating the tooth boundary curve includes the step of generating an octagonal contour passing through the extreme points based on the extreme points. Further, generating the tooth boundary curve includes displacing the octagonal contour within the mapping data by the curvature value to generate the boundary curve.
[0020] The first artificial intelligence neural network receives the three-dimensional scan data and outputs three-dimensional tooth detection information. A program for causing a computer to execute the method for automatically separating teeth from 3D scan data using the tooth boundary curve is recorded on a computer-readable recording medium. Effect of the Invention
[0021] According to the method for automatically separating teeth from 3D scan data using tooth boundary curves according to the present invention, the process is automatically performed through an AI neural network for tooth detection, mesh parameterization, and an AI neural network for tooth boundary generation, thereby reducing the effort required to separate teeth from scan data.
[0022] Using the tooth detection AI neural network, the mesh parameterization, and the tooth boundary generation AI neural network, tooth boundary curves, not tooth surfaces, are accurately generated, which can prevent problems such as a partial region of a tooth not being masked, a false region that is not a tooth being determined as a tooth, a mask not being formed within one tooth and thus affecting adjacent teeth or gums, multiple teeth being determined as one tooth, etc. Thus, the accuracy of the tooth separation can be improved.
[0023] In addition, the artificial intelligence neural network for generating tooth boundary can generate the tooth boundary curve that separates the teeth and gums at once by modifying the initially input contour data, thereby improving the accuracy of the tooth separation.
[0024] In addition, even if scan data for partial teeth rather than the entire teeth is input, the tooth detection artificial intelligence neural network is used to accurately determine the area of each tooth and generate boundary curves for each tooth-specific scan data corresponding to the tooth area, so that tooth separation can be performed using the same model regardless of whether the scan data represents the entire teeth or partial teeth. [Brief description of the drawings]
[0025] [Figure 1] FIG. 1 is a diagram showing an example of a tooth separation error that occurs in conventional automatic tooth separation from 3D scan data. [Diagram 2] FIG. 2 is a diagram showing an example of a tooth separation error that occurs in conventional automatic tooth separation of 3D scan data. [Diagram 3] FIG. 3 is a diagram showing an example of a tooth separation error that occurs in conventional automatic tooth separation of 3D scan data. [Figure 4] FIG. 4 is a diagram showing an example of a tooth separation error that occurs in conventional automatic tooth separation of 3D scan data. [Diagram 5] FIG. 5 is a diagram showing an example of a tooth separation error that occurs in conventional automatic tooth separation of 3D scan data. [Figure 6] FIG. 6 is a diagram showing an example of a tooth separation error that occurs in conventional automatic tooth separation of 3D scan data. [Figure 7] FIG. 7 is a simplified flowchart illustrating a method for automatically separating teeth from 3D scan data using tooth boundary curves according to an embodiment of the present invention. [Figure 8] FIG. 8 is a detailed flow chart of the automatic tooth separation method of FIG. [Figure 9] FIG. 9 is a conceptual diagram showing the tooth detection process of FIG. [Figure 10] FIG. 10 is a diagram showing an example of input for the mesh imaging of FIG. [Figure 11] FIG. 11 is a diagram showing an example of the mesh imaging output of FIG. [Figure 12] FIG. 12 is a diagram showing an example of an output of the first artificial intelligence neural network of FIG. [Figure 13] FIG. 13 is a diagram showing an example of an output of the first artificial intelligence neural network of FIG. [Figure 14] FIG. 14 is a conceptual diagram illustrating the parameterization process of FIG. [Figure 15] FIG. 15 illustrates the curvature data used in the parameterization process of FIG. [Figure 16] FIG. 16 is a diagram showing tooth scan data generated by extracting a portion of the mesh in FIG. [Figure 17] FIG. 17 is a diagram showing tooth mapping data generated by parameterizing the curvature data of the tooth scan data in the manner shown in FIG. [Figure 18] FIG. 18 is a conceptual diagram showing the boundary curve generation process of FIG. [Figure 19] FIG. 19 is a conceptual diagram specifically illustrating the operation of the second artificial intelligence neural network of FIG. [Figure 20] FIG. 20 shows the final tooth segmentation result produced by the boundary 3D mapping of FIG. [Figure 21] FIG. 21 is a detailed flowchart illustrating a method for automatically separating teeth from 3D scan data using tooth boundary curves according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0026] With respect to the embodiments of the present invention set forth herein, specific structural or functional descriptions are merely exemplary for purposes of describing the embodiments of the present invention, and the example embodiments of the present invention may be embodied in various forms and should not be construed as being limited to the embodiments set forth herein.
[0027] The present invention can be modified in various ways and can have various forms, and specific embodiments are shown by way of example in the drawings and will be described in detail herein, but it is to be understood that this is not intended to limit the invention to the specific disclosed forms, but rather to include all modifications, equivalents, and alternatives falling within the spirit and technical scope of the present invention.
[0028] Terms such as "first" and "second" are used to describe various components, but the components should not be limited by the terms. The terms are used only to distinguish one component from another. For example, a first component can be referred to as a second component, and similarly, a second component can be referred to as a first component, without departing from the scope of the present invention.
[0029] When an element is said to be "coupled" or "connected" to another element, it should be understood that it may be directly coupled or connected to the other element, but there may also be other elements in between. On the other hand, when an element is said to be "directly coupled" or "directly connected" to another element, it should be understood that there are no other elements in between. Other expressions describing the relationship between elements, such as "between" and "immediately between," or "adjacent to" and "directly adjacent to," should be interpreted similarly.
[0030] The terms used in this application are merely used to describe certain embodiments and are not intended to limit the present invention. A singular expression includes a plural expression unless the context clearly indicates otherwise. In this application, the terms "include" or "have" are intended to specify the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, and should be understood not to preclude the presence or additional possibility of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.
[0031] Unless otherwise defined, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art to which this invention pertains. Terms as defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning they have in the context of the relevant art, and not as being idealized or overly formal, unless expressly defined in this application.
[0032] However, when an embodiment can be implemented differently, the functions or acts specified in a particular block may occur differently from the sequence specified in the flowchart. For example, two consecutive blocks may actually be performed substantially simultaneously, or the blocks may be performed in reverse depending on the functions or acts involved.
[0033] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The same components in the drawings are designated by the same reference numerals, and duplicated descriptions of the same components will be omitted.
[0034] 7 is a flow chart showing a method for automatically separating teeth from 3D scan data using tooth boundary curves according to an embodiment of the present invention, and FIG. 8 is a flow chart showing the method for automatically separating teeth from FIG.
[0035] As shown in Figures 7 and 8, the present invention uses mesh parameterization and deep learning to fully automatically segment individual teeth from 3D scan data, for example, the scan data may consist of a polygon mesh, and therefore the scan data may also be referred to as a mesh.
[0036] The method for automatic tooth segmentation from 3D scan data according to the present embodiment allows for fully automatic segmentation of individual teeth from the scan data using a first artificial intelligence neural network (AI1) for tooth detection and a second artificial intelligence neural network (AI2) for mesh parameterization and tooth border generation.
[0037] The method for automatically separating teeth from 3D scan data according to this embodiment includes a tooth detection process for detecting teeth from an image of 3D oral cavity scan data, a process for parameterizing a mesh of a certain area based on the teeth found in the tooth detection process, and a process for generating a boundary from the parameterized image to generate a boundary curve for the scan data.
[0038] In general, a deep learning network can only receive data of a certain size (shaped data). The scan data is represented as a 3D polygon mesh, and there is a problem that the number of mesh points, edges, and cells varies from data to data. Therefore, in order to improve the accuracy of a deep learning network that receives the scan data as input, a process of converting the scan data into shaped data is necessary. In the present invention, mesh parameterization can be used to convert the scan data into shaped data.
[0039] The mesh parameterization means one-to-one mapping of points constituting a 3D mesh to another predetermined space. In order to map from a space in which a mesh exists to the other predetermined space (parameter space), the boundary of the space in which the mesh exists is first mapped to the boundary of the parameter space. The internal points excluding the boundary of the space in which the mesh exists are mapped to the inside of the boundary of the parameter space, and a method of maintaining the topology of the mesh as much as possible using an energy function can be used.
[0040] The mesh parameterization allows the scan data to be mapped into two-dimensional space, thereby properly forming the shaped data for use in the deep learning.
[0041] The mesh parameterization can improve the accuracy of the deep learning network of the second artificial intelligence neural network (AI2) that generates tooth boundary curves. As shown in FIG. 8, the method for automatically separating teeth from 3D scan data using a tooth boundary curve according to this embodiment includes the steps of detecting teeth in the scan data using a first artificial intelligence neural network (S200), extracting tooth scan data from the scan data based on the tooth detection result (S300), generating tooth mapping data corresponding to a predetermined space based on the tooth scan data (S400), inputting the tooth mapping data into a second artificial intelligence neural network to generate a tooth boundary curve (S500), and mapping the tooth boundary curve to the scan data (S600).
[0042] For example, the method for automatically separating teeth from 3D scan data using tooth boundary curves further includes converting the 3D scan data into a 2D image (S100). The method for automatically separating teeth from 3D scan data according to the present embodiment is performed by a computing device.
[0043] Fig. 9 is a conceptual diagram showing the tooth detection process of Fig. 7. Fig. 10 is a diagram showing an example of an input for mesh imaging of Fig. 9. Fig. 11 is a diagram showing an example of an output of mesh imaging (S200) of Fig. 9. Fig. 12 is a diagram showing an example of an output of the first artificial intelligence neural network (AI1) of Fig. 9. Fig. 13 is a diagram showing an example of an output of the first artificial intelligence neural network (AI1) of Fig. 9.
[0044] 7 to 13, the tooth detection process includes converting the tooth scan data into a two-dimensional image (S100), which is also called mesh imaging.
[0045] In this embodiment, the input of the first artificial intelligence neural network (AI1) is a two-dimensional image, so the process includes converting the three-dimensional scan data into the two-dimensional image.
[0046] For example, the two-dimensional image is RGB grayscale data including red, green, and blue tones. For example, the two-dimensional image is black and white grayscale data. For example, the two-dimensional image is a depth map including depth information. The depth map is generated based on any one of the maximum value, minimum value, standard deviation, and average in the two-dimensional coordinates of the 3D scan data.
[0047] The step of converting the scan data into the two-dimensional image includes a principal axis normalization step, which is to set a spatial orientation by principal axis analysis, in which the principal axes formed by points in the scan data are analyzed to determine first, second, and third principal axes that are perpendicular to each other.
[0048] The longest axis among the first, second and third principal axes extracted in the principal axis analysis is determined to be the left-right direction of the U-shape of the tooth. The shortest axis among the first, second and third principal axes is determined to be the up-down direction of the U-shape. The second longest axis among the first, second and third principal axes is determined to be the front-back direction of the U-shape.
[0049] The principal axis normalization step aligns the scan data to maximize visibility of the occlusal surfaces of the teeth. The step of detecting the teeth in the scan data includes a step (S220) of the first artificial intelligence neural network (AI1) receiving the 2D image and outputting an output 2D image including 2D tooth detection information, and a step (S240) of converting the 2D tooth detection information into 3D tooth detection information.
[0050] Here, the tooth detection information is a tooth bounding box surrounding each tooth. Here, the tooth detection information is tooth feature points representing the surface center points of each tooth.
[0051] That is, the first artificial intelligence neural network (AI1) receives the two-dimensional image of FIG. 11, generates a two-dimensional bounding box (2D Point) (S220), and converts the two-dimensional bounding box (2D Point) into three dimensions (S240) to generate the three-dimensional bounding box (3D Point) in FIG. 12.
[0052] Separately, the first artificial intelligence neural network (AI1) receives the two-dimensional image of FIG. 11, generates two-dimensional tooth feature points (2D Points) (S220), and converts the two-dimensional tooth feature points (2D Points) into three dimensions (S240) to generate three-dimensional tooth feature points (3D Points) in FIG. 13.
[0053] As described above, the tooth feature points may be automatically generated through the first artificial intelligence neural network (AI1), or may be generated using the bounding box automatically generated through the first artificial intelligence neural network (AI1). For example, the tooth feature points may be generated by mapping the center point of the tooth bounding box onto the tooth surface.
[0054] If the scan data includes 14 teeth, 14 bounding boxes are generated by the first artificial intelligence neural network (AI1), the size of the bounding boxes varies depending on the tooth. Similarly, if the scan data includes 14 teeth, 14 feature points are generated by the first artificial intelligence neural network (AI1).
[0055] The first artificial intelligence neural network (AI1) detects individual teeth by learning a bounding box that surrounds the tooth based on a one-stage object detector. Here, the ROI (region of interest) of the bounding box input as learning data to the first artificial intelligence neural network (AI1) can be set to a size that fits the tooth properly. If the size of the ROI is too small or too large, the performance of the tooth detection will be reduced.
[0056] Fig. 14 is a conceptual diagram showing the parameterization process of Fig. 7. Fig. 15 is a diagram showing curvature data used in the parameterization process of Fig. 7. Fig. 16 is a diagram showing tooth scan data generated by extracting a mesh portion of Fig. 14. Fig. 17 is a diagram showing tooth mapping data generated by parameterizing the curvature data of the tooth scan data of Fig. 14. As shown in Figures 7 to 14, the parameterization process includes a step of extracting tooth scan data from the scan data based on the tooth detection result (S300), and a step of generating tooth mapping data corresponding to a predetermined space based on the tooth scan data (S400).
[0057] For example, when a tooth bounding box surrounding one tooth is generated in the step of detecting the tooth in the scan data, the step of extracting the tooth scan data extracts the tooth scan data corresponding to one tooth using a tooth cutting box corresponding to the tooth bounding box. Here, the tooth cutting box includes the tooth bounding box and is larger than the tooth bounding box. As shown in FIG. 16, it can be seen that the tooth cutting box is larger than the tooth bounding box. The tooth bounding box is an area that closely corresponds to the boundary of the tooth in order to detect the tooth, and the tooth cutting box is an area for generating the boundary curve after cutting out the scan data in tooth units. Therefore, the tooth cutting box does not necessarily need to be the same as the tooth bounding box, and it is preferable that the tooth cutting box includes the bounding box and is slightly larger than the tooth bounding box.
[0058] For example, when a tooth feature point indicating a surface center point of one tooth is generated in the step of detecting the teeth in the scan data, the step of extracting the tooth scan data extracts the tooth scan data corresponding to one tooth using a tooth cut area formed based on the tooth feature point, for example, the tooth cut area is a spherical area having the tooth feature point at its center.
[0059] The parameterization process includes converting the dental scan data into curvature data representing a curvature value for each point in the dental scan data. 15 shows curvature data indicating the curvature value of each point in the scan data. For example, the curvature data may indicate a maximum curvature value, a minimum curvature value, a Gaussian curvature value, and a mean curvature value. In FIG. 15, the curvature data indicates the minimum curvature value of each point.
[0060] The curvature value is relatively constant for the upper surface of the tooth. In contrast, the curvature value varies greatly at the boundary between teeth. Therefore, the curvature value represents the area where the tooth and gums come into contact, and the tooth boundary curve can be generated with high accuracy using the curvature value.
[0061] In conventional curvature data, the larger the curvature value, the more black the color, and the smaller the curvature value, the more white the color. In this embodiment, a step of inverting the gradation of the curvature data may be further performed, such that the large minimum curvature value of each point is represented by white, and the small minimum curvature value of each point is represented by black. Figure 15 shows a diagram in which the gradation of the curvature data is inverted, in which the part with the large minimum curvature value is represented by white, and the part with the small minimum curvature value is represented by black. For example, if the gradation of the curvature data is not inverted, the large minimum curvature value will be represented by black, and the small minimum curvature value will be represented by white, as opposed to Figure 15.
[0062] That is, if the curvature data has a value that does not invert the minimum curvature value, meaningful parts (parts with a large minimum curvature value) such as the part where the teeth and gums touch will appear black. However, in this embodiment, the grayscale inversion causes meaningful parts (parts with a large minimum curvature value) such as the part where the teeth and gums touch to appear white.
[0063] For example, if there is a hole in the scan data and there is no point in the corresponding portion, so that there is no curvature value, the conventional curvature data displays the portion without curvature value in black. If the gradation of the curvature data is not inverted, the meaningful portion (e.g., the portion where the tooth and gums touch) will appear in black, and in this case, the portion without curvature value will also be displayed in black, resulting in a problem that the portion without curvature value will be mistaken for a meaningful portion. In contrast, if the gradation of the curvature data is inverted, the meaningful portion (e.g., the portion where the tooth and gums touch) will appear in white, and in this case, the hole portion without curvature value will not be recognized as a meaningful portion, so that it is possible to eliminate erroneous judgment due to the portion without curvature value.
[0064] The tooth mapping data generating step (S400) maps the curvature data of the tooth scan data corresponding to each tooth into the predetermined space to generate the tooth mapping data. Figure 17 shows an example of the tooth mapping data. The tooth mapping data can be understood as the curvature data of the tooth scan data corresponding to one tooth being mapped into a predetermined space and converted into two-dimensional data.
[0065] For example, the predetermined space is a regular square. Since the tooth bounding box and the tooth cutting box have different shapes and sizes depending on the tooth, the tooth scan data has different shapes and sizes depending on the tooth. In contrast, the shape and size of the predetermined space can be set to be constant regardless of the tooth. In this case, the shape and size of the tooth mapping data is constant regardless of the tooth.
[0066] The tooth mapping data is preferably set to the same size since it is the input for the second artificial intelligence neural network (AI2). Fig. 18 is a conceptual diagram showing the boundary curve generation process of Fig. 7. Fig. 19 is a conceptual diagram specifically showing the operation of the second artificial intelligence neural network of Fig. 18. Fig. 20 is a diagram showing the final tooth segmentation result generated by the boundary 3D mapping of Fig. 18.
[0067] As shown in Figures 7 to 20, the inputs of the second artificial intelligence neural network (AI2) are the mapping data and the tooth bounding box, and the output of the second artificial intelligence neural network (AI2) is the tooth boundary curve.
[0068] Here, the tooth bounding box is used as initial contour information (initial box definition in FIG. 19) in the second artificial intelligence neural network (AI2). For example, the step of generating the tooth boundary curve (S500) includes a step of rotating the tooth bounding box to generate a diamond contour, for example, rotating the tooth bounding box by 45 degrees to generate the diamond contour.
[0069] For example, the step of generating the tooth boundary curve (S500) further includes a step of generating an extreme point by moving the apex position of the diamond contour in the mapping data (deformation). The extreme point can be moved according to the position where the curvature value is small or large in the mapping data.
[0070] For example, the step of generating the tooth boundary curve (S500) further includes a step of generating an octagonal contour passing through the extreme points based on the extreme points. For example, the step of generating the tooth boundary curve (S500) further includes a step of moving the octagonal contour within the mapping data by the curvature value to generate the boundary curve (Object shape).
[0071] The tooth boundary curve, which is the output of the second artificial intelligence neural network (AI2), is 3D mapped onto the scan data (S600) to complete the tooth segmentation. Figure 20 shows the final tooth segmentation result.
[0072] According to this embodiment, the effort required to separate teeth from scan data can be reduced by automatically performing the process through an artificial intelligence neural network for tooth detection (AI1) and an artificial intelligence neural network for mesh parameterization and tooth boundary generation (AI2). Since tooth boundary curves, not tooth surfaces, are accurately generated using the tooth detection artificial intelligence neural network (AI1), the mesh parameterization, and the tooth boundary generation artificial intelligence neural network (AI2), problems such as partial tooth regions not being masked, incorrect regions that are not teeth being determined as teeth, masks not being formed within one tooth and thus invading adjacent teeth or gums, multiple teeth being determined as one tooth, etc. can be prevented, thereby improving the accuracy of the tooth separation.
[0073] In addition, the artificial intelligence neural network for tooth boundary generation (AI2) can generate the tooth boundary curve that separates the teeth and gums at once by modifying the initially input contour data, thereby improving the accuracy of the tooth separation.
[0074] In addition, even if scan data for partial teeth rather than the entire teeth is input, the tooth detection artificial intelligence neural network (AI1) is used to accurately determine the area of each tooth and generate boundary curves for each tooth-specific scan data corresponding to the tooth area, so that tooth separation can be performed using the same model (AI1, AI2) regardless of whether the scan data represents the entire teeth or partial teeth.
[0075] FIG. 21 is a detailed flowchart illustrating a method for automatically separating teeth from 3D scan data using tooth boundary curves according to an embodiment of the present invention. The automatic tooth separation method according to this embodiment is similar to the automatic tooth separation method of Figures 1 to 20, except for the input of the first artificial intelligence neural network, so the same components are given the same reference symbols and duplicated descriptions are omitted.
[0076] As shown in FIG. 21 , the method for automatically separating teeth from 3D scan data using a tooth boundary curve according to this embodiment includes the steps of detecting teeth in the scan data using a first artificial intelligence neural network (S250), extracting tooth scan data from the scan data based on the tooth detection result (S300), generating tooth mapping data corresponding to a predetermined space based on the tooth scan data (S400), inputting the tooth mapping data into a second artificial intelligence neural network to generate a tooth boundary curve (S500), and mapping the tooth boundary curve to the scan data (S600).
[0077] The method for automatically separating teeth from 3D scan data according to the present embodiment is performed by a computing device. In the automatic tooth separation method of Figures 1 to 20, the input of the first artificial intelligence neural network (AI1) is shown to be a two-dimensional image, but in this embodiment, the input of the first artificial intelligence neural network (AI1) is shown to be three-dimensional scan data.
[0078] In this embodiment, the first artificial intelligence neural network (AI1) receives the 3D scan data and outputs 3D tooth detection information, where the tooth detection information is a tooth bounding box or tooth feature points.
[0079] According to this embodiment, the effort required to separate teeth from scan data can be reduced by automatically performing the process using an artificial intelligence neural network for tooth detection (AI1) and an artificial intelligence neural network for mesh parameterization and tooth boundary generation (AI2).
[0080] Since tooth boundary curves, not tooth surfaces, are accurately generated using the tooth detection artificial intelligence neural network (AI1), the mesh parameterization, and the tooth boundary generation artificial intelligence neural network (AI2), problems such as partial tooth regions not being masked, incorrect regions that are not teeth being determined as teeth, masks not being formed within one tooth and thus invading adjacent teeth or gums, multiple teeth being determined as one tooth, etc. can be prevented, thereby improving the accuracy of the tooth separation.
[0081] In addition, the artificial intelligence neural network for tooth boundary generation (AI2) can generate the tooth boundary curve that separates the teeth and gums at once by modifying the initially input contour data, thereby improving the accuracy of the tooth separation.
[0082] In addition, even if scan data for partial teeth rather than the entire teeth is input, the tooth detection artificial intelligence neural network (AI1) is used to accurately determine the area of each tooth and generate boundary curves for each tooth-specific scan data corresponding to the tooth area, so that tooth separation can be performed using the same model (AI1, AI2) regardless of whether the scan data represents the entire teeth or partial teeth.
[0083] According to an embodiment of the present invention, there is provided a computer readable recording medium having a program recorded thereon for executing the method for automatically separating teeth from 3D scan data using tooth boundary curves according to the embodiment. The method can be written as a computer executable program and can be implemented in a general-purpose digital computer that runs the program using a computer readable medium. In addition, data structures used in the above-mentioned method can be recorded in the computer readable medium through various means. The computer readable medium can include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded in the medium can be those specifically designed and constructed for the present invention, or those that are known and usable by those of ordinary skill in the art 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 floptical disks, and hardware devices specifically configured to store and execute program instructions, such as ROMs, RAMs, flash memories, and the like. Examples of program instructions include not only machine language codes, such as those created by a compiler, but also high-level language codes executed by a computer using an interpreter, and the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of this invention. The method for automatically separating teeth from 3D scan data using tooth boundary curves may also be implemented in the form of a computer program or application that is stored in a recording medium and executed by a computer.
[0084] [Industrial Applicability] The present invention relates to a method for automatically separating teeth from 3D scan data using tooth boundary curves, and a computer-readable recording medium having a program for executing the method on a computer, which can reduce the effort required for separating teeth and improve accuracy.
[0085] Although the present invention has been described above with reference to preferred embodiments, those skilled in the art will appreciate that various modifications and variations of the present invention may be made without departing from the spirit and scope of the present invention as set forth in the following claims.
Claims
1. detecting teeth in the scan data using a first artificial intelligence neural network; extracting tooth scan data from the scan data based on the tooth detection result; generating dental mapping data corresponding to a predetermined space based on the dental scan data; inputting the tooth mapping data into a second artificial intelligence neural network to generate tooth boundary curves; mapping the tooth boundary curve onto the scan data.
2. converting the three-dimensional scan data into a two-dimensional image; The two-dimensional image is input to the first artificial intelligence neural network.
2. A method for automatically separating teeth from three-dimensional scanned data using the tooth boundary curve according to claim 1.
3. The step of detecting teeth in the scan data includes: the first artificial intelligence neural network receiving the two-dimensional image and outputting an output two-dimensional image including two-dimensional tooth detection information; and converting the two-dimensional tooth detection information into three-dimensional tooth detection information.
3. A method for automatically separating teeth from three-dimensional scan data using the tooth boundary curve according to claim 2.
4. and detecting the teeth in the scan data includes generating a tooth bounding box that encloses a tooth.
2. A method for automatically separating teeth from three-dimensional scanned data using the tooth boundary curve according to claim 1.
5. The step of extracting tooth scan data comprises extracting the tooth scan data corresponding to one tooth using a tooth cutting box corresponding to the tooth bounding box. A method for automatically separating teeth from three-dimensional scan data using the tooth boundary curve according to claim 4.
6. The tooth cutting box includes the tooth bounding box and is larger than the tooth bounding box. A method for automatically separating teeth from three-dimensional scan data using the tooth boundary curve according to claim 5.
7. and detecting the teeth in the scan data includes mapping center points of the tooth bounding boxes onto a tooth surface to generate tooth feature points. A method for automatically separating teeth from three-dimensional scan data using the tooth boundary curve according to claim 4.
8. The step of extracting the tooth scan data includes extracting the tooth scan data corresponding to one tooth using a tooth cut area formed based on the tooth feature points. A method for automatically separating teeth from three-dimensional scan data using the tooth boundary curve according to claim 7.
9. converting the dental scan data into curvature data representing a curvature value for each point in the dental scan data. A method for automatically separating teeth from three-dimensional scan data using the tooth boundary curve according to claim 4.
10. The curvature data represents a minimum curvature value for each of the points. A method for automatically separating teeth from three-dimensional scan data using the tooth boundary curve according to claim 9.
11. 11. The method of claim 10, further comprising a step of inverting the gradation of the curvature data so that the minimum curvature value of each point is large and is represented in white, and the minimum curvature value of each point is small and is represented in black.
12. and generating the tooth mapping data by mapping the curvature data of the tooth scan data into the predetermined space to generate the tooth mapping data. A method for automatically separating teeth from three-dimensional scan data using the tooth boundary curve according to claim 9.
13. an input of the second artificial intelligence neural network is the mapping data and the tooth bounding box, and an output of the second artificial intelligence neural network is the tooth boundary curve. A method for automatically separating teeth from 3D scan data using the tooth boundary curve according to claim 12.
14. The step of generating the tooth boundary curve includes a step of rotating the tooth bounding box to generate a diamond outline. A method for automatically separating teeth from 3D scan data using the tooth boundary curve according to claim 13.
15. The step of generating the tooth boundary curve further includes a step of moving a position of a vertex of the diamond contour in the mapping data by the curvature value to generate a pole point. A method for automatically separating teeth from 3D scan data using the tooth boundary curve according to claim 14.
16. The step of generating the tooth boundary curve further includes a step of generating an octagonal contour passing through the extreme points based on the extreme points. A method for automatically separating teeth from 3D scan data using the tooth boundary curve according to claim 15.
17. The step of generating the tooth boundary curve further includes a step of moving the octagonal contour within the mapping data by the curvature value to generate the boundary curve. A method for automatically separating teeth from 3D scan data using the tooth boundary curve according to claim 16.
18. The first artificial intelligence neural network receives the three-dimensional scan data and outputs three-dimensional tooth detection information.
2. A method for automatically separating teeth from three-dimensional scanned data using the tooth boundary curve according to claim 1.
19. A computer-readable recording medium having a program recorded thereon for causing a computer to execute the method according to any one of claims 1 to 18.
Citation Information
Patent Citations
Method and system for hybrid mesh segmentation
JP2019524367A
Protection cover for electricity charging inlet of vehicle
KR1020230040789A
Apparatus and method for automatic registration of oral scan data and computed tomography image using crown segmentation of oral scan data
KR102273438B1
Teeth segmentation using neural networks
WO2021155230A1
