WO2024071571A1

WO2024071571A1 - Method for segmenting three-dimensional oral model and system therefor

Info

Publication number: WO2024071571A1
Application number: PCT/KR2023/008155
Authority: WO
Inventors: 최병선; 김완; 용태훈; 안홍기
Original assignee: 오스템임플란트 주식회사
Priority date: 2022-09-27
Filing date: 2023-06-14
Publication date: 2024-04-04
Also published as: KR20240043443A

Abstract

Provided are a method for segmenting a three-dimensional oral model, and an apparatus to which the method is applied. The method for segmenting a three-dimensional oral model, according to one embodiment, may comprises steps in which: a first artificial neural network having received three-dimensional teeth scan data outputs region segmentation data representing a region segmentation three-dimensional oral model in which a gum region and a crown region of a three-dimensional oral model represented by the three-dimensional teeth scan data are identified; the region segmentation three-dimensional oral model according to the region segmentation data is analyzed to identify an occlusal surface of the region segmentation three-dimensional oral model, and generates an occlusal crown image representing only the crown region of the occlusal surface; a second artificial neural network having received the occlusal crown image outputs center point data indicating the center point of a crown region of each tooth in the occlusal crown image; and a third artificial neural network having received the three-dimensional teeth scan data and the center point data outputs segmentation data representing a three-dimensional segmentation model in which the crown region of each tooth included in the three-dimensional oral model is segmented by tooth.

Description

Segmentation method and system for 3D oral model

The present disclosure relates to a segmentation method of a three-dimensional oral model. More specifically, it relates to a segmentation method and device that is applied to a 3D oral model and partitions a 3D area for each tooth.

Digital dentistry refers to IT technology that helps treat patients who visit the dentist by digitizing and analyzing the patient's oral information. For example, digital dentistry can also be used in orthodontic treatment. When establishing a dental treatment plan using digital dentistry, an improved tooth arrangement can be predicted reliably and conveniently compared to the current tooth condition. Additionally, when using a tooth model provided in a digital dentistry environment, teeth can be freely moved in all directions in a virtual space implemented as a three-dimensional space, allowing various orthodontic treatment plans to be established.

In a digital dentistry environment, a 3D oral model in which gums or other anatomical structures and teeth are realized in 3D is provided to the surgeon, and the surgeon can select teeth to be treated using the 3D oral model. Additionally, the surgeon can edit the teeth to be treated in the 3D oral model.

However, in order for the operator to edit a specific tooth in a 3D oral model, the 3D area of each crown included in the 3D oral model must be distinguished from the gums or other anatomical structures. In other words, 3D segmentation of the tooth area for the 3D oral model is required.

Existing automated tooth segmentation technology has many cases where segmentation is applied to 2D images such as CT cross-sectional images, but it is difficult to apply to 3D oral models created as a result of 3D oral scans. In other words, technology for automatically segmenting a 3D oral model is not available, so segmenting a 3D model requires considerable time and effort, such as having the operator manually input points indicating the area of each tooth on the 3D oral model. must be invested.

The technical problem that the present disclosure aims to solve is to provide a method for automatically segmenting the crown area of a 3D oral model and a device to which the method is applied. The technical problems of the present disclosure are not limited to the technical problems mentioned above, and other technical problems not mentioned can be clearly understood by those skilled in the art from the description below.

A segmentation method of a three-dimensional oral model according to an embodiment for solving the above technical problem is performed by a computing system, comprising the steps of acquiring at least one tooth scan data, and the three-dimensional tooth scan data. A step of outputting region separation data representing the three-dimensional oral model, separating the gum region and crown region of the three-dimensional oral model generated by the input first artificial neural network using at least one three-dimensional tooth scan data; , analyzing the region-separated three-dimensional oral model to identify the occlusal surface of the three-dimensional oral model, generating an occlusal crown image representing only the crown region of the occlusal surface, and receiving the occlusal crown image as input. 2 An artificial neural network outputs center point data indicating the center point of the crown area of each tooth on the occlusal crown image, and a third artificial neural network receiving the 3D tooth scan data and the center point data outputs the 3D oral cavity. It may include outputting segmentation data representing a 3D segmentation model in which the crown area of each tooth included in the model is segmented for each tooth.

In one embodiment, the step of outputting the region separation data includes downsampling the 3D tooth scan data, and inputting the downsampled 3D tooth scan data into the first artificial neural network. It may include steps.

In one embodiment, the step of outputting the region separation data may include masking the gum area of the region separation 3D oral model to have the first attribute, or masking the crown area of the region separation 3D oral model to have the second attribute. It may include performing masking processing to have , adjusting the region separation data so that the region separation data reflects the results of the masking process, and outputting the adjusted region separation data.

In one embodiment, the step of generating the occlusal crown image includes analyzing the region-separated 3D oral cavity model according to the region separation data to identify the occlusal surface of the region-separated 3D oral cavity model, and the occlusal surface It will include positioning a virtual camera to have a field-of-view (FOV) perpendicular to the You can. At this time, the step of outputting the region separation data includes masking the gum region of the region separation 3D oral model to have a first attribute, or masking the crown region of the region separation 3D oral model to have a second attribute. It may include performing masking processing and adjusting the region separation data so that the region separation data reflects the results of the masking process. In addition, the step of generating the occlusal crown image using the image of the occlusal surface captured by the virtual camera includes transparently processing the gum area in the image of the occlusal surface using the result of the masking process, It may include generating the occlusal surface crown image.

In one embodiment, the step of outputting the center point data may include selecting the second artificial neural network from among a plurality of candidate artificial neural networks using the shape of the crown area of the occlusal crown image.

In one embodiment, the step of outputting the center point data includes selecting the second artificial neural network from among a plurality of candidate artificial neural networks using the shape of the crown area of the pre-designated area of the occlusal crown image. can do.

In one embodiment, the step of outputting the center point data includes calculating reliability of the center point data using output data of the second artificial neural network, and when the reliability is less than a reference value, the center point on the occlusal crown image It may include outputting a user interface for manual input.

In one embodiment, the step of outputting the segmentation data includes down-sampling the 3D tooth scan data, and combining the downsampled 3D tooth scan data and the center point data with the third artificial It may include inputting into a neural network.

In one embodiment, the step of outputting the segmentation data includes masking each crown region so that each crown region of the three-dimensional segmentation model according to the segmentation data is visually distinguished from each other, and the segmentation data is masked. It may include adjusting the segmentation data to reflect a result of processing, and outputting the adjusted segmentation data. At this time, the step of outputting the adjusted segmentation data includes identifying each crown position using the result of the masking process, and based on the identified crown position, a dental formula number for each crown region. It may include labeling and outputting labeling result data for the dental formula number for each crown region along with the adjusted segmentation data.

In one embodiment, the segmentation method of the 3D oral model may further include generating an output screen for 3D orthodontic simulation using the segmentation data.

A segmentation method of a 3D oral model according to another embodiment for solving the above technical problem includes acquiring at least one 3D tooth scan data, and a first artificial neural network receiving the 3D tooth scan data 3D A step of outputting region separation data representing a region-separated three-dimensional oral model in which the gum region and crown region of the oral model are separated, analyzing the region-separated three-dimensional oral model to determine the occlusal surface of the region-separated three-dimensional oral model. identifying and generating an occlusal crown image representing only the crown area of the occlusal surface, and a second artificial neural network receiving the occlusal crown image determines the center point of the crown area of each tooth on the occlusal crown image. A step of outputting pointing center point data, and a third artificial neural network receiving the region separation data and the center point data representing a 3D segmentation model in which the crown area of each tooth included in the 3D oral model is segmented for each tooth. It may include the step of outputting segmentation data. A segmentation device for a three-dimensional oral model according to another embodiment for solving the above technical problem includes at least one three-dimensional tooth scan data, a first artificial neural network, a second artificial neural network, a third artificial neural network, and a three-dimensional oral model. It may include a memory into which a segmentation program is loaded, and a processor through which the 3D oral model segmentation program is executed. At this time, in the 3D tooth segmentation program, the first artificial neural network that receives the 3D tooth scan data separates the area where the gum area and the crown area of the 3D oral model are separated and separates the area representing the 3D oral model. Instructions for outputting data, analyzing the region-separated 3D oral model to identify the occlusal surface of the 3D oral model, and generating an occlusal crown image representing only the crown area of the occlusal surface; , an instruction for the second artificial neural network receiving the occlusal crown image to output center point data indicating the center point of the crown area of each tooth on the occlusal crown image, and the three-dimensional tooth scan data and the center point data. The received third artificial neural network may include instructions for outputting segmentation data representing a 3D segmentation model in which the crown region of each tooth included in the 3D oral model is segmented for each tooth.

In one embodiment, the 3D tooth segmentation program may further include instructions for generating an output screen for 3D orthodontic simulation using the output segmentation data.

According to some embodiments of the present disclosure, segmentation results for each tooth region of a 3D oral model can be provided in a highly automated manner.

According to some embodiments of the present disclosure, by combining a plurality of machine-learned artificial neural networks and an image processing algorithm, accurate segmentation results for each tooth region of a 3D oral model can be provided.

According to some embodiments of the present disclosure, while using high-resolution 3D tooth scan data, the complexity of the hyper parameter aspect of the artificial neural network is prevented from excessively increasing, thereby improving the machine learning requirements of the artificial neural network. costs can be reduced.

According to this embodiment, segmentation errors can be minimized by inputting data that has completed preprocessing required for each step in each step.

1 illustrates an example environment in which a segmentation system for a three-dimensional oral model can be applied, according to an embodiment of the present disclosure.

Figure 2 is a flowchart of a segmentation method of a 3D oral cavity model according to another embodiment of the present disclosure.

Figure 3 is a flowchart of a segmentation method of a 3D oral cavity model according to another embodiment of the present disclosure.

FIG. 4 is a diagram illustrating down-sampling of 3D scan data performed in some embodiments of the present disclosure.

FIG. 5 is a diagram for explaining the operation of a first artificial neural network performed in some embodiments of the present disclosure.

6 to 7 are diagrams for explaining an occlusal crown image generation operation performed in some embodiments of the present disclosure.

8 to 9 are diagrams for explaining a center point data generation operation performed in some embodiments of the present disclosure.

10 to 11 are flowcharts for explaining the operation of a third artificial neural network performed in some embodiments.

FIG. 12 is a conceptual diagram of the segmentation method of a 3D oral model described with reference to FIG. 2.

Figure 13 is a hardware configuration diagram of an oral model segmentation device according to another embodiment of the present disclosure.

Hereinafter, preferred embodiments of the present disclosure will be described in detail with reference to the attached drawings. The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the technical idea of the present invention is not limited to the following embodiments and may be implemented in various different forms. The following examples are merely intended to complete the technical idea of the present invention and to be used in the technical field to which the present invention pertains. It is provided to fully inform those skilled in the art of the scope of the present invention, and the technical idea of the present invention is only defined by the scope of the claims. In describing the present disclosure, if it is determined that a detailed description of a related known configuration or function may obscure the gist of the present invention, the detailed description will be omitted.

A segmentation method of a 3D oral model according to an embodiment of the present disclosure will be described with reference to FIG. 1. As shown in FIG. 1, the oral model segmentation system 100 according to this embodiment can perform segmentation of the oral cavity model through interaction with the user device 220.

The oral model segmentation system 100 may acquire 3D tooth scan data stored from the scan data storage device 230, analyze the oral model, and output segmentation data using the analyzed results. The 3D tooth scan data may be scan data captured by a 3D scanner.

Additionally, as the segmentation function of the 3D oral model is implemented on a cloud computing node, the segmentation function of the 3D oral model may be divided and performed on a plurality of cloud computing nodes for each detailed module.

Hereinafter, in describing the method according to this embodiment, description of the subject performing some operations may be omitted. At this time, it should be understood that the subject performing the corresponding operation is the computing device. Hereinafter, in this specification, the computing device will be referred to as an ‘oral model segmentation device.’

Figure 2 is a flowchart of a 3D oral model segmentation method according to an embodiment of the present invention.

Referring to Figure 2, the oral model segmentation system 100 can acquire 3D tooth scan data (S110). The 3D tooth scan data may be received from a 3D scanner connected to the oral model segmentation system 100. The 3D scanner may be an oral scanner that photographs the patient's oral cavity. The 3D tooth scan data may be understood as data for rendering a 3D oral model.

The 3D tooth scan data may be input to a first machine-learned artificial neural network. The first artificial neural network is machine-learned to distinguish between the gum area and the crown area of the 3D oral model. For example, the first artificial neural network may be supervised learning using learning data in which the crown region is masked from a 3D oral model.

Meanwhile, if the 3D tooth scan data is for rendering a high-resolution 3D oral model, the data size will be significant. If so, the first artificial neural network that receives the 3D tooth scan data should have a high structural complexity. For example, the first artificial neural network that receives the 3D tooth scan data with a large data size will have a large number of layers and a large number of nodes in each layer.

That is, the first artificial neural network will have high complexity in terms of hyperparameters. In that case, in order to machine learn the first artificial neural network to have a level of performance that can be commercialized, a very large amount of learning data will be needed. Considering that it costs a considerable amount of money to prepare learning data with the crown area masked in a 3D oral model, and that machine learning a large number of learning data also results in a significant computing load, the first It is undesirable for artificial neural networks to have a high level of structural complexity in terms of economic efficiency.

In order to solve the above-mentioned problems caused by high-resolution 3D tooth scan data, the 3D tooth scan data may be down-sampled and then input to the first artificial neural network. FIG. 4 shows an exemplary first 3D oral cavity model 310 represented by original 3D scan data and an exemplary second 3D oral cavity model 320 represented by downsampled 3D scan data. Compared to the first 3D oral model 310, the second 3D oral model 320 expresses the tooth scan results relatively simply by changing the curved area to a straight area and the curved area to a flat area. You will understand.

The degree of downsampling corresponds to the structural complexity of the first artificial neural network. The more complex the structure of the first artificial neural network is, the more simple the tooth is by reducing the number of lines and faces in the straight and flat areas changed from the curved area and curved area. The degree to which scan results are derived can be determined.

Since the diversity of pixels in the scanned image included in the 3D tooth scan data decreases depending on the degree of downsampling, the size of the 3D tooth scan data may be reduced.

Considering this, in some embodiments, a plurality of first artificial neural networks having different structural complexities are machine learned and stored, and the structural complexity of a first artificial neural network selected from among the plurality of first artificial neural networks is stored. The 3D tooth scan data may be downsampled to a corresponding level.

The 3D tooth scan data whose data size has been reduced as a result of the above-described downsampling will reduce the amount of computation related to the first artificial neural network, and as a result, the time required for computation related to the first artificial neural network may also be reduced. This has the effect of reducing the overall time required for 3D oral model segmentation. The oral model segmentation device may input the 3D tooth scan data into the first artificial neural network and obtain region separation data output from the first artificial neural network (S120). The region separation data may be understood as representing a region separation 3D oral model in which the gum region and crown region of the 3D oral model expressed by the 3D tooth scan data are separated.

As mentioned above, it is mentioned again that the oral model segmentation device may input downsampled 3D tooth scan data into the first artificial neural network.

Referring to FIG. 5, the input and output data for the first artificial neural network will be described again. The first artificial neural network 330 may receive downsampled 3D tooth scan data and output the region separation data. FIG. 5 shows the results of rendering downsampled 3D dental scan data as an exemplary 3D oral cavity model 320 and the results of rendering the region separation data as an exemplary 3D oral cavity model 340.

Hereinafter, in this specification, the 3D oral cavity model 340 in which the crown and gums are separated will be referred to as the ‘region-separated 3D oral cavity model’.

In some embodiments, the region-separated 3D oral model may be one in which the gum region is masked to have the first attribute. For example, the region-separated 3D oral model may be masked so that the gum region has a first color, a first pattern, or a first transparency. The first color, the first pattern, and the first transparency may be colors, patterns, and transparency that are difficult for the original three-dimensional oral model to have.

In some other embodiments, the region-separated 3D oral model may be one in which the crown region is masked to have a second attribute. For example, the region-separated 3D oral model may be masked so that the crown region has a second color, a second pattern, or a second transparency. The second color, the second pattern, and the second transparency may also be colors, patterns, and transparency that are difficult for the original three-dimensional oral model to have.

The masking process for the region-separated three-dimensional oral model may be a visual masking process. This is because the second and third artificial neural networks, which will be described later, are CNN (Convolutional Neural Network)-based artificial neural networks for processing visual information.

The region-separated three-dimensional oral cavity model 340 of FIG. 5 may be understood as masking the crown region in a manner that is colored with a second color.

Let's return to Figure 2 again for explanation.

The oral model segmentation device analyzes the region-separated three-dimensional oral model according to the region separation data, uses the results of the analysis to identify the occlusal surface of the region-separated three-dimensional oral model, and expresses only the coronal area of the occlusal surface. An occlusal crown image can be created (S130).

The oral model segmentation device can generate an occlusal crown image, which is a two-dimensional image looking at the crown from a perspective parallel to the occlusal surface, by executing logic based on a 3D model viewer. The operation of the oral model segmentation device performing S130 will be described in more detail with reference to FIGS. 6 and 7.

Referring to FIG. 6, the oral model segmentation device analyzes the region-separated three-dimensional oral model 350 viewed from the facial side to the lingual direction, and determines the occlusal line 351 of the maxillary crown and mandibular crown and the crown. The center line 352 can be confirmed.

The occlusal line of the crown may be a line extending in the left and right directions from the point where the upper and lower front teeth meet. The center line of the crown may be a line extending upward and downward from the point where the upper and lower front teeth meet.

In addition, the oral model segmentation device analyzes the region-separated three-dimensional oral model 360 viewed from the right posterior tooth in the lingual direction, and can further confirm the occlusal line 361 of the maxillary crown and mandibular crown and the center line 362 of the crown. .

In addition, the oral model segmentation device analyzes the region-separated three-dimensional oral model 370 viewed from the left posterior tooth in the lingual direction, and can further confirm the occlusal line 371 of the maxillary crown and mandibular crown and the center line 372 of the crown. . The oral model segmentation device will be able to uniquely specify the occlusal surface of the region-separated three-dimensional oral model (340) using the analysis results of the region-separated three-dimensional oral model (350, 360, 370) viewed from three directions. . Figure 6 shows an occlusal crown image 510 representing only the crown area of the occlusal surface specified in this way.

Referring to FIG. 7 , a method for generating an occlusal crown image 510 will be described.

The oral model segmentation device matches the visual line of the virtual camera 530 with the vertical axis passing through the center of the occlusal image 520 specified by the above-described method, and also ensures that the virtual camera 530 does not delete part of the occlusal image. The virtual camera 530 can be positioned to accommodate a predefined size.

As a result, the virtual camera 530 can capture the occlusal crown image 510, which is a snapshot image of the occlusal surface. The captured occlusal crown image 510 is input to the second artificial neural network 540.

Again, we will return to FIG. 2 for explanation.

In the next operation, the oral model segmentation device uses the second artificial neural network to select the center point of the crown image with similar individual characteristics among the previously learned crown images as the center point, thereby creating the center point of each crown region on the occlusal crown image 510. Identify (S140).

The second artificial neural network may also be supervised using learning data from occlusal crown images in which the center point of each crown region is labeled, similar to the first artificial neural network.

Meanwhile, the second artificial neural network can be machine-learned by grouping based on the arrangement and shape of the crown. For example, a first type of second artificial neural network machine-learned using a first group of learning data having a first type of crown arrangement and shape, and a second type of second artificial neural network having a second type of crown arrangement and shape. A second type of second artificial neural network machine-learned using learning data and a third type of second artificial neural network machine-learned using a third group of learning data having a third type of crown arrangement and shape, respectively. It can exist.

There may be a plurality of types grouped based on the arrangement and shape of the crown. The types can be grouped according to the similarity in arrangement and shape of the crowns being less than a predefined standard. The oral model segmentation device may be able to select one second artificial neural network among the first to third types of second artificial neural networks using the shape of the crown area of the occlusal crown image.

In addition, the oral model segmentation device may select a second artificial neural network among the first to third types of second artificial neural networks using the shape of the crown area of the pre-designated area of the occlusal crown image. .

As shown in FIG. 8, the second artificial neural network 540 receives the occlusal crown image 510 and outputs center point data 550 indicating the center point of the crown area of each tooth on the occlusal crown image. can do. 9 shows the result 560 where each

crown center point

560a, 560b, 560c, 560d according to the center point data 550 is displayed on the occlusal crown image 550.

Meanwhile, in some embodiments, the oral model segmentation device calculates the reliability of the center point data using the output data of the second artificial neural network, and when the reliability is less than a reference value, manually inputs the center point on the occlusal crown image. You can also print the user interface for this. The reliability may be lower as the central point data deviates from the standard data of the central point data collected by the artificial neural network by more than a certain level. In other words, in cases where the artificial neural network cannot accurately identify the center point of each crown region, the oral model segmentation device requests the user to manually point the center point, so that the 3D segmentation can be completed accurately through the subsequent operations. make it possible

Let's return to Figure 2 again for explanation.

The oral model segmentation device inputs the 3D tooth scan data into a machine-learned third artificial neural network (S160) and inputs the center point data into the machine-learned third artificial neural network, and as a result, the third artificial neural network is Segmentation data can be output (S150). The segmentation data may be understood as representing a 3D segmentation model in which the crown area of each tooth included in the 3D oral model is segmented for each tooth.

Similar to the first artificial neural network and the second artificial neural network, the third artificial neural network also uses learning data in which the 3D area of each tooth crown is distinguished on the 3D oral model and the center point on the occlusal crown image is labeled. It may be supervised learning.

The oral model segmentation device uses a third artificial neural network to select the crown state of a crown image with similar individual characteristics among pre-learned crown images as the crown state of the acquired 3D tooth scan data, and applies a second artificial neural network to the selected crown state. Segmentation data can be generated and output by inputting the output center point data (S150).

In Figure 10, the third artificial neural network 910 receives downsampled 3D tooth scan data 320 and center point data 560, and the crown region 920a to 920d of each tooth is segmented to create a 3D segmentation model ( It shows outputting segmentation data representing 920).

In some embodiments, the oral model segmentation device masks each crown region of the three-dimensional segmentation model 920 so that each crown region is visually distinguished from each other, and the segmentation data reflects the result of the masking process. Segmentation data can be adjusted, and the adjusted segmentation data can be output. Figure 9 shows a three-dimensional segmentation model 920 of segmentation data adjusted as described above.

In some embodiments, the oral model segmentation device uses the results of the masking process to identify each crown location, labels a tooth number for each crown region based on the identified crown location, and labels the adjusted crown location with a tooth number for each crown region. Labeling result data for the dental formula number for each crown region may be output along with the segmentation data.

So far, the segmentation method of the 3D oral model according to this embodiment has been described. As described above, it can be seen that no user manipulation is required to obtain a 3D oral model with complete segmentation using 3D scan data. Meanwhile, in the segmentation method of a 3D oral model according to another embodiment of the present disclosure, as shown in FIG. 11, the fourth artificial neural network 930 creates a 3D oral model 340 with the crown and gums separated and a center point. Data 560 may be input, and segmentation data representing a 3D segmentation model 940 in which the crown regions 940a to 940d of each tooth are segmented may be output.

Similar to the first to third artificial neural networks, the fourth artificial neural network also divides the 3D region of each tooth crown on a 3D oral model in which the crown and gums are separated, and the center point on the occlusal crown image is labeled. It may be supervised learning using trained learning data.

The flowchart of the three-dimensional oral model according to this embodiment is shown in FIG. 2. To explain the different operation compared to FIG. 1, the fourth artificial neural network receives region separation data as a result of step S120 (S170). That is, the fourth artificial neural network receives the region separation data and the center point data and outputs the segmentation data.

The fourth artificial neural network according to this embodiment receives area separation data in which the crown area and gum area have been processed as a kind of preprocessing by the first artificial neural network instead of the 3D tooth scan data, which can be considered original data. The region separation data may be understood to contain more information because information about the distinction between the crown region and the gum region is added compared to the 3D tooth scan data. Therefore, the fourth artificial neural network receives more information than the third artificial neural network, and this may have an advantage in segmentation accuracy.

Hereinafter, the segmentation method of the 3D oral model described with reference to FIG. 1 will be summarized and explained with reference to FIG. 12.

The downsampled 3D tooth scan data 320 is input to the first artificial neural network 330, and the first artificial neural network 330 is region classification data representing a 3D oral model 340 with the crown and gums separated. outputs.

Next, an occlusal crown image 520 is created using a 3D oral model 340 in which the crown and gums are separated. The occlusal crown image 520 is input to the second artificial neural network 540, and the second artificial neural network 540 outputs center point data indicating the center point 560 of each crown region.

Next, the 3D tooth scan data 320 and the center point data indicating the center point 560 of each tooth crown area are input to the third artificial neural network 910, and the third artificial neural network 910 is the crown area of each tooth. Segmentation data representing this segmented 3D segmentation model 920 is output.

Hereinafter, the segmentation method of the 3D oral model described with reference to FIG. 2 will be summarized and explained with reference to FIG. 13.

The down-sampled 3D tooth scan data 320 is input to the first artificial neural network 330, and the first artificial neural network 330 is region classification data representing a 3D oral model 340 with the crown and gums separated. outputs.

Next, region classification data representing the three-dimensional oral model 340 in which the crown and gums are separated, and center point data indicating the center point 560 of each crown region are input to the fourth artificial neural network 930, and the fourth The artificial neural network 930 outputs segmentation data representing a 3D segmentation model 940 in which the crown region of each tooth is segmented.

Meanwhile, segmentation data obtained according to the embodiments described so far with reference to FIGS. 1 to 13 can be used to generate an output screen for 3D orthodontic simulation. For example, the segmentation data may be used in the process of displaying the results of each tooth placement movement for a 3D orthodontic simulation.

Hereinafter, an oral model segmentation system 100 according to another embodiment of the present disclosure will be described with reference to FIG. 14 . As shown in FIG. 13, the oral model segmentation system 100 of this embodiment includes one or more processors 1100, a system bus 1600, a communication interface 1200, and a computer program 1500 performed by the processor 1100. ) may include a memory 1400 for loading a computer program 1500 and a storage 1300 for storing a computer program 1500.

The processor 1100 controls the overall operation of each component of the oral model segmentation system 100. The processor 1100 may perform operations on at least one application or program to execute methods/operations according to various embodiments of the present disclosure. The memory 1400 stores various data, commands and/or information. The memory 1400 may load one or more computer programs 1500 from the storage 1300 to execute methods/operations according to various embodiments of the present disclosure. The bus 1600 provides communication between components of the simulation device 200. The communication interface 1200 supports Internet communication of the 3D oral model segmentation system 100. Additionally, the communication interface 1200 may be connected to a 3D scanner device (not shown). Storage 1300 may non-temporarily store one or more computer programs 1500. The computer program 1500 may include one or more instructions implementing methods/operations according to various embodiments of the present disclosure. When the computer program 1500 is loaded into the memory 1400, the processor 1100 can perform methods/operations according to various embodiments of the present disclosure by executing the one or more instructions.

The memory 1400 includes the patient's 3D dental scan data, data defining a first artificial neural network, data defining a second artificial neural network, data defining a third artificial neural network, and a computer program that performs 3D oral model segmentation. (1500) can be loaded.

The computer program 1500 may include instructions for performing one or more operations in which methods/operations according to various embodiments of the present disclosure are implemented.

For example, the computer program 1500 separates the gum area and crown area of the 3D oral model represented by the 3D tooth scan data by the first artificial neural network receiving the 3D tooth scan data. 3 An instruction for outputting region separation data representing a dimensional oral model, analyzing the region separation 3D oral model according to the region separation data to identify the occlusal surface of the region separation 3D oral model, and the crown area of the occlusal surface. an instruction for generating an occlusal crown image representing only an occlusal crown image, and an instruction for the second artificial neural network receiving the occlusal crown image to output center point data indicating the center point of the crown area of each tooth on the occlusal crown image; , the third artificial neural network that receives the 3D tooth scan data and the center point data outputs segmentation data representing a 3D segmentation model in which the crown area of each tooth included in the 3D oral model is segmented for each tooth. May contain instructions.

So far, various embodiments of the present disclosure and effects according to the embodiments have been mentioned with reference to FIGS. 1 to 14 . The effects according to the technical idea of the present disclosure are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below.

The technical ideas of the present disclosure described so far can be implemented as computer-readable code on a computer-readable medium. The computer program recorded on the computer-readable recording medium can be transmitted to another computing device through a network such as the Internet, installed on the other computing device, and thus used on the other computing device.

Although operations are shown in the drawings in a specific order, it should not be understood that the operations must be performed in the specific order shown or sequential order or that all illustrated operations must be performed to obtain the desired results. In certain situations, multitasking and parallel processing may be advantageous. Although embodiments of the present disclosure have been described above with reference to the attached drawings, those skilled in the art will understand that the present invention can be implemented in other specific forms without changing the technical idea or essential features. I can understand that there is. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. The scope of protection of the present invention should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be construed as being included in the scope of rights of the technical ideas defined by this disclosure.

Claims

In a method performed by a computing system,

Acquiring at least one 3D tooth scan data;

An area representing a separated 3D oral model in which the gum area and the crown area of the 3D oral model generated by the first artificial neural network that received the 3D tooth scan data using at least one 3D tooth scan data are separated. outputting separated data;

Analyzing the region-separated three-dimensional oral model to identify an occlusal surface of the three-dimensional oral model, and generating an occlusal crown image representing only the crown region of the occlusal surface;

A second artificial neural network receiving the occlusal crown image outputs center point data indicating the center point of the crown area of each tooth on the occlusal crown image; and

A third artificial neural network receiving the 3D tooth scan data and the center point data outputs segmentation data representing a 3D segmentation model in which the crown area of each tooth included in the 3D oral model is segmented for each tooth. containing,

Segmentation method for 3D oral model.
According to claim 1,

The step of outputting the region separation data is,

Down-sampling the 3D tooth scan data; and

Comprising the step of inputting the downsampled 3D tooth scan data into the first artificial neural network,

Segmentation method for 3D oral model.
According to claim 1,

The step of outputting the region separation data is,

masking the gum area of the region-separated 3D oral model to have a first attribute, or masking the crown region of the region-separated 3D oral model to have a second attribute;

adjusting the region separation data so that the region separation data reflects a result of the masking process; and

Including outputting the adjusted region separation data,

Segmentation method for 3D oral model.
According to claim 1,

The step of generating the occlusal crown image is,

Analyzing the region-separated 3D oral cavity model to identify an occlusal surface of the region-separated 3D oral cavity model;

Positioning a virtual camera to have a field-of-view (FOV) perpendicular to the occlusal surface; and

Comprising the step of generating the occlusal surface crown image using the image of the occlusal surface captured by the virtual camera,

Segmentation method for 3D oral model.
According to clause 4,

The step of outputting the region separation data is,

masking the gum area of the region-separated 3D oral model to have a first attribute, or masking the crown region of the region-separated 3D oral model to have a second attribute; and

A step of adjusting the region separation data so that the region separation data reflects the result of the masking process,

The step of generating the occlusal crown image using the image of the region-separated three-dimensional oral model captured by the virtual camera,

Comprising the step of generating the occlusal crown image by transparently processing the gum area in the image of the occlusal surface using the result of the masking process,

Segmentation method for 3D oral model.
According to claim 1,

The step of outputting the center point data is,

Comprising the step of selecting the second artificial neural network from among a plurality of candidate artificial neural networks using the shape of the crown area of the occlusal crown image,

Segmentation method for 3D oral model.
According to claim 1,

The step of outputting the center point data is,

Comprising the step of selecting the second artificial neural network from among a plurality of candidate artificial neural networks using the shape of the crown area of the pre-designated area of the occlusal crown image,

Segmentation method for 3D oral model.
According to claim 1,

The step of outputting the center point data is,

calculating reliability of the central point data using output data of the second artificial neural network; and

If the reliability is less than the reference value, outputting a user interface for manual input of the center point on the occlusal crown image,

Segmentation method for 3D oral model.
According to claim 1,

The step of outputting the segmentation data is,

Down-sampling the 3D tooth scan data; and

Including the step of inputting the downsampled 3D tooth scan data and the center point data into the third artificial neural network,

Segmentation method for 3D oral model.
According to claim 1,

The step of outputting the segmentation data is,

masking each crown region of the three-dimensional segmentation model according to the segmentation data so that each crown region is visually distinguished from each other; and

adjusting the segmentation data so that the segmentation data reflects the results of the masking process; and

Including outputting the adjusted segmentation data,

Segmentation method for 3D oral model.
According to claim 10,

The step of outputting the adjusted segmentation data includes:

Using the result of the masking process, identifying each crown position;

Labeling a dental formula number for each crown region based on the identified crown position; and

Comprising the step of outputting labeling result data for the dental formula number for each crown region together with the adjusted segmentation data,

Segmentation method for 3D oral model.
According to claim 1,

Further comprising generating an output screen for 3D orthodontic simulation using the segmentation data,

Segmentation method for 3D oral model.
In a method performed by a computing system,

Acquiring at least one 3D tooth scan data;

A first artificial neural network receiving the 3D tooth scan data outputs area separation data representing a 3D oral model in which the gum area and the crown area of the 3D oral model are separated;

Analyzing the region-separated three-dimensional oral model to identify an occlusal surface of the three-dimensional oral model, and generating an occlusal crown image representing only the crown region of the occlusal surface;

A second artificial neural network receiving the occlusal crown image outputs center point data indicating the center point of the crown area of each tooth on the occlusal crown image; and

A third artificial neural network receiving the region separation data and the center point data outputs segmentation data representing a 3D segmentation model in which the crown region of each tooth included in the 3D oral model is segmented for each tooth. ,

Segmentation method for 3D oral model.
a memory into which at least one three-dimensional tooth scan data, a first artificial neural network, a second artificial neural network, a third artificial neural network, and a three-dimensional oral model segmentation program are loaded; and

Includes a processor that executes the 3D oral model segmentation program,

The 3D tooth segmentation program,

Instructions for the first artificial neural network receiving the 3D tooth scan data to output region separation data representing a region separation 3D oral model in which the gum region and the crown region of the 3D oral model are separated;

Instructions for analyzing the region-separated three-dimensional oral model to identify an occlusal surface of the three-dimensional oral model and generating an occlusal crown image representing only the crown region of the occlusal surface;

Instructions for the second artificial neural network receiving the occlusal crown image to output center point data indicating the center point of the crown area of each tooth on the occlusal crown image; and

An instruction in which the third artificial neural network, which has received the 3D tooth scan data and the center point data, outputs segmentation data representing a 3D segmentation model in which the crown area of each tooth included in the 3D oral cavity model is segmented for each tooth. Including,

Segmentation device for 3D oral model.
According to claim 14,

The 3D tooth segmentation program,

Further comprising instructions for generating an output screen for 3D orthodontic simulation using the output segmentation data,

Segmentation device for 3D oral model.