KR102714028B1 - System for orthodontic treatment information provision and method for thereof - Google Patents

Abstract

An operating method of an orthodontic treatment information providing system operated by at least one processor, wherein when a plurality of two-dimensional teeth images of a patient are received, the received plurality of two-dimensional teeth images are input into a pre-trained deep learning network to generate one three-dimensional teeth data. A tooth is segmented from the three-dimensional teeth data generated by the deep learning network to generate three-dimensional teeth data, and the initial three-dimensional teeth data of the patient and the generated three-dimensional teeth data are aligned to obtain a three-dimensional change amount.

Description

{System for orthodontic treatment information provision and method for its}

The present invention relates to an orthodontic treatment information providing system and an operating method thereof, which analyzes a patient's two-dimensional teeth into three dimensions for orthodontic treatment of the patient and creates an orthodontic treatment plan based on the analysis.

Orthodontic treatment is an essential treatment for improving the functions of the teeth themselves, such as pronunciation, chewing, and temporomandibular joints, in addition to simply aesthetic reasons. For orthodontic treatment, it is essential to analyze the patient's dental condition and obtain 3D dental data.

At this time, the movement of teeth due to orthodontic treatment is different for each patient. Therefore, if the patient visits the hospital 10 to 20 times during the orthodontic treatment period, the medical staff analyzes the patient's current dentition and updates the treatment plan based on the analyzed dentition. Therefore, the medical staff's accurate analysis of the patient's dentition can improve the accuracy of the orthodontic results and patient satisfaction.

During orthodontic treatment, teeth rotate and translate in three dimensions, so in order to accurately analyze the teeth, the movement of the teeth must be analyzed in three dimensions. Accordingly, three-dimensional analysis of teeth in general is largely divided into two methods.

The first is an analysis method using an oral scanner. This is a method of obtaining 3D dental data of a patient using an oral scanner and analyzing the obtained 3D dental data in 3D. This can obtain accurate dental data, but has the disadvantage of requiring high labor and a long time due to expensive equipment and complex photographing procedures.

The second method is to look at two-dimensional dental images taken with a general camera and have medical staff interpret them three-dimensionally based on their empirical knowledge. This has the disadvantage of low dental analysis accuracy because the accuracy of the analysis varies depending on the experience of the medical staff.

Korean Patent No. 2462185 (Registration Date: 2022.10.28.) U.S. Patent Publication No. 2023-0066220 (Published: 2023.03.02.)

Accordingly, the present invention provides an orthodontic treatment information providing system and an operating method thereof, which generates three-dimensional dental data using two-dimensional dental images and analyzes the generated three-dimensional dental data to provide information for orthodontic treatment.

As an operating method of an orthodontic treatment information providing system operated by at least one processor, which is one feature of the present invention for achieving the technical task of the present invention,

The method comprises the steps of: receiving a plurality of two-dimensional dental images of a patient; inputting the received plurality of two-dimensional dental images into a pre-trained deep learning network to generate one three-dimensional dental data; segmenting teeth from the three-dimensional dental data generated by the deep learning network to generate three-dimensional dental data; and aligning the initial three-dimensional dental data of the patient with the generated three-dimensional dental data to obtain a three-dimensional variation.

Before the step of receiving the two-dimensional teeth images, the step of training the deep learning network so that a set of positions where light has passed through each of a plurality of pixels constituting the original two-dimensional teeth image is input to the deep learning network and teeth geometry information including the set of positions and the color and density values of the positions is output.

The above deep learning network may be a deep learning network based on NeRF (Neural Radiance Fields).

The above learning step may further include a step of defining a pixel value for each pixel by adding up the tooth geometry information for each pixel, and a step of generating a rendered two-dimensional tooth image by adding up the pixel values of each of the plurality of pixels.

The step of generating the rendered two-dimensional teeth image includes the step of calculating a mean square error between the original two-dimensional teeth image and the rendered two-dimensional teeth image using a loss function, and the loss function can add a difference in outlines of the rendered two-dimensional teeth image and the original two-dimensional teeth image to the calculated mean square error.

The above multiple two-dimensional teeth images may be images collected through an image collection device such as a digital camera or a mobile phone.

The step of generating the above three-dimensional teeth data may include the step of inputting a density value included in the teeth geometry information generated from the two-dimensional teeth image into a three-dimensional mesh model generation function, and generating the three-dimensional teeth data using a marching cube function.

The step of acquiring the above three-dimensional change amount may further include the step of surface-aligning the crown of each tooth included in the initial three-dimensional tooth data with the crown of each tooth included in the generated three-dimensional tooth data.

After the surface alignment step, a step of extracting the three-dimensional translation and rotation values of each tooth as the three-dimensional change amount may be included.

Another feature of the present invention for achieving the technical problem of the present invention is an interface for receiving a plurality of two-dimensional dental images of a patient, a memory, and at least one processor for executing instructions of a program loaded in the memory, wherein the processor generates one dental geometry information from the received plurality of two-dimensional dental images, generates three-dimensional dental data from three-dimensional dental data generated from the dental geometry information, and aligns the initial three-dimensional dental data of the patient with the generated three-dimensional dental data to obtain a three-dimensional variation.

The above processor trains the deep learning network to output tooth geometry information including the set of positions and the color and density values of the positions by inputting a set of positions where light has passed through each of a plurality of pixels constituting the original two-dimensional tooth image into the deep learning network, and the deep learning network may be a deep learning network based on NeRF (Neural Radiance Fields).

The above processor can define a pixel value for each pixel by adding up a plurality of pieces of tooth geometry information for each pixel, and generate a rendered two-dimensional tooth image by adding up the pixel values of each of the plurality of pixels.

The above processor can calculate a mean square error between the original two-dimensional teeth image and the rendered two-dimensional teeth image, and add a difference in outlines between the rendered two-dimensional teeth image and the original two-dimensional teeth image to the calculated mean square error to generate the rendered two-dimensional teeth image.

The above processor can generate the three-dimensional tooth data using a marching cube function by inputting the density value included in the tooth geometry information into a three-dimensional mesh model generation function.

The above processor can surface-align the crown of each tooth included in the initial 3D tooth data with the crown of each tooth included in the generated 3D tooth data, and extract the 3D translation value and rotation value of each tooth as the 3D change amount.

According to the present invention, since three-dimensional dental data is obtained from two-dimensional dental images, three-dimensional dental data can be easily obtained without expensive equipment or complex photographing procedures.

In addition, since the three-dimensional change amount of the teeth is tracked and analyzed using three-dimensional alignment, it is possible to track the three-dimensional change amount of the teeth using three-dimensional surface alignment between the teeth of the initial and current dentition.

In addition, through the three-dimensional change in each tooth, the dental condition and treatment effect can be accurately analyzed, ultimately improving the accuracy of orthodontic treatment and patient satisfaction.

Figure 1 is an exemplary diagram of an environment in which an orthodontic treatment information providing system according to an embodiment of the present invention is implemented.
Figure 2 is a flow chart of the operation of an orthodontic treatment information providing system according to an embodiment of the present invention.
FIG. 3 is an example of a two-dimensional tooth image obtained according to an embodiment of the present invention.
FIG. 4 is an example diagram of three-dimensional dental data generated according to an embodiment of the present invention.
Figure 5 is an exemplary diagram showing the learning process of a deep learning network according to an embodiment of the present invention.
FIG. 6 is an example diagram showing segmentation of three-dimensional tooth data according to an embodiment of the present invention.
FIG. 7 is an example of a three-dimensional change in a tooth between an initial point in time and a current point in time obtained according to an embodiment of the present invention.
Figure 8 is a structural diagram of an orthodontic treatment information providing system according to an embodiment of the present invention.

Hereinafter, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily practice the present invention. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts that are not related to the description are omitted, and similar parts are assigned similar drawing reference numerals throughout the specification.

Throughout the specification, whenever a part is said to "include" a component, this does not mean that it excludes other components, but rather that it may include other components, unless otherwise stated.

Referring to the drawings below, a system for providing orthodontic treatment information and its operation method according to an embodiment of the present invention will be described.

Figure 1 is an exemplary diagram of an environment in which an orthodontic treatment information providing system according to an embodiment of the present invention is implemented.

As illustrated in Fig. 1, the orthodontic treatment information providing system (100) collects a plurality of two-dimensional teeth images. Here, the two-dimensional teeth images are explained by using as an example the fact that they are collected through various types of image collection devices (not illustrated in the drawing) that can collect images, such as a digital camera and a mobile phone.

The orthodontic treatment information providing system (100) generates 3D dentition data from the input 2D dentition images. Then, the orthodontic treatment information providing system (100) segments teeth from the generated 3D dentition data and compares the segmented teeth with the teeth of the original patient to obtain the 3D change amount of the teeth.

The orthodontic treatment information providing system (100) provides tooth analysis information including the three-dimensional change amount of the acquired teeth. In addition, the orthodontic treatment information providing system (100) provides orthodontic treatment information so that medical staff can establish an orthodontic treatment plan to advance to the final teeth state based on the three-dimensional change amount.

The operation method of the orthodontic treatment information providing system (100) is described with reference to FIG. 2.

Figure 2 is a flow chart of the operation of an orthodontic treatment information providing system according to an embodiment of the present invention.

As illustrated in Fig. 2, the orthodontic treatment information providing system (100) creates and trains a deep learning network to create a two-dimensional dental image into three-dimensional dental data (S100).

That is, the orthodontic treatment information provision system (100) constructs a deep learning network based on NeRF (Neural Radiance Fields). Each pixel constituting a two-dimensional image is expressed as the sum of information passed by one ray. NeRF is a deep learning network that imitates this physical phenomenon.

That is, when the NeRF-based deep learning network receives as input a set of locations (p={x, y, z}) where a ray passed, it provides as output 3D geometric information of teeth (infor={r, g, b, d, a}) (hereinafter referred to as 'teeth geometric information') including the color and density values of the input locations. Here, the parameters r, g, and b included in the teeth geometric information are values representing the color of the input locations, d represents depth information, and a represents density. The density plays a role similar to a weight when expressing the color of a pixel. For example, air has a low density, so color information is slightly reflected, whereas teeth have a high density, so color information is highly reflected. In the embodiment of the present invention, for the convenience of explanation, teeth geometric information may also be referred to as deep learning network output information.

The NeRF-based deep learning network uses a positional encoding technique to increase the dimension of input information. NeRF is a deep learning network developed for general objects, and uses a frequency-based positional encoding technique that can robustly express a high-frequency range. However, since the high-frequency range is extremely rare in biological tissues such as teeth, the frequency-based positional encoding technique is inefficient, and therefore, in the embodiment of the present invention, an interpolation-based positional encoding technique suitable for biological tissues is used.

The orthodontic treatment information providing system (100) extracts multiple grids of various sizes that include one point (p _n ) of an input ray in order to apply an interpolation-based position encoding technique. The size and number of grids can be defined in advance, and in the embodiment of the present invention, the size and number are not limited to either one.

The orthodontic treatment information providing system (100) obtains a feature vector for expressing a point (p _n ) in each grid. A plurality of grids are extracted, and the feature vector obtained from each grid is a parameter that changes each time a deep learning network is trained, and is optimized as training continues. The orthodontic treatment information providing system (100) can provide continuity by flattening a plurality of feature vectors by linear interpolation. This is a known technology, and the embodiment of the present invention does not limit the flattening or continuity of the feature vectors to any one method.

). 딥러닝 네트워크는 2차원 이미지의 모든 픽셀에 대하여 광선의 위치를 입력으로 받아 네트워크 출력 정보를 합산함으로써, 랜더링 된 2차원의 치열 이미지(

)를 생성한다.The orthodontic treatment information providing system (100) trains the constructed deep learning network. To train the deep learning network, the orthodontic treatment information providing system (100) inputs one ray position (p _n ) for generating one pixel value into the deep learning network, and defines the pixel value through the sum of the network output information (info _n ).

) The deep learning network receives the position of the light beam for each pixel of the two-dimensional image as input and sums the network output information to produce a rendered two-dimensional tooth image (

) is created.

Here, the basic loss function of NeRF calculates the mean squared error between the synthetic image and the original image. However, since the dental data must clearly express the distinction between teeth, the NeRF-based deep learning network according to the embodiment of the present invention applies contour information by adding it to the loss function as in the following mathematical expression 1.

[Mathematical formula 1]

는 딥러닝 네트워크에서 출력된 정보로, 랜더링 된 2차원 치열 이미지를 의미한다. 그리고,

는 원본 이미지, mse는 두 이미지 간 평균제곱오차를 의미하고, edge는 입력된 두 이미지의 윤곽선 차이를 의미한다.Here,

is the information output from the deep learning network, which means a rendered two-dimensional tooth image. And,

is the original image, mse is the mean square error between the two images, and edge is the difference in outlines between the two input images.

The orthodontic treatment information providing system (100) trains the constructed deep learning network, and then receives multiple two-dimensional teeth images of the patient from an image collection device (S110). In the embodiment of the present invention, a description is made by using an image collection device such as a digital camera, a mobile phone, etc. to receive multiple two-dimensional teeth images taken at various angles so that the patient's teeth can be seen as an example. However, the orthodontic treatment information providing system (100) may also receive a video recorded so that the patient's teeth can be seen as two-dimensional teeth data.

The orthodontic treatment information providing system (100) uses the deep learning network trained in step S100 to generate 2D teeth images into 3D teeth data (S120).

That is, the positions of all rays derived from two-dimensional tooth images are input into the pre-trained deep learning network, and tooth geometry information of the space corresponding to the positions of the rays is obtained.

Here, among the tooth geometry information, the density value is mapped to a value according to the object characteristics. For example, air is mapped to a low value because it has a low density, and teeth are mapped to a high value. In the embodiment of the present invention, the criteria for the high and low values to be mapped are not limited to a single numerical value.

The orthodontic treatment information providing system (100) inputs a density value included in the dental geometry information into a three-dimensional mesh model generating function to generate three-dimensional dental data. In the embodiment of the present invention, the use of a marching cube function to generate a three-dimensional dental mesh model, i.e., three-dimensional dental data, is described as an example, but various mesh model generating functions can be used. That is, when the orthodontic treatment information providing system (100) inputs a density value into the marching cube function, it obtains three-dimensional dental data as a result.

After generating the 3D dental data in this way, the orthodontic treatment information providing system (100) segments the teeth from the 3D dental data (S130). In other words, the orthodontic treatment information providing system (100) is explained by dividing each tooth from the 3D dental data as an example, but various artificial intelligence-based tooth segmentation networks may be used. In addition, since the method of segmenting teeth from the 3D dental data is already a known technology, the embodiment of the present invention does not describe it by being limited to any one method.

The orthodontic treatment information providing system (100) performs 3D surface alignment of the patient's initial dentition and the dentition of teeth segmented at step S130 to obtain a 3D change amount (S140). That is, in order to obtain a 3D change amount of each tooth from the initial dentition to the current time point, the orthodontic treatment information providing system (100) performs surface alignment of the teeth of the initial dentition and the teeth obtained at step S130, respectively.

During orthodontic treatment, the movement of the teeth or the shape of the tooth head (crown) itself corresponds to a rigid body that does not change. Therefore, the results of surface alignment of the initial teeth and recent teeth can be considered to have high reliability. The method of the orthodontic treatment information providing system (100) for surface alignment of the initial teeth and recent teeth is a known technology, and a detailed description thereof is omitted in the embodiment of the present invention.

The orthodontic treatment information providing system (100) can extract the three-dimensional translational value and rotational value of each tooth through surface alignment. Accordingly, the orthodontic treatment information providing system (100) can provide the translational value and rotational value of each extracted tooth to the medical staff as tooth analysis information.

In addition, the orthodontic treatment information providing system (100) can provide treatment plan information for correcting the current teeth to the final teeth state during orthodontic treatment. Here, the treatment plan information can include translational values and rotational values to move from the current teeth position to the final teeth position, and other information necessary to obtain the translational values and rotational values.

An example of two-dimensional dental images collected by an image collection device during the above-described procedure is described with reference to FIG. 3, and three-dimensional dental data generated by an orthodontic treatment information providing system (100) using the two-dimensional dental images is described with reference to FIG. 4.

FIG. 3 is an example diagram of a two-dimensional dental image obtained according to an embodiment of the present invention, and FIG. 4 is an example diagram of three-dimensional dental data generated according to an embodiment of the present invention.

First, as illustrated in FIG. 3, in the embodiment of the present invention, six two-dimensional dental images of a patient taken from different angles are collected and used as inputs for the orthodontic treatment information providing system (100). However, the orthodontic treatment information providing system (100) may also receive as input images taken by an image collecting device so that all of the patient's teeth are shown, and generate the images into a plurality of two-dimensional dental images. Since the method by which the orthodontic treatment information providing system (100) generates images into two-dimensional dental images can be implemented using various technologies, a detailed description thereof will be omitted in the embodiment of the present invention.

When the orthodontic treatment information providing system (100) receives a plurality of two-dimensional teeth images as illustrated in FIG. 3, it generates three-dimensional teeth data using a pre-trained deep learning network.

That is, when a set of locations (p={x, y, z}) where light has passed through each pixel constituting each of a plurality of two-dimensional teeth images, as illustrated in Fig. 4, is input to a deep learning network, the deep learning network obtains deep learning network output information for the teeth space.

Here, among the deep learning network output information, the density value is mapped according to the characteristics of the object. The deep learning network inputs the density value into the 3D mesh model generation function to generate a 3D dental mesh model. The 3D dental mesh model generated in this way corresponds to the 3D dental data illustrated in Fig. 4. The 3D dental data includes both the teeth part (①) and the gum part (②).

Additionally, if the 3D dental data is generated from a 2D dental image of the patient's maxillary sinus, the palate may be included. And if the 3D dental data is generated from a 2D dental image of the patient's mandibular sinus, the patient's tongue may also be included.

Here, the learning process of a deep learning network that outputs two-dimensional dental images as three-dimensional dental data is described with reference to Fig. 5.

Figure 5 is an exemplary diagram showing the learning process of a deep learning network according to an embodiment of the present invention.

As illustrated in Fig. 5, a pixel (③) that constitutes a two-dimensional image is expressed as the sum of information passed by a single ray. Then, when a NeRF-based deep learning network receives as input a set of locations (p={x, y, z}) where a ray passed, it provides as output the three-dimensional tooth geometry information (info={r, g, b, d, a}) including the color and density values of the input locations.

Here, the parameters r, g, and b included in the tooth geometry information are values representing the color of the input location, d represents depth information, and a represents a density value. In the embodiment of the present invention, for the convenience of explanation, the geometric information is referred to as deep learning network output information.

NeRF-based deep learning networks use positional encoding techniques to increase the dimensionality of input information. NeRF is a deep learning network developed for general objects, and uses a frequency-based positional encoding technique that can robustly express high-frequency regions. However, since high-frequency regions are extremely rare in biological tissues such as teeth, the frequency-based positional encoding technique is inefficient, and therefore, in the embodiment of the present invention, an interpolation-based positional encoding technique suitable for biological tissues is used.

). 딥러닝 네트워크는 2차원 이미지의 모든 픽셀에 대하여 광선의 위치를 입력으로 받아 네트워크 출력 정보를 합산함으로써, 랜더링 된 2차원의 치열 이미지(

)를 생성한다.The orthodontic treatment information providing system (100) trains the constructed deep learning network. To train the deep learning network, the orthodontic treatment information providing system (100) inputs one ray position (p _n ) for generating one pixel value into the deep learning network, and defines the pixel value through the sum of the network output information (info _n ).

) The deep learning network receives the position of the light beam for each pixel of the two-dimensional image as input and sums the network output information to produce a rendered two-dimensional tooth image (

) is created.

Here, the basic loss function of NeRF calculates the mean squared error between the synthetic image and the original image. However, since the dental data must clearly express the distinction between teeth, the NeRF-based deep learning network according to the embodiment of the present invention applies contour information by adding it to the loss function as in the mathematical expression 1 described above.

FIG. 6 is an example diagram showing the division of three-dimensional tooth data according to an embodiment of the present invention.

As illustrated in FIG. 6, when the orthodontic treatment information providing system (100) acquires 3D dental data, it segments only the teeth from the acquired 3D dental data and creates 3D dental data.

The orthodontic treatment information providing system (100) can use various methods to divide 3D tooth data from 3D dental data. As one example, the orthodontic treatment information providing system (100) can use an artificial intelligence technique for dividing tooth data.

That is, if 3D dental data is input into a learned artificial intelligence model, individual 3D dental data can be obtained. Here, 3D dental data is composed of points and surfaces. Therefore, if points of 3D dental data are input into an artificial intelligence model, it is possible to classify which tooth each point is, and as a result, individual 3D teeth can be segmented.

In addition to one embodiment like this, the embodiment of the present invention does not limit the method of segmenting 3D tooth data from 3D dental data to any one method. In addition, the orthodontic treatment information providing system (100) can also utilize various artificial intelligence-based tooth segmentation networks, but is not limited to any one network.

The following describes the three-dimensional change between the teeth acquired at the initial point in time and the teeth at the current point in time with reference to Figure 7.

FIG. 7 is an example of a three-dimensional change in a tooth between an initial point in time and a current point in time obtained according to an embodiment of the present invention.

Figure 7 (a) shows the shape of each tooth acquired at the initial point in time of the patient, and Figure 7 (b) shows the shape of each tooth head (also referred to as 'crown') at the current point in time.

The orthodontic treatment information provision system (100) applies surface alignment and extracts 3D translation/rotation values to obtain the 3D change amount of each tooth from the initial orthodontic treatment time to the present time.

During orthodontic treatment, teeth move, but the shape of the tooth head itself is a rigid body that does not change, so high reliability of surface alignment can be expected. In other words, since the teeth have the same tooth head at the initial and current time points, high reliability of surface alignment and high-precision change amount can be obtained.

Figure 8 is a structural diagram of an orthodontic treatment information providing system according to an embodiment of the present invention.

Referring to FIG. 8, in a correction treatment information providing system (100) operated by at least one processor, a program including instructions described to execute the operation of the present invention is executed. The program can be stored in a computer-readable storage medium and can be distributed.

The hardware of the orthodontic treatment information providing system (100) may include at least one processor (110), memory (120), storage (130), and communication interface (140), and may be connected via a bus. In addition, hardware such as input devices and output devices may be included. The orthodontic treatment information providing system (100) may be equipped with various software, including an operating system capable of running a program.

The processor (110) is a device that controls the operation of the orthodontic treatment information providing system (100), and may be a processor of various types that processes commands included in the program, for example, a CPU (Central Processing Unit), an MPU (Micro Processor Unit), an MCU (Micro Controller Unit), a GPU (Graphics Processing Unit), etc.

The memory (120) loads the corresponding program so that the commands described to execute the operation of the present invention are processed by the processor (110). The memory (120) may be, for example, a ROM (read only memory), a RAM (random access memory), etc. The storage (130) stores various data, programs, etc. required to execute the operation of the present invention. The communication interface (140) may be a wired/wireless communication module.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements made by those skilled in the art using the basic concept of the present invention defined in the following claims also fall within the scope of the present invention.

Claims (15)

A method of operating an orthodontic treatment information providing system operated by at least one processor,
The above orthodontic treatment information provision system is,
A step of receiving multiple two-dimensional dental images of a patient,
A step of inputting the received multiple two-dimensional teeth images into a pre-trained deep learning network and generating one three-dimensional teeth data based on the density value included in the teeth geometry information generated from the two-dimensional teeth images.
A step of generating 3D tooth data by segmenting teeth from the 3D tooth data generated by the above deep learning network, and
A step of obtaining a 3D change amount by aligning the initial dental data of the above patient with the 3D dental data generated above.
A method of operation, comprising: In the first paragraph,
Before the step of receiving the above two-dimensional dental images,
A step of training the deep learning network so that when a set of positions where light has passed through each of a plurality of pixels constituting the original two-dimensional teeth image is input to the deep learning network, teeth geometry information including the set of positions and the color and density value of the positions is output.
A method of operation, comprising: In the second paragraph,
The above deep learning network is a deep learning network based on NeRF (Neural Radiance Fields), and its operation method. In the second paragraph,
The above learning steps are:
A step of defining a pixel value for each pixel by summing the above-mentioned tooth geometry information for each pixel, and
A step of generating a rendered two-dimensional teeth image by adding up the pixel values of each of the above plurality of pixels.
A method of operation, further comprising: In paragraph 4,
The step of generating the above rendered two-dimensional teeth image is:
A step of calculating the mean square error between the original two-dimensional teeth image and the rendered two-dimensional teeth image using a loss function,
Including,
The above loss function is an operating method that adds the difference in outline between the rendered two-dimensional tooth image and the original two-dimensional tooth image to the calculated mean square error. In the first paragraph,
A method of operation, wherein the above plurality of two-dimensional teeth images are images collected through an image collection device, either a digital camera or a mobile phone. In the first paragraph,
The step of generating the above 3D dental data is:
A step of inputting the above density value into a 3D mesh model generation function and generating the 3D tooth data using a marching cube function.
A method of operation, comprising: In the first paragraph,
The step of obtaining the above three-dimensional change amount is:
A step of surface-aligning the crown of each tooth included in the initial dental data with the crown of each tooth included in the generated 3D dental data.
A method of operation, further comprising: In Article 8,
After the above surface alignment step,
Step of extracting the 3D translation and rotation values of each tooth as the 3D change amount
A method of operation, comprising: An interface for receiving multiple two-dimensional dental images of a patient,
memory, and
comprising at least one processor that executes instructions of a program loaded into the memory;
The above processor
An orthodontic treatment information providing system that generates single dental geometric information from the plurality of received two-dimensional dental images, generates three-dimensional dental data from three-dimensional dental data generated from density values included in the dental geometric information, and obtains a three-dimensional change amount by aligning the patient's initial three-dimensional dental data with the generated three-dimensional dental data. In Article 10,
The above processor,
When a set of locations where light has passed through each of a plurality of pixels constituting the original two-dimensional teeth image is input to a deep learning network, the deep learning network is trained to output teeth geometry information including the set of locations, the color of the locations, and the density value.
The above deep learning network is a deep learning network based on NeRF (Neural Radiance Fields), an orthodontic treatment information providing system. In Article 11,
The above processor,
An orthodontic treatment information providing system that defines a pixel value for each pixel by adding up multiple pieces of dental geometric information for each pixel, and generates a rendered two-dimensional dental image by adding up the pixel values of each of the multiple pixels. In Article 12,
The above processor,
An orthodontic treatment information providing system, which calculates the mean square error between the original two-dimensional teeth image and the rendered two-dimensional teeth image, and generates the rendered two-dimensional teeth image by adding the difference in outlines between the rendered two-dimensional teeth image and the original two-dimensional teeth image to the calculated mean square error. In Article 13,
The above processor,
An orthodontic treatment information providing system that inputs the density value included in the above dental geometric information into a three-dimensional mesh model generation function and generates the above three-dimensional dental data using a marching cube function. In Article 14,
The above processor,
An orthodontic treatment information providing system that surfaces-aligns the crown of each tooth included in the initial 3D dental data with the crown of each tooth included in the generated 3D dental data, and extracts the 3D translation value and rotation value of each tooth as the 3D change amount.

Images