KR102608827B1 - Method of classifying implant classes for AI learning - Google Patents

Abstract

An implant class classification method for AI learning according to the present invention is provided. The implant class classification method for AI learning includes a structure recognition process that recognizes the coronal part, middle part, and apical part of the implant fixture; A pattern analysis process of analyzing the pattern of the shapes of the coronal part, middle part, and apical part; It includes an AI learning data set construction process of configuring an AI learning data set based on the shapes of the coronal part, middle part, and apical part.

Description

{Method of classifying implant classes for AI learning}

The present invention relates to an implant class classification method. More specifically, it is about an implant class classification method and device for AI learning.

Implants may include an implant fixture, an abutment, and a crown. In this regard, an implant refers to a fixture that is installed in the alveolar bone so as to support the alveolar bone with a prosthesis for the restoration of missing teeth. Meanwhile, an abutment refers to a pillar coupled to an implant for connection between the implant and a prosthesis. Additionally, a prosthesis means that it is manufactured in a shape similar to the patient's teeth.

Meanwhile, it is necessary to perform a search and automatic identification of the corresponding model from a camera image or X-ray image including the gum on which this implant fixture is mounted. In particular, users such as dentists need to search for and automatically identify the model from the patient's X-ray image. However, if the patient does not know about the implant model, there is a problem in that it is impossible to know which implant model it is.

Additionally, even if the patient knows about the implant model or can know the implant model from previous procedure information, verification of the model is required through search and automatic identification of the model from the patient's X-ray image.

Additionally, detailed information needs to be provided on various models and implant models of various sizes manufactured by various manufacturers. Even if a user, such as a dentist, can know the implant model from previous procedure information, there is a problem of having to search for detailed information about the implant model again. Additionally, there is a problem of having to separately acquire know-how on surgical methods for the corresponding implant model.

In this regard, artificial intelligence learning can be performed by classifying the entire implant fixture into one basic unit. However, the artificial intelligence learning method that classifies the entire implant fixture into one basic unit has a problem in that it is difficult to apply a wide range of models and construct learning data.

Therefore, the purpose of the present invention is to provide an implant class classification method and device for AI learning in order to solve the above-mentioned problems.

Additionally, the purpose of the present invention is to provide a function to automatically identify and classify implants based on implant X-ray images and search for a corresponding model from the implant X-ray images.

To solve the above problems, an implant class classification method for AI learning according to the present invention is provided. The implant class classification method for AI learning includes a structure recognition process that recognizes the coronal part, middle part, and apical part of the implant fixture; A pattern analysis process of analyzing the pattern of the shapes of the coronal part, middle part, and apical part; It includes an AI learning data set construction process of configuring an AI learning data set based on the shapes of the coronal part, middle part, and apical part.

According to one embodiment, the method extracts each shape of the coronal part, middle part, and apical part from the input image of a specific implant, and based on each extracted shape and the AI learning data set. An implant type/characteristic determination process for determining the type and characteristics of the input specific implant may be further included.

According to one embodiment, in the process of configuring the AI learning data set, the coronal part is classified into one of a plurality of classes according to the thread design of the upper part of the implant fixture, and the middle part is classified into the implant It is classified into a plurality of classes according to the body design and thread shape of the middle part of the fixture, and the apical part can be classified into a plurality of classes according to the body shape and hole shape of the lower part of the implant fixture. .

According to one embodiment, in the pattern analysis process, if the implant image including the implant fixture is an X-ray image, the connection feature, connection color, and connection type of the coronal part are It is possible to analyze the presence or absence of micro threads in the coronal part and the pattern of the shape of the micro threads without analyzing the pattern.

According to one embodiment, in the AI learning data set construction process and the implant type/characteristic determination process, if it is determined that there is a micro thread, the coronal part is V-shaped, rounded, square, or buttress with a micro thread. , reverse buttress, fin, no threads, and if it is determined that there are no micro threads, the coronal part is V-shaped, rounded, no threads, fin, square, buttress, without micro threads. It can be classified into any one class among reverse buttresses.

According to one embodiment, in the AI learning data set construction process and the implant type/characteristic determination process, the body shape of the middle part is classified as one of straight and tapered, and the thread shape of the middle part is V-shaped. It can be classified into one of the following classes: rounded, no threads, fin, buttress, reverse buttress, and square.

According to one embodiment, in the process of configuring the AI learning data set and the process of determining the type/characteristics of the implant, the groove of the apical part is classified into one class of presence or absence, and the groove of the apical part is classified into one class of presence or absence. The hole is classified into one class among no hole, round, and oblong, the part shape of the apical part is classified into one class among straight and tapered, and the apex shape of the apical part is flat, cone, and dome. , semi-dome, or flared.

According to one embodiment, the implant type/characteristic determination process includes an implant image input process of receiving an image of the specific implant; From the input image of the specific implant, the shape of the coronal part according to the presence or absence of the micro thread of the implant fixture, the body shape and thread shape of the middle part, and the presence or absence of groove, hole shape, and body shape of the apical part. A class-based feature extraction process that extracts class-based features; And it may include a class-based implant type/characteristic determination process of determining the type and characteristics of the input specific implant based on the extracted class-based characteristics and the AI learning data set.

According to one embodiment, in the implant image input process, a screen area where an X-ray image of the implant to be searched can be input is displayed on the display of the user terminal, and a user input is applied on the screen area, When a specific X-ray image is selected, the specific X-ray image including one or more implant images is displayed on the display, and when a second implant image is placed within the guide box among the one or more implant images, the guide box The image of the second implant located within can be captured and controlled to be transmitted to the artificial intelligence server.

According to one embodiment, in the implant image input process, each feature element is searched for each part from the second implant image and each feature element is synthesized to generate a final composite image, and the second implant image and each part are generated. A final composite image according to the identification information may be displayed on the display, and the final composite image may be displayed so that each feature element for each part is distinguished from other parts.

According to one embodiment, after performing the class-based implant type/characteristic determination process, when the search button displayed on the display is selected, the selected specific X-ray is based on the second implant image and the final composite image. An implant search process may be further included to search for a specific implant that optimally matches the image. During the implant search process, the optimally matching specific implant and related information about the specific implant may be displayed on the display.

According to at least one embodiment of the present invention, there is an advantage of providing an implant class classification method and device for AI learning.

In addition, according to at least one embodiment of the present invention, the advantage is that the type and characteristics of the implant can be determined by analyzing the pattern of the shapes of the coronal part 410, the middle part 420, and the apical part 430. there is.

The features and effects of the present invention described above will become more apparent through the following detailed description in conjunction with the accompanying drawings, and accordingly, those skilled in the art will be able to easily implement the technical idea of the present invention. You will be able to.

1A to 1C show screens before and after execution of an application program capable of performing implant search and automatic identification according to the present invention.
FIGS. 2A and 2B show screens associated with image acquisition and image transmission when the scan implant module corresponding to the automatic implant identification function is executed.
Figure 3 shows the detailed configuration of an implant search and automatic identification device and a system including the same according to the present invention.
Figure 4a shows the structure of the implant fixture classified into a coronal part 410, a middle part 420, and an apical part 430.
Figure 4b shows an example of a classification method according to structural subdivision of an implant fixture according to the present invention.
Figure 5 shows a conceptual diagram of the structural segmentation process of the implant fixture and the class classification process through pattern analysis according to the present invention.
Figures 6a and 6b show the detailed structure of the coronal part 410 of the implant fixture according to the present invention and the classes defined accordingly.
Figure 7 shows the thread shape of the coronal part 410 that can correspond to each class when there is a micro thread according to the present invention.
Figures 8a and 8b show the detailed structure of the middle part 420 of the implant fixture according to the present invention and the classes defined accordingly.
Figure 9 shows the thread shape of the middle part 420 that can correspond to each class classification when the middle part 420 of the implant fixture according to the present invention is a straight type.
Figure 10 shows the thread shape of the middle part 420 that can correspond to each class when the middle part 420 of the implant fixture according to the present invention is tapered.
Figures 11a and 11b show the detailed structure of the apical part 430 of the implant fixture according to the present invention and the classes defined accordingly.
Figure 12 shows the case where the apical part 430 of the implant fixture according to the present invention has a groove of yes and no hole and each shape is flat, cone, dome, or flared.
Figure 13 shows some examples of models that can be produced depending on the presence or absence of a groove, the presence or absence of a hole, and the shape of the hole and the shape of the apical part of the apical part 430 according to the present invention.
Figure 14 shows a flowchart of the implant class classification method for AI learning according to the present invention.

The features and effects of the present invention described above will become more apparent through the following detailed description in conjunction with the accompanying drawings, and accordingly, those skilled in the art will be able to easily implement the technical idea of the present invention. You will be able to.

Since the present invention can be modified in various ways and can have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all changes, equivalents, and substitutes included in the spirit and technical scope of the present invention.

When describing each drawing, similar reference numerals are used for similar components.

Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another.

For example, a first component may be named a second component, and similarly, the second component may also be named a first component without departing from the scope of the present invention. The term and/or includes any of a plurality of related stated items or a combination of a plurality of related stated items.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by a person of ordinary skill in the technical field to which the present invention pertains.

Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and unless clearly defined in the present application, should not be interpreted in an ideal or excessively formal sense. It shouldn't be.

The suffixes module, block, and part for components used in the following description are given or used interchangeably only considering the ease of writing the specification, and do not have distinct meanings or roles in themselves.

Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings so that those skilled in the art can easily implement them. In the following description of embodiments of the present invention, if a detailed description of a related known function or known configuration is judged to unnecessarily obscure the gist of the present invention, the detailed description will be omitted.

Hereinafter, the implant search and automatic identification device according to the present invention and the system for performing implant search and automatic identification based on implant images will be described.

In this regard, Figures 1A to 1C show screens before and after execution of an application program capable of performing implant search and automatic identification according to the present invention. Specifically, Figure 1a shows a state in which an application program capable of performing implant search and automatic identification according to the present invention is displayed on the background screen of a terminal.

Meanwhile, Figure 1b shows the initial screen when the application program capable of performing implant search and automatic identification according to the present invention is executed. Two images may be displayed on the initial screen where the application program that can perform implant search and automatic identification is executed. As an example, a first image showing a crown inserted into an implant and a second image related to this application program may be displayed together. The first image may be associated with an implant search and automatic identification module according to the invention, which may be referred to as an implant master. The second image may be associated with a clinical experience and knowledge sharing module. When the clinical experience and knowledge sharing module is executed, articles, photos, videos, etc. about clinical cases can be posted, and survey and evaluation functions can be performed.

When the implant master associated with the first image is executed, the initial screen and subsequent process are described as follows. In this regard, Figure 1c shows the initial screen when the implant master associated with the first image is executed.

Referring to FIG. 1C, when the implant master is executed, three icons may be displayed on the screen on the display. The first icon performs an automatic implant identification function and may be referred to as a scan implant icon. When the first icon is selected, structural classification of implant X-ray images according to machine learning techniques and automatic implant reading functions using deep learning can be implemented.

The second icon performs an implant search function and may be referred to as a pick implant features icon. When the second icon is selected, an intuitive search function of UI/UX techniques according to the implant structural characteristic image can be implemented.

The third icon allows you to check saved implant information and may be referred to as the saved implants icon. When the third icon is selected, an information confirmation function can be implemented to check basic information and detailed information about the stored implant.

As described above, a detailed description of the case in which the scan implant module corresponding to the automatic implant identification function according to the present invention is executed is as follows. In this regard, FIGS. 2A and 2B show screens associated with image acquisition and image transmission when the scan implant module corresponding to the automatic implant identification function is executed. Meanwhile, Figure 3 shows the detailed configuration of an implant search and automatic identification device and a system including the same according to the present invention. The implant search and automatic identification device of FIG. 3 may be a terminal (electronic device) in which an implant search and automatic identification module is installed, for example, a mobile terminal or a fixed terminal. In this case, the implant search and automatic identification module may be an application program that can perform implant search and automatic identification.

Referring to FIG. 3, the implant search and automatic identification device 100 may include a display 110 and a control unit 120. The implant search and automatic identification device 100 may include a wireless communication unit 130 and a memory 140. The implant search and automatic identification device 100 may be configured to interoperate with the artificial intelligence server 200. The artificial intelligence server 200 may be referred to as an artificial intelligence (AI) recognition server.

The wireless communication unit 130 may be operably coupled to the control unit 120, and the control unit 120 may be configured to control the wireless communication unit 130. The wireless communication unit 130 may provide a wireless interface to operably couple the device 100 and the artificial intelligence server 200. The wireless communication unit 130 may be configured to transmit information to or receive information from the artificial intelligence server 200 through an access point (AP). Additionally, the wireless communication unit 130 may be configured to transmit information to or receive information from the artificial intelligence server 200 through a base station.

The memory 140 may be operably coupled to the control unit 120 , and the control unit 120 may be configured to control the memory 140 . Memory 140 may be configured to store relevant information during the implant search and automatic identification process.

When the application program icon capable of performing implant search and automatic identification is selected on the desktop of the display 110 as shown in FIG. 1A, the initial screen of the application program may be displayed on the display 110 as shown in FIG. 1B. . When the first image is selected in the initial screen of FIG. 1B or the implant master is executed with basic settings, the application program may be executed as shown in FIG. 1C. Specifically, when the implant master is executed, the implant master module associated with the implant master may be executed. In this case, the control unit 120 corresponding to the processor may be linked with the application program 150 through an application program interface (API). Additionally, some execution modules of the application program 150 may be executed through the processor of the control unit 120.

Referring to FIG. 2A, an implant image may be acquired through a camera. Implant image acquisition can be accomplished in several ways: 1) Only implant images that require identification can be captured separately from images taken by dental X-ray equipment. In this case, the implant search and automatic identification module can be configured to directly interface with the X-ray imaging solution. 2) Implant images can be configured to be captured through a separate agent program. 3) Additionally, as shown in Figure 2a, the implant image may be configured to be captured through a camera on a screen or printed photo.

Meanwhile, the input implant image can be enlarged and reduced, rotated, and moved with two fingers so that the implant image to be identified is located within the guide box 110a of FIG. 2A. In this regard, the position of the line 110b at the top of the guide box 110a can be adjusted according to the judgment of the user (dentist) to define the line at the top of the implant fixture.

Referring to FIGS. 2A, 2B, and 3, the terminal 100 may be configured to capture an implant image located within the guide box 110a, transmit it to the artificial intelligence server 200, and wait for the transmission result. While waiting for the transmission result, separate notices, advertisement images, videos, etc. can be displayed on the display as shown in FIG. 2B.

Meanwhile, the implant class classification method for AI learning according to the present invention can be performed in conjunction with the AI server 200 by the application program 150 installed in the terminal 100. As an example, the application program 150 is driven by the control unit 120 corresponding to the processor of the terminal 100, and an implant class classification method for AI learning can be performed.

In relation to the present invention, the entire implant fixture may be defined as one class. However, the traditional artificial intelligence learning method that classifies the entire implant fixture into one basic unit may have difficulty applying all of the extensive models to construct learning data. Additionally, in the case of this method, if the shape is different in detail, it must all be classified into a new class, reflecting this.

For example, as of 2018, there are over 500 manufacturers, over 4,000 models, and tens of thousands of derivative models. Meanwhile, in deep learning algorithms, as the number of classes increases, sensitivity to class distinction decreases and the amount of training data required for each class increases significantly, which can significantly reduce accuracy. It can be said that it is almost impossible to configure all models in the world (including past discontinued models) with artificial intelligence learning data. In addition, when a new implant product is released, a new class must be added to the existing learning data set and learned, but it is difficult to learn quickly and the recognition rate can vary greatly depending on the situation.

Therefore, the present invention proposes a learning method through class classification according to the structural subdivision of the implant fixture. Meanwhile, the learning method through class classification according to the structural subdivision of the implant fixture of the present invention may be applied in combination with a manual search method performed through a user interface.

In relation to the learning method through class classification according to the structural subdivision of this implant fixture, Figure 4a shows the implant fixture as a coronal part (410), a middle part (420), and an apical part (430). )(apical part). Meanwhile, Figure 4b shows an example of a classification method according to structural subdivision of an implant fixture according to the present invention.

Referring to FIG. 4A, the top part, middle part, and bottom part of the implant fixture may be referred to as a coronal part 410, a middle part 420, and an apical part 430, respectively. Referring to FIG. 4B, the connection feature of the coronal part 410 is the upper connection part of the implant and cannot be confirmed in the X-ray image, so it can be excluded from classification. Additionally, the connection color cannot be confirmed in the X-ray image as it is the upper connection part of the implant, so it can be excluded from classification. Additionally, the connection type can be excluded from classification because it is difficult to distinguish in the image. In addition, the level classified as bone or tissue can be excluded from classification because it is difficult to distinguish due to changes depending on the condition of the gums.

Accordingly, in the present invention, the presence or absence of micro threads of the coronal part 410 and the pattern of the shape of the micro threads can be analyzed. Additionally, it can be classified according to the shape and design of the middle part 420. Additionally, the apical part 430 can be classified according to the presence or absence of a groove, hole shape, body shape, and apex shape.

The implant class classification method for AI learning according to the present invention divides the implant fixture structure into three stages through analysis of image patterns and recognizes the shape of the extracted pattern (screw line and body shape, etc.) using machine learning techniques. It can be. In this regard, Figure 5 shows a conceptual diagram of the structural segmentation process of the implant fixture and the class classification process through pattern analysis according to the present invention.

Referring to Figure 5, the implant fixture is subdivided into a coronal part 410, a middle part 420, and an apical part 430 corresponding to the upper part, middle part, and lower part. Afterwards, pattern analysis such as shape extraction can be performed for each part. Accordingly, an AI learning dataset can be constructed for a plurality of classes, for example, 35 classes (coronal 14, middle 9, apical 12).

First, the coronal part 410 can be defined into 14 classes according to the thread design of the upper part of the implant. In this regard, FIGS. 6A and 6B show the detailed structure of the coronal part 410 of the implant fixture according to the present invention and the classes defined accordingly.

Referring to FIG. 6A, image analysis can be performed focusing on the part marked with a different color in the coronal part 410 at the top of the implant, that is, the part where threads are formed on the side, which is the micro-thread part. Accordingly, class classification is possible without analyzing the image of the entire coronal part 410 of the implant fixture. Therefore, it is possible to shorten image analysis time and dataset construction and learning time for AI learning, while also enabling accurate model search.

In this regard, classification can be made preferentially according to the presence or absence of micro threads.

Referring to FIG. 6B, if it is determined that there are micro threads, the coronal part 410 may be classified into one of the following classes: V-shaped, rounded, square, buttress, reverse buttress, fin, and no threads. On the other hand, if it is determined that there are no micro threads, the coronal part 410 may be classified into any one of the following classes: V-shaped, rounded, no threads, fin, square, buttress, and reverse buttress.

Meanwhile, Figure 7 shows the thread shape of the coronal part 410 that can correspond to each class when there is a micro thread according to the present invention. Referring to FIG. 7, threads of some other shapes and/or forms may be classified into the same class classification, that is, one of V-shaped, rounded, square, buttress, reverse buttress, fin, and no threads. For example, the dimensions and some structures of the coronal part 410 of the implant fixture are different for the two models classified as V-shaped, but they can be classified into the same class because there is a micro thread and the thread shape is V-shaped. .

Next, the middle part 420 can be defined into 9 classes based on the body design and thread design of the middle part of the implant. In this regard, FIGS. 8A and 8B show the detailed structure of the middle part 420 of the implant fixture according to the present invention and the classes defined accordingly.

Referring to FIG. 8A, image analysis can be performed focusing on the parts displayed in different colors in the middle part 420 of the middle part of the implant, that is, the body shape and the thread shape. Accordingly, class classification is possible without analyzing the image of the entire middle part 420 of the implant fixture. Therefore, it is possible to shorten image analysis time and dataset construction and learning time for AI learning, while also enabling accurate model search.

In this regard, classification may be preferentially made according to body design.

Referring to FIG. 8B, the body shape of the middle part 420 may be classified as either straight or tapered. Additionally, the thread shape of the middle part 420 may be classified into one of the following classes: V-shaped, rounded, no threads, fin, buttress, reverse buttress, and square.

Meanwhile, Figure 9 shows the thread shape of the middle part 420 that can correspond to each class when the middle part 420 of the implant fixture according to the present invention is a straight type. Referring to FIG. 9, threads of some different shapes and/or forms may also be classified into the same class. For example, two models each classified as no threads, V-shaped, rounded, square, buttress, or reverse buttress may be classified into the same class. In this regard, the dimensions and some structures of the middle part 420 of the implant fixture may be different, for example the direction of thread formation. However, the thread shape is the same as no threads, the same as V-shaped, or the same as rounded. Alternatively, they may be identical to square and classified into the same class despite some dimensional or structural differences. Alternatively, the thread shape may be the same as a buttress or reverse buttress, so they can be classified into the same class despite some dimensional or structural differences.

In addition, Figure 10 shows the thread shape of the middle part 420 that can correspond to each class when the middle part 420 of the implant fixture according to the present invention is tapered. Referring to FIG. 10, threads of some different shapes and/or forms may also be classified into the same class. For example, two models each classified as no threads, V-shaped, rounded, square, buttress, reverse buttress, or fin may be classified into the same class. In this regard, the dimensions and some structures of the middle part 420 of the implant fixture may differ, for example in that some breaks are formed on the threads. However, because the thread shape is the same as no threads, the same as V-shaped, the same as rounded , or the same as square, they can be classified into the same class despite some dimensional or structural differences. Alternatively, the thread shape may be the same as buttress, the same as reverse buttress, or the same as fin, so they can be classified into the same class despite some dimensional or structural differences.

Next, the apical part 430 can be defined into 12 classes based on the presence or absence of a groove at the bottom of the implant, hole, part shape, and apex shape. In this regard, Figures 11a and 11b show the detailed structure of the apical part 430 of the implant fixture according to the present invention and the classes defined accordingly.

Referring to FIG. 11A, image analysis can be performed focusing on the parts displayed in different colors in the apical part 430 at the bottom of the implant, that is, the presence or absence of grooves, holes, part shape, and apex shape. Accordingly, class classification is possible without image analysis of the entire apical part 430 of the implant fixture. Therefore, it is possible to shorten image analysis time and dataset construction and learning time for AI learning, while also enabling accurate model search.

In this regard, classification can be made preferentially according to body design corresponding to the presence or absence of grooves and apex shape.

Referring to FIGS. 4B and 11B , the apical part 430 may be classified into one class of having a groove or not having a groove. Additionally, the hole of the apical part 430 may be classified into any one of no hole, round, and oblong classes. Additionally, the part shape of the apical part 430 may be classified into one of straight and tapered classes. Additionally, the apex shape of the apical part 430 may be classified into one of the following classes: cone, flat, dome, semi-dome, and flared.

Meanwhile, Figure 12 shows the case where the apical part 430 of the implant fixture according to the present invention has a shape of flat, cone, dome, semi-dome, or flared when the groove is yes and no hole is present. Referring to FIG. 12, all models may have something in common in that the apical part 430 has yes grooves and no holes. Meanwhile, two models classified as having the same apex shape may be equally classified as having a flat shape. Alternatively, two models classified as the same apex shape may be equally classified as cone shapes. Alternatively, two models classified as the same apex shape may be classified as having the same dome shape. Alternatively, two models classified as having the same apex shape may be equally classified as having a flared shape.

Referring to FIGS. 11B and 12 , threads of some other shapes and/or forms may be classified into the same class classification, that is, one of flat, cone, dome, semi-dome, and flared classes. For example, the dimensions and some structures, such as thread shapes, of the apical part 430 of the implant fixture may be different for two models classified as flat, cone, dome, semi-dome, and flared. However, the apex shapes have common features such as flat, cone, dome, semi-dome, and flared, so they can be classified into the same class.

Meanwhile, Figure 13 shows a model that is classified into different classes because the groove presence and hole shape of the apical part 430 according to the present invention are the same and the body shape is different. Referring to FIG. 13, a groove exists and the hole shape is round, but the apex shape is different into flat and dome, so they can be classified into different classes. As another example, there is no groove and there is no hole in common, but the body shape is different, such as flat, cone, and dome, so it can be classified into different classes. As another example, a groove exists and the hole shape is oblong, but the apex shape is different, such as flat and dome, so they can be classified into different classes. On the other hand, even if there is no groove and the apex shape is common to the dome, if the hole shape is different, it can be classified into different classes. Referring to FIG. 13, there is no groove and the apex shape is common as a dome, but in the case of no hole and round hole, they can be classified into different classes.

Considering the above-described class classification characteristics, an implant class classification method for AI learning according to the present invention will be described. In this regard, Figure 14 shows a flowchart of the implant class classification method for AI learning according to the present invention. Referring to FIG. 14, the implant class classification method may be configured to include a structure recognition process (S100), a pattern analysis process (S200), and an AI learning data set construction process (S300). Additionally, the implant class classification method may be configured to further include an implant type/characteristic determination process (S400). Meanwhile, the implant class classification method for AI learning may be performed by a processor of a device that performs the implant class classification method. As an example, it may be performed by an application program running on the control unit 120 of the terminal 100 of FIG. 3, and the application program may be linked to the AI server 200.

In the structure recognition process (S100), the coronal part 410, middle part 420, and apical part 430 of the implant fixture are recognized. In the pattern analysis process (S200), pattern analysis of the shapes of the coronal part 410, middle part 420, and apical part 430 may be performed through the operator's work. In the AI learning data set construction process (S300), pattern analysis is performed by the operator on the shapes of the coronal part 410, middle part 420, and apical part 430.

In this regard, the structure recognition process (S100), the pattern analysis process (S200), and the AI learning data set construction process (S300) can be classified as processes for learning, but are not limited thereto. As an example, a structure recognition process (S100), a pattern analysis process (S200), and an AI learning data set construction process (S200) may be performed to determine the type of implant.

Meanwhile, the type and characteristics of the input specific implant can be determined based on each shape and AI learning data set extracted in the implant type/characteristic determination process (S400). In this regard, considering the learning process and the discrimination process separately, from the image of the specific implant input in the implant type/characteristic determination process (S400), the coronal part 410, the middle part 420, and the AP Each shape of the knife part 430 can be extracted. Thereafter, in the implant type/characteristic determination process (S400), the type and characteristics of the input specific implant can be determined based on each extracted shape and the AI learning data set.

In the AI learning data set construction process (S300), the coronal part 410 may be classified into one of a plurality of classes according to the thread design of the upper part of the implant fixture. Additionally, the middle part 420 may be classified into a plurality of classes depending on the body design and thread shape of the middle portion of the implant fixture. Additionally, the apical part 430 may be classified into a plurality of classes depending on the body shape and hole shape of the lower part of the implant fixture.

In the pattern analysis process (S200), if the implant image including the implant fixture is an X-ray image, the pattern of the connection feature, connection color, and connection type of the coronal part 410 is It can be configured not to analyze. In addition, in the pattern analysis process (S200), it may be configured to analyze the presence or absence of micro threads of the coronal part 410 and the pattern of the shape of the micro threads.

In the AI learning data set construction process (S300) and the implant type/characteristic determination process (S400), class classification may be performed in a different way for each part.

If the coronal part 410 is determined to have micro threads, it may be classified into one of the following classes: V-shaped, rounded, square, buttress, reverse buttress, fin, or no threads with micro threads. Meanwhile, if it is determined that there are no micro threads, the coronal part 410 may be classified into any one of the following classes: V-shaped, rounded, no threads, fin, square, buttress, and reverse buttress without micro threads. .

Meanwhile, the body shape of the middle part 420 may be classified as straight or tapered. The thread shape of the middle part 420 may be classified into one of the following classes: V-shaped, rounded, no threads, fin, buttress, reverse buttress, and square.

On the other hand, the apical part 430 may be classified into one of two classes: presence of groove or absence of groove. The hole of the apical part 430 may be classified into one of the following classes: no hole, round, and oblong. The part shape of the apical part 430 may be classified into one of straight and tapered classes. The apex shape of the apical part 430 may be classified into one of flat, cone, dome, semi-dome, and flared classes.

The implant type/characteristics determination process (S400) may be configured to include an implant image input process (S410), a class-based characteristic extraction process (S420), and a class-based implant type/characteristics determination process (S430).

In the implant image input process (S410), an image of a specific implant can be input. In the implant image input process (S410), a screen area where the X-ray image of the implant to be searched can be input may be displayed on the display of the user terminal. Additionally, when a user input is applied to the screen area and a specific X-ray image is selected, the specific X-ray image including one or more implant images may be displayed on the display. also. When a second implant image is placed in the guide box among the one or more implant images, the second implant image located in the guide box can be captured and controlled to be transmitted to the artificial intelligence server.

In the implant image input process (S410), each feature element can be searched for each part from the second implant image, and each feature element can be synthesized to generate a final composite image. Additionally, the second implant image and a final composite image according to the identification information of each part can be displayed on the display. In this case, the final composite image may be displayed so that each feature element for each part is distinguished from other parts (see FIGS. 6A, 8A, and 11A).

In the class-based feature extraction process (S420), the shape of the coronal part 410, the body shape and thread shape of the middle part 420 according to the presence or absence of the micro thread of the implant fixture from the input image of the specific implant, Class-based characteristics regarding the presence or absence of grooves, hole shape, and body shape of the apical part 430 can be extracted.

In the class-based implant type/characteristic determination process (S430), the type and characteristics of the input specific implant can be determined based on the extracted class-based characteristics and the AI learning data set.

After performing the class-based implant type/characteristic determination process (S430), an implant search process (S440) may be further included. When the search button displayed on the display is selected, a specific implant can be searched based on the second implant image and the final composite image in the implant search process (S440). In this case, the specific searched implant is selected to optimally match a specific X-ray image.

In relation to the implant type/characteristic determination process (S400), the information identified for each part of the coronal part, middle part, and apical part can be compared with the information in the existing implant database to search for the corresponding implant model. . Specifically, as described above, in the implant search process (S440), a specific implant may be searched based on the second implant image and the final composite image. In this case, the specific searched implant is selected to optimally match a specific X-ray image.

In the implant type/characteristic determination process (S400), the information identified for each part of the coronal part, middle part, and apical part from the implant image including the implant fixture is compared with the information in the previously established implant database to determine the corresponding You can search for implant models. For example, in the implant search process (S440), the information identified for each part can be compared with the information in a previously established implant database to search for the corresponding implant model.

In addition, in the implant type/characteristic determination process (S400), for example, in the implant search process (S440), additional details associated with the searched implant model can be checked to finally determine whether the searched implant model matches the implant image. there is.

As described above, the terminal 100 may be configured to include a display 110 and a control unit 120. The terminal 100 may include a wireless communication unit 130 and a memory 140.

The terminal 100 may be configured to interoperate with the artificial intelligence server 200. Meanwhile, the terminal 100 may be configured to interface with the dental X-ray equipment 300.

Display 110 may be configured to display information associated with an implant image. When the captured implant image is located within the guide box 110a on the display 110, the control unit 120 may control the captured implant image located within the guide box to be transmitted to the artificial intelligence server 200.

In response, the artificial intelligence server 200 may generate an implant identification result for the captured implant image based on AI and transmit it to the terminal 100. Alternatively, the application program running on the terminal 100 may link with the artificial intelligence server 200 to generate an implant identification result for the captured implant image based on AI. In this regard, the control unit 120 may display information associated with the captured implant image and the implant image with the highest matching result according to the identification result on the display 110.

In the above, we looked at the implant class classification method for AI learning according to the present invention. The technical effects according to the present invention are as follows.

According to at least one embodiment of the present invention, there is an advantage of providing an implant class classification method and device for AI learning.

In addition, according to at least one embodiment of the present invention, the advantage is that the type and characteristics of the implant can be determined by analyzing the pattern of the shapes of the coronal part 410, the middle part 420, and the apical part 430. there is.

The features and effects of the present invention described above will become more apparent through the following detailed description in conjunction with the accompanying drawings, and accordingly, those skilled in the art will be able to easily implement the technical idea of the present invention. You will be able to.

Since the present invention can be modified in various ways and can have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all changes, equivalents, and substitutes included in the spirit and technical scope of the present invention.

According to software implementation, the procedures and functions described in this specification, as well as design and parameter optimization for each component, can be implemented as separate software modules. Software code can be implemented as a software application written in an appropriate programming language. The software code may be stored in memory and executed by a controller or processor.

Claims (10)

In the implant class classification method for AI learning, the method is performed by a processor of a device that performs the implant class classification method,
A structure recognition process that recognizes the coronal part, middle part, and apical part of the implant fixture;
A pattern analysis process of analyzing the pattern of the shapes of the coronal part, middle part, and apical part;
An AI learning data set construction process of configuring an AI learning data set based on the shapes of the coronal part, middle part, and apical part; and
From the image of the input specific implant, the respective shapes of the coronal part, middle part, and apical part are extracted, and the type and characteristics of the input specific implant based on each extracted shape and the AI learning data set. Includes an implant type/characteristic determination process that determines,
In the process of constructing the AI learning data set,
The coronal part is classified into one of a plurality of classes depending on the thread design of the upper part of the implant fixture,
The middle part is classified into a plurality of classes according to the body design and thread shape of the middle part of the implant fixture,
The apical part is classified into a plurality of classes according to the body shape and hole shape of the lower part of the implant fixture,
In the pattern analysis process,
If the implant image containing the implant fixture is an X-ray image, the pattern of the connection feature, connection color, and connection type of the coronal part is not analyzed. Analyzing the presence or absence of micro threads and the pattern of the shape of the micro threads,
In the AI learning data set construction process and the implant type/characteristic determination process,
If it is determined that there is a micro thread, the coronal part is classified into one of the following classes: V-shaped, rounded, square, buttress, reverse buttress, fin, and no threads with micro thread,
If it is determined that there are no micro threads, the coronal part is an implant class for AI learning that is classified into any one of the following classes without micro threads: V-shaped, rounded, no threads, fin, square, buttress, and reverse buttress. Classification method. delete delete According to claim 1,
In the AI learning data set construction process and the implant type/characteristic determination process,
The body shape of the middle part is classified as straight or tapered,
An implant class classification method for AI learning in which the thread shape of the middle part is classified into one of the following classes: V-shaped, rounded, no threads, fin, buttress, reverse buttress, and square. According to clause 4,
In the AI learning data set construction process and the implant type/characteristic determination process,
The apical part is classified into one of two classes: presence or absence of a groove,
The hole of the apical part is classified into one of the following classes: no hole, round, and oblong,
The part shape of the apical part is classified into one of straight and tapered classes,
An implant class classification method for AI learning in which the apex shape of the apical part is classified into one of flat, cone, dome, semi-dome, and flared classes. According to clause 5,
The implant type/characteristic determination process is,
An implant image input process of receiving an image of the specific implant;
From the input image of the specific implant, the shape of the coronal part according to the presence or absence of the micro thread of the implant fixture, the body shape and thread shape of the middle part, and the presence or absence of groove, hole shape, and body shape of the apical part. A class-based feature extraction process that extracts class-based features; and
An implant class classification method for AI learning, comprising a class-based implant type/characteristic determination process for determining the type and characteristics of the input specific implant based on the extracted class-based characteristics and the AI learning data set. According to clause 6,
In the implant image input process,
Displays a screen area on the display of the user terminal where the X-ray image of the implant to be searched can be entered,
When a user input is applied to the screen area and a specific X-ray image is selected, the specific X-ray image including one or more implant images is displayed on the display,
An implant class classification method for AI learning, wherein when a second implant image is placed in a guide box among the one or more implant images, the second implant image located in the guide box is captured and controlled to be transmitted to an artificial intelligence server. According to clause 7,
In the implant image input process,
Searching for each feature element for each part from the second implant image and synthesizing each feature element to generate a final composite image,
Displaying the second implant image and a final composite image according to the identification information of each part on the display,
The final composite image is displayed so that each feature element for each part is distinguished from other parts. An implant class classification method for AI learning. According to clause 8,
After performing the class-based implant type/characteristic determination process,
When the search button displayed on the display is selected, further comprising an implant search process of searching for a specific implant that optimally matches the selected specific X-ray image based on the second implant image and the final composite image,
During the implant search process,
An implant class classification method for AI learning, wherein the optimally matched specific implant and related information about the specific implant are displayed on the display. According to claim 1,
In the process of determining the implant type/characteristics,
Search for the corresponding implant model by comparing information identified for each part of the coronal part, middle part, and apical part from the implant image including the implant fixture with information in a previously constructed implant database,
An implant class classification method for AI learning, which checks additional details associated with the retrieved implant model and finally determines whether the retrieved implant model matches the implant image.