WO2021215582A1

WO2021215582A1 - Periodontitis automatic diagnosis method and program for implementing same

Info

Publication number: WO2021215582A1
Application number: PCT/KR2020/008388
Authority: WO
Inventors: 이원진; 이상정; 장혁준
Original assignee: 서울대학교산학협력단
Priority date: 2020-04-21
Filing date: 2020-06-26
Publication date: 2021-10-28
Also published as: KR102380166B1; KR20210130326A; KR102380166B9

Abstract

The present invention relates to a method for automatic diagnosis by a computing apparatus operated by at least one processor, the method comprising the steps of: detecting individual tooth regions comprising a tooth crown region and a tooth root region, an alveolar region supporting a tooth root region, and a tooth crown region, on the basis of the anatomical shape in a received dental radiographic image; setting the major axis passing through the center of each tooth, on the basis of the detected individual tooth regions, setting, as a tooth root point, an intersection of the major axis and the boundaries of individual tooth regions in the tooth root region, extracting, as a first intersection point, an intersection of the major axis and the boundary of an alveolar region, and extracting, as a second intersection point, an intersection of the major axis and the boundary in contact with the tooth root region in the tooth crown region; and calculating a subsidence rate of the alveolar bone for each individual tooth from a ratio of the length between the tooth root point and the first intersection point to the length between the tooth root point and the second intersection point.

Description

Periodontitis automatic diagnosis method and program implementing the same

It relates to the automatic diagnosis technology for periodontitis.

Periodontal disease is a disease that occurs in the tissues surrounding the teeth, such as the gingiva, periodontal ligaments, and alveolar bone surrounding the teeth. In detail, a form limited to the gums, that is, soft tissue, in a relatively light and fast recovery form is called gingivitis, and a case in which such inflammation has progressed to the gums and around the gums is called periodontitis.

Although periodontal disease is the most common disease, it can cause alveolar bone loss, tooth loss, damage to hemorrhoids (tooth enamel, dentin), chewing disorder, and decreased nutritional intake.

In general, a diagnosis method for periodontal disease is to use a manual instrument such as a periodontal probe to check the length of the probe that has entered the patient's gums and whether bleeding is caused, or to diagnose by reading a dental radiographic image of the patient's gums. There is a way.

However, when a periodontal probe is used, bleeding occurs easily due to periodontitis, so that the depth of the periodontal pocket can be measured more deeply, taking a lot of time, and causing great inconvenience to the patient. In addition, when a radiographic image is used, since it is greatly influenced by the subjective experience and career of a doctor who reads the radiographic image, there is a possibility that an inaccurate diagnosis may occur.

In addition, if a typical intraoral radiograph is performed during apical radiography, one or several teeth and the tissues around the teeth (alveolar bone) can be clearly checked, but it is difficult to check the condition of the entire tooth, and it is difficult to check the can On the other hand, panoramic radiography is a type of tomography that shows the upper jaw bone, the mandible bone, and the facial structures as one continuous radiograph. Panoramic radiography is relatively simple and can obtain a lot of necessary information from the entire tooth and alveolar bone through a small amount of x-ray exposure in a short time.

For accurate diagnosis, a specific tooth area is limited through a panoramic radiographic image, and the degree of periodontal disease progression is checked through intraoral radiographic imaging of the limited specific tooth area again. Accordingly, it takes a lot of radiographs, time and money to confirm periodontal disease.

Therefore, there is a need for a technology that automatically detects anatomical structures and regions in a panoramic radiographic image without relying on the experience of a surgeon and automatically provides objectively diagnosed information on the stage of periodontitis.

One embodiment of the present invention is an automatic periodontitis diagnosis technology that automatically detects an anatomical structure and area necessary for diagnosis from a radiographic image, and quantitatively calculates the degree of subsidence of alveolar bone in the anatomical structure and area to provide a degree of periodontitis progression is to provide

In addition to the above tasks, it may be used to achieve other tasks not specifically mentioned.

In a method for automatically diagnosing periodontitis by a computing device operated by at least one processor according to an embodiment of the present invention, an individual tooth including a crown region and a root region based on an anatomical shape in a received dental radiographic image detecting the region, the alveolar region supporting the root region, and the crown region, setting a long axis passing through the center for each tooth based on the detected individual tooth region, and the intersection of the long axis and the boundary line of the individual tooth regions in the root region Step of setting as the root point, extracting the intersection of the long axis and the boundary line of the alveolar region as the first intersection, and extracting the intersection of the long axis and the boundary line in contact with the root region in the crown region as the second intersection, And calculating the subsidence rate of the alveolar bone for each individual tooth through the ratio of the length between the root point and the first intersection to the length between the root point and the second intersection point.

One that collects learning data including individual tooth regions, alveolar bone regions, and crown regions of a corresponding dental radiographic image corresponding to a plurality of dental radiographic images so that each individual tooth region, alveolar bone region, and crown region are output from the dental radiographic image The method further includes training the above detection model.

The detection model may be implemented as a convolutional neural network that detects an individual tooth region, an alveolar bone region, and a crown region from a dental radiographic image as an individual image having the same resolution as a dental radiographic image.

The detecting may include pre-processing the collected dental radiographic image according to the input of the detection model, inputting the pre-processed input image to the learned detection model, and obtaining individual images for the detected area from the learned detection model. have.

In the step of setting the long axis, if the long axis passing through the center point of the tooth is set by applying the principal component analysis algorithm to each individual tooth region, image coordinates corresponding to the set long axis may be automatically extracted.

In the extracting step, after converting each individual tooth region, alveolar bone region, and crown region into a binarized image, the Moore neighbor tracking algorithm is applied to the transformed image to obtain a set of image coordinates for the boundary line of each region.

The extracting step is to detect the intersection between the image coordinates corresponding to the major axis and the image coordinate sets for each boundary line of each region, and the image coordinates for the apical point for each tooth, the image coordinates for the first intersection, and the image of the second intersection point Coordinates can be extracted.

In the step of calculating the subsidence rate of the alveolar bone, a value obtained by dividing the length between the image coordinates of the apical point and the image coordinates of the second intersection point by the length between the image coordinates of the root point and the image coordinates of the first intersection point is converted into a percentage to convert the subsidence of the alveolar bone It is possible to calculate the rate and provide a diagnosis step of periodontitis corresponding to the subsidence rate of the alveolar bone.

The step of calculating the subsidence rate of the alveolar bone is displayed by superimposing the boundary line of the alveolar bone region and the boundary line in contact with the root region in the crown region on the dental radiographic image, and based on the major axis for each tooth, the root point, the first intersection, and the second The alveolar bone subsidence rate for each intersection and tooth may be displayed and provided to the interlocking terminal.

An individual tooth region including a crown region and a root region based on an anatomical shape in the received dental radiographic image as a program stored in a computer-readable storage medium according to an embodiment of the present invention and executed by a processor , the operation of detecting the alveolar region supporting the root region, and the crown region, the operation of setting the long axis passing through the center for each tooth based on the detected individual tooth region, the intersection of the long axis and the boundary line of the individual tooth regions in the root region The operation of setting as the root point, extracting the intersection of the long axis and the boundary line of the alveolar region as the first intersection point, and extracting the intersection of the long axis and the boundary line in contact with the root region in the crown region as the second intersection point, And it includes instructions for executing an operation of calculating the subsidence rate of the alveolar bone for each individual tooth through the ratio of the length between the root point and the first intersection point to the length between the root point and the second intersection point.

One embodiment of the present invention can provide accurate periodontitis diagnosis information without relying on the experience of a doctor by automatically diagnosing and providing an objectively quantified periodontitis progression level in a dental radiographic image.

One embodiment of the present invention may provide periodontitis diagnosis information for each individual tooth through a panoramic radiographic image obtained with relatively low x-ray exposure without taking multiple radiographic images.

One embodiment of the present invention diagnoses periodontitis by automatically acquiring individual tooth regions, alveolar bone regions, and crown regions from a panoramic dental radiographic image, thereby reducing the time required for diagnosis and providing convenience to users and doctors. have

1 is a cross-sectional conceptual view schematically showing a mandibular tooth in periodontal disease.

2 is a block diagram of an automatic periodontitis diagnosis apparatus according to an embodiment of the present invention.

3 is a flowchart illustrating a method for diagnosing periodontitis using a learned detection model according to an embodiment of the present invention.

4 is an exemplary diagram illustrating a detection model according to an embodiment of the present invention.

5 is an exemplary view showing a boundary line for each tooth, an alveolar bone region boundary, and a boundary line between a root and a crown according to an embodiment of the present invention.

6 is a cross-sectional conceptual view schematically illustrating a mandibular tooth showing intersections of boundary lines according to an embodiment of the present invention.

7 is an exemplary view showing the subsidence rate of the alveolar bone compared to the boundary line between the root and the crown for each tooth according to an embodiment of the present invention.

8 is an exemplary view showing the stage of periodontitis diagnosed for each tooth according to an embodiment of the present invention.

9 is a hardware configuration diagram of a computing device according to an embodiment of the present invention.

With reference to the accompanying drawings, the embodiments of the present invention will be described in detail so that those of ordinary skill in the art to which the present invention pertains can easily implement them. The present invention may be embodied in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and the same reference numerals are used for the same or similar components throughout the specification. In addition, in the case of a well-known known technology, a detailed description thereof will be omitted.

Throughout the specification, when a part "includes" a certain element, it means that other elements may be further included, rather than excluding other elements, unless otherwise stated.

In the specification, the learning image and the input image refer to a dental radiographic image, and specifically refer to a panoramic radiographic image, but are not limited thereto.

Figure 1 (a) shows the structure of the mandible, (b) shows the tooth structure changed according to the stage of periodontal disease.

As shown in (a) of FIG. 1 , a tooth is formed of a crown positioned above the tooth and a tooth root, which means the root of the tooth. Specifically, a tooth is composed of pulp, dentin surrounding the root canal, enamel surrounding the upper side of the dentin, and cementum surrounding the lower side of the upper and lower dentin, and the crown represents the enamel area and the root represents the cementum area. And the chalk enamel boundary line or alveolar line means the boundary between the crown and the root, and corresponds to the actual boundary line between enamel and cementum.

In addition, the alveolar bone refers to the protruding parts of the maxilla (upper jawbone) and the mandible (mandibular bone), and refers to a protrusion-shaped bone tissue that serves to support the root of the tooth.

Periodontal disease is a generic term for lesions occurring in periodontal tissue, and inflammation spreads from the gingiva (gum) to the alveolar bone, causing the alveolar bone to become inflamed and melt or disappear.

Figure 1 (b) shows the outside and the inside of the tooth, and is an exemplary view listing the dental state changed according to the incidence of periodontal disease from the healthy state of the tooth.

Comparing the state of the left tooth and the state of the right tooth in FIG. 1( b ), it can be seen that the most characteristic feature of the periodontal disease is that the alveolar bone is lost and the root of the tooth is exposed to the outside.

Therefore, it is possible to quantitatively determine the degree of loss of alveolar bone (the rate of subsidence of alveolar bone) by calculating the ratio of the length of the root-alveolar junction to the length of the junction of the tooth root - alveolar enamel boundary.

As shown in FIG. 2 , the automatic periodontitis diagnosis apparatus 100 includes a pre-processing unit 110 that pre-processes a learning image according to an input format of a detection model, and a learning that trains one or more detection models using the pre-processed learning image, respectively. The unit 120 includes a detection unit 130 that extracts regions of teeth, alveolar bone, and crown by using a detection model from the pre-processed learning image. In addition, the automatic periodontitis diagnosis apparatus 100 sets the apical point as a reference, the control unit 140 for extracting intersection points between the extracted areas, calculates the length between the apical point and the intersection point, and according to the ratio of the calculated length between the intersection points It further includes a diagnosis unit 150 for outputting periodontitis diagnosis information by determining the subsidence rate of the alveolar bone.

For the sake of explanation, the preprocessor 110 , the learning unit 120 , the detection unit 130 , and the diagnosis unit 150 are named and called, but these are computing devices operated by at least one processor. Here, the preprocessing unit 110 , the learning unit 120 , the detecting unit 130 , the control unit 140 , and the diagnosis unit 150 may be implemented in one computing device or distributed in separate computing devices. When distributed in a separate computing device, the preprocessor 110 , the learner 120 , the detector 130 , the controller 140 , and the diagnosis unit 150 may communicate with each other through a communication interface. The computing device may be any device capable of executing a software program written to carry out the present invention, and may be, for example, a server, a laptop computer, or the like.

Here, the learning unit 120 may be implemented separately from the automatic periodontitis diagnosis apparatus 100 or may not be used as the learning of the detection model is completed.

The preprocessor 110 collects and classifies dental radiographic images (hereinafter, a learning image) for deep learning model learning, and converts them into learning data of the detection model. The preprocessor 110 may convert the training image into training data based on the input format of each detection model with respect to one or more detection models.

In this case, when one or more training images are received, the preprocessor 110 may exclude the training image when the noise of the training image is equal to or greater than the sleep sound threshold or the sharpness is less than or equal to the sharpness threshold. In addition, the preprocessor 110 may exclude the learning image including the mixed dentition (a state in which primary and permanent teeth are mixed).

And the preprocessor 110 is the amount of learning data through image enhancement such as left and right inversion or image rotation of the learning image for learning the detection model, linear movement and contrast change, Gaussian-blurring, etc. It can be amplified by about 64 times, etc. Here, the amount of amplified learning data can be easily changed by a user later.

Meanwhile, the preprocessor 110 may collect tooth or implant area data, alveolar bone area data, and chalk enamel boundary area data for the corresponding dental radiographic image together with the learning image.

Meanwhile, when the learning is completed and the real image is collected, the preprocessor 110 may convert the format of the real image based on the input format to the learned detection model.

The learning unit 120 uses the converted learning data as an input value to learn the detection model so that the collected tooth or implant area data, alveolar bone area data, and crown area data are obtained.

Here, the detection model can be implemented as Mask R-CNN proposed to detect and segment pixel-level objects in an image by improving the conventional Faster R-CNN.

Mask R-CNN consists of a branch that predicts an object mask in parallel with the existing branch to recognize the bounding box. Binary prediction.

As such, the detection model is an artificial intelligence model implemented by one deep learning or machine learning, and it is possible to acquire a tooth or implant area, an alveolar bone area, and a crown area through one artificial intelligence model, and each acquired area is independent. AI models can be implemented. Accordingly, one or a plurality of artificial intelligence models corresponding to the above-described configurations may be implemented by one or a plurality of computing devices.

The learning unit 120 trains the detection model using a graphics processing unit (GPU) server, and after learning once, calculates a loss value for the learning result using a loss function. Then, the learning unit 120 calculates a weight at which the calculated loss value is minimized, and retrains the detection model by applying the calculated weight.

The learning unit 120 may repeat learning for a preset number of times or end learning when the calculated function value reaches a preset threshold.

In other words, when receiving data on a dental radiographic image, the learning unit 120 repeatedly learns one or more detection models so that one or more regions among a tooth or implant region, an alveolar bone region, and a crown region in the dental radiographic image are output. .

The detector 130 applies the input image converted by the preprocessor 110 to one or more learned detection models to obtain a tooth or implant region, an alveolar bone region, and a crown region for the corresponding input image, respectively.

In addition, the detector 130 may convert the acquired tooth or implant region, alveolar bone region, and chalk enamel boundary region into a binarized image to extract outlines from the binarized image.

For example, the detector 130 may apply a contour tracking algorithm such as a Moore-neighbor tracking algorithm to the binarized image.

As such, the detection unit 130 extracts the boundary line for each individual tooth or implant, the alveolar bone area boundary line, and the chalk enamel boundary line. In this case, the detector 130 may extract the corresponding boundary lines in the form of a coordinate set in the input image.

The control unit 140 determines the long axis of the individual tooth or the long axis of the implant for the area of the tooth or implant, respectively.

Here, the long axis refers to the central axis of individual teeth or implants, and the controller 140 may derive it by applying a principal component analysis algorithm to the area of each tooth or implant.

And the control unit 140 is the long axis of each extracted individual tooth or implant, the boundary line for each individual tooth or implant, the boundary line of the alveolar bone area, and the boundary line that is in contact with the root area in the crown area (hereinafter referred to as the chalk enamel boundary line). to decide

The control unit 140 determines a point where the individual long axis and the corresponding tooth or implant boundary line intersect each other as the root point.

In this case, in the case of the mandible, the control unit 140 determines that the lowest point intersecting the long axis is the root point, and in the case of the upper jaw, the uppermost point intersecting the long axis is the root point.

And the controller 140 extracts the root point of the individual tooth or implant as image coordinates.

In this case, the controller 140 may provide the long axis and the root point of the individual tooth or implant, the alveolar bone region boundary line, and the chalk enamel boundary line over one image. In this case, the image may be an input image or a separately generated image.

In addition, the control unit 140 detects a point that intersects each other between the individual long axis and the alveolar bone region boundary as a first intersection point, and detects a point that crosses the individual long axis and the chalk enamel boundary line as a second intersection point.

At this time, the boundary line of the alveolar bone region is a connection structure between the maxilla and the mandible, and one line is created and the point that intersects the individual long axes is detected as one point. The intersecting point is detected as two different points.

Therefore, the diagnosis unit 150 may detect a point having a closer distance to an individual root point among the two detected points as the second intersection point.

As such, the diagnosis unit 150 detects the apical point, the first intersection, and the second intersection for each individual tooth and implant based on the individual long axis, and these points may be detected as coordinates of the image.

The diagnosis unit 150 may quantify the degree of subsidence of the alveolar bone for each individual tooth and implant by calculating the length between the root point and the first intersection point as a ratio to the length between the root point and the second intersection point.

In detail, the diagnosis unit 150 may provide diagnostic information by determining the progression stage of periodontal disease (periodontitis) divided into N stages according to the degree of subsidence of the alveolar bone.

For example, the progression stage of periodontitis is the area of alveolar bone compared to the cementum enamel boundary line according to the criteria determined at the 2017 World Workshop on the Classification of Periodontal and Peri-implant Diseases and Conditions. When the subsidence rate of the boundary line is 15% or less, it can be determined in Stage 1, when it is 15% or more and 33% or less, in Stage 2, and when it exceeds 33%, it can be determined in Stage 3. The stage of progression of periodontitis can be easily changed and designed according to a change in criteria such as subdivision of treatment stage and subdivision of diagnosis of periodontitis.

First, the automatic periodontitis diagnosis apparatus 100 learns one or more detection models so that each individual tooth region, alveolar bone region, and crown region are output from the dental radiographic image that is the collected learning data (S110)

The automatic periodontitis diagnosis apparatus 100 may design and build a detection model to detect a tooth, an implant, an area according to an oral structure, etc. based on an anatomical structure.

Accordingly, the automatic periodontitis diagnosis apparatus 100 collects learning data to build a detection model.

In collecting dental radiographic images as learning data, the automatic periodontitis diagnosis apparatus 100 determines the degree of noise, sharpness, etc., or determines whether shape distortion, mixed dentition, indwelling, etc. are included to obtain a dental radiographic image as a learning image. can be selected.

In addition, when a dental radiographic image is taken as an input value, the automatic periodontitis diagnosis apparatus 100 may collect data of an individual tooth region, an alveolar bone region, and a crown region that can be obtained as a result value together.

The automatic periodontitis diagnosis apparatus 100 calculates a weight at which the loss value is minimized by calculating a loss function for the learning result value when learning of the detection model is completed once. And the automatic periodontitis diagnosis apparatus 100 repeats learning through the detection model to which the calculated weight value is applied.

As such, the automatic periodontitis diagnosis apparatus 100 may complete the learning by building a detection model to which a weight determined through N repetition learning is applied. (N is a natural number)

Such a detection model may be implemented as a convolutional neural network that detects individual tooth regions, alveolar bone regions, and crown regions from a dental radiographic image as individual images having the same resolution as a dental radiographic image. In detail, the detection model may be implemented as an artificial intelligence model such as Mask R-CNN, but is not limited thereto.

In addition, more than one detection model can be implemented, and there are independent detection models such as a first detection model outputting each individual tooth region, a second detection model outputting an alveolar bone region, and a third detection model outputting a crown region. can be implemented. In this case, when outputting individual tooth regions, the teeth and implants may be output as independent detection results.

As such, step S110 is a step of learning the detection model, and after the learning of the detection model is completed, it may start from step S120.

Next, the automatic periodontitis diagnosis apparatus 100 detects an individual tooth area, an alveolar bone area, and a crown area based on the anatomical shape from the received dental radiographic image ( S120 ).

The automatic periodontitis diagnosis apparatus 100 pre-processes a dental radiographic image input according to the input format of the detection model in order to apply the received dental radiographic image to the detection model, and then inputs the pre-processed input image to the learned detection model. have.

The automatic periodontitis diagnosis apparatus 100 may detect an individual tooth region, an alveolar bone region, and a crown region from an image obtained by photographing all teeth by using one or more learned detection models.

In this case, each detected area may have the same image coordinates with the same resolution as that of the dental radiation area.

Meanwhile, the automatic periodontitis diagnosis apparatus 100 may convert each detected individual tooth region, alveolar bone region, and crown region into a binarized image, and then detect a boundary line of each region.

As shown in FIG. 4 , when the automatic periodontitis diagnosis apparatus 100 collects the user's dental radiographic image 10 , it inputs to the learned detection model 200 , The implant area, the alveolar bone area, and the crown area are acquired.

In this case, the automatic periodontitis diagnosis apparatus 100 has the same resolution as the inputted dental radiographic image 10 and the acquired images.

In detail, the automatic periodontitis diagnosis apparatus 100 outputs an independent detection result (mask) of the tooth and the implant in the process of individually detecting the area of the tooth or implant through the CNN 1 (210), and the output detection result Images can be superimposed. In other words, by setting the automatic periodontitis diagnosis apparatus 100 to have the same resolution, it is possible to overlap detection result images or to overlap an input dental radiographic image and detection result images without a separate process.

And the periodontitis automatic diagnosis apparatus 100 detects the alveolar bone region as one region connected to the mandible and the maxilla through the CNN 2 220 .

In the process of detecting the crown region corresponding to the crown of the tooth through the CNN 3 230 , the automatic periodontitis diagnosis apparatus 100 detects two regions for each of the maxillary and mandibular teeth.

As such, the automatic periodontitis diagnosis apparatus 100 may acquire a tooth or implant area, an alveolar bone area, and a crown area through the individually learned

detection models

210 , 220 and 230 , respectively.

In addition, the automatic periodontitis diagnosis apparatus 100 converts a region acquired through each detection model into a binary image such as a tooth or implant region 11 , an alveolar bone region 12 , and a crown region 13 .

If each region is converted into a binarized image, the shape of each region can be confirmed more clearly.

Next, the automatic periodontitis diagnosis apparatus 100 may detect the boundary line of each area by applying the Moore-Neighbor algorithm in the converted binarized image. In addition, the automatic periodontitis diagnosis apparatus 100 may acquire a set of image coordinates for a boundary line of each detected area.

5 (a), (b), and (c) show images from which boundary lines are derived after binarization of different dental radiographic images, respectively.

As shown in Figure 5, the boundary line 14 for each tooth or implant is represented by an individual boundary line for each tooth or implant, and the alveolar bone region boundary line 15 is shown as one boundary line based on the line of alveolar bone connected to the maxilla and mandible, The boundary line 16 between the root and the crown is divided into a maxillary line and a mandibular line, and is represented by two boundary lines.

As shown in (a), (b), and (c) of FIG. 5 , since each detected image has the same image coordinates with the same resolution, the images can be overlapped and displayed without a separate additional image processing process.

In detail, the automatic periodontitis diagnosis apparatus 100 may detect a boundary line corresponding to the root region from among individual teeth or implant boundary lines, and detect the chalk enamel boundary line in contact with the root region as a boundary line for the crown region.

Next, the automatic periodontitis diagnosis apparatus 100 sets the long axis passing through the center of the tooth based on the individual tooth area ( S130 ).

The automatic periodontitis diagnosis apparatus 100 may automatically extract image coordinates corresponding to the set major axis when a major axis passing through the center point of the tooth is set by applying a principal axes of inertia to each tooth area.

Here, the long axis is set on a one-to-one basis not only for each tooth but also for each implant, and has the same meaning as the central axis.

Next, the automatic periodontitis diagnosis apparatus 100 sets the intersection point of the set long axis and the boundary line of the individual tooth area as the root point ( S140 ).

The automatic periodontitis diagnosis apparatus 100 may set an intersection point with a long axis set in the root region among the boundaries of individual tooth regions as the root point.

In this case, the root point is set on a one-to-one basis for each tooth or implant, like the long axis, and serves as a reference point in the diagnosis of periodontitis.

Next, the automatic periodontitis diagnosis apparatus 100 detects the intersection of the long axis and the boundary line of the alveolar region as the first intersection point and detects the intersection of the long axis and the boundary line of the crown region as the second intersection point (S150).

Here, the boundary line of the alveolar bone region is one alveolar bone region boundary, and one intersection is detected for each major axis of each tooth or implant.

In addition, the automatic periodontitis diagnosis apparatus 100 detects an intersection point with the long axis based on the chalk enamel boundary line, which is a boundary line that is in contact with the root area, among the boundary lines of the crown area.

In other words, the automatic periodontitis diagnosis apparatus 100 detects the intersection between the image coordinates corresponding to the long axis and the image coordinate sets for each detected boundary line of each region, and the image coordinates for the root point for each tooth and the image of the first intersection point The image coordinates of the coordinates and the second intersection point may be extracted.

Next, the automatic periodontitis diagnosis apparatus 100 calculates the subsidence rate of the alveolar bone for individual teeth through the ratio of the length between the root point and the first intersection to the length between the root point and the second intersection (S160).

The automatic periodontitis diagnosis apparatus 100 converts a value obtained by dividing the length between the image coordinates of the root point and the image coordinates of the second intersection by the length between the image coordinates of the root point and the image coordinate of the first intersection into a percentage to determine the subsidence rate of the alveolar bone. can be calculated.

As shown in FIG. 6 , the automatic periodontitis diagnosis apparatus 100 generates a long axis passing through the center for each individual tooth, a boundary line for each tooth including a crown and a tooth root, an alveolar bone area boundary line, and a boundary line between the crown and the tooth root (the cementum enamel boundary line). ) to calculate the intersection point.

At this time, when a plurality of tooth roots are formed like a molar, the long axis is not set with respect to the center for each tooth root, and as shown in FIG. can be calculated.

In addition, the automatic periodontitis diagnosis apparatus 100 calculates the ratio of the first intersection point of the root point and the boundary line of the alveolar bone region to the second intersection point of the root point and the boundary line between the crown and the tooth root.

In other words, the automatic periodontitis diagnosis apparatus 100 divides the length (A) between the image coordinates of the root point and the image coordinates of the second intersection by the length (B) between the image coordinates of the root point and the image coordinates of the first intersection. , to calculate the alveolar bone settlement rate of the corresponding tooth or implant.

As such, the automatic periodontitis diagnosis apparatus 100 may calculate an alveolar bone settlement rate for each tooth, and classify the diagnosis stage of periodontitis based on the calculated alveolar bone settlement ratio.

For example, the automatic periodontitis diagnosis apparatus 100 can diagnose in stage 1 when the subsidence rate of alveolar bone compared to the chalk enamel boundary is 15% or less, in stage 2 when it is 15% or more and less than 33%, and in stage 3 when it exceeds 33%. .

Here, the diagnosis stage of periodontitis can be easily changed and set according to the treatment process or the progress of the study.

On the other hand, the automatic periodontitis diagnosis apparatus 100 superimposes and displays the boundary line of the alveolar bone area and the boundary line in contact with the root area in the crown area on the received dental radiographic image, and based on the major axis for each tooth, the root point, the first intersection point, and The second intersection may be displayed and provided to the interworking terminal.

Here, the terminal may include a doctor's terminal, a display, and the like, and the automatic periodontitis diagnosis apparatus 100 may store periodontitis diagnosis information in a server or database interworking in addition to the terminal.

As shown in FIG. 7 or FIG. 8 , a diagnosis image may be generated by overlapping and displaying an intersection or boundary line on a dental radiographic image, and also displaying periodontitis diagnosis information for each tooth or implant.

7 is an exemplary view showing the subsidence rate of alveolar bone compared to the chalk enamel boundary for each tooth according to an embodiment of the present invention, and FIG. 8 is a periodontitis stage diagnosed by tooth according to an embodiment of the present invention. It is also an example.

As shown in Fig. 7, the boundary line of the alveolar bone region and the cementum enamel boundary line are displayed on the dental radiographic image, and the settling rate for each major axis of each tooth or implant is shown in the image showing the long axis, the root point, the first intersection point, and the second intersection point. can be displayed and provided.

At this time, as described above, the settlement rate is quantified by converting it into a percentage, and a larger number means a larger settlement rate.

On the other hand, as shown in FIG. 8, it is possible to display the stage according to the progression stage of periodontitis, not the subsidence rate of the alveolar bone for each major axis.

As such, the automatic periodontitis diagnosis apparatus 100 has been exemplified as providing the periodontitis progression stage or the subsidence rate of the alveolar bone as a diagnosis result in a visual way at the root part of the long axis, but it is not necessarily limited to the corresponding position, and it is easy by the reader or the user can be changed and set.

9 is a hardware configuration diagram of a computing device according to an embodiment.

Referring to FIG. 9 , the pre-processing unit 110 , the learning unit 120 , the detecting unit 130 , the control unit 140 , and the diagnosis unit 150 are shown in the computing device 300 operated by at least one processor. Executes a program including instructions described to carry out the operations of the invention.

The hardware of the computing device 300 may include at least one processor 310 , a memory 320 , a storage 330 , and a communication interface 340 , and may be connected through a bus. In addition, hardware such as an input device and an output device may be included. The computing device 300 may be loaded with various software including an operating system capable of driving a program.

The processor 310 is a device that controls the operation of the computing device 300 , and may be a processor 310 of various types for processing instructions included in a program, for example, a central processing unit (CPU), an MPU ( It may be a micro processor unit), a micro controller unit (MCU), a graphic processing unit (GPU), or the like. The memory 320 loads the corresponding program so that the instructions described to execute the operation of the present invention are processed by the processor 310 . The memory 320 may be, for example, read only memory (ROM), random access memory (RAM), or the like. The storage 330 stores various data, programs, etc. required for executing the operation of the present invention. The communication interface 340 may be a wired/wireless communication module.

As such, the automatic periodontitis diagnosis apparatus 100 diagnoses and provides an objectively quantified periodontitis progression level in a dental radiographic image, thereby providing accurate periodontitis diagnosis information quickly and conveniently without relying on the experience of a reader.

In addition, it can be effectively used for establishing a treatment plan in the dental treatment area such as periodontal disease as well as implant surgery, and for comparative analysis before and after treatment.

A program for executing the method according to an embodiment of the present invention may be recorded in a computer-readable recording medium.

The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The media may be specially designed and configured, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include hard disks, magnetic media such as floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floppy disks, and ROMs, RAMs, flash memories, and the like. Hardware devices specially configured to store and execute the same program instructions are included. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

Although one preferred embodiment of the present invention has been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements by those skilled in the art using the basic concept of the present invention defined in the following claims are also It belongs to the scope of the present invention.

Claims

A method for automatically diagnosing periodontitis by a computing device operated by at least one processor, the method comprising:

Detecting an individual tooth region including a crown region and a root region, an alveolar region supporting the root region, and the crown region from the received dental radiographic image, based on the anatomical shape;

setting a long axis passing through the center for each tooth based on the detected individual tooth area, and setting the intersection point of the long axis and the boundary line of the individual tooth area in the root area as the root point;

Extracting the intersection of the long axis and the boundary line of the alveolar region as a first intersection, and extracting the intersection of the long axis and the boundary line in contact with the root region in the crown region as a second intersection, and

Calculating the subsidence rate of the alveolar bone for each individual tooth through the ratio of the length between the root point and the first intersection point to the length between the root point and the second intersection point;

A method of automatic diagnosis of periodontitis comprising a.
In claim 1,

Collecting learning data including individual tooth regions, alveolar bone regions, and crown regions of a corresponding dental radiographic image corresponding to a plurality of dental radiographic images so that each individual tooth region, alveolar bone region, and crown region are output from the dental radiographic image training one or more detection models;

Periodontitis automatic diagnosis method further comprising.
In claim 2,

The detection model is

An automatic periodontitis diagnosis method implemented by a convolutional neural network for detecting the individual tooth region, the alveolar bone region, and the crown region from the dental radiographic image as an individual image having the same resolution as the dental radiographic image.
In claim 2,

The detecting step is

After pre-processing the collected dental radiographic images according to the input of the detection model, the pre-processed input image is input to the learned detection model, and an individual image of the detected area is obtained from the learned detection model. Automatic diagnosis of periodontitis Way.
In claim 1,

The step of setting the long axis is,

An automatic periodontitis diagnosis method for automatically extracting image coordinates corresponding to the set long axis when a long axis passing through the center point of a tooth is set by applying a principal component analysis algorithm to each individual tooth area.
In claim 5,

The extraction step is

After converting each of the individual tooth regions, the alveolar bone region, and the crown region into a binarized image, a Moore neighbor tracking algorithm is applied to the transformed image to obtain a set of image coordinates for the boundary lines of each region.
In claim 6,

The extraction step is

By detecting the intersection between the image coordinates corresponding to the long axis and the image coordinate sets for each boundary line of each region, the image coordinates for the apical point for each tooth, the image coordinates of the first intersection and the image coordinates of the second intersection are obtained. Automatic diagnosis of periodontitis by extraction.
In claim 7,

The step of calculating the subsidence rate of the alveolar bone,

Calculate the subsidence rate of the alveolar bone by converting a value obtained by dividing the length between the image coordinates of the apical point and the image coordinates of the second intersection by the length between the image coordinates of the apical point and the image coordinates of the first intersection into a percentage, and the alveolar bone An automatic diagnosis method of periodontitis that provides a diagnostic step of periodontitis corresponding to the settlement rate of
In claim 7,

The step of calculating the subsidence rate of the alveolar bone,

The boundary line of the alveolar bone region and the boundary line in contact with the root region in the crown region are superimposed and displayed on the dental radiographic image, and based on the major axis of each tooth, the root point, the first intersection, the second intersection, and the alveolar bone for each tooth An automatic diagnosis method of periodontitis that displays the settlement rate and provides it to the interlocking terminal.
A program stored in a computer-readable storage medium and executed by a processor, comprising:

Detecting an individual tooth region including a crown region and a root region, an alveolar region supporting the root region, and the crown region from the received dental radiographic image based on an anatomical shape;

An operation of setting a long axis passing through the center for each tooth based on the detected individual tooth area;

The operation of setting the intersection of the long axis and the boundary line of the individual tooth area in the root area as the root point;

Extracting the intersection of the long axis and the boundary line of the alveolar region as a first intersection, and extracting the intersection of the long axis and the boundary line in contact with the root region in the crown region as a second intersection, and

Calculating the subsidence rate of the alveolar bone for each individual tooth through the ratio of the length between the root point and the first intersection point to the length between the root point and the second intersection point;

A program containing instructions to execute
In claim 10,

Collecting learning data including individual tooth regions, alveolar bone regions, and crown regions of a corresponding dental radiographic image corresponding to a plurality of dental radiographic images so that each individual tooth region, alveolar bone region, and crown region are output from the dental radiographic image The program further comprising the operation of training one or more detection models.
in paragraph 11

The detecting operation is

After preprocessing the collected dental radiographic image according to the input of the detection model, the preprocessed input image is input to the learned detection model, and the same resolution as the dental radiographic image for the area detected from the learned detection model is obtained. A program that acquires individual images.
in paragraph 12

The extraction operation is

By detecting the intersection between the image coordinates corresponding to the long axis and the image coordinate sets for each boundary line of each region, the image coordinates for the apical point for each tooth, the image coordinates of the first intersection and the image coordinates of the second intersection are obtained. extracting program.
In claim 13,

The operation of calculating the subsidence rate of the alveolar bone is,

Calculate the subsidence rate of the alveolar bone by converting a value obtained by dividing the length between the image coordinates of the apical point and the image coordinates of the second intersection by the length between the image coordinates of the apical point and the image coordinates of the first intersection into a percentage, and the alveolar bone A program that classifies the diagnostic stages of periodontitis based on the settlement rate of