Dental analysis device, dental analysis method, and dental analysis program

Abstract

To estimate the degree of progression of periodontal disease from a dental X-ray image using machine learning, and to output an analysis result including the estimation result.SOLUTION: According to a dental analysis device (10) of the present invention, the degree of progression of periodontal disease for each individual tooth is estimated from an image (110) of the individual tooth contained in a panoramic X-ray image (100) using a machine learning model. Then, an analysis result for each individual tooth, including the estimation result of the degree of progression of periodontal disease for the individual tooth, is output, and specifically, an analysis result screen 400 is displayed on a display (38).SELECTED DRAWING: Figure 12

Description

[Technical Field]

[0001] The present invention relates to a dental analysis device, a dental analysis method, and a dental analysis program, and more particularly to a dental analysis device, a dental analysis method, and a dental analysis program that perform dental analysis using machine learning. [Background technology]

[0002] In recent years, many dental analysis technologies using machine learning such as deep learning have been proposed. For example, Patent Document 1 discloses a technology that uses deep learning including the YOLO (You Only Look Once) system to detect lesions in dental panoramic X-ray images and determine the name of the disease. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2019-208831 Summary of the Invention [Problem to be solved by the invention]

[0004] In the field of dentistry, one of the major themes is how to combat periodontal disease, which is said to be a national disease in Japan. Therefore, if machine learning could be used to estimate (predict) the progression of periodontal disease from dental X-ray images, it is expected to make a significant contribution to supporting the diagnosis of periodontal disease.

[0005] Therefore, an object of the present invention is to provide a novel dental analysis device, dental analysis method, and dental analysis program that can estimate the progression of periodontal disease from dental X-ray images using machine learning. [Means for solving the problem]

[0006] To achieve this object, the present invention includes a first invention relating to a dental analysis device, a second invention relating to a dental analysis method, and a third invention relating to a dental analysis program.

[0007] A first aspect of the present invention, which relates to a dental analysis device, includes an estimation means and an information output means. The estimation means estimates the progression of periodontal disease for each tooth from an image of the tooth included in a dental X-ray image using a machine learning model. The information output means outputs analysis result information for each tooth, including the estimation result by the estimation means.

[0008] The machine learning model may detect each tooth, its cement-enamel junction (hereinafter referred to as "CEJ"), and its alveolar crest from an image of the tooth. The estimation means may estimate the progression of periodontal disease for each tooth based on the relative ratio between a first distance, which is the distance between the apex of the tooth detected by the machine learning model and the CEJ of the tooth, and a second distance, which is the distance between the CEJ of the tooth detected by the machine learning model and the alveolar crest of the tooth.

[0009] In this case, the machine learning models may include a first model, a second model, and a third model. The first model detects each individual tooth from an image of the tooth. The second model detects the CEJ of each individual tooth from an image of the tooth. And the third model detects the alveolar bone crest of each individual tooth from an image of the tooth.

[0010] The machine learning model may also be constructed by supervised learning based on multiple dental radiographs.

[0011] Furthermore, the information output means may output the analysis result information by displaying the analysis result information for each tooth.

[0012] In this case, the information output means may display a list of the analysis result information for each tooth.

[0013] The information output means may display the estimation result by the estimation means using at least one of a character string and a mark.

[0014] A second aspect of the present invention, which relates to a dental analysis method, includes an estimation step and an information output step. In the estimation step, a machine learning model is used to estimate the progression of periodontal disease for each tooth from an image of the tooth included in a dental X-ray image. In the information output step, analysis result information for each tooth including the estimation result obtained in the estimation step is output.

[0015] A third aspect of the present invention, which relates to a dental analysis program, causes a computer to execute an estimation step and an information output step. In the estimation step, a machine learning model is used to estimate the progression of periodontal disease for each tooth from an image of the tooth included in a dental X-ray image. In the information output step, analysis result information for each tooth including the estimation result obtained in the estimation step is output. [Effects of the Invention]

[0016] According to the present invention, it is possible to estimate the progression of periodontal disease from dental X-ray images using machine learning, which is particularly useful for assisting in the diagnosis of periodontal disease. [Brief explanation of the drawings]

[0017] [Figure 1] FIG. 1 is a diagram showing the configuration of a dental analysis system according to a first embodiment of the present invention. [Figure 2] FIG. 2 is a diagram showing an example of a panoramic X-ray image in the first embodiment. [Figure 3] FIG. 3 is a diagram for explaining how the dental analysis system according to the first embodiment estimates the progression of periodontal disease. [Figure 4]FIG. 4 is a diagram showing a detailed configuration of the AI ​​processing unit in the first embodiment. [Figure 5] FIG. 5 is a diagram for explaining how to detect individual teeth using the individual tooth detection model in the first embodiment. [Figure 6] FIG. 6 is a diagram showing an example of a detection result using the individual tooth detection model in the first embodiment. [Figure 7] FIG. 7 is a diagram showing a special example of how to detect individual teeth using the individual tooth detection model in the first embodiment. [Figure 8] FIG. 8 is a diagram showing an example of a detection result using the CEJ line detection model in the first embodiment. [Figure 9] FIG. 9 is a diagram showing an example of a detection result using the alveolar bone line detection model in the first embodiment. [Figure 10] FIG. 10 is a diagram showing an example of an analysis image in the first embodiment. [Figure 11] FIG. 11 is a diagram for explaining a method for estimating the progression stage of periodontal disease in the first embodiment. [Figure 12] FIG. 12 is a diagram showing an example of an analysis result screen in the first embodiment. [Figure 13] FIG. 13 is a diagram showing an example of another aspect of the analysis result screen in the first embodiment. [Figure 14] FIG. 14 is a flowchart showing the flow of processing by the control unit in the first embodiment. [Figure 15] FIG. 15 is a flow diagram showing the flow of learning of the machine learning model in the first embodiment. [Figure 16] FIG. 16 is a diagram showing another example of the analysis result screen in the first embodiment. [Figure 17] FIG. 17 is a diagram showing the configuration of a dental analysis system according to a second embodiment of the present invention. [Figure 18] FIG. 18 is a diagram showing the configuration of a dental analysis system according to a third embodiment of the present invention. [Figure 19] FIG. 19 is a diagram showing a conceptual configuration of an analysis result table in the fourth embodiment of the present invention. [Figure 20] FIG. 20 is a diagram showing an example of a diagnostic priority list screen in the fourth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0018] [First Example] A first embodiment of the present invention will be described with reference to FIGS.

[0019] As shown in Fig. 1, a dental analysis system 10 according to the first embodiment includes a dental panoramic X-ray apparatus 20 and a dental analysis apparatus 30. This dental analysis system 10 is installed, for example, in a dental clinic or a dental examination facility where dental examinations are available. The examination facility referred to here includes a mobile facility such as a dental examination bus.

[0020] The dental panoramic X-ray imaging device 20 comprehensively X-rays an area including the upper and lower jaws and their dentition of a subject (not shown), to obtain a panoramic X-ray image 100 as shown in Fig. 2. This panoramic X-ray image 100 (strictly speaking, digital data of the panoramic X-ray image 100) is input to a dental analysis device 30. Note that the dental panoramic X-ray imaging device 20 is a well-known device, and therefore a detailed description thereof will be omitted.

[0021] The dental analysis device 30 is a dedicated device, or so-called built-in equipment, that analyzes the panoramic X-ray images 100 input from the dental panoramic X-ray device 20, estimates the progression of periodontal disease for each tooth of the subject, and outputs analysis information for each tooth including the estimation results. This dental analysis device 30 has a control unit 32, a memory unit 34, an operation unit 36, a display 38, etc.

[0022] The control unit 32 is a control means that controls the entire dental analysis device 30. Although not shown in detail, the control unit 32 has a computer, such as a CPU, as a control execution means. The control unit 32 also has a main memory unit as a main storage means that the CPU can directly access. The main memory unit includes, for example, a ROM and a RAM. The ROM stores a control program (firmware) for controlling the operation of the CPU. The RAM forms a working area and a buffer area when the CPU executes processing in accordance with the control program. Furthermore, the control unit 32 has an AI processing unit 32a, or in other words, forms the AI ​​processing unit 32a. The AI ​​processing unit 32a will be described in detail later.

[0023] The storage unit 34 is an auxiliary storage means and includes, for example, a hard disk drive (not shown in detail). Various data, such as the panoramic X-ray image 100, are stored in the storage unit 34 as appropriate. In particular, the panoramic X-ray image 100 is stored in a compiled form in an electronic medical record 34a. The storage unit 34 may include a rewritable nonvolatile memory, such as a flash memory, in addition to or instead of the hard disk drive.

[0024] The operation unit 36 ​​is an operation receiving means for receiving operations by an operator (such as an analyst) (not shown), and may be, for example, a keyboard. The operation unit 36 ​​may also include a pointing device such as a mouse or a touchpad.

[0025] The display 38 is a display means, and various screens such as an analysis result screen 400, which will be described later, are displayed on the display 38. The display 38 is built into the dental analysis device 30, but may also be provided separately from the dental analysis device 30 as one of the external devices.

[0026] As described above, the dental analysis device 30 analyzes the panoramic X-ray image 100 input from the dental panoramic X-ray device 20 to estimate the progression of periodontal disease for each tooth of the subject, and outputs analysis information for each tooth including the estimation results. Of these, the estimation of the progression of periodontal disease is performed as follows.

[0027] That is, the dental analysis device 30 detects each tooth from an image (portion) 110 of the tooth included in the panoramic X-ray image 100, and identifies a baseline 120 (described later) that serves as a reference line for the tooth. Additionally, the dental analysis device 30 detects the CEJ of each tooth from the image 110 of the tooth included in the panoramic X-ray image 100. Furthermore, the dental analysis device 30 detects the alveolar bone crest of each tooth from the image of the tooth included in the panoramic X-ray image 100. As shown in FIG. 3, the CEJ is the boundary between the enamel and cementum of the tooth. The alveolar bone crest is the apex of the alveolar bone.

[0028] Then, the dental analysis device 30 estimates the progression of periodontal disease for each tooth based on the relative ratio (Lb / La) of the distance La (first distance) between the root apex, which is the portion of the base line 120 corresponding to the root apex, and the CEJ, to the distance Lb (second distance) between the CEJ and the alveolar bone crest, specifically based on an index value α calculated by the following formula 1. Note that the distance La between the root apex and the CEJ corresponds to the root length, and therefore this distance La will be referred to as the "root length" hereinafter. Furthermore, the distance Lb between the CEJ and the alveolar bone crest correlates with CAL (Clinical Attachment Level: the distance between the CEJ and the pocket bottom), and therefore this distance Lb will be referred to as the "CAL-correlated dimension" hereinafter.

[0029] 《Formula 1》 α=1-(Lb / La)

[0030] According to the index value α calculated by this formula 1, for example, the larger the index value α, the lower the progression of the periodontal disease, i.e., the milder it is estimated to be. On the other hand, the smaller the index value α, the higher the progression of the periodontal disease, i.e., the more severe it is estimated to be.

[0031] Specifically, two thresholds are set in advance: a first threshold β1 and a second threshold β2 (<β1) that is smaller than the first threshold β1. For example, if the index value α is larger than the first threshold β1 (α>β1), it is estimated that the stage of progression of periodontal disease is mild or that the patient is healthy. If the index value α is larger than the second threshold β2 and equal to or smaller than the first threshold β1 (β2<α≦β1), it is estimated that the stage of progression of periodontal disease is moderate (moderate). Furthermore, if the index value α is equal to or smaller than the second threshold β2 (α≦β2), it is estimated that the stage of progression of periodontal disease is severe.

[0032] At the World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions, held in November 2017 in Chicago, USA, co-sponsored by the American Academy of Periodontology and the European Federation of Periodontology, a new classification of periodontal disease was established. According to this classification, if the ratio of CAL to root length La (CAL / La) is 15% or less, the periodontal disease is classified as stage I, i.e., mild. If this ratio is 15% or greater but less than 33%, the periodontal disease is classified as stage II, i.e., moderate. Furthermore, if this ratio is greater than 33%, the periodontal disease is classified as stage III or IV, i.e., severe. Thus, this classification establishes two thresholds: 15% and 33%. Accordingly, the first threshold β1 is set to a value corresponding to (consistent with) the 15% threshold in this classification, e.g., 0.5. The second threshold value β2 is set to a value that corresponds to (matches) the threshold value of 33% in the relevant category, for example, 0.2.

[0033] In order to estimate the progression of periodontal disease in this manner, the AI ​​processing unit 32a has three machine learning models, namely, an individual tooth detection model 322 (first model), a CEJ line detection model 324 (second model), and an alveolar bone line detection model 326 (third model), as shown in Figure 4.

[0034] The individual tooth detection model 322 is, for example, a Mask_R-CNN model, and detects each tooth from the tooth image 110 included in the panoramic X-ray image 100 by instance segmentation, that is, extracts the tooth image 110. Note that the individual tooth detection model 322 is not limited to the Mask_R-CNN model, and may be any other appropriate model.

[0035] After each tooth is detected by the individual tooth detection model 322 in this manner, the tooth image 100 is surrounded by a rectangular frame 130, as shown in FIG. 5 . More specifically, the rectangular frame 130 circumscribes the contour of the tooth image 110 and has the smallest area. A straight line connecting the midpoints Pa and Pb of a pair of short sides 130a and 130b of the rectangular frame 130 is identified as the baseline 120. Identification of the baseline 120 using the rectangular frame 130 is performed by the control unit 32 (or by an appropriate element other than the AI ​​processing unit 32a), but may also be performed by an appropriate machine learning model. Each of the midpoints Pa and Pb corresponds to an end of the baseline 120.

[0036] The result of identifying the baseline 120 is shown in FIG. 6. Note that in FIG. 6, for ease of viewing, parts included in the panoramic X-ray image 100, such as the individual tooth images 110, are indicated by dashed lines. The individual tooth detection model 322 detects individual teeth based on the shading (light and dark) of the panoramic X-ray image 100, or is trained to do so. In particular, as shown in FIG. 7, when adjacent tooth images 110 and 110 overlap each other, there is a risk that these teeth may be erroneously detected as a single tooth. However, the individual tooth detection model 322 is trained to properly detect each individual tooth by focusing on the fact that the overlapping portion 110a is brighter than the individual (single) tooth image 110.

[0037] The CEJ line detection model 324, which may be a U-Net model, uses semantic segmentation to detect the CEJ of each tooth from each tooth image 110 included in the panoramic radiographic image 100. Specifically, as shown in FIG. 8, the CEJ line detection model 324 is trained to detect (identify) the CEJ line 140 connecting the CEJs of each tooth for each of the upper and lower jaws. Note that, for ease of viewing, parts included in the panoramic radiographic image 100, such as the individual tooth images 110, are indicated by dashed lines in FIG. 8. The CEJ line detection model 324 detects the CEJ line 140 based on the shading of the panoramic radiographic image 100. The CEJ line detection model 324 also detects the CEJ line 140 by taking into account the fact that each tooth is constricted near the CEJ. In Figure 8, the maxillary CEJ line 140 is formed in a loop shape that connects the CEJs of each maxillary tooth as well as the heads (tips of the crowns) of each of the maxillary teeth, but this is not limited to this and may be formed as a single line that connects only the CEJs of each maxillary tooth. Similarly, the mandibular CEJ line 140 is formed in a loop shape that connects the CEJs of each mandibular tooth as well as the heads of each of the mandibular teeth, but this is not limited to this and may be formed as a single line that connects only the CEJs of each mandibular tooth. The CEJ line detection model 324 is not limited to semantic segmentation and may employ instance segmentation, and is not limited to a U-Net model but may be any other appropriate model.

[0038] The alveolar bone line detection model 326 is also, for example, a U-Net model, which detects the alveolar bone crest of each tooth from the tooth image 110 included in the panoramic X-ray image 100 by semantic segmentation, and detects (identifies) the alveolar bone line 150 connecting the alveolar bone crests of each tooth, as shown in detail in Figure 9. Note that, in Figure 9 as well, for ease of viewing, parts included in the panoramic X-ray image 100, such as the tooth image 110, are indicated by dashed lines. The alveolar bone line detection model 326 also detects the alveolar bone line 150 based on the shading of the panoramic X-ray image 100. 9, the alveolar bone line 150 is formed in a loop shape so that the portion connecting the alveolar bone crests of the upper jaw teeth and the portion connecting the alveolar bone crests of the lower jaw teeth are continuous, but this is not limited to this, and the alveolar bone line connecting the alveolar bone crests of the upper jaw teeth and the alveolar bone line connecting the alveolar bone crests of the lower jaw teeth may be formed separately (separately).The alveolar bone line detection model 326 is also not limited to semantic segmentation and may employ instance segmentation, and may be not limited to a U-Net model but any other appropriate model.

[0039] Then, the control unit 32 superimposes the baselines 120 of the individual teeth detected by the individual tooth detection model 322, the CEJ lines 140 detected by the CEJ line detection model 324, and the alveolar bone line 150 detected by the alveolar bone line detection model 326 onto the panoramic X-ray image 100 to form an analysis image 100a as shown in Fig. 10. Note that in Fig. 10, for ease of viewing and convenience of explanation, parts included in the panoramic X-ray image 100, such as the individual tooth images 110, are indicated by solid lines, the baselines 120 of the individual teeth are indicated by short-dashed lines, the CEJ lines 140 are indicated by long-dashed lines, and the alveolar bone line 150 is indicated by a dash-dot line. However, in reality, the baselines 120, the CEJ lines 140, and the alveolar bone line 150 are indicated by solid lines of different colors, such as blue, green, and red, respectively.

[0040] 11, the control unit 32 identifies the end Pb of the base line 120 on the root side as the apex, which is the portion of the base line 120 that corresponds to the root apex. The control unit 32 also identifies the intersection Px of the base line 120 with the CEJ line 140 as the CEJ, and identifies the intersection Py of the base line 120 with the alveolar bone line 150 as the alveolar bone crest. The control unit 32 then calculates the root length La, which is the distance between the end Pb identified as the root apex and the intersection Px identified as the CEJ, and calculates the CAL correlation dimension Lb, which is the distance between the intersection Px identified as the CEJ and the intersection Py identified as the alveolar bone crest. The control unit 32 then calculates the index value α by substituting the calculated root length La and CAL correlation dimension Lb into the aforementioned Equation 1.

[0041] Additionally, the control unit 32 estimates the progression level of periodontal disease by comparing the calculated index value α with the aforementioned first threshold value β1 and second threshold value β2. That is, for example, if the index value α is greater than the first threshold value β1 (α>β1), the control unit 32 estimates that the progression level of periodontal disease is mild or that the subject is healthy. If the index value α is greater than the second threshold value β2 and less than or equal to the first threshold value β1 (β2<α≦β1), the control unit 32 estimates that the progression level of periodontal disease is moderate. Furthermore, if the index value α is less than or equal to the second threshold value β2 (α≦β2), the control unit 32 estimates that the progression level of periodontal disease is severe.

[0042] After estimating the degree of progression of periodontal disease for each tooth in this manner, the control unit 32 outputs the analysis results for each tooth, including the estimated results of the degree of progression of the periodontal disease, and displays an analysis result screen 400, as shown in detail in Figure 12, on the display 38.

[0043] 12, an original image 402 of an appropriate size representing the panoramic X-ray image 100 shown in Fig. 2 that was the subject of analysis is placed in the upper left position on the analysis result screen 400. In addition, an analysis image 404 of an appropriate size for reference representing the analysis image 100a shown in Fig. 10 is placed to the right of the original image 402 on the analysis result screen 400, that is, in the upper right position on the analysis result screen 400. Therefore, the operator can compare the original image 402 and the analysis image 404 with each other and, in particular, can confirm the baseline (short-dashed line), CEJ line (long-dashed line), and alveolar bone line (dash-dotted line) of each tooth from the analysis image 404.

[0044] Furthermore, below the original image 402 and the analysis image 404 on the analysis result screen 400, i.e., at the bottom of the analysis result screen 400, a list 406 showing the analysis results including the estimated results of the progression of periodontal disease for each tooth is arranged. In this list 406, a rectangular display field 408 is provided for each tooth, and the arrangement of the display fields 408 corresponds to the actual arrangement of the teeth. That is, the display fields 408 numbered "11" to "18" correspond to (a maximum of) eight teeth on the right side of the upper jaw, and the display fields 408 numbered "21" to "28" correspond to (a maximum of) eight teeth on the left side of the upper jaw. The display fields 408 numbered "31" to "38" correspond to (a maximum of) eight teeth on the left side of the lower jaw, and the display fields 408 numbered "41" to "48" correspond to (a maximum of) eight teeth on the right side of the lower jaw. In addition, the value in the ones digit of each number represents the type of tooth; that is, the eight values ​​from "1" to "8" represent central incisors, lateral incisors, canines, first premolars, second premolars, first molars, second molars, and third molars (wisdom teeth), respectively.

[0045] Each display field 408 has a mark 410 that indicates the estimated result of the progression of periodontal disease for the tooth corresponding to that display field 408. Then, below the mark 410 in each display field 408, a character string 412 that indicates the estimated result of the progression of periodontal disease for the tooth corresponding to that display field 408 is arranged. Furthermore, below the character string 412 in each display field 408, a number 414 that indicates the index value α for the tooth corresponding to that display field 408 is arranged. Therefore, by referring to the mark 410 and character string 412 in each display field 408, the operator can recognize (the estimated result of) the progression of periodontal disease for each tooth, and by referring to the mark 410 in particular, the operator can intuitively recognize the progression of the periodontal disease.

[0046] The mark 410 for a tooth estimated to have mild periodontal disease or to be healthy has the appearance of a check mark within a circle, and although it is not apparent from FIG. 12 , it is colored green, for example. The mark 410 for a tooth estimated to have moderate periodontal disease has the appearance of a triangular shape with an extension mark within it, and although it is not apparent from FIG. 12 , it is colored yellow, for example. The mark 410 for a tooth estimated to have severe periodontal disease has the appearance of a cross mark within a circle, and although it is not apparent from FIG. 12 , it is colored red, for example. This aspect of the mark 410 is merely an example and is not limiting. In addition, although it is not apparent from FIG. 12 , each character string 412 is colored the same as the mark 410 corresponding to that character string 412.

[0047] Additionally, by referring to the numbers 414 in each display field 408, the operator can recognize the index value α for each tooth, thereby more precisely ascertaining the stage of periodontal disease for each tooth. As previously mentioned, a larger index value α indicates a lower stage of periodontal disease, or in other words, a healthier (or healthier) tooth. Therefore, the operator can easily determine the healthiness of each tooth from the magnitude of the index value α. For example, if the stage of periodontal disease is estimated based on the relative ratio (Lb / La) between the root length La and the CAL correlation dimension Lb, a larger ratio indicates a higher stage of periodontal disease, i.e., a poorer tooth. Therefore, the operator will need some practice to determine the healthiness of each tooth from the magnitude of the ratio. Although not visible in FIG. 12 , the number 414 representing the index value α is colored the same as the mark 410 and character string 412 corresponding to the number 414.

[0048] Furthermore, at an appropriate position on the analysis result screen 400 between the original image 402, the analysis image 404, and the table 406, for example, toward the right, a switch button 416 is provided, which is an operator that accepts an operation to switch the display mode of the analysis result screen 400. By operating this switch button 416, the display mode of the analysis result screen 400 is selectively switched between a mode based on individual tooth examination and a mode based on block examination. FIG. 12 shows an example of the analysis result screen 400 in the mode based on individual tooth examination. When the switch button 416 is operated to switch from the mode based on individual tooth examination to the mode based on block examination in the state shown in FIG. 12, the analysis result screen 400 transitions to the mode shown in FIG. 13.

[0049] According to the analysis result screen 400 shown in Fig. 13, a block-based list 420 is arranged in place of the list 406 on the analysis result screen 400 shown in Fig. 12. In this block-based list 420, the teeth are divided into a plurality of blocks, for example, six blocks, and each block is provided with a display column 422. Specifically, the teeth of the upper jaw are divided into three blocks, a central block, a right block, and a left block, and the teeth of the lower jaw are divided into three blocks, a central block, a right block, and a left block. More specifically, a total of six teeth numbered "11" to "13" and "21" to "23" (i.e., the central incisors, lateral incisors, and canines on both the left and right sides of the maxilla) are classified into the central block of the maxilla, a maximum of five teeth numbered "14" to "18" (i.e., the first premolar, second premolar, first molar, second molar, and third molar on the right side of the maxilla) are classified into the right block of the maxilla, and a maximum of five teeth numbered "24" to "28" (i.e., the first premolar, second premolar, first molar, second molar, and third molar on the left side of the maxilla) are classified into the left block of the maxilla. A total of six teeth numbered "31" to "33" and "41" to "43" (i.e., the central incisors, lateral incisors, and canines on both the left and right sides of the lower jaw) are divided into a central block of the lower jaw, a maximum of five teeth numbered "34" to "38" (i.e., the first premolar, second premolar, first molar, second molar, and third molar on the left side of the lower jaw) are divided into a left block of the lower jaw, and a maximum of five teeth numbered "44" to "48" (i.e., the first premolar, second premolar, first molar, second molar, and third molar on the right side of the lower jaw) are divided into a right block of the lower jaw. Then, a display column 422 is provided for each block.

[0050] Each display column 422 is provided with a mark 424 indicating the overall estimated result of the degree of progression of periodontal disease for the teeth classified into the block corresponding to that display column 422, for example, the estimated result for the tooth with the most advanced periodontal disease among the teeth classified into that block. That is, for example, if the degree of progression for the tooth with the most advanced periodontal disease among the teeth classified into the block corresponding to that display column 422 is severe, a mark 424 corresponding to the severe degree of progression is provided in that display column 422. Furthermore, if the degree of progression for the tooth with the most advanced periodontal disease among the teeth classified into the block corresponding to that display column 422 is moderate, a mark 424 corresponding to the moderate degree of progression is provided in that display column 422. Furthermore, if the tooth with the most advanced stage of periodontal disease among the teeth divided into blocks corresponding to each display column 422 has a mild (or healthy) stage of periodontal disease, that is, if all the teeth divided into that block have a mild stage of periodontal disease, a mark 424 corresponding to the mild stage of periodontal disease is placed in that display column 422.

[0051] Therefore, the operator can intuitively recognize the overall stage of periodontal disease for the teeth divided into each block by referring to the marks 424 in each display field 422. The analysis result screen 400 according to the block examination shown in Fig. 13 is suitable for roughly recognizing the stage of periodontal disease for each tooth, for example.

[0052] For example, the appearance of the mark 424 on the analysis result screen 400 in the mode according to the block examination is the same as the mark 410 on the analysis result screen 400 in the mode according to the individual tooth examination shown in Fig. 12, but the size of the mark 424 on the analysis result screen 400 in the mode according to the block examination is larger than the size of the mark 410 on the analysis result screen 400 in the mode according to the individual tooth examination. Furthermore, the analysis result screen 400 in the mode according to the block examination does not have the character string representing the progression level of periodontal disease and the number representing the index value α described above, as on the analysis result screen 400 in the mode according to the individual tooth examination. However, the analysis result screen 400 in the mode according to the block examination may also have the character string representing the overall progression level of periodontal disease for each block and the number representing the index value α described above, as on the analysis result screen 400 in the mode according to the individual tooth examination. Then, when the mode shown in FIG. 13 is switched from the mode according to the block examination to the mode according to the individual tooth examination by operating the switching button 416, the analysis result screen 400 transitions to the mode shown in FIG.

[0053] 14 is a flowchart showing the processing flow of the control unit 32 in the first embodiment. As shown in FIG. 14, in step S1, the control unit 32 first uses the individual tooth detection model 322 to detect each tooth from the image 110 of the tooth included in the panoramic X-ray image 100. Note that before step S1, appropriate preprocessing may be performed, such as removing (trimming) portions of the panoramic X-ray image 100 that are unnecessary for estimating the progression of periodontal disease. After executing step S1, the control unit 32 advances the processing to step S2.

[0054] In step S2, the control unit 32 identifies the baseline 120 of each tooth in the manner described with reference to Fig. 5. Then, in the following step S3, the control unit 32 uses the CEJ line detection model 324 to detect the CEJ line 140 from each tooth image 110 included in the panoramic X-ray image 100. Furthermore, in the following step S5, the control unit 32 uses the alveolar bone line detection model 326 to detect the alveolar bone line 150 from each tooth image 110 included in the panoramic X-ray image 100. Thereafter, the control unit 32 proceeds to step S7.

[0055] In step S7, the control unit 32 calculates the root length La in the manner described with reference to Figures 10 and 11, and then in the following step S9, calculates the CAL correlation dimension Lb in the manner described with reference to Figures 10 and 11. Then, in the following step S11, the control unit 32 calculates the index value α based on the above-mentioned equation 1, that is, by substituting the root length La calculated in step S7 and the CAL correlation dimension Lb calculated in step 9 into equation 1. Thereafter, the control unit 32 proceeds to step S13.

[0056] In step S13, the control unit 32 estimates the progression level of periodontal disease by comparing the calculated index value α with the first threshold value β1 and the second threshold value β2. That is, for example, if the index value α is greater than the first threshold value β1 (α>β1), the control unit 32 estimates that the progression level of periodontal disease is mild or that the subject is healthy. If the index value α is greater than the second threshold value β2 and less than or equal to the first threshold value β1 (β2<α≦β1), the control unit 32 estimates that the progression level of periodontal disease is moderate. Furthermore, if the index value α is less than or equal to the second threshold value β2 (α≦β2), the control unit 32 estimates that the progression level of periodontal disease is severe. Thereafter, the control unit 32 proceeds to step S15.

[0057] In step S15, the control unit 32 displays the analysis result screen 400 on the display 38. Specifically, the analysis result screen 400 reflects the calculation result of the index value α in step S11 and the estimation result of the progression of periodontal disease for each tooth in step S13. The control unit 32 then completes the series of processes. As described above, when the switch button 416 on the analysis result screen 400 is operated, the mode of the analysis result screen 400 switches in response to this operation; in other words, the control unit 32 performs appropriate processing to achieve this. The control unit 32 also stores the analysis results, such as the analysis result screen 400, in the memory unit 34, and more specifically, compiles them in the electronic medical record 34a.

[0058] FIG. 15 is a flow diagram showing the learning flow for each of the individual tooth detection model 322, the CEJ line detection model 324, and the alveolar bone line detection model 326. As shown in FIG. 15, first, in step S101, annotation is performed, i.e., multiple labeled data are prepared as training data. For example, for the individual tooth detection model 322, data in which the contours of individual tooth images 110 contained in each of the multiple panoramic X-ray images 100 are precisely labeled (traced) is prepared as training data for each of the multiple panoramic X-ray images 100. For the CEJ line detection model 324, data in which the ideal CEJ line 140 is labeled for each of the multiple panoramic X-ray images 100 is prepared as training data. Furthermore, for the alveolar bone line detection model 326, data in which the ideal alveolar bone line 150 is labeled for each of the multiple panoramic X-ray images 100 is prepared as training data. After the annotation in step S101, data augmentation is performed in the following step S103. Note that the panoramic X-ray image 100 to be annotated may also be subjected to appropriate preprocessing, such as deleting unnecessary portions from the panoramic X-ray image 100.

[0059] In data augmentation in step S103, training data is augmented. For example, images in which the shading, tilt, left-right direction, etc. of each of the plurality of panoramic X-ray images 100 have been appropriately changed are used as training data, thereby augmenting the training data. Then, in the following step S105, learning (supervised learning) is performed for each of the individual tooth detection model 322, the CEJ line detection model 324, and the alveolar bone line detection model 326 using the plurality of training data including the augmented data.

[0060] The learning in step S105 constructs a practical individual tooth detection model 322, a CEJ line detection model 324, and an alveolar bone line detection model 326. Furthermore, repeated learning, particularly by providing more accurate training data, improves the detection accuracy of each of the individual tooth detection model 322, the CEJ line detection model 324, and the alveolar bone line detection model 326, and ultimately improves the accuracy of estimating the progression of periodontal disease.

[0061] For example, when the number of training data sets including padding was 100, the detection accuracy (pixel accuracy) of the individual tooth detection model 322, the CEJ line detection model 324, and the alveolar bone line detection model 326 was 92%, 95%, and 97%, respectively. Therefore, the final estimation accuracy of the progression of periodontal disease was an extremely high value of approximately 85% (= 92% × 95% × 97%).

[0062] Among the detection accuracies of the individual tooth detection model 322, the CEJ line detection model 324, and the alveolar bone line detection model 326, the detection accuracy of the individual tooth detection model 322 (92%) is slightly lower than the other detection accuracies (95% and 97%). One reason for this is thought to be a lack of variety in the training data provided to the individual tooth detection model 322. That is, human teeth vary in shape, size, alignment, and other characteristics, some of which are unique. Therefore, by providing the individual tooth detection model 322 with more accurate training data based on panoramic X-ray images 100 of various teeth, including teeth with unique characteristics, the detection accuracy of the individual tooth detection model 322 is improved, and ultimately, it is thought that the accuracy of estimating the progression of periodontal disease is further improved.

[0063] As described above, according to the first embodiment, it is possible to estimate the stage of periodontal disease for each tooth from the panoramic X-ray image 100 using machine learning, i.e., it is possible to estimate the stage of periodontal disease without performing a probing test, which significantly contributes to supporting the diagnosis of periodontal disease.

[0064] Furthermore, according to the first embodiment, the progression of periodontal disease is estimated based on the index value α calculated by the above-mentioned formula 1, in other words, based on the ratio (Lb / La) between the tooth root length La and the CAL correlation dimension Lb, and therefore is not affected by individual differences in the panoramic X-ray images 100, particularly by differences in the size of the tooth images 110 between the panoramic X-ray images 100. In other words, although the size of the tooth images 110 may differ between the panoramic X-ray images 100 (which is almost always the case), according to the first embodiment, the progression of periodontal disease can be accurately estimated without being affected by this.

[0065] In the first embodiment, the control unit 32 is responsible for estimating the progression of periodontal disease for each tooth, and this control unit 32 is an example of an estimation means according to the present invention. The control unit 32 also displays the analysis result screen 400 on the display 38, and this control unit 32, in cooperation with the display 38, constitutes an example of an analysis information output means according to the present invention.

[0066] Furthermore, in the first embodiment, the AI ​​processing unit 32a is configured by the control unit 32, but the AI ​​processing unit 32a may be provided separately from the control unit 32 (as a separate element).

[0067] Furthermore, in addition to or instead of displaying the analysis result screen 400 on the display 38, the control unit 32 may output, by voice, information similar to that displayed by the analysis result screen 400. In this case, a voice output means such as a speaker will be provided.

[0068] In addition, instead of the analysis result screen 400 shown in FIG. 12, the analysis result screen 400 shown in FIG. 16 may be displayed on the display 38.

[0069] According to the analysis result screen 400 shown in Fig. 16, instead of the table 406 on the analysis result screen 400 shown in Fig. 12, a list 430 is provided which is a collection of roughly rectangular display fields 432 for each individual tooth. The arrangement of the display fields 432 included in this list corresponds to the arrangement of the actual teeth, similar to the display field 408 in Fig. 12. Then, for example, at the top of each display field 432, a character string 434 similar to the character string 412 in Fig. 12 is arranged, i.e., the character string 434 represents the estimated result of the progression of periodontal disease for the tooth corresponding to the display field 432. Then, below the character string 434, a mark 436 similar to the mark 410 in Fig. 12 is arranged, i.e., the mark 436 represents the estimated result of the progression of periodontal disease for the tooth corresponding to the display field 432. Furthermore, below mark 436, in other words, at the bottom of display field 432, another bar graph-like mark, i.e., bar graph-like mark 438, is placed, which indicates the estimated result of the progression of periodontal disease for the tooth corresponding to that display field 432. Furthermore, below each display field 432, a number 440 similar to number 414 in Fig. 12, i.e., a number 440 indicating the index value α for the tooth corresponding to that display field 432, is placed.

[0070] Therefore, the operator can more intuitively recognize the progression of periodontal disease for each tooth by referring to the bar graph mark 438 in addition to the character string 434 and mark 436 in each display field 432. In addition, the operator can recognize the index value α for each tooth by referring to the numbers 440 arranged below each display field 432, that is, the operator can recognize the progression of periodontal disease for each tooth in more detail.

[0071] For example, the bar graph mark 438 for a tooth estimated to have mild periodontal disease or be healthy is the longest (tallest) and is colored the lightest. The bar graph mark 438 for a tooth estimated to have moderate periodontal disease is slightly shorter (shorter) and is colored the darkest. The bar graph mark 438 for a tooth estimated to have severe periodontal disease is the shortest and is colored the darkest. Therefore, the operator can intuitively recognize the stage of periodontal disease for each tooth from not only the length (height) of each bar graph mark 438 but also the color intensity (lightness) of the bar graph mark 438. Although not apparent from FIG. 16 , while each bar graph mark 438 is colored red, the color is not limited to red and may be different colors depending on the stage of periodontal disease.

[0072] Furthermore, in the first embodiment, the rectangular frame 130 is used to specify the aforementioned base line 120, but this is not limiting. That is, the base line 120 of each tooth may be specified in a manner different from that described with reference to Fig. 5. In extreme terms, it is sufficient to specify the root length La and the CAL correlation dimension Lb even if the base line 120 is not specified.

[0073] [Second Example] Next, a second embodiment of the present invention will be described with reference to FIG.

[0074] The dental analysis system 10a according to the second embodiment includes a dental analysis device 50 that performs the same processing as the dental analysis device 30 in the first embodiment, specifically, that estimates the progression of periodontal disease for each tooth from a panoramic X-ray image 100. The dental analysis device 50 is provided as a cloud server. In a facility such as a dental clinic where the dental panoramic X-ray imaging device 20 is installed, a management device 60 and a storage device 70 are provided instead of the dental analysis device 30 in the first embodiment. The management device 60 is, for example, a personal computer (hereinafter referred to as a "PC") and has a display 62. The storage device 70 is provided external to the management device 60, but may also be built into the management device 60.

[0075] According to the dental analysis system 10a of the second embodiment, the panoramic X-ray image 100 (strictly speaking, digital data of the panoramic X-ray image 100) obtained by the dental panoramic X-ray imaging apparatus 20 is input to the management device 60. The management device 60 stores the panoramic X-ray image 100 input from the dental panoramic X-ray imaging apparatus 20 in the storage device 70, and more specifically, compiles the panoramic X-ray image 100 into an electronic medical record 70a stored in the storage device 70. At the same time, the management device 60 sends, i.e., uploads, the panoramic X-ray image 100 to the dental analysis device 50.

[0076] The dental analysis device 50 includes a control unit 52 having an AI processing unit 52a similar to that in the first embodiment. The dental analysis device 50 analyzes the panoramic X-ray images 100 uploaded from the management device 60 to estimate the progression of periodontal disease for each tooth. The dental analysis device 50 then sends data to the management device 60 for displaying the analysis results for each tooth, including the estimated progression of periodontal disease for that tooth, on the display 38 of the management device 60 as an analysis result screen 400, which is exactly the same as that in the first embodiment.

[0077] When the management device 60 receives the analysis results sent from the dental analysis device 50, it outputs the analysis results, i.e., displays the analysis result screen 400 similar to that in the first embodiment on the display 38. In addition, the management device 60 stores the analysis results received from the dental analysis device 50 in the storage device 70, and more specifically, compiles them in an electronic medical record 70a.

[0078] According to the second embodiment, it is possible to provide, in the form of a so-called web service, an analysis that includes a process for estimating the stage of periodontal disease for each tooth from the panoramic X-ray image 100. This greatly contributes to improving the applicability and flexibility of the present invention.

[0079] [Third Example] Next, a third embodiment of the present invention will be described with reference to FIG.

[0080] The dental analysis system 10b according to the third embodiment is configured by a general-purpose PC 80, which performs the same processing as the dental analysis device 30 according to the first embodiment. That is, the PC 80 functions as a device performing the same processing as the dental analysis device 30 according to the first embodiment by installing a dental analysis application (application software) therein. The dental analysis application is downloaded from a predetermined web server or provided via an appropriate storage medium such as a CD or DVD.

[0081] According to the third embodiment, by providing a dental analysis application, a general-purpose PC 80 can function as a dental analysis device, which also greatly contributes to improving the applicability and flexibility of the present invention.

[0082] [Fourth Example] Next, a fourth embodiment of the present invention will be described with reference to FIGS.

[0083] The fourth embodiment is based on the first embodiment described above, and is particularly suitable for screening in mass medical examinations such as school health checkups. In other words, in mass medical examinations, there are many subjects, and it is necessary to select from among these many subjects those who need to undergo a detailed diagnosis (examination), and the fourth embodiment is extremely suitable for performing this efficiently.

[0084] For this reason, in the fourth embodiment, an analysis result table 500 as shown in Fig. 19 is provided, and is specifically stored in the storage unit 34 (see Fig. 1). The analysis result table 500 shown in Fig. 19 is an example of an analysis result table 500 for a school (university) health checkup.

[0085] This analysis result table 500 is linked to the electronic medical record 34a (see FIG. 1). The analysis table 500 stores various data, such as each student's student ID number, name, affiliation, number of severe cases, number of moderate cases, and evaluation score, using the management number (No.) for the analysis table 500 as a key. The number of severe cases refers to the total number of severe cases of periodontal disease, and the number of moderate cases refers to the total number of moderate cases of periodontal disease. The evaluation score is a score (score) that serves as a comprehensive indicator of the degree of progression of periodontal disease. For example, a tooth with mild (or healthy) periodontal disease is assigned a score of 1, a tooth with moderate periodontal disease is assigned a score of 2, and a tooth with severe periodontal disease is assigned a score of 3. The sum of these scores is the evaluation score. Therefore, the higher the evaluation score, the higher the overall progression of periodontal disease, and therefore the greater the need for a detailed diagnosis, or in other words, the higher the priority for having that detailed diagnosis. Note that the method for assigning evaluation points is not limited to this.

[0086] 20 is displayed on the display 38 (see FIG. 1) based on the contents of the analysis result table 500. Strictly speaking, when the operation unit 36 ​​(see FIG. 1) accepts an operation by the operator to display the diagnostic priority list screen 600 on the display 38, the diagnostic priority list screen 600 is displayed on the display 38. The control unit 32 is responsible for displaying the diagnostic priority list screen 600 on the display 38.

[0087] The diagnostic priority list screen 600 displays a list 602, located approximately in the center, showing students who require a detailed diagnosis. A student who requires a detailed diagnosis here refers to a student who satisfies one or all of the following conditions: a first condition that the severe score is equal to or greater than the threshold for the severe score; a second condition that the moderate score is equal to or greater than the threshold for the moderate score; and a third condition that the evaluation score is equal to or greater than the threshold for the evaluation score. The list 602 displays information about the student who requires a detailed diagnosis, such as their student ID number, name, affiliation, severe score, moderate score, and evaluation score, using the management number (No.) for the list 602 as a key. Additionally, a drop-down list 604 is located at an appropriate position on the diagnostic priority list screen 600, for example, near the upper right corner.

[0088] The drop-down list 604 is an operator that accepts an operation to select the order (sort order) of the information to be displayed in the list 602. By operating the drop-down list 604, it is possible to arbitrarily select any of the order of severe severity, moderate severity, and evaluation score as the order of the information to be displayed in the list 602. Note that Fig. 20 shows an example of the state in which the evaluation score order has been selected.

[0089] The contents of this list 602 are also stored in the storage unit 34. The contents of the list 602 can be printed by a printer (not shown), or can be sent to an external device such as a PC (different from the one shown in FIG. 18).

[0090] As described above, the fourth embodiment allows for efficient selection of students who need to undergo a detailed diagnosis, making it extremely suitable for screening in group medical examinations such as school health checkups. Moreover, the first embodiment, on which the fourth embodiment is based, allows for estimation of the progression of periodontal disease without the need for a probing test, as described above, dramatically speeding up screening. This is extremely beneficial for group medical examinations involving a large number of subjects, and the more subjects there are, the more significant this becomes.

[0091] Although the fourth embodiment uses a school health checkup as an example, the present invention is not limited to this and can also be applied to health checkups for various organizations such as companies and other group health checkups. Furthermore, the fourth embodiment is based on the first embodiment, but is not limited to this and may be based on either the second or third embodiment.

[0092] [Other application examples] The above-described embodiments are specific examples of the present invention and do not limit the technical scope of the present invention. The present invention can also be applied to aspects other than these embodiments.

[0093] For example, in each embodiment, the panoramic X-ray image 100 obtained by the dental panoramic X-ray imaging device 20 is the subject of analysis, but this is not limited to this, and general dental X-ray images (dental X-ray images) and dental CT images may also be the subject of analysis.

[0094] Furthermore, in each embodiment, the progression level of periodontal disease is estimated based on the index value α calculated using the above-described formula 1, and the first threshold value β1 and the second threshold value β2 are determined as criteria for determining the progression level of periodontal disease in accordance with the new classification established at a workshop held in November 2017. However, this is not limited to this. Since the criteria for determining the progression level of periodontal disease vary depending on the country or region of the world, the first threshold value β1 and the second threshold value β2 may be determined accordingly, or a number other than two (i.e., one or three or more) threshold values ​​may be determined as the judgment criteria. In other words, the threshold values ​​as the judgment criteria, including the first threshold value β1 and the second threshold value β2, may be set (variable) as appropriate.

[0095] Furthermore, in each embodiment, the progression level of periodontal disease is estimated based on the index value α calculated using Equation 1 as described above, but this is not limiting. For example, the progression level of periodontal disease may be estimated based on the ratio (Lb / La) between the root length La and the CAL correlation dimension Lb. In this case, since the larger the value of the ratio (Lb / La), the higher the progression level of periodontal disease is estimated to be, it is necessary to set a threshold value as an appropriate judgment criterion according to the ratio (Lb / La). Furthermore, the progression level of periodontal disease may be estimated based on the reciprocal (La / Lb) of the ratio (Lb / La) referred to here. In this case, it is also important to set a threshold value appropriate for the reciprocal (La / Lb).

[0096] In addition, if the actual value (actual dimension) of the CAL correlation dimension Lb is known, the progression level of periodontal disease may be estimated based on the actual value of the CAL correlation dimension Lb. For example, if the dimensions of any tooth are actually measured, the actual value of the CAL correlation dimension Lb can be determined based on the measurement. Furthermore, the actual dimensions of the subject's teeth can be estimated from several parameters including the subject's height and weight, and the actual value of the CAL correlation dimension Lb can also be determined based on the estimated values. The progression level of periodontal disease may then be estimated based on the actual value of the CAL correlation dimension Lb determined in this manner.

[0097] In each embodiment, the analysis was performed using three machine learning models, namely, the individual tooth detection model 322, the CEJ line detection model 324, and the alveolar bone line detection model 326. However, the analysis may be performed using a number of machine learning models other than three. This is not limited to this. That is, two or more of the detection of individual teeth (baselines 120), the detection of the CEJ line 140, and the detection of the alveolar bone line 150 may be achieved by a single machine learning model. Furthermore, instead of a supervised learning model, a model such as unsupervised learning, semi-supervised learning, or self-supervised learning may be adopted as the machine learning model.

[0098] Furthermore, the present invention can be provided in the form of an apparatus such as the dental analysis apparatus 30 in the first embodiment or the dental analysis apparatus 50 in the second embodiment, or in the form of a method called a dental analysis method, or in the form of a program called a dental analysis program such as the dental analysis app in the third embodiment.

[0099] Furthermore, the present invention can also be provided in the form of a non-transitory computer-readable storage medium storing the dental analysis program. The storage medium referred to here includes disk-type media such as CDs and DVDs, and semiconductor-type media such as USB memory and SD memory cards. Furthermore, instead of portable media, device-embedded (built-in) media such as ROMs and hard disk drives can also be used as the storage medium referred to here. [Explanation of symbols]

[0100] 10...Network System 20…Multifunction device 30...PC 50...Cloud Server 202 ... Control section 202a...CPU 202b... Main memory section 208 ... Image processing section 210 ... Image forming unit

Claims

1. an estimation means for estimating the progression of periodontal disease for each tooth from an image of the tooth included in a dental X-ray image using a machine learning model; and A dental analysis device comprising: information output means for outputting analysis result information for each tooth including the estimation result by the estimation means.

2. the machine learning model detects the individual tooth, the cementoenamel junction of the individual tooth, and the alveolar crest of the individual tooth from the image of the individual tooth; 2. The dental analysis device of claim 1, wherein the estimation means estimates the progression of periodontal disease for each tooth based on a relative ratio between a first distance, which is the distance between the root apex of the individual tooth detected by the machine learning model and the cement-enamel junction of the individual tooth, and a second distance, which is the distance between the cement-enamel junction of the individual tooth detected by the trained model and the alveolar bone crest of the individual tooth.

3. The machine learning model is a first model for detecting individual teeth from the images of the individual teeth; a second model for detecting the cement-enamel junction of each tooth from the image of the individual tooth; and The dental analysis device of claim 2 , further comprising a third model for detecting the alveolar bone crest of each tooth from the image of the individual tooth.

4. The dental analysis device of claim 1 , wherein the machine learning model is constructed by supervised learning based on a plurality of the dental radiograph images.

5. The dental analysis device according to claim 1 , wherein the information output means outputs the analysis result information by displaying the analysis result information for each tooth.

6. The dental analysis device according to claim 5 , wherein said information output means displays the analysis result information for each tooth in a list.

7. The dental analysis device according to claim 5 , wherein the information output means displays the estimation result by the estimation means by at least one of a character string and a mark.

8. an estimation step of estimating the progression of periodontal disease for each tooth from an image of the tooth included in the dental x-ray image using a machine learning model; and A dental analysis method comprising an information output step of outputting analysis result information for each tooth including the estimation result obtained by the estimation step.

9. an estimation procedure for estimating the progression of periodontal disease for each tooth from an image of the tooth included in a dental X-ray image using a machine learning model; and A dental analysis program that causes a computer to execute an information output procedure that outputs analysis result information for each tooth including an estimation result obtained by the estimation procedure.

Images