Learning methods, identification methods, and implant identification systems for implants

By using X-ray simulators and artificial neural network technology, 3D implant images are converted into 2D implant images and a learning model is generated. This solves the problems of difficulty in identifying implant model information and inaccurate measurement of bone loss, and achieves more efficient implant information identification and abnormality detection.

CN123124935APending Publication Date: 2026-11-17OSSTEMIMPLANT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202580025171.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2024-09-26
Filing Date
2025-05-30
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, implant model information is difficult to maintain access over a long period, and there is a lack of automated tools for measuring the degree of surrounding bone loss using radiographic images, resulting in low accuracy and consistency.

Method used

3D implant images are converted into 2D implant images using an X-ray simulator, and image effects are applied to generate an implant learning model. The model is then trained using an artificial neural network to determine implant type information and abnormalities.

Benefits of technology

It improves the accuracy of implant model information and the reliability of anomaly detection, reduces dependence on imaging angle and image quality, and enables more accurate measurement of peri-implant bone loss.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123124935A_ABST
    Figure CN123124935A_ABST
Patent Text Reader

Abstract

This invention discloses an implant learning method, an implant determination method, and an implant determination system. The invention applies an X-ray simulator to 3D implant images to convert them into 2D implant images, applies various image effects to enhance the 2D implant images, and generates various implant learning models. Furthermore, this invention uses implant learning models to determine the implanted implant type information from actual patient images.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to image data processing technology, and more specifically, to technology for analyzing and processing image data. Background Technology

[0002] Since the introduction of dental implant procedures, various implant models have been developed. Dental prostheses include the crown that serves as a tooth and the implant itself, which is embedded in the alveolar bone below the gum line and acts as the root. Additionally, an abutment supporting the crown and fixing screws for coupling the abutment can be added. In other words, because the implant is embedded in the alveolar bone, dentists cannot directly visually inspect it; visual identification via X-ray images is required.

[0003] Implant components are not yet standardized, and different manufacturers are developing implants with varying specifications. Therefore, compatibility between implant models is low. For long-term maintenance, it is necessary to retain or maintain access to information about implanted implants. Patients often struggle to remember the implant model information, and there are also numerous cases of hospitals closing down or losing relevant medical records. Consequently, patients who have previously had implants often find it difficult to identify their implant model, leading to difficulties in maintenance and repair. While such problems can be addressed through clinician experience or trial and error, this requires a high level of proficiency, thus limiting the ability of less experienced clinicians to correctly identify implants.

[0004] Furthermore, implants are designed for long-term use, and regular monitoring is necessary to ensure comfortable use over extended periods. Regular monitoring of implants utilizes radiographic imaging to determine the extent of bone loss around the implanted bone. However, due to the lack of automated tools for measuring the degree of surrounding bone loss from radiographic images, researchers or clinicians manually measure the degree of bone loss using a graphical program with a dedicated radiographic image viewer after radiographic imaging.

[0005] In measurements relying on manual operation, the following problems arise: The difference in scale between the actual object imaged and the radiographic image during radiographic imaging leads to measurement errors, reducing the accuracy of peripheral bone loss. Furthermore, errors caused by the measurer's visual perception and operational skills make it difficult to consider the measurement of objectively accurate peripheral bone loss. In other words, it depends on the operator's skill level, resulting in low reproducibility and consistency, making accurate measurement of peripheral bone loss difficult. Summary of the Invention Technical issues

[0006] According to the embodiments, the present invention proposes an implant learning method and an implant determination system, which can convert 3D implant images into 2D images by applying an X-ray simulator to 3D implant images, enhance 2D implants by applying various image effects to 2D implants, and generate implant learning models.

[0007] In addition, the present invention proposes an implant identification method and system that can more accurately determine implant type information and identify implant abnormalities such as peri-implantitis by using an implant learning model that includes enhanced implant images, regardless of the patient's imaging angle and image quality. Technical solutions

[0008] The implant learning method according to the embodiments includes: converting a 3D implant image into a 2D implant image by applying an X-ray simulator, wherein the X-ray simulator changes the angle or position at which X-rays are irradiated onto the 3D implant image; generating an enhanced image by applying an image effect to the 2D implant image; and training an artificial neural network by using at least one of the 2D implant image and the enhanced image to generate an implant learning model for determining implant type information.

[0009] Image effects may include at least one of the following: light source vector, direction vector, pixel size, blur, gamma effect, brightness effect, contrast effect, rotation effect, and cropping effect.

[0010] Implant learning methods may also include generating training images by changing at least one of the imaging angle and imaging position relative to a virtual phantom.

[0011] Implant learning methods may also include: segmenting edentulous regions in training images; and replacing the numbered regions corresponding to the segmented edentulous regions with 2D implant images generated by applying an X-ray simulator to 3D implant images.

[0012] The implant learning method may further include training the generated implant learning model, and the training may include: segmenting multiple enhanced 2D implant images; extracting multiple implant feature vectors from the segmented images of the multiple 2D implant images using a feature extraction model; clustering the multiple extracted implant feature vectors according to implant type; classifying each of the multiple clustered implant types into multiple subtypes by subdivision; and performing implant matching using the classified implant feature vectors.

[0013] An implant determination method includes: acquiring an implant learning model generated by training an artificial neural network using a 2D implant image converted by applying an X-ray simulator to a 3D implant image; inputting an actual patient image into the implant learning model; and determining implant model information from the actual patient image using the implant learning model and outputting the determined implant model information.

[0014] Implant model information may include at least one of the following: the implant manufacturer, the manufacturer's system, the length and diameter of the implant, and the number and pitch of the implant threads.

[0015] The output of implant model information may include: matching the diameter, length, and rotation information between the implant segmented in the actual patient image and the implant in the 2D implant image; extracting feature points from both the implant segmented in the actual patient image and the implant in the 2D implant image; matching the shape of the external threads of the two implants using the extracted feature points; and matching the internal shape of the two implants when the external thread shapes are the same.

[0016] In the output of implant model information, the length and diameter of the implant in the actual patient image can be calculated using the scaling ratio between the implant in the actual patient image and the implant in the 2D implant image.

[0017] Implant identification methods may also include: determining implant abnormalities based on whether the bone level and the implant are in contact with each other in actual patient images.

[0018] Determining implant abnormalities may include: determining that the implant is normal when the bone level is not in contact with the implant in the actual patient image; and determining that the implant is abnormal and calculating the bone loss height when the bone level is in contact with the implant and deviates from it by at least a predetermined angle.

[0019] Implant identification methods may also include: when, in an actual patient image, the contact area between the bone level and the implant is located below the area where the abutment is coupled, using tools recommended by the implant manufacturer and system for removing the implant.

[0020] Implant identification methods may also include recommending antibiotic or surgical treatment based on the degree of bone loss when, in actual patient images, the contact area between the bone level and the implant is located above the area where the abutment is coupled relative to the uppermost part of the implant.

[0021] The implant identification method may also include: operating in conjunction with a patient management program on a display device to provide implant model information to the patient management program.

[0022] According to another embodiment, an implant determination system includes: a learning device configured to: convert a 3D implant image into a 2D implant image by applying an X-ray simulator, wherein the X-ray simulator changes the angle or position at which X-rays are irradiated onto the 3D implant image; generate an enhanced image by applying an image effect to the converted 2D implant image; and train an artificial neural network using at least one of the 2D implant image and the enhanced image to generate an implant learning model for determining implant type information; a determination device configured to: input an actual patient image into the trained implant learning model, determine implant type information of the implanted implant from the actual patient image using the implant learning model, and output the implant type information; and a display device configured to: receive the implant type information from the determination device and display the implant type information. Beneficial effects

[0023] This invention can convert 3D implant images into 2D implant images by applying an X-ray simulator, wherein the X-ray simulator changes the angle or position of the X-ray irradiation on the 3D implant image, enhances the 2D implant image by applying various image effects, and generates an implant learning model for recognizing various implants.

[0024] This invention can more accurately determine implant model information and identify implant abnormalities such as peri-implantitis by using an implant learning model that includes enhanced implant images, regardless of the patient's imaging angle and image quality. Attached Figure Description

[0025] Figure 1 This is a diagram illustrating the configuration of an implant determination system according to an embodiment of the present invention;

[0026] Figure 2 This is a flowchart illustrating an implant learning method according to an embodiment of the present invention;

[0027] Figure 3 This is a flowchart illustrating an implant determination method according to an embodiment of the present invention;

[0028] Figure 4 This is a diagram illustrating the generation of a 2D implant image by applying an X-ray simulator to a 3D implant image according to an embodiment of the present invention.

[0029] Figure 5 This is a diagram illustrating the generation of enhanced images through various image processing operations according to an embodiment of the present invention;

[0030] Figure 6This is a diagram showing the segmentation of an anatomical structure according to an embodiment of the present invention;

[0031] Figure 7 This is a diagram showing the assignment of numbered regions to teeth and implants in a training image according to an embodiment of the present invention;

[0032] Figure 8 and Figure 9 This is a diagram illustrating the generation of training images according to various embodiments of the present invention;

[0033] Figure 10 This is a diagram showing a scene where a edentulous area in a training image is replaced with a 2D implant image according to an embodiment of the present invention;

[0034] Figure 11 This is a diagram illustrating 2D implant images used for implant matching during implant learning model training according to an embodiment of the present invention.

[0035] Figure 12 This is a diagram illustrating the clustering of multiple implant feature vectors by implant type during implant learning model training according to an embodiment of the present invention.

[0036] Figure 13 This is a diagram illustrating how, during implant learning model training according to an embodiment of the present invention, each of multiple clustered implant types is classified into multiple subtypes through subdivision.

[0037] Figure 14 This is a diagram illustrating the implant feature vectors during implant learning model training according to an embodiment of the present invention;

[0038] Figure 15 This is a diagram showing a screen where, according to an embodiment of the present invention, the determined implant model information is matched with an actual patient image.

[0039] Figure 16 This is a diagram illustrating the determination of implant abnormalities according to an embodiment of the present invention; and

[0040] Figure 17 This is a diagram illustrating a screen that provides recommendation information based on whether the implant has abnormalities, according to an embodiment of the present invention. Detailed Implementation

[0041] The advantages and features of the present invention, as well as the ways in which these advantages and features are realized, will become apparent from the following detailed description of the embodiments taken in conjunction with the accompanying drawings. However, the invention can be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. These embodiments are provided to make this disclosure thorough and complete, and to fully convey the scope of the invention to those skilled in the art, and the invention is defined only by the scope of the appended claims. Throughout the specification, the same or similar reference numerals denote the same or similar elements.

[0042] In describing exemplary embodiments of the invention, detailed descriptions of related known configurations or functions included herein will be omitted where it is determined that such detailed descriptions may unnecessarily obscure the subject matter of the invention. The terminology described below is defined in consideration of the functionality within the invention and may vary depending on the user, the user's intent, or convention. Therefore, the definitions of the terminology used herein should follow the context disclosed herein.

[0043] In the following, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the exemplary embodiments can be modified in various different ways, and the scope of the present invention is not limited to the exemplary embodiments described below. Embodiments of the present invention are provided to assist those skilled in the art in explaining and understanding the invention.

[0044] Figure 1 This is a diagram illustrating the configuration of an implant determination system according to an embodiment of the present invention.

[0045] Reference Figure 1 The implant determination system 1 includes a learning device 11, a determination device 12, and a display device 13. The learning device 11 and the determination device 12 may be computing devices.

[0046] The learning device 11 uses training images to generate an implant learning model for determining implant type information. For example, the learning device 11 can convert a 3D implant image into a 2D implant image using an X-ray simulator, where the X-ray simulator changes the angle or position of the X-rays irradiating the 3D implant image, and can enhance the 2D implant image by applying various image effects, thereby generating the implant learning model. The following will refer to… Figure 2 Describe the implant learning method of learning device 11.

[0047] The determining device 12 acquires the implant learning model that has been trained by the learning device 11, and uses the acquired implant learning model to determine the implant type information from the actual patient image. Furthermore, it can determine whether the implant has any abnormalities. The following will refer to... Figure 3 Describe the implant determination method of the determination device 12.

[0048] When the learning device 11 generates an implant learning model by training with training images, the determining device 12 can acquire the implant learning model from the learning device 11. The implant learning model can be used simultaneously in the learning device 11 or the determining device 12, or it can be stored in a separate external device, such as an artificial intelligence (AI) device.

[0049] The learning device 11 and the determining device 12 may each include a processor and a database. The learning device 11 and the determining device 12 may be in the form of a server.

[0050] The processor executes the program. The processor can be, for example, a central processing unit (CPU), a processing device, a computing device, a processor, a microprocessor, a microcomputer, a digital signal processor (DSP), etc.

[0051] The database can store the information required for training and the implant learning model generated through training, and can provide information and implant learning models in response to user requests.

[0052] Display device 13 is an electronic device capable of executing patient management procedures. The electronic device may include a personal computer (PC), laptop computer, tablet computer, smartphone, etc. The patient management procedure may include a viewer program, image editing program, etc. The patient management procedure may also be incorporated into programs used for other general image processing.

[0053] Display device 13 displays information on the screen of the patient management program. For example, display device 13 can receive implant model information determined by the determining device 12 using the implant learning model generated through training by the learning device 11, and display the received implant model information.

[0054] Figure 2 This is a flowchart illustrating an implant learning method according to an embodiment of the present invention.

[0055] Reference Figure 1 and Figure 2 The learning device 11 converts a 3D implant image into a 2D implant image by applying an X-ray simulator, wherein the X-ray simulator changes the angle or position of the X-ray irradiation onto the 3D implant image (S210). (Refer to the following...) Figure 4 This describes an implementation method for converting 3D implant images into 2D implant images. Here, in addition to images obtained by imaging actual patients, 3D implant images may also include training images generated for training implant learning models.

[0056] Subsequently, the learning device 11 generates an enhanced image by applying image effects to the 2D implant image (S220). Image effects may include at least one of the following: light source vector, direction vector, pixel size, blur, gamma effect, brightness effect, contrast effect, rotation effect, and cropping effect. (Refer to below...) Figure 5 Describes an implementation method for generating enhanced images.

[0057] Subsequently, the learning device 11 trains an artificial neural network using at least one of a 2D implant image and an enhanced image, thereby generating an implant learning model (S230). The following will refer to... Figures 12 to 14 Describe the implementation method of generating implant learning models.

[0058] In addition, the learning device 11 can train the generated implant learning model (S240).

[0059] Training the implant learning model can include image segmentation, feature extraction, clustering, and implant matching. For example, in step S240 of training the implant learning model, the learning device 11 can segment multiple enhanced 2D implant images and extract multiple implant feature vectors from the segmented images using a feature extraction model. Furthermore, the learning device 11 can cluster the extracted implant feature vectors according to implant type and further subdivide each of the clustered implant types into multiple subtypes. Subsequently, the learning device 11 can perform implant matching using the classified implant feature vectors.

[0060] Figure 3 This is a flowchart illustrating an implant determination method according to an embodiment of the present invention.

[0061] Reference Figure 1 and Figure 3 The determining device 12 acquires an implant learning model generated by training an artificial neural network using a 2D implant image converted by applying an X-ray simulator to a 3D implant image (S310).

[0062] Subsequently, the determining device 12 inputs the actual patient image into the implant learning model that has been trained (S320).

[0063] Subsequently, the determining device 12 determines the implant model information from the actual patient image using the implant learning model and outputs the determined implant model information (S330). The implant model information may include at least one of the following: the implant manufacturer, the manufacturer's system, the length and diameter of the implant, and the number and pitch of the implant threads.

[0064] In step S330, when outputting implant model information, the determining device 12 can use the scaling ratio between the implant in the actual patient image and the implant in the 2D implant image to calculate the length and diameter of the implant in the actual patient image.

[0065] Subsequently, the determining device 12 can determine implant abnormalities based on whether the bone level and the implant are in contact with each other in the actual patient image (S340). In this case, when the bone level and the implant are not in contact with each other in the actual patient image, the determining device 12 can determine that the implant is normal. When the bone level and the implant are in contact with each other and deviate at least a predetermined angle, the determining device 12 can determine that the implant is abnormal and calculate the bone loss height. (Refer to the following...) Figure 16 and Figure 17 Describes the implementation methods for identifying implant abnormalities.

[0066] Furthermore, when in actual patient images, the contact area between the bone level and the implant is located below the area where the abutment is coupled relative to the uppermost part of the implant, the determining device 12 can recommend tools for removing the implant based on the implant manufacturer and system.

[0067] Furthermore, when in actual patient images, the contact area between the bone level and the implant is located above the area where the abutment is coupled relative to the uppermost part of the implant, the determining device 12 can recommend antibiotic treatment or surgical treatment based on the height of bone loss.

[0068] In addition, the determining device 12 can be operated in conjunction with the patient management program of the display device 13 to provide implant model information to the patient management program.

[0069] Figure 4 This is a diagram illustrating the generation of a 2D implant image by applying an X-ray simulator to a 3D implant image according to an embodiment of the present invention.

[0070] Reference Figure 1 and Figure 4 The learning device 11 can generate a 2D implant image 420 by applying an X-ray simulator to a 3D implant image 410, wherein the X-ray simulator changes the angle or position of the X-ray irradiation. The X-ray simulator is a tool that virtually irradiates the 3D implant image 410 with X-rays 430 to generate the 2D implant image 420, thereby allowing the shape of the implant or the shape of the abutment coupled to the implant to be observed in the 2D implant image 420.

[0071] Figure 5 This is a diagram illustrating the generation of enhanced images through various image processing operations according to an embodiment of the present invention.

[0072] Reference Figure 1and Figure 5 The learning device 11 can learn by feeding the implant model from a 2D implant image, for example... Figure 5 The first or second model is used to apply image effects to generate an enhanced image. For example, the learning device 11 can apply light source vectors, direction vectors, pixel size adjustments, blur effects, gamma effects, brightness effects, contrast effects, rotation effects, or cropping effects. Each of the vectors and effects can be applied individually or in combination.

[0073] As an example of applying image effects, reference numerals 510 and 520 indicate an enhanced image obtained by applying the same pixel size, blur effect, brightness effect, and contrast effect, while applying different values ​​to the light source vector, detector vector, gamma effect, and rotation effect.

[0074] Therefore, by applying various image effects, a large amount of enhanced data can be generated, which makes it less susceptible to the influence of various imaging angles and image quality when acquiring panoramic images or intraoral sensor images.

[0075] Figure 6 This is a diagram showing the segmentation of an anatomical structure according to an embodiment of the present invention.

[0076] Reference Figure 1 and Figure 6 The learning device 11 acquires training images 60 and segments the anatomical structures in the acquired training images 60. Anatomical structures may include teeth, bones, implants, etc. Training images 60 may be 2D panoramic images, periapical radiographic images, etc.

[0077] The learning device 11 can use an implant learning model generated through AI learning to segment anatomical structures in the training image 60. For example, the learning device 11 can segment the implant 620, tooth 610, and bone level 630 in the training image 60. The learning device 11 can mask and display the segmented implant 620, tooth 610, and bone level 630 separately, or it can integrate the implant 620, tooth 610, and bone level 630, and then mask and display the integrated result.

[0078] AI learning can be performed through computations using artificial neural networks. Artificial neural networks can include convolutional neural networks (CNNs), recurrent neural networks (RNNs), deep belief networks, restricted Boltzmann machines, and so on.

[0079] Figure 7 This is a diagram showing the assignment of numbered regions to teeth and implants in a training image according to an embodiment of the present invention.

[0080] Reference Figure 1 and Figure 7Individual teeth and implants can be numbered in the training image 70 of the segmented teeth and implants, thereby generating numbered boxes 701 and 702. Numbered boxes 701 and 702 refer to the regions in the data represented as rectangular boxes enclosing the corresponding segmented individual teeth and implants in terms of their maximum horizontal and vertical dimensions.

[0081] Figure 8 and Figure 9 This is a diagram illustrating the generation of training images according to various embodiments of the present invention.

[0082] Reference Figure 1 , Figure 8 and Figure 9 In addition to the acquired training images, the learning device 11 can also segment anatomical structures, including teeth, bones and implants, from the generated training images after they have been virtually generated.

[0083] The learning device 11 can generate various training images by changing the imaging angle or imaging position relative to the virtual phantom. The training images can be dental images such as virtual panoramic data and periapical radiographic data.

[0084] For example, such as Figure 8 As shown, the learning device 11 can virtually generate image 810 in which the patient bites an occlusal block, wherein the occlusal block is positioned deeper in the oral cavity, or the virtual phantom is positioned forward, thereby providing a reduction effect on the anterior tooth region. Alternatively, the learning device 11 can virtually generate image 820 in which the patient bites an occlusal block, wherein the occlusal block is positioned closer to the front of the oral cavity, or the virtual phantom is positioned backward, thereby providing a magnification effect on the anterior tooth region.

[0085] As another example, such as Figure 9 As shown, the learning device 11 can generate a V-shaped image 910 by providing an imaging effect in which the virtual phantom is tilted downwards. Alternatively, the learning device 11 can obtain a straight line image 920 by providing an imaging effect in which the virtual phantom is tilted upwards.

[0086] As yet another example, the learning device 11 can obtain an image tilted to the left or right by providing the effect of imaging in a way that the virtual phantom is tilted to one side.

[0087] In addition, the learning device 11 can generate various training images by changing settings such as tube voltage (kVp) and tube current (mA) relative to the virtual phantom.

[0088] In addition to the acquired training images, the learning device 11 can further enhance the data by adjusting the image effects applied to the training images. In this case, image effects may include brightness (windowing level), contrast (windowing width), inverting black and white (invert), sharpness (sharpening), etc.

[0089] Figure 10 This is a diagram showing a scene where a edentulous area in a training image is replaced with a 2D implant image according to an embodiment of the present invention.

[0090] Reference Figure 1 and Figure 10 The training image 111 may include a edentulous region. The learning device 11 may segment the edentulous region in the training image 111 and then replace the edentulous region with a 2D implant image obtained by applying an X-ray simulator to the 3D implant image.

[0091] According to the method of replacing the edentulous area with a 2D implant image, the learning device 11 assigns a numbered region 1110 to the edentulous area of ​​the training image 111. Subsequently, the learning device 11 replaces the assigned numbered region 1110 with the corresponding 2D implant image 1120 to generate a training image 112 with the 2D implant image 1120 added. For example, as... Figure 10 As shown, the learning device 11 can form a 2D implant image 1120 in “.png” format into a rectangular model with the same horizontal and vertical dimensions as the numbered region 1110 corresponding to the edentulous area in the training image 111, and then add the 2D implant image 1120 to the numbered region 1110.

[0092] The learning device 11 can perform training by changing orientation vectors, rotation vectors, etc., using virtual training images generated from a virtual phantom and 2D implant images at various angles or positions. Furthermore, since the learning model can be generated during training with the correct implant identity known, it is not necessary to determine implant type information from training images of actual patients with implants. Therefore, the time required for such determination can be eliminated, and accuracy can be improved.

[0093] Figure 11 This is a diagram illustrating 2D implant images used for implant matching during implant learning model training according to an embodiment of the present invention.

[0094] Reference Figure 1 and Figure 11The learning device 11 performs implant matching using feature points. For example, the learning device 11 can match two implants using feature points from both the implant in the training image generated by adding a 2D implant image to a virtual phantom image and the implant in the 2D implant image.

[0095] The learning device 11 can segment 2D implant images into frames with preset shapes. For example, the learning device 11 can distinguish between image 1210, which only includes the implant, and images 1220 and 1230, which are coupled to the abutment to the implant, and can segment each image into a rectangular frame according to the size of the corresponding image.

[0096] Figure 12 This is a diagram illustrating the clustering of multiple implant feature vectors by implant type during implant learning model training according to an embodiment of the present invention.

[0097] Reference Figure 1 and Figure 12 The learning device 11 generates a feature extraction model by performing contrastive learning on multiple enhanced 2D implant images. The feature extraction model can be an artificial neural network-based model used to extract implant feature vectors from segmented images of multiple enhanced 2D implant images. "Contrastive learning" refers to training such that the distance between feature vectors for implant images of the same type decreases (i.e., the feature vectors become more similar), while the distance between feature vectors for implant images of different types increases.

[0098] Learning device 11 can cluster multiple implant feature vectors extracted using a trained feature extraction model according to implant type (see [link]). Figure 12 Furthermore, the learning device 11 can further subdivide each of the multiple clusters of implant types into multiple subtypes by further subdividing the corresponding clusters of implant types (see [link]). Figure 13 ).

[0099] More specifically, once the feature extraction model is generated, implant feature vectors of implant images of the same type can have similar distributions, while implant feature vectors of implant images of different types can have different distributions.

[0100] For example, such as Figure 12As shown, the learning device 11 can extract the implant feature vector of each of the multiple segmented implant images a, b, c, d, e, and f through a feature extraction model. The implant feature vectors corresponding to the first implant image (a: TS3), the second implant image (b: TS4), the third implant image (c: KS3), the fourth implant image (d: SS3), the fifth implant image (e: US2), and the sixth implant image (f: US3), respectively, can be distributed in the feature space 1300.

[0101] Subsequently, the learning device 11 can cluster the first implant image (a: TS3) and the second implant image (b: TS4), which are implants of the same type, into a first group; cluster the fifth implant image (e: US2) and the sixth implant image (f: US3) into a fourth group; cluster the fourth implant image (d: SS3), which belong to different implant types and whose feature vector distributions are close to each other, into a third group; and cluster the third implant image (c: KS3) into a second group. That is, training is performed so that the feature vectors are distributed differently according to the implant type (e.g., implant types TS and KS are classified as the same type, while SS and US are classified as different types). Multiple implant feature vectors can be clustered by type according to distance (e.g., using K-Means algorithm) or according to density (e.g., using DBSCAN).

[0102] Figure 13 This is a diagram illustrating how, during implant learning model training according to an embodiment of the present invention, each of multiple clustered implant types is classified into multiple subtypes through subdivision.

[0103] Reference Figure 1 and Figure 13 Since even within the same type, there are distinct patterns, the learning device 11 can classify and cluster the implant feature vector input into subtypes with similar patterns. For example, such as Figure 13 As shown, even within the first group, the learning device 11 can also classify implant types TS3(a) and TS4(b).

[0104] The learning device 11 can cluster images with similar patterns using an image clustering engine, or classify images by extracting image features using an image classification engine that includes an image feature extraction network. In this case, at least one of the image clustering engine and the image classification engine that includes an image feature extraction network can be configured at least partially as an artificial neural network trained according to a machine learning algorithm.

[0105] Figure 14This is a diagram illustrating the implant feature vectors during implant learning model training according to an embodiment of the present invention.

[0106] Reference Figure 1 and Figure 14 The learning device 11 can extract the implant feature vector of each of the multiple segmented implant images. For example... Figure 14 As shown, the implant feature vector can correspond to the implant's external thread shape 1510, the implant's external thread spacing (number of external threads) 1520, the implant's external vertex shape 1530, the distance from the top of the implant (top) to the implant's internal thread 1540, the distance from the top of the implant to the area coupled to the abutment 1550, the implant's internal thread diameter 1560, the implant's internal angle 1570, the implant's external angle 1580, etc.

[0107] Figure 15 This is a diagram showing a screen where, according to an embodiment of the present invention, the determined implant model information is matched with an actual patient image.

[0108] Reference Figure 1 and Figure 15 The determining device 12 can use an implant learning model that clusters the extracted feature vectors through virtual data augmentation and contrastive learning to determine implant model information from actual patient images. Implant model information may include the implant manufacturer and system, as well as the implant's length and diameter.

[0109] For example, the determining device 12 can input an actual patient image 1600 into the implant learning model to obtain and output implant model information. More specifically, the implant learning model can segment the various anatomical structures in the actual patient image 1600, including teeth, bone, and implants 1610, determine implant model information including the manufacturer and system of the segmented implants 1610, and output the determined implant model information.

[0110] Subsequently, as Figure 15 As shown, the determining device 12 matches the implants in the actual patient image 1600 with the implants in the 2D implant image 1620. For example, the determining device 12 can set the implant 1610 segmented from the actual patient image 1600 as a region of interest, and match the two implants between the set region of interest and the 2D implant image 1620 corresponding to the implant model information output by the implant learning model. During the matching, the determining device 12 can match the implant 1610 segmented from the actual patient image 1600 and the implants in the 2D implant image 1620 for diameter, length, rotation information, etc.

[0111] Subsequently, the determining device 12 extracts features from both the implant 1610 segmented from the actual patient image 1600 and the implant in the 2D implant image 1620, compares the extracted features, and searches for corresponding feature point pairs. Then, using the found feature point pairs, the determining device 12 can first match the shape (thread spacing and protrusion) of the external threads of the two implants, and when the external thread shapes are the same, it can further match the internal shapes of the two implants.

[0112] Subsequently, the determining device 12 can calculate the accurate length and diameter of the implant 1610 using a scaling factor 1630 between the implant 1610 segmented from the actual patient image 1600 and the implant in the 2D implant image 1620. For example, the determining device 12 can calculate the accurate length and diameter of the implant 1610 by applying the scaling factor of the actual patient image to the 2D implant image and matching the two implants based on the applied scaling factor.

[0113] When a 3D implant is imaged as a 2D panoramic image, the implant may not be positioned with its root tip parallel to the y-axis, but rather tilted towards the z-axis. In contrast, according to the present invention, since the 2D implant image 1620 generated by applying an X-ray simulator to the 3D implant image at various angles has been used for training, the implant can be accurately matched in two dimensions, and the accurate length and diameter of the implant can be determined using the scaling factor 1630.

[0114] The determining device 12 can calculate the distance from the top of the implant to the area where the abutment is coupled, based on matching information and information about the implant's manufacturer, system, diameter, and length.

[0115] Figure 16 This is a diagram illustrating the determination of implant abnormalities according to an embodiment of the present invention.

[0116] Reference Figure 1 and Figure 16 After determining the implant model information from actual patient images, the determining device 12 can use the segmented bone level to determine abnormalities of the implant and surrounding teeth.

[0117] The determining device 12 can segment the bone and implant 1610 from the actual patient image 1600, and then identify the contact area between the bone level 1710 and the implant 1610. The bone level 1710 is expected to have a gentle curve relative to the implant 1610. When the bone level 1710 deviates by at least a predetermined angle, the determining device 12 can determine that bone loss exists relative to the implant 1610 and calculate the contact area between the implant 1610 and the bone level 1710. In the case of consecutive implants, for example, when teeth, implant 1, implant 2, implant 3, and teeth are arranged in this order, if the bone level deviates by at least a predetermined angle at implant 1, remains straight at implant 2, and deviates again by at least a predetermined angle at implant 3, then the determining device 12 can determine that bone loss exists relative to implant 1 and implant 2.

[0118] As shown in (a), when the bone level 1710 and the implant 1610 are not in contact with each other in the actual patient image 1600, the determining device 12 determines that the implant 1610 is normal and does not calculate the bone loss height. Conversely, as shown in (b), when the bone level 1710 and the implant 1610 are in contact with each other and the bone level 1710 is deviated at at least a predetermined angle, the determining device 12 determines that the implant 1610 is abnormal, with the possibility of peri-implantitis, and calculates the bone loss height.

[0119] Figure 17 This is a diagram illustrating a screen that provides recommendation information based on whether the implant has abnormalities, according to an embodiment of the present invention.

[0120] Reference Figure 1 , Figure 16 and Figure 17 As shown in (a), when, in an actual patient image, the contact area between the bone level 1710 and the implant 1610 is located below the abutment-coupled area 1820 relative to the uppermost end 1810 of the implant, the determining device 12 can recommend tools for removing the implant 1610 based on the implant manufacturer and system specifications. This is because there is a risk of implant breakage when the bone level 1710 is located below the abutment-coupled area 1820 relative to the uppermost end 1810 of the implant.

[0121] When recommending an implant removal tool, the determining device 12 can recommend an implant removal tool based on information values ​​indicating the height from the top of the implant to the area where the abutment is coupled.

[0122] When, as shown in (b), the bone level 1710 is located above the abutment-coupled region 1820 relative to the uppermost end 1810 of the implant, the determining device 12 calculates the bone loss height based on scaling information and detected implant length information. Subsequently, when the calculated bone loss height is less than or equal to a predetermined height, for example, 2 mm, the determining device 12 may recommend antibiotic treatment.

[0123] Conversely, when the bone loss height is greater than or equal to the predetermined height and the bone level 1710 is above the area 1820 where the abutment is coupled relative to the uppermost end 1810 of the implant, the determining device 12 may recommend surgical treatment (such as smoothing the exposed surface of the implant and performing bone grafting).

[0124] Furthermore, the determining device 12 can be operated in conjunction with the patient management program of the display device 13 to provide the patient management program of the display device 13 with information about the tools used to remove the implant, bone loss height information, and implant model information (manufacturer, system, length, and diameter).

[0125] Embodiments of the present invention have been described above. Those skilled in the art will understand that the invention can be implemented in modified forms without departing from its essential characteristics. Therefore, the disclosed embodiments should be considered descriptive rather than restrictive. The scope of the invention is indicated by the claims rather than the foregoing description, and all differences within the equivalent scope should be interpreted as included within the invention.

Claims

1. An implant learning method executed by a computing device, comprising: A 3D implant image is converted into a 2D implant image by applying an X-ray simulator, wherein the X-ray simulator changes the angle or position at which X-rays are irradiated onto the 3D implant image; An enhanced image is generated by applying image effects to the 2D implant image; as well as An artificial neural network is trained using at least one of the 2D implant image and the enhanced image to generate an implant learning model for determining implant type information.

2. The implant learning method according to claim 1, wherein, The image effects include at least one of the following: light source vector, direction vector, pixel size, blur, gamma effect, brightness effect, contrast effect, rotation effect, and cropping effect.

3. The implant learning method according to claim 1 further includes: Training images are generated by changing at least one of the imaging angle and imaging position relative to the virtual phantom.

4. The implant learning method according to claim 3 further includes: Segment the missing tooth region in the training image; as well as The numbered regions corresponding to the segmented edentulous areas are replaced with the 2D implant images generated by applying the X-ray simulator to the 3D implant images.

5. The implant learning method according to claim 1, further comprising: The generated implant learning model is trained. The training includes: Segmenting multiple enhanced 2D implant images; Multiple implant feature vectors are extracted from the segmented images of the multiple 2D implant images using a feature extraction model; Cluster the extracted implant feature vectors according to implant type; Each of the multiple clustered implant types is further subdivided into multiple subtypes through detailed classification; and Implant matching is performed using the classified implant feature vectors.

6. A method for determining implants executed by a computing device, comprising: An implant learning model is obtained, which is generated by training an artificial neural network using 2D implant images transformed by applying an X-ray simulator to 3D implant images; The actual patient image is input into the implant learning model that has been trained. as well as The implant learning model determines the implant model information from the actual patient image and outputs the determined implant model information.

7. The implant determination method according to claim 6, wherein, The implant model information includes at least one of the following: the implant manufacturer, the manufacturer's system, the length and diameter of the implant, and the number and pitch of the threads of the implant.

8. The implant determination method according to claim 7, wherein, The output of the implant model information includes: The diameter, length, and rotation information of the implant segmented in the actual patient image are matched with the implant in the 2D implant image; Extract feature points from both the implant segmented from the actual patient image and the implant in the 2D implant image; The extracted feature points were used to match the shape of the external threads of the two implants; and When the external threads of the two implants are the same, the internal shapes of the two implants are matched.

9. The implant determination method according to claim 7, wherein, The output of the implant model information includes: calculating the length and diameter of the implant in the actual patient image using the scaling ratio between the implant in the actual patient image and the implant in the 2D implant image.

10. The implant determination method according to claim 7, further comprising: An abnormality of the implant is determined based on whether the bone level and the implant are in contact with each other in the actual patient image.

11. The implant determination method according to claim 10, wherein, The determination of abnormalities in the implant includes: When the bone level in the actual patient image is not in contact with the implant, the implant is determined to be normal; and When the bone level contacts the implant and deviates from it at at least a predetermined angle, the implant abnormality is determined and the bone loss height is calculated.

12. The implant determination method according to claim 10, further comprising: When, in the actual patient image, the contact area between the bone level and the implant is located below the area where the abutment is coupled relative to the uppermost part of the implant, the tool recommended by the implant manufacturer and system for removing the implant is used.

13. The implant determination method according to claim 10, further comprising: When, in the actual patient image, the contact area between the bone level and the implant is located above the area where the abutment is coupled relative to the uppermost part of the implant, antibiotic treatment or surgical treatment is recommended based on the degree of bone loss.

14. The implant determination method according to claim 7, further comprising: It operates in conjunction with the patient management program of the display device to provide the implant model information to the patient management program.

15. An implant identification system, comprising: A learning device configured to: convert a 3D implant image, which serves as a training image and represents an actual implanted implant, into a 2D implant image by applying an X-ray simulator, wherein the X-ray simulator changes the angle or position at which X-rays are irradiated onto the 3D implant image; generate an enhanced image by applying image effects to the converted 2D implant image; and train an artificial neural network using at least one of the 2D implant image and the enhanced image to generate an implant learning model for determining implant type information; A determining device is configured to: input an actual patient image into the implant learning model that has been trained, determine the implant model information of the implanted implant from the actual patient image through the implant learning model, and output the implant model information; as well as A display device configured to receive and display the implant model information from the determining device.