A three-dimensional intelligent jaw bone resorption rate dynamic change monitoring method based on CBCT

Abstract

The application relates to the field of artificial intelligence medical image processing technology and discloses a three-dimensional intelligent jaw bone absorption rate dynamic change monitoring method based on CBCT. CBCT data is used for three-dimensional image analysis, a deep learning model is used to automatically segment teeth, tooth enamel, maxillary bones and mandibular bones, the enamel-dentin junction and the top base of periodontal bone defects are automatically identified, the position of tooth root tips is automatically located, and the local bone absorption rate is calculated.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence medical image processing technology, and in particular to a three-dimensional intelligent method for monitoring dynamic changes in jawbone resorption rate based on CBCT. Background Technology

[0002] Periodontal disease refers to diseases occurring in the supporting tissues of the teeth (periodontal tissues), including two main categories: gingivitis of the gingival tissue and periodontitis affecting deeper periodontal tissues. Severe periodontitis has become the sixth leading cause of chronic non-communicable disease worldwide.

[0003] During the examination, specific identification, labeling, and comparison of periodontal tissues are required, a task primarily performed by computers. However, existing methods have several problems, such as:

[0004] The existing Chinese invention patent CN118864405A discloses an artificial intelligence-based method for detecting periodontitis. However, this method relies solely on two-dimensional panoramic images, which lack sufficient image resolution and depth information, making it difficult to accurately assess the three-dimensional resorption of periodontal bone. The annotation of key points in this method depends on the doctor's experience, and the subsequent bone resorption calculation is only a rough geometric measurement, lacking refined quantitative indicators. Furthermore, the annotation of key points in this method lacks a unified anatomical reference, which may lead to instability of key points and affect the reliability of subsequent analysis. This method is based on analysis of only one image and cannot support long-term follow-up and dynamic change analysis.

[0005] Existing Chinese invention patent CN117351293A discloses a joint learning method and device for classifying periodontal disease images. This scheme only outputs category labels for periodontal disease (such as "healthy / affected / mild to moderate severe"), lacking descriptions of quantitative indicators of the lesion area (such as bone resorption rate), resulting in coarse diagnostic granularity and weak interpretability. This scheme is based on traditional two-dimensional images (such as oral photographs or panoramic films), which are difficult to fully reflect the true degree of alveolar bone loss in the buccal / lingual / vertical directions. This proposal is based on CBCT images for three-dimensional reconstruction, segmenting teeth, enamel, and maxillae to achieve visualization and analysis of the complete spatial structure of periodontal tissues. This scheme is a single image analysis and does not have a spatial alignment mechanism between multi-time point data, making it impossible to assess the progression trend of periodontal disease. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a CBCT-based three-dimensional intelligent jawbone aspiration system.

[0007] The method for monitoring the dynamic changes in resorption rate uses CBCT data for three-dimensional image analysis and automatically segments teeth, enamel, maxilla and mandible through a deep learning model. It automatically identifies the cementoenamel junction and the base of the periodontal bone defect apex, and automatically locates the root apex position to calculate the local bone resorption rate.

[0008] To achieve the above objectives, this invention provides the following technical solution: a three-dimensional intelligent method for monitoring dynamic changes in jawbone resorption rate based on CBCT, comprising the following steps:

[0009] S1. Obtain an initial three-dimensional image of the periodontium, segment the initial three-dimensional image, and automatically extract the three-dimensional structures of teeth, enamel, maxilla, and mandible;

[0010] S2. Establish standardized anatomical reference points on the surface of each tooth;

[0011] S3. For each reference point, automatically identify the corresponding cementoenamel boundary and the base point of the periodontal bone defect apex;

[0012] S4. Based on the surface data points of a single tooth, the number and location of the root apex of the current tooth are obtained through principal component analysis, clustering, and projection methods.

[0013] S5. Based on the cementoenamel boundary and the base of the periodontal bone defect apex at each reference point, as well as the nearest root apex, calculate the jawbone resorption rate at that reference point.

[0014] S6. For CBCT images of the same patient acquired at different time points, spatial alignment of the three-dimensional tooth model is achieved through point cloud registration, and historical reference points are mapped to dynamically monitor the trend of jawbone resorption.

[0015] Furthermore, the segmentation method in S1 is as follows:

[0016] S11. Perform four-category segmentation at low resolution to locate the upper dentition, lower dentition, maxilla, and mandible regions.

[0017] S12. Based on the segmentation results, the upper and lower dentition regions are cropped out. The centroid localization, tooth classification and tooth segmentation of the teeth are performed at low resolution. The position of each tooth is determined according to the tooth segmentation results at low resolution.

[0018] S13. Cropping each tooth region at high resolution for individual tooth and enamel segmentation;

[0019] S14. Based on the enamel segmentation results, the high-resolution image is cropped into upper and lower regions with the center of all enamel regions as the boundary, and the maxilla and mandible are segmented in the two regions respectively.

[0020] Furthermore, the reference points are the six-point method, which includes the buccal midpoint B, buccal mesial point MB, buccal distal point DB, lingual midpoint L, lingual mesial point ML, and lingual distal point DL. The specific location method is as follows:

[0021] S21. Perform principal component analysis on the voxel coordinates of all tooth regions, and define the direction of the longest principal component as the tooth direction vector. If the calculation result is If the direction is downward, then the x, y, and z coordinates of that term are multiplied by... ;

[0022] S22. When adjacent teeth are present, the teeth are connected by the left and right adjacent teeth. and teeth Calculate the nearest neighbors of the two teeth, denoted as P and N respectively, and denote the line connecting them as the mesial-distal direction. ;

[0023] S23. When there are no adjacent teeth, create an internal point NC. Its x and y coordinates are derived from the center of the upper / lower jawbone, and its z coordinate is derived from the average of the center points of the enamel of all upper / lower teeth. Draw a line perpendicular to the center point EC of the current tooth's enamel. The straight line connecting the NC-EC points intersects the tooth surface at points P and N; the line connecting these two points is denoted as the mesial-distal direction. ;

[0024] S24, passing through point C, the midpoint of P and N, and using... Define a plane for the normal vector , in plane Teeth trimming Obtain the trimming outline, and starting from points P and N on the outline, move a distance towards the tongue and cheek sides along the outline. The points are obtained as ML, MB, DL, and DB points;

[0025] S25, in plane Draw a perpendicular line from C to PN, intersecting the lingual and buccal sides of the tooth cutting contour at two points, denoted as L and B respectively.

[0026] Furthermore, the nearest neighbor point calculation method for two teeth in S22 is as follows: Extract the three-dimensional surface model of the tooth using the segmentation results of teeth t and t-1, and traverse the teeth... Each point in the surface data, and in the teeth In the surface data, the corresponding nearest neighbor index and point are found using KDTree, and the tooth is obtained through the index. The surface data at that point is the normal vector. The dot product of the difference between that point and the nearest neighbor and the normal vector is the directed distance from the nearest neighbor of the adjacent tooth to that point.

[0027] After traversing all points of the current tooth, find the point with the minimum directed distance, which is the nearest neighbor of the adjacent tooth;

[0028] Current teeth With the next tooth The method for calculating the nearest neighbor N is the same.

[0029] Furthermore, based on the near-, medium-, and far-field directions Further refine the tooth direction vector, and the result is corrected. The calculation method is as follows:

[0030]

[0031]

[0032] In the formula, This is the corrected direction vector of the tooth's long axis (the final Z-axis). Let be the initial major axis direction vector of the tooth. Let be the mesial-distal direction vector of the tooth. Let be the initial buccal-lingual vector of the tooth.

[0033] Furthermore, S3 specifically refers to:

[0034] Based on the reference point, determine the CEJ and ABD points at this location;

[0035] Taking point L as an example, its corresponding CEJ and ABD points are denoted as CEJ. L and ABD L Passing through point C and passing through and Construct a plane in two directions ;

[0036] Calculate the contour formed by the intersection of the plane and the boundary of the enamel region, and take half of the contour closest to the reference point. The set of points within this contour is denoted as . ;

[0037] When the current tooth is the upper tooth, The intersection point with the smallest z-value is denoted as CEJ. L Otherwise, the intersection point with the largest z-value is denoted as CEJ. L ;

[0038] For the jawbone where the current tooth is located, in plane The jawbone's cross-sectional outline is obtained by trimming the jawbone boundary, and the M point and... The constructed straight line divides it into two segments, the segment closer to point L is denoted as... ,exist Searching in CEJ L The closest point is denoted as ABD. L ;

[0039] Based on the above method, the same operation can be performed on other reference points to obtain the corresponding CEJ and ABD points.

[0040] Furthermore, S4 specifically refers to:

[0041] Obtain surface data D from the current tooth, and calculate the PCA principal axis direction for points in D. Make the points in D Projecting the direction yields the projection set. ;

[0042] When the current tooth is an upper tooth, a certain proportion of the points in the projection set before the projected value is taken as the root region. When the current tooth is a lower tooth, a certain proportion of the points after projecting the values ​​from the projection point set is taken as the root region. ;

[0043] Clustering is performed on the root regions, and the number of clusters is equal to the number of roots. This is based on the in-cell point set. Directional projection determines the apex point;

[0044] For the upper teeth, the point with the largest projection value is the root apex; for the lower teeth, the point with the smallest projection value is the root apex.

[0045] Furthermore, the method for calculating the jawbone resorption rate is as follows:

[0046] For the selected reference point, first select the root apex R0 that is closest to it. This point can be selected from the root apex points located in the previous step by calculation to find the closest point.

[0047] The line connecting the reference point and the corresponding ABD points is: , passing through CEJ point towards Draw a perpendicular line, and denote the foot of the perpendicular as point J; similarly, extend R0 to... The foot of the perpendicular line is denoted as R;

[0048] Define the jawbone resorption rate at this reference point.

[0049]

[0050] Where a is the distance between points ABD and J, b is the distance between points R and J, and d is set to 2mm by default;

[0051] The average absorption rate of all points is denoted as:

[0052]

[0053] in and Let a and b be the average values ​​of a and b at each point.

[0054] Furthermore, the reference point is the 30-point method, and the specific positioning method is as follows:

[0055] Determine the P and N points on each tooth and Plane, passing through the center point of the enamel of the target tooth ,by Construct a plane for the axis of rotation ;

[0056] The initial position of the rotation is a plane passing through E and parallel to the line containing PN; the resulting plane is denoted as . When rotated, a plane is formed every 12°, denoted as arrive Calculate using Rodriguez's rotation formula normal vector as follows:

[0057]

[0058] Points P and N in the near-mid-far direction, with Let the cross section of the normal vector pass through the midpoint C of P and N be denoted as . ,flat The contour formed by the intersection with the tooth boundary is denoted as ,all It intersects with it to form 30 points. Based on the arrangement of these points on the contour, they are respectively labeled as points TP0 to TP in a counterclockwise direction. 29 ;

[0059] For each reference point The corresponding CEJ and ABD points are obtained, and the jawbone resorption rate at these points and the average resorption rate of all reference points are calculated accordingly.

[0060] Furthermore, S6 specifically refers to:

[0061] The data from the two time points were segmented into teeth to obtain tooth surface models T0 and T1 at the two time points.

[0062] Point clouds are extracted from two surface models, and principal component analysis is performed on each model to obtain the principal orientation of the point cloud. The two point clouds are initially aligned based on the principal orientation of the model and the midpoint of the model.

[0063] By using the iterative nearest point algorithm, fine alignment of two point clouds is achieved. Through point cloud registration, the corresponding periodontal analysis reference point in T0 is mapped to T1, enabling accurate comparison of changes in jawbone resorption.

[0064] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0065] 1. This method uses CBCT three-dimensional image data as input and utilizes a deep learning model to accurately segment and reconstruct the teeth, enamel, maxilla, and mandible in three dimensions, constructing a complete spatial structure to provide anatomical support for absorption rate calculation.

[0066] 2. This method establishes a standardized reference point system (6 or 30 points), locates the CEJ and ABD points at each reference point, and then combines the root apex information to calculate a continuous "absorption rate" index, replacing subjective grading and realizing an objective and verifiable analysis system.

[0067] 3. This method constructs a point cloud registration algorithm to achieve spatial alignment of teeth at different time points; maps historical and current reference points to analyze the temporal evolution trend of bone resorption; and supports the visual tracking of alveolar bone recovery or progression during patient follow-up.

[0068] 4. This proposal outputs precise lesion location information (such as which tooth, which surface, and the depth of bone resorption), and provides a spatial relationship diagram of CEJ–ABD–apex at the structural level, which facilitates doctors to assess periodontal status and develop individualized treatment plans by combining images.

[0069] 5. This method uses high-resolution CBCT data to reconstruct structures and is highly adaptable to the variability of bone defects through deep learning models. It can be applied to the diagnosis and evaluation of periodontal disease at all stages, and is especially suitable for preoperative planning and postoperative efficacy evaluation in clinical practice.

[0070] 6. This method divides the entire evaluation process into six stages: image segmentation, point localization, structure recognition, root apex localization, ratio calculation, and dynamic registration. Each module can be deployed independently or integrated into an existing CBCT reading system or oral management platform, and has good scalability and engineering implementation value. Attached Figure Description

[0071] Appendix Figure 1 This is a flowchart of the present invention;

[0072] Appendix Figure 2 This is a flowchart of tooth segmentation;

[0073] Appendix Figure 3 This is a diagram showing the results of enamel segmentation;

[0074] Appendix Figure 4 This is a diagram showing the segmentation results of the maxilla and mandible;

[0075] Appendix Figure 5 This is a diagram showing the positions of the six reference points;

[0076] Appendix Figure 6 This is a schematic diagram of the PCA axis direction, nearest neighbor points, and outline.

[0077] Appendix Figure 7 This is a schematic diagram illustrating the determination of MB, ML, DB, and DL points;

[0078] Appendix Figure 8 This is a schematic diagram showing the determination of points B and L;

[0079] Appendix Figure 9 This is a schematic diagram of the cross-sectional profile of tooth enamel and the CEJ point;

[0080] Appendix Figure 10 This is a schematic diagram of ABD point generation;

[0081] Appendix Figure 11 This is a flowchart for locating the apex of the tooth root;

[0082] Appendix Figure 12 This is a schematic diagram of the root apex localization results;

[0083] Appendix Figure 13 These are 15 sets of cross-sectional normal vector diagrams;

[0084] Appendix Figure 14 This is a schematic diagram with 30 reference points;

[0085] Appendix Figure 15 This is a schematic diagram of 30 CEJ points;

[0086] Appendix Figure 16 This is a schematic diagram of 30 ABD points;

[0087] Figure 17 Tooth models at different times: (a) initial state of the tooth model; (b) tooth model aligned using PCA; (c) tooth model finely registered using ICP. Detailed Implementation

[0088] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined in this application.

[0089] The CEJ (Cemento-Enamel Junction) shown in this invention refers to the junction between the crown and root of the tooth. It is an important reference point for judging the degree of periodontal bone resorption. In CBCT, it appears as the boundary between the enamel and cementum at the tooth margin and has certain image recognition characteristics.

[0090] (2) ABD (Apical Base of the Periodontal Bone Defect): This indicates the location of the deepest part of the alveolar bone defect that is close to the root surface. It is a key point for determining the depth of vertical bone resorption.

[0091] (3) Point Cloud Registration: This refers to the process of aligning three-dimensional tooth point cloud data from two or more CBCT images acquired at different times through spatial transformation. It is the core of dynamic tracking of changes.

[0092] The AAP / EFP 2018 classification criteria (American Academy of Periodontology / European Federation of Periodontology) explicitly proposes using "bone resorption depth" as a criterion for classifying the severity of periodontitis; the resorption rate quantification system proposed in this paper corresponds exactly to this.

[0093] (2) CBCT application guidelines (such as AAP 2020): emphasize the advantages of CBCT in assessing the morphology and distribution of periodontal bone defects; this proposal is based on CBCT for three-dimensional structural identification, which is in line with the technical path of international clinical guidelines.

[0094] like Figure 1 As shown, a three-dimensional intelligent method for monitoring dynamic changes in jawbone resorption rate based on CBCT is disclosed, including the following steps:

[0095] S1. Obtain an initial three-dimensional image of the periodontium, segment the initial three-dimensional image, and automatically extract the three-dimensional structures of teeth, enamel, maxilla, and mandible;

[0096] S2. Establish standardized anatomical reference points on the surface of each tooth;

[0097] S3. For each reference point, automatically identify the corresponding cementoenamel boundary and the base point of the periodontal bone defect apex;

[0098] S4. Based on the surface data points of a single tooth, the number and location of the root apex of the current tooth are obtained through principal component analysis, clustering, and projection methods.

[0099] S5. Based on the cementoenamel boundary and the base of the periodontal bone defect apex at each reference point, as well as the nearest root apex, calculate the jawbone resorption rate at that reference point.

[0100] S6. For CBCT images of the same patient acquired at different time points, spatial alignment of the three-dimensional tooth model is achieved through point cloud registration, and historical reference points are mapped to dynamically monitor the trend of jawbone resorption.

[0101] Image segmentation involves tooth segmentation, enamel segmentation, maxillary segmentation, and mandibular segmentation. Because the original CBCT has high resolution, direct processing would be computationally intensive. Therefore, this approach employs a multi-resolution, multi-stage segmentation process. For example... Figure 2As shown, the specific segmentation process is as follows: the image is downsampled to a low resolution (e.g., voxel size 0.5mm). 0.5mm First, four-category segmentation is performed at a resolution of 0.5mm to locate the upper and lower dentition, maxilla and mandible regions. Based on the segmentation results, the upper and lower dentition regions are cropped, and tooth centroid localization, tooth classification, and tooth segmentation are performed at low resolution. The position of each tooth is determined based on the segmentation results at low resolution. Then, at high resolution (e.g., voxel size of 0.15mm), the segmentation is performed. 0.15mm Each tooth region was cropped at 0.15mm for individual tooth and enamel segmentation. Based on the enamel segmentation results, the high-resolution image was cropped into upper and lower regions, with the center of all enamel regions as the boundary. Maxillary and mandibular segmentation was then performed separately within each of the two regions. The enamel segmentation results and the maxillary and mandibular segmentation results are shown below. Figure 3 , 4 As shown.

[0102] The segmentation network uses a 3D Unet model, which can be replaced with any feasible network model structure to achieve automatic segmentation of teeth, enamel, and maxillae.

[0103] To analyze periodontal and jawbone resorption, the first step is to determine the reference points for analysis. A common method is to divide the tooth into four surfaces: buccal, lingual, mesial, and distal, and then select specific points on these surfaces. For example, using the six-point method... Figure 5 As shown, the six points are: midpoint of the buccal side (B), mesial buccal side (MB), distal buccal side (DB), midpoint of the lingual side (L), mesial lingual side (ML), and distal lingual side (DL). B and L are the midpoints of the buccal and lingual sides, respectively, while the other four points are the buccal and lingual points on the mesial and distal surfaces, respectively. The positions of the six reference points are as follows: Figure 5 As shown.

[0104] To determine the location of each tooth reference point, principal component analysis (PCA) is first performed on the voxel coordinates of all tooth regions. The direction of the longest-distributed principal component is defined as the tooth direction vector. If the calculation result is If the direction is downward, then the x, y, and z coordinates of that term are multiplied by... .

[0105] To further determine the mesial-distal orientation of the teeth, two cases are considered: those with adjacent teeth and those without. When adjacent teeth are present, the orientation is determined by the left and right adjacent teeth. and teeth Calculate the nearest neighbors of the two teeth, denoted as P and N respectively, and denote the line connecting them as the mesial-distal direction. .

[0106] The nearest neighbor between two teeth is calculated as follows: The three-dimensional surface model of the tooth, i.e., the tooth boundary, is extracted based on the segmentation results of teeth t and t-1. The teeth are then traversed. Each point in the surface data, and in the teeth In the surface data, the corresponding nearest neighbor index and point are found using a KDTree (K-Dimensional Tree), and the tooth is retrieved using the index. The surface data at a given point is the normal vector. The dot product of the difference between this point and its nearest neighbor and the normal vector is the directed distance from the nearest neighbor of the adjacent tooth to this point. After traversing all points of the current tooth, the point with the minimum directed distance is the nearest neighbor of the adjacent tooth. With the next tooth The method for calculating the nearest neighbor N is the same.

[0107] Because of missing teeth, or when a tooth is located at the end of the dentition and has fewer than two adjacent teeth, it is impossible to determine the mesial-distal direction based on the nearest point of the adjacent teeth. Therefore, an internal point NC is created, whose x and y coordinates are derived from the center of the maxilla / mandible, and whose z coordinate is derived from the average of the center points of the enamel of all maxilla / mandible teeth. A perpendicular line is drawn from the enamel center point EC of the current tooth to the center point of the enamel of the tooth. The straight line connecting the NC-EC points intersects the tooth surface at points P and N; the line connecting these two points is denoted as the mesial-distal direction. .

[0108] Based on near-mid- and far-mid directions This can further refine the tooth direction vector, and the result will be corrected. The calculation method is as follows:

[0109]

[0110]

[0111] In the formula, This is the corrected direction vector of the tooth's long axis (the final Z-axis). Let be the initial major axis direction vector of the tooth. Let be the mesial-distal direction vector of the tooth. Let be the initial buccal-lingual vector of the tooth.

[0112] Passing through the midpoint C of P and N, and with Define a plane for the normal vector In plane Teeth trimming Obtain the trimming outline. The PCA axis direction, nearest neighbor points, and outline lines are as follows: Figure 6 As shown. Starting from points P and N on the outline, move a distance towards the tongue and cheek along the outline. The points obtained are used as ML, MB, DL, and DB points. A value of 2mm is typically used. The result is as follows: Figure 7 As shown.

[0113] Finally, on the plane Draw a perpendicular line from C to PN, intersecting the lingual and buccal sides of the tooth trimming contour at two points, denoted as L and B respectively. At this point, the six reference points are located. The result is as follows... Figure 8 As shown.

[0114] Based on the reference point, the CEJ and ABD points at that location can be determined. Taking point L as an example, its corresponding CEJ and ABD points are denoted as CEJL and ABDL. Passing through point C, and traversing... and Construct a plane in two directions Calculate the contour formed by the intersection of the plane and the boundary of the enamel region, and take half of the contour closest to the reference point; the set of points in this contour is denoted as . When the current tooth is the upper tooth, The intersection point with the smallest z-value is denoted as CEJL; otherwise, the intersection point with the largest z-value is denoted as CEJL.

[0115] For the jawbone where the current tooth is located, in plane The jawbone's cross-sectional outline is obtained by trimming the jawbone boundary, and the M point and... The constructed straight line divides it into two segments, the segment closer to point L is denoted as... .exist Find the point closest to CEJL in the middle, and denote it as ABDL.

[0116] Based on the above method, the same operation can be performed on other reference points to obtain the corresponding CEJ and ABD points.

[0117] The procedure for locating the apex of the tooth root is as follows: Figure 11 As shown. Surface data D is obtained from the current tooth, and the principal axis directions of PCA are calculated for points in D. Make the points in D Projecting the direction yields the projection set. When the current tooth is an upper tooth, a certain proportion of the points in the projection point set are selected from the projection point set, such as selecting the points at 0.15 above the upper tooth as the root region. When the current tooth is a lower tooth, a certain proportion of the projection values ​​are taken from the projection point set, such as taking the point set of the lower 0.15 as the root region. Clustering is performed on the root regions, and the number of clusters is equal to the number of roots. The number of clusters is based on the in-cell point set. The apex is determined by directional projection. For the upper teeth, the point with the largest projection value is the apex, and for the lower teeth, the point with the smallest projection value is the apex.

[0118] Test the teeth using the methods described above, such as Figure 12 As shown, the correct root position can be obtained for teeth with 1, 2, 3, or 4 root apexes.

[0119] For the selected reference point, first select the nearest root apex point R0. This point can be selected from the root apex points located in the previous step through calculation to find the closest point. Let the line connecting the reference point and the corresponding ABD point be... , passing through CEJ point towards Draw a perpendicular line, and denote the foot of the perpendicular as point J. Similarly, extend R0 to... The foot of the perpendicular line is denoted as R.

[0120] Define the jawbone resorption rate at this reference point.

[0121]

[0122] Where a is the distance between points ABD and J, b is the distance between points R and J, and d is set to 2mm by default.

[0123] The average absorption rate of all points is denoted as:

[0124]

[0125] in and Let a and b be the average values ​​of a and b at each point.

[0126] To more comprehensively examine the jawbone resorption around the teeth, the reference points can be selected more densely, not limited to the aforementioned six reference points. Therefore, 30 reference points are collected evenly over a week. The location of the 30 reference points is as follows.

[0127] Following the method described above, determine the P and N points on each tooth. Plane. Passing through the center point of the enamel of the target tooth. ,by Construct a plane for the axis of rotation The initial position of the rotation is a plane passing through E and parallel to the line containing PN; the resulting plane is denoted as . During rotation, a plane is formed every 12°, denoted as . arrive Plane normal vectors are as follows Figure 13 As shown. Therefore, the Rodriguez rotation formula is applied to calculate... normal vector as follows:

[0128]

[0129] The calculation method for points P and N in the near-mid-far direction is the same as that of the 6-point method, with... Let the cross section of the normal vector pass through the midpoint C of P and N be denoted as . .flat The contour formed by the intersection with the tooth boundary is denoted as .all It intersects with it to form 30 points, such as Figure 14 As shown. Based on the arrangement of these points on the contour, they are denoted as points TP0 to TP29 in a counter-clockwise direction.

[0130] For each reference point (in Following the method described above, the corresponding CEJ and ABD points can be obtained, and the jawbone resorption rate at each point and the average resorption rate of all reference points can be calculated accordingly. The generated CEJ and ABD points are shown below. Figure 15 and 16 As shown.

[0131] For data from multiple time points, the progression of a patient's periodontal condition can be monitored by comparing the bone resorption of corresponding teeth. To this end, tooth segmentation was performed on data from two time points to obtain tooth surface models T0 and T1. Point clouds were extracted from the two surface models, and principal component analysis was performed on each to obtain the principal orientation of the point clouds. The two point clouds were initially aligned based on the principal orientation and the midpoint of the model. Then, the Iterative Closest Point (ICP) algorithm was used to achieve fine alignment of the two point clouds. The tooth surface models before and after registration are shown below. Figure 17 As shown, point cloud registration can map the corresponding periodontal analysis reference points in T0 to T1, enabling precise comparison of jawbone resorption changes.

[0132] The above embodiments of this application can complement each other without causing conflict.

[0133] It should be noted that the dimensions of layers and regions may be exaggerated in the accompanying drawings for clarity. Furthermore, it is understood that when an element or layer is referred to as being "on" another element or layer, it can be directly on the other element, or there may be intermediate layers. Additionally, it is understood that when an element or layer is referred to as being "below" another element or layer, it can be directly below the other element, or there may be more than one intermediate layer or element. Furthermore, it is also understood that when a layer or element is referred to as being "between" two layers or two elements, it can be the only layer between the two layers or two elements, or there may be more than one intermediate layer or element. Similar reference numerals throughout indicate similar elements.

[0134] The term “multiple” means two or more, unless otherwise expressly defined.

[0135] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0136] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A three-dimensional intelligent method for monitoring dynamic changes in jawbone resorption rate based on CBCT, characterized in that, Includes the following steps: S1. Obtain an initial three-dimensional image of the periodontium, segment the initial three-dimensional image, and automatically extract the three-dimensional structures of teeth, enamel, maxilla, and mandible; S2. Establish standardized anatomical reference points on the surface of each tooth; S3. For each reference point, automatically identify the corresponding cementoenamel boundary and the base point of the periodontal bone defect apex; S4. Based on the surface data points of a single tooth, the number and location of the root apex of the current tooth are obtained through principal component analysis, clustering, and projection methods. S5. Based on the cementoenamel boundary and the base of the periodontal bone defect apex at each reference point, as well as the nearest root apex, calculate the jawbone resorption rate at that reference point. S6. For CBCT images of the same patient acquired at different time points, spatial alignment of the three-dimensional tooth model is achieved through point cloud registration, and historical reference points are mapped to dynamically monitor the trend of jawbone resorption.

2. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 1, characterized in that, The specific segmentation method in S1 is as follows: S11. Perform four-category segmentation at low resolution to locate the upper dentition, lower dentition, maxilla, and mandible regions. S12. Based on the segmentation results, the upper and lower dentition regions are cropped out. The centroid localization, tooth classification and tooth segmentation of the teeth are performed at low resolution. The position of each tooth is determined according to the tooth segmentation results at low resolution. S13. Cropping each tooth region at high resolution for individual tooth and enamel segmentation; S14. Based on the enamel segmentation results, the high-resolution image is cropped into upper and lower regions with the center of all enamel regions as the boundary, and the maxilla and mandible are segmented in the two regions respectively.

3. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 1, characterized in that, The reference points are the six-point method, which includes the buccal midpoint B, buccal mesial point MB, buccal distal point DB, lingual midpoint L, lingual mesial point ML, and lingual distal point DL. The specific location method is as follows: S21. Perform principal component analysis on the voxel coordinates of all tooth regions, and define the direction of the longest principal component as the tooth direction vector. If the calculation result is If the direction is downward, then the x, y, and z coordinates of that term are multiplied by... ; S22. When adjacent teeth are present, the teeth are connected by the left and right adjacent teeth. and teeth Calculate the nearest neighbors of the two teeth, denoted as P and N respectively, and denote the line connecting them as the mesial-distal direction. ; S23. When there are no adjacent teeth, create an internal point NC. Its x and y coordinates are derived from the center of the upper / lower jawbone, and its z coordinate is derived from the average of the center points of the enamel of all upper / lower teeth. Draw a line perpendicular to the center point EC of the current tooth's enamel. The straight line connecting the NC-EC points intersects the tooth surface at points P and N; the line connecting these two points is denoted as the mesial-distal direction. ; S24, passing through point C, the midpoint of P and N, and using... Define a plane for the normal vector , in plane Teeth trimming Obtain the trimming outline, and starting from points P and N on the outline, move a distance towards the tongue and cheek sides along the outline. The points are obtained as ML, MB, DL, and DB points; S25, in plane Draw a perpendicular line from C to PN, intersecting the lingual and buccal sides of the tooth cutting contour at two points, denoted as L and B respectively.

4. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 3, characterized in that, The nearest neighbor between two teeth in S22 is calculated as follows: The three-dimensional surface model of the tooth is extracted using the segmentation results of teeth t and t-1, and then the teeth are traversed... Each point in the surface data, and in the teeth In the surface data, the corresponding nearest neighbor index and point are found using KDTree, and the tooth is obtained through the index. The surface data at that point is the normal vector. The dot product of the difference between that point and the nearest neighbor and the normal vector is the directed distance from the nearest neighbor of the adjacent tooth to that point. After traversing all points of the current tooth, find the point with the minimum directed distance, which is the nearest neighbor of the adjacent tooth; Current teeth With the next tooth The method for calculating the nearest neighbor N is the same.

5. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 3, characterized in that, Based on near-mid- and far-mid directions Further refine the tooth direction vector, and the result is corrected. The calculation method is as follows: ； ； In the formula, This is the corrected direction vector of the tooth's long axis (the final Z-axis). Let be the initial major axis direction vector of the tooth. Let be the mesial-distal direction vector of the tooth. Let be the initial buccal-lingual vector of the tooth.

6. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 1, characterized in that, S3 specifically refers to: Based on the reference point, determine the CEJ and ABD points at this location; Taking point L as an example, its corresponding CEJ and ABD points are denoted as CEJ. L and ABD L Passing through point C and passing through and Construct a plane in two directions ; Calculate the contour formed by the intersection of the plane and the boundary of the enamel region, and take half of the contour closest to the reference point. The set of points within this contour is denoted as . ; When the current tooth is the upper tooth, The intersection point with the smallest z-value is denoted as CEJ. L Otherwise, the intersection point with the largest z-value is denoted as CEJ. L ; For the jawbone where the current tooth is located, in plane The jawbone's cross-sectional outline is obtained by trimming the jawbone boundary, and the M point and... The constructed straight line divides it into two segments, the segment closer to point L is denoted as... ,exist Searching in CEJ L The closest point is denoted as ABD. L ; Based on the above method, the same operation can be performed on other reference points to obtain the corresponding CEJ and ABD points.

7. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 1, characterized in that, S4 specifically refers to: Obtain surface data D from the current tooth, and calculate the PCA principal axis direction for points in D. Make the points in D Projecting the direction yields the projection set. ; When the current tooth is an upper tooth, a certain proportion of the points in the projection set before the projected value is taken as the root region. When the current tooth is a lower tooth, a certain proportion of the points after projecting the values ​​from the projection point set is taken as the root region. ; Clustering is performed on the root regions, and the number of clusters is equal to the number of roots. This is based on the in-cell point set. Directional projection determines the apex point; For the upper teeth, the point with the largest projection value is the root apex; for the lower teeth, the point with the smallest projection value is the root apex.

8. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 1, characterized in that, The method for calculating jawbone resorption rate is as follows: For the selected reference point, first select the root apex R0 that is closest to it. This point can be selected from the root apex points located in the previous step by calculation to find the closest point. The line connecting the reference point and the corresponding ABD points is: , passing through CEJ point towards Draw a perpendicular line, and denote the foot of the perpendicular as point J; Similarly, R0 to The foot of the perpendicular line is denoted as R; The jawbone resorption rate at this reference point is defined as follows: ； Where a is the distance between points ABD and J, b is the distance between points R and J, and d is set to 2mm by default; The average absorption rate of all points is denoted as: ； in and Let a and b be the average values ​​of a and b at each point.

9. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 1, characterized in that, The reference point is the 30-point method, and the specific positioning method is as follows: Determine the P and N points on each tooth and Plane, passing through the center point of the enamel of the target tooth ,by Construct a plane for the axis of rotation ; The initial position of the rotation is a plane passing through E and parallel to the line containing PN; the resulting plane is denoted as . When rotated, a plane is formed every 12°, denoted as arrive Calculate using Rodriguez's rotation formula normal vector as follows: ； Points P and N in the near-mid-far direction, with Let the cross section of the normal vector pass through the midpoint C of P and N be denoted as . ,flat The contour formed by the intersection with the tooth boundary is denoted as ,all It intersects with it to form 30 points. Based on the arrangement of these points on the contour, they are respectively labeled as points TP0 to TP in a counterclockwise direction. 29 ; For each reference point The corresponding CEJ and ABD points are obtained, and the jawbone resorption rate at these points and the average resorption rate of all reference points are calculated accordingly.

10. The method for monitoring the dynamic changes of three-dimensional intelligent jawbone resorption rate based on CBCT according to claim 1, characterized in that, S6 specifically refers to: The data from the two time points were segmented into teeth to obtain tooth surface models T0 and T1 at the two time points. Point clouds are extracted from two surface models, and principal component analysis is performed on each model to obtain the principal orientation of the point cloud. The two point clouds are initially aligned based on the principal orientation of the model and the midpoint of the model. By using the iterative nearest point algorithm, fine alignment of two point clouds is achieved. Through point cloud registration, the corresponding periodontal analysis reference point in T0 is mapped to T1, enabling accurate comparison of jawbone resorption changes.

Images