Methods for detecting bone resorption areas in periapical radiographs for the identification of periodontitis

By spatially aligning and analyzing the degree of imaging anomalies in full-mouth periapical radiographs, the problem of misjudging bone resorption areas in existing technologies has been solved, achieving more stable and reliable detection of bone resorption areas.

CN123071230APending Publication Date: 2026-11-10THE FIRST AFFILIATED HOSPITAL OF MEDICAL COLLEGE OF XIAN JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611025252.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-10
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing whole-mouth periapical radiograph image analysis methods lack effective judgment on whether areas with localized grayscale reduction or bone contour interruption belong to high-risk imaging areas in whole-mouth images, which can easily lead to misjudgment.

Method used

By acquiring periapical radiographs from multiple historical and current patients, spatial alignment, grayscale normalization, noise reduction, and orientation correction are performed to identify tooth instance regions. The degree of projection overlap and difference is analyzed, and confidence weights are set based on the degree of imaging anomalies for weighted analysis to eliminate the interference of imaging differences on the detection results.

Benefits of technology

It improves the stability and reliability of periapical radiograph image analysis results, reduces the risk of misjudgment, and enhances the credibility of bone resorption area detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123071230A_ABST
    Figure CN123071230A_ABST
Patent Text Reader

Abstract

This invention relates to the field of medical imaging technology, specifically to a method for detecting bone resorption regions in full-mouth periapical radiographs for identifying periodontitis. This invention maps dental images from different patients and in different shooting postures onto a unified structural coordinate system to obtain aligned tooth instance regions. By analyzing the degree of projection overlap of each tooth instance region, the degree of difference between each tooth instance region and its adjacent tooth instance regions, and the imaging variability of tooth instance regions at the same location across multiple historical patients, the degree of imaging anomaly at each location is obtained. Then, confidence weights are set to weight the projection overlap and difference degrees for the current patient, respectively, to obtain the imaging risk level of each tooth instance region for the current patient. This invention reduces the interference of imaging differences on the detection results and improves the reliability, stability, and interpretability of periapical radiograph image analysis results.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of medical imaging technology, specifically to a method for detecting bone resorption areas in periapical radiographs for the identification of periodontitis. Background Technology

[0002] Clinically, periapical radiographs are often used to analyze periodontal bone resorption. These radiographs typically consist of multiple two-dimensional X-rays taken from various positions in the maxilla and mandible. Each radiograph covers only the root of one or a few adjacent teeth and the surrounding alveolar bone. The shooting angle, exposure intensity, and tooth anatomy often differ between different positions in the same patient. For example, the posterior teeth are more prone to skewed shooting angles due to their posterior location, while the anterior teeth are more likely to show overexposure or periapical stretching. This difference directly affects the visualization of the alveolar bone margins, resulting in some areas having clear bone boundaries while others appear blurred or overlapped.

[0003] Periapical radiographs are two-dimensional projection images. In areas of bone resorption commonly seen in periodontitis, the buccal and lingual bone plates, root images, and adjacent structures often overlap, easily appearing as localized grayscale reduction or bone contour interruption on periapical radiographs. These appearances may stem from actual bone resorption or are simply imaging results caused by imaging conditions or structural overlap. Current detection methods mostly analyze single periapical radiographs directly, lacking the ability to determine whether areas of localized grayscale reduction or bone contour interruption constitute high-risk imaging areas within the full-mouth imaging, easily misinterpreting imaging problems as structural changes. Summary of the Invention

[0004] To address the above-mentioned technical problems, the present invention aims to provide a method for detecting bone resorption areas in a full-mouth periapical radiograph for the identification of periodontitis.

[0005] According to an embodiment of the present invention, a method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis is provided, and the specific technical solution adopted is as follows:

[0006] Acquire periapical radiograph data sets corresponding to multiple historical patients and the current patient respectively, and spatially align the tooth instance regions in the periapical radiograph data sets corresponding to each historical patient and the current patient to obtain multiple aligned tooth instance regions corresponding to each historical patient and the current patient.

[0007] For each historical patient and current patient, the degree of projection overlap of each tooth instance region is analyzed, and the degree of difference between each tooth instance region and its adjacent tooth instance regions is analyzed;

[0008] For multiple historical patients, based on the degree of projection overlap and the degree of difference, the imaging variability of the tooth instance region at the same location among multiple historical patients is analyzed to obtain the degree of imaging abnormality of the tooth instance region at each location.

[0009] Based on the degree of imaging abnormality, confidence weights are set, and the degree of projection overlap and the degree of difference of the current patient are weighted respectively to obtain the imaging risk of each tooth instance region of the current patient.

[0010] In some embodiments of the present invention, acquiring a set of periapical radiographs corresponding to multiple historical patients and the current patient includes:

[0011] The hospital database is used to obtain sets of original periapical radiographs corresponding to multiple historical patients and current patients. The sets of original periapical radiographs include multiple original periapical radiographs.

[0012] For each original periapical radiograph, grayscale normalization, noise reduction, size unification, and orientation correction are performed to obtain a set of periapical radiographs for multiple historical patients, each set of which includes multiple periapical radiographs.

[0013] In some embodiments of the present invention, spatial alignment is performed on the tooth instance regions in the periapical radiograph datasets corresponding to each historical patient and the current patient, including:

[0014] For each of the aforementioned periapical radiographs, the tooth regions are identified and located to obtain all tooth instance regions in the periapical radiograph data set corresponding to each historical patient;

[0015] For each tooth instance region, the root main axis and root apex position are extracted. Using the root main axis and root apex position as alignment references, the tooth instance regions in all periapical radiographs of the same historical patient and the current patient are spatially aligned.

[0016] In some embodiments of the present invention, for each periapical radiograph, tooth regions are identified and located to obtain all tooth instance regions in the periapical radiograph data set corresponding to each historical patient, including:

[0017] For each of the aforementioned periapical radiographs, the image region where the tooth is located is identified and marked as the tooth region using a tooth structure recognition model;

[0018] Within the tooth region, different teeth are divided into instance-level regions, so that each tooth corresponds to an independent tooth instance region, thus obtaining all tooth instance regions in the periapical radiograph data set corresponding to each historical patient.

[0019] In some embodiments of the present invention, for each historical patient and the current patient, the degree of projection overlap of each said tooth instance region is analyzed, including:

[0020] For each historical and current patient, the number of root apex regions in each tooth instance is identified using a tooth structure recognition model;

[0021] Analyze the continuity of the bone structure margins in the root region of each of the tooth instances;

[0022] Extract the bone structure edge pixels of the root region in each tooth instance area, determine the direction of each bone structure edge pixel, and obtain the angle between the direction and the main axis of the tooth root.

[0023] The degree of projection overlap of each tooth instance region is obtained based on the number of root apexes, the continuity of the bone structure edges, and the included angle.

[0024] In some embodiments of the present invention, analyzing the continuity of the bone structure margins of the root region in each of the said tooth instance regions includes:

[0025] The maximum length of the continuous bone margin of the root region in each of the tooth instance regions is detected, and combined with the theoretical continuous bone margin length of the root region in each of the tooth instance regions, the degree of continuity of the bone structure margin of the root region in each of the tooth instance regions is obtained.

[0026] In some embodiments of the present invention, for each historical patient and the current patient, the degree of difference between each tooth instance region and its adjacent tooth instance regions is analyzed, including:

[0027] For each historical patient and current patient, the height of the root region in each tooth instance area is obtained, and then the height difference between each tooth instance area and its adjacent teeth is analyzed;

[0028] For each historical patient and current patient, the differences in the angle between each tooth instance region and its adjacent teeth are analyzed;

[0029] Based on the height difference and the included angle difference, the degree of difference between each tooth instance region and its adjacent tooth instance regions is obtained.

[0030] In some embodiments of the present invention, analyzing the angle difference between each tooth instance region and its adjacent teeth includes:

[0031] Match all bone structure edge pixels in the root region of each tooth instance region with all bone structure edge pixels in the root region of its adjacent tooth instance regions to obtain several matching pixel pairs;

[0032] Obtain the included angles corresponding to all bone structure edge pixels in the root region of each tooth instance region, calculate the absolute value of the difference between the included angles corresponding to the two bone structure edge pixels in the matching pixel pair, traverse all the matching pixel pairs, calculate the mean of the absolute values ​​of the included angle differences, and obtain the angle difference between each tooth instance region and its adjacent teeth.

[0033] In some embodiments of the present invention, for multiple historical patients, based on the degree of projection overlap and the degree of difference, the imaging variability of the tooth instance region at the same location among multiple historical patients is analyzed to obtain the degree of imaging abnormality of the tooth instance region at each location, including:

[0034] Align the aligned tooth instance regions corresponding to multiple historical patients across patients to obtain tooth instance regions corresponding to multiple historical patients at the same location.

[0035] Based on the degree of projection overlap, the variance of the degree of projection overlap of tooth instance regions at the same location for multiple historical patients is calculated.

[0036] Based on the degree of difference, the variance of the degree of difference in the tooth instance area at the same location for multiple historical patients is calculated.

[0037] Based on the variance of the projection overlap and the variance of the difference, and combining the projection overlap and the difference, the imaging anomaly degree of the tooth instance region at each location is obtained.

[0038] In some embodiments of the present invention, confidence weights are set according to the degree of imaging anomaly, including:

[0039] Set a threshold for the degree of imaging anomalies;

[0040] Compare whether the degree of imaging anomaly is greater than the threshold for the degree of imaging anomaly;

[0041] If so, then set the confidence weight to the degree of imaging anomaly;

[0042] If not, set the confidence weight to 0.

[0043] Compared with existing technologies, the method for detecting bone resorption areas in full-oral periapical radiographs for identifying periodontitis provided by this invention has the following beneficial effects:

[0044] This invention effectively eliminates the influence of tooth tilt, scale differences, and imaging angle changes on image analysis results by mapping dental images of different patients and in different shooting postures to a unified structural coordinate system.

[0045] This invention, by quantitatively analyzing the consistency of structural edge direction in tooth instance regions and the structural change trends between adjacent tooth instance regions, can objectively characterize the structural continuity, structural overlap risk, and imaging stability of different tooth instance regions during periapical radiograph imaging. Furthermore, by analyzing the differences in imaging stability across patients, it quantifies the reliability of different tooth instance regions, reducing the interference of imaging differences on the detection results. This avoids the risk of misjudgment caused by relying solely on grayscale changes or local thresholds, and improves the ability to identify interference-prone areas from the perspective of imaging geometry and structural behavior, thereby enhancing the stability and interpretability of periapical radiograph image analysis results.

[0046] The bone region images of the same tooth position in periapical radiographs of different patients show varying degrees of difference. By combining the degree of projection overlap and difference with the analysis of the imaging variability of the same tooth position across multiple historical patients, a dual judgment can be made as to whether a certain area belongs to the "image distortion-prone area". The imaging risk of different tooth positions in full-mouth periapical radiographs is not consistent. Without introducing diagnostic conclusions, the reliability of the detection results of bone resorption areas is increased based solely on the information of the images themselves, thereby improving the reliability of periapical radiograph image analysis results. Attached Figure Description

[0047] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram of the basic process of a method for detecting bone resorption areas in a periapical radiograph for identifying periodontitis, provided in one embodiment of the present invention. Detailed Implementation

[0049] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the whole-mouth periapical radiograph bone resorption region detection method for periodontitis identification proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms such as “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of additional identical elements in the article or device that includes the element.

[0051] The following description, in conjunction with the accompanying drawings, details a specific scheme for detecting bone resorption areas in a full-mouth periapical radiograph for the identification of periodontitis, provided by the present invention.

[0052] Please see Figure 1 This illustrates the basic flowchart of a method for detecting bone resorption areas in a full-mouth periapical radiograph for identifying periodontitis, provided by an embodiment of the present invention.

[0053] like Figure 1 As shown, an embodiment of the present invention provides a method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis, specifically including:

[0054] S100: Obtain the periapical radiograph data sets corresponding to multiple historical patients and the current patient respectively, and spatially align the tooth instance regions in the periapical radiograph data sets corresponding to each historical patient and the current patient to obtain multiple aligned tooth instance regions corresponding to each historical patient and the current patient.

[0055] To detect bone resorption areas in the current patient's full-mouth periapical radiograph, it is first necessary to acquire periapical radiograph datasets from multiple historical patients and the current patient. Further steps include:

[0056] First, the hospital database is used to obtain sets of original periapical radiographs for multiple historical patients and current patients. These sets of original periapical radiographs include multiple original periapical radiographs. Specifically, the data sets of original periapical radiographs from multiple historical and current patients are obtained from the hospital's database. Each original periapical radiograph is a two-dimensional grayscale image, covering one or more adjacent teeth, their corresponding roots, and the surrounding alveolar bone region. The acquired original periapical radiographs are categorized at the patient level, with multiple images belonging to the same historical patient grouped into the same original periapical radiograph set. Each historical patient corresponds to one original periapical radiograph set, which covers at least multiple tooth positions in the maxilla and mandible of the historical patient. The imaging order and basic imaging parameters of the original periapical radiographs are preserved, establishing a basic data structure of "patient-image set," rather than single-image analysis. Furthermore, there may be differences in imaging angle, exposure conditions, and coverage between multiple periapical radiographs from the same historical patient.

[0057] Then, for each original periapical radiograph image data, grayscale normalization, noise reduction, size unification and orientation correction are performed to obtain periapical radiograph image data sets corresponding to multiple historical patients. The periapical radiograph image data set includes multiple periapical radiograph images. Specifically, the following preprocessing operations were performed on each original root tip image data: grayscale normalization was performed on each original root tip image data using maximum / minimum value normalization methods, normalizing the grayscale values ​​to the range of [0,1] to reduce the overall brightness difference caused by different shooting conditions and ensure that subsequent feature extraction is based on a unified grayscale scale rather than the original exposure difference; noise suppression was performed on each original root tip image data using mean filtering, median filtering, bilateral filtering methods, etc., to reduce the impact of high-frequency noise on subsequent structure recognition; resampling was performed on each original root tip image data using SimpleITK to unify the size of each original root tip image data, and orientation correction based on a physical model was performed on each original root tip image data to ensure that different images are consistent in pixel resolution and basic orientation, avoiding errors in subsequent spatial analysis due to inconsistent image scales.

[0058] To eliminate imaging differences caused by different shooting angles in the periapical radiograph data set corresponding to the same patient, after obtaining the periapical radiograph data sets corresponding to multiple historical patients and the current patient, the tooth instance regions in the periapical radiograph data sets corresponding to each historical patient and the current patient are spatially aligned to obtain multiple aligned tooth instance regions corresponding to each historical patient and the current patient. Further steps include:

[0059] First, for each periapical radiograph, the tooth regions are identified and located, resulting in all tooth instance regions in the periapical radiograph dataset for each historical patient. Specifically, for each periapical radiograph, a tooth structure recognition model is used to analyze the image, identify the image region where the tooth is located, and mark the corresponding region as a tooth region. This yields multiple aligned tooth instance regions for each historical and current patient. The tooth region includes the crown region, the root region, and the adjacent alveolar bone region around the root. It should be noted that this tooth region only identifies the "structural location," not the disease. Within the tooth region, different teeth are divided into instance-level segments, so that each tooth corresponds to an independent tooth instance region. Each tooth instance region includes at least the complete root image of the tooth and the alveolar bone region directly adjacent to the tooth. This yields all tooth instance regions in the periapical radiograph dataset for each historical patient, providing the smallest analytical unit for subsequent alignment and analysis "on a tooth-by-tooth basis."

[0060] Then, for each tooth instance region, the root principal axis and root apex position are extracted. Using the root principal axis and root apex position as alignment benchmarks, spatial alignment is performed on tooth instance regions in all periapical radiographs of the same historical patient and the current patient. Specifically, for each tooth instance region in the same historical patient image set, the root principal axis and root apex position are extracted. The root principal axis can be obtained by fitting the tooth contour after skeletalization processing, or by statistically analyzing the direction of the edge points of the root region. The direction of the root principal axis is from the root apex to the crown. The root apex position can be obtained by comparing a preset root apical structure template with the local area of ​​the image. Using the root principal axis and root apex position as alignment benchmarks, spatial alignment processing is performed on corresponding tooth instance regions in different periapical radiographs of the same historical patient and the current patient, so that they have consistent positions and orientations under a unified coordinate system. It should be noted that this alignment process allows for local morphological differences, but ensures the consistency of the root principal axis direction and relative spatial relationship. By spatially aligning the tooth instance areas, the imaging differences caused by different shooting angles of the same patient are eliminated, thus constructing logical "full mouth consistency" rather than simple image stitching.

[0061] In addition, consistency checks are performed on the aligned tooth instance areas. When obvious deviations are detected, such as obvious overlaps or obvious interruptions, the alignment parameters (alignment benchmark) are adjusted to ensure the reasonable matching of the root structure in different images and to avoid incorrect alignment affecting subsequent stability and reliability analysis.

[0062] At this point, a set of aligned tooth instance region image data corresponding to each historical patient has been obtained; each tooth instance region contains the tooth root and its surrounding alveolar bone image region in a unified coordinate system. This aligned tooth instance region image data can be directly used for subsequent cross-patient stability analysis and imaging confidence weight construction.

[0063] Because periapical radiographs are two-dimensional projection images, there are inherent differences in the imaging process for different tooth positions. For example, the posterior tooth region has a complex structure and significant overlap in root bifurcation areas. Even between the same tooth position in different patients, the imaging appearance of the posterior tooth region can easily fluctuate significantly. If the imaging differences between different tooth instance regions are not distinguished, unstable imaging areas can easily be mistaken for structural changes. In step S100, the aligned tooth instance regions are statistically analyzed for tooth instance regions at the same position from multiple historical patients. This analysis includes aspects such as grayscale distribution, bone margin morphology, and structural overlap to assess the stability of different tooth instance regions in periapical radiographs. Simultaneously, combining tooth structural features and tooth position attributes, the credibility of different tooth instance regions in the current imaging scenario is quantified, and a corresponding credibility weight is assigned to each region. The specific implementation steps are S200 to S400.

[0064] S200: For each historical patient and current patient, analyze the degree of projection overlap of each tooth instance region and the degree of difference between each tooth instance region and its adjacent tooth instance regions.

[0065] When analyzing periapical radiographs, each tooth region can be analyzed individually for each historical and current patient. Therefore, in embodiments of this invention, for each historical and current patient, the degree of projection overlap of each tooth region and the degree of difference between each tooth region and its adjacent tooth regions are analyzed. Wherein:

[0066] For each historical and current patient, the degree of projection overlap of each tooth instance area was analyzed, further including:

[0067] First, for each historical and current patient, the number of root apexes in each tooth instance region is identified using a tooth structure recognition model. Specifically, the tooth structure recognition model identifies the number of root apexes in each tooth instance region as 1, 2, or 3, etc. The more root apexes, the more complex the current tooth instance region is, and the more susceptible it is to the influence of overlapping imaging.

[0068] Then, projection overlap and angular deviation directly lead to a decrease in the continuous length of the edge breakage. Therefore, it is necessary to analyze the continuity of the bone structure edge in the root region of each tooth instance area. Specifically, edge detection is performed on the tooth instance area to obtain the bone structure edge pixels of the root region in the tooth instance area. The maximum length of the continuous bone edge in the root region of each tooth instance area is obtained. Existing technology will not be elaborated here. Based on historical patient data, the average value of the continuous bone edge in the root region of the tooth instance area at the same location is obtained as the theoretical continuous bone edge length of the root region in the tooth instance area at that location. The maximum length of the continuous bone edge is combined with the theoretical continuous bone edge length, that is, the ratio of the maximum length of the continuous bone edge to the theoretical continuous bone edge length is calculated to obtain the continuity of the bone structure edge in the root region of each tooth instance area. Then, the continuity of the bone structure edge in the root region of each tooth instance area is constructed. The first patient's The formula for calculating the continuity of the bone structure margin in the root region of a single tooth instance area is as follows:

[0069]

[0070] In the formula, Indicates the first The first patient's The degree of continuity of the bone structure margins in the root region of a single tooth instance area; Indicates the first The first patient's The length of the continuous bone margin in the root region of a single tooth instance area; This represents the function that takes the maximum value. Indicates the first The first patient's The maximum length of the continuous bone margin in the root region of a single tooth instance area; express The theoretical continuous bone margin length of the root region in a given tooth instance area.

[0071] The closer the ratio is to 1, the better its continuity. Using "proportion" instead of absolute length eliminates differences in tooth size, making it suitable for cross-patient statistics.

[0072] Furthermore, the bone structure edge pixels of the root region in each tooth instance area are extracted, the direction of each bone structure edge pixel is determined, and the angle between the direction and the main axis of the tooth root is obtained. Specifically, edge detection is performed on the tooth instance area to obtain the bone structure edge pixels of the root region in the tooth instance area. Based on the local gradient direction or edge tangent direction of the bone structure edge pixels, the direction of each bone structure edge pixel is determined, and then the angle between the direction of the bone structure edge pixel and the main axis of the tooth root is obtained.

[0073] Finally, based on the number of root apexes, the continuity of bone structure edges, and the included angle, the degree of projection overlap for each tooth instance region is obtained. Specifically, the first... The first patient's The formula for calculating the projection overlap of a tooth instance region is:

[0074]

[0075] In the formula, Indicates the first The first patient (historical or current patient) The degree of projection overlap of the tooth instance region; Indicates the first The first patient's The number of edge pixels of the bone structure in the root region of a tooth instance area; Indicates the first The first patient's The root region of the first tooth instance area The angle between the direction of each bone structure edge pixel and the main axis of the tooth root; This represents 180 degrees in radians. Indicates the first The first patient's Number of root apex regions in a single tooth instance; Indicates the first The first patient's The degree of continuity of the bone structure margins in the root region of a single tooth instance area; This represents a linear normalization function, such as a max-min normalization function, used to normalize the projected overlap values ​​to a certain value. Within the range, the maximum value in the maximum-minimum normalization function is the preset upper limit of the range, which is set according to the maximum value of the projection overlap that appears in the historical data; Represented by natural constant An exponential function with base 0.

[0076] The more consistent the direction of the edge pixels of the bone structure with the direction of the main axis of the tooth root, i.e., the angle... The smaller the value, the more likely the pixels at the edge of the bone structure are to have projection overlap, resulting in a higher imaging risk. Additionally, the shape of the teeth must be considered; the more roots a tooth has, the higher the risk of physical overlap. The continuity of the bone structure edges... The smaller the value, the worse the continuity of the bone structure edge in the root region of the tooth instance area, indicating that the pixels at the edge of this bone structure are more likely to have projection overlap, i.e., the higher the risk of overlap; ultimately, the degree of projection overlap... The smaller the value, the more concentrated, continuous, and stable the bone structure's edge structure is, and the less likely there is projection overlap or imaging interference; the degree of projection overlap... The larger the value, the more dispersed the direction of the bone structure edge, the poorer the continuity and the poorer the stability, and the greater the possibility of projection overlap or imaging interference.

[0077] For each historical and current patient, the degree of difference between each tooth instance region and its adjacent tooth instance regions was analyzed, further including:

[0078] First, for each historical and current patient, the height of the root region in each tooth instance area is obtained, and then the height difference between each tooth instance area and its adjacent teeth is analyzed. Specifically, the root principal axis of the root region in each tooth instance area is obtained by skeletalization processing and fitting or by directional statistics of the edge points of the root region. The length of the root principal axis is measured as the height of the root region in each tooth instance area. In addition, for the convenience of comparison and analysis, the length of the root principal axis is normalized here, such as by maximum-minimum normalization, to normalize the length of the root principal axis to... Within the range, the maximum value in the maximum-minimum normalization function is the preset upper limit of the range. This upper limit is set based on the maximum length of the root main axis that appeared in historical data, thus obtaining the normalized height. The absolute value of the height (normalized height) difference between each tooth instance region and its adjacent teeth (the two teeth on each side of the tooth instance region are its adjacent teeth) is calculated to obtain the height difference between each tooth instance region and its adjacent teeth.

[0079] Then, for each historical and current patient, the differences in the angles between each tooth instance region and its adjacent teeth are analyzed. Specifically, using the Dynamic Time Warping (DTW) algorithm, all bone structure edge pixels in the root region of each tooth instance region are paired with all bone structure edge pixels in the root regions of its adjacent tooth instance regions, resulting in several matched pixel pairs. The angles corresponding to all bone structure edge pixels in the root region of each tooth instance region are obtained (the angle between the direction corresponding to the bone structure edge pixel and the main axis of the tooth root). In addition, for ease of comparison and analysis, the angles are normalized here, such as by performing max-min normalization to normalize the angles to... Within the range, the maximum value in the maximum-minimum normalization function is the preset upper limit of the range. This upper limit is set based on the maximum angle of the included angle that appears in historical data, thus obtaining the normalized included angle. The absolute value of the difference between the included angles (normalized included angles) corresponding to the two bone structure edge pixels within the matched pixel pair is calculated. By traversing all matched pixel pairs, the mean of the absolute values ​​of the differences in included angles (normalized included angles) is calculated to obtain the angle difference between each tooth instance region and its adjacent teeth.

[0080] Based on the differences in height and angle, the degree of difference between each tooth instance region and its adjacent tooth instance regions is obtained. Specifically, the first... The first patient's The formula for calculating the degree of difference between a tooth instance region and its adjacent tooth instance regions is as follows:

[0081]

[0082] In the formula, Indicates the first The first patient (historical or current patient) The degree of difference between a tooth instance region and its adjacent tooth instance regions; Indicates the first The first patient's The number of adjacent tooth instance regions on the left and right sides of a tooth instance region (number on each side), in this embodiment The value of is 2; Indicates the first The first patient's The tooth instance area and the left or right side of the first The angle difference between adjacent tooth instance regions (mean of the absolute value of the normalized angle difference of all matching pixel pairs); Indicates the first The first patient's The tooth instance area and the left or right side of the first Height difference between adjacent tooth regions (normalized absolute height difference); and Indicates the weighting coefficient. and The value can be 0.4 or 0.6, used to balance the dimensional differences between angle and height.

[0083] When the difference in angle and height between a tooth instance region and its adjacent tooth instance region is large, the degree of difference between the two corresponding tooth instance regions is large.

[0084] S300: For multiple historical patients, based on the degree of projection overlap and difference, the imaging variability of the tooth instance area at the same location is analyzed among multiple historical patients to obtain the degree of imaging abnormality of the tooth instance area at each location.

[0085] Step S200 analyzed the degree of projection overlap and the degree of difference between adjacent tooth instance regions for each historical patient. However, this analysis of projection overlap and difference is performed on a single tooth instance region within the same historical patient, reflecting the differences in projection overlap and difference between images of different tooth positions within the same historical patient. Furthermore, the image appearance of areas with high projection overlap and difference tends to fluctuate significantly between different patients. Therefore, in addition to analyzing the projection overlap and difference, further analysis of the imaging variability of tooth instance regions at the same location across multiple historical patients is needed to make a dual judgment: different tooth instance regions may indicate imaging abnormalities rather than actual bone resorption.

[0086] Based on the above analysis, in the embodiments of the present invention, for multiple historical patients, the imaging variability of the tooth instance region at the same location is analyzed among multiple historical patients based on the degree of projection overlap and difference, to obtain the degree of imaging abnormality of the tooth instance region at each location. Further, this includes:

[0087] First, the aligned tooth instance regions corresponding to multiple historical patients are aligned across patients to obtain tooth instance regions corresponding to the same location for multiple historical patients. Specifically, using the FDI (Fédération Dentaire Internationale, World Dental Federation) tooth position marking method or anatomical tooth arrangement relative sequence number, cross-patient alignment matching is performed on tooth instance regions corresponding to the same tooth position number from different historical patients. That is, the individual tooth instance regions in the historical patient mapping space are aligned with the periapical radiographs of different historical patients according to the tooth position. It should be noted that the current cross-patient tooth instance region matching operation does not consider features such as tooth shape and tooth intervals, but only the positional relationship of the tooth instance regions in the spatial mapping.

[0088] Then, based on the degree of projection overlap, the variance of the projection overlap of tooth instance regions at the same location from multiple historical patients is calculated; and based on the degree of difference, the variance of the degree of difference of tooth instance regions at the same location from multiple historical patients is calculated; these are used to characterize the imaging instability of this location in the population. By statistically analyzing the variances of projection overlap and difference for historical patients, the larger the variance, the greater the imaging instability of the tooth data within the current tooth instance region due to the influence of its own structure, and the lower the reliability of subsequent results.

[0089] Finally, based on the variances of projection overlap and difference, and combining these variances, the degree of imaging anomaly in the tooth instance region at each location is obtained. Specifically, the... The tooth instance area at the location (corresponding to the first) The formula for calculating the degree of imaging abnormality in a single tooth instance region is as follows:

[0090]

[0091] In the formula, Indicates the first The tooth instance area at the location (corresponding to the first) The degree of imaging abnormalities in a single tooth instance region; This indicates that all historical patients were in the first... The tooth instance area at the location (corresponding to the first) Variance of projection overlap (for each tooth instance region); This indicates that all historical patients were in the first... The tooth instance area at the location (corresponding to the first) The mean degree of projection overlap of each tooth instance region; This indicates that all historical patients were in the first... The tooth instance area at the location (corresponding to the first) Variance of the degree of difference in (each tooth instance region); This indicates that all historical patients were in the first... The tooth instance area at the location (corresponding to the first) The mean degree of difference in (one tooth instance region); This represents a linear normalization function, such as a max-min normalization function, used to normalize imaging anomaly values ​​to a normal value. Within the range, the maximum value in the maximum-minimum normalization function is the preset upper limit of the range, which is set based on the maximum value of the imaging anomaly that occurred in the historical data.

[0092] S400: Based on the degree of imaging abnormality, set confidence weights and weight the degree of projection overlap and difference of the current patient respectively to obtain the imaging risk of each tooth instance area of ​​the current patient.

[0093] By constructing a standardized imaging interference distribution in the tooth space using periapical radiographs from multiple historical patients, this method reflects the long-term stability characteristics of different tooth instance regions in periapical radiographs. When analyzing the periapical radiographs of the current patient, this historical stability information is introduced as a scene prior. Combined with the grayscale, structural orientation, and overlap features in the current patient's images, the imaging reliability of different tooth instance regions is comprehensively evaluated, thereby avoiding the bias caused by relying solely on historical statistics or single image features.

[0094] Based on the above analysis, in the embodiments of the present invention, confidence weights are set according to the degree of imaging abnormality, and the degree of projection overlap and difference of the current patient are weighted respectively to obtain the imaging risk of each tooth instance area of ​​the current patient.

[0095] Based on the above operations, the projection overlap and difference values ​​for each tooth instance region of the current patient are calculated. These features are only applicable to analyzing whether there are specific anomalies in the tooth data within the current tooth instance region. Large values ​​within the current tooth instance region may be due to disease or the original tooth structure. Therefore, the imaging anomaly degree within the current tooth instance region is used as the weight for calculating the confidence score of the current tooth instance region based on historical data. Thus, a confidence score weight is set based on the imaging anomaly degree. Specifically, an imaging anomaly degree threshold is set, which can be 0.6; the imaging anomaly degree is compared to whether it exceeds the imaging anomaly degree threshold; if it does, i.e. Then set the confidence weight of imaging risk. The degree of imaging anomaly, i.e. If not, that is Then set the confidence weight of imaging risk. The value is 0. This means that if the current tooth instance area has historically been stable (with a small degree of imaging abnormality), even if the current patient has a slight abnormality, the confidence weight of the imaging risk will not be easily increased; if the current location has historically been a high-interference area, the confidence weight of the imaging risk will be significantly increased as long as similar signs appear now.

[0096] The imaging risk of each tooth instance region in the current patient is obtained by weighting the projection overlap and difference levels using confidence weights. Specifically, the adjusted projection overlap level is obtained by weighting the projection overlap level of each tooth instance region in the current patient using confidence weights; the adjusted difference level is obtained by weighting the difference level of each tooth instance region in the current patient using confidence weights; and the sum of the adjusted projection overlap level and the adjusted difference level is calculated to obtain the imaging risk of each tooth instance region in the current patient. Therefore, the imaging risk of the current patient's [number]th tooth instance region is calculated. The formula for calculating the imaging risk of a single tooth instance region is as follows:

[0097]

[0098] In the formula, Indicates the current patient's number Imaging risk level of a single tooth instance region; Indicates the current patient's number The degree of difference in the region of each tooth instance; Indicates the current patient's number The degree of projection overlap of the tooth instance region; The confidence weights represent the risk of imaging. This represents a linear normalization function, such as a max-min normalization function, used to normalize imaging risk values ​​to a certain value. Within the range, the maximum value in the maximum-minimum normalization function is the preset upper limit of the range, which is set based on the maximum value of imaging risk that appears in historical data.

[0099] In periapical radiographs, even if bone structural changes are detected, it is necessary to determine whether these changes are located in areas where imaging conditions are relatively reliable; otherwise, they are easily affected by projection or structural overlap. Imaging risk is incorporated into the detection results of bone resorption areas. The detection results for different areas are modulated and labeled. When the imaging risk corresponding to a certain area is high, such as exceeding a set risk threshold (which can be calibrated based on the error tolerance rate of clinical review or historical statistical data, for example, 0.5), that area is marked as an imaging risk area to indicate that its results are significantly affected by the imaging scenario.

[0100] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0101] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis, characterized in that, The method includes: Acquire periapical radiograph data sets corresponding to multiple historical patients and the current patient respectively, and spatially align the tooth instance regions in the periapical radiograph data sets corresponding to each historical patient and the current patient to obtain multiple aligned tooth instance regions corresponding to each historical patient and the current patient. For each historical patient and current patient, the degree of projection overlap of each tooth instance region is analyzed, and the degree of difference between each tooth instance region and its adjacent tooth instance regions is analyzed; For multiple historical patients, based on the degree of projection overlap and the degree of difference, the imaging variability of the tooth instance region at the same location among multiple historical patients is analyzed to obtain the degree of imaging abnormality of the tooth instance region at each location. Based on the degree of imaging abnormality, confidence weights are set, and the degree of projection overlap and the degree of difference of the current patient are weighted respectively to obtain the imaging risk of each tooth instance region of the current patient.

2. The method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis according to claim 1, characterized in that, Obtain apical radiographs of multiple historical and current patients, including: The hospital database is used to obtain sets of original periapical radiographs corresponding to multiple historical patients and current patients. The sets of original periapical radiographs include multiple original periapical radiographs. For each original periapical radiograph, grayscale normalization, noise reduction, size unification, and orientation correction are performed to obtain a set of periapical radiographs for multiple historical patients, each set of which includes multiple periapical radiographs.

3. The method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis according to claim 2, characterized in that, Spatial alignment of tooth instance regions in the periapical radiograph datasets corresponding to each historical and current patient is performed, including: For each of the aforementioned periapical radiographs, the tooth regions are identified and located to obtain all tooth instance regions in the periapical radiograph data set corresponding to each historical patient; For each tooth instance region, the root main axis and root apex position are extracted. Using the root main axis and root apex position as alignment references, the tooth instance regions in all periapical radiographs of the same historical patient and the current patient are spatially aligned.

4. The method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis according to claim 3, characterized in that, For each periapical radiograph, the tooth region is identified and located to obtain all tooth instance regions in the periapical radiograph data set corresponding to each historical patient, including: For each of the aforementioned periapical radiographs, the image region where the tooth is located is identified and marked as the tooth region using a tooth structure recognition model; Within the tooth region, different teeth are divided into instance-level regions, so that each tooth corresponds to an independent tooth instance region, thus obtaining all tooth instance regions in the periapical radiograph data set corresponding to each historical patient.

5. The method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis according to claim 1, characterized in that, For each historical patient and current patient, the degree of projection overlap of each tooth instance region is analyzed, including: For each historical and current patient, the number of root apex regions in each tooth instance is identified using a tooth structure recognition model; Analyze the continuity of the bone structure margins in the root region of each of the tooth instances; Extract the bone structure edge pixels of the root region in each tooth instance area, determine the direction of each bone structure edge pixel, and obtain the angle between the direction and the main axis of the tooth root. The degree of projection overlap of each tooth instance region is obtained based on the number of root apexes, the continuity of the bone structure edges, and the included angle.

6. The method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis according to claim 5, characterized in that, The analysis included the degree of continuity of the bone structure margins in the root region of each of the tooth instance areas, including: The maximum length of the continuous bone margin of the root region in each of the tooth instance regions is detected, and combined with the theoretical continuous bone margin length of the root region in each of the tooth instance regions, the degree of continuity of the bone structure margin of the root region in each of the tooth instance regions is obtained.

7. The method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis according to claim 5, characterized in that, For each historical and current patient, the degree of difference between each stated tooth instance region and its adjacent tooth instance regions is analyzed, including: For each historical patient and current patient, the height of the root region in each tooth instance area is obtained, and then the height difference between each tooth instance area and its adjacent teeth is analyzed; For each historical patient and current patient, the differences in the angle between each tooth instance region and its adjacent teeth are analyzed; Based on the height difference and the included angle difference, the degree of difference between each tooth instance region and its adjacent tooth instance regions is obtained.

8. The method for detecting bone resorption areas in a full-mouth periapical radiograph for identifying periodontitis according to claim 7, characterized in that, Analyze the differences in the angle between each tooth instance region and its adjacent teeth, including: Match all bone structure edge pixels in the root region of each tooth instance region with all bone structure edge pixels in the root region of its adjacent tooth instance regions to obtain several matching pixel pairs; Obtain the included angles corresponding to all bone structure edge pixels in the root region of each tooth instance region, calculate the absolute value of the difference between the included angles corresponding to the two bone structure edge pixels in the matching pixel pair, traverse all the matching pixel pairs, calculate the mean of the absolute values ​​of the included angle differences, and obtain the angle difference between each tooth instance region and its adjacent teeth.

9. The method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis according to claim 1, characterized in that, For multiple historical patients, based on the degree of projection overlap and the degree of difference, the imaging variability of the tooth instance region at the same location is analyzed among multiple historical patients to obtain the degree of imaging abnormality of the tooth instance region at each location, including: Align the aligned tooth instance regions corresponding to multiple historical patients across patients to obtain tooth instance regions corresponding to multiple historical patients at the same location. Based on the degree of projection overlap, the variance of the degree of projection overlap of tooth instance regions at the same location for multiple historical patients is calculated. Based on the degree of difference, the variance of the degree of difference in the tooth instance area at the same location for multiple historical patients is calculated. Based on the variance of the projection overlap and the variance of the difference, and combining the projection overlap and the difference, the imaging anomaly degree of the tooth instance region at each location is obtained.

10. The method for detecting bone resorption areas in a full-oral periapical radiograph for identifying periodontitis according to claim 1, characterized in that, Based on the degree of imaging anomaly, confidence weights are set, including: Set a threshold for the degree of imaging anomalies; Compare whether the degree of imaging anomaly is greater than the threshold for the degree of imaging anomaly; If so, then set the confidence weight to the degree of imaging anomaly; If not, set the confidence weight to 0.