Tooth preparation quality evaluation method and system based on image recognition

Abstract

The invention relates to the technical field of image analysis, in particular to a tooth preparation quality evaluation method and system based on image recognition. According to the method, the three-dimensional form of the tooth preparation area is subjected to point cloud reconstruction, so that the form characteristics of the tooth surface are accurately obtained, and the integrity and fineness of data are ensured. And in combination with light intensity reflection calculation of an illumination incident angle, the matching degree of the point cloud data is improved, so that the tooth surface information is more accurate. On the basis of curvature gradient analysis, the local change trend in the preparation area is identified, and the more sensitive capture capability for morphological details of different areas is ensured. The depth deviation distribution screening mechanism enhances the accuracy of the regional deviation, so that the morphological measurement is more stable. And an analysis model is established through spatial distribution of regional angle deviation in a normal vector calculation mode, so that the curved surface deviation can be compared according to a real anatomical form, and the evaluation of the tooth preparation quality better meets physiological requirements.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and particularly to a method and system for evaluating the quality of tooth preparation based on image recognition. Background Art

[0002] The technical field of image analysis includes technical methods for processing, analyzing, and interpreting image data. The core content of this technical field includes links such as image acquisition, preprocessing, feature extraction, pattern recognition, and classification, and involves methods such as computer vision, machine learning, and deep learning. Image analysis is widely used in multiple fields such as medical image analysis, automatic target recognition, quality inspection, and security monitoring. By analyzing the image information, functions such as target recognition, morphological measurement, and feature classification are realized.

[0003] Among them, the method for evaluating the quality of tooth preparation based on image recognition refers to a method for analyzing and evaluating the quality during the tooth preparation process using image recognition technology. This method covers specific technical means such as image acquisition of the tooth preparation area, image denoising, edge detection, three-dimensional reconstruction, morphological feature extraction, and standardized comparison. By quantitatively calculating feature parameters such as the morphology of tooth preparation, cutting depth, and edge smoothness, it is evaluated whether the tooth preparation meets the standard requirements.

[0004] The prior art has limitations in the accuracy of morphological analysis of the tooth preparation area. The image denoising and edge detection methods have insufficient processing capabilities for complex curved surface morphologies, which may lead to the loss of details of tooth morphological features, making the evaluation of the preparation quality in local areas inaccurate. During the three-dimensional reconstruction process, the acquisition method of point cloud data is limited by the illumination and projection matching of a single angle, affecting the integrity of the data and resulting in errors in the morphological reconstruction of some areas. The morphological feature extraction lacks more refined analysis of curvature and depth changes, making it difficult to accurately reflect the subtle differences in edge smoothness and cutting depth during the tooth preparation process, and affecting the reliability of the evaluation results. During the standardized comparison process, the morphological matching method relies on a fixed template and cannot be dynamically adjusted according to individual anatomical differences, resulting in low adaptability of the evaluation results. Overall, the prior art has limited accuracy in aspects such as data acquisition, morphological reconstruction, feature extraction, and quality evaluation, resulting in certain errors in the evaluation of the quality of tooth preparation and affecting the accurate assessment of the quality of clinical operations. Summary of the Invention

[0005] The purpose of the present invention is to solve the drawbacks existing in the prior art, and to propose a method and system for evaluating the quality of tooth preparation based on image recognition.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: A method for evaluating the quality of tooth preparation based on image recognition, including the following steps:

[0007] S1: Obtain an image of the buccal fossa preparation area of the mandibular first molar, reconstruct the tooth surface morphology through point cloud, extract the point cloud data after reconstruction, calculate the curvature gradient of the buccal fossa preparation area, and obtain the curvature gradient distribution information;

[0008] S2: Extract the adjacent point sets of the first region from the curvature gradient distribution information, compare the depth change trends of adjacent regions in each region, screen the second regions where the depth deviation exceeds the deviation threshold, and obtain the depth deviation distribution information;

[0009] S3: Extract the normal vectors of the second regions in the depth deviation distribution information, calculate the regions corresponding to the normal angle deviation, and perform constrained fitting of the local surface to obtain the surface deviation analysis result;

[0010] S4: Based on the surface deviation analysis result, calculate the entropy value features of the entire buccal fossa preparation area of the mandibular first molar and classify the regions with the same type of deviation, and calculate the deviation accumulation factors corresponding to each type of region after classification to obtain the regional deviation classification result;

[0011] S5: Based on the deviation accumulation factors corresponding to each type of region in the regional deviation classification result, calculate the overall quality evaluation value of the buccal fossa preparation area of the mandibular first molar to obtain the tooth preparation quality evaluation result.

[0012] As a further solution of the present invention, the curvature gradient distribution information includes local gradient values, gradient thresholds, and the boundaries of the first region. The depth deviation distribution information includes reference depths, depth deviations, and the boundaries of the second region. The normal angle deviation distribution map includes normal vectors, normal angle deviations, and various normal angle deviation regions. The surface deviation analysis result includes local surface constrained fitting results, surface deviation ranges, and surface deviation trends. The regional deviation classification result includes entropy value feature distributions, regions with the same type of deviation, and deviation accumulation factors. The tooth preparation quality evaluation result includes overall quality evaluation values, evaluation grading criteria, and quality grades for each type of region.

[0013] As a further solution of the present invention, the specific steps for obtaining the curvature gradient distribution information are as follows:

[0014] S111: Obtain a three-dimensional image of the buccal fossa preparation area of the mandibular first molar, collect the reflected images of blue light at multiple incident angles, calculate the light intensity difference based on the reflection data of the incident angle illumination, extract the point cloud projection relationship under multi-angle illumination, and use the coordinate matching method to fuse the multi-viewpoint cloud data to reconstruct the tooth surface morphology and generate the tooth surface point cloud data;

[0015] S112: Based on the tooth surface point cloud data, extract the point cloud data to calculate the principal curvature and average curvature of each point in the buccal fossa preparation area. According to the curvature change rate of adjacent points, use the formula:

[0016]

[0017] Calculate the local gradient distribution value G of the buccal fossa at point i i ;

[0018] where K i is the principal curvature at point i, K j is the principal curvature at the adjacent point j, H i is the mean curvature at point i, H j is the mean curvature at point j, N(i) is the set of neighborhood points of point i, d ij is the Euclidean distance between point i and point j, ∈ and δ are positive numbers to prevent division by zero;

[0019] S113: Call the local gradient distribution value of the buccal fossa, screen the first region where the local gradient exceeds the gradient threshold, analyze the change of the local gradient in the first region, and obtain the curvature gradient distribution information.

[0020] As a further solution of the present invention, the obtaining step of the depth deviation distribution information is specifically as follows:

[0021] S211: Based on the curvature gradient distribution information, extract the set of adjacent points in the first region, call the reference depth of the standard mandibular first molar morphology data, and use the formula:

[0022]

[0023] Calculate the depth deviation M1 of the first region, and obtain the depth deviation measurement result of the first region;

[0024] where Z1 is the reference depth of the first region, Q is the number of adjacent points in the first region, Z p is the reference depth of the point p in the set of adjacent points in the first region, is the average depth value of the set of adjacent points in the first region, and Q + 1 is the normalization factor;

[0025] S212: According to the depth deviation measurement result of the first region, compare the depth change trend of each region with that of the adjacent region, screen the second region where the depth deviation exceeds the deviation threshold, and obtain the depth deviation distribution information.

[0026] As a further solution of the present invention, the obtaining step of the surface deviation analysis result is specifically as follows:

[0027] S311: Based on the depth deviation distribution information, extract the normal vector of the second region, calculate the normal angle deviation between the normal vector of the second region and the normal of the standard mandibular first molar morphology, and establish a normal angle deviation distribution map according to the spatial distribution of the normal angle deviation values

[0028] S312: Based on the normal angle deviation distribution map, set an angle deviation threshold, divide into multiple normal angle deviation regions, and based on the spatial distribution relationship of the data points, adopt a local surface constraint fitting method to obtain the local surface deviation value of each region, integrate the deviation information of all regions, and generate a surface deviation analysis result.

[0029] As a further solution of the present invention, the specific steps for obtaining the regional deviation classification result are as follows:

[0030] S411: Based on the surface deviation analysis result, extract the deviation information of all the buccal fossa preparation regions of the first mandibular molar, and use the formula:

[0031]

[0032] Calculate the entropy value feature A of the corresponding region, integrate the entropy value information of all data points, and establish an entropy value feature distribution;

[0033] Among them, P r represents the proportion of data points in the rth deviation interval, D t represents the deviation value of the tth data point, D avg represents the average deviation of all data points in the corresponding region, v represents the total number of data points, and s represents the total number of deviation intervals;

[0034] S412: Based on the entropy value feature distribution, classify according to the entropy value range of the data points, determine the same-class deviation regions, calculate the deviation accumulation factor of the corresponding same-class deviation regions, the deviation accumulation factor represents the overall accumulation of deviations in this region, integrate the accumulation factors of all classified regions, and obtain the regional deviation classification result.

[0035] As a further solution of the present invention, the specific steps for obtaining the tooth preparation quality evaluation result are as follows:

[0036] S511: Based on the deviation accumulation factor corresponding to each type of region in the regional deviation classification result, calculate the overall quality evaluation value of the buccal fossa preparation region of the first mandibular molar, extract the deviation accumulation factors of all classified regions, and establish a quality evaluation value;

[0037] S512: According to the set evaluation criteria, referring to the quality evaluation value, divide the grade interval of the overall quality of the tooth preparation region, integrate the grading information of all evaluation regions, and obtain the tooth preparation quality evaluation result.

[0038] A tooth preparation quality evaluation system based on image recognition, according to the tooth preparation quality evaluation method based on image recognition described above, the system includes:

[0039] The three-dimensional reconstruction and curvature analysis module acquires images of the buccal fossa preparation area of the first mandibular molar, reconstructs the tooth surface morphology through point cloud reconstruction, extracts the point cloud data after reconstruction to calculate the curvature gradient of the buccal fossa preparation area, and obtains the curvature gradient distribution information;

[0040] The depth deviation detection module extracts the adjacent point sets of the first region from the curvature gradient distribution information, compares the depth change trends of adjacent regions in each region, screens the second regions with depth deviations exceeding the deviation threshold, and obtains the depth deviation distribution information;

[0041] The normal analysis and surface fitting module extracts the normal vectors of the second regions in the depth deviation distribution information, calculates the regions corresponding to the normal angle deviations, and performs constrained fitting of the local surfaces to obtain the surface deviation analysis results;

[0042] The deviation feature extraction and classification module, based on the surface deviation analysis results, calculates the entropy value features of all the buccal fossa preparation areas of the first mandibular molar and classifies the same type of deviation regions, and calculates the deviation accumulation factors corresponding to each type of region after classification to obtain the regional deviation classification results;

[0043] The tooth preparation quality evaluation module calculates the overall quality evaluation value of the buccal fossa preparation area of the first mandibular molar based on the deviation accumulation factors corresponding to each type of region in the regional deviation classification results, and obtains the tooth preparation quality evaluation results.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0045] In the present invention, through point cloud reconstruction of the three-dimensional morphology of the tooth preparation area, the morphological features of the tooth surface are accurately obtained, ensuring the integrity and fineness of the data. Combining the calculation of the light intensity reflection of the light incident angle improves the matching degree of the point cloud data, making the tooth surface information more accurate. Based on the curvature gradient analysis, the local change trends in the preparation area are identified, ensuring a more sensitive capture ability for the morphological details of different regions. The screening mechanism of the depth deviation distribution enhances the accuracy of the regional deviation, making the morphological measurement more stable. The calculation method of the normal vector establishes an analysis model through the spatial distribution of the regional angle deviations, ensuring that the surface deviation can be compared according to the real anatomical morphology, making the evaluation of the tooth preparation quality more in line with the physiological requirements. Through the calculation of entropy value features, the quantification of data differences is realized, and the deviation accumulation analysis of the classified regions can accurately measure the processing consistency of different regions, improving the scientific nature of the quality evaluation. Finally, through the integration of multi-level deviation information, the comprehensive evaluation of the tooth preparation quality is realized, and the overall quality is graded according to the quality standard, which helps to improve the accuracy of the tooth preparation operation and provides more valuable data support for clinical applications. Brief Description of the Drawings

[0046] Figure 1 Schematic diagram of the main steps of the present invention;

[0047] Figure 2 Flow chart of step S1 of the present invention;

[0048] Figure 3 Flow chart of step S2 of the present invention;

[0049] Figure 4 Flow chart of step S3 of the present invention;

[0050] Figure 5 Flow chart of step S4 of the present invention;

[0051] Figure 6 Flow chart of step S5 of the present invention. Detailed implementation manners

[0052] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0053] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more unless otherwise specifically defined.

[0054] Please refer to Figure 1 , the present invention provides a technical solution: a method for evaluating the quality of tooth preparation based on image recognition, including the following steps:

[0055] S1: Obtain the three-dimensional image of the buccal fossa preparation area of the first mandibular molar and the reflection images at multiple incident angles of blue light projection, reconstruct the tooth surface morphology by combining point clouds, extract the point cloud data to calculate the principal curvature and average curvature of each point in the buccal fossa preparation area, calculate the local gradient value based on the curvature change rate of adjacent points, screen the first area where the local gradient exceeds the gradient threshold, and obtain the curvature gradient distribution information;

[0056] S2: Extract the adjacent point set of the first region from the curvature gradient distribution information, call the standard mandibular first molar morphology data to calculate the reference depth of this region, calculate the depth deviation of each region according to the reference depth, compare the depth change trends of adjacent regions of each region, and screen the second regions with depth deviations exceeding the deviation threshold to obtain the depth deviation distribution information;

[0057] S3: Extract the normal vectors of the second regions in the depth deviation distribution information, calculate the normal angle deviation between the normal vectors of the second regions and the standard mandibular first molar morphology, establish a normal angle deviation distribution map, divide the regions with various normal angle deviations in the normal angle deviation distribution map and perform constrained fitting of the local surfaces to obtain the surface deviation analysis results;

[0058] S4: Based on the surface deviation analysis results, calculate the entropy value features of all the buccal fossa preparation regions of the mandibular first molars, classify the regions with the same type of deviation according to the entropy value feature distribution, and calculate the deviation accumulation factors corresponding to each type of region after classification to obtain the regional deviation classification results;

[0059] S5: Based on the deviation accumulation factors corresponding to each type of region in the regional deviation classification results, calculate the overall quality assessment value of the buccal fossa preparation regions of the mandibular first molars, set the grading criteria according to the evaluation value to obtain the tooth preparation quality assessment results;

[0060] The curvature gradient distribution information includes local gradient values, gradient thresholds, and the boundaries of the first regions. The depth deviation distribution information includes reference depths, depth deviations, and the boundaries of the second regions. The normal angle deviation distribution map includes normal vectors, normal angle deviations, and regions with various normal angle deviations. The surface deviation analysis results include local surface constrained fitting results, surface deviation ranges, and surface deviation trends. The regional deviation classification results include entropy value feature distributions, regions with the same type of deviation, and deviation accumulation factors. The tooth preparation quality assessment results include overall quality assessment values, evaluation grading criteria, and quality grades of various regions.

[0061] Please refer to Figure 2 , and the specific steps for obtaining the curvature gradient distribution information are as follows:

[0062] S111: Obtain the three-dimensional image of the buccal fossa preparation region of the mandibular first molar, collect the reflection images of the blue light projected at multiple incident angles, calculate the light intensity difference based on the reflection data of the incident angle illumination, extract the point cloud projection relationship under multi-angle illumination, and use the coordinate matching method to fuse the multi-view point cloud data to reconstruct the tooth surface morphology and generate the tooth surface point cloud data;

[0063] To obtain the three-dimensional image of the buccal fossa preparation area of the mandibular first molar, a high-precision scanning device is required. This device uses the principle of structured light scanning, projects blue light onto the tooth surface, and obtains high-resolution point cloud data by using the light reflection information at multiple angles. In practical applications, a blue light three-dimensional scanner can be used to continuously scan the tooth surface to obtain complete morphological data. For example, for a buccal fossa area with a size of about 10mm×10mm, a scanning method with a resolution of 0.01mm can be used to obtain high-precision point cloud data. Each point contains three-dimensional coordinates (X, Y, Z) and reflected light intensity information. Preprocess these point cloud data, including noise removal and outlier rejection, to ensure the reconstruction quality. In actual operation, if there are a large number of high-frequency noises in the point cloud data (for example, surface mutation points caused by measurement errors), the denoising process can be carried out through the neighborhood median filtering method. After that, in order to enhance the integrity of the point cloud data, it is necessary to match the light reflection images at different incident angles. By matching the projection data at different angles, the reconstruction accuracy can be improved. For example, through comparative analysis of the point cloud data obtained from a single scanning perspective, if there are incomplete areas, the multi-view stitching method is used to supplement the data. The point cloud data stitching process involves coordinate transformation and optimal alignment. Usually, the iterative closest point (ICP) method is used to register the point clouds from different perspectives. Through rotation and translation transformations, the point cloud error in the overlapping area is minimized. For example, for point cloud data with an initial error of 0.2mm, after ICP alignment optimization, the error can be reduced to less than 0.05mm. After completing the fusion of the point cloud data, the Delaunay triangulation method is used for meshing to form complete tooth surface morphological data, and finally, the tooth surface point cloud data is generated.

[0064] S112: Based on the tooth surface point cloud data, extract the point cloud data and calculate the principal curvature and mean curvature of each point in the buccal fossa preparation area. According to the curvature change rate of adjacent points, use the formula:

[0065]

[0066] Calculate the local gradient distribution value G of the buccal fossa at point i i ;

[0067] where, K i is the principal curvature at point i, K j is the principal curvature at the adjacent point j, H i is the mean curvature at point i, H j is the mean curvature at point j, N(i) is the set of neighborhood points of point i, d ij is the Euclidean distance between point i and point j, ∈ and δ are small positive numbers to prevent division by zero, usually set to 10 -6 to ensure calculation stability, ∑ j∈N(i)denotes the summation operation over all points j in the neighborhood point set N(i) of point i, and max(H j ,H i ) denotes taking the maximum value of the mean curvatures at points i and j.

[0068] Given the neighborhood point set N(i) of a certain point P i , assuming it contains 4 neighborhood points (i.e., j = 1, 2, 3, 4), the principal curvature K j , mean curvature H j and its Euclidean distance d i from P ij are as follows:

[0069] Set up Data Table 1 (Neighborhood Point Data)

[0070] Neighboring point j <![CDATA[K j (Principal curvature)]]> <![CDATA[H j (Mean curvature)]]> <![CDATA[d ij (mm)]]> 1 0.14 0.10 0.25 2 0.12 0.11 0.30 3 0.16 0.09 0.20 4 0.11 0.08 0.35

[0071] If K i = 0.15 (the principal curvature at the current point P i ), H i = 0.12 (the mean curvature at the current point P i ), set small positive numbers to prevent division by zero: ∈ = 10 -6 , δ = 10 -6 .

[0072] Calculate the term for each neighborhood point. For

[0073] For

[0074] For

[0075] For

[0076] Perform the summation calculation:

[0077] Calculate the square root

[0078] The calculation result G i = 0.0443 represents the local gradient value of this point. In actual operation, it can be used to characterize the curvature change situation in this area and for subsequent morphological analysis. If this value is higher than the set threshold G th , it indicates that the curvature change at this point is drastic and may belong to an area with obvious structural features.

[0079] The advantage of the formula is that by jointly calculating the principal curvature and the mean curvature, the rate of change of the local geometric shape is considered, thereby providing a more accurate local curvature change rate for more accurately identifying morphological features in subsequent analyses.

[0080] S113: Call the local gradient distribution value of the buccal fossa, screen the first region where the local gradient exceeds the gradient threshold, analyze the local gradient change of the first region, and obtain the curvature gradient distribution information;

[0081] Call the local gradient distribution value of the buccal fossa, set a gradient threshold to screen out the regions with relatively drastic curvature changes, and set the threshold G th Determined based on statistical analysis, its calculation method can be set by the mean weighted standard deviation method: G th = μ G + λσ G , where: G th : Local gradient screening threshold; μ G : Average gradient value of all points; σ G : Standard deviation of the gradient value; λ: Adjustment coefficient, usually with a value range of [1.5, 2.5].

[0082] Suppose the average gradient value calculated from the point cloud data of a certain region is μ G = 0.05, and the standard deviation σ G =

[0083] 0.01. When λ = 2, the threshold is calculated as follows: G th = 0.05 + 2×0.01 = 0.07. For each point G i , if G i > G th , then this point belongs to the region with drastic curvature changes. Finally, the curvature gradient distribution information is obtained.

[0084] This result shows that the high-gradient regions screened through statistical analysis can accurately reflect the local change characteristics of the tooth morphology and can be used for further surface reconstruction or morphological analysis.

[0085] Please refer to Figure 3 , and the specific steps for obtaining the depth deviation distribution information are as follows:

[0086] S211: Based on the curvature gradient distribution information, extract the adjacent point set of the first region, call the reference depth of the standard mandibular first molar morphology data, and use the formula:

[0087]

[0088] Calculate the depth deviation M1 of the first region to obtain the depth deviation measurement result of the first region;

[0089] Among them, Z1 is the reference depth of the first region, Q is the number of adjacent points in the first region, and Z p is the reference depth of the point p in the set of adjacent points in the first region, represents the summation operation on all adjacent points in the first region, (Z p - Z1) 2 is the squared deviation of the adjacent point depth from the reference depth, which is used to measure the local mean square error, is the average depth value of the set of adjacent points in the first region, is the third-order deviation cumulant of the adjacent point depth relative to its average depth value, which is used to measure the skewness effect. Q + 1 is the normalization factor to avoid the influence of small sample errors in skewness calculation.

[0090] First, it is necessary to determine the position of the first region from the curvature gradient distribution dataset. Assuming the center point coordinates of this region are (x1, y1, z1), then, according to the given adjacent point determination rule, all the points that meet the adjacency condition around this region are screened out. Usually, the screening of adjacent points can be based on the Euclidean distance or Voronoi diagram division. For example, if the adjacency radius is set to 1 mm, then all the points that meet the following conditions are classified as adjacent points in the first region: Among them, (x p , y p , z p ) represents the coordinates of a certain point in the curvature gradient distribution, and x1, y1, z1 are the three-dimensional coordinates of the center point of the first region, representing the spatial position of this region. Next, call the morphological database of the standard mandibular first molar, extract the reference depth of this region from the database, and set this reference depth as Z1. By calculating the average depth of its adjacent points as the comparison benchmark, the reference depth is calculated as follows: Among them, Z p is the depth value of the adjacent point, and Q is the number of adjacent points in this region. For example, assuming the depths of five adjacent points in a certain region are 2.5 mm, 2.7 mm, 2.6 mm, 2.8 mm, and 2.4 mm respectively, then its average reference depth is:

[0091]

[0092] Then calculate the depth deviation of the first region, using the following formula:

[0093]

[0094] Assuming Z1 = 2.6 mm, the depth deviation is calculated as follows:

[0095]

[0096] Obtain the depth deviation metric of the first region.

[0097] S212: According to the depth deviation measurement results of the first region, compare the depth change trends of each region with its adjacent regions, screen the second regions where the depth deviation exceeds the deviation threshold, and obtain the depth deviation distribution information;

[0098] First, obtain the depth change of the first region and calculate its depth change trend T1:

[0099]

[0100] where M 1,prev is the depth deviation value of the first region at the previous time step. Assume the depth deviation value at the previous time step is 0.018, then:

[0101]

[0102] Then, compare the depth change trends of adjacent regions. Assume the depth deviation calculation method of the second region is the same, and obtain its trend T2. If the change amount of T2 is much higher than that of the first region, and its depth deviation value M2 exceeds the set deviation threshold M thresh , then this region is determined as a region with abnormal deviation. Assume the deviation threshold is 0.025. If it is calculated that M2 = 0.03, then: M2 > M thresh , so this region is screened as the second region where the depth deviation exceeds the deviation threshold, and finally the depth deviation distribution information is obtained.

[0103] The advantage of this formula is that by combining the square term and the third-order deviation term, it can measure both the mean square error change and the skewness degree of the local region, so as to more accurately evaluate local anomalies while ensuring overall stability and avoid misjudgments that may be caused by relying solely on the mean square error. In addition, the normalization factor Q + 1 is used to avoid the influence of small sample data and improve the result stability.

[0104] Please refer to Figure 4 , and the specific steps for obtaining the surface deviation analysis results are as follows:

[0105] S311: Based on the depth deviation distribution information, extract the normal vectors of the second regions, calculate the normal angle deviation between the normal vectors of the second regions and the normal of the standard mandibular first molar morphology, and establish a normal angle deviation distribution map according to the spatial distribution of the normal angle deviation values;

[0106] Extract the normal vectors of multiple data points within the second region. The normal vector of each data point represents the local surface direction of that point in three-dimensional space. For example, in oral scan data, each molar surface point corresponds to a normal vector. Suppose the normal vector of point A is (0.2, 0.8, -0.6), and the normal vector of point B is (0.3, 0.7, -0.5). It is necessary to further calculate these normal vectors of the data points to obtain their deviation angles. Calculate the angles between all the normal vectors in the second region and the normal vectors of the standard mandibular first molar morphology. The normal vectors of the standard mandibular first molar morphology usually come from the ideal morphology model in the anatomical database. Suppose the standard normal vector is (0.1, 0.9, -0.4), then the deviation angles of each data point can be obtained by calculating the included angles of the normal vectors.

[0107] Suppose the calculation method for point A is as follows. The included angle calculation uses the vector dot product method, and the calculation results are as follows:

[0108]

[0109] Among them, represents the normal vector of the l-th data point, represents the standard normal vector, and are the magnitudes of the corresponding vectors respectively. If calculating the normal angle deviation of point A:

[0110]

[0111] Similarly, the normal angle deviations of other data points can be calculated. The normal angle deviation values of all data points will be used to construct a normal angle deviation distribution map. The drawing of the distribution map is based on a three-dimensional coordinate system, where the X and Y axes represent the spatial positions on the tooth surface, and the Z axis represents the normal angle deviation value. Suppose some calculated normal angle deviations of data points are listed:

[0112] Table 2 Data table of normal angle deviations

[0113]

[0114] As shown in Table 2, after calculating the normal angle deviation values of each data point, they can be used to draw a normal angle deviation distribution map. The distribution map can visually display the normal angle deviation conditions of each region on the tooth surface, and finally obtain a normal angle deviation distribution map.

[0115] S312: Based on the normal angle deviation distribution map, set an angle deviation threshold, divide multiple normal angle deviation regions, and based on the spatial distribution relationship of the data points, adopt a local surface constraint fitting method to obtain the local surface deviation values of each region, integrate the deviation information of all regions, and generate a surface deviation analysis result;

[0116] Call up the normal angle deviation distribution map, set the angle deviation threshold value to judge the deviation degree of different regions. The set threshold value ranges are 5°, 10°, and 15°, corresponding to the low deviation, medium deviation, and high deviation regions respectively. For example, in Table 1, the deviation value of point C is 7.1°, belonging to the medium deviation region; the deviation value of point B is 9.8°, also belonging to the medium deviation region; while the deviation value of point A is 12.5°, belonging to the high deviation region. Classify all data points according to the deviation region, determine the boundaries of each region, and perform local surface constraint fitting on the data points after region division.

[0117] During the local surface constraint fitting process, the deviation calculation method for each region adopts the average deviation value of the data points to represent the overall deviation of the region. For example, for a certain region containing five data points, their deviation values are 6.3°, 7.8°, 8.5°, 10.2°, and 9.1° respectively, then calculate their average deviation value: By calculating the local deviation values of all regions, the trend of the overall surface deviation can be obtained. If the deviation values of a certain region are all higher than the set deviation threshold, it indicates that there is a large deviation in the shape of this region; if the region deviation value is small, it indicates that the tooth surface is relatively uniform. Finally, calculate the surface deviation of all regions, integrate the deviation information, and obtain the surface deviation analysis result.

[0118] Please refer to Figure 5 , and the specific steps for obtaining the regional deviation classification result are as follows:

[0119] S411: Based on the surface deviation analysis result, extract the deviation information of all buccal fossa preparation regions of the first mandibular molar, and use the formula:

[0120]

[0121] Calculate the entropy value feature A of the corresponding region, which is a comprehensive index used to describe the deviation characteristics of a certain region. The larger the value, the more complex and discrete the deviation distribution of this region. Integrate the entropy value information of all data points to establish the entropy value feature distribution;

[0122] Among them, P r represents the proportion of data points in the r-th deviation interval, D t represents the deviation value of the t-th data point, D avg represents the average deviation of all data points in this region, v represents the total number of data points, and s represents the total number of deviation intervals.

[0123] Extract the deviation information of the entire buccal fossa preparation area of the mandibular first molar. The deviation value of each data point represents the degree of spatial offset of that point relative to the standard tooth morphology. For example, if the deviation value of a certain data point is 0.15 mm, it means that this point is offset 0.15 mm outward from the standard tooth surface, while the deviation value of another data point is -0.08 mm, indicating that it is offset 0.08 mm inward from the standard surface. The deviation information of all data points constitutes the deviation information set of this area. Calculate these data points to obtain their entropy value characteristics, and the entropy value characteristics are used to measure the dispersion degree of the deviation data in this area. First, discretize the deviation data in this area, set the deviation interval. For example, divide the deviation value range into several intervals, and set the interval range as [0, 0.1] mm, [0.1, 0.2] mm, [0.2, 0.3] mm, etc. Calculate the probability of the deviation value of each data point according to the interval it belongs to. For example, if a certain area contains 10 data points, among which 3 data points' deviation values fall into the [0, 0.1] mm interval, 4 data points' deviation values fall into the [0.1, 0.2] mm interval, and 3 data points' deviation values fall into the [0.2, 0.3] mm interval, then calculate the proportion of data points in each interval. The probability calculation is as follows:

[0124] Table 3 Deviation data distribution table

[0125] Deviation interval (mm) Number of data points in this interval <![CDATA[The probability P of data points in this interval r > [0，0.1] 3 3 / 10＝0.3 [0.1，0.2] 4 4 / 10＝0.4 [0.2，0.3] 3 3 / 10＝0.3

[0126] Calculate the entropy value characteristics of this area according to the proportion of data points in each interval, using the formula:

[0127]

[0128] The calculated entropy value characteristic A = 0.57 represents the distribution of deviation data in this area.

[0129] S412: Based on the entropy value characteristic distribution, classify according to the entropy value range of the data points to determine the same-type deviation areas, calculate the deviation cumulative factor of the corresponding same-type deviation areas, and the deviation cumulative factor represents the overall cumulative situation of the deviation in this area. Integrate the cumulative factors of all classified areas to obtain the regional deviation classification result;

[0130] Call the entropy value feature distribution, classify according to the entropy value range of data points, determine the same-class deviation area, and set the entropy value classification thresholds as low entropy (A < 0.3), medium entropy (0.3 ≤ A < 0.6), and high entropy (A ≥ 0.6). For example, if the entropy value of a certain area is 0.57, it belongs to the medium entropy area, while if the entropy value of another area is 0.72, it belongs to the high entropy area. All data points are divided according to the entropy value range to determine the same-class deviation area. For each type of area, calculate the corresponding deviation cumulative factor. The deviation cumulative factor represents the total cumulative situation of deviations within this area. The calculation method is the sum of the deviations of all data points multiplied by the reciprocal of the number of data points. For example, within a certain area, the deviation values of data points are 0.12mm, 0.08mm, 0.14mm, 0.18mm, and 0.09mm respectively. The total deviation value is: 0.12 + 0.08 + 0.14 + 0.18 + 0.09 = 0.61. Calculate the deviation cumulative factor F of this area: F = 0.61 × 5 -1 = 0.122. The deviation cumulative factor F = 0.122 represents the deviation cumulative degree of this area. Integrate the cumulative factors of all classification areas to obtain the area deviation classification result.

[0131] Please refer to Figure 6 , and the specific steps for obtaining the evaluation result of the tooth preparation quality are as follows:

[0132] S511: Based on the deviation cumulative factors corresponding to each type of area in the area deviation classification result, calculate the overall quality evaluation value of the buccal fossa preparation area of the first mandibular molar, extract the deviation cumulative factors of all classification areas, and establish the quality evaluation value;

[0133] First, determine the deviation cumulative factor data of all classification areas. For example, a buccal fossa preparation area of the first mandibular molar contains five classification areas, and their deviation cumulative factors are 0.12, 0.15, 0.08, 0.10, and 0.09 respectively. These values are obtained from the area deviation characteristics calculated in the previous step. When calculating the overall quality evaluation value, directly sum all the deviation cumulative factors. The calculation formula is as follows: G = F1 + F2 + F3 + F4 + F5, where G represents the overall quality evaluation value, and F1, F2, F3, F4, F5 represent the deviation cumulative factors of each classification area. Substitute the values: G = 0.12 + 0.15 + 0.08 + 0.10 + 0.09 = 0.54. The calculated quality evaluation value is 0.54, and this value represents the overall deviation situation of this tooth preparation area, and is further used for subsequent quality evaluation grading, as shown in Table 1.

[0134] Table 4 Data table of deviation cumulative factors of classification areas

[0135] Region number Deviation accumulation factor 1 0.12 2 0.15 3 0.08 4 0.10 5 0.09

[0136] Table 4 lists the deviation cumulative factor data for each classification area of the tooth preparation area, with a total of 0.54. Integrate the calculation results to obtain the quality assessment value.

[0137] S512: According to the set evaluation criteria, refer to the quality assessment value, divide the grade interval of the overall quality of the tooth preparation area, integrate the grading information of all evaluation areas, and obtain the tooth preparation quality assessment result;

[0138] Call the quality assessment value, according to the set evaluation criteria, divide the grade interval of the overall quality assessment value of the tooth preparation area, set the grading threshold, classify the evaluation value, and obtain the corresponding quality grading for each area. The specific evaluation criteria are set as follows. The evaluation grade interval is divided into three categories: high quality (Q ≤ 0.3), qualified quality (0.3 < Q ≤ 0.6), unqualified quality (Q > 0.6). For example, for the calculated quality assessment value Q = 0.54, this value falls into the interval of 0.3 < Q ≤ 0.6, so it is classified as the qualified quality grade, as shown in Table 5.

[0139] Table 5 Tooth Preparation Quality Assessment Grading Criteria

[0140] Quality assessment value Q Quality grade Q≤0.3 High quality 0.3<Q≤0.6 Qualified quality Q>0.6 Unqualified quality

[0141] Table 5 lists the criteria for quality grading based on the quality assessment value Q. According to the calculated Q = 0.54, the grading result corresponding to this value is qualified quality. Integrate the grading information of all evaluation areas to finally obtain the tooth preparation quality assessment result.

[0142] The tooth preparation quality assessment system based on image recognition, according to the tooth preparation quality assessment method based on image recognition, the system includes:

[0143] The three-dimensional reconstruction and curvature analysis module obtains the image of the buccal fossa preparation area of the first mandibular molar, reconstructs the tooth surface morphology through point cloud, extracts the point cloud data after reconstruction to calculate the curvature gradient of the buccal fossa preparation area, and obtains the curvature gradient distribution information;

[0144] The depth deviation detection module extracts the adjacent point set of the first area from the curvature gradient distribution information, compares the depth change trend of adjacent areas in each area, and screens the second area with a depth deviation exceeding the deviation threshold to obtain the depth deviation distribution information;

[0145] The normal analysis and surface fitting module extracts the normal vector of the second area in the depth deviation distribution information, calculates the area corresponding to the normal angle deviation, and performs constrained fitting of the local surface to obtain the surface deviation analysis result;

[0146] Based on the results of the surface deviation analysis, the deviation feature extraction and classification module calculates the entropy value features of all the buccal fossa preparation areas of the mandibular first molar and classifies the deviation areas of the same type, and calculates the deviation accumulation factor corresponding to each type of area after classification to obtain the regional deviation classification result;

[0147] Based on the deviation accumulation factor corresponding to each type of area in the regional deviation classification result, the tooth preparation quality assessment module calculates the overall quality assessment value of the buccal fossa preparation area of the mandibular first molar to obtain the tooth preparation quality assessment result.

[0148] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still belong to the protection scope of the technical solution of the present invention.

Claims

1. A method for evaluating the quality of tooth preparation based on image recognition, characterized in that, It includes the following steps: S1: Obtain an image of the buccal fossa preparation area of the mandibular first molar, reconstruct the tooth surface morphology through point cloud, extract the point cloud data after reconstruction, calculate the curvature gradient of the buccal fossa preparation area, and obtain the curvature gradient distribution information; S2: Extract the adjacent point sets of the first region from the curvature gradient distribution information, compare the depth change trends of adjacent regions in each region, screen the second regions with depth deviations exceeding the deviation threshold, and obtain the depth deviation distribution information; S3: Extract the normal vectors of the second regions in the depth deviation distribution information, calculate the regions corresponding to the normal angle deviations, and perform constrained fitting of the local surface to obtain the surface deviation analysis result; S4: Based on the surface deviation analysis result, calculate the entropy value features of all the buccal fossa preparation areas of the mandibular first molar and classify the regions with the same type of deviation, and calculate the deviation accumulation factor corresponding to each type of region after classification to obtain the region deviation classification result; S5: Based on the deviation accumulation factors corresponding to each type of region in the region deviation classification result, calculate the overall quality evaluation value of the buccal fossa preparation area of the mandibular first molar to obtain the tooth preparation quality evaluation result.

2. The method for evaluating the quality of tooth preparation based on image recognition according to claim 1, wherein The curvature gradient distribution information includes local gradient values, gradient thresholds, and the boundaries of the first region. The depth deviation distribution information includes reference depths, depth deviations, and the boundaries of the second region. The normal angle deviation distribution map includes normal vectors, normal angle deviations, and multiple normal angle deviation regions. The surface deviation analysis result includes local surface constrained fitting results, surface deviation ranges, and surface deviation trends. The region deviation classification result includes entropy value feature distributions, regions with the same type of deviation, and deviation accumulation factors. The tooth preparation quality evaluation result includes overall quality evaluation values, evaluation grading criteria, and quality grades for each type of region.

3. The method for evaluating the quality of tooth preparation based on image recognition according to claim 1, wherein The specific steps for obtaining the curvature gradient distribution information are as follows: S111: Obtain a three-dimensional image of the buccal fossa preparation area of the mandibular first molar, collect the reflected images of blue light at multiple incident angles, calculate the light intensity difference based on the reflected data of the incident angle illumination, extract the point cloud projection relationship under multi-angle illumination, and use the coordinate matching method to fuse the multi-viewpoint cloud data to reconstruct the tooth surface morphology and generate the tooth surface point cloud data; S112: Based on the tooth surface point cloud data, extract the point cloud data to calculate the principal curvature and average curvature of each point in the buccal fossa preparation area. According to the curvature change rate of adjacent points, use the formula: Calculate the local gradient distribution value G of the buccal fossa at point i i ; Among them, K i is the principal curvature at point i, and K j is the principal curvature at the adjacent point j, and H i is the mean curvature at point i, and H j is the mean curvature at point j. N(i) is the set of neighborhood points of point i, and d ij is the Euclidean distance between point i and point j. ∈ and δ are positive numbers to prevent division by zero; S113: Call the local gradient distribution value of the buccal fossa, screen the first regions with local gradients exceeding the gradient threshold, and analyze the local gradient changes in the first regions to obtain the curvature gradient distribution information.

4. The method for evaluating the quality of tooth preparation based on image recognition according to claim 1, wherein The specific steps for obtaining the depth deviation distribution information are as follows: S211: Based on the curvature gradient distribution information, extract the adjacent point sets of the first region, call the reference depth of the standard mandibular first molar morphology data, and use the formula: Calculate the depth deviation M1 of the first region to obtain the depth deviation measurement result of the first region; where Z1 is the reference depth of the first region, Q is the number of adjacent points in the first region, and Z p is the reference depth of the point p in the adjacent point set of the first region, is the average depth value of the adjacent point set of the first region, and Q + 1 is the normalization factor; S212: According to the depth deviation measurement results of the first region, compare the depth change trends of each region with those of adjacent regions, screen out the second regions with depth deviations exceeding the deviation threshold, and obtain the depth deviation distribution information.

5. The method for evaluating the quality of tooth preparation based on image recognition according to claim 1, wherein The specific steps for obtaining the surface deviation analysis results are as follows: S311: Based on the depth deviation distribution information, extract the normal vectors of the second regions, calculate the normal angle deviations between the normal vectors of the second regions and the normal of the standard mandibular first molar morphology, and establish a normal angle deviation distribution map according to the spatial distribution of the normal angle deviation values; S312: Based on the normal angle deviation distribution map, set an angle deviation threshold, divide it into multiple normal angle deviation regions, and adopt a local surface constraint fitting method based on the spatial distribution relationship of data points to obtain the local surface deviation values of each region, integrate the deviation information of all regions, and generate the surface deviation analysis results.

6. The method for evaluating the quality of tooth preparation based on image recognition according to claim 1, wherein, The specific steps for obtaining the regional deviation classification results are as follows: S411: Based on the surface deviation analysis results, extract the deviation information of all the buccal fossa preparation regions of the mandibular first molar, and use the formula: Calculate the entropy value feature A of the corresponding region, integrate the entropy value information of all data points, and establish an entropy value feature distribution; where P r represents the proportion of data points in the r-th deviation interval, D t represents the deviation value of the t-th data point, D avg represents the average deviation of all data points in the corresponding area, v represents the total number of data points, and s represents the total number of deviation intervals; S412: Based on the entropy value feature distribution, classify according to the entropy value range of the data points, determine the same type of deviation regions, calculate the deviation cumulative factor of the corresponding same type of deviation regions, where the deviation cumulative factor represents the overall cumulative situation of the deviations within the region, integrate the cumulative factors of all classified regions, and obtain the regional deviation classification results.

7. The method for evaluating the quality of tooth preparation based on image recognition according to claim 1, wherein The specific steps for obtaining the tooth preparation quality evaluation results are as follows: S511: Based on the deviation cumulative factors corresponding to each type of region in the regional deviation classification results, calculate the overall quality evaluation value of the buccal fossa preparation region of the mandibular first molar, extract the deviation cumulative factors of all classified regions, and establish a quality evaluation value; S512: According to the set evaluation criteria, refer to the quality evaluation value, divide the grade intervals of the overall quality of the tooth preparation region, and integrate the classification information of all evaluation regions to obtain the tooth preparation quality evaluation results.

8. A tooth preparation quality assessment system based on image recognition, characterized in that, According to the tooth preparation quality evaluation method based on image recognition according to any one of claims 1-7, the system includes: The three-dimensional reconstruction and curvature analysis module acquires the image of the buccal fossa preparation region of the mandibular first molar, reconstructs the tooth surface morphology through point cloud, extracts the point cloud data after reconstruction to calculate the curvature gradient of the buccal fossa preparation region, and obtains the curvature gradient distribution information; The depth deviation detection module extracts the adjacent point sets of the first region from the curvature gradient distribution information, compares the depth change trends of each region with those of adjacent regions, and screens out the second regions with depth deviations exceeding the deviation threshold to obtain the depth deviation distribution information; The normal analysis and surface fitting module extracts the normal vectors of the second regions in the depth deviation distribution information, calculates the regions corresponding to the normal angle deviations, and performs local surface constraint fitting to obtain the surface deviation analysis results; Based on the surface deviation analysis results, the deviation feature extraction and classification module calculates the entropy value features of all the buccal fossa preparation areas of the mandibular first molar and classifies the deviation areas of the same type, and calculates the deviation accumulation factor corresponding to each type of area after classification to obtain the regional deviation classification result; Based on the deviation accumulation factor corresponding to each type of area in the regional deviation classification result, the tooth preparation quality evaluation module calculates the overall quality evaluation value of the buccal fossa preparation area of the mandibular first molar to obtain the tooth preparation quality evaluation result.

Images