Automatic tooth arrangement method and system based on artificial intelligence
Through the artificial intelligence adaptive regional growth method, the problem of tooth dirt affecting edge detection is solved, and more accurate tooth contour edge extraction and efficient generation of tooth elimination schemes is achieved.
Patent Information
- Application Number
- CN202411194597.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-08-29
AI Technical Summary
In the existing automatic dental discharge technology, the presence of dirt on the teeth causes incomplete or inaccurate contours of the teeth extracted by the edge detection algorithm, which affects the accuracy of the dental discharge scheme.
Adaptive regional growth method based on artificial intelligence is adopted to gradually obtain the teeth contour edge by obtaining the initial seed points and growth points, improving the accuracy and completeness of edge detection.
It realizes a relatively accurate and complete tooth contour edge extraction, which improves the accuracy and reference value of automatic tooth elimination schemes.
Smart Images

Figure CN118697499B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and in particular to an automatic tooth arrangement method and system based on artificial intelligence. Background Art
[0002] Automatic tooth arrangement refers to the process of automatically designing and making dentures, braces, etc. in the dental field using computer-aided design and computer-aided manufacturing technology to achieve the purpose of tooth correction, denture restoration, production of cosmetic crowns and veneers, etc. At present, in order to provide patients with better services and to improve the accuracy and efficiency of automatic tooth arrangement, artificial intelligence has also begun to be widely used in automatic tooth arrangement. That is, the combination of artificial intelligence and tooth arrangement is crucial to improve the accuracy and efficiency of tooth arrangement.
[0003] Currently, the surface images of the patient's teeth and gums are usually input into the trained AI model first. The AI model will identify the shape and position information of the teeth by capturing the tooth area in the surface image, that is, by extracting the tooth edge contour information in the image, so as to automatically generate a tooth arrangement plan. Currently, traditional edge detection algorithms are usually used to extract the tooth edge contours in the image, that is, the edge image corresponding to the surface image is usually obtained based on traditional edge detection algorithms. However, some patients have dirt on their teeth, such as black stains, and the presence of tooth dirt will cause the tooth edge contour extracted by the edge detection algorithm to appear incomplete or inaccurate. That is, when there is dirt on the teeth, the edge image obtained based on the edge detection algorithm is relatively inaccurate or incomplete, which leads to lower accuracy of the subsequent automatically generated tooth arrangement plan. Therefore, how to accurately and completely obtain the edge image corresponding to the surface image of the patient's teeth and gums is an urgent problem to be solved. Summary of the invention
[0004] In order to solve the above problems, the present invention provides an automatic tooth arrangement method and system based on artificial intelligence, and the technical solutions adopted are as follows:
[0005] In a first aspect, an embodiment of the present invention provides an automatic tooth arrangement method based on artificial intelligence, comprising the following steps:
[0006] Acquire a target surface image of the teeth and gums of a target patient, and obtain an initial seed point set corresponding to the target surface image;
[0007] Acquire an initial set of points to be grown for each initial seed point in the initial seed point set, obtain a growth point of each initial seed point according to the initial set of points to be grown, obtain each first growth edge, a first seed point corresponding to each first growth edge, and a first growth sequence corresponding to each first growth edge according to the growth point of each initial seed point;
[0008] Acquire a first set of points to be grown of the first seed point, obtain a growth point of the first seed point corresponding to each first growth edge according to the first set of points to be grown and the first growth sequence, obtain each second growth edge, a second seed point corresponding to each second growth edge, and a second growth sequence corresponding to each second growth edge according to the growth point of the first seed point;
[0009] Obtaining a target tooth contour edge on the target surface image according to the second growth edges, the second seed point and the second growth sequence;
[0010] According to the target tooth contour edge, a target edge image corresponding to the target surface image is obtained, and the target edge image is input into the trained AI model to output the tooth arrangement plan for the target patient.
[0011] In a second aspect, an embodiment of the present invention provides an automatic tooth arrangement system based on artificial intelligence, comprising a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the above-mentioned automatic tooth arrangement method based on artificial intelligence.
[0012] Beneficial effects: The present invention obtains the growth points of the first seed points corresponding to each first growth edge according to the first set of points to be grown and the first growth sequence, and obtains each second growth edge, the second seed point corresponding to each second growth edge and the second growth sequence corresponding to each second growth edge according to the growth points of the first seed point, and then obtains the target tooth contour edge on the target surface image according to each second growth edge, the second seed point and the second growth sequence; finally, according to the target tooth contour edge, obtains the target edge image corresponding to the target surface image, and inputs the target edge image into the trained AI model to output the tooth arrangement plan for the target patient. The present invention can obtain the target tooth contour edge on the target surface image more accurately and completely according to each second growth edge, the second seed point and the second growth sequence, and the target edge image determined based on the target tooth contour edge can improve the accuracy of automatically generating the tooth arrangement plan for the target patient, that is, the tooth arrangement plan automatically generated based on the target edge image has a higher accuracy, and the tooth arrangement plan generated based on the target edge image also has a higher reference value. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0014] Figure 1 A flowchart of an automatic tooth arrangement method based on artificial intelligence of the present invention;
[0015] Figure 2 Schematic diagram of the target surface image. DETAILED DESCRIPTION
[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the embodiments of the present invention.
[0017] Unless defined otherwise, 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.
[0018] This embodiment provides an automatic tooth arrangement method based on artificial intelligence, which is described in detail as follows:
[0019] like Figure 1 As shown, the automatic tooth arrangement method based on artificial intelligence includes the following steps:
[0020] Step S001, obtaining a target surface image of teeth and gums of a target patient, and obtaining an initial seed point set corresponding to the target surface image.
[0021] Because some patients have dirt, such as black stains, on their teeth, the presence of tooth dirt will cause the tooth edge contour extracted by the edge detection algorithm to appear incomplete or inaccurate. That is, when there is dirt on the teeth, the edge image obtained based on the edge detection algorithm is relatively inaccurate or the extracted tooth contour edge will appear inaccurate or incomplete, which will lead to low accuracy of the subsequent automatically generated tooth arrangement plan. In order to improve the accuracy of subsequent automatic tooth arrangement, this embodiment proposes an automatic tooth arrangement method based on artificial intelligence. The method is mainly based on the characteristics of the dirt area, and performs adaptive regional growth to obtain more accurate tooth contour edges, thereby achieving the purpose of improving the accuracy of automatic tooth arrangement.
[0022] Therefore, through the above analysis, it can be known that the present embodiment needs to perform region growing later, and the region growing first needs to determine the initial seed point on the surface image of the patient's teeth and gums, and before determining the initial seed point, the surface image of the patient's teeth and gums needs to be obtained first; in addition, for ease of understanding, the present embodiment will take any patient who needs teeth arrangement as an example for analysis, and record it as the target patient, and there is dirt on the teeth of the target patient, such as black stains; therefore, the present embodiment will obtain the surface image of the target patient's teeth and gums next, and the specific acquisition process is:
[0023] Since teeth are not arranged completely in a straight line, it is difficult to ensure that the surfaces of all teeth and gums of the patient are captured in a single image. Therefore, this embodiment will use a medical oral photography lens to capture images of the surfaces of the target patient's teeth and gums from different angles, and multiple surface images of the target patient's teeth and gums can be obtained. This embodiment requires that all the captured surface images be stitched together to obtain a stitched image that can include the surfaces of all teeth and gums of the target patient.
[0024] In addition, in order to ensure the reliability of the collected images and the accuracy of the subsequent tooth edge contour extraction, this embodiment requires that the light source of the shooting environment is sufficient when the image is collected.
[0025] Then, all the collected surface images are grayed out, and the grayed out surface images are recorded as target surface images of the target patient's teeth and gums; for example, Figure 2 is any target surface image of the target patient's teeth and gums.
[0026] Next, this embodiment will analyze each target surface image, and then obtain the initial seed point set corresponding to each target surface image. Since the method for obtaining the initial seed point set corresponding to all target surface images in this embodiment is the same, this embodiment will take the process of obtaining the initial seed point set corresponding to any target surface image as an example for ease of understanding and description, specifically:
[0027] First, the target surface image is edge detected using a bidirectional edge detection algorithm, and all detected edges are recorded as initial edges on the target surface image. Since there is dirt on the target patient's teeth, the initial edge obtained is not completely the tooth edge contour, and the initial edge obtained may not represent the complete tooth edge contour, so the initial edge may be a part of the tooth contour edge or a tooth dirt edge, that is, the initial edge may belong to the tooth contour edge or the tooth dirt edge. In addition, the process of edge detection of an image using a bidirectional edge detection algorithm is a well-known technology, so it will not be described in detail in this embodiment.
[0028] In order to ensure the reliability of subsequent adaptive region growing and to accurately obtain the target tooth contour edge, this embodiment requires that the selected initial seed point cannot fall into the tooth stain area. In order to achieve the purpose that the initial seed point cannot fall into the tooth stain area, this embodiment will analyze the characteristics of the obtained initial edge, and determine the placement position of the initial seed point based on the analysis result. The specific process of analyzing the initial edge includes:
[0029] First, the edge pixel point sequence corresponding to each initial edge on the target surface image is obtained. Since there are closed edges and non-closed edges in the initial edges obtained at this time, and the closed edges and non-closed edges have different methods in determining the edge pixel point sequence, that is, for any initial edge a on the target surface image:
[0030] If the initial edge a is determined to be a closed edge, then any edge pixel point on the initial edge a is selected as the starting point of the initial edge a, and starting from the starting point of edge a, the edge pixel points on the initial edge a are traversed in a clockwise direction until the starting point of the initial edge a is reached, and all edge pixel points on edge a are sorted in the order of traversal to obtain the edge pixel point sequence corresponding to the initial edge a.
[0031] If it is determined that the initial edge a does not belong to a closed edge, it means that the initial edge a has two endpoints at this time. Therefore, the two endpoints of the initial edge a are obtained and recorded as the starting endpoint and the ending endpoint respectively. Then, the traversal is started from the starting endpoint of the initial edge a until the ending endpoint of the initial edge a is reached. The traversal is stopped, and all edge pixel points on the initial edge a are sorted in the order of traversal to obtain the edge pixel point sequence corresponding to the initial edge a.
[0032] Therefore, through the above process, the edge pixel point sequence corresponding to each initial edge can be obtained. Then, this embodiment analyzes the edge pixel point sequence corresponding to each initial edge to obtain the appropriate placement degree of each initial edge, and the appropriate placement degree is an important basis for determining the initial seed point later. Therefore, the specific process of obtaining the appropriate placement degree is:
[0033] For the initial edge a: the slope sequence corresponding to the initial edge a is obtained, and the bth slope in the slope sequence is the slope between the bth edge pixel point and the b+1th edge pixel point in the edge pixel point sequence corresponding to the initial edge a; then the difference sequence of the slope sequence is obtained, and the kth difference in the difference sequence is the absolute value of the difference between the kth slope and the k+1th slope in the slope sequence; then, the cumulative sum of all the differences in the difference sequence is calculated and recorded as the comprehensive difference, and then the reciprocal of the value obtained by adding the preset first constant and the comprehensive difference is calculated and recorded as the first index value of the initial edge a; then the area of the minimum circumscribed rectangle corresponding to the initial edge a and the total number of edge pixels on the initial edge a are obtained, which are recorded as the characteristic area and the characteristic quantity value respectively, and then the ratio of the characteristic quantity value to the characteristic area is calculated and recorded as the second index value of the initial edge a; finally, the normalized value of the product of the first index value and the second index value of the initial edge a is calculated and recorded as the appropriate placement degree of the initial edge a.
[0034] In a specific application, the implementer may set the value of the preset first constant according to actual conditions. For example, in this embodiment, the value of the preset first constant is set to 1.
[0035] In addition, in this embodiment, the specific expression for obtaining the appropriate placement degree of the initial edge a is:
[0036]
[0037] in, is the appropriate placement degree of the initial edge a, is the total number of differences in the difference sequence of the slope sequence corresponding to the initial edge a, is the kth difference in the difference sequence, is the area of the minimum circumscribed rectangle corresponding to the initial edge a, is the total number of edge pixels on the initial edge a, The reciprocal of is the first index value of the initial edge a, is the second index value of the initial edge a, and Norm() is the normalization function.
[0038] And when the first index value and the second index value of the initial edge a are larger, it indicates that The larger the value of is, the smaller the first index value and the second index value of the initial edge a are, indicating that The smaller the value of .
[0039] And because the shape of the tooth stain edge during the extension process is more tortuous and the overall distribution of pixels on the edge is more chaotic, that is, the linear distribution characteristics of the pixels on the tooth stain edge are weaker, and the extension distance is relatively shorter; on the contrary, the shape of the tooth contour edge is relatively gentle during the extension process compared to the tooth stain edge, and the overall distribution of pixels on the edge is less chaotic, that is, the linear distribution characteristics of the pixels on the tooth contour edge are stronger, and the extension distance is relatively longer.
[0040] Since the first index value of the initial edge a can reflect the change characteristics of the initial edge a, the second index value of the initial edge a can reflect the linear distribution characteristics or extended continuity characteristics of the pixels located on the initial edge, and when the first index value and the second index value are larger, that is, The larger the value of is, the smoother the change of the pixels on the initial edge a is, the more linear the distribution of the pixels on the initial edge a is, and the longer the extension distance of the initial edge a is. This also indicates that the initial edge a is more likely to belong to the edge of the tooth contour, and the subsequent initial edge a is more likely to be the initial seed point placement area. On the contrary, when the first index value and the second index value are smaller, that is, The smaller the value is, the more tortuous the change of the pixels on the initial edge a is, the less linear the distribution of the pixels on the initial edge a is, and the shorter the extension distance of the initial edge a is. This also indicates that the possibility that the initial edge a belongs to the tooth stain edge is greater, and the possibility that the subsequent initial edge a is the initial seed point placement area is smaller.
[0041] Therefore, in this embodiment, the suitable placement degree of all initial edges can be obtained according to the above-mentioned acquisition of the suitable placement degree of the initial edge a; then the edge to be placed is determined based on the suitable placement degree, specifically: whether the suitable placement degree of each initial edge is not less than the preset first threshold value is judged, if so, the corresponding initial edge is determined to belong to the tooth contour edge, and the corresponding initial edge is recorded as the edge to be placed; then the set constructed by all the acquired edges to be placed is recorded as the edge set to be placed, and one edge pixel point is selected on each edge to be placed in the edge set to be placed as the initial seed point of the corresponding edge to be placed, and then the set constructed by the initial seed points of all the edges to be placed is recorded as the initial seed point set corresponding to the target surface image.
[0042] In a specific application, the implementer needs to set the preset first threshold according to the actual situation. For example, in this embodiment, the preset first threshold is set to 0.8.
[0043] Therefore, this embodiment obtains the initial seed point set corresponding to the target surface image through the above process.
[0044] Step S002, obtaining an initial set of points to be grown of each initial seed point in the initial seed point set, obtaining the growth points of each initial seed point based on the initial set of points to be grown, obtaining each first growth edge, the first seed point corresponding to each first growth edge, and the first growth sequence corresponding to each first growth edge based on the growth points of each initial seed point; obtaining the first set of points to be grown of the first seed point, obtaining the growth point of the first seed point corresponding to each first growth edge based on the first set of points to be grown and the first growth sequence, obtaining each second growth edge, the second seed point corresponding to each second growth edge, and the second growth sequence corresponding to each second growth edge based on the growth point of the first seed point; obtaining the target tooth contour edge on the target surface image based on the second growth edges, the second seed point, and the second growth sequence.
[0045] If there is dirt on the surface of a tooth, and one side edge of the dirt area is adjacent to the gums, then this embodiment requires that when the region growing algorithm is subsequently used to perform region growing, if the growth reaches the dirt area on the tooth, the growth should not continue along the dirt-free area on the tooth, that is, the growth should continue along the edge of the dirt area adjacent to the gums, so as to ensure that a more accurate and complete tooth contour edge is extracted; for example, for Figure 2 For the 4th tooth in the upper tooth area, and the 4th tooth is determined in order from left to right, in the process of obtaining the contour edge of the 4th tooth based on the region growing algorithm, if it grows to the dirt area on the 4th tooth, then it should continue to grow along the edge of the dirt area adjacent to the gum above the 4th tooth.
[0046] Therefore, in order to meet the growth requirements in the regional growth process analyzed above, this embodiment will next determine the growth point of the seed point by analyzing the grayscale difference between the seed point and the pixels in its neighborhood and the direction of the vector formed by the initial seed point and the pixels in its neighborhood, so as to facilitate the subsequent determination of the target tooth contour edge, wherein the target tooth contour edge refers to the edge of a relatively accurate and complete tooth contour.
[0047] In order to obtain the target tooth contour edge, this embodiment first needs to obtain the initial set of points to be grown for each initial seed point in the initial seed point set, and the method for obtaining the initial set of points to be grown is: for any initial seed point C in the initial seed point set, in the target surface image, a set constructed by all pixel points in the eight neighborhoods of the initial seed point C is recorded as the initial set of points to be grown for the initial seed point C.
[0048] Then, according to the initial set of points to be grown of each initial seed point, the growth point of each initial seed point is obtained. In order to facilitate description and analysis, the present embodiment will take the process of obtaining the growth point of the initial seed point C as an example for description, that is, the specific process of obtaining the growth point of the initial seed point C is as follows:
[0049] First, the suitable growth degree corresponding to each initial point to be grown in the initial set of points to be grown of the initial seed point C is obtained, and then the growth point is determined based on the suitable growth degree. For ease of understanding, this embodiment will take the process of obtaining the suitable growth degree corresponding to any initial point to be grown D in the initial set of points to be grown of the initial seed point C as an example for description, that is, the process of obtaining the suitable growth degree corresponding to the initial point to be grown D is:
[0050] First, the absolute value of the difference between the grayscale value of the initial seed point C and the grayscale value of the initial point D to be grown is calculated and recorded as the first difference value, and then the ratio of the first difference value to the preset second constant is calculated and recorded as the first ratio; then, the first ratio is multiplied by the preset third constant and then subtracted from the value of the preset first constant and recorded as the first operation value, and the normalized value of the absolute value of the first operation value is recorded as the first characterization value of the initial point D to be grown.
[0051] In a specific application, the implementer needs to set the preset second constant and the preset third constant according to the actual situation. In this embodiment, the preset second constant and the preset third constant are set to 255 and 2 respectively.
[0052] In addition, in this embodiment, the specific expression for obtaining the first characterization value of the initial point D to be grown is:
[0053]
[0054] in, is the first characterization value of the initial growth point D, is the gray value of the initial seed point C, is the gray value of the initial growth point D; and The larger or smaller the value of The value of is close to 1 or -1, and The closer the value is to 1 or -1, the The closer the value is to 1.
[0055] Since the color of the dirt area is darker than that of the tooth area, it means that the grayscale value of the tooth area is larger and the grayscale value of the dirt area is lower. In order to ensure that when growing to the dirt area on the tooth, it should not continue to grow along the dirt-free area on the corresponding tooth, but should continue to grow along the edge of the dirt area adjacent to the gums of the corresponding tooth, it should be required not only The smaller the value, the greater the possibility that the initial growth point D is a growth point, and it is also required The larger the value of is, the greater the possibility that the initial growth point D is the growth point. The larger the value of , the greater the possibility that the initial growth point D is the growth point, which also indicates that the initial growth point D is more suitable for growth.
[0056] When there is no dirt area on the teeth, the grayscale difference in the areas on both sides of the tooth contour edge is relatively large. When there is dirt near the position of the tooth contour edge, the grayscale difference in the areas on both sides of the tooth contour edge is not only relatively large, but also the grayscale confusion degree difference in the areas on both sides of the tooth contour edge is relatively large, that is, since the grayscale in the dirt area is relatively chaotic and the grayscale in the oral tissue area such as the teeth or gums is relatively uncluttered, the grayscale confusion degree difference in the areas on both sides of the tooth contour edge is relatively large; therefore, in order to further achieve the purpose of not continuing to grow along the dirt-free area on the corresponding teeth when growing to the dirt area on the teeth, but continuing to grow along the edge of the dirt area adjacent to the gums of the corresponding teeth, this embodiment will next obtain the second target value of the initial point to be grown D based on the grayscale features in the areas on both sides of the vector formed by the initial seed point C and the initial point to be grown D, that is, the second target value can also reflect the appropriate growth degree of the initial point to be grown D, so the specific process of obtaining the second target value of the initial point to be grown D is:
[0057] First, the vector formed by the initial seed point C to the initial point to be grown D is recorded as the first vector; then, a straight line perpendicular to the direction of the first vector is drawn through the initial point to be grown D, and recorded as the first straight line; then, on the first straight line, the left adjacent pixel points and the right adjacent pixel points of the initial point to be grown D are obtained, and recorded as the first pixel point and the second pixel point respectively; then, on the first straight line, the left adjacent pixel points of the first pixel point and the right adjacent pixel points of the second pixel point are obtained, and recorded as the first target point and the second target point respectively; that is, on the first straight line, the first target point is located on the left side of the initial point to be grown D, and the second target point is located on the right side of the initial point to be grown D, and the first target point and the second target point are each separated from the initial point to be grown D by one pixel.
[0058] Then, on the target surface image, all the pixels belonging to the eight neighborhoods of the first target point and all the pixels belonging to the eight neighborhoods of the second target point are obtained, and the set constructed by the grayscale values of all the pixels belonging to the eight neighborhoods of the first target point is recorded as the first set, and the set constructed by the grayscale values of all the pixels belonging to the eight neighborhoods of the second target point is recorded as the second set; then, the absolute value of the difference between the mean of the first set and the mean of the second set is calculated, and recorded as the mean difference, and then the absolute value of the difference between the variance of the first set and the variance of the second set is continued to be calculated, and recorded as the variance difference, and the value obtained by adding the normalized value of the mean difference to the normalized value of the variance difference is recorded as the second target value of the initial point D to be grown.
[0059] In addition, when the mean difference and variance difference are larger, it indicates that the second target value of the initial growth point D is larger. The larger the second target value, the greater the possibility that the initial growth point D is a growth point, which also indicates that the initial growth point D is more suitable for growth.
[0060] Then, the first characterization value of the initial growth point D is added to the second target value of the initial growth point D, and the added value is recorded as the suitable growth degree corresponding to the initial growth point D. The larger the value of the suitable growth degree is, the greater the possibility that the initial growth point D is the growth point of the initial seed point C. Since the initial seed point C is a point on the edge of the tooth contour, the position of the growth point of the initial seed point C is more likely to be a certain position on the edge of the tooth contour when there is no dirt.
[0061] Therefore, this embodiment can obtain the suitable growth degree corresponding to each initial point to be grown in the set of initial points to be grown of the initial seed point C according to the above-mentioned method of obtaining the suitable growth degree corresponding to the initial point to be grown D, and then select the initial point to be grown corresponding to the largest suitable growth degree in the set of initial points to be grown of the initial seed point C as the growth point of the initial seed point C.
[0062] So far, through the above process, this embodiment can obtain the growth points of all initial seed points. After obtaining the growth points of the initial seed points, this embodiment will obtain each first growth edge, the first seed point corresponding to each first growth edge, and the first growth sequence corresponding to each first growth edge according to the growth point of each initial seed point. The specific process of obtaining each first growth edge, the first seed point corresponding to each first growth edge, and the first growth sequence corresponding to each first growth edge is:
[0063] For the initial seed point C, the edge formed by the initial seed point C and the growth point of the initial seed point C is recorded as the first growth edge L0 of the initial seed point C, that is, the first growth edge L0 is the curve obtained by connecting the initial seed point C and the growth point of the initial seed point C, and then the growth point of the initial seed point C is used as the first seed point corresponding to the first growth edge L0, and the sequence constructed by the initial seed point C and the growth point of the initial seed point C is recorded as the first growth sequence corresponding to the first growth edge L0; and in the first growth sequence corresponding to the first growth edge L0, the growth point of the initial seed point C is located behind the initial seed point C.
[0064] It can be known from the above description that an initial seed point and its corresponding growth point can obtain a first growth edge, and a first growth edge corresponds to a first seed point and a first growth sequence respectively.
[0065] At this point, this embodiment has completed the first growth, and then continues with the second growth, that is, continues to obtain the first set of points to be grown corresponding to the first seed points of each first growth edge, and the method for obtaining the first set of points to be grown is: for the first seed point g1 corresponding to any first growth edge L1, in the target surface image, a set constructed by all pixel points belonging to the eight neighborhoods of the first seed point g1 is recorded as the first set of points to be grown of the first seed point g1.
[0066] Then, according to the first set of to-be-grown points of the first seed points corresponding to each first growing edge and the first growth sequence corresponding to each first growing edge, the growth point of the first seed point corresponding to each first growing edge is obtained. For the sake of ease of understanding and analysis, this embodiment will take the process of obtaining the growth point of the first seed point g1 corresponding to the first growing edge L1 as an example for analysis. Then, the process of obtaining the growth point of the first seed point g1 is as follows:
[0067] First, the suitable growth degree corresponding to each first to-be-grown point in the first to-be-grown point set of the first seed point g1 is obtained, and then the growth point of the first to-be-grown point is determined based on the suitable growth degree. For ease of understanding, this embodiment will take the process of obtaining the suitable growth degree corresponding to any first to-be-grown point D0 in the first to-be-grown point set of the first seed point g1 as an example for description, that is, the process of obtaining the suitable growth degree corresponding to the first to-be-grown point D0 is:
[0068] The first characterization value and the second target value of the first growth point D0 are obtained, and the method for obtaining the first characterization value and the second target value of the first growth point D0 is the same as the method for obtaining the first characterization value and the second target value of the initial growth point D mentioned above, so this embodiment will not be described in detail.
[0069] Since the first characterization value and the second target value are both obtained by analyzing the grayscale condition, and the growth direction is easily offset based on the grayscale condition alone, that is, it is easy to grow inside the tooth area or inside the dirt, which leads to the inability to accurately or completely extract the tooth contour edge in the subsequent process. In order to avoid this phenomenon, this embodiment will further determine the growth point of the first seed point g1 by analyzing the direction of the grown path, that is, by analyzing the direction of the grown path to obtain the second characterization value of the first point to be grown D0, and the specific acquisition process of the second characterization value is as follows:
[0070] First, the vector formed by the first seed point g1 to the first point to be grown D0 is recorded as the second vector, and the feature vector set of the first growth sequence corresponding to the first growth edge L1 is obtained, and the f-th feature vector in the feature vector set refers to the vector formed by the f-th pixel point to the f+1-th pixel point in the sequence Z1, and the sequence Z1 is the first growth sequence corresponding to the first growth edge L1.
[0071] Then, by analyzing the difference between the direction of the vector in the feature vector set and the direction of the second vector, the similar vector corresponding to the second vector is obtained, and the specific acquisition process is: for any feature vector f0 in the feature vector set, determine whether the absolute value of the difference between the direction angle of the feature vector f0 and the direction angle of the second vector is not greater than the preset second threshold. If so, it is determined that the direction of the feature vector f0 is the same as the direction of the second vector, and the feature vector f0 is recorded as the similar vector corresponding to the second vector.
[0072] In a specific application, the implementer may set the preset second threshold value according to actual conditions. For example, in this embodiment, the preset second threshold value may be set to 10 degrees or 0 degrees.
[0073] Then, the number of similar vectors corresponding to the second vector is counted and recorded as the first quantitative characterization value, and the normalized value of the first quantitative characterization value is recorded as the second characterization value of the first growth point D0; and the larger the first quantitative characterization value, the more eigenvectors there are in the same direction as the second vector, which further indicates that the possibility that the first growth point D0 is a growth point is greater, and also indicates that the first growth point D0 is more suitable for growth.
[0074] Then, the value obtained by adding the second characterization value of the first growth point D0 to the first characterization value of the first growth point D0 is recorded as the first target value of the first growth point D0, and then the value obtained by adding the first target value of the first growth point D0 to the second target value of the first growth point D0 is recorded as the suitable growth degree corresponding to the first growth point D0; and similarly, the larger the value of the suitable growth degree is, the greater the possibility that the first growth point D0 is the growth point of the initial seed point C, and also indicates that the position of the first growth point D0 is more likely to be a certain position on the edge of the tooth contour when there is no dirt.
[0075] Therefore, through the above process, the suitable growth degree corresponding to each first growth point in the first growth point set of the first seed point g1 can be obtained, and then all the first growth points that do not belong to the first growth edge L1 are obtained in the first growth point set of the first seed point g1, and the set constructed by all the first growth points that do not belong to the first growth edge L1 is recorded as the first feature set, and then in the first feature set, the first growth point corresponding to the largest suitable growth degree is selected as the growth point of the first seed point g1, and according to the method of obtaining the growth point of the first seed point g1, the growth point of the first seed point corresponding to each first growth edge can also be obtained; in addition, the point that does not belong to the growth edge in this embodiment refers to a pixel point that is not located on the corresponding growth edge.
[0076] After obtaining the growth point of the first seed point corresponding to each first growth edge, then according to the growth point of the first seed point corresponding to each first growth edge, each second growth edge, the second seed point corresponding to each second growth edge, and the second growth sequence corresponding to each second growth edge are obtained, and the specific acquisition process of each second growth edge, the second seed point corresponding to each second growth edge, and the second growth sequence corresponding to each second growth edge is:
[0077] For the first growth edge L1, the edge formed by the first growth edge L1 and the growth point t1 is recorded as the second growth edge L2 of the first growth edge L1, the growth point t1 is the growth point of the first seed point corresponding to the first growth edge L1, and then the growth point t1 is used as the second seed point corresponding to the second growth edge L2, and the new sequence constructed by the first growth sequence corresponding to the first growth edge L1 and the growth point t1 is recorded as the second growth sequence corresponding to the second growth edge L2; and the second growth edge L2 refers to the curve obtained by connecting the first growth edge L1 and the growth point t1. When connecting, the growth point t1 is connected to the last pixel point in the first growth sequence corresponding to the first growth edge L1; and in the second growth sequence corresponding to the second growth edge L2, the growth point t1 is located behind the first growth sequence corresponding to the first growth edge L1.
[0078] In addition, a first growing edge and a growing point of its corresponding first seed point can obtain a second growing edge, and a second growing edge corresponds to a second seed point and a second growing sequence respectively.
[0079] After obtaining each second growth edge, the second seed point corresponding to each second growth edge, and the second growth sequence corresponding to each second growth edge, the present embodiment will continue to grow, that is, the present embodiment will obtain all target tooth contour edges on the target surface image according to each second growth edge, the second seed point, and the second growth sequence, specifically:
[0080] For the second growth edge L2: obtain the second set of points to be grown of the second seed point corresponding to the second growth edge L2, and obtain the first feature point and the second feature point of the second seed point corresponding to the second growth edge L2 according to the second set of points to be grown of the second seed point corresponding to the second growth edge L2 and the second growth sequence corresponding to the second growth edge L2, and then determine whether the first feature point of the second seed point corresponding to the second growth edge L2 is the initial seed point on the second growth edge L2; if so, record the second growth edge L2 as the target tooth contour edge on the target surface image; otherwise, use the second feature point of the second seed point corresponding to the second growth edge L2 as the growth point of the second seed point corresponding to the second growth edge L2, and obtain the third growth edge L3 of the second growth edge L2, the third seed point corresponding to the third growth edge L3, and the third growth sequence corresponding to the third growth edge L3 according to the growth point of the second seed point corresponding to the second growth edge L2; and in this embodiment, there is an initial seed point on any growth edge.
[0081] In addition, if the first feature point of the second seed point corresponding to the second growth edge L2 is determined to be the initial seed point on the second growth edge L2, it also indicates that the first feature point of the second seed point corresponding to the second growth edge L2 is the first point in the second growth sequence corresponding to the second growth edge L2.
[0082] Then, the third set of points to be grown of the third seed point corresponding to the third growth edge L3 is obtained, and the first feature point and the second feature point of the third seed point corresponding to the third growth edge L3 are obtained according to the third set of points to be grown of the third seed point corresponding to the third growth edge L3 and the third growth sequence corresponding to the third growth edge L3; and it is further determined whether the first feature point of the third seed point corresponding to the third growth edge L3 is the initial seed point on the third growth edge L3; if so, the third growth edge L3 is recorded as the target tooth contour edge on the target surface image; otherwise, the second feature point of the third seed point corresponding to the third growth edge L3 is used as the growth point of the third seed point corresponding to the third growth edge L3.
[0083] According to the above process, the growth points of the third seed points corresponding to all the third growth edges can be obtained, and then the fourth growth edges of each third growth edge, the fourth seed points corresponding to the fourth growth edge and the fourth growth sequence corresponding to the fourth growth edge can be obtained according to the growth points of the third seed points corresponding to each third growth edge, and then it is continued to be judged whether the fourth growth edge is the target tooth contour edge on the target surface image, and so on, until there is no growth edge that needs to be judged, and all the target tooth contour edges on the acquired target surface image are statistically obtained.
[0084] In addition, the method for obtaining the second set of points to be grown of the second seed point and the third set of points to be grown of the third seed point in this embodiment is as follows: for the second seed point g2 corresponding to the second growth edge L2, in the target surface image, a set constructed by all pixel points belonging to the eight neighborhoods of the second seed point g2 is recorded as the second set of points to be grown of the second seed point g2; for the third seed point g3 of the third growth edge L3, in the target surface image, a set constructed by all pixel points belonging to the eight neighborhoods of the second seed point g3 is recorded as the second set of points to be grown of the third seed point g3. Since the tooth contour edge is a closed edge, in the growth process of this embodiment, when the point corresponding to the maximum suitable growth degree is the initial seed point, it indicates that a relatively complete tooth contour edge has been extracted, so the growth needs to be stopped.
[0085] The specific process of obtaining the first characteristic point and the second characteristic point of the second seed point corresponding to the second growth edge L2 is as follows: for the second seed point g2 of the second growth edge L2: obtain the suitable growth degree corresponding to each second growth point in the second growth point set Q of the second seed point g2, and the method for obtaining the suitable growth degree corresponding to the second growth point is the same as the method for obtaining the suitable growth degree corresponding to the first growth point D0 mentioned above, so this embodiment will not be described in detail; then, in the second growth point set Q, select the second growth point corresponding to the largest suitable growth degree as the first characteristic point of the second seed point g2. Then, the set constructed by all the second pixel points in the second growth point set Q that do not belong to the second growth edge L2 is recorded as the second characteristic set, and in the second characteristic set, select the second growth point corresponding to the largest suitable growth degree as the second characteristic point of the second seed point g2.
[0086] In addition, in the present embodiment, the method for obtaining the third growth edge L3 of the second growth edge L2, the third seed point corresponding to the third growth edge L3, and the third growth sequence corresponding to the third growth edge L3, and the method for obtaining the fourth growth edge of each third growth edge, the fourth seed point corresponding to the fourth growth edge, and the fourth growth sequence corresponding to the fourth growth edge are the same as the above-mentioned method for obtaining each second growth edge, the second seed point corresponding to each second growth edge, and the second growth sequence corresponding to each second growth edge, and therefore will not be described in detail again; and in the present embodiment, the method for obtaining the first feature point and the second feature point of the third seed point corresponding to the third growth edge L3 is the same as the above-mentioned method for obtaining the first feature point and the second feature point of the second seed point corresponding to the second growth edge L2, and therefore will not be described in detail again.
[0087] So far, the present embodiment has completed the region growing and obtained all the target tooth contour edges on the target surface image, and the target tooth contour edges are relatively accurate and complete tooth contour edges.
[0088] Step S003, obtaining a target edge image corresponding to the target surface image according to the target tooth contour edge, and inputting the target edge image into the trained AI model to output a tooth arrangement plan for the target patient.
[0089] This embodiment then sets the grayscale values of all pixels on the edge of the target tooth contour on the target surface image to 255, and sets the grayscale values of pixels that do not belong to the edge of the target tooth contour on the target surface image to 0, thereby obtaining a target edge image corresponding to the target surface image; that is, on the target edge image corresponding to the target surface image, the grayscale values of the pixels located on the edge of the target tooth contour are all 255, and the grayscale values of the pixels that are not located on the edge of the target tooth contour are all 0.
[0090] In this embodiment, there are multiple target surface images of the target patient's teeth and gums. Based on the above process, the target edge image corresponding to each target surface image can be obtained; then the target edge image is input into the trained AI model, and the tooth arrangement plan of the target patient is output, and the final tooth arrangement plan is converted into a CAD model; and in specific applications, the implementer can select a suitable AI model according to the actual situation, for example, a convolutional network CNN, a generative adversarial network GAN, etc., and the training process of the AI model is a well-known technology, so this embodiment will not be described in detail.
[0091] An automatic tooth arrangement system based on artificial intelligence in this embodiment includes a memory and a processor, and the processor executes a computer program stored in the memory to implement the above-mentioned automatic tooth arrangement method based on artificial intelligence.
[0092] In summary, this embodiment obtains the growth points of the first seed points corresponding to each first growth edge based on the first set of growth points and the first growth sequence, and obtains each second growth edge, the second seed point corresponding to each second growth edge and the second growth sequence corresponding to each second growth edge based on the growth points of the first seed point, and then obtains the target tooth contour edge on the target surface image based on each second growth edge, the second seed point and the second growth sequence; finally, the target edge image corresponding to the target surface image is obtained based on the target tooth contour edge, and the target edge image is input into the trained AI model to output the tooth arrangement plan for the target patient. This embodiment can obtain the target tooth contour edge on the target surface image more accurately and completely based on each second growth edge, the second seed point and the second growth sequence, and the target edge image subsequently determined based on the target tooth contour edge can improve the accuracy of automatically generating the tooth arrangement plan for the target patient, that is, the tooth arrangement plan automatically generated based on the target edge image has a higher accuracy, and the tooth arrangement plan generated based on the target edge image also has a higher reference value.
[0093] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. An automatic tooth arrangement method based on artificial intelligence, It is characterized in that The method comprises the following steps: Acquire a target surface image of the teeth and gums of a target patient, and obtain an initial seed point set corresponding to the target surface image; Acquire an initial set of points to be grown for each initial seed point in the initial seed point set, obtain a growth point of each initial seed point according to the initial set of points to be grown, obtain each first growth edge, a first seed point corresponding to each first growth edge, and a first growth sequence corresponding to each first growth edge according to the growth point of each initial seed point; Acquire a first set of points to be grown of the first seed point, obtain a growth point of the first seed point corresponding to each first growth edge according to the first set of points to be grown and the first growth sequence, obtain each second growth edge, a second seed point corresponding to each second growth edge, and a second growth sequence corresponding to each second growth edge according to the growth point of the first seed point; Obtaining a target tooth contour edge on the target surface image according to the second growth edges, the second seed point and the second growth sequence; According to the target tooth contour edge, a target edge image corresponding to the target surface image is obtained, and the target edge image is input into the trained AI model to output the tooth arrangement plan for the target patient.
2. The automatic tooth arrangement method based on artificial intelligence as claimed in claim 1, It is characterized in that The method for obtaining the initial seed point set includes: Perform edge detection on the target surface image using an edge detection algorithm, and record all detected edges as initial edges on the target surface image; obtain edge pixel point sequences corresponding to the initial edges, and obtain appropriate placement degrees of each initial edge according to the edge pixel point sequences corresponding to each initial edge on the target surface image; It is determined whether the degree of suitability for placement of each of the initial edges is not less than a preset first threshold value. If so, the corresponding initial edge is recorded as an edge to be placed; a set constructed by all the edges to be placed is recorded as an edge set to be placed, and one edge pixel point is selected on each edge to be placed in the edge set to be placed as an initial seed point of the corresponding edge to be placed, and a set constructed by the initial seed points of all the edges to be placed is recorded as an initial seed point set corresponding to the target surface image.
3. The automatic tooth arrangement method based on artificial intelligence as claimed in claim 2, It is characterized in that The method for obtaining the appropriate placement degree of each initial edge includes: For any initial edge a on the target surface image: Obtain a difference sequence of the slope sequence corresponding to the initial edge a, wherein the bth slope in the slope sequence is the slope between the bth edge pixel point and the b+1th edge pixel point in the edge pixel point sequence corresponding to the initial edge a; and the kth difference in the difference sequence is the absolute value of the difference between the kth slope and the k+1th slope in the slope sequence; The cumulative sum of all differences in the difference sequence is recorded as a comprehensive difference, and the reciprocal of the sum of a preset first constant and the comprehensive difference is recorded as a first index value of the initial edge a; the area of the minimum circumscribed rectangle corresponding to the initial edge a is obtained, and the ratio of the total number of edge pixels on the initial edge a to the area is recorded as a second index value of the initial edge a; the normalized value of the product of the first index value and the second index value is recorded as the appropriate placement degree of the initial edge a.
4. The automatic tooth arrangement method based on artificial intelligence as claimed in claim 1, It is characterized in that The method for obtaining the growth points of the initial seed points comprises: For any initial to-be-grown point D in the initial to-be-grown point set of any initial seed point C in the initial seed point set: The absolute value of the difference between the grayscale value of the initial seed point C and the grayscale value of the initial point D to be grown is recorded as a first difference value; the ratio of the first difference value to the preset second constant is recorded as a first ratio; the first ratio is multiplied by the preset third constant and then subtracted from the value of the preset first constant, which is recorded as a first operation value; the normalized value of the absolute value of the first operation value is recorded as a first characterization value of the initial point D to be grown; The vector formed from the initial seed point C to the initial point to be grown D is recorded as the first vector; a straight line perpendicular to the direction of the first vector is drawn through the initial point to be grown D, and recorded as the first straight line; on the first straight line, the left adjacent pixel points and the right adjacent pixel points of the initial point to be grown D are obtained, and recorded as the first pixel point and the second pixel point respectively; on the first straight line, the left adjacent pixel points of the first pixel point and the right adjacent pixel points of the second pixel point are obtained, and recorded as the first target point and the second target point respectively; on the target surface image, The set constructed by the grayscale values of all the pixels in the eight neighborhoods of the first target point is recorded as the first set, and the set constructed by the grayscale values of all the pixels in the eight neighborhoods of the second target point is recorded as the second set; the absolute value of the difference between the mean of the first set and the mean of the second set is recorded as the mean difference; the absolute value of the difference between the variance of the first set and the variance of the second set is recorded as the variance difference; the value obtained by adding the normalized value of the mean difference to the normalized value of the variance difference is recorded as the second target value of the initial point D to be grown; The sum of the first characterization value of the initial point D to be grown and the second target value of the initial point D to be grown is recorded as the suitable growth degree corresponding to the initial point D to be grown; In the set of initial points to be grown of the initial seed point C, the initial points to be grown corresponding to the largest suitable growth degree are selected as the growth points of the initial seed point C.
5. The automatic tooth arrangement method based on artificial intelligence as claimed in claim 1, It is characterized in that The method for obtaining each first growing edge, a first seed point corresponding to each first growing edge, and a first growing sequence corresponding to each first growing edge comprises: For any initial seed point C in the initial seed point set, the edge formed by the initial seed point C and the growth point of the initial seed point C is recorded as the first growth edge L0 of the initial seed point C, and the growth point of the initial seed point C is recorded as the first seed point corresponding to the first growth edge L0; the sequence constructed by the initial seed point C and the growth point corresponding to the initial seed point C is recorded as the first growth sequence corresponding to the first growth edge L0.
6. The automatic tooth arrangement method based on artificial intelligence as claimed in claim 4, It is characterized in that The method for obtaining the growth point of the first seed point corresponding to each first growth edge includes: For any first to-be-grown point D0 in the first to-be-grown point set corresponding to the first seed point g1 of any first growing edge L1: Obtaining a first characterization value and a second target value of the first growth point D0; The vector formed by the first seed point g1 to the first growth point D0 is recorded as the second vector; the feature vector set corresponding to the first growth sequence corresponding to the first growth edge L1 is obtained, and the f-th feature vector in the feature vector set is the vector formed by the f-th pixel point to the f+1-th pixel point in the first growth sequence corresponding to the first growth edge L1; for any feature vector f0 in the feature vector set, if the absolute value of the difference between the direction angle of the feature vector f0 and the direction angle of the second vector is not greater than the preset second threshold value, the feature vector f0 is recorded as the similarity vector corresponding to the second vector; the number of similar vectors corresponding to the second vector is counted and recorded as the first quantity characterization value, and the normalized value of the first quantity characterization value is recorded as the second characterization value of the first growth point D0; the sum of the second characterization value of the first growth point D0 and the first characterization value of the first growth point D0 is recorded as the first target value of the first growth point D0; the sum of the first target value of the first growth point D0 and the second target value of the first growth point D0 is recorded as the suitable growth degree corresponding to the first growth point D0; For the first seed point g1, a set constructed by all first growth points in the first growth point set of the first seed point g1 that do not belong to the first growth edge L1 is recorded as a first feature set. In the first feature set, the first growth point corresponding to the maximum suitable growth degree is selected as the growth point of the first seed point g1.
7. The automatic tooth arrangement method based on artificial intelligence as claimed in claim 5, It is characterized in that The method for obtaining the target tooth contour edge on the target surface image according to the second growth edges, the second seed points and the second growth sequence comprises: For any second growth edge L2: obtain the second set of points to be grown of the second seed point corresponding to the second growth edge L2, and obtain the first feature point and the second feature point of the second seed point corresponding to the second growth edge L2 according to the second set of points to be grown of the second seed point corresponding to the second growth edge L2 and the second growth sequence corresponding to the second growth edge L2; determine whether the first feature point of the second seed point corresponding to the second growth edge L2 is the initial seed point on the second growth edge L2, and if so, record the second growth edge L2 as the target tooth contour edge on the target surface image; otherwise, use the second feature point of the second seed point corresponding to the second growth edge L2 as the growth point of the second seed point corresponding to the second growth edge L2, and obtain the third growth edge L3 of the second growth edge L2, the third seed point corresponding to the third growth edge L3, and the third growth sequence corresponding to the third growth edge L3 according to the growth point of the second seed point corresponding to the second growth edge L2; Obtain a third set of points to be grown of the third seed point corresponding to the third growth edge L3, and obtain a first feature point and a second feature point of the third seed point corresponding to the third growth edge L3 according to the third set of points to be grown of the third seed point corresponding to the third growth edge L3 and the third growth sequence corresponding to the third growth edge L3; and continue to determine whether the first feature point of the third seed point corresponding to the third growth edge L3 is the initial seed point on the third growth edge L3; if so, record the third growth edge L3 as the target tooth contour edge on the target surface image.
8. The automatic tooth arrangement method based on artificial intelligence as claimed in claim 7, It is characterized in that The method for acquiring the set of points to be grown includes: For any seed point, in the target surface image, a set constructed by all pixel points in the eight neighborhoods of the seed point is recorded as the set of points to be grown of the seed point.
9. The automatic tooth arrangement method based on artificial intelligence as claimed in claim 7, It is characterized in that The method for obtaining the first characteristic point and the second characteristic point of the second seed point corresponding to the second growing edge L2 includes: For the second set Q of points to be grown of the second seed point g2 corresponding to the second growth edge L2, obtain the suitable growth degree corresponding to each second point to be grown in the second set Q of points to be grown, and the method for obtaining the suitable growth degree of the second point to be grown is the same as the method for obtaining the suitable growth degree of the first point to be grown D0; in the second set Q of points to be grown, select the second point to be grown corresponding to the largest suitable growth degree as the first feature point of the second seed point g2; construct a set of all second points to be grown in the second set Q of points to be grown that do not belong to the second growth edge L2, and record it as the second feature set; in the second feature set, select the second point to be grown corresponding to the largest suitable growth degree as the second feature point of the second seed point g2.
10. An automatic tooth arrangement system based on artificial intelligence, comprising a memory and a processor, It is characterized in that The processor executes the computer program stored in the memory to implement the artificial intelligence-based automatic tooth arrangement method as described in any one of claims 1-9.
