A method and device for identifying maxillary anterior tooth alveolar bone defects

Abstract

The application discloses a kind of maxillary anterior tooth alveolar bone defect identification method and device, the method comprises: S1, obtains global voxel matrix and voxel size information;Global voxel matrix is segmented and handled, and N tooth information is obtained;S2, based on voxel size information, each tooth information is handled, and the corresponding vector set of each tooth information is obtained;S3, each tooth information and corresponding vector set are handled, and the corresponding classification image pair set of each tooth information is obtained;S4, using the multi-scale classification identification model that is pre-trained, the classification image pair set corresponding to each tooth information is handled, and the probability vector set corresponding to each tooth information is obtained;S5, the probability vector set corresponding to each tooth information is handled, and the classification identification result corresponding to each tooth information is obtained.The method of the application can automatically complete the high-precision classification identification of alveolar bone defect.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of tooth damage detection, and particularly relates to a maxillary anterior tooth alveolar bone defect recognition method and device. BACKGROUND

[0002] Alveolar bone defects (ABD) are widely present in people without orthodontic treatment history. For patients who need to be treated by tooth extraction and orthodontic treatment, it is crucial to check the root bone morphology of the anterior teeth before orthodontic treatment. Orthodontists need to develop individualized orthodontic treatment plans according to the root bone morphology of the anterior teeth of the patients, to ensure the safe movement of the tooth roots within the alveolar bone range and reduce the occurrence of alveolar bone defects and tooth root absorption of the anterior teeth. For the alveolar bone defects of the anterior teeth after orthodontic treatment, orthodontists should diagnose and find out the defects in time, closely observe the mild and moderate alveolar bone defects for a long time, and perform periodontal bone graft surgery for severe alveolar bone defects. Therefore, accurately detecting and recognizing the alveolar bone windowing, cracking and normal region from the CBCT image of the maxillary anterior teeth is crucial for ensuring the safe movement range of the alveolar bone during the orthodontic surgery and reducing the damage of the alveolar bone.

[0003] However, the existing recognition method has the following problems: first, high dependence on manual work: since the traditional CBCT image analysis needs to be manually reconstructed by experienced physicians in the sagittal plane and the coronal plane, it is time-consuming and low in efficiency; second, limited recognition accuracy: two-dimensional images cannot comprehensively evaluate the three-dimensional morphology of the alveolar bone, and the existing convolutional neural network (CNN) model is difficult to effectively fuse multi-view features due to the limitation of the local receptive field; third, insufficient automation: the existing recognition method needs to manually define the region of interest (ROI), and cannot realize end-to-end automatic recognition. SUMMARY

[0004] The technical problem to be solved by the application is to provide a maxillary anterior tooth alveolar bone defect recognition method and device, which can automatically complete high-precision classification and recognition of the alveolar bone defect condition.

[0005] To solve the above technical problems, a maxillary anterior tooth alveolar bone defect recognition method is disclosed in the first aspect of the embodiments of the application, and the method comprises:

[0006] S1, acquiring a global voxel matrix and voxel size information; performing segmentation processing on the global voxel matrix to obtain N tooth information; N is an integer from 1 to 4;

[0007] The voxel size information includes a first voxel size, a second voxel size and a third voxel size;

[0008] The tooth information includes a tooth position number, a local voxel matrix and a mask matrix;

[0009] The global voxel matrix, the local voxel matrix and the mask matrix are all three-dimensional matrices, and the sizes of the local voxel matrix and the mask matrix are completely consistent; the values of the elements in the mask matrix are 0 or 1;

[0010] S2, based on the voxel size information, processing each tooth information to obtain a vector set corresponding to each tooth information;

[0011] The vector set includes a first vector, a second vector and a third vector.

[0012] S3, processing each tooth information and the corresponding vector set to obtain a classification image pair set corresponding to each tooth information;

[0013] The classification image pair set includes a plurality of classification image pairs; the classification image pair includes a first classification image and a second classification image.

[0014] S4, using a pre-trained multi-scale classification recognition model, processing the classification image pair set corresponding to each tooth information to obtain a probability vector set corresponding to each tooth information.

[0015] The probability vector set includes a probability vector corresponding to each classification image pair in the corresponding classification image pair set; the dimension of the probability vector is 5.

[0016] S5, processing the probability vector set corresponding to each tooth information to obtain a classification recognition result corresponding to each tooth information; the value of the classification recognition result is an integer from 1 to 5.

[0017] As an optional implementation, in the first aspect of the embodiment of the application, based on the voxel size information, processing each tooth information to obtain a vector set corresponding to each tooth information, including:

[0018] S21, presetting a template information set; the template information set includes M template information; the template information includes the tooth number and a two-dimensional template point set.

[0019] S22, based on the voxel size information, processing the mask matrix of each tooth information to obtain a compressed index matrix and the first vector corresponding to each tooth information;

[0020] S23, based on the voxel size information, processing the compressed index matrix and the two-dimensional template point set corresponding to each tooth information to obtain a three-dimensional feature point corresponding to each tooth information.

[0021] S24. Process the first vector and the three-dimensional feature point corresponding to each tooth information to obtain the second vector and the third vector corresponding to each tooth information; the first vector, the second vector and the third vector corresponding to each tooth information constitute the corresponding vector set.

[0022] As an optional implementation, in the first aspect of the present invention, processing the mask matrix of each tooth information based on the voxel size information to obtain a compressed index matrix and the first vector corresponding to each tooth information includes:

[0023] S221. Compress the mask matrix to obtain the corresponding compressed index matrix;

[0024] S222. Construct a model using a standard mask matrix, process the compressed index matrix and the voxel size information, and construct the corresponding standard mask matrix.

[0025] The standard mask matrix construction model is as follows:

[0026]

[0027] In the formula, and The standard mask matrix and the compressed index matrix are respectively, the first... i Line 1 j Column elements; i From 1 to I integers, I The total number of rows in the compressed index matrix; j It can be 1, 2, or 3; , and These are the first voxel size, the second voxel size, and the third voxel size, respectively.

[0028] S223. Using the feature matrix calculation model, the standard mask moments are processed to obtain the corresponding feature matrix;

[0029] The feature matrix calculation model is as follows:

[0030]

[0031] In the formula, C is the feature matrix, and B is the standard mask matrix;

[0032] S224. Perform eigenvalue decomposition on the feature matrix to obtain three feature pairs; each feature pair includes an eigenvalue and a corresponding eigenvector.

[0033] S225, normalize the feature vector corresponding to the largest feature value in the three feature pairs to obtain the first vector corresponding to the feature vector.

[0034] As an optional implementation, in the first aspect of the embodiment of the present application, the processing of the compressed index matrix and the two-dimensional template point set corresponding to each tooth information based on the voxel size information to obtain the three-dimensional feature point corresponding to each tooth information comprises:

[0035] S231, processing the compressed index matrix and the voxel size information corresponding to the tooth information to obtain the conversion coordinate information and the two-dimensional coordinate point set corresponding to the tooth information;

[0036] The conversion coordinate information comprises a selected origin, a first unit vector and a second unit vector; and the two-dimensional coordinate point set comprises J a two-dimensional coordinate point; J is an integer greater than 1;

[0037] S232, registering the J two-dimensional coordinate points to the two-dimensional template point set corresponding to the two-dimensional coordinate points to obtain a rotation matrix corresponding to the two-dimensional coordinate points;

[0038] S233, multiplying the J two-dimensional coordinate points by the rotation matrix respectively to obtain J two-dimensional registration coordinate points corresponding to the two-dimensional coordinate points;

[0039] S234, performing convex hull detection processing on the J two-dimensional registration coordinate points to obtain a two-dimensional vertex set corresponding to the two-dimensional registration coordinate points; the two-dimensional vertex set comprises a plurality of two-dimensional vertices;

[0040] S235, setting the two-dimensional feature point as the lowermost two-dimensional vertex in the two-dimensional vertex set;

[0041] S236, processing the rotation matrix, the two-dimensional feature point and the conversion coordinate information by using a three-dimensional feature point calculation model to obtain the three-dimensional feature point corresponding to the tooth information;

[0042] The three-dimensional feature point calculation model is:

[0043]

[0044]

[0045] In the formula, Q is a vector composed of three-dimensional coordinate values of the three-dimensional feature point; is an inverse matrix of the rotation matrix; A two-dimensional column vector consisting of the two-dimensional coordinates of the two-dimensional feature points; O, and , which are the column vector, the first unit vector, and the second unit vector formed by the three-dimensional coordinates of the selected origin, respectively.

[0046] As an optional implementation, in the first aspect of the present invention, processing the compressed index matrix and the voxel size information corresponding to the tooth information to obtain the corresponding transformed coordinate information and two-dimensional coordinate point set includes:

[0047] S2311. Process the compressed index matrix and the voxel size information to obtain a mask point set; the mask point set includes... I One mask point;

[0048] S2312, will I Each of the mask points is projected onto the line containing the first vector to obtain... I The first projection point;

[0049] S2313, Set the cutting point to I The midpoint between the two farthest first projection points;

[0050] S2314, will I In each of the aforementioned mask points, the corresponding voxel intersects with the cutting plane. J The mask points are combined to form a set of cutting points; the cutting surface is a plane that passes through the cutting points and whose normal vector is the first vector.

[0051] S2315. Set the cutting points together. J Each of the mask points is projected onto the cutting surface to obtain a corresponding second projection point;

[0052] S2316. Based on the cutting plane, respectively... J The second projection point is subjected to coordinate transformation to obtain the corresponding transformed coordinate information and a two-dimensional coordinate point set.

[0053] As an optional implementation, in the first aspect of the present invention, processing each piece of tooth information and the corresponding vector set to obtain a set of classification image pairs corresponding to each piece of tooth information includes:

[0054] S31. Process the local voxel matrix and voxel size information of the tooth information to obtain the three-dimensional center point;

[0055] S32, N1 sampling points are arranged at equal intervals on a main axis; the main axis is a straight line along the first vector and passing through the three-dimensional center point; N1 is an integer greater than 1;

[0056] S33, N1 sampling plane pairs are arranged based on the N1 sampling points, the second vector and the third vector; the sampling plane pair includes a first sampling plane and a second sampling plane;

[0057] S34, the three-dimensional voxel region corresponding to the local voxel matrix is sampled by using the N1 sampling plane pairs respectively, to obtain N1 initial image pairs; the initial image pair includes a first initial image and a second initial image;

[0058] S35, N1 enhanced image pairs are obtained by using an image enhancement model to enhance the N1 initial image pairs; the enhanced image pair includes a first enhanced image and a second enhanced image;

[0059] S36, N1 classification image pairs are obtained by respectively performing size adjustment processing on the N1 enhanced image pairs; the N1 classification image pairs constitute the classification image pair set corresponding to the tooth information.

[0060] As an optional implementation, in the first aspect of the embodiment, the multi-scale classification and recognition model includes an image preprocessing module, a global feature generation module, a local feature generation module, a feature fusion module and a classification module;

[0061] The image preprocessing module is in data connection with the global feature generation module and the local feature generation module, and is configured to preprocess the classification image pair to obtain a preprocessed image;

[0062] The global feature generation module is in data connection with the feature fusion module, and is configured to process the preprocessed image to obtain a global feature map;

[0063] The local feature generation module is in data connection with the feature fusion module, and is configured to process the preprocessed image to obtain a local feature map;

[0064] The feature fusion module is in data connection with the classification module, and is configured to perform feature fusion processing on the global feature map and the local feature map to obtain a fusion feature map;

[0065] The classification module is configured to perform classification processing on the fusion feature map to obtain the probability vector.

[0066] As an optional implementation, in the first aspect of the embodiment of the present application, the feature fusion module comprises an initial fusion unit and R intermediate fusion units; R is an integer greater than 2;

[0067] The initial fusion unit is configured to process the global feature map and the local feature map to obtain an initial fusion feature map.

[0068] The intermediate fusion unit is configured to process the global feature map, the local feature map and a previous stage feature map to obtain an intermediate fusion feature map.

[0069] The previous stage feature map of the first intermediate fusion unit is the initial fusion feature map, and the previous stage feature map of the last R-1 intermediate fusion units is the intermediate fusion feature map output by the previous intermediate fusion unit.

[0070] As an optional implementation, in the first aspect of the embodiment of the present application, the intermediate fusion unit comprises a channel attention subunit, a first multiplication subunit, a spatial attention subunit, a second multiplication subunit, a first splicing subunit, a first convolution subunit, an average pooling subunit, a second splicing subunit, a second convolution subunit, a residual subunit and an addition subunit.

[0071] The input end of the channel attention subunit inputs the global feature map; the output end of the channel attention subunit is connected to the first input end of the first multiplication subunit; the second input end of the first multiplication subunit inputs the global feature map; and the output end of the first multiplication subunit is connected to the first input end of the first splicing subunit.

[0072] The input end of the spatial attention subunit inputs the local feature map; the output end of the spatial attention subunit is connected to the first input end of the second multiplication subunit; the second input end of the second multiplication subunit inputs the local feature map; and the output end of the second multiplication subunit is connected to the second input end of the first splicing subunit.

[0073] The input end of the first convolution subunit of the first intermediate fusion unit is connected to the output end of the initial fusion unit; and the input end of the first convolution subunit of the last R-1 intermediate fusion units is connected to the output end of the previous intermediate fusion unit.

[0074] The output end of the first convolution subunit is connected to the input end of the average pooling subunit; the output end of the average pooling subunit is respectively connected to the first input end of the second splicing subunit and the first input end of the addition subunit; the second input end and the third input end of the second splicing subunit input the global feature map and the local feature map respectively.

[0075] The output end of the second splicing subunit is connected to the input end of the second convolution subunit; the output end of the second convolution subunit is connected to the third input end of the first splicing subunit; the output end of the first splicing subunit is connected to the input end of the residual subunit; the output end of the residual subunit is connected to the second input end of the addition subunit; and the output end of the addition subunit is the output end of the feature fusion module.

[0076] The second aspect of the present application discloses a computer storage medium, the computer storage medium stores computer instructions, when the computer instructions are called, part or all steps of the maxillary anterior tooth alveolar bone defect identification method disclosed in the first aspect of the embodiment of the present application are executed.

[0077] Compared with the prior art, the embodiment of the present application has the following beneficial effects:

[0078] The present application processes each tooth information based on voxel size information, obtains a corresponding vector set, and further obtains a corresponding classification image pair set; a multi-scale classification identification model is used to process each classification image pair set, obtain a corresponding probability vector set, and further obtain a corresponding classification identification result, so as to get rid of artificial dependence and automatically complete high-precision classification identification of alveolar bone defects. BRIEF DESCRIPTION OF DRAWINGS

[0079] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be activated in the embodiment description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0080] Figure 1 is a flowchart of a maxillary anterior tooth alveolar bone defect identification method disclosed by the embodiment of the present application.

[0081] Figure 2 is a structural diagram of a multi-scale classification identification model of a maxillary anterior tooth alveolar bone defect identification method disclosed by the embodiment of the present application.

[0082] Figure 3 is a structural diagram of a global feature generation module of a multi-scale classification identification model of a maxillary anterior tooth alveolar bone defect identification method disclosed by the embodiment of the present application.

[0083] Figure 4 is a structural diagram of a local feature generation module of a multi-scale classification identification model of a maxillary anterior tooth alveolar bone defect identification method disclosed by the embodiment of the present application.

[0084] Figure 5 is a structural schematic view of a feature fusion module of a multi-scale classification recognition model of a maxillary anterior tooth alveolar bone defect recognition method disclosed by the embodiment of the application.

[0085] Figure 6 is a structural schematic view of an intermediate fusion unit of a multi-scale classification recognition model of a maxillary anterior tooth alveolar bone defect recognition method disclosed by the embodiment of the application.

[0086] Figure 7 is a structural schematic view of a maxillary anterior tooth alveolar bone defect recognition device disclosed by the embodiment of the application.

[0087] Figure 8 is a structural schematic view of another maxillary anterior tooth alveolar bone defect recognition device disclosed by the embodiment of the application. DETAILED DESCRIPTION

[0088] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the application.

[0089] In the description of the application, it should be noted that the orientations or positional relationships indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. are based on the orientations or positional relationships shown in the drawings, and are only for the purpose of facilitating the description of the application and simplifying the description, and therefore cannot be understood as indicating or implying that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the application. In addition, the terms "first", "second", "third" are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0090] In the description of the application, it should be noted that, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the communication inside two elements. For a person of ordinary skill in the art, the specific meaning of the above terms in the application can be understood according to the specific circumstances.

[0091] Embodiment one

[0092] Please refer to Figures 1-6 . Figure 1is a flowchart of a maxillary anterior tooth alveolar bone defect recognition method disclosed by the embodiment of the present application. Figure 2 is a structural diagram of a multi-scale classification recognition model of a maxillary anterior tooth alveolar bone defect recognition method disclosed by the embodiment of the present application. Figure 3 is a structural diagram of a global feature generation module of a multi-scale classification recognition model of a maxillary anterior tooth alveolar bone defect recognition method disclosed by the embodiment of the present application. Figure 4 is a structural diagram of a local feature generation module of a multi-scale classification recognition model of a maxillary anterior tooth alveolar bone defect recognition method disclosed by the embodiment of the present application. Figure 5 is a structural diagram of a feature fusion module of a multi-scale classification recognition model of a maxillary anterior tooth alveolar bone defect recognition method disclosed by the embodiment of the present application. Figure 6 is a structural diagram of an intermediate fusion unit of a multi-scale classification recognition model of a maxillary anterior tooth alveolar bone defect recognition method disclosed by the embodiment of the present application. Figure 1 The maxillary anterior tooth alveolar bone defect recognition method described is applied to the field of tooth damage detection, such as recognition of the defect condition of alveolar bone, and the embodiment of the present application is not limited thereto. Figure 1 As shown, the method comprises:

[0093] S1, obtain a global voxel matrix and voxel size information; segment the global voxel matrix to obtain N pieces of tooth information; N is an integer from 1 to 4.

[0094] It should be noted that, in order to obtain the global voxel matrix and the voxel size information, first, the maxillary anterior teeth of a patient are subjected to CBCT scanning to obtain a set of two-dimensional gray scale tomographic images (such as axial, coronal and sagittal images), which are stored in DICOM format. Then, through a reconstruction algorithm, the two-dimensional tomographic images are converted into a global voxel matrix, each element of the matrix corresponding to a cuboid voxel, and the element value representing the X-ray attenuation coefficient of the human tissue in the cuboid voxel.

[0095] The voxel size information includes a first voxel size, a second voxel size and a third voxel size.

[0096] It should be noted that the first voxel size, the second voxel size and the third voxel size represent the length of each cuboid voxel in the X, Y and Z axis directions of the preset global coordinate system, respectively.

[0097] The tooth information includes tooth position number, local voxel matrix and mask matrix.

[0098] It should be noted that the above tooth position number is a standardized numbering system for accurately identifying the position of the tooth in stomatology. Its core function is to clearly and uniformly refer to each tooth in medical records, diagnosis and treatment communication, academic exchanges and other scenarios to avoid confusion. The above tooth position number can use the ISO system, ADA system or Palmer system, etc. The embodiments of the present application are not limited.

[0099] The above global voxel matrix, the above local voxel matrix and the above mask matrix are all three-dimensional matrices, and the size of the above local voxel matrix and the above mask matrix is completely consistent; the value of the element in the above mask matrix is 0 or 1.

[0100] It should be noted that the size of the above local voxel matrix and the mask matrix is completely consistent, that is, the number of rows, columns and depth of the two are equal.

[0101] It should be noted that the local voxel matrix in the above tooth information is a sub-matrix corresponding to the region where the tooth corresponding to the tooth position number of the tooth information in the global voxel matrix is located; when the element value of the above mask matrix is 1 or 0, it respectively represents that the cuboid voxel corresponding to the corresponding element in the local voxel matrix belonging to the same tooth information represents tooth tissue or does not represent tooth tissue.

[0102] Optionally, the above segmentation processing can be realized by processing the global voxel matrix based on a neural network based on a 3D U-Net architecture. This method is a known method, and the embodiments of the present application will not be repeated.

[0103] S2, based on the above voxel size information, processing each of the above tooth information to obtain a vector set corresponding to each of the above tooth information.

[0104] It should be noted that the N tooth information and the N vector set are one-to-one corresponding.

[0105] The above vector set includes a first vector, a second vector and a third vector.

[0106] It should be noted that the above first vector is used as the normal vector of the transverse section of the corresponding tooth; the above second vector is used as the normal vector of the coronal section of the corresponding tooth; and the above third vector is used as the normal vector of the sagittal section of the corresponding tooth.

[0107] S3, processing each of the above tooth information and the corresponding above vector set to obtain a classification image pair set corresponding to each of the above tooth information.

[0108] The above classification image pair set includes a plurality of classification image pairs; and the above classification image pair includes a first classification image and a second classification image.

[0109] S4, processing each of the classification image pair set corresponding to the above tooth information by using the pre-trained multi-scale classification recognition model to obtain a probability vector set corresponding to each of the tooth information.

[0110] The probability vector set includes a probability vector corresponding to each of the classification image pairs in the classification image pair set; and the probability vector has a dimension of 5.

[0111] It should be noted that the multi-scale classification recognition model processes each of the classification image pairs in the classification image pair set to obtain a probability vector corresponding to each of the classification image pairs.

[0112] S5, processing each of the probability vector set corresponding to the tooth information to obtain a classification recognition result corresponding to each of the tooth information; and the classification recognition result has an integer value of 1 to 5.

[0113] It should be noted that the classification recognition result is used to represent the degree of cracking of the maxillary anterior teeth, thereby reflecting the alveolar bone defect. When the recognition result is 1, 2, 3, 4 or 5, it respectively represents that the tooth corresponding to the tooth position number in the corresponding tooth information exists normally, has mild bone cracking, has moderate bone cracking, has severe bone cracking, and has bone windowing.

[0114] In an optional embodiment, the processing of each of the tooth information based on the voxel size information to obtain a vector set corresponding to each of the tooth information includes:

[0115] S21, a set of preset template information; the template information set includes M template information; and the template information includes a tooth position number and a two-dimensional template point set.

[0116] Preferably, M is 4. The tooth position numbers in the four template information correspond to four maxillary anterior teeth.

[0117] It should be noted that the two-dimensional template point set in each template information is a set of a plurality of two-dimensional points on a cross-sectional (i.e., horizontal cross-sectional) image at a distance of 1 / 2 from the root tip of the tooth corresponding to the tooth position number of the template information of any patient.

[0118] It should be noted that the two-dimensional coordinate system (defined by mutually perpendicular u-axis and v-axis) corresponding to each two-dimensional template point set is different, wherein the v-axis is the left-right symmetry axis of the corresponding cross-sectional image, and points away from the lingual side of the corresponding tooth.

[0119] S22, processing the mask matrix of each of the tooth information based on the voxel size information to obtain a compressed index matrix and the first vector corresponding to each of the tooth information.

[0120] S23, based on the voxel size information, processing the compression index matrix corresponding to each tooth information and the two-dimensional template point set to obtain the three-dimensional feature point corresponding to each tooth information.

[0121] It should be noted that a certain tooth information corresponds to a certain two-dimensional template point set, which means that the tooth position numbers contained by the two are the same.

[0122] S24, processing the first vector corresponding to each tooth information and the three-dimensional feature point to obtain the second vector and the third vector corresponding to each tooth information; the first vector, the second vector and the third vector corresponding to each tooth information constitute the corresponding vector set.

[0123] In another optional embodiment, based on the voxel size information, the mask matrix of each tooth information is processed to obtain the compression index matrix corresponding to each tooth information and the first vector, comprising:

[0124] S221, compressing the mask matrix to obtain the corresponding compression index matrix.

[0125] It should be noted that the compression processing of the mask matrix is to splice the indexes of all elements with value 1 in the mask matrix in X, Y and Z axis directions as row vectors to obtain the compression index matrix. ix , iy , iz

[0126] S222, using the standard mask matrix construction model to process the compression index matrix and the voxel size information to construct the corresponding standard mask matrix.

[0127] The standard mask matrix construction model is:

[0128]

[0129] In the formula, and are the elements of the first row and the first column of the standard mask matrix and the compression index matrix, respectively; i is an integer from 1 to j i is an integer from 1 to I I is the total number of rows of the compression index matrix; j is 1, 2 or 3; , and are the first voxel size, the second voxel size and the third voxel size, respectively.​​​

[0130] S223, using the feature matrix calculation model, processing the above standard mask matrix to obtain the corresponding feature matrix.

[0131] The feature matrix calculation model is:

[0132]

[0133] In the formula, C is the feature matrix, and B is the standard mask matrix.

[0134] S224, performing eigenvalue decomposition processing on the feature matrix to obtain three feature pairs; the feature pair includes an eigenvalue and a corresponding feature vector.

[0135] S225, normalizing the corresponding feature vector of the largest eigenvalue in the three feature pairs to obtain the corresponding first vector.

[0136] In another optional embodiment, based on the voxel size information, the compressed index matrix corresponding to each tooth information and the two-dimensional template point set are processed to obtain the three-dimensional feature point corresponding to each tooth information, including:

[0137] S231, processing the compressed index matrix corresponding to the tooth information and the voxel size information to obtain the corresponding conversion coordinate information and two-dimensional coordinate point set.

[0138] The conversion coordinate information includes a selected origin, a first unit vector and a second unit vector; the two-dimensional coordinate point set includes J two-dimensional coordinate points; J is an integer greater than 1.

[0139] It should be noted that the selected origin is a three-dimensional coordinate point in the global coordinate system; the first unit vector and the second unit vector are both three-dimensional vectors in the global coordinate system.

[0140] S232, registering J two-dimensional coordinate points to the corresponding two-dimensional template point set to obtain a rotation matrix.

[0141] Optionally, the J two-dimensional coordinate points are registered to the corresponding two-dimensional template point set by a 2D-ICP algorithm to obtain a rotation matrix.

[0142] It should be noted that the rotation matrix is a 2x2 matrix.

[0143] S233, registering JThe two-dimensional coordinate points are respectively multiplied by the rotation matrix to obtain corresponding J two-dimensional registration coordinate points.

[0144] It should be noted that the multiplication of the two-dimensional coordinate points and the rotation matrix means that the two-dimensional coordinate values of the two-dimensional coordinate points form a two-dimensional column vector, which is multiplied by the rotation matrix. After rotation, the coordinate system of the two-dimensional registration coordinate points is the same as that of the two-dimensional template point set.

[0145] S234, the convex hull detection processing is performed on the J two-dimensional registration coordinate points to obtain a corresponding two-dimensional vertex set; the two-dimensional vertex set includes a plurality of two-dimensional vertices.

[0146] It should be noted that the convex hull detection can use the Graham algorithm, and the embodiments of the present application are not limited. The vertex set is a set of vertices of the smallest polygon (i.e. convex hull) surrounding the J two-dimensional registration coordinate points.

[0147] S235, the two-dimensional feature point is set as the lowermost two-dimensional vertex in the two-dimensional vertex set.

[0148] It should be noted that the lowermost vertex in the two-dimensional vertex set refers to the two-dimensional vertex with the smallest v-axis coordinate.

[0149] S236, using a three-dimensional feature point calculation model, the rotation matrix, the two-dimensional feature point and the conversion coordinate information are processed to obtain corresponding three-dimensional feature points.

[0150] The three-dimensional feature point calculation model is:

[0151]

[0152]

[0153] In the formula, Q is a vector formed by the three-dimensional coordinate values of the three-dimensional feature points; is an inverse matrix of the rotation matrix; is a two-dimensional column vector formed by the two-dimensional coordinate values of the two-dimensional feature points; O, and are column vectors formed by the three-dimensional coordinate values of the selected origin, the first unit vector and the second unit vector, respectively.

[0154] In another optional embodiment, the processing of the compression index matrix corresponding to the tooth information and the voxel size information to obtain the conversion coordinate information and the two-dimensional coordinate point set includes:

[0155] S2311, processing the compressed index matrix and the voxel size information to obtain a mask point set; the mask point set includes I one mask point.

[0156] It should be noted that the processing of the compressed index matrix and the voxel size information is to multiply the first column, the second column and the third column of the compressed index matrix by the first voxel size, the second voxel size and the third voxel size respectively to obtain a local coordinate matrix D, each row of D corresponds to one mask point in the mask point set.

[0157] S2312, projecting the I mask points onto the straight line where the first vector is located to obtain I first projection points.

[0158] S2313, setting the sectioning point as the midpoint of the two I first projection points that are farthest apart.

[0159] S2314, combining the I mask points in the mask point set, which correspond to voxels intersecting the sectioning plane, as a sectioning point set; the sectioning plane is a plane passing through the sectioning points and having a normal vector being the first vector. J

[0160] It should be noted that the voxel corresponding to the mask point refers to a cuboid voxel with the mask point as the center.

[0161] S2315, projecting the J mask points in the sectioning point set onto the sectioning plane to obtain corresponding second projection points.

[0162] S2316, based on the sectioning plane, performing coordinate conversion processing on the J second projection points to obtain corresponding conversion coordinate information and a two-dimensional coordinate point set.

[0163] In another optional embodiment, based on the sectioning plane, performing coordinate conversion processing on the J second projection points to obtain corresponding conversion coordinate information and a two-dimensional coordinate point set, including:

[0164] S23161, setting the selected origin as any point on the sectioning plane that is different from the J second projection points.

[0165] S23162, selecting two orthogonal unit vectors and on the sectioning plane.​, respectively as the first unit vector and the second unit vector.

[0166] S23163, using a two-dimensional coordinate calculation model, processing the selected origin, the first unit vector, the second unit vector, and each second projection point to obtain a two-dimensional coordinate point corresponding to each second projection point.

[0167] The two-dimensional coordinate calculation model is:

[0168]

[0169] In the formula, is the two-dimensional coordinate value of the i-th two-dimensional coordinate point; P is a vector composed of the three-dimensional coordinate values of the i-th second projection point; O is a vector composed of the three-dimensional coordinate values of the selected origin; ii ii represents the inner product operation of the vector.

[0170] In another optional embodiment, the processing of each tooth information corresponding to the first vector and the three-dimensional feature point to obtain the second vector and the third vector corresponding to each tooth information comprises:

[0171] S241, setting the construction point as any point on the straight line where the first vector is located.

[0172] S242, using a second vector calculation model to process the construction point and the three-dimensional feature point to obtain the corresponding second vector.

[0173] The second vector calculation model is:

[0174]

[0175] In the formula, and are the second vector and the first vector corresponding to the tooth information, respectively; and are column vectors composed of the three-dimensional coordinate values of the three-dimensional feature point and the construction point corresponding to the tooth information, respectively.

[0176] S243, calculating the cross product of the first vector and the corresponding second vector to obtain the corresponding third vector.

[0177] In another optional embodiment, the processing of each tooth information and the corresponding vector set to obtain a classification image pair set corresponding to each tooth information comprises:

[0178] ​​S31, processing the local voxel matrix and the voxel size information of the tooth information to obtain a three-dimensional center point.

[0179] Optionally, the three-dimensional center point coordinate is (ia, ib, ic), where , , , ia, ib, and ic are respectively the element number of the local voxel matrix along the X, Y, and Z axis directions. II , JJ and KK .

[0180] S32, setting N sampling points on the main axis at equal intervals; the main axis is a straight line along the first vector and passing through the three-dimensional center point; N is an integer greater than 1.

[0181] Preferably, the interval of the adjacent sampling points is 0.25mm.

[0182] Preferably, the N sampling points include the three-dimensional center point, and N sampling points on both sides of the three-dimensional center point, N is an integer greater than 1. .

[0183] S33, based on the N sampling points, the second vector, and the third vector, setting N sampling plane pairs; the sampling plane pair includes a first sampling plane and a second sampling plane. It should be noted that the first sampling plane and the second sampling plane of the nth sampling plane pair both pass through the nth sampling point, and respectively take the corresponding second vector and the third vector as the normal vector.

[0184] S34, respectively using the N sampling plane pairs to sample the three-dimensional voxel region corresponding to the local voxel matrix to obtain N initial image pairs; the initial image pair includes a first initial image and a second initial image.

[0185] It should be noted that the three-dimensional voxel region corresponding to the local voxel matrix is a cuboid region, including a cuboid voxel with each element in the local voxel matrix as the center, and the center coordinate of the cuboid voxel corresponding to the element with index (ia, ib, ic) in the local voxel matrix is . In addition, the voxel value of each cuboid voxel is the corresponding element value in the local voxel matrix.

[0186]

[0187] ​​​​​​​It should be noted that the above sampling process utilizes trilinear interpolation, which slices the three-dimensional voxel region corresponding to the local voxel matrix based on the first and second sampling planes of each sampling plane pair, respectively, to obtain an initial image pair corresponding to each sampling plane pair. The first initial image is obtained by slicing from the first sampling plane, and the second initial image is obtained by slicing from the second sampling plane. The above trilinear interpolation method is a known method, and will not be described further in this embodiment.

[0188] S35. Using image enhancement models, for The above initial image pairs are enhanced to obtain The above-mentioned enhanced image pair includes a first enhanced image and a second enhanced image.

[0189] S36, respectively for The above-mentioned enhanced image pairs are resized to obtain... The above-described image pairs; The aforementioned image pairs constitute the set of the aforementioned image pairs corresponding to the aforementioned tooth information.

[0190] It should be noted that the above-described resizing process involves scaling all the first and second enhanced images to obtain first and second classification images of the same size. For example, the first enhanced image, after scaling, yields a first classification image with a size of 150×224 pixels; the second enhanced image, after scaling, yields a second classification image with the same size of 150×224 pixels.

[0191] In yet another optional embodiment, the expression for the above image enhancement model is:

[0192]

[0193]

[0194]

[0195]

[0196]

[0197]

[0198] In the formula, l 2 to L Integers of -1 ; and The first ll The first initial image and the second initial image of the aforementioned initial image pair; and are respectively a first coefficient, a second coefficient and a third coefficient preset, and ll is an integer from 1 to 3. ll , L , , and are respectively a first coefficient, a second coefficient and a third coefficient preset, and .

[0199] Preferably, the first coefficient, the second coefficient and the third coefficient are respectively 0.3, 0.4 and 0.3. , , are respectively a first coefficient, a second coefficient and a third coefficient preset, and

[0200] In yet another optional embodiment, as shown in FIG. 6, the multi-scale classification recognition model comprises an image preprocessing module, a global feature generation module, a local feature generation module, a feature fusion module and a classification module. Figure 2

[0201] The image preprocessing module is in data connection with the global feature generation module and the local feature generation module, and is configured to preprocess the classification image pair to obtain a preprocessed image.

[0202] It should be noted that the preprocessing of the classification image pair is to firstly splice the first classification image and the second classification image in the classification image pair left and right to obtain a spliced image, and then scale the spliced image to obtain the preprocessed image. For example, the first classification image and the second classification image each having a size of 150x224 pixels are spliced left and right to obtain a spliced image having a size of 300x224 pixels, and the spliced image is further scaled to obtain a preprocessed image having a size of 224x224 pixels.

[0203] The global feature generation module is in data connection with the feature fusion module, and is configured to process the preprocessed image to obtain a global feature map.

[0204] The local feature generation module is in data connection with the feature fusion module, and is configured to process the preprocessed image to obtain a local feature map.

[0205] The feature fusion module is in data connection with the classification module, and is configured to perform feature fusion processing on the global feature map and the local feature map to obtain a fused feature map.

[0206] The classification module is configured to perform classification processing on the fused feature map to obtain the probability vector.

[0207] ​It should be noted that the multi-scale classification recognition model has an accuracy of 92.9% and an F1 score of 89.8% on the test set, which is 5%-8% higher than traditional models (CNN, Swin Transformer, etc.).

[0208] In yet another optional embodiment, as shown in Figure 3 The global feature generation module includes a first downsampling unit and a global feature extraction unit.

[0209] The first downsampling unit is data-connected to the global feature extraction unit and is configured to perform first downsampling processing on the spliced image to obtain a first downsampling image.

[0210] It should be noted that the first downsampling is achieved through patch merging. For example, by performing 4 times first downsampling on a spliced image with a size of 224x224 pixels, a first downsampling image with a size of 56x56 pixels is obtained.

[0211] The global feature extraction unit is configured to process the first downsampling image to obtain the global feature map.

[0212] Optionally, the global feature extraction unit is constructed based on a Transformer architecture.

[0213] It should be noted that the global feature extraction unit can dynamically model the global interaction between different patches, capture the dependency relationship of long-distance pixels / regions, and achieve global semantic extraction (such as the boundary continuity of the sagittal root form and the coronal alveolar bone).

[0214] In yet another optional embodiment, as shown in Figure 4 The local feature generation module includes a second downsampling unit and a local feature extraction unit.

[0215] The second downsampling unit is data-connected to the local feature extraction unit and is configured to perform second downsampling processing on the spliced image to obtain a second downsampling image.

[0216] It should be noted that the second downsampling is achieved through convolution operation (kernel size 4x4, step 2). For example, by performing second downsampling on a spliced image with a size of 224x224 pixels, a local feature map with a size of 112x112 pixels is obtained.

[0217] The local feature extraction unit is configured to process the second downsampling image to obtain the local feature map.

[0218] Optionally, the local feature extraction unit is constructed based on a ResNet architecture, thereby efficiently extracting multi-scale local features while maintaining gradient flow, enhancing detail capture and preventing degradation.

[0219] In yet another optional embodiment, as shown in Figure 5 The feature fusion module includes an initial fusion unit and R intermediate fusion units, where R is an integer greater than 2.

[0220] The initial fusion unit is configured to process the global feature map and the local feature map to obtain an initial fusion feature map.

[0221] The intermediate fusion unit is configured to process the global feature map, the local feature map, and a previous-level feature map to obtain an intermediate fusion feature map.

[0222] The previous-level feature map of the first intermediate fusion unit is the initial fusion feature map, and the previous-level feature map of the last R-1 intermediate fusion units is the intermediate fusion feature map output by the previous intermediate fusion unit.

[0223] Preferably, R is 3.

[0224] As can be seen, the feature fusion module gradually improves the expression ability of the output feature by multiple hierarchical fusions of the global feature map and the local feature map.

[0225] In yet another optional embodiment, as shown in Figure 6 The intermediate fusion unit includes a channel attention subunit, a first multiplication subunit, a spatial attention subunit, a second multiplication subunit, a first splicing subunit, a first convolution subunit, an average pooling subunit, a second splicing subunit, a second convolution subunit, a residual subunit, and an addition subunit.

[0226] The input end of the channel attention subunit inputs the global feature map, and the output end of the channel attention subunit is connected to the first input end of the first multiplication subunit.

[0227] It should be noted that the channel attention subunit can be constructed by a Squeeze-and-Excitation module, and the embodiments of the present application are not limited thereto. The channel attention subunit weights the channel dimension of the input feature map through global average pooling and a fully connected layer to generate channel attention weights, thereby enhancing the feature correlation between channels.

[0228] An input end of the spatial attention subunit is connected with the local feature map.

[0229] It should be noted that the spatial attention subunit can be constructed by a SAM structure in a CBAM (Convolutional Block Attention Module), and the embodiments of the present application are not limited thereto. The spatial attention subunit generates a spatial attention map by performing average pooling and maximum pooling operations along the channel dimension, so as to highlight the importance of key spatial positions in the feature map.

[0230] An input end of the first convolution subunit of the first intermediate fusion unit is connected with an output end of the initial fusion unit, and input ends of the first convolution subunits of the R-1 intermediate fusion units are connected with output ends of the previous intermediate fusion units.

[0231] It should be noted that the first convolution subunit is used for performing a 1x1 convolution operation, which can adjust the channel number of the input early-stage feature, reduce the calculation amount, and increase the feature expression capability.

[0232] An output end of the first convolution subunit is connected with an input end of the average pooling subunit, an output end of the average pooling subunit is connected with a first input end of the second concatenation subunit and a first input end of the addition subunit respectively, and a second input end and a third input end of the second concatenation subunit are respectively input with the global feature map and the local feature map.

[0233] It should be noted that the average pooling subunit is used for performing an average pooling operation on the feature map after the 1x1 convolution, so as to reduce the spatial resolution of the feature map and further extract global information.

[0234] An output end of the second concatenation subunit is connected with an input end of the second convolution subunit, an output end of the second convolution subunit is connected with a third input end of the first concatenation subunit, an output end of the first concatenation subunit is connected with an input end of the residual subunit, an output end of the residual subunit is connected with a second input end of the addition subunit, and an output end of the addition subunit is used as an output end of the feature fusion module.

[0235] It should be noted that the residual subunit is constructed based on an Inverted Residual MLP (IRMLP) and is used to enhance the feature expression capability and relieve gradient vanishing or explosion, and an input-output relationship thereof is as follows, where LN represents a layer normalization operation, and Conv1 and Conv3 represent 1x1 and 3x3 convolution operations respectively:

[0236]

[0237] It should be noted that the second convolution subunit is used to perform a 3x3 convolution operation.

[0238] It can be seen that the intermediate fusion unit accurately mines important information of different dimensions of features through global (channel attention) and local (spatial attention) branches, and improves the feature discrimination. In addition, through a hierarchical fusion mechanism, the original multi-source features are first spliced (through the second splicing subunit) and refined through convolution; and the weighted enhanced features are then spliced (through the first splicing subunit) and deeply fused through the IRMLP to ensure the feature richness and expression.

[0239] In yet another optional embodiment, the initial fusion unit is obtained by removing the first convolution subunit, the average pooling subunit, the second splicing subunit, and the second convolution subunit in the intermediate fusion unit.

[0240] In yet another optional embodiment, the classification module includes, in sequence, a pooling layer, a layer normalization layer, a fully connected layer, and a Softmax layer.

[0241] The pooling layer is used to perform global average pooling (GAP) processing on the fusion feature map to obtain a compressed vector. For example, after GAP processing, the dimension of the compressed vector can be 1x1x512.

[0242] The layer normalization layer is used to perform layer normalization (Layer Normalization) processing on the compressed vector to obtain a layer normalization vector.

[0243] The fully connected layer is used to perform class mapping processing on the layer normalization vector to obtain a recognition vector with a dimension of 5.

[0244] The Softmax layer is used to process the recognition vector to obtain a probability vector with a dimension of 5.

[0245] It should be noted that the Softmax layer processes the recognition vector through the following expression, where is the i-th element of the probability vector, representing the probability value that the recognition result is jj . jj ​and The first and second are the recognition vectors respectively. jj The and the first kk Each element.

[0246]

[0247] In yet another optional embodiment, the loss function expression used for pre-training the above-mentioned multi-scale classification and recognition model is:

[0248]

[0249] In the formula, L This represents the total loss value. The probability vector is the first k Each element value; For all training samples, the corresponding classification and recognition result is actually: The quantity.

[0250] As can be seen, by using the above loss function, when the number of samples for the five classification results is unbalanced (for example, there are few samples of moderate bone fracture), the weight of the recognition result with fewer training samples can be increased, thereby increasing its contribution to the total loss value and thus improving the classification ability of the multi-scale classification recognition model after pre-training.

[0251] In another optional embodiment, the above-mentioned processing of the probability vector set corresponding to each of the above-mentioned tooth information to obtain the classification and recognition result corresponding to each of the above-mentioned tooth information includes:

[0252] S51. For each of the above-mentioned tooth information corresponding to the above-mentioned probability vector set, execute S52~S53.

[0253] S52. Take each of the above probability vectors in the above probability vector set as a row vector and concatenate them to obtain a probability matrix.

[0254] S53. Set the above classification and recognition result as the column number of the largest element in the above probability matrix.

[0255] It is evident that the method for identifying alveolar bone defects in maxillary anterior teeth described in the embodiments of the present invention can eliminate the need for manual intervention and automatically complete the high-precision classification and identification of alveolar bone defects.

[0256] Example 2

[0257] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of a maxillary anterior alveolar bone defect identification device disclosed in an embodiment of the present invention. Figure 7 The described device can be applied to the field of tooth damage detection, such as the identification of alveolar bone defects; however, the embodiments of this invention are not limited to this.Figure 7 As shown in the figure, the device can include:

[0258] An information acquisition module 201 is configured to acquire a global voxel matrix and voxel size information, and perform segmentation processing on the global voxel matrix to obtain N pieces of tooth information.

[0259] A vector generation module 202 is configured to process each piece of tooth information based on the voxel size information to obtain a vector set corresponding to each piece of tooth information.

[0260] A classification image generation module 203 is configured to process each piece of tooth information and the corresponding vector set to obtain a classification image pair set corresponding to each piece of tooth information.

[0261] A probability vector generation module 204 is configured to process the classification image pair set corresponding to each piece of tooth information by using a pre-trained multi-scale classification recognition model to obtain a probability vector set corresponding to each piece of tooth information.

[0262] A classification result generation module 205 is configured to process the probability vector set corresponding to each piece of tooth information to obtain a classification recognition result corresponding to each piece of tooth information.

[0263] As can be seen, the maxillary anterior tooth alveolar bone defect recognition device described in the embodiment of the present application can automatically complete high-precision classification and recognition of alveolar bone defect conditions without relying on artificial.

[0264] Embodiment three

[0265] Please refer to Figure 8 , Figure 8 is a structural schematic diagram of another maxillary anterior tooth alveolar bone defect recognition device disclosed in the embodiment of the present application. Among them, Figure 8 The maxillary anterior tooth alveolar bone defect recognition device described in the embodiment of the present application can be applied to the field of tooth damage detection, such as recognition of alveolar bone defect conditions, and the embodiment of the present application is not limited to this. For example Figure 8 As shown in the figure, the maxillary anterior tooth alveolar bone defect recognition device can include the following parts:

[0266] A memory 301 in which executable program codes are stored.

[0267] A processor 302 coupled with the memory 301.

[0268] The processor 302 invokes the executable program codes stored in the memory 301, and is configured to execute the steps in the maxillary anterior tooth alveolar bone defect recognition method described in the embodiment one.

[0269] Embodiment four

[0270] The embodiment of the present application discloses a computer readable storage medium, which stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute steps in the maxillary anterior tooth alveolar bone defect identification method described in the embodiment one.

[0271] The above-described device embodiments are only illustrative, wherein the modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical modules, i.e., can be located in one place or distributed to multiple network modules. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0272] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and the necessary general hardware platform, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, and the computer software product can be stored in a computer readable storage medium, including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disk storage, a magnetic disk storage, a magnetic tape storage, or any other computer readable medium that can be used to carry or store data.

[0273] It should be finally pointed out that: the maxillary anterior tooth alveolar bone defect recognition method and device disclosed by the embodiment of the application disclosed only the preferred embodiment of the application, only for the description of the technical solutions of the application, not to limit; although the application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand; the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application.

Claims

1. A method of identifying maxillary anterior dental alveolar bone defects, characterized in that, The method comprises the following steps: S1, obtaining a global voxel matrix and voxel size information; segmenting the global voxel matrix to obtain N pieces of tooth information; N is an integer from 1 to 4; each element in the global voxel matrix corresponds to a cuboid voxel; the voxel size information comprises a first voxel size, a second voxel size and a third voxel size; the first voxel size, the second voxel size and the third voxel size respectively represent the length of each cuboid voxel in the X, Y and Z axis directions of a preset global coordinate system; the tooth information comprises a tooth position number, a local voxel matrix and a mask matrix; the global voxel matrix, the local voxel matrix and the mask matrix are all three-dimensional matrices, and the size of the local voxel matrix and the mask matrix is completely consistent; the value of an element in the mask matrix is 0 or 1; S2, based on the voxel size information, processing each piece of tooth information to obtain a vector set corresponding to each piece of tooth information; the vector set comprises a first vector, a second vector and a third vector; the first vector is used as the normal vector of the transverse section of the corresponding tooth; the second vector is used as the normal vector of the coronal section of the corresponding tooth; and the third vector is used as the normal vector of the sagittal section of the corresponding tooth; S3, processing each piece of tooth information and the corresponding vector set to obtain a classification image pair set corresponding to each piece of tooth information; the classification image pair set comprises a plurality of classification image pairs; each classification image pair comprises a first classification image and a second classification image; S4, using a pre-trained multi-scale classification recognition model to process the classification image pair set corresponding to each piece of tooth information to obtain a probability vector set corresponding to each piece of tooth information; the probability vector set comprises a probability vector corresponding to each classification image pair in the corresponding classification image pair set; and the dimension of the probability vector is 5; S5, processing the probability vector set corresponding to each piece of tooth information to obtain a classification recognition result corresponding to each piece of tooth information; the value of the classification recognition result is an integer from 1 to 5.

2. The maxillary anterior alveolar bone defect identification method of claim 1, wherein, The processing of each piece of tooth information based on the voxel size information to obtain a vector set corresponding to each piece of tooth information comprises: S21, presetting a template information set; the template information set comprises M pieces of template information; each piece of template information comprises a tooth position number and a two-dimensional template point set; S22, based on the voxel size information, processing the mask matrix of each piece of tooth information to obtain a compressed index matrix and the first vector corresponding to each piece of tooth information; S23, based on the voxel size information, processing the compressed index matrix and the two-dimensional template point set corresponding to each piece of tooth information to obtain a three-dimensional feature point corresponding to each piece of tooth information. S24, processing the first vector and the three-dimensional feature point corresponding to each tooth information to obtain the second vector and the third vector corresponding to each tooth information; the first vector, the second vector and the third vector corresponding to each tooth information constitute a corresponding vector set. 3.The maxillary anterior alveolar bone defect identification method of claim 2, wherein, The processing of the mask matrix of each tooth information based on the voxel size information to obtain the compressed index matrix and the first vector corresponding to each tooth information comprises: S221, compressing the mask matrix to obtain a corresponding compressed index matrix; S222, using a standard mask matrix construction model to process the compressed index matrix and the voxel size information to obtain a corresponding standard mask matrix; The standard mask matrix construction model is: wherein and are the standard mask matrix and the compressed index matrix, respectively, the element in the i-th row and the j-th column of the standard mask matrix is i , j , i is an integer from 1 to I , I is the total number of rows of the compressed index matrix; j is 1, 2 or 3; , and are the first, second and third voxel size, respectively. S223, using a feature matrix calculation model to process the standard mask matrix to obtain a corresponding feature matrix; The feature matrix calculation model is: In the formula, C is the feature matrix, and B is the standard mask matrix; S224, performing eigenvalue decomposition processing on the feature matrix to obtain three feature pairs; the feature pair comprises an eigenvalue and a corresponding feature vector; S225, normalizing the largest eigenvalue and the corresponding feature vector in the three feature pairs to obtain the first vector. 4.The method of identifying maxillary anterior tooth alveolar bone defects according to claim 2, characterized in that, The processing of the compressed index matrix and the two-dimensional template point set corresponding to each tooth information based on the voxel size information to obtain the three-dimensional feature point corresponding to each tooth information comprises: S231, processing the compressed index matrix and the voxel size information corresponding to the tooth information to obtain corresponding conversion coordinate information and a two-dimensional coordinate point set; The conversion coordinate information includes a selected origin, a first unit vector, and a second unit vector; and the two-dimensional coordinate point set includes J a two-dimensional coordinate point; J is an integer greater than 1. S232, will J The two-dimensional coordinate points are registered to the corresponding two-dimensional template point set to obtain the corresponding rotation matrix; S233, will J Each of the given two-dimensional coordinate points is multiplied by the rotation matrix to obtain the corresponding... J Two-dimensional registration coordinate points; S234、to J a convex hull detection process is performed on the two-dimensional registration coordinate points to obtain a corresponding two-dimensional vertex set; the two-dimensional vertex set includes a plurality of two-dimensional vertices; S235, setting a two-dimensional feature point as the lowest two-dimensional vertex in the two-dimensional vertex set; S236, using a three-dimensional feature point calculation model to process the rotation matrix, the two-dimensional feature point and the conversion coordinate information to obtain the three-dimensional feature point; The three-dimensional feature point calculation model is: wherein Q is a vector of three-dimensional coordinate values of the three-dimensional feature points; is an inverse matrix of the rotation matrix; is a two-dimensional column vector of two-dimensional coordinate values of the two-dimensional feature points, and are a column vector of three-dimensional coordinate values of the selected origin, the first unit vector and the second unit vector, respectively.

5. The maxillary anterior alveolar bone defect identification method of claim 4, wherein, The processing of the compressed index matrix and the voxel size information corresponding to the tooth information to obtain the conversion coordinate information and the two-dimensional coordinate point set comprises: S2311. Process the compressed index matrix and the voxel size information to obtain a mask point set; the mask point set includes... I One mask point; S2312、the I masking points are projected onto the straight line where the first vector is located respectively to obtain I first projection points; S2313, set the sectioning point as I a midpoint of two first projection points farthest away from each other in the first projection points. S2314、the I one of the mask points, corresponding to a voxel intersecting the section plane J one of the mask points, combined as a section point set; the section plane is a plane passing through the section point and the normal vector is the first vector; S2315. Set the cutting points together. J Each of the mask points is projected onto the cutting surface to obtain a corresponding second projection point; S2316, based on the section, respectively, to J second projection point coordinate conversion processing, get the corresponding conversion coordinate information and two-dimensional coordinate point set. 6.The maxillary anterior alveolar bone defect identification method of claim 1, wherein, The processing of each tooth information and the corresponding vector set to obtain a classification image pair set corresponding to each tooth information comprises: S31, processing the local voxel matrix and the voxel size information of the tooth information to obtain a three-dimensional center point; S32, equally spacing N1 sampling points on the principal axis; the principal axis is a straight line along the first vector and passing through the three-dimensional center point; N1 is an integer greater than 1; S33, based on N1 sampling points, the second vector and the third vector, N1 sampling plane pairs are set; the sampling plane pair comprises a first sampling plane and a second sampling plane; S34, respectively, using N1 sampling planes, the corresponding three-dimensional voxel region of the local voxel matrix is sampled to obtain N1 initial image pairs; the initial image pair includes a first initial image and a second initial image; S35, using an image enhancement model, N1 initial image pairs are enhanced to obtain N1 enhanced image pairs; the enhanced image pair includes a first enhanced image and a second enhanced image; S36, respectively, N1 enhanced image pairs are size adjusted to obtain N1 classification image pairs; N1 classification image pairs constitute the classification image pair set corresponding to the tooth information. 7.The maxillary anterior alveolar bone defect recognition method according to claim 1, characterized in that, The multi-scale classification recognition model includes an image preprocessing module, a global feature generation module, a local feature generation module, a feature fusion module, and a classification module; The image preprocessing module is data connected with the global feature generation module and the local feature generation module, and is used for preprocessing the classification image pair to obtain a preprocessed image; The global feature generation module is data connected with the feature fusion module, and is used for processing the preprocessed image to obtain a global feature map; The local feature generation module is data connected with the feature fusion module, and is used for processing the preprocessed image to obtain a local feature map; The feature fusion module is data connected with the classification module, and is used for performing feature fusion processing on the global feature map and the local feature map to obtain a fusion feature map; The classification module is used for classifying the fusion feature map to obtain the probability vector. 8.The maxillary anterior alveolar bone defect identification method of claim 7, wherein, The feature fusion module includes an initial fusion unit and R intermediate fusion units; R is an integer greater than 2; The initial fusion unit is used for processing the global feature map and the local feature map to obtain an initial fusion feature map; The intermediate fusion unit is used for processing the global feature map, the local feature map, and a previous stage feature map to obtain an intermediate fusion feature map; The previous stage feature map of the first intermediate fusion unit is the initial fusion feature map; the previous stage feature map of the last R-1 intermediate fusion units is the intermediate fusion feature map output by the previous intermediate fusion unit. 9.The maxillary anterior alveolar bone defect identification method of claim 8, wherein, The intermediate fusion unit includes a channel attention subunit, a first multiplication subunit, a spatial attention subunit, a second multiplication subunit, a first splicing subunit, a first convolution subunit, an average pooling subunit, a second splicing subunit, a second convolution subunit, a residual subunit, and an addition subunit; The input end of the channel attention subunit inputs the global feature map; the output end of the channel attention subunit is connected with the first input end of the first multiplication subunit; the second input end of the first multiplication subunit inputs the global feature map; and the output end of the first multiplication subunit is connected with the first input end of the first splicing subunit; An input end of the spatial attention sub-unit inputs the local feature map; an output end of the spatial attention sub-unit is connected to a first input end of the second multiplication sub-unit; a second input end of the second multiplication sub-unit inputs the local feature map; An output end of the second multiplication sub-unit is connected to a second input end of the first concatenation sub-unit; An input end of the first convolution sub-unit of the first intermediate fusion unit is connected to an output end of the initial fusion unit; input ends of the first convolution sub-units of the following R-1 intermediate fusion units are connected to output ends of the previous intermediate fusion units; An output end of the first convolution sub-unit is connected to an input end of the average pooling sub-unit; an output end of the average pooling sub-unit is connected to a first input end of the second concatenation sub-unit and a first input end of the addition sub-unit, respectively; a second input end and a third input end of the second concatenation sub-unit input the global feature map and the local feature map, respectively; An output end of the second concatenation sub-unit is connected to an input end of the second convolution sub-unit; an output end of the second convolution sub-unit is connected to a third input end of the first concatenation sub-unit; an output end of the first concatenation sub-unit is connected to an input end of the residual sub-unit; an output end of the residual sub-unit is connected to a second input end of the addition sub-unit; an output end of the addition sub-unit is an output end of the feature fusion module.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which are invoked to execute the maxillary anterior tooth alveolar bone defect identification method according to any one of claims 1-9. The computer storage medium stores computer instructions, which are invoked to execute the maxillary anterior tooth alveolar bone defect identification method according to any one of claims 1-9.

Images