Condylar process detection model and method and device for obtaining mandibular movement trajectory using the same

Abstract

This application proposes a condylar detection model and a method and apparatus for obtaining mandibular movement trajectories using it, comprising the following steps: acquiring CBCT data, the condylar position, and TMD diagnostic results from the CBCT data by constructing a condylar detection model; obtaining the condylar axis from the CBCT data based on the condylar position, and then obtaining the condylar axis from the CBCT data based on the condylar axis; constructing a three-dimensional coordinate system based on the condylar axis; performing three-dimensional modeling based on the CBCT data to obtain a mandibular model; and using the mandibular model to perform rotational motion around the axis in the three-dimensional coordinate system to obtain the mandibular movement trajectory. This scheme can correct the mandibular movement trajectory according to the patient's TMD type to obtain a more accurate movement trajectory, facilitating subsequent diagnosis by dentists.

Description

Technical Field

[0001] This application relates to the field of oral medicine, and in particular to a condylar detection model and a method and apparatus for obtaining mandibular movement trajectories using it. Background Technology

[0002] The condyle is one of the main growth centers of the mandible. Temporomandibular joint disorder (TMD) is a common disease related to the condyle. According to the "Expert Consensus on Cone-Beam CT Examination and Diagnostic Criteria for Temporomandibular Joint Disorders", the symptoms of TMD can be roughly divided into joint space narrowing, bone destruction, bone hyperplasia, bone sclerosis, condyle shortening or morphological changes, and bone cystic changes.

[0003] Current technology typically uses data from oral scan models for modeling and obtains the condylar axis between the two condyles using oral scan data. Based on the condylar axis, the occlusal trajectory of the mandible is simulated to assist doctors in making other judgments. However, due to the changes in the condyles, the occlusal trajectory of patients with temporomandibular joint disorder often differs from that of a normal mandible. Therefore, the mandibular occlusal trajectory generated by current technology often contains errors when assisting dentists in making judgments.

[0004] In conclusion, there is an urgent need for a method that can predict the mandibular occlusal trajectory based on the different condylar conditions of different patients, thereby more accurately assisting dentists in disease diagnosis. Summary of the Invention

[0005] This application provides a condylar detection model and a method and apparatus for obtaining mandibular movement trajectories using it. By constructing a condylar detection model, the position and TMD type of the condyle in CBCT data can be obtained. Then, the condylar axis can be fine-tuned according to the patient's TMD type, thereby obtaining a more accurate mandibular movement trajectory, which is convenient for doctors to make subsequent diagnoses.

[0006] In a first aspect, embodiments of this application provide a method for constructing a condyle detection model, the method comprising:

[0007] Obtain at least one CBCT data point labeled with condylar location and TMD type as a training sample;

[0008] A condylar detection model is constructed, which consists of an embedding module and a segmentation module. The embedding module segments each training sample into at least one sagittal plane, adds order information to each sagittal plane, divides each sagittal plane into multiple sub-images, obtains tensor information of each sub-image, and the tensor information of all sub-images included in each sagittal plane constitutes the initial tensor of the sagittal plane. Position information is added to each sub-image to obtain a position encoding matrix, a projection matrix is ​​constructed, and the tensor information of each sub-image in the initial tensor of each sagittal plane is multiplied by the projection matrix to obtain a projection vector matrix. The position encoding matrix is ​​fused into the projection vector matrix to obtain the word embedding tensor of each sagittal plane.

[0009] The segmentation module consists of multiple attention modules with the same structure connected in series. After normalizing the word embedding tensors of all sagittal planes, the tensors are fed into multiple connected attention modules to obtain attention feature maps. The attention feature maps include condylar positions. A multilayer perceptron is added after the last attention module, and the output of the last attention module is input into the multilayer perceptron to obtain the TMD type.

[0010] The obtained condylar position and TMD type are backpropagated to the condylar detection model to update the parameters in the condylar detection model so as to minimize the loss function, thus completing the construction of the condylar detection model.

[0011] Secondly, embodiments of this application provide a method for obtaining mandibular occlusal trajectory, including:

[0012] Acquire CBCT data, input the CBCT data into a trained condylar detection model to obtain two condylar locations corresponding to the CBCT data and TMD diagnostic results, the TMD diagnostic results including different TMD types;

[0013] The initial condylar axis of each condyle is obtained based on the location of each condyle. A correction coefficient table is constructed based on the TMD diagnosis results. The correction coefficient table contains correction coefficients for correcting the normal condylar axis to the condylar axis of different TMD types. Each initial condylar axis is corrected according to the correction coefficient table to obtain the final condylar axis. The two final condylar axes are connected to obtain the condylar axis.

[0014] A mandibular model is obtained by performing three-dimensional modeling based on the CBCT data. A three-dimensional coordinate system is constructed with the condylar axis as the X-axis and the center point of the condylar axis as the origin O. The centroids of the two condyles on the mandibular model are obtained and aligned with the condylar axis to place the mandibular model into the three-dimensional coordinate system.

[0015] The mandibular model rotates around the condylar axis in the three-dimensional coordinate system OXYZ, and the motion trajectory is corrected by a three-dimensional rotation and translation matrix to obtain the mandibular motion trajectory.

[0016] Thirdly, embodiments of this application provide an apparatus for constructing a condylar detection model, comprising:

[0017] Acquisition module: Acquire at least one CBCT data point labeled with condylar location and TMD type as a training sample;

[0018] Embedding Module: Constructs a condylar detection model, which consists of an embedding module and a segmentation module. The embedding module segments each training sample into at least one sagittal plane, adds order information to each sagittal plane, divides each sagittal plane into multiple sub-images, obtains tensor information of each sub-image, and the tensor information of all sub-images included in each sagittal plane forms the initial tensor of the sagittal plane. Position information is added to each sub-image to obtain a position encoding matrix, a projection matrix is ​​constructed, and the tensor information of each sub-image in the initial tensor of each sagittal plane is multiplied by the projection matrix to obtain a projection vector matrix. The position encoding matrix is ​​fused into the projection vector matrix to obtain the word embedding tensor of each sagittal plane.

[0019] Segmentation Module: The segmentation module consists of multiple attention modules with the same structure connected in series. After normalizing the word embedding tensors of all sagittal planes, the information is fed into multiple connected attention modules to obtain an attention feature map. The attention feature map includes condylar positions. A multilayer perceptron is added after the last attention module, and the output of the last attention module is input into the multilayer perceptron to obtain the TMD type.

[0020] Validation module: The obtained condylar position and TMD type are backpropagated to the condylar detection model to update the parameters in the condylar detection model so as to minimize the loss function, thus completing the construction of the condylar detection model.

[0021] Fourthly, embodiments of this application provide a mandibular movement trajectory acquisition device, comprising:

[0022] Acquisition module: Acquires CBCT data, inputs the CBCT data into a trained condylar detection model to obtain two condylar locations corresponding to the CBCT data and TMD diagnostic results, the TMD diagnostic results including different TMD types;

[0023] Correction module: Obtain the initial condylar axis of each condyle based on the location of each condyle, construct a correction coefficient table based on the TMD diagnosis results, the correction coefficient table contains correction coefficients for correcting the normal condylar axis to the condylar axis of different TMD types, correct each initial condylar axis according to the correction coefficient table to obtain the final condylar axis, and connect the two final condylar axes to obtain the condylar axis;

[0024] Construction module: Based on the CBCT data, a three-dimensional model of the mandible is obtained. A three-dimensional coordinate system is constructed with the condylar axis as the X-axis and the center point of the condylar axis as the origin O. The centroids of the two condyles on the mandible model are obtained and aligned with the condylar axis to place the mandible model into the three-dimensional coordinate system.

[0025] Trajectory simulation module: The mandibular model rotates around the condylar axis in the three-dimensional coordinate system OXYZ, and the motion trajectory is corrected by a three-dimensional rotation and translation matrix to obtain the mandibular motion trajectory.

[0026] Fifthly, embodiments of this application provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a device for constructing a condylar detection model or a method for obtaining mandibular occlusion and trajectory.

[0027] In a sixth aspect, embodiments of this application provide a readable storage medium storing a computer program, the computer program including program code for controlling a process to execute the process, the process including a method for obtaining mandibular movement trajectory and an apparatus for constructing a condylar detection model.

[0028] The main contributions and innovations of this invention are as follows:

[0029] This application embodiment obtains the condylar position and TMD type in CBCT data by constructing a condylar detection model. The condylar detection model uses a multi-layer attention mechanism for feature extraction, which can more accurately obtain the condylar position. This scheme constructs a correction coefficient table according to different TMD types and uses the correction coefficient table to correct the condylar axis, thereby obtaining a more accurate condylar axis. This scheme uses a three-dimensional rotation and translation matrix to constrain rigid body transformation, so as to better simulate the small-amplitude translational movements that the mandible can perform in reality.

[0030] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0031] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0032] Figure 1 This is a flowchart illustrating a method for constructing a condylar detection model according to an embodiment of this application;

[0033] Figure 2 This is a schematic diagram of a center positioning according to an embodiment of this application;

[0034] Figure 3 This is a schematic diagram of a mandibular model according to an embodiment of this application;

[0035] Figure 4 This is a schematic diagram of the structure of an attention module according to an embodiment of this application;

[0036] Figure 5 This is a schematic diagram of the structure of a segmentation module according to an embodiment of this application;

[0037] Figure 6 This is a schematic flowchart of a device for constructing a condylar detection model according to an embodiment of this application;

[0038] Figure 7 This is a schematic flowchart of a mandibular movement trajectory acquisition device according to an embodiment of this application;

[0039] Figure 8 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0040] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0041] It should be noted that the steps of the corresponding methods are not necessarily performed in the order shown and described in this specification in other embodiments. In some other embodiments, the methods may include more or fewer steps than described in this specification. Furthermore, a single step described in this specification may be broken down into multiple steps in other embodiments; and multiple steps described in this specification may be combined into a single step in other embodiments.

[0042] To better understand this plan, the terms mentioned in this plan will be explained again:

[0043] CBCT data: Cone-beam computed tomography (CBCT) of the oral and maxillofacial region works by using an X-ray generator to perform circular DR (digital radiography) around the subject with a low dose of radiation (typically around 10 mA). The data obtained from the intersection of multiple digital radiographs around the subject (180-360 times, depending on the product) is then reconstructed in a computer to obtain a three-dimensional image.

[0044] TMD: Short for Temporomandibular Joint Disorder. Temporomandibular joint disorder syndrome is the most common disease of the oral and maxillofacial region, and its pathogenesis is not fully understood. The main clinical manifestations of this syndrome are pain in the joint area, clicking sounds during movement, and mandibular movement disorders. Most cases are due to joint dysfunction and have a good prognosis; however, a very small number of cases may also develop organic changes.

[0045] Example 1

[0046] A method for constructing a condyle detection model, comprising:

[0047] Obtain at least one CBCT data point labeled with condylar location and TMD type as a training sample;

[0048] A condylar detection model is constructed, which consists of an embedding module and a segmentation module. The embedding module segments each training sample into at least one sagittal plane, adds order information to each sagittal plane, divides each sagittal plane into multiple sub-images, obtains tensor information of each sub-image, and the tensor information of all sub-images included in each sagittal plane constitutes the initial tensor of the sagittal plane. Position information is added to each sub-image to obtain a position encoding matrix, a projection matrix is ​​constructed, and the tensor information of each sub-image in the initial tensor of each sagittal plane is multiplied by the projection matrix to obtain a projection vector matrix. The position encoding matrix is ​​fused into the projection vector matrix to obtain the word embedding tensor of each sagittal plane.

[0049] The segmentation module consists of multiple attention modules with the same structure connected in series. After normalizing the word embedding tensors of all sagittal planes, the tensors are fed into multiple connected attention modules to obtain attention feature maps. The attention feature maps include condylar positions. A multilayer perceptron is added after the last attention module, and the output of the last attention module is input into the multilayer perceptron to obtain the TMD type.

[0050] The obtained condylar position and TMD type are backpropagated to the condylar detection model to update the parameters in the condylar detection model so as to minimize the loss function, thus completing the construction of the condylar detection model.

[0051] In some specific embodiments, medical professionals mark the positions of the left and right condyles in the acquired CBCT data, perform TMD diagnosis on the temporal joint where the condyles are located, and classify the CBCT data with TMD tags based on the diagnosis results.

[0052] In some specific embodiments, the sagittal plane is a lateral section. Sequence information Pos1 is added to each sagittal plane; for example, the leftmost sagittal plane is numbered 1, the adjacent sagittal plane to its right is numbered 2, and so on. Each sagittal plane is divided into n sub-images of size p*p, and the tensor information of each sub-image is Xi. The n sub-images are flattened to obtain the tensor information F of the sagittal plane, denoted as F[X1,X2,X3,…,Xn]. Position information is added to each sub-image to obtain a position encoding matrix Epos. For example, the position of the top-left sub-image is denoted as 1, and the position of the bottom-right sub-image is denoted as n. A projection matrix E is constructed, with a size of (p*p)*d. Each sub-image in F is multiplied by the projection matrix E to obtain a d-dimensional projection vector (Xi*E). The projection vectors are merged into a matrix to obtain a projection vector matrix. The position encoding matrix Epos and the projection vector matrix are fused to obtain the word embedding tensor Z of each sagittal plane. The specific formula is as follows:

[0053] Z=[X1*E,X2*E,X3*E,…,Xn*E]+Epos

[0054] Specifically, Epos contains sequence information Pos1 and position information Pos2, meaning that the position encoding matrix can be used to determine which part of the CBCT data the current sub-image belongs to.

[0055] In some specific embodiments, the structure of the attention module is as follows: Figure 4 As shown, the attention module first normalizes the word embedding tensors of all sagittal planes to obtain a normalized result. This normalized result is then input into a parallel attention layer. The attention layer performs a cubic linear transformation on the normalized result to obtain matrices Q, K, and V. Q and K are then multiplied by a dot product to obtain a similarity matrix. This similarity matrix is ​​normalized to obtain a weight matrix. The weight matrix is ​​multiplied by V to obtain the attention tensor. The attention tensor of each attention module is calculated using the following formula:

[0056]

[0057] Y = softmax(a(Q, K))V

[0058] Where Y represents the attention tensor and A represents the similarity matrix, which represents the similarity of elements in the normalization result.

[0059] Specifically, the attention tensor is fed into a multilayer perceptron to obtain the output of each attention module, with an output size of n*d. The attention module adopts a residual structure, and the attention layers are parallelized to prevent overfitting.

[0060] In some specific embodiments, the structure of the segmentation module is as follows: Figure 5 As shown, in the segmentation module, multiple attention modules can be sequentially connected. The output of the i-th attention module is denoted as Zi. The larger i is, the more deep semantic information the output of the attention module contains, which can be used to determine the condyle position. The smaller i is, the more obvious the shallow position information and fine-grained features are, which can be used to restore the resolution of image upsampling and improve the segmentation effect. The output results of each attention module are stacked in reverse order to obtain the stacked result. The stacked result is fused with the input CBCT data to obtain the attention feature map.

[0061] For example, the output of the last attention module is Zi, which has a size of n*d. After convolution and upsampling operations on Zi, it is superimposed on Zi-1. The above steps are repeated to obtain the attention feature map.

[0062] In some specific embodiments, the same method used to obtain the training samples is used to obtain the validation set and test set. After training with the training samples, the validation set is used to evaluate the model performance. The values ​​of the manually set hyperparameters are optimized based on the changes in the loss function and various indicators during training to prevent overfitting and other problems. Finally, the test set is used to test the model and obtain performance indicators such as condyle segmentation accuracy and TMD type diagnosis accuracy to evaluate the model's generalization ability. Finally, the trained model is applied to the acquisition of condyle location and the diagnosis of TMD type. The input of the condyle detection model is CBCT data, and the output is CBCT data containing condyle location and TMD type diagnosis results.

[0063] Example 2

[0064] This application provides a method for obtaining the trajectory of mandibular movements, specifically, referring to... Figure 1 The method includes:

[0065] Acquire CBCT data, input the CBCT data into a trained condylar detection model to obtain two condylar locations corresponding to the CBCT data and TMD diagnostic results, the TMD diagnostic results including different TMD types;

[0066] The initial condylar axis of each condyle is obtained based on the location of each condyle. A correction coefficient table is constructed based on the TMD diagnosis results. The correction coefficient table contains correction coefficients for correcting the normal condylar axis to the condylar axis of different TMD types. Each initial condylar axis is corrected according to the correction coefficient table to obtain the final condylar axis. The two final condylar axes are connected to obtain the condylar axis.

[0067] A mandibular model is obtained by performing three-dimensional modeling based on the CBCT data. A three-dimensional coordinate system is constructed with the condylar axis as the X-axis and the center point of the condylar axis as the origin O. The centroids of the two condyles on the mandibular model are obtained and aligned with the condylar axis to place the mandibular model into the three-dimensional coordinate system.

[0068] The mandibular model rotates around the condylar axis in the three-dimensional coordinate system OXYZ, and the motion trajectory is corrected by a three-dimensional rotation and translation matrix to obtain the mandibular motion trajectory.

[0069] In some embodiments, in the step of "obtaining the initial condylar axis of each condyle based on each condylar position", histogram statistics are performed on the pixels contained in the condylar position, the gray value of each pixel is counted, a threshold is applied to the gray value of each pixel to obtain the specific condylar position, a horizontal line of the specific condylar position is determined, a first straight line perpendicular to the horizontal line is drawn at the highest point of the specific condylar position, a second straight line is drawn at the narrowest point of the anterior joint space of the specific condylar position, the second straight line forms a 45-degree angle with the horizontal line, and the intersection of the first straight line and the second straight line is the initial condylar axis.

[0070] Specifically, the narrowest point of the anterior joint space is the narrowest part of the anterior half of the joint space, such as... Figure 2 The position of the letter 'b' is where the anterior joint space is narrowest.

[0071] Specifically, the condyle detection model determines the condyle position by defining the bounding box of the condyle. That is, one of the outputs of the condyle detection model is a bounding box B1, the position of which is the condyle position, and the data within which is the data contained in the condyle position.

[0072] Specifically, a histogram is used to statistically analyze the data surrounding the condyle to obtain statistical results. Based on these results, a threshold is used to filter the data to determine the specific condyle position. The specific condyle position is within a bounding box B2 within a bounding box B1. The threshold value for threshold filtering is set according to the specific application. Furthermore, the threshold can be set manually or adaptively. Determining the specific condyle position based on the condyle position can improve the accuracy of subsequent positioning of the initial condyle axis.

[0073] Specifically, the method for defining the initial condylar axis is as follows: Figure 2 As shown, Figure 2 In the diagram, the outer ring represents bounding box B1, and the inner ring represents bounding box B2. First, the horizontal line of the specific condylar position is determined. The horizontal line of the specific condylar position is a line that is parallel to the ground in the normal state of the condyle. The horizontal line is used to keep the condyle in a horizontal state to facilitate the positioning of the initial condylar axis. In the horizontal state of the condyle, the highest point of the condyle is selected. At the highest point of the condyle, a first straight line a is drawn perpendicular to the horizontal line. A second straight line b is drawn at the narrowest point of the anterior joint space at the specific condylar position. The second straight line b forms a 45-degree angle with the horizontal line. The intersection point O of the first straight line a and the second straight line b is the initial condylar axis.

[0074] Furthermore, the initial condylar axes at the two condylar positions are obtained using the methods described above.

[0075] In some embodiments, in the step of "correcting each initial condylar axis according to the correction coefficient table to obtain the final condylar axis", a third line is drawn through the intersection of the first line and the second line. The third line is at a 45-degree angle to the horizontal line and is perpendicular to the second line. The lengths of the second line and the third line within the specific condylar position are calculated respectively. The correction coefficients in the correction coefficient table are used to adjust the lengths of the second line and the third line to correct each initial condylar axis to obtain the final condylar axis.

[0076] Specifically, since the initial condylar axis is the intersection of the second and third lines, the position and direction of the initial condylar axis can be corrected by adjusting the lengths of the second and third lines.

[0077] Specifically, the correction coefficient table is constructed based on the TMD type obtained from the condyle detection model, and the correction coefficient table contains correction coefficients for correcting the normal condyle axis to the condyle axis of different TMD types.

[0078] Specifically, the correction coefficient table sets different correction coefficients for different TMD types based on the doctor's clinical experience. That is to say, based on the doctor's judgment, we quantitatively determine a coefficient for each TMD type, and then discuss the final coefficient value with the doctor to ensure the rationality of the coefficient setting.

[0079] Furthermore, a corrected bias is obtained based on the second line length, the third line length, and the correction coefficient, and the final condylar axis is obtained by adding the corrected bias to the initial condylar axis.

[0080] For example, such as Figure 2 As shown, a third line c is drawn through the intersection of the first line a and the second line b, such that the third line c forms a 45-degree angle with the horizontal line and is perpendicular to the second line b. The lengths Lb of the second line and Lc of the third line within the specific condylar position are calculated respectively. Then, the lengths Lb and Lc of the second line are adjusted using the correction coefficient α in the correction coefficient table through the correction bias formula. The correction bias formula is as follows:

[0081] bias = log(Lc / Lb) * α

[0082] Where Lc represents the length of the third line, Lb represents the length of the second line, α represents the correction coefficient, and bias represents the correction bias.

[0083] Specifically, the correction coefficient α is affected by the type of TMD. If the patient does not have TMD, there is no need to use the correction bias formula to correct the initial condylar axis. It is only necessary to calculate the centroids of the two condyles and connect the two centroids to obtain the condylar axis.

[0084] Specifically, this solution can generate different condylar axes based on different TMD types.

[0085] Specifically, since TMD can cause abnormal mandibular movement trajectory, thus affecting subsequent diagnosis of the disease by doctors, different correction coefficients are obtained according to the type of TMD, and then the initial condylar axis is corrected by the correction coefficients.

[0086] In some specific embodiments, in the step of "connecting the two final condylar axes to obtain the condylar axis", the above method is used to obtain the two final condylar axes respectively, thereby obtaining the condylar axis.

[0087] In some embodiments, in the step of "obtaining a mandibular model by performing three-dimensional modeling based on the CBCT data", an adaptive threshold is used to segment the CBCT data to obtain coarse segmented mandibular data, region growth is performed on the coarse segmented mandibular data to obtain fine segmented mandibular data, and three-dimensional modeling is performed based on the fine segmented mandibular data to obtain a mandibular model.

[0088] Furthermore, the Otsu algorithm is used to determine the adaptive threshold of the CBCT data, and the adaptive threshold is used to segment the CBCT data to obtain a first segmentation result, which includes bone tissue and soft tissue. A narrow threshold is obtained, and the narrow threshold is used to segment the bone tissue to obtain a second segmentation result, which includes the mandible. Finally, the morphological dilation algorithm is used to calculate the second segmentation result to obtain coarse segmentation mandibular bone data.

[0089] Specifically, the narrow threshold is a finely adjusted threshold that allows for better segmentation of bone tissue.

[0090] Specifically, the advantage of using adaptive thresholds to segment the CBCT data is that it is computationally efficient and fast.

[0091] Specifically, segmenting the bone tissue using a narrow threshold allows for the initial separation of the mandible.

[0092] Specifically, since using a narrow threshold to segment bone tissue leads to the loss of boundary information, a morphological dilation algorithm is used to compensate for the loss of boundary information and obtain coarsely segmented mandibular bone data.

[0093] Furthermore, at least one sub-point is selected on the coarsely segmented mandibular data. The seed point is a voxel. A first range is defined for each seed point, and a first region is set for each seed point. The similarity of each voxel in the first range with respect to the corresponding seed point is calculated. A similarity threshold is set, and voxels with a similarity greater than the similarity threshold with the corresponding seed point are added to the first region. When the mean and variance in the first region both exceed the set threshold, the calculation is stopped, and the finely segmented mandibular data is obtained.

[0094] Specifically, the similarity between a voxel and its corresponding seed point can be measured by features such as grayscale difference and gradient.

[0095] Furthermore, the marching cube algorithm was used to perform 3D modeling on the finely segmented mandibular bone data, and the modeling result is as follows. Figure 3 As shown.

[0096] Specifically, each voxel has 8 vertices, so each voxel can form an 8-bit binary number. Based on different binary numbers, 256 different cases are preset. For each voxel, the values ​​of its 8 vertices are compared with the threshold of the isosurface to determine whether the voxel intersects with the isosurface. If they intersect, it means that the voxel is near the mandibular surface, and the intersecting vertices are marked as 1; otherwise, they are marked as 0. Based on the marking results, the binary number of each voxel is obtained and compared with the preset 256 cases to determine the shape and position of the isosurface in the voxel. Finally, the edges on the isosurface are triangulated to generate a triangular mesh model, thus completing the mandibular modeling and obtaining the mandibular model.

[0097] In some specific embodiments, in the step of "establishing a three-dimensional coordinate system with the condylar axis as the X-axis and the center point of the condylar axis as the origin O", a Z-axis is drawn directly above the origin O in the direction of the X-axis, and a Y-axis is drawn perpendicular to the plane ZOX through the origin O to obtain the three-dimensional coordinate system OXYZ. The centroids of the two condyles on the mandibular model are obtained and aligned with the condylar axis. The farthest point on the mandibular model from the condylar axis is obtained and aligned with the Y-axis. The purpose of doing this is to locate the position of the mandibular model in the three-dimensional coordinate system using the centroids of the two condyles and the farthest point, so as to place the mandibular model into the three-dimensional coordinate system.

[0098] In some embodiments, in the step of "correcting the motion trajectory by a three-dimensional rotation and translation matrix to obtain the mandibular motion trajectory", the angle and translation amplitude of the rotation around the axis are set, the coordinates of any point on the mandible are obtained, and the coordinates of the point after the motion are obtained by combining the angle, translation amplitude and coordinates of the rotation around the axis with the coordinates of the point. The mandibular motion trajectory is obtained by traversing the above operations.

[0099] Specifically, the rotational motion around the axis is a rigid body transformation motion. To mimic the small-amplitude translational motion of the mandible in reality, a three-dimensional rotation and translation matrix is ​​used to correct the motion trajectory. The formula for the three-dimensional rotation and translation matrix is ​​as follows:

[0100]

[0101] Where x, y, z are the point coordinates, p is the point index, R is the rotation matrix, and T is the translation matrix. In this scheme, the rotation matrix only rotates around the condylar axis, so the rotation matrix of this scheme is: θ is the rotation angle, and the translation matrix is... t is the translation distance.

[0102] Example 3

[0103] Based on the same concept, referencing Figure 6 This application also proposes a mandibular movement trajectory acquisition device, comprising:

[0104] Acquisition module: Acquires CBCT data, inputs the CBCT data into a trained condylar detection model to obtain two condylar locations corresponding to the CBCT data and TMD diagnostic results, wherein the TMD diagnostic results include different TMD types;

[0105] Correction module: Obtain the initial condylar axis of each condyle based on the location of each condyle, construct a correction coefficient table based on the TMD diagnosis results, the correction coefficient table contains correction coefficients for correcting the normal condylar axis to the condylar axis of different TMD types, correct each initial condylar axis according to the correction coefficient table to obtain the final condylar axis, and connect the two final condylar axes to obtain the condylar axis;

[0106] Construction Module: Based on the CBCT data, a 3D model of the mandible is obtained. The condylar axis is taken as the X-axis, the center point of the condylar axis is taken as the origin O, the Z-axis is drawn through the origin O in the direction directly above the X-axis, and the perpendicular line from the origin O to the plane XOZ is drawn as the Y-axis to obtain the 3D coordinate system OXYZ. The centroids of the two condyles on the mandible model are obtained and aligned with the condylar axis. The farthest point on the mandible model from the condylar axis is obtained and aligned with the Y-axis.

[0107] Trajectory simulation module: The mandibular model rotates around the condylar axis in the three-dimensional coordinate system OXYZ, and the motion trajectory is corrected by a three-dimensional rotation and translation matrix to obtain the mandibular motion trajectory.

[0108] Example 4

[0109] Based on the same concept, referencing Figure 7 This application also proposes a device for constructing a condyle detection model, comprising:

[0110] Acquisition module: Acquire at least one CBCT data point labeled with condylar location and TMD type as a training sample;

[0111] Embedding Module: Constructs a condylar detection model, which consists of an embedding module and a segmentation module. The embedding module segments each training sample into at least one sagittal plane, adds order information to each sagittal plane, divides each sagittal plane into multiple sub-images, obtains tensor information of each sub-image, and the tensor information of all sub-images included in each sagittal plane forms the initial tensor of the sagittal plane. Position information is added to each sub-image to obtain a position encoding matrix, a projection matrix is ​​constructed, and the tensor information of each sub-image in the initial tensor of each sagittal plane is multiplied by the projection matrix to obtain a projection vector matrix. The position encoding matrix is ​​fused into the projection vector matrix to obtain the word embedding tensor of each sagittal plane.

[0112] Segmentation Module: The segmentation module consists of multiple attention modules with the same structure connected in series. After normalizing the word embedding tensors of all sagittal planes, the information is fed into multiple connected attention modules to obtain an attention feature map. The attention feature map includes condylar positions. A multilayer perceptron is added after the last attention module, and the output of the last attention module is input into the multilayer perceptron to obtain the TMD type.

[0113] Validation module: The obtained condylar position and TMD type are backpropagated to the condylar detection model to update the parameters in the condylar detection model so as to minimize the loss function, thus completing the construction of the condylar detection model.

[0114] Example 5

[0115] This embodiment also provides an electronic device, see reference. Figure 8 It includes a memory 404 and a processor 402, wherein the memory 404 stores a computer program and the processor 402 is configured to run the computer program to perform the steps in any of the above method embodiments.

[0116] Specifically, the processor 402 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0117] The memory 404 may include a mass storage device for data or instructions. For example, and not limitingly, the memory 404 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 404 may include removable or non-removable (or fixed) media. Where appropriate, the memory 404 may be internal or external to a data processing device. In a particular embodiment, the memory 404 is non-volatile memory. In a particular embodiment, the memory 404 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0118] The memory 404 can be used to store or cache various data files that need to be processed and / or communicated, as well as possible computer program instructions executed by the processor 402.

[0119] The processor 402 reads and executes computer program instructions stored in the memory 404 to implement any of the mandibular motion trajectory acquisition methods or condylar detection model construction methods in the above embodiments.

[0120] Optionally, the electronic device may further include a transmission device 406 and an input / output device 408, wherein the transmission device 406 is connected to the processor 402, and the input / output device 408 is connected to the processor 402.

[0121] The transmission device 406 can be used to receive or send data via a network. Specific examples of the network described above may include wired or wireless networks provided by the communication provider of the electronic device. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 406 may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0122] The input / output device 408 is used to input or output information. In this embodiment, the input information may be CBCT data, condylar position markings, TMD type, etc., and the output information may be the bite movement trajectory of the mandible, etc.

[0123] Optionally, in this embodiment, the processor 402 can be configured to perform the following steps via a computer program:

[0124] S101. Obtain at least one CBCT data point labeled with condylar location and TMD type as a training sample;

[0125] S102. Construct a condylar detection model. The condylar detection model consists of an embedding module and a segmentation module. The embedding module segments each training sample into at least one sagittal plane, adds order information to each sagittal plane, divides each sagittal plane into multiple sub-images on an average basis, obtains tensor information of each sub-image, and the tensor information of all sub-images included in each sagittal plane constitutes the initial tensor of the sagittal plane. Position information is added to each sub-image to obtain a position encoding matrix. A projection matrix is ​​constructed. The tensor information of each sub-image in the initial tensor of each sagittal plane is multiplied by the projection matrix to obtain a projection vector matrix. The position encoding matrix is ​​fused into the projection vector matrix to obtain the word embedding tensor of each sagittal plane.

[0126] S103. The segmentation module is composed of multiple attention modules with the same structure connected in sequence. After normalizing the word embedding tensors of all sagittal planes, they are fed into multiple connected attention modules to obtain an attention feature map. The attention feature map includes condylar positions. A multilayer perceptron is added after the last attention module. The output of the last attention module is input into the multilayer perceptron to obtain the TMD type.

[0127] S104. The obtained condylar position and TMD type are backpropagated to the condylar detection model to update the parameters in the condylar detection model so as to minimize the loss function, thus completing the construction of the condylar detection model.

[0128] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0129] Generally, various embodiments can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects of the invention can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device, but the invention is not limited thereto. Although various aspects of the invention may be shown and described as block diagrams, flowcharts, or using some other graphical representation, it should be understood that, by way of non-limiting example, these blocks, apparatuses, systems, techniques, or methods described herein can be implemented in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0130] Embodiments of the present invention can be implemented by computer software, which may be executable by a data processor of a mobile device, such as a processor entity, or by hardware, or by a combination of software and hardware. Computer software or programs (also referred to as program products) including software routines, applets, and / or macros can be stored in any device-readable data storage medium, and they include program instructions for performing specific tasks. The computer program product may include one or more computer-executable components configured to perform the embodiments when the program is run. The one or more computer-executable components may be at least one piece of software code or a portion thereof. Additionally, it should be noted in this respect that, as Figure 8 Any box in the logical flow can represent a program step, or interconnected logic circuits, boxes and functions, or a combination of program steps and logic circuits, boxes and functions. Software can be stored on physical media such as memory chips or blocks of storage implemented within a processor, magnetic media such as hard disks or floppy disks, and optical media such as DVDs and their data variants, CDs, etc. The physical medium is a non-transient medium.

[0131] Those skilled in the art should understand that the technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0132] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for obtaining the trajectory of mandibular movement, characterized in that, Includes the following steps: Obtain at least one CBCT data point labeled with condylar location and TMD type as a training sample; A condylar detection model is constructed, which consists of an embedding module and a segmentation module. The embedding module segments each training sample into at least one sagittal plane, adds order information to each sagittal plane, divides each sagittal plane into multiple sub-images, obtains tensor information of each sub-image, and the tensor information of all sub-images included in each sagittal plane constitutes the initial tensor of the sagittal plane. Position information is added to each sub-image to obtain a position encoding matrix, a projection matrix is ​​constructed, and the tensor information of each sub-image in the initial tensor of each sagittal plane is multiplied by the projection matrix to obtain a projection vector matrix. The position encoding matrix is ​​fused into the projection vector matrix to obtain the word embedding tensor of each sagittal plane. The segmentation module consists of multiple attention modules with identical structures connected in series. After normalizing the word embedding tensors of all sagittal planes, the data is fed into the series of attention modules to obtain an attention feature map, which includes condylar positions. A multilayer perceptron is added after the last attention module, and the output of the last attention module is input into the multilayer perceptron to obtain the TMD type. The obtained condylar positions and TMD types are backpropagated to the condylar detection model to update the parameters of the condylar detection model to minimize the loss function, thus completing the construction of the condylar detection model. CBCT data is acquired and input into the trained condylar detection model to obtain two condylar positions corresponding to the CBCT data and the TMD diagnostic results, which include different TMD types. The initial condylar axis of each condyle is obtained based on the location of each condyle. A correction coefficient table is constructed based on the TMD diagnosis results. The correction coefficient table contains correction coefficients for correcting the normal condylar axis to the condylar axis of different TMD types. Each initial condylar axis is corrected according to the correction coefficient table to obtain the final condylar axis. The two final condylar axes are connected to obtain the condylar axis. A mandibular model is obtained by performing three-dimensional modeling based on the CBCT data. A three-dimensional coordinate system is constructed with the condylar axis as the X-axis and the center point of the condylar axis as the origin O. The centroids of the two condyles on the mandibular model are obtained and aligned with the condylar axis to place the mandibular model into the three-dimensional coordinate system. The mandibular model rotates around the condylar axis in the three-dimensional coordinate system OXYZ, and the motion trajectory is corrected by a three-dimensional rotation and translation matrix to obtain the mandibular motion trajectory.

2. The method for obtaining the mandibular movement trajectory according to claim 1, characterized in that, In the step of "obtaining the initial condylar axis of each condyle based on each condylar position", histogram statistics are performed on the pixels contained in the condylar position, the gray value of each pixel is counted, and the specific condylar position is obtained by thresholding based on the gray value of each pixel. The horizontal line of the specific condylar position is determined, a first straight line perpendicular to the horizontal line is drawn at the highest point of the specific condylar position, and a second straight line is drawn at the narrowest point of the anterior joint space of the specific condylar position. The second straight line forms a 45-degree angle with the horizontal line, and the intersection of the first straight line and the second straight line is the initial condylar axis.

3. The method for obtaining the mandibular movement trajectory according to claim 2, characterized in that, In the step of "correcting each initial condylar axis according to the correction coefficient table to obtain the final condylar axis", a third line is drawn through the intersection of the first line and the second line. The third line is at a 45-degree angle to the horizontal line and is perpendicular to the second line. The lengths of the second line and the third line within the specific condylar position are calculated respectively. The correction coefficients in the correction coefficient table are used to adjust the lengths of the second line and the third line.

4. The method for obtaining the mandibular movement trajectory according to claim 1, characterized in that, In the step of "obtaining a mandibular model by performing three-dimensional modeling based on the CBCT data", an adaptive threshold is used to segment the CBCT data to obtain coarse segmented mandibular data, region growth is performed on the coarse segmented mandibular data to obtain fine segmented mandibular data, and three-dimensional modeling is performed based on the fine segmented mandibular data to obtain a mandibular model.

5. The method for obtaining the mandibular movement trajectory according to claim 1, characterized in that, In the step of "correcting the motion trajectory by a three-dimensional rotation and translation matrix to obtain the mandibular motion trajectory", the angle of rotation around the axis and the translation range are set, the coordinates of any point on the mandible are obtained, and the coordinates of the point after the motion are obtained by using the three-dimensional rotation and translation matrix in combination with the angle of rotation around the axis, the translation range and the coordinates of the point. The mandibular motion trajectory is obtained by traversing the above operations.

6. A device for acquiring the trajectory of mandibular movement, characterized in that, include: Acquisition module: Acquire at least one CBCT data point labeled with condylar location and TMD type as a training sample; A condylar detection model is constructed, which consists of an embedding module and a segmentation module. The embedding module segments each training sample into at least one sagittal plane, adds order information to each sagittal plane, divides each sagittal plane into multiple sub-images, obtains tensor information of each sub-image, and the tensor information of all sub-images included in each sagittal plane constitutes the initial tensor of the sagittal plane. Position information is added to each sub-image to obtain a position encoding matrix, a projection matrix is ​​constructed, and the tensor information of each sub-image in the initial tensor of each sagittal plane is multiplied by the projection matrix to obtain a projection vector matrix. The position encoding matrix is ​​fused into the projection vector matrix to obtain the word embedding tensor of each sagittal plane. The segmentation module consists of multiple attention modules with identical structures connected in series. After normalizing the word embedding tensors of all sagittal planes, the data is fed into the series of attention modules to obtain an attention feature map, which includes condylar positions. A multilayer perceptron is added after the last attention module, and the output of the last attention module is input into the multilayer perceptron to obtain the TMD type. The obtained condylar positions and TMD types are backpropagated to the condylar detection model to update the parameters of the condylar detection model to minimize the loss function, thus completing the construction of the condylar detection model. CBCT data is acquired and input into the trained condylar detection model to obtain two condylar positions corresponding to the CBCT data and the TMD diagnostic results, which include different TMD types. Correction module: Obtain the initial condylar axis of each condyle based on the location of each condyle, construct a correction coefficient table based on the TMD diagnosis results, the correction coefficient table contains correction coefficients for correcting the normal condylar axis to the condylar axis of different TMD types, correct each initial condylar axis according to the correction coefficient table to obtain the final condylar axis, and connect the two final condylar axes to obtain the condylar axis; Construction module: Based on the CBCT data, a three-dimensional model of the mandible is obtained. A three-dimensional coordinate system is constructed with the condylar axis as the X-axis and the center point of the condylar axis as the origin O. The centroids of the two condyles on the mandible model are obtained and aligned with the condylar axis to place the mandible model into the three-dimensional coordinate system. Trajectory simulation module: The mandibular model rotates around the condylar axis in the three-dimensional coordinate system OXYZ, and the motion trajectory is corrected by a three-dimensional rotation and translation matrix to obtain the mandibular motion trajectory.

7. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform a method for obtaining a mandibular movement trajectory as described in any one of claims 1-5.

8. A readable storage medium, characterized in that, The readable storage medium stores a computer program, the computer program including program code for controlling the process to execute the process, the process including a method for obtaining mandibular movement trajectory as described in any one of claims 1-5.

Images