Tooth root angle three-dimensional intelligent measurement method based on oral cavity CBCT image

Abstract

The invention relates to the technical field of tooth root angle measurement, and provides a tooth root angle three-dimensional intelligent measurement method based on an oral cavity CBCT image, and the method comprises the steps: constructing a Teeth-MambaSeg tooth segmentation neural network model for carrying out tooth segmentation on a to-be-measured oral cavity CBCT image; a training database is constructed, and the Teeth-MambaSeg tooth segmentation neural network model is trained; performing tooth segmentation on a to-be-measured oral cavity CBCT image through the trained Teeth-MambaSeg tooth segmentation neural network model, and projecting a single segmented tooth to a simulated two-dimensional projection plane to obtain a tooth simulated projection image; according to the tooth simulation projection image, feature points related to tooth root angle calculation are extracted, and three-dimensional space coordinates of the feature points are obtained; and obtaining a tooth root angle three-dimensional measurement value according to the three-dimensional space coordinates of the feature points. According to the invention, personal errors of measurement point positioning can be effectively reduced, and the stability and accuracy of tooth root angle measurement are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tooth root angle measurement, and in particular to a three-dimensional intelligent measurement method for tooth root angles based on oral CBCT images. Background Art

[0002] The tooth root angle measured based on oral cone-beam computed tomography (CBCT) is an important clinical reference indicator in orthodontic surgery. CBCT is a low-radiation, non-invasive, high-resolution imaging technology. Fig.16 As shown, the X-rays in the device scan around the patient along a circular orbit, and then the reconstruction algorithm inside the device can obtain a three-dimensional image model of the oral and maxillofacial area, that is, the oral CBCT image. The coronal plane, sagittal plane, and cross-section in the figure are three sets of two-dimensional section images obtained by slicing the oral CBCT three-dimensional image in three mutually perpendicular directions. At present, the measurement method based on oral CBCT images has higher accuracy than the traditional image measurement method of dental films and is widely used in the measurement of tooth angles. The Pickert measurement method and the Ivanova measurement method are the image measurement methods of root angles commonly used in clinical practice. Both methods manually select clinically significant measurement surfaces and measurement points in the two-dimensional section based on the doctor's experience, and calculate the root-related clinically significant angles based on the straight lines constructed from the measurement points.

[0003] When using measurement software for manual measurement, it is necessary to manually locate the landmark points before measuring. This method has human errors in the positioning of landmark points. The doctor's experience level will directly affect the accuracy of the measurement results. Therefore, the above two conventional clinical measurement methods cannot accurately and stably evaluate the tooth root angle measurement values.

[0004] Therefore, it is necessary to develop a three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images to improve the accuracy and stability of tooth root angle measurement. Summary of the invention

[0005] Based on this, in order to improve the accuracy and stability of tooth root angle measurement, the present invention provides a three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images, and its specific technical solution is as follows:

[0006] A three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images, comprising the following steps:

[0007] Construct a Teeth-MambaSeg neural network model for teeth segmentation in oral CBCT images to be measured;

[0008] Constructing a training database to train the Teeth-MambaSeg teeth segmentation neural network model;

[0009] Performing tooth segmentation on the oral CBCT image to be measured by using the trained Teeth-MambaSeg tooth segmentation neural network model, and projecting a single segmented tooth onto a simulated two-dimensional projection plane to obtain a tooth simulation projection image;

[0010] Extracting feature points related to tooth root angle calculation according to the tooth simulation projection image, and obtaining the three-dimensional spatial coordinates of the feature points;

[0011] According to the three-dimensional spatial coordinates of the feature points, a three-dimensional measurement value of the tooth root angle is obtained.

[0012] The three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images is based on oral CBCT images. By automatically calibrating the measurement points in the three-dimensional images of teeth, it can effectively reduce the human error in the positioning of the measurement points and improve the stability of the measurement. In addition, the three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images obtains the three-dimensional measurement value of the root angle according to the three-dimensional spatial coordinates of the feature points. The root angle measurement in three-dimensional space is to calculate the angle of the three-dimensional space straight line. This method itself is more in line with the real medical definition and is more accurate than the projection straight line angle measurement method based on two-dimensional sections in clinical routine. It avoids the calculation error caused by inappropriate selection of two-dimensional sections and improves the accuracy of root angle measurement.

[0013] Preferably, the teeth are segmented on the oral CBCT image to be measured by using the trained Teeth-MambaSeg teeth segmentation neural network model, and a single segmented tooth is projected onto a simulated two-dimensional projection plane. The specific method for obtaining a simulated tooth projection image comprises the following steps:

[0014] Performing tooth segmentation on the oral CBCT image to be measured by using the trained Teeth-MambaSeg tooth segmentation neural network model to obtain tooth number classification and segmentation results;

[0015] Extract the three-dimensional structure of the single tooth with the corresponding number, and set the minimum contour area f of the three-dimensional space voxel of the single tooth with the corresponding number mask ;

[0016] Creating a tooth simplified distance projection simulation operator, and moving the single tooth with the corresponding number to the center of the three-dimensional coordinate axis of the tooth simplified distance projection simulation operator;

[0017] According to the tooth simplified distance projection simulation operator and the minimum contour area fmask A system matrix is ​​obtained, and the tooth simulation projection image is obtained according to the system matrix.

[0018] Preferably, according to the tooth simplified distance projection simulation operator and the minimum contour area f mask The specific method of obtaining the system matrix includes the following steps:

[0019] Get the three-dimensional space voxel f j The projection point coordinates of the eight vertices;

[0020] Get the minimum rectangular area containing the eight projection points and the area A of the minimum rectangular area τ ;

[0021] Get the area A of the region overlapping with the minimum rectangular area p ;

[0022] Get the coefficients a of the system matrix θ [i,j]=A p / A τ ;

[0023] Where i = 1,...N u ×N v ,j=1,...N x ×N y ×N z , θ represents the sampling angle, N u 、N v Indicates the resolution of the projected image, N x 、N y 、N z The resolution of volume data representing three-dimensional space.

[0024] Preferably, the three-dimensional space voxel f is obtained j The specific formulas for the projection point coordinates of the eight vertices include:

[0025]

[0026] in, f j =(X j ,Y j ,Z j ) represents the three-dimensional voxel f j The coordinate value of .

[0027] Preferably, the specific method for obtaining the three-dimensional space coordinates of the feature point includes the following steps:

[0028] Select any two different visions and obtain the projection matrix M of the two visions 1 and M2 ;

[0029] According to the two visual projection matrices M 1 and M 2 , obtain the three-dimensional space coordinates of the feature point.

[0030] Preferably, the formulas of the two visual projection matrices are respectively as well as

[0031] Among them, (u 1 ,v 1 ,1) and (u 2 ,v 2 , 1) respectively represent the first projection point p of the feature point under the two visions 1 And the second projection point p 2 The homogeneous coordinates in each projection image, (X, Y, Z, 1) represents the homogeneous coordinates of the feature point P in the world coordinate system, Represents the i-th row and j-th column element of the projection matrix Mk, k=1,2.

[0032] Preferably, according to the two visual projection matrices M 1 and M 2 The specific method for obtaining the three-dimensional space coordinates of the feature points includes the following steps:

[0033] According to the two visual projection matrix formulas, four linear equations about X, Y, and Z are obtained

[0034] The three-dimensional spatial coordinates (X, Y, Z) of the feature point are obtained according to the four linear equations.

[0035] Preferably, the specific method for obtaining the three-dimensional measurement value of the tooth root angle according to the three-dimensional spatial coordinates of the feature point comprises the following steps:

[0036] The three-dimensional spatial coordinates of each of the feature points are obtained, and the three-dimensional measurement value of the tooth root angle is obtained according to a spatial straight line formed by connecting the three-dimensional spatial coordinates of each of the feature points.

[0037] Preferably, the three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images also includes the following steps:

[0038] Obtaining multiple sets of three-dimensional space coordinates corresponding to the two different visions;

[0039] The average value of the three-dimensional space coordinates is calculated, and the average value is used as the final three-dimensional space coordinate value.

[0040] Preferably, the tooth simulation projection image proj=A θ *Xr teeth Among them, Xr teeth Represents the grayscale vector value of a single tooth, A θ represents the system matrix, and θ represents the sampling angle. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0042] Figure 1 It is a schematic diagram of the overall process of a three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images in one embodiment of the present invention;

[0043] Figure 2 is a flowchart of a specific method for obtaining a tooth simulation projection image in one embodiment of the present invention;

[0044] Figure 3 is a flowchart of a specific method for obtaining a system matrix in one embodiment of the present invention;

[0045] Figure 4 is a flowchart of a specific method for obtaining the three-dimensional space coordinates of feature points in one embodiment of the present invention;

[0046] Figure 5 It is a schematic diagram of the overall process of a three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images in another embodiment of the present invention;

[0047] Figure 6 It is a schematic diagram of the offline training and online application process of the three-dimensional intelligent measurement method of tooth root angle in one embodiment of the present invention;

[0048] Figure 7 It is a schematic diagram of the overall framework structure of the Teeth-MambaSeg teeth segmentation neural network model in one embodiment of the present invention;

[0049] Figure 8 It is a schematic diagram of the structure of a tooth gated convolution block and a multi-directional scanning feature fusion block in one embodiment of the present invention;

[0050] Fig. 9 is a calculation schematic diagram of a tooth simplified distance projection simulation operator in one embodiment of the present invention;

[0051] Fig.10 is a schematic diagram of the effect of a two-dimensional simulation projection sequence of a single tooth in one embodiment of the present invention;

[0052] Fig.11 is a schematic diagram showing the effect of extracting feature points from a tooth projection image in one embodiment of the present invention;

[0053] Fig.12 is a schematic diagram of binocular vision projection geometric relationship in one embodiment of the present invention;

[0054] Fig.13 is a schematic diagram of an oral CBCT image in one embodiment of the present invention;

[0055] Fig.14 is a schematic diagram of a tooth segmentation result in one embodiment of the present invention;

[0056] Fig.15 is a schematic diagram of the effect of three-dimensional measurement of tooth root angles in one embodiment of the present invention;

[0057] Fig.16 It is a schematic diagram of the standard display mode of oral CBCT in the prior art. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with its embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the protection scope of the present invention.

[0059] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only implementation method.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0061] The “first” and “second” mentioned in the present invention do not represent specific quantities and orders, but are merely used to distinguish names.

[0062] Before describing the specific embodiments of the present invention in detail, a brief introduction to the prior art is first given.

[0063] The tooth root angle measured based on oral cone-beam computed tomography (CBCT) is an important clinical reference indicator in orthodontic surgery. CBCT is a low-radiation, non-invasive, high-resolution imaging technology. Fig.16 As shown, the X-rays in the device scan around the patient along a circular orbit, and then the reconstruction algorithm inside the device can obtain a three-dimensional image model of the oral and maxillofacial area, that is, the oral CBCT image. The coronal plane, sagittal plane, and cross-section in the figure are three sets of two-dimensional section images obtained by slicing the oral CBCT three-dimensional image in three mutually perpendicular directions. At present, the measurement method based on oral CBCT images has higher accuracy than the traditional image measurement method of dental films and is widely used in the measurement of tooth angles. The Pickert measurement method and the Ivanova measurement method are the image measurement methods of root angles commonly used in clinical practice. Both methods manually select clinically significant measurement surfaces and measurement points in the two-dimensional section based on the doctor's experience, and calculate the root-related clinically significant angles based on the straight lines constructed from the measurement points.

[0064] However, the current conventional clinical measurement methods cannot accurately and stably evaluate the root angle measurement values ​​of teeth. When using measurement software for manual measurement, it is necessary to manually locate the landmarks before measuring. This method has human errors in landmark positioning, and the doctor's experience level will directly affect the accuracy of the measurement results. The Ivanova measurement method and the Pickert measurement method are widely used in clinical practice. However, these two clinical measurement methods use the two-dimensional section of CBCT to measure the root angle, which loses the three-dimensional information. The measured angle value is not accurate enough, and the manually selected measurement surface and measurement point cannot guarantee the stability of the measurement results.

[0065] The three-dimensional intelligent measurement technology of teeth based on oral CBCT is still a technical problem to be solved. Its difficulty lies in the following two links: First, the accurate segmentation of individual teeth from oral CBCT images is an important basis for intelligent measurement technology. There are many artifacts in the CBCT imaging process, such as metal artifacts, motion artifacts and scattering artifacts. Severe artifacts greatly reduce the imaging quality and greatly increase the difficulty of image segmentation. In addition, when facing complex clinical cases of deformed teeth, teeth have variable morphological characteristics, which makes it difficult to achieve accurate segmentation. Although many scholars have conducted a lot of meaningful research on the automatic segmentation algorithm of teeth, it is still a practical and technically challenging task, and there is currently no clinically applicable system. In the past decade, many methods have been explored to design manual features for tooth segmentation (such as level sets, graph cuts or template fitting), but these methods have low feature description capabilities and are very sensitive to the complex appearance of tooth CBCT images, so cumbersome manual intervention is required for initialization or correction. Second, intelligently calibrate the measurement points from the three-dimensional model of the teeth to obtain the three-dimensional measurement values ​​of the tooth root angle. How to measure the root angle in three-dimensional space based on the segmented three-dimensional tooth image is still a difficult technical problem to achieve.

[0066] Deep learning methods are expected to achieve accurate and stable three-dimensional intelligent measurement of tooth root angles based on oral CBCT. Deep learning networks have strong feature learning capabilities and have achieved great success in the fields of computer vision and medical image segmentation. In order to cope with the challenges brought by artifacts and complex cases, and to further improve the accuracy and stability of intelligent three-dimensional measurement of teeth, a series of studies have attempted to use deep neural networks for tooth segmentation. Zhiming Cui et al. were the first to apply deep learning networks to the automated three-dimensional segmentation of teeth. Thanks to the largest dataset to date for training, the system's tooth segmentation accuracy exceeded that of radiologists. However, this method still has defects in reconstructing details. Ma Xuechun et al. proposed to use nnU-Net neural networks to intelligently segment various anatomical structures of the oral cavity, and then intelligently measure the diagnostic parameters of the implant area based on the three-dimensional model of the anatomical structure. The experimental results show that the intelligent three-dimensional measurement results based on neural networks are accurate and reliable. However, in the face of more complex tooth morphology, the accuracy and stability of the algorithm are still insufficient.

[0067] In order to solve the problem that the existing conventional clinical measurement methods cannot accurately and stably measure the tooth root angle, the present invention provides a three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images. Figure 1 As shown, the following steps are included:

[0068] S1, constructing a Teeth-MambaSeg neural network model for teeth segmentation in the oral CBCT image to be measured. Fig.13 shown.

[0069] like Figure 7 As shown in the figure, the Teeth-MambaSeg tooth segmentation neural network model mainly consists of three parts: ① a 3D feature encoder with multiple oral three-dimensional Mamba modules, which is used to model global information of different scales; ② a 3D decoder based on residual network blocks, which is used to predict segmentation results; ③ jump connections, which are used to connect global multi-scale features to the decoder for feature reuse.

[0070] ①3D feature encoder

[0071] The 3D feature encoder consists of a depth-wise convolution and multiple dental 3D Mamba blocks (D3DMamba). The first-scale features are extracted by depth-wise separable convolution and then input into the D3DMamba block and the downsampling block in sequence. In the D3DMamba block, the three-dimensional spatial relationship features are extracted by the teeth gated convolution block (TGC), and then the multi-directional scanning feature fusion block (MDScan) is responsible for flattening the extracted three-dimensional spatial relationship features into a one-dimensional feature sequence and fusing them. After the fused features are layer normalized (LayerNorm), they enter the multi-layer perception layer (MLP) to obtain multi-scale feature representation.

[0072] Teeth Gated Convolution Block (TGC): Figure 8 (a) is a schematic diagram of the tooth gated convolution block. For tooth segmentation, TGC uses the inherent multi-layer morphological prior information of teeth to extract accurate tooth features.

[0073] Multi-directional scanning feature fusion block (MD Scan): Figure 8 (b) is a schematic diagram of the multi-directional scanning feature fusion module. The MD Scan block uses three methods to scan the features extracted by the TGC block to achieve feature interaction and obtain 3D features after dynamic fusion. Figure 8As shown in (b), the three methods are forward feature interaction (Forward Features Interaction), backward feature interaction (Reverse Features Interaction) and alternating feature interaction (Inter-slice Features Interaction).

[0074] ②3D decoder

[0075] The feature encoder uses D3DMamba blocks to extract multi-scale features. The 3D decoder refers to the U-shaped network structure of UNETR and designs a decoder network structure based on residual blocks to restore the size of the 3D image layer by layer and finally predict the segmentation result.

[0076] ③ Skip Connection

[0077] The skip connection connects the global multi-scale features extracted by the 3D feature encoder to the corresponding layer of the 3D decoder, thereby enabling feature reuse and enhancing the expressiveness of the segmentation network.

[0078] S2, constructing a training database and training the Teeth-MambaSeg teeth segmentation neural network model.

[0079] The training database can be derived from clinical data provided by hospitals and / or cooperating doctors, and the diversity of cases is sought during collection. Data labels are produced using 3DSlicer, an open source software for medical image processing. The production of three-dimensional data labels can outline the corresponding tissue position of each two-dimensional section with the help of clinicians, so as to finely outline the three-dimensional segmentation label data of the alveolar bone and teeth.

[0080] Preferably, the training database includes a training set, a validation set and a test set. The test set data is segmented by the Teeth-MambaSeg teeth segmentation neural network model and the test set label result is qualitatively and quantitatively compared to evaluate the generalization ability of the Teeth-MambaSeg teeth segmentation neural network model. At the same time, the segmentation effect of the tooth morphology-specific case is tested to evaluate the robustness of the Teeth-MambaSeg teeth segmentation neural network model.

[0081] S3, performing tooth segmentation on the oral CBCT image to be measured by using the trained Teeth-MambaSeg tooth segmentation neural network model, and projecting a single segmented tooth onto a simulated two-dimensional projection plane to obtain a tooth simulation projection image.

[0082] like Fig.14As shown in the schematic diagram of tooth segmentation results, the colors represent the tooth numbers.

[0083] S4, extracting feature points related to tooth root angle calculation according to the tooth simulation projection image, and acquiring the three-dimensional spatial coordinates of the feature points.

[0084] Edge detection, gradient calculation and other technologies can be used to extract feature points related to root angle calculation, such as Fig.11 As shown, the red dots in the figure are the extracted feature points.

[0085] S5, obtaining the three-dimensional measurement value of the tooth root angle according to the three-dimensional spatial coordinates of the feature point, such as Fig.15 Preferably, the specific method of step S5 comprises the following steps: obtaining the three-dimensional spatial coordinates of each of the feature points, and obtaining the three-dimensional measurement value of the tooth root angle according to the spatial straight line formed by connecting the three-dimensional spatial coordinates of each of the feature points.

[0086] The three-dimensional intelligent measurement method for tooth root angles based on oral CBCT images is based on oral CBCT images and can automatically calibrate measurement points directly in the three-dimensional tooth images using a set computer program, thereby effectively reducing human errors in positioning measurement points and improving measurement stability. On the other hand, the three-dimensional intelligent measurement method for tooth root angles based on oral CBCT images obtains three-dimensional measurement values ​​of the root angles according to the three-dimensional spatial coordinates of feature points. The root angle measurement in three-dimensional space is to calculate the angle of a straight line in three-dimensional space. This method itself is more in line with the real medical definition and is more accurate than the projection straight line angle measurement method based on two-dimensional sections in clinical routine. It avoids calculation errors caused by inappropriate selection of two-dimensional sections and improves the accuracy of root angle measurement.

[0087] In summary, the three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images can improve the accuracy and stability of tooth root angle measurement, thereby saving doctors' treatment time, improving doctors' treatment efficiency, and reducing the occurrence of misdiagnosis. It can be used as a reliable clinical tooth root angle measurement method.

[0088] As a preferred technical solution, Figure 6 As shown, the method of the present invention mainly includes two parts: offline training and online application.

[0089] (1) Offline training part: mainly data set labeling, construction of Teeth-MambaSeg tooth segmentation neural network model and clinical applicability evaluation. First, the collected clinical oral CBCT data is used to produce the label data required for the segmentation network, and the data is divided into training set, validation set and test set. Then, the Teeth-MambaSeg tooth segmentation neural network model is constructed, and the specific network structure is designed, and the optimal weights are trained; finally, the data set is input into the Teeth-MambaSeg tooth segmentation neural network model for training to obtain the model parameters. On the other hand, it is necessary to evaluate the clinical applicability of the three-dimensional intelligent measurement method, including statistical comparison using clinical routine measurement values, such as qualitatively and quantitatively comparing the test set data after segmentation by the network model with the test set label results, evaluating the generalization ability of the segmentation network model, and then improving the Teeth-MambaSeg tooth segmentation neural network model based on the evaluation results.

[0090] (2) Online application: mainly includes clinical data preprocessing, tooth 3D segmentation and root angle 3D intelligent measurement. First, generate sub-voxels of the tooth area, and perform preprocessing operations including denoising and enhancement on the input oral CBCT image data; then, input the preprocessed data into the trained Teeth-MambaSeg tooth segmentation neural network model, segment the tooth structure one by one according to the tooth number, and obtain the accurate tooth 3D segmentation results output according to the tooth number classification; finally, perform 3D intelligent measurement of the tooth root angle based on the segmentation results, including intelligent calibration of measurement points and calculation of root angle.

[0091] As a preferred technical solution, in step S3, as Figure 2 As shown, the teeth are segmented on the oral CBCT image to be measured by using the trained Teeth-MambaSeg teeth segmentation neural network model, and a single segmented tooth is projected onto a simulated two-dimensional projection plane. The specific method for obtaining a simulated tooth projection image includes the following steps:

[0092] S30, performing tooth segmentation on the oral CBCT image to be measured by using the trained Teeth-MambaSeg tooth segmentation neural network model to obtain tooth number classification and segmentation results;

[0093] S31, extracting the three-dimensional structure of a single tooth with a corresponding number, and setting the minimum contour area f of the three-dimensional space voxel (Volume Pixel, volume element) of the single tooth with the corresponding number mask ;

[0094] S32, creating a tooth simplified distance projection simulation operator, and moving the single tooth with the corresponding number to the center of the three-dimensional coordinate axis of the tooth simplified distance projection simulation operator;

[0095] S33, based on the tooth simplified distance projection simulation operator and the minimum contour area f mask Obtain the system matrix, and obtain the tooth simulation projection image according to the system matrix, such as Fig.10 shown.

[0096] Preferably, in step S33, Figure 3 as well as Fig. 9 As shown, according to the tooth simplified distance projection simulation operator and the minimum contour area f mask The specific method of obtaining the system matrix includes the following steps:

[0097] S330, obtaining the three-dimensional space voxel f j The projection point coordinates of the eight vertices.

[0098] Among them, Fig. 9 As shown in 2.a, the three-dimensional space voxel f is obtained. j The specific formulas for the projection point coordinates of the eight vertices include:

[0099]

[0100] in, f j =(X j ,Y j ,Z j ) represents the three-dimensional voxel f j The coordinate value of ΔX voxel , ΔY voxel , ΔZ voxel It means 0.5 times the voxel size in the X, Y, and Z directions. s=1,...8 represents voxel f j The eight vertex coordinates, P θ represents the projection matrix under the sampling angle θ, express The two-dimensional projection point coordinates of .

[0101] S331, such as Fig. 9 As shown in 2.b, the minimum rectangular area containing the eight projection points and the area A of the minimum rectangular area are obtained. τ ; The minimum rectangular area is marked with a blue rectangle. Assume that a single photon passes through voxel f j And the probability of being detected in the minimum rectangular area is 1.

[0102] S332, obtaining the area A of the region overlapping the minimum rectangular region p .

[0103] Assume that the coordinates of the i-th pixel of the defined projection image are (u k ,v l ).like Fig. 9 As shown in 2.c, the area A p exist Fig. 9 The 2.c part is marked by a red rectangle.

[0104] S333, obtain the coefficients a of the system matrix θ [i,j]=A p / A τ .

[0105] Where i = 1,...N u ×N v ,j=1,...N x ×N y ×N z , θ represents the sampling angle, N u 、N v Indicates the resolution of the projected image, N x 、N y 、N z Represents the resolution of volume data in three-dimensional space, the system matrix A θ N u ×N v Row and N x ×N y ×N z List.

[0106] The volume data is the minimum contour area f of a three-dimensional voxel (Volume Pixel, volume element) mask The tooth simulation projection image proj = A θ *Xr teeth Among them, Xr teeth Represents the grayscale vector value of a single tooth, A θ represents the system matrix, and θ represents the sampling angle.

[0107] As a preferred technical solution, Figure 4 As shown, the specific method for obtaining the three-dimensional space coordinates of the feature point includes the following steps:

[0108] S40, select any two different visions and obtain the projection matrix M of the two visions 1 and M 2 ;

[0109] S41, according to the two visual projection matrices M1 and M 2 , obtain the three-dimensional space coordinates of the feature point.

[0110] According to the mathematical relations of projective geometry, such as Fig.12 As shown, for any point P in space, if the camera C 1 (The present invention corresponds to the X-ray source projected at the viewing angle 1) projected to the point p on the plane 1 , using camera C 2 (The present invention corresponds to the X-ray source projected at the viewing angle 2) projected to the point p on the plane 2 The feature points related to the root angle calculation under different viewing angles are extracted by using edge detection, gradient calculation and other technologies. Any two different viewing angles are selected, and the projection matrices of the two viewing angles can be obtained according to the set geometric parameters. That is, the projection matrices of the two visual angles are known to be M 1 and M 2 , we can get the two visual projection matrix formulas as follows: as well as

[0111] Among them, (u 1 ,v 1 ,1) and (u 2 ,v 2 , 1) respectively represent the first projection point p of the feature point under the two visions 1 And the second projection point p 2 The homogeneous coordinates in each projection image, (X, Y, Z, 1) represents the homogeneous coordinates of the feature point P in the world coordinate system, Represents the projection matrix M k The element in the i-th row and j-th column of , k=1,2.

[0112] As a preferred technical solution, according to the two visual projection matrices M 1 and M 2 The specific method for obtaining the three-dimensional space coordinates of the feature points includes the following steps:

[0113] According to the two visual projection matrix formulas, four linear equations about X, Y, and Z are obtained:

[0114] The three-dimensional spatial coordinates (X, Y, Z) of the feature point are obtained according to the four linear equations.

[0115] According to the two visual projection matrices M 1 and M 2 formula, while eliminating Z c1 and Zc2 , you can get the above four linear equations about X, Y, and Z.

[0116] like Fig.12 As shown, since the space point P is O 1 p 1 With O 2 p 2 The intersection point of , it must satisfy the above four linear equations at the same time, so the values ​​of the three-dimensional space coordinates (X, Y, Z) can be obtained according to the above four linear equations.

[0117] Preferably, in practical applications, since data always has noise, the three coordinate values ​​X, Y, and Z of the spatial point P can be calculated using the least squares method.

[0118] like Figure 5 As shown, the three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images also includes the following steps:

[0119] S6, obtaining multiple sets of three-dimensional space coordinates corresponding to the two different visions;

[0120] S7, calculating the average of the plurality of three-dimensional space coordinates, and taking the average as the final three-dimensional space coordinate value.

[0121] Specifically, according to the projection point p under two different perspectives 1 、p 2 And the corresponding projection geometric parameters are used to calculate the spatial point P, which is also the three-dimensional spatial coordinates (X, Y, Z) of the feature point P.

[0122] In step S7, the average of the plurality of three-dimensional space coordinates is calculated, and the purpose of taking the average as the final three-dimensional space coordinate value is to reduce errors.

[0123] As a preferred technical solution, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the three-dimensional intelligent measurement method of tooth root angles based on oral CBCT images.

[0124] In summary, the three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images has the following beneficial effects:

[0125] 1. Based on the most advanced visual Mamba neural network architecture, a Teeth-MambaSeg tooth segmentation neural network model suitable for tooth segmentation in oral CBCT images was constructed. A tooth number ID recognition module was added to the Teeth-MambaSeg tooth segmentation neural network model, which can output tooth segmentation results one by one according to the number.

[0126] 2. An innovative tooth simplified distance projection simulation operator suitable for CBCT imaging equipment is created and applied to the tooth projection of oral CBCT. The output classification results of the tooth segmentation network are used to extract the three-dimensional spatial area of ​​a single tooth and move it to the center of the projection field of view to generate a simulated two-dimensional projection sequence for each tooth.

[0127] 3. Based on the oral CBCT image, using the set computer level, the measurement points can be automatically calibrated directly in the three-dimensional image of the teeth, thereby effectively reducing the human error in the positioning of the measurement points and improving the stability of the measurement.

[0128] 4. Obtain the three-dimensional measurement value of the root angle based on the three-dimensional spatial coordinates of the feature points. The root angle measurement in three-dimensional space is to calculate the angle of the three-dimensional straight line. This method itself is more in line with the real medical definition and is more accurate than the projection straight line angle measurement method based on two-dimensional sections in clinical routine. It avoids calculation errors caused by inappropriate selection of two-dimensional sections and improves the accuracy of root angle measurement.

[0129] 5. It can save doctors' treatment time, improve doctors' treatment efficiency, reduce the occurrence of misdiagnosis, and can be used as a reliable clinical tooth root angle measurement method.

[0130] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are 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.

[0131] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images, characterized in that: The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images comprises the following steps: Construct a Teeth-MambaSeg neural network model for teeth segmentation in oral CBCT images to be measured; Constructing a training database to train the Teeth-MambaSeg teeth segmentation neural network model; Performing tooth segmentation on the oral CBCT image to be measured by using the trained Teeth-MambaSeg tooth segmentation neural network model, and projecting a single segmented tooth onto a simulated two-dimensional projection plane to obtain a tooth simulation projection image; Extracting feature points related to tooth root angle calculation according to the tooth simulation projection image, and obtaining the three-dimensional spatial coordinates of the feature points; According to the three-dimensional spatial coordinates of the feature points, a three-dimensional measurement value of the tooth root angle is obtained.

2. A three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images as claimed in claim 1, characterized in that: The specific method of performing tooth segmentation on the oral CBCT image to be measured by using the trained Teeth-MambaSeg tooth segmentation neural network model and projecting a single segmented tooth onto a simulated two-dimensional projection plane to obtain a tooth simulation projection image comprises the following steps: Performing tooth segmentation on the oral CBCT image to be measured by using the trained Teeth-MambaSeg tooth segmentation neural network model to obtain tooth number classification and segmentation results; Extract the three-dimensional structure of the single tooth with the corresponding number, and set the minimum contour area f of the three-dimensional space voxel of the single tooth with the corresponding number mask ; Creating a tooth simplified distance projection simulation operator, and moving the single tooth with the corresponding number to the center of the three-dimensional coordinate axis of the tooth simplified distance projection simulation operator; According to the tooth simplified distance projection simulation operator and the minimum contour area f mask A system matrix is ​​obtained, and the tooth simulation projection image is obtained according to the system matrix.

3. The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images according to claim 2, characterized in that: According to the tooth simplified distance projection simulation operator and the minimum contour area f mask The specific method of obtaining the system matrix includes the following steps: Get the three-dimensional space voxel f j The projection point coordinates of the eight vertices; Get the minimum rectangular area containing the eight projection points and the area A of the minimum rectangular area τ ; Get the area A of the region overlapping with the minimum rectangular area p ; Get the coefficients a of the system matrix θ [i,j]=A p / A τ ; Where i = 1,...N u ×N v ,j=1,...N x ×N y ×N z , θ represents the sampling angle, N u 、N v Indicates the resolution of the projected image, N x 、N y 、N z The resolution of volume data representing three-dimensional space.

4. The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images according to claim 3, characterized in that: Get the three-dimensional space voxel f j The specific formulas for the projection point coordinates of the eight vertices include: in, f j =(X j ,Y j ,Z j ) represents the three-dimensional voxel f j The coordinate value of .

5. The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images according to claim 1, characterized in that: The specific method for obtaining the three-dimensional space coordinates of the feature point includes the following steps: Select any two different visions and obtain the projection matrices M1 and M2 of the two visions; According to the two visual projection matrices M1 and M2, the three-dimensional space coordinates of the feature points are obtained.

6. The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images according to claim 5, characterized in that: The formulas of the two visual projection matrices are Wherein, (u1, v1, 1) and (u2, v2, 1) represent the homogeneous coordinates of the first projection point p1 and the second projection point p2 of the feature point in the two visual systems in their respective projection images, and (X, Y, Z, 1) represents the homogeneous coordinates of the feature point P in the world coordinate system. Represents the i-th row and j-th column element of the projection matrix Mk, k=1,2.

7. The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images according to claim 6, characterized in that: According to the two visual projection matrices M1 and M2, a specific method for obtaining the three-dimensional space coordinates of the feature points comprises the following steps: According to the two visual projection matrix formulas, four linear equations about X, Y, and Z are obtained The three-dimensional spatial coordinates (X, Y, Z) of the feature point are obtained according to the four linear equations.

8. The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images according to claim 7, characterized in that: According to the three-dimensional spatial coordinates of the feature points, a specific method for obtaining the three-dimensional measurement value of the tooth root angle comprises the following steps: The three-dimensional spatial coordinates of each of the feature points are obtained, and the three-dimensional measurement value of the tooth root angle is obtained according to a spatial straight line formed by connecting the three-dimensional spatial coordinates of each of the feature points.

9. The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images according to claim 8, characterized in that: The three-dimensional intelligent measurement method of tooth root angle based on oral CBCT images also includes the following steps: Obtaining multiple sets of three-dimensional space coordinates corresponding to the two different visions; The average value of the three-dimensional space coordinates is calculated, and the average value is used as the final three-dimensional space coordinate value.

10. The three-dimensional intelligent measurement method for tooth root angle based on oral CBCT images according to claim 2, wherein the tooth simulation projection image proj = A θ *Xr teeth ;in, XR teeth Represents the grayscale vector value of a single tooth, A θ represents the system matrix, and θ represents the sampling angle.

Images