An immediate implant evaluation method for upper anterior teeth based on alveolar bone segmentation

Abstract

The application discloses an evaluation method for immediate implant of upper anterior teeth based on alveolar bone segmentation, which comprises the following steps: obtaining a region of interest of the whole oral cavity through an oral cavity segmentation model to reduce the interference information of other bone tissues in a CBCT image; obtaining the segmentation result and tooth features of the teeth by using a tooth segmentation model; then using the tooth features to assist the alveolar bone segmentation model to make the segmentation result of the alveolar bone more accurate; and automatically evaluating the comprehensive structure parameters of the labial side and the lingual side of the alveolar bone according to the segmentation result of the teeth and the alveolar bone, finally comparing the comprehensive structure parameters with the standard value of the immediate implant of the upper anterior teeth to obtain the evaluation result of the immediate implant. The application realizes the segmentation of the alveolar bone of the upper anterior teeth based on the CBCT data and the automatic evaluation of the immediate implant, and does not need to rely on the subjective experience of doctors, so that the evaluation is more comprehensive, the precision is higher and the time consumption is shorter.

Description

Technical Field

[0001] This invention belongs to the field of medical image processing, and in particular relates to a method for immediate implant assessment of maxillary anterior teeth based on alveolar bone segmentation. Background Technology

[0002] Dental health plays a crucial role in oral function and appearance. In modern dental practice, immediate implantation, as a cutting-edge dental restoration technique, not only restores the function and aesthetics of missing teeth but also reduces patients' waiting time after tooth extraction, thus providing a more convenient oral rehabilitation solution. It is widely used in the anterior maxillary region. However, the feasibility of immediate implantation surgery directly depends on the health of the alveolar bone. Therefore, achieving precise segmentation and automated assessment of the alveolar bone is essential for improving the efficiency and accuracy of preoperative decision-making in immediate implantation surgery.

[0003] Chinese invention patent application CN115661141A, published on January 31, 2023, entitled "A Method and System for Segmenting Teeth and Alveolar Bone Based on CBCT Images," discloses a method and system for segmenting teeth and alveolar bone based on CBCT images. This method first uses a deep learning model to extract the Region of Interest (ROI) and tooth centroids from CBCT images. Then, it inputs the ROI image and the tooth centroid information into a deep learning-based tooth and alveolar bone model to obtain the segmentation results. While this method achieves tooth and alveolar bone segmentation, it does not use targeted deep learning methods to improve the segmentation accuracy of the tooth and alveolar bone contours. It does not fully consider the constraints of tooth features on the alveolar bone contours and fails to address the problem of abnormal alveolar bone contour segmentation at corresponding tooth positions. This results in inaccurate intersection edges between the alveolar bone and teeth in the segmentation results, affecting the accuracy of subsequent immediate implantation decisions.

[0004] Chinese invention patent application CN113112477A, ​​published on July 13, 2021, entitled "An Artificial Intelligence-Based Measurement and Analysis Method for Immediate Anterior Tooth Implantation," discloses an artificial intelligence-based method for immediate anterior tooth implantation measurement and analysis. This method uses CBCT data and oral surface STL data as input into a neural network, which directly outputs tissue index measurement parameters for designated points on the anterior teeth. While this method can directly assess the tissue index parameters of the anterior teeth, it suffers from measurement accuracy errors because it uses a neural network to generate these parameters directly from image data. Furthermore, the fixed measurement points introduce randomness and cannot represent the overall condition of the alveolar bone. Moreover, the output parameters require the dentist to combine them with experience to determine whether immediate implantation is feasible, impacting the efficiency and accuracy of immediate implantation assessment. Summary of the Invention

[0005] To overcome the shortcomings of existing preoperative assessments for immediate implantation, such as reliance on doctors' subjective experience, incomplete assessments, insufficient accuracy, and time consumption, this invention provides an immediate implantation assessment method for maxillary anterior teeth based on alveolar bone segmentation. Building upon existing technologies, this method further enables automatic assessment of immediate implantation by segmenting the alveolar bone of maxillary anterior teeth based on CBCT data. It does not rely on doctors' subjective experience, provides a more comprehensive assessment, higher accuracy, and shorter time consumption.

[0006] The technical solution adopted by this invention to solve its technical problem is:

[0007] An immediate implant assessment method for maxillary anterior teeth based on alveolar bone segmentation includes the following steps:

[0008] Step S1: Input the original head CBCT image into the oral cavity segmentation model to obtain the segmentation result of the oral cavity. Then, crop the original CBCT image according to the segmentation result to obtain the oral cavity region of interest I. ROI ;

[0009] Step S2: I of the oral cavity ROI Input into the tooth segmentation module to obtain tooth segmentation result I t and dental features F t ;

[0010] Step S3: I of the oral cavity ROI With dental features F t Input into the alveolar bone segmentation module to obtain alveolar bone segmentation result I e ;

[0011] Step S4: Segment the teeth into segments I t Alveolar bone segmentation results I e Input the data into the immediate planting assessment module to obtain the immediate planting assessment results.

[0012] Furthermore, in step S1, the oral cavity segmentation model includes a feature encoding network and a feature decoding network. The feature encoding network contains five convolutional blocks and a downsampling operation, and the feature decoding network contains four convolutional blocks, an upsampling operation, and skip connections.

[0013] Preferably, step S2 includes the following sub-steps:

[0014] Step 1.1: Input the original head CBCT image into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F1, whose dimension is...

[0015] Step 1.2: Perform max pooling on feature map F1 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F2, whose dimension is...

[0016] Step 1.3: Perform max pooling on feature map F2 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F3, whose dimension is...

[0017] Step 1.4: Perform max pooling on feature map F3 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F4, whose dimension is...

[0018] Step 1.5: Perform max pooling on feature map F4 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F5, whose dimension is...

[0019] Step 1.6: Upsample feature map F5, then concatenate it with feature map F4 at the channel level, and input it into a convolution block with a kernel size of 3×3×3 to obtain feature map F. up1 Its dimension size is the same as that of F4;

[0020] Step 1.7: Transfer the feature map F up1 Upsampling is performed, followed by channel-level concatenation with feature map F3, and then input into a convolutional block with a kernel size of 3×3×3 to obtain feature map F. up2 Its dimension size is the same as that of F3;

[0021] Step 1.8: Transfer feature map F up2 Upsampling is performed, followed by channel-level concatenation with feature map F2, and then input into a convolutional block with a kernel size of 3×3×3 to obtain feature map F. up3 Its dimension size is the same as that of F2;

[0022] Step 1.9: Transfer the feature map F up3 Upsampling is performed, followed by channel-level concatenation with feature map F1, and then input into a convolutional block with a kernel size of 3×3×3 to obtain feature map F. up4 Its dimension size is the same as that of F1;

[0023] Step 1.10: Transfer the feature map F up4 Inputting the data into a convolutional block with a kernel size of 1×1×1 yields the segmentation result R. oral ;

[0024] Step 1.11: Based on the segmentation result R oral The corresponding original images are cropped to obtain the region of interest (ROI) in the oral cavity from the head CBCT images. ROI .

[0025] Furthermore, in step S2, the tooth segmentation module includes seven convolutional blocks, four downsampling operations, and two upsampling operations. Each convolutional block consists of two concatenated convolutional layers. Step S2 includes the following sub-steps:

[0026] Step 2.1: I of the oral cavity ROI The input is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F1, which has a dimension of C. t1 ×D t1 ×H t1 ×W t1 ;

[0027] Step 2.2: Perform max pooling on feature map F1 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F2, whose dimension is C. t2 ×D t2 ×H t2 ×W t2 ;

[0028] Step 2.3: Perform max pooling on feature map F2 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F3, whose dimension is C. t3 ×D t3 ×H t3 ×W t3 ;

[0029] Step 2.4: Perform max pooling on feature map F3 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F4, whose dimension is C. t4 ×D t4 ×H t4 ×W t4 ;

[0030] Step 2.5: Upsample feature map F4 and input it into a convolution block with a kernel size of 1×1×1 to obtain tooth feature map F. tooth Its dimension size is the same as that of F1.

[0031] Step 2.6: Perform max pooling on feature map F4 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F5, whose dimension is C. t4 ×D t4 ×H t4 ×W t4 ;

[0032] Step 2.7: Upsample the feature map F5 and input it into a convolutional block with a kernel size of 1×1×1 to obtain the tooth segmentation result I. t .

[0033] Furthermore, in step S3, the alveolar bone segmentation module includes a multi-scale feature encoding module and a feature fusion decoding module; step S3 includes the following sub-steps:

[0034] Step 3.1: I of the oral cavity ROI The input is fed into the multi-scale feature encoding module to obtain the multi-scale feature map F. en ;

[0035] Step 3.2: Convert the multi-scale feature map F en And tooth feature diagram F tooth The input is fed into the feature fusion decoding module to obtain the alveolar bone segmentation result I. e .

[0036] Preferably, in step 3.1, the multi-scale feature encoding module includes five convolutional blocks and a downsampling method. Each convolutional block consists of two concatenated convolutional layers, and each convolutional layer consists of a convolution, a normalization layer, and a ReLU activation function. The process of step 3.1 is as follows:

[0037] Step 3.1.1: I of the oral cavity ROI The input is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en1 Its dimensions are C1×D1×H1×W1;

[0038] Step 3.1.2: Transfer the feature map F en1 Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en2 Its dimensions are C2×D2×H2×W2;

[0039] Step 3.1.3: Transfer the feature map F en2 Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en3 Its dimensions are C3×D3×H3×W3;

[0040] Step 3.1.4: Transfer feature map F en3 Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en4 Its dimensions are C4×D4×H4×W4;

[0041] Step 3.1.5: Transfer the feature map F en4 Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en5 Its dimensions are C5×D5×H5×W5, and the feature maps obtained from each convolutional block form feature maps F of different scales. en .

[0042] More preferably, in step 3.2, the feature fusion decoding module includes four convolutional blocks, skip connections, an upsampling method, and a tooth feature attention module. The process of step 3.2 is as follows:

[0043] Step 3.2.1: Transfer the feature map F en5 Feature map F en4 and dental features F tooth The input is fed into the tooth feature attention module to obtain the tooth-guided alveolar bone feature F. te4 Its dimensions are C4×D4×H4×W4;

[0044] Step 3.2.2: Transfer the feature map F te4 Feature map F en3 and dental features F tooth The input is fed into the tooth feature attention module to obtain the tooth-guided alveolar bone feature F. te3 Its dimensions are C3×D3×H3×W3;

[0045] Step 3.2.3: Transfer the feature map F te3 Feature map F en2 and dental features F tooth The input is fed into the tooth feature attention module to obtain the tooth-guided alveolar bone feature F. te2 Its dimensions are C2×D2×H2×W2;

[0046] Step 3.2.4: Transfer the feature map F te2 Feature map F en1 and dental features F tooth The input is fed into the tooth feature attention module to obtain the tooth-guided alveolar bone feature F. te1 Its dimensions are C1×D1×H1×W1;

[0047] Step 3.2.5: Transfer the feature map F te1 Inputting into a convolution block with a kernel size of 1×1×1 yields the alveolar bone segmentation result I. e .

[0048] More preferably, in step 3.2.1, the tooth feature F tooth The input is fed into a fully connected layer with a 1×1×1 convolutional kernel, the size of which is equal to that of the feature map F.en5 Consistent with feature map F en5 The data is concatenated by channel and upsampled to match the feature map F. en4 With the same size, it is compared with feature map F en4 The data is concatenated along channels and input into a 3×3×3 convolutional block. After normalization and ReLU activation, the tooth-guided alveolar bone features F are obtained. te4 Its dimensions are C4×D4×H4×W4;

[0049] In step 3.2.2, the tooth feature F tooth The input is fed into a fully connected layer with a 1×1×1 convolutional kernel, the size of which is equal to that of the feature map F. te4 Consistent with feature map F te4 The data is concatenated by channel and upsampled to match the feature map F. en3 With the same size, it is compared with feature map F en3 The data is concatenated along channels and input into a 3×3×3 convolutional block. After normalization and ReLU activation, the tooth-guided alveolar bone features F are obtained. te3 Its dimensions are C3×D3×H3×W3;

[0050] In step 3.2.3, the tooth feature F tooth The input is fed into a fully connected layer with a 1×1×1 convolutional kernel, the size of which is equal to that of the feature map F. te3 Consistent with feature map F te3 The data is concatenated by channel and upsampled to match the feature map F. en2 With the same size, it is compared with feature map F en2 The data is concatenated along channels and input into a 3×3×3 convolutional block. After normalization and ReLU activation, the tooth-guided alveolar bone features F are obtained. te2 Its dimensions are C2×D2×H2×W2;

[0051] In step 3.2.4, the tooth feature F tooth The input is fed into a fully connected layer with a 1×1×1 convolutional kernel, the size of which is equal to that of the feature map F. te2 Consistent with feature map F te2 The data is concatenated by channel and upsampled to match the feature map F. en1 With the same size, it is compared with feature map F en4 The data is concatenated along channels and input into a 3×3×3 convolutional block. After normalization and ReLU activation, the tooth-guided alveolar bone features F are obtained. te1 Its dimensions are C1×D1×H1×W1.

[0052] Furthermore, step S4 includes the following sub-steps:

[0053] Step 4.1: Segment the teeth into sections I t Alveolar bone segmentation results I e The non-zero points in the set are denoted as the tooth point set D. t With alveolar bone point set D e ;

[0054] Step 4.2: Set the alveolar bone point D e Based on the formula for calculating the major axis vector, the alveolar bone major axis vector v is obtained. e The formula is as follows:

[0055]

[0056] Step 4.3: Set the tooth point set D t The highest point p on the z-axis t As the highest point of the tooth foramen, combined with the alveolar bone long axis vector v e Create a cross section B parallel to the y-axis. e , by section B e Alveolar bone point set D e Divided into the labial alveolar bone point set D ef With the set of alveolar bone points near the lingual side D eb ;

[0057] Step 4.4: Based on the alveolar bone point set D e Extracting the point set D from the outer surface of the alveolar bone ew D, the point set on the outer surface of the alveolar bone ew respectively with the set of alveolar bone points D on the labial side ef and the set of alveolar bone points near the lingual side D eb Take the union of the sets to obtain the point set D on the outer surface of the alveolar bone near the labial side. efw Point set D of the lingual alveolar bone outer surface ebw ;

[0058] Step 4.5: Based on the tooth point set D t With alveolar bone point set D e Based on the coordinate adjacency relationship, the point set D on the tooth fork surface is extracted. h ;

[0059] Step 4.6, Point set D on the near-labial alveolar bone outer surface efw With the point set D on the tooth fork surface h The thinnest thickness d of the alveolar bone wall near the labial side is obtained by calculating the Euclidean distance at each point. f The formula is as follows:

[0060]

[0061] Step 4.7: Take the alveolar bone point set D near the lingual side. eb The lowest point p on the z-axis b p, the lowest point of the alveolar bone near the labial side b , and the highest point p of the tooth hole t Connect the lines, and take point p at the 1 / 3 mark of the line. q Take section B perpendicular to the alveolar bone. e normal direction vector v c From point p q Starting from vector v c Create a ray and determine whether the elements ray passes through are contained in the near-lingual alveolar bone point set D. eb The point p at the 1 / 3 of the palatal root was obtained. e ;

[0062] Step 4.8, point p on the lateral root of the palate (1 / 3 of the distance) e Starting with the alveolar bone long axis vector v t Create a ray and determine whether the elements ray passes through are contained in the point set D on the outer surface of the alveolar bone near the lingual side. ebw The point p on the outer surface through which the ray passes is obtained. eb1 According to the highest point p of the tooth hole t With the outer surface point p eb1 The coordinates are used to obtain the bone wall thickness d along the long axis of the lingual alveolar bone. e1 ;

[0063] Step 4.9, point p on the lateral root of the palate (1 / 3 of the distance) e Starting from section B perpendicular to the alveolar bone e normal direction vector v c Create a ray and determine whether the elements ray passes through are contained in the point set D on the outer surface of the alveolar bone near the lingual side. ebw The point p on the outer surface through which the ray passes is obtained. eb2 According to the highest point p of the tooth hole t With the outer surface point p eb3 The coordinates are used to obtain the bone wall thickness d in the vertical section direction of the alveolar bone near the lingual side. e2 ;

[0064] Step 4.10: Based on the thinnest thickness d of the alveolar bone wall near the labial side. f Whether it exceeds the standard value for the near-labial side, and the bone wall thickness d along the long axis of the near-lingual alveolar bone. e1 The bone wall thickness d in the vertical cross-section of the alveolar bone near the lingual side e2 Whether it exceeds the standard value near the lingual side, the immediate implantation assessment result is obtained.

[0065] The main advantages of this invention are: it enables the automatic segmentation of alveolar bone in the anterior position based on CBCT data and immediate implant assessment, without relying on the doctor's subjective experience, and the assessment is more comprehensive, accurate, and time-saving. Attached Figure Description

[0066] Figure 1 This is a flowchart of the assessment method for immediate implantation of maxillary anterior teeth based on alveolar bone segmentation;

[0067] Figure 2 This is a technical framework diagram of an immediate implant assessment method for maxillary anterior teeth based on alveolar bone segmentation;

[0068] Figure 3 This is a schematic diagram of the tooth segmentation module structure;

[0069] Figure 4 This is a schematic diagram of the alveolar bone segmentation module structure;

[0070] Figure 5 This is a sample image of immediate implant assessment based on CBCT tooth and alveolar bone segmentation. Detailed Implementation

[0071] The present invention will now be further described with reference to the accompanying drawings.

[0072] Reference Figures 1-5 An immediate implant assessment method for maxillary anterior teeth based on alveolar bone segmentation includes the following steps:

[0073] Step S1: Input the original head CBCT image into the oral cavity segmentation model to obtain the segmentation result of the oral cavity. Then, crop the original CBCT image according to the segmentation result to obtain the oral cavity region of interest I. ROI ;

[0074] In step S1, the oral cavity segmentation model includes a feature encoding network and a feature decoding network. The feature encoding network contains five convolutional blocks and a downsampling operation, and the feature decoding network contains four convolutional blocks, an upsampling operation, and skip connections.

[0075] Step S2 includes the following sub-steps:

[0076] Step 1.1: Input the original head CBCT image into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F1, whose dimension is...

[0077] Step 1.2: Perform max pooling on feature map F1 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F2, whose dimension is...

[0078] Step 1.3: Perform max pooling on feature map F2 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F3, whose dimension is...

[0079] Step 1.4: Perform max pooling on feature map F3 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F4, whose dimension is...

[0080] Step 1.5: Perform max pooling on feature map F4 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F5, whose dimension is...

[0081] Step 1.6: Upsample feature map F5, then concatenate it with feature map F4 at the channel level, and input it into a convolution block with a kernel size of 3×3×3 to obtain feature map F. up1 Its dimension size is the same as that of F4;

[0082] Step 1.7: Transfer the feature map F up1 Upsampling is performed, followed by channel-level concatenation with feature map F3, and then input into a convolutional block with a kernel size of 3×3×3 to obtain feature map F. up2 Its dimension size is the same as that of F3;

[0083] Step 1.8: Transfer feature map F up2 Upsampling is performed, followed by channel-level concatenation with feature map F2, and then input into a convolutional block with a kernel size of 3×3×3 to obtain feature map F. up3 Its dimension size is the same as that of F2;

[0084] Step 1.9: Transfer the feature map F up3 Upsampling is performed, followed by channel-level concatenation with feature map F1, and then input into a convolutional block with a kernel size of 3×3×3 to obtain feature map F. up4 Its dimension size is the same as that of F1;

[0085] Step 1.10: Transfer the feature map F up4 Inputting the data into a convolutional block with a kernel size of 1×1×1 yields the segmentation result R. oral ;

[0086] Step 1.11: Based on the segmentation result R oral The corresponding original images are cropped to obtain the region of interest (ROI) in the oral cavity from the head CBCT images. ROI .

[0087] Step S2: I of the oral cavity ROI Input into the tooth segmentation module to obtain tooth segmentation result I t and dental features F t ;

[0088] like Figure 3 As shown, in step S2, the tooth segmentation module includes seven convolutional blocks, four downsampling operations, and two upsampling operations. Each convolutional block consists of two concatenated convolutional layers. Step S2 includes the following sub-steps:

[0089] Step 2.1: I of the oral cavity ROI The input is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F1, which has a dimension of C. t1 ×D t1 ×H t1 ×W t1 ;

[0090] Step 2.2: Perform max pooling on feature map F1 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F2, whose dimension is C. t2 ×D t2 ×H t2 ×W t2 ;

[0091] Step 2.3: Perform max pooling on feature map F2 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F3, whose dimension is C. t3 ×D t3 ×H t3 ×W t3 ;

[0092] Step 2.4: Perform max pooling on feature map F3 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F4, whose dimension is C. t4 ×D t4 ×H t4 ×W t4 ;

[0093] Step 2.5: Upsample feature map F4 and input it into a convolution block with a kernel size of 1×1×1 to obtain tooth feature map F. tooth Its dimension size is the same as that of F1;

[0094] Step 2.6: Perform max pooling on feature map F4 and input it into a convolutional block with a kernel size of 3×3×3 to obtain output feature map F5, whose dimension is C. t4 ×D t4 ×H t4 ×Wt4 ;

[0095] Step 2.7: Upsample the feature map F5 and input it into a convolutional block with a kernel size of 1×1×1 to obtain the tooth segmentation result I. t .

[0096] Step S3: I of the oral cavity ROI With dental features F t Input into the alveolar bone segmentation module to obtain alveolar bone segmentation result I e ;

[0097] like Figure 4 As shown, in step S3, the alveolar bone segmentation module includes a multi-scale feature encoding module and a feature fusion decoding module; step S3 includes the following sub-steps:

[0098] Step 3.1: I of the oral cavity ROI The input is fed into the multi-scale feature encoding module to obtain the multi-scale feature map F. en ;

[0099] Step 3.2: Convert the multi-scale feature map F en And tooth feature diagram F tooth The input is fed into the feature fusion decoding module to obtain the alveolar bone segmentation result I. e .

[0100] Preferably, in step 3.1, the multi-scale feature encoding module includes five convolutional blocks and a downsampling method. Each convolutional block consists of two concatenated convolutional layers, and each convolutional layer consists of a convolution, a normalization layer, and a ReLU activation function. The process of step 3.1 is as follows:

[0101] Step 3.1.1: I of the oral cavity ROI The input is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en1 Its dimensions are C1×D1×H1×W1;

[0102] Step 3.1.2: Transfer the feature map F en1 Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en2 Its dimensions are C2×D2×H2×W2;

[0103] Step 3.1.3: Transfer the feature map F en2 Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en3 Its dimensions are C3×D3×H3×W3;

[0104] Step 3.1.4: Transfer feature map F en3 Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en4 Its dimensions are C4×D4×H4×W4;

[0105] Step 3.1.5: Transfer the feature map F en4 Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map F. en5 Its dimensions are C5×D5×H5×W5, and the feature maps obtained from each convolutional block form feature maps F of different scales. en .

[0106] More preferably, in step 3.2, the feature fusion decoding module includes four convolutional blocks, skip connections, an upsampling method, and a tooth feature attention module. The process of step 3.2 is as follows:

[0107] Step 3.2.1: Transfer the feature map F en5 Feature map F en4 and dental features F tooth The input is fed into the tooth feature attention module to obtain the tooth-guided alveolar bone feature F. te4 Its dimensions are C4×D4×H4×W4;

[0108] Step 3.2.2: Transfer the feature map F te4 Feature map F en3 and dental features F tooth The input is fed into the tooth feature attention module to obtain the tooth-guided alveolar bone feature F. te3 Its dimensions are C3×D3×H3×W3;

[0109] Step 3.2.3: Transfer the feature map F te3 Feature map F en2 and dental features F tooth The input is fed into the tooth feature attention module to obtain the tooth-guided alveolar bone feature F. te2 Its dimensions are C2×D2×H2×W2;

[0110] Step 3.2.4: Transfer the feature map F te2 Feature map F en1 and dental features F tooth The input is fed into the tooth feature attention module to obtain the tooth-guided alveolar bone feature F. te1 Its dimensions are C1×D1×H1×W1;

[0111] Step 3.2.5: Transfer the feature map F te1Inputting into a convolution block with a kernel size of 1×1×1 yields the alveolar bone segmentation result I. e .

[0112] More preferably, in step 3.2.1, the tooth feature F tooth The input is fed into a fully connected layer with a 1×1×1 convolutional kernel, the size of which is equal to that of the feature map F. en5 Consistent with feature map F en5 The data is concatenated by channel and upsampled to match the feature map F. en4 With the same size, it is compared with feature map F en4 The data is concatenated along channels and input into a 3×3×3 convolutional block. After normalization and ReLU activation, the tooth-guided alveolar bone features F are obtained. te4 Its dimensions are C4×D4×H4×W4;

[0113] In step 3.2.2, the tooth feature F tooth The input is fed into a fully connected layer with a 1×1×1 convolutional kernel, the size of which is equal to that of the feature map F. te4 Consistent with feature map F te4 The data is concatenated by channel and upsampled to match the feature map F. en3 With the same size, it is compared with feature map F en3 The data is concatenated along channels and input into a 3×3×3 convolutional block. After normalization and ReLU activation, the tooth-guided alveolar bone features F are obtained. te3 Its dimensions are C3×D3×H3×W3;

[0114] In step 3.2.3, the tooth feature F tooth The input is fed into a fully connected layer with a 1×1×1 convolutional kernel, the size of which is equal to that of the feature map F. te3 Consistent with feature map F te3 The data is concatenated by channel and upsampled to match the feature map F. en2 With the same size, it is compared with feature map F en2 The data is concatenated along channels and input into a 3×3×3 convolutional block. After normalization and ReLU activation, the tooth-guided alveolar bone features F are obtained. te2 Its dimensions are C2×D2×H2×W2;

[0115] In step 3.2.4, the tooth feature F tooth The input is fed into a fully connected layer with a 1×1×1 convolutional kernel, the size of which is equal to that of the feature map F. te2 Consistent with feature map F te2 The data is concatenated by channel and upsampled to match the feature map F.en1 With the same size, it is compared with feature map F en4 The data is concatenated along channels and input into a 3×3×3 convolutional block. After normalization and ReLU activation, the tooth-guided alveolar bone features F are obtained. te1 Its dimensions are C1×D1×H1×W1.

[0116] Step S4: Segment the teeth into segments I t Alveolar bone segmentation results I e Input the data into the immediate planting assessment module to obtain the immediate planting assessment results.

[0117] Step S4 includes the following sub-steps:

[0118] Step 4.1: Segment the teeth into sections I t Alveolar bone segmentation results I e The non-zero points in the set are denoted as the tooth point set D. t With alveolar bone point set D e ;

[0119] Step 4.2: Set the alveolar bone point D e Based on the formula for calculating the major axis vector, the alveolar bone major axis vector v is obtained. e The formula is as follows:

[0120]

[0121] Step 4.3: Set the tooth point set D t The highest point p on the z-axis t As the highest point of the tooth foramen, combined with the alveolar bone long axis vector v e Create a cross section B parallel to the y-axis. e , by section B e Alveolar bone point set D e Divided into the labial alveolar bone point set D ef With the set of alveolar bone points near the lingual side D eb ;

[0122] Step 4.4: Based on the alveolar bone point set D e Extracting the point set D from the outer surface of the alveolar bone ew D, the point set on the outer surface of the alveolar bone ew respectively with the set of alveolar bone points D on the labial side ef and the set of alveolar bone points near the lingual side D eb Take the union of the sets to obtain the point set D on the outer surface of the alveolar bone near the labial side. efw Point set D of the lingual alveolar bone outer surface ebw ;

[0123] Step 4.5: Based on the tooth point set D t With alveolar bone point set De Based on the coordinate adjacency relationship, the point set D on the tooth fork surface is extracted. h ;

[0124] Step 4.6, Point set D on the near-labial alveolar bone outer surface efw With the point set D on the tooth fork surface h The thinnest thickness d of the alveolar bone wall near the labial side is obtained by calculating the Euclidean distance at each point. f The formula is as follows:

[0125]

[0126] Step 4.7: Take the alveolar bone point set D near the lingual side. eb The lowest point p on the z-axis b p, the lowest point of the alveolar bone near the labial side b , and the highest point p of the tooth hole t Connect the lines, and take point p at the 1 / 3 mark of the line. q Take section B perpendicular to the alveolar bone. e normal direction vector v c From point p q Starting from vector v c Create a ray and determine whether the elements ray passes through are contained in the near-lingual alveolar bone point set D. eb The point p at the 1 / 3 of the palatal root was obtained. e ;

[0127] Step 4.8, P at the 1 / 3 of the palatal root. e Starting with the alveolar bone long axis vector v t Create a ray and determine whether the elements ray passes through are contained in the point set D on the outer surface of the alveolar bone near the lingual side. ebw The point p on the outer surface through which the ray passes is obtained. eb1 According to the highest point p of the tooth hole t With the outer surface point p eb1 The coordinates are used to obtain the bone wall thickness d along the long axis of the lingual alveolar bone. e1 ;

[0128] Step 4.9, point p on the lateral root of the palate (1 / 3 of the distance) e Starting from section B perpendicular to the alveolar bone e normal direction vector v c Create a ray and determine whether the elements ray passes through are contained in the point set D on the outer surface of the alveolar bone near the lingual side. ebw The point p on the outer surface through which the ray passes is obtained. eb2 According to the highest point p of the tooth hole t With the outer surface point p eb3 The coordinates are used to obtain the bone wall thickness d in the vertical section direction of the alveolar bone near the lingual side. e2 ;

[0129] Step 4.10: Based on the thinnest thickness d of the alveolar bone wall near the labial side. f Whether it exceeds the standard value for the near-labial side, and the bone wall thickness d along the long axis of the near-lingual alveolar bone. e1 The bone wall thickness d in the vertical cross-section of the alveolar bone near the lingual side e2 Whether it exceeds the standard value near the lingual side, the immediate implantation assessment result is obtained.

[0130] It should be noted that the example image of immediate implant assessment based on CBCT tooth and alveolar bone segmentation in this embodiment is as follows: Figure 5 As shown.

[0131] It should be noted that in this embodiment, C is the number of channels, D is the depth of the image, H is the height of the image, W is the width of the image, and the subscript of the letter indicates the sequence number, which is used to distinguish the dimensions of different feature maps.

[0132] This embodiment uses an oral cavity segmentation model to obtain the region of interest (ROI) of the entire oral cavity as input for the alveolar bone segmentation model, significantly reducing interference from other bone tissues in CBCT images. Since immediate implantation is based on observing alveolar bone structural parameters, a tooth segmentation model is used to obtain tooth segmentation results and tooth features. These tooth features are then used to assist the alveolar bone segmentation model, making the alveolar bone segmentation more accurate. Based on the tooth and alveolar bone segmentation results, the comprehensive structural parameters of the prolabial and prolingual sides of the alveolar bone are automatically evaluated, and finally compared with the standard values ​​for immediate implantation of the maxillary anterior teeth to obtain the immediate implantation evaluation result.

[0133] The embodiments described in this specification are merely examples of implementations of the inventive concept and are for illustrative purposes only. The scope of protection of this invention should not be considered limited to the specific forms described in these embodiments; rather, it extends to equivalent technical means conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for evaluating immediate implant placement in maxillary anterior teeth based on alveolar bone segmentation, characterized in that, The method includes the following steps: Step S1: Input the original head CBCT image into the oral cavity segmentation model to obtain the segmentation result of the oral cavity, and crop the original CBCT image according to the segmentation result to obtain the region of interest in the oral cavity. ; Step S2: Select the region of interest in the oral cavity Input the data into the tooth segmentation module to obtain the tooth segmentation results. and dental features ; Step S3: Select the region of interest in the oral cavity With dental features Input the data into the alveolar bone segmentation module to obtain the alveolar bone segmentation results. The alveolar bone segmentation module includes a multi-scale feature encoding module and a feature fusion decoding module; it includes the following sub-steps: Step 3.1: Identify the region of interest in the oral cavity. The input is fed into the multi-scale feature encoding module to obtain a multi-scale feature map. ; Step 3.2: Convert the multi-scale feature map and tooth feature diagram The input is fed into the feature fusion decoding module to obtain the alveolar bone segmentation result. ; Step S4: Segment the teeth Results of alveolar bone segmentation Input the data into the immediate planting assessment module to obtain the immediate planting assessment results.

2. The immediate implant assessment method for maxillary anterior teeth based on alveolar bone segmentation as described in claim 1, characterized in that, In step S1, the oral cavity segmentation model includes a feature encoding network and a feature decoding network. The feature encoding network contains five convolutional blocks and a downsampling operation, and the feature decoding network contains four convolutional blocks, an upsampling operation, and skip connections.

3. The immediate implant assessment method for maxillary anterior teeth based on alveolar bone segmentation as described in claim 2, characterized in that, Step S1 includes the following sub-steps: Step 1.1: Input the original head CBCT image into a convolution block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 1.2: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 1.3: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 1.4: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 1.5: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 1.6: Transfer the feature map Upsampling is performed, followed by comparison with the feature map. Channel-level concatenation is performed and input into a convolutional block with a kernel size of 3×3×3 to obtain the feature map. Its dimensional size is Consistent; Step 1.7: Transfer the feature map Upsampling is performed, followed by comparison with the feature map. Channel-level concatenation is performed and input into a convolutional block with a kernel size of 3×3×3 to obtain the feature map. Its dimensional size is Consistent; Step 1.8: Transfer the feature map Upsampling is performed, followed by comparison with the feature map. Channel-level concatenation is performed and input into a convolutional block with a kernel size of 3×3×3 to obtain the feature map. Its dimensional size is Consistent; Step 1.9: Transfer the feature map Upsampling is performed, followed by comparison with the feature map. Channel-level concatenation is performed and input into a convolutional block with a kernel size of 3×3×3 to obtain the feature map. Its dimensional size is Consistent; Step 1.10: Transfer the feature map Inputting the data into a convolutional block with a kernel size of 1×1×1 yields the segmentation result. ; Step 1.11: Based on the segmentation results The corresponding original images are cropped to obtain the region of interest in the oral cavity from the head CBCT images. .

4. A method for immediate implant assessment of maxillary anterior teeth based on alveolar bone segmentation as described in any one of claims 1 to 3, characterized in that, In step S2, the tooth segmentation module includes seven convolutional blocks, four downsampling operations, and two upsampling operations. Each convolutional block consists of two concatenated convolutional layers. Step S2 includes the following sub-steps: Step 2.1: Identify the region of interest in the oral cavity. The input is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 2.2: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 2.3: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 2.4: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 2.5: Transfer the feature map Upsampling is performed and the data is input into a convolutional block with a kernel size of 1×1×1 to obtain the tooth feature map. Its dimensional size is Consistent; Step 2.6: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 2.7: Transfer the feature map Upsampling is performed and the result is fed into a convolutional block with a kernel size of 1×1×1 to obtain the tooth segmentation result. .

5. The immediate implant assessment method for maxillary anterior teeth based on alveolar bone segmentation as described in claim 1, characterized in that, In step 3.1, the multi-scale feature encoding module includes five convolutional blocks and a downsampling method. Each convolutional block consists of two concatenated convolutional layers, and each convolutional layer consists of a convolution, a normalization layer, and a ReLU activation function. The process of step 3.1 is as follows: Step 3.1.1: Identify the region of interest in the oral cavity. The input is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 3.1.2: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 3.1.3: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 3.1.4: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is ; Step 3.1.5: Transfer the feature map Max pooling is performed, and the result is fed into a convolutional block with a kernel size of 3×3×3 to obtain the output feature map. Its dimension size is The feature maps obtained from each convolutional block form feature maps of different scales. .

6. The immediate implant assessment method for maxillary anterior teeth based on alveolar bone segmentation as described in claim 5, characterized in that, In step 3.2, the feature fusion decoding module includes four convolutional blocks, skip connections, an upsampling method, and a tooth feature attention module. The process of step 3.2 is as follows: Step 3.2.1: Transfer the feature map Feature map and tooth feature diagram The input is fed into the tooth feature attention module to obtain a tooth-guided alveolar bone feature map. Its dimension size is ; Step 3.2.2: Transfer the feature map Feature map and tooth feature diagram The input is fed into the tooth feature attention module to obtain a tooth-guided alveolar bone feature map. Its dimension size is ; Step 3.2.3: Transfer the feature map Feature map and tooth feature diagram The input is fed into the tooth feature attention module to obtain a tooth-guided alveolar bone feature map. Its dimension size is ; Step 3.2.4: Transfer the feature map Feature map and tooth feature diagram The input is fed into the tooth feature attention module to obtain a tooth-guided alveolar bone feature map. Its dimension size is ; Step 3.2.5: Transfer the feature map Inputting the data into a convolutional block with a kernel size of 1×1×1 yields alveolar bone segmentation results. .

7. The immediate implant assessment method for maxillary anterior teeth based on alveolar bone segmentation as described in claim 6, characterized in that, In step 3.2.1, the tooth feature map is... The input is fed into a fully connected layer with a 1×1×1 kernel, the size of which is equal to the feature map. Consistent with the feature map The data is concatenated by channel and then upsampled to match the feature map. With the same size, compare it with the feature map The data is concatenated along channels and fed into a 3×3×3 convolutional block. After normalization and ReLU activation, a tooth-guided alveolar bone feature map is obtained. Its dimension size is ; In step 3.2.2, the tooth feature map is... The input is fed into a fully connected layer with a 1×1×1 kernel, the size of which is equal to the feature map. Consistent with the feature map The data is concatenated by channel and then upsampled to match the feature map. With the same size, compare it with the feature map The data is concatenated along channels and fed into a 3×3×3 convolutional block. After normalization and ReLU activation, a tooth-guided alveolar bone feature map is obtained. Its dimension size is ; In step 3.2.3, the tooth feature map is... The input is fed into a fully connected layer with a 1×1×1 kernel, the size of which is equal to the feature map. Consistent with the feature map The data is concatenated by channel and then upsampled to match the feature map. With the same size, compare it with the feature map The data is concatenated along channels and fed into a 3×3×3 convolutional block. After normalization and ReLU activation, a tooth-guided alveolar bone feature map is obtained. Its dimension size is ; In step 3.2.4, the tooth feature map is... The input is fed into a fully connected layer with a 1×1×1 kernel, the size of which is equal to the feature map. Consistent with the feature map The data is concatenated by channel and then upsampled to match the feature map. With the same size, compare it with the feature map The data is concatenated along channels and fed into a 3×3×3 convolutional block. After normalization and ReLU activation, a tooth-guided alveolar bone feature map is obtained. Its dimension size is .

8. A method for immediate implant assessment of maxillary anterior teeth based on alveolar bone segmentation as described in any one of claims 1 to 3, characterized in that, Step S4 includes the following sub-steps: Step 4.1: Segment the teeth Alveolar bone segmentation results The non-zero points in the set are denoted as the tooth point set. Alveolar bone point set ; Step 4.2: Set alveolar bone points Based on the formula for calculating the major axis vector, the major axis vector of the alveolar bone is obtained. The formula is as follows: ； Step 4.3: Set the tooth points The highest point of the z-axis As the highest point of the dental foramen, combined with the long axis vector of the alveolar bone Create a cross section parallel to the y-axis. , by cross section Alveolar bone point set Divided into the set of alveolar bone points on the near-labial side set of alveolar bone points near the lingual side ; Step 4.4: Based on the alveolar bone point set Extraction of point set from the outer surface of alveolar bone Points on the outer surface of the alveolar bone Set together with the alveolar bone points on the near-labial side respectively and the set of alveolar bone points on the lingual side By taking the union of the sets, we can obtain the point sets on the outer surface of the alveolar bone near the labial side. and the set of points on the outer surface of the lingual alveolar bone ; Step 4.5: Based on the tooth point set Alveolar bone point set Based on the coordinate adjacency relationship, the point set on the tooth fork surface is extracted. ; Step 4.6: Point set on the outer surface of the near-labial alveolar bone Point set with tooth fork surface The thinnest thickness of the alveolar bone wall near the labial side is obtained by calculating the Euclidean distance at each point. The formula is as follows: ； Step 4.7: Take a sample of alveolar bone near the lingual side. The lowest point of the z-axis As the lowest point of the alveolar bone near the labia , and the highest point of the tooth foramen Connect the lines, and take the point at 1 / 3 of the line. Take a section perpendicular to the alveolar bone. normal direction vector From point Starting along vector Create a ray and determine whether the elements ray passes through are contained in the near-lingual alveolar bone point set. The 1 / 3 point on the lateral root of the palate was obtained. ; Step 4.8, move the palatal root 1 / 3 point Starting with the vector along the long axis of the alveolar bone... Create a ray and determine whether the elements traversed by the ray are contained in the point set on the outer surface of the alveolar bone near the lingual side. The points on the outer surface through which the ray passes are obtained. Based on the highest point of the tooth hole With outer surface points The coordinates are used to obtain the bone wall thickness along the long axis of the alveolar bone on the near-lingual side. ; Step 4.9, locate the 1 / 3 point on the lateral aspect of the palate. Starting from a point perpendicular to the alveolar bone section... normal direction vector Create a ray and determine whether the elements traversed by the ray are contained in the point set on the outer surface of the alveolar bone near the lingual side. The points on the outer surface through which the ray passes are obtained. Based on the highest point of the tooth hole With outer surface points The coordinates are used to obtain the bone wall thickness in the vertical section of the alveolar bone near the lingual side. ; Step 4.10: Based on the thinnest thickness of the alveolar bone wall on the near-labial side. Whether it exceeds the standard value on the near-labial side, and the bone wall thickness along the long axis of the near-lingual alveolar bone. Thickness of the bone wall in the vertical section of the alveolar bone near the lingual side Whether it exceeds the standard value near the lingual side, the immediate implantation assessment result is obtained.

Images