A method, mold and device for fabricating an immediate digitalization implant guide

Abstract

This invention proposes a method for fabricating an immediate digital implant guide, comprising: S1: acquiring a patient's dental impression and taking CBCT images; S2: identifying the positioning anchor point information on the CBCT image, registering it with the positioning anchor point coordinates of the dental model tray, placing the simulated implant according to the missing tooth location, and obtaining its spatial coordinate position information; S3: pouring quick-setting plaster to fabricate a plaster positive mold; S4: fixing the plaster positive mold onto a guide ring placement device, calculating the movement parameters of each axis of the guide ring placement device, and operating the guide ring placement device to move the guide pin and plaster positive mold to the required positions and angles respectively; S5: curing the guide ring onto the surface of the plaster positive mold using light curing, removing the guide pin, fabricating a film on the surface of the plaster positive mold, shaping and trimming, polishing, and sterilizing before use. This method improves the implantation accuracy and is simple, fast, economical, and practical. This invention also proposes a corresponding coordinate transformation mold and guide ring placement device.

Description

Technical Field

[0001] This invention relates to the field of dental technology, and more specifically to a method for fabricating an immediate digital implant guide. Background Technology

[0002] Dental implants, also known as implant-supported prostheses, are biomimetic teeth made of biomaterials. They consist of two parts: the implant (root portion) which is inserted into the jawbone for support and fixation, and the crown which performs chewing functions. Based on the changes in the alveolar bone and jawbone after tooth loss, the dentist selects an implant of a specific shape and inserts it into the jawbone as an artificial tooth root. Then, a crown is installed on the implant post that protrudes into the oral cavity, achieving a shape and function similar to a natural tooth. With the continuous development of dental implant technology, its clinical success rate and restorative effects have greatly improved, leading to its acceptance by more and more doctors and patients. However, for sites adjacent to important anatomical structures such as the nerve canal and maxillary sinus, and in cases with poor bone conditions in the implant area, accurate implant placement places high demands on the dentist. Traditional drilling methods for dental implants cannot accurately locate the holes; relying solely on the dentist's experience results in insufficient drilling precision and stability, posing surgical risks. To overcome the above problems, a simplified implant guide technique is often used. Based on information from radiographic imaging, the implantation point is located on a plaster model, an implantation channel is created, and the guide is made using self-curing plastic or thermoforming via a molding machine. This type of guide is simple to manufacture and inexpensive, and can generally guarantee the post-restoration effect. However, its fabrication only references the surface information of the model and lacks consideration of the three-dimensional structure of the bone in the implantation area, especially precise information on the width and height of the alveolar ridge in the edentulous area, as well as the distribution of nerves and sinuses in the jawbone. This results in poor accuracy and safety, and it cannot be combined with CT technology.

[0003] Meanwhile, in the field of oral medicine, cone-beam computed tomography (CBCT) is widely used in the diagnosis and reconstructive surgery of oral and maxillofacial diseases due to its advantages of low radiation, high precision, and provision of three-dimensional oral information. For example, it can be used to locate and extract impacted teeth, measure the mandibular nerve canal, and diagnose and treat jawbone lesions. In clinical practice, dentists can intuitively understand the patient's dentofacial morphology through CBCT images and use tools to measure oral and maxillofacial parameters to assist in judging the condition and developing personalized, precise treatment and reconstructive surgery plans. In particular, in some prosthetic oral surgeries, dentists need to place a tray with metal balls into the patient's mouth, then identify anchor points based on CT images and precisely register them with points in the mold's spatial coordinate system to complete accurate calculations of oral data.

[0004] The purpose of this invention is to provide a method and tool for creating digital implantation guides by combining CBCT image positioning anchor point recognition and matching. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention proposes a method for fabricating an immediate digital implantation guide plate, which combines CBCT image anchor point recognition and coordinate transfer matching to improve implantation accuracy and has the advantages of being simple, fast, economical and practical.

[0006] To achieve the above objectives, the present invention provides a method for manufacturing an immediate digital implantation guide plate, comprising the following steps:

[0007] S1: Place the prepared oral impression material into the dental impression tray with positioning anchors, put it into the patient's mouth, instruct the patient to take a centric bite, and after the impression has solidified, take the patient to take a CBCT scan. After the CBCT image scan is completed, remove the dental impression tray.

[0008] S2: Identify the positioning anchor point information on the CBCT image and register it with the positioning anchor point coordinates of the dental model tray. Based on the missing tooth location, place the simulated implant simulated by the software at the implantation location. The spatial coordinate information of the simulated implant is associated with the coordinate information of the positioning anchor point on the CBCT image to obtain the spatial coordinate position information.

[0009] S3: After pouring quick-setting plaster into the dental mold tray, place it into the mold body. Apply plaster to the top cover of the mold as well. Then, fasten the top cover of the mold onto the mold body. After the plaster has solidified, demold the plaster positive mold. At this time, the top cover of the mold, the dental mold tray, and the plaster positive mold are integrated into one piece.

[0010] S4: After flipping the top cover of the mold with the plaster positive mold, fix it on the guide ring placement device. The positioning anchor point information on the dental mold tray is associated with each axis of the guide ring placement device. Calculate the spatial coordinate information of the simulated implant to be expressed, which is equivalent to the movement parameters of each axis of the guide ring placement device. Operate the guide ring placement device to move the guide pin and the plaster positive mold to the required position and angle respectively.

[0011] S5: After assembling the guide ring and guide pin, apply semi-fluid photosensitive adhesive to the bottom cap of the guide ring. Press down the guide pin to contact the surface of the plaster positive mold, then apply the adhesive under light. After the adhesive solidifies, remove the guide pin. At this point, the guide ring is fixed to the plaster positive mold at a predetermined angle. Make a film on the surface of the plaster positive mold. After molding, trim the film containing the guide ring, remove the top and bottom caps of the guide ring, grind and polish, and sterilize for later use.

[0012] Furthermore, in step S2, the process of identifying the positioning anchor point information on the CBCT image and registering it with the positioning anchor point coordinates of the dental mold tray also includes the following steps:

[0013] S201: Data acquisition and annotation. CBCT images were acquired by placing a dental model tray with positioning anchors in the mouths of multiple patients as training data. The positioning anchors in the CBCT images were labeled to construct the training set for the deep learning model.

[0014] S202: Annotation optimization: Based on morphological algorithms and geometric calculations, the center point location information of each positioning anchor point region is extracted, and a circular region is automatically generated on the center coordinates as the detection mark of the positioning anchor point;

[0015] S203: Training data preprocessing, based on digital image processing algorithms, performs data augmentation and normalization on each layer of two-dimensional slices of CBCT images in the training set to improve the diversity and quantity of training data, thereby improving the generalization performance of deep learning models;

[0016] S204: Deep network model construction and training. Based on the PyTorch deep learning platform, a multi-object detection convolutional neural network model is implemented. The input is a CBCT image and the corresponding localization anchor detection label. The model loss function is minimized by using the built-in Adam optimization algorithm of PyTorch to obtain the optimal deep network model parameters and the trained CBCT image localization anchor detection.

[0017] S205: Optimization of anchor point detection results. Based on the trained CBCT image anchor point detection model and isosurface extraction algorithm, all CBCT image training data are processed to obtain the 3D segmentation results of anchor points for each subject. The anchor point detection results are then optimized based on the post-processing algorithm and saved as ".stl" format. The spatial information of each anchor point is further calculated to reduce the false positive rate of the detection results.

[0018] S206: Based on PCA coarse registration, CBCT positioning anchor points are transferred to the dental mold spatial coordinate system. The 3D coordinate information of the positioning anchor points in the dental mold tray is extracted using a 3D mesh processing algorithm. Principal component analysis is then used to calculate the mapping vector between the anchor points and the CBCT image positioning anchor points from step S205, achieving preliminary registration from the CBCT image spatial coordinate system to the dental mold tray spatial coordinate system.

[0019] S207: Based on ICP fine registration of CBCT positioning anchor points to the mold space coordinate system, the positioning anchor points of the coarsely registered CBCT image and the positioning anchor points of the dental mold tray are re-registered based on the iterative nearest neighbor algorithm, and the registration results and parameters are saved. Finally, the three-dimensional surface of the virtual implant is registered to the dental mold tray space.

[0020] Further, in step S204, the multi-object detection convolutional neural network model built based on the PyTorch deep learning platform includes:

[0021] The feature learning network module is constructed using ResNet and includes 28 convolutional layers with 3×3 kernels, 3 pooling layers, 3 dropout layers, and 4 batch normalization layers.

[0022] The feature fusion network module is constructed using a feature pyramid network (FPN), which includes six convolutional layers with 3×3 kernels, three upsampling layers, three channel connection layers, two dropout layers with a dropout rate of 0.6, and one spatial attention mechanism module.

[0023] The pixel classifier consists of a convolutional layer with a 3×3 kernel and a flexible maximum function.

[0024] Further, in step S204, the construction process of the multi-object detection convolutional neural network model includes: designing the loss function of the multi-object detection convolutional neural network, then inputting the CBCT image and the corresponding localization anchor point markers, and using PyTorch's built-in Adam optimization algorithm to minimize the loss function to obtain the optimal model parameters. The loss function is shown below:

[0025]

[0026] In the formula, P L (y = G(x)|x, W) is the probability predicted by the deep network model that pixel x belongs to the anchor point category, G(x) is the true label of the anchor point region of pixel x, with a value of 0 or 1, I is the slice image data of each layer in the CBCT image training data, λ is the weight value of the second loss importance, N is the number of pixels in the image, Ω is the spatial coordinate domain of the image, the learning rate is set to 0.00001, the number of iterations is 500, and the number of training images in the batch is 20.

[0027] Further, in step S205, the trained multi-object detection convolutional neural network model is first used to process all CBCT image training data to obtain the localization anchor point classification probability map P for each data point. L (x), set the threshold to T L Using logical operation P L (x)>T L The probability map is converted into a binary result, and then a morphological processing algorithm is used to extract the three-dimensional connected regions in the segmentation result. If the number of connected regions is greater than 8, the spatial position information of each connected region is calculated to obtain the average coordinates of all anchor points. The distance between each candidate anchor point and the center point is calculated, and the 8 results with the closest distance are retained as the localization anchor point detection results. Then, the surface information of the three-dimensional localization anchor point detection results is extracted using the isosurface extraction algorithm, and the grid data is generated and saved in ".stl" format.

[0028] Further, in step S506, the process of calculating the transfer matrix and distance deviation between the positioning anchor points on the CBCT image and the positioning anchor points on the dental mold tray using the PCA algorithm includes:

[0029] First, the PCA algorithm is used to calculate the point cloud V of the anchor point surface on the CBCT image. B ={[x1, y1, z1], ..., [x m y m , z m The mean v of ]} C =[x c y c , z c The point cloud data is processed to remove the mean, resulting in... Then, the covariance matrix is ​​calculated using the Singular Value Decomposition (SVD) method. Given the eigenvalues ​​and eigenvectors, sort the eigenvalues ​​in descending order and rearrange the eigenvectors in descending order to obtain the eigenvector matrix P. B ;

[0030] The PCA algorithm was used again to calculate the point cloud V of the positioning anchor point surface on the dental mold tray. M ={[x1, y1, z1], ..., [x n y n , z n The mean v of ]} mc =[x mc y mc , z mc ], and the eigenvector matrix P M ;

[0031] Based on the above parameters, coarse registration is achieved between the positioning anchor points on the CBCT image and the positioning anchor points on the dental mold tray, as shown in the following formula:

[0032]

[0033] v t =v m -v m R C

[0034] V BT =V B R C +v T

[0035] Among them, R C Let v be the transition matrix. t For the displacement offset term, V BT Point cloud results of coarse registration of anchor points in CBCT images to the spatial coordinate system of the dental model tray.

[0036] The present invention also proposes a coordinate transformation mold for manufacturing a digital implantation guide plate, comprising a mold body and a mold top cover, wherein a dental mold tray is provided between the mold body and the mold top cover, and a number of positioning anchor points are provided on the dental mold tray.

[0037] The present invention also proposes a guide ring placement device for fabricating a digital implant guide plate. The guide ring placement device is equipped with a guide pin, which is used to place the guide ring onto the surface of the plaster positive mold on the dental mold tray. The guide ring placement device has two linear axes and two rotation axes, and the guide pin has one linear axis.

[0038] Furthermore, the guide ring includes a guide ring body, and the two ends of the guide ring body are respectively provided with a top cover and a bottom cover.

[0039] The beneficial effects of this invention are:

[0040] 1. By introducing prior image information, the accuracy and precision of the algorithm in identifying the location of anchor points are greatly improved, and the requirements of the model on the accuracy of data annotation are reduced. In addition, based on the isosurface extraction algorithm and the iterative nearest neighbor algorithm, the three-dimensional coordinate information of the anchor points in the CBCT image is generated and automatically registered with the anchor point coordinates of the dental mold tray as a standard mold. Experiments show that the algorithm has the advantages of high efficiency, accuracy and robustness, and can achieve good detection and registration accuracy in noisy and complex CBCT images.

[0041] 2. The residual network model and the feature pyramid network model are fused to extract the image features of the CBCT image. A pixel-wise classifier is used to efficiently identify the location information of the anchor points in the CBCT image. Furthermore, the PCA and ICP algorithms are used to register the CBCT image anchor point detection results to the origin coordinate system of the dental mold tray, which facilitates the subsequent fabrication of the implant guide.

[0042] In summary, the method of this invention combines CBCT image anchor point recognition and coordinate transfer matching, improving the accuracy of guide ring setting for implant slabs, thereby enhancing implant placement accuracy. It also boasts advantages such as simplicity, speed, and cost-effectiveness. Using this method, the guide plate fabrication process for simple cases can be shortened from several hours to several days using traditional 3D-printed guide plates to 10-15 minutes. Immediate implantation can be completed according to the case design plan. Furthermore, because the spatially matching anchor points are not based on tooth shape registration, it is particularly convenient and quick for edentulous cases and cases where metal crowns interfere with the abutment teeth. Attached Figure Description

[0043] The invention will now be further described and explained with reference to the accompanying drawings.

[0044] Figure 1 This is a flowchart of the detection of positioning anchor points on CBCT images and their registration with positioning anchor points on the dental mold tray.

[0045] Figure 2 This is an architecture diagram of a network model for identifying anchor points on CBCT images.

[0046] Figure 3 This is a schematic diagram of CBCT image localization anchor point annotation, detection results, and 3D reconstruction.

[0047] Figure 4 This is a schematic diagram of the coordinate transformation mold.

[0048] Figure 5 This is a schematic diagram of the guide ring placement device.

[0049] Figure 6 This is a schematic diagram of the guide ring structure.

[0050] Reference numerals: 11. Mold body; 12. Mold top cover; 13. Tooth mold tray; 14. Positioning anchor point; 15. Conversion interface; 21. Base; 22. Longitudinal beam; 23. Support plate; 24. Y-axis electric slide; 25. X-axis electric slide; 26. A-axis rotary stepper motor; 261. Working platform; 27. B-axis rotary stepper motor; 271. Reversing plate; 272. Guide pin; 31. Guide ring body; 32. Guide ring top closed; 33. Guide ring bottom closed. Detailed Implementation

[0051] The technical solution of the present invention will be more clearly and completely explained below with reference to the accompanying drawings and through the description of preferred embodiments of the present invention.

[0052] Example 1: A method for manufacturing an immediate digital implantation guide plate according to a preferred embodiment of the present invention includes the following steps:

[0053] S1: Place the prepared oral impression material into the dental impression tray with positioning anchors, put it into the patient's mouth, instruct the patient to perform centric occlusion, and after the impression solidifies, take a CBCT scan of the patient. After the CBCT image is scanned, remove the dental impression tray. The positioning anchors are high-density radiopaque anchors. In this embodiment, metal anchors are used as positioning anchors.

[0054] S2: The information collected at this time includes the patient's intraoral (occlusal) bone data and the positioning anchor point data on the dental model tray. The positioning anchor point information on the CBCT image is identified and registered with the positioning anchor point coordinates on the dental model tray. Based on the missing tooth location, the doctor places the simulated implant in the correct position according to the conditions on the CBCT image. The spatial coordinate information of the simulated implant is associated with the coordinate information of the positioning anchor point on the CBCT image to obtain the relevant spatial coordinate position information.

[0055] like Figure 1As shown, the process of identifying the positioning anchor point information on the CBCT image and registering it with the positioning anchor point coordinates of the dental mold tray specifically includes the following steps:

[0056] S201: Data Acquisition and Labeling. Fifty CBCT images containing dental mold trays with positioning anchors were collected from the mouths of multiple patients as training data. Professionals with knowledge of oral medicine used medical image labeling software to label the positioning anchors in the CBCT images to construct the training set for the deep learning model. The labeling rules are as follows: Each CBCT image has a total of 8 metal anchors, which are labeled layer by layer from the mandible to the maxilla. Each metal anchor is labeled with a different label ID number (1-8) to distinguish different metal anchors from each other.

[0057] S202: Annotation optimization. Based on morphological algorithms and geometric calculations, the center point position information of each positioning anchor point region is extracted, and a circular region with a radius of 3 pixels is automatically generated on the center coordinate as the detection mark of the positioning anchor point. In order to improve the accuracy of annotation, the central moment of the annotation result of each positioning anchor point is calculated, the center point position information of the positioning anchor point is extracted, and a circular region with a radius of 3 pixels is generated at the center position as the detection mark of the metal anchor point. Finally, it is saved as a binary image.

[0058] S203: Training data preprocessing. Based on digital image processing algorithms, data augmentation and normalization are performed on each layer of two-dimensional slices of CBCT images in the training set to increase the diversity and quantity of training data, thereby improving the generalization performance of the deep learning model. Image enhancement algorithms such as image flipping (vertical flipping), angle rotation (clockwise 90 degrees, 180 degrees, and 270 degrees), and random cropping (cropped image size of 320×320) are used to process CBCT images and corresponding detection markers, increasing the number of images by 32 times and improving data diversity. Then, the data is subjected to max-min normalization and mean removal processing to reduce the pixel distribution differences of the data.

[0059] S204: Deep network model construction and training. Based on the PyTorch deep learning platform, a multi-object detection convolutional neural network model is implemented. The input is a CBCT image and the corresponding localization anchor detection label. The model loss function is minimized by using the built-in Adam optimization algorithm of PyTorch to obtain the optimal deep network model parameters and obtain the trained CBCT image localization anchor detection.

[0060] like Figure 2 As shown, the multi-object detection convolutional neural network model built based on the PyTorch deep learning platform includes:

[0061] The feature learning network module is constructed using ResNet and includes 28 convolutional layers with 3×3 kernels, 3 pooling layers, 3 dropout layers, and 4 batch normalization layers.

[0062] The feature fusion network module is constructed using a feature pyramid network (FPN), which includes six convolutional layers with 3×3 kernels, three upsampling layers, three channel connection layers, two dropout layers with a dropout rate of 0.6, and one spatial attention mechanism module.

[0063] The pixel classifier consists of a convolutional layer with a 3×3 kernel and a flexible maximum function.

[0064] The construction process of a multi-object detection convolutional neural network model includes: designing the loss function of the multi-object detection convolutional neural network, then inputting CBCT images and corresponding localization anchor point labels, and using PyTorch's built-in Adam optimization algorithm to minimize the loss function to obtain the optimal model parameters. The loss function is shown below:

[0065]

[0066] In the formula, P L (y = G(x)|x, W) is the probability predicted by the deep network model that pixel x belongs to the anchor point category, G(x) is the true label of the anchor point region of pixel x, with a value of 0 or 1, I is the slice image data of each layer in the CBCT image training data, λ is the weight value of the second loss importance, N is the number of pixels in the image, Ω is the spatial coordinate domain of the image, the learning rate is set to 0.00001, the number of iterations is 500, and the number of training images in the batch is 20.

[0067] S205: Optimization of anchor point detection results. First, a trained multi-object detection convolutional neural network model is used to process all CBCT image training data to obtain the anchor point classification probability map P for each data point. L (x), set the threshold to T L Using logical operation P L (x)>T L The probability map is converted into a binary result, and then a morphological processing algorithm is used to extract the three-dimensional connected regions in the segmentation result. If the number of connected regions is greater than 8, the spatial position information of each connected region is calculated to obtain the average coordinates of all anchor points. The distance between each candidate anchor point and the center point is calculated, and the 8 results with the closest distance are retained as the localization anchor point detection results. Then, the surface information of the three-dimensional localization anchor point detection results is extracted using the isosurface extraction (marching cubes) algorithm, and the grid data is generated and saved in ".stl" format. The spatial information of each localization anchor point is further calculated to reduce the false positive rate of the detection results.

[0068] S206: As Figure 3 As shown, the positioning anchor points of the CBCT image and the positioning anchor points on the dental mold tray have a corresponding spatial position relationship. However, there are directional differences and relative displacements in the coordinate systems of the two. The PCA algorithm can be used to calculate the transfer matrix and distance deviation between the two. Based on the PCA coarse registration, the CBCT positioning anchor points are transferred to the mold spatial coordinate system. The three-dimensional coordinate information of the positioning anchor points in the dental mold tray is extracted based on the three-dimensional mesh processing algorithm. The mapping vector between the anchor points and the positioning anchor points of the CBCT image in step S205 is calculated by the principal component analysis algorithm, so as to achieve the initial registration from the spatial coordinate system of the CBCT image to the spatial coordinate system of the dental mold tray.

[0069] The process of calculating the transfer matrix and distance deviation between the positioning anchor points on the CBCT image and the positioning anchor points on the dental mold tray using the PCA algorithm includes:

[0070] First, the PCA algorithm is used to calculate the point cloud V of the anchor point surface on the CBCT image. B ={[x1, y1, z1], ..., [x m y m , z m The mean v of ]} C =[x c y c , z c The point cloud data is processed to remove the mean, resulting in... Then, the covariance matrix is ​​calculated using the Singular Value Decomposition (SVD) method. Given the eigenvalues ​​and eigenvectors, sort the eigenvalues ​​in descending order and rearrange the eigenvectors in descending order to obtain the eigenvector matrix P. B ;

[0071] The PCA algorithm was used again to calculate the point cloud V of the positioning anchor point surface on the dental mold tray. M ={[x1, y1, z1], ..., [x n y n , z n The mean v of ]} mc =[x mc y mc , z mc ], and the eigenvector matrix P M ;

[0072] Based on the above parameters, coarse registration is achieved between the positioning anchor points on the CBCT image and the positioning anchor points on the dental mold tray, as shown in the following formula:

[0073]

[0074] v t =v m -v m R C

[0075] V BT =V B R C +v T

[0076] Among them, R C Let v be the transition matrix. t For the displacement offset term, V BT Point cloud results of coarse registration of anchor points in CBCT images to the spatial coordinate system of the dental model tray.

[0077] S207: Based on ICP fine registration, CBCT anchor points are located to the mold space coordinate system. Although the PCA algorithm can efficiently register CBCT metal anchor points to the mold space coordinate system, it is easily affected by noise, which reduces its registration accuracy. A local registration method is needed to optimize its accuracy. The iterative nearest neighbor algorithm is used to locate the anchor point cloud V of the coarsely registered CBCT image. BT And dental mold tray positioning anchor point cloud V M A second registration was performed, with the number of iterations set to 500. The algorithm was also prohibited from performing any size changes or flipping operations on the CBCT crown point cloud, resulting in the point cloud transition matrix M. B Therefore, based on the transition matrix M B After obtaining the precise registration of the CBCT image positioning anchor points, the registration results and parameters are saved. Finally, the three-dimensional surface of the virtual implant is registered to the dental model tray space.

[0078] S3: After pouring quick-setting plaster into the dental mold tray, place it into the mold body. Apply plaster to the top cover of the mold as well, and then fasten the top cover of the mold onto the mold body. After the plaster has solidified, demold the plaster positive mold. At this time, the top cover of the mold, the dental mold tray, and the plaster positive mold are integrated. The positioning anchor point information on the dental mold tray is consistent with the positioning anchor point mapping point in the CBCT image. Thus, the coordinate information of the simulated implant on the CBCT image is associated with the positioning anchor point on the dental mold tray. Since the positions of the top cover of the mold and the dental mold tray are relatively fixed, the coordinate information of the positioning anchor point on the dental mold tray can be associated with the top cover of the mold, and finally the association between the top cover of the mold and the coordinates of the virtual implant is established.

[0079] S4: After flipping the mold top cover with the plaster positive mold, fix it on the guide ring placement device. The reference planes of each axis of the guide ring placement device and the reference plane of the mold top cover are fixed parameters. Therefore, the positioning anchor point information on the mold top cover can be associated with each axis of the guide ring placement device. Calculate the spatial coordinate information of the simulated implant to be expressed, which is equivalent to the movement parameters of each axis of the guide ring placement device. Then operate the guide ring placement device to move the guide pin and the plaster positive mold to the required position and angle respectively.

[0080] S5: After assembling the guide ring and guide pin, apply semi-fluid photosensitive adhesive to the bottom cap of the guide ring. Press the guide pin down at the calculated angle until it contacts the surface of the plaster mold, then apply the adhesive under light. After the adhesive has solidified, remove the guide pin. At this point, the guide ring is fixed to the plaster mold at the predetermined angle. Referring to the orthodontic retainer procedure, use a guide ring diaphragm for thermoplastic negative pressure molding, or cover it with thermoplastic wax, or accumulate self-curing material to create a diaphragm on the surface of the plaster mold. After molding, trim the diaphragm containing the guide ring, remove the top and bottom caps of the guide ring, grind and polish, and sterilize for later use.

[0081] Example 2: A coordinate transformation mold for a method of manufacturing an instant digital planting guide plate, such as... Figure 4 As shown, the mold includes a mold body 11 and a mold top cover 12. The top surface of the mold body 11 is provided with several connecting posts, and the bottom surface of the mold top cover 12 is provided with several countersunk holes that mate with the connecting posts. A toothed mold tray 13 is provided between the mold body 11 and the mold top cover 12, and the toothed mold tray 13 is provided with eight positioning anchor points 14. The top surface of the mold top cover 12 is provided with a conversion interface 15 for fixing it to a guide ring placement device.

[0082] Example 3: A guide ring placement device for a method of manufacturing an immediate digital implantation guide plate, such as... Figure 5 As shown, the system includes a base 21, on which a pair of longitudinal beams 22 are mounted, and a support plate 23 is mounted on the top of the longitudinal beams 22. A Y-axis electric slide 24 is mounted on the base 21, and an X-axis electric slide 25 is mounted on the slider of the Y-axis electric slide 24. The sliding directions of the sliders of the Y-axis electric slide 24 and the X-axis electric slide 25 are perpendicular to each other in the horizontal plane. An A-axis rotary stepper motor 26 is mounted on the X-axis electric slide 25, and a work platform 261 is mounted on the A-axis rotary stepper motor 26. The work platform 261 rotates with the rotation axis of the A-axis rotary stepper motor 26, and has a mounting slot for a matching conversion interface 15. A B-axis rotary stepper motor 27 is mounted on the support plate 23, and a reversing plate 271 rotates with its rotation axis. A guide pin 272 is mounted on the reversing plate 271, and the guide pin 272 is pressable. The guide ring is positioned by pressing the guide pin 272 down onto the surface of the plaster male mold on the working platform 261.

[0083] like Figure 6As shown, the guide ring includes a guide ring body 31, and the two ends of the guide ring body 31 are respectively provided with a top cover and a bottom cover.

[0084] The above-described specific embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Various modifications, substitutions, and improvements made by those skilled in the art to the technical solutions of the present invention based on the provided textual description and drawings, without departing from the design concept and spirit of the present invention, should all fall within the scope of protection of the present invention. The scope of protection of the present invention is determined by the claims.

Claims

1. A method of fabricating an immediate digitalization implant guide, characterized by, Includes the following steps: S1: Place the prepared oral impression material into the dental impression tray with positioning anchors, put it into the patient's mouth, instruct the patient to take a centric bite, and after the impression has solidified, take the patient to take a CBCT scan. After the CBCT image scan is completed, remove the dental impression tray. S2: Identify the positioning anchor point information on the CBCT image and register it with the positioning anchor point coordinates of the dental model tray. Based on the location of the missing tooth, place the simulated implant simulated by the software at the implantation position. The spatial coordinate information of the simulated implant is correlated with the coordinate information of the positioning anchor points on the CBCT image to obtain the spatial coordinate position information, including the following steps: S201: Data acquisition and annotation. CBCT images were acquired by placing a dental model tray with positioning anchors in the mouths of multiple patients as training data. The positioning anchors in the CBCT images were labeled to construct the training set for the deep learning model. S202: Annotation optimization: Based on morphological algorithms and geometric calculations, the center point location information of each positioning anchor point region is extracted, and a circular region is automatically generated on the center coordinates as the detection mark of the positioning anchor point; S203: Training data preprocessing, based on digital image processing algorithms, performs data augmentation and normalization on each layer of two-dimensional slices of CBCT images in the training set to improve the diversity and quantity of training data, thereby improving the generalization performance of deep learning models; S204: Deep network model construction and training. Based on the PyTorch deep learning platform, a multi-object detection convolutional neural network model is implemented. The input is a CBCT image and the corresponding localization anchor detection label. The model loss function is minimized by using the built-in Adam optimization algorithm of PyTorch to obtain the optimal deep network model parameters and the trained CBCT image localization anchor detection. S205: Optimization of anchor point detection results. Based on the trained CBCT image anchor point detection model and isosurface extraction algorithm, all CBCT image training data are processed to obtain the 3D segmentation results of anchor points for each subject. The anchor point detection results are then optimized based on the post-processing algorithm and saved in ".stl" format. The spatial information of each anchor point is further calculated to reduce the false positive rate of the detection results. S206: Based on PCA coarse registration, CBCT positioning anchor points are transferred to the dental mold spatial coordinate system. The 3D coordinate information of the positioning anchor points in the dental mold tray is extracted using a 3D mesh processing algorithm. Principal component analysis is then used to calculate the mapping vector between the anchor points and the CBCT image positioning anchor points from step S205, achieving preliminary registration from the CBCT image spatial coordinate system to the dental mold tray spatial coordinate system. S207: Based on ICP fine registration of CBCT positioning anchor points to the mold space coordinate system, the positioning anchor points of the coarsely registered CBCT image and the positioning anchor points of the dental mold tray are re-registered based on the iterative nearest neighbor algorithm, and the registration results and parameters are saved. Finally, the three-dimensional surface of the virtual implant is registered to the dental mold tray space. S3: After pouring quick-setting plaster into the dental mold tray, place it into the mold body. Apply plaster to the top cover of the mold as well. Then, fasten the top cover of the mold onto the mold body. After the plaster has solidified, demold the plaster positive mold. At this time, the top cover of the mold, the dental mold tray, and the plaster positive mold are integrated into one piece. S4: After flipping the top cover of the mold with the plaster positive mold, fix it on the guide ring placement device. The positioning anchor point information on the dental mold tray is associated with each axis of the guide ring placement device. Calculate the spatial coordinate information of the simulated implant to be expressed, which is equivalent to the movement parameters of each axis of the guide ring placement device. Operate the guide ring placement device to move the guide pin and the plaster positive mold to the required position and angle respectively. S5: After assembling the guide ring and guide pin, apply semi-fluid photosensitive adhesive to the bottom cap of the guide ring. Press down the guide pin to contact the surface of the plaster positive mold, then apply the adhesive under light. After the adhesive solidifies, remove the guide pin. At this point, the guide ring is fixed to the plaster positive mold at a predetermined angle. Make a film on the surface of the plaster positive mold. After molding, trim the film containing the guide ring, remove the top and bottom caps of the guide ring, grind and polish, and sterilize for later use.

2. The method of claim 1, wherein, In step S204, the multi-object detection convolutional neural network model built based on the PyTorch deep learning platform includes: The feature learning network module is constructed using ResNet and includes 28 convolutional layers with 3×3 kernels, 3 pooling layers, 3 dropout layers, and 4 batch normalization layers. The feature fusion network module is constructed using a feature pyramid network (FPN), which includes six convolutional layers with 3×3 kernels, three upsampling layers, three channel connection layers, two dropout layers with a dropout rate of 0.6, and one spatial attention mechanism module. The pixel classifier consists of a convolutional layer with a 3×3 kernel and a flexible maximum function.

3. The method of claim 2, wherein, In step S204, the construction process of the multi-object detection convolutional neural network model includes: designing the loss function of the multi-object detection convolutional neural network, then inputting the CBCT image and the corresponding localization anchor point markers, and using PyTorch's built-in Adam optimization algorithm to minimize the loss function to obtain the optimal model parameters. The loss function is shown below: ； In the formula, These are pixels predicted by the deep network model. The probability of belonging to the anchor point category. It is a pixel The actual label of the positioning anchor point area, with a value of 0 or 1. For the training data, each slice image from the CBCT images is used. This is the weight value for the importance of the second loss. It is the number of pixels in the image. The image's spatial coordinates are used, the learning rate is set to 0.00001, the number of iterations is 500, and the number of training images in a batch is 20.

4. The method for manufacturing an immediate digital planting guide plate according to claim 1, characterized in that, In step S205, the trained multi-object detection convolutional neural network model is first used to process all CBCT image training data to obtain the localization anchor point classification probability map for each data point. Set the threshold to Using logical operations The probability map is converted into a binary result, and then a morphological processing algorithm is used to extract the three-dimensional connected regions in the segmentation result. If the number of connected regions is greater than 8, the spatial position information of each connected region is calculated to obtain the average coordinates of all anchor points. The distance between each candidate anchor point and the center point is calculated, and the 8 results with the closest distance are retained as the localization anchor point detection results. Then, the surface information of the three-dimensional localization anchor point detection results is extracted using the isosurface extraction algorithm, and the mesh data is generated and saved in ".stl" format.

5. The method for manufacturing an immediate digital implantation guide plate according to claim 1, characterized in that, In step S206, the process of calculating the transfer matrix and distance deviation between the positioning anchor points on the CBCT image and the positioning anchor points on the dental mold tray using the PCA algorithm includes: First, the PCA algorithm is used to calculate the point cloud of the anchor points on the CBCT image. mean The point cloud data is processed to remove the mean, resulting in... Then, the covariance matrix is ​​calculated using the Singular Value Decomposition (SVD) method. The eigenvalues ​​and eigenvectors are obtained by sorting the eigenvalues ​​in descending order and rearranging the eigenvectors in descending order to obtain the eigenvector matrix. ; The PCA algorithm was used again to calculate the point cloud of the positioning anchor points on the dental mold tray. mean and eigenvector matrix ; Based on the above parameters, coarse registration is achieved between the positioning anchor points on the CBCT image and the positioning anchor points on the dental mold tray, as shown in the following formula: ； ； ； in, Let be the transition matrix. For displacement offset term, Point cloud results of coarse registration of anchor points in CBCT images to the spatial coordinate system of the dental model tray.

6. A coordinate transformation mold for a method of manufacturing an immediate digital implantation guide plate according to any one of claims 1-5, characterized in that, It includes a mold body and a mold top cover, and a toothed mold tray is provided between the mold body and the mold top cover. The toothed mold tray is provided with several positioning anchor points.

7. A guide ring placement device for a method of fabricating an immediate digital implantation guide plate according to any one of claims 1-5, characterized in that, The guide ring placement device is equipped with a guide pin, which is used to place the guide ring onto the surface of the plaster positive mold on the dental mold tray. The guide ring placement device has two linear axes and two rotation axes, and the guide pin has one linear axis.

8. The guide ring placement device for fabricating an immediate digital implantation guide plate according to claim 7, characterized in that, The guide ring includes a guide ring body, and the two ends of the guide ring body are respectively provided with a top cover and a bottom cover.

Images