Oral implant surgery planning method and system and storage medium

Abstract

The invention provides an oral implant surgery planning method and system and a storage medium. The method comprises the steps that an oral CT three-dimensional model is segmented through an oral model segmentation deep learning model to obtain a tooth CT three-dimensional model, a jaw bone CT three-dimensional model, a neural tube CT three-dimensional model and a blood vessel CT three-dimensional model; generating a dental arch curve based on the tooth CT three-dimensional model; missing tooth coordinate data, missing adjacent tooth coordinate data and implantation area coordinate data are obtained through analysis; analyzing to obtain the axial data of the missing adjacent teeth, and analyzing to obtain the axial data of the implant and the coordinate data of the implantation point according to the axial data of the missing adjacent teeth; performing analysis to obtain safe boundary area coordinate data of the implant; and screening in the implant database to obtain the implant type. A key anatomical structure in an oral cavity CT three-dimensional model is automatically identified and segmented through an oral cavity model segmentation deep learning model, a safety boundary area is calculated, an optimal implant scheme is screened out based on an implant library, and decision support is provided for doctors.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical equipment, and relates to a method, a system and a storage medium for planning an oral implant surgery before performing an oral treatment surgery. BACKGROUND

[0002] Oral implant surgery is one of the most ideal methods for repairing tooth loss at present, and can effectively restore the mastication function and facial aesthetics of a patient. Whether the implant surgery is successful or not depends largely on the accurate design of the implant. Accurate implant design can ensure that the implant tooth can normally perform the mastication function, bear the occlusal force and have a reasonable force distribution after the implant tooth surgery is performed, thereby avoiding excessive load and at the same time maximizing the preservation of the original tissue, promoting the combination of the bone and the implant, and ensuring that the implant tooth can exist stably in the oral cavity for a long time.

[0003] In the implant design process, the operator needs to design the implantation point, implantation depth and axial direction of the implant according to the alveolar bone, dental arch line and missing tooth of the patient. In the prior art, the operator needs to first obtain the oral CT three-dimensional model of the patient through CT scanning, and then obtains the implantation scheme by manually analyzing the oral CT three-dimensional model for empirical planning. The manual analysis of the oral CT three-dimensional model pays more attention to the state of the teeth and the jaw bone, and cannot comprehensively and multi-angulary consider the combined condition of the two, and pays less attention to the nerve canal and blood vessels. Therefore, the implant design technical solution of the prior art has problems such as low efficiency, strong subjectivity and large anatomical structure recognition error.

[0004] There is an urgent need in the industry to propose a new oral implant surgery planning scheme to improve the recognition accuracy of the teeth, jaw bone and other soft tissues in the oral cavity, so as to better improve the success rate of the implant surgery. SUMMARY

[0005] The technical problem to be solved by the present application is to provide an oral implant surgery planning method, system and storage medium to effectively improve the success rate of the implant surgery by comprehensively and accurately analyzing the oral structure of the patient through the oral model segmentation deep learning model and providing data support for implant design, in view of the above defects of the prior art.

[0006] The technical solution adopted by the present application to solve the technical problem is as follows:

[0007] An oral implant surgery planning method comprises the following steps:

[0008] S1. Construct an oral model segmentation deep learning model, and segment the oral CT three-dimensional model to obtain a tooth CT three-dimensional model, a jaw bone CT three-dimensional model, a nerve canal CT three-dimensional model and a blood vessel CT three-dimensional model according to the oral model segmentation deep learning model;

[0009] S2. generating an arch curve based on the tooth CT three-dimensional model;

[0010] S3. obtaining missing tooth coordinate data, missing adjacent tooth coordinate data and implant region coordinate data by analyzing the tooth CT three-dimensional model, the jaw CT three-dimensional model, the nerve duct CT three-dimensional model, the blood vessel CT three-dimensional model and the arch curve;

[0011] S4. obtaining the axial data of the missing adjacent tooth based on the tooth CT three-dimensional model analysis, and obtaining the axial data of the implant and the implant point coordinate data based on the axial data of the missing adjacent tooth;

[0012] S5. obtaining the safe boundary region coordinate data of the implant by combining the implant region coordinate data and the axial data of the implant;

[0013] S6. screening the implant type in the implant database according to the axial data of the implant and the safe boundary region coordinate data.

[0014] Compared with the prior art, the beneficial effects of the technical scheme are: the key anatomical structures in the oral CT three-dimensional model are automatically identified and segmented by the oral model segmentation deep learning model, the safe boundary region of the implant is calculated, and the optimal implant scheme is screened based on the implant library, thereby providing decision support for clinicians.

[0015] Further, the step S1 specifically comprises:

[0016] S101. obtaining a plurality of sets of oral CBCT basic data, and performing gray scale normalization processing on the plurality of sets of oral CBCT basic data;

[0017] S102. labeling the gray scale normalized oral CBCT basic data to obtain corresponding tooth label model, jaw label model, nerve duct label model and blood vessel label model;

[0018] S103. dividing the gray scale normalized oral CBCT basic data into a data set to obtain an oral model segmentation training set, an oral model segmentation verification set and an oral model segmentation test set;

[0019] S104. defining a combined loss function , wherein, is a total loss coefficient, is a dice loss coefficient, is a cross-entropy loss coefficient, is a boundary loss coefficient, is a loss function parameter;

[0020] S105. Based on the SA-UNet model as the network architecture, input the dental model segmentation training set of the dental CBCT basic data into the SA-UNet model, analyze the probability of each three-dimensional voxel in the dental CBCT basic data belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model, and output the highest one as the judgment result;

[0021] S106. Compare the judgment result with the tooth label model, the jawbone label model, the neural tube label model and the blood vessel label model obtained by labeling the dental CBCT basic data, calculate the total loss coefficient corresponding to the judgment result through the combination loss function, and update the loss function parameters according to the total loss coefficient;

[0022] S107. Input the dental model segmentation training set into the SA-UNet model after updating the loss function parameters, and iterate until , to obtain the dental model segmentation deep learning model;

[0023] S108. Obtain the dental CT three-dimensional model and input the dental CT three-dimensional model into the dental model segmentation deep learning model, analyze the dental CT three-dimensional model through the dental model segmentation deep learning model, calculate the probability of each three-dimensional voxel in the dental CT three-dimensional model belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model; when the probability of the three-dimensional voxel belonging to the tooth label model is the highest, it is judged as the three-dimensional voxel of the tooth CT three-dimensional model, when the probability of the three-dimensional voxel belonging to the jawbone label model is the highest, it is judged as the three-dimensional voxel of the jawbone CT three-dimensional model, when the probability of the three-dimensional voxel belonging to the neural tube label model is the highest, it is judged as the three-dimensional voxel of the neural tube CT three-dimensional model, and when the probability of the three-dimensional voxel belonging to the blood vessel label model is the highest, it is judged as the three-dimensional voxel of the blood vessel CT three-dimensional model;

[0024] S109. Each three-dimensional voxel in the dental CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model;

[0025] S110. Smooth the segmentation boundary of the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model, and remove the isolated area.

[0026] The beneficial effects of adopting the above scheme are as follows: by unifying the grayscale scale through grayscale normalization, the model generalization ability of oral CBCT basic data can be effectively improved, making it easier for operators to annotate multiple sets of oral CBCT basic data, define different label models in oral CBCT basic data, and then optimize the oral model segmentation deep learning model to an ideal state through a combination loss function. Then, the oral CT 3D model obtained by CT scan is input into the oral model segmentation deep learning model for automatic recognition, thereby obtaining the corresponding tooth CT 3D model, jawbone CT 3D model, nerve canal CT 3D model and blood vessel CT 3D model, improving the efficiency and accuracy of data analysis.

[0027] Furthermore, step S2 specifically includes:

[0028] S201. Obtain the center point coordinates of each tooth based on the 3D CT model of the teeth. Calculate the average coordinates of all teeth. ;

[0029] S202. Using the average coordinates of all teeth Using the origin as the coordinates, calculate the center point coordinates of each tooth. The coordinate distance from the origin is used to obtain the centralized data point set;

[0030] S203. Calculate the covariance matrix of the centered data point set, take the direction with the largest data variance as the dental arch width direction, and take the direction with the second largest data variance that is perpendicular to the dental arch width direction as the dental arch depth direction.

[0031] S204. Based on the width direction of the dental arch, the depth direction of the dental arch, and the coordinates of the center point of each tooth, the weighted least squares method is used to fit the dental arch curve;

[0032] S205. Extract points on the fitted dental arch curve at equal intervals as control points, generate spline curves through the control points, and smooth the fitted curves through spline interpolation to obtain the smoothed dental arch curves.

[0033] The beneficial effects of adopting the above scheme are: after finding the average coordinates of all teeth, subtracting the average coordinates from the coordinates of each tooth yields a new set of points centered on the origin of the average coordinates, eliminating the influence of positional offset. By fitting the dental arch curve using the weighted least squares method, the dental arch curve can be made to conform to the overall tooth arrangement.

[0034] Furthermore, step S3 specifically includes:

[0035] S301. Obtain a standard dentition model and identify the coordinate data of missing teeth by comparing the 3D CT model of teeth with the standard dentition model;

[0036] S302. Obtain missing adjacent tooth coordinate data according to the tooth CT three-dimensional model and the missing tooth coordinate data;

[0037] S303. Obtain the horizontal direction range of the implant according to the tooth CT three-dimensional model, the missing tooth coordinate data, and the missing adjacent tooth coordinate data;

[0038] S304. Obtain the vertical direction range of the implant according to the jawbone CT three-dimensional model;

[0039] S305. Determine the implant area range of the implant according to the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model, in combination with the horizontal direction range and the vertical direction range of the implant.

[0040] The beneficial effects of the above scheme are: the missing tooth coordinate data is obtained by comparing the tooth CT three-dimensional model and the standard dentition model, the missing adjacent tooth coordinate data is further obtained in combination with the tooth CT three-dimensional model, the implant area coordinate data is analyzed in combination with the jawbone CT three-dimensional model, the neural tube CT three-dimensional model, and the blood vessel CT three-dimensional model, and it is clear that the implant tooth operation is performed in which area range.

[0041] Further, the step S4 specifically includes:

[0042] S401. Extract the axial data of each tooth in the tooth CT three-dimensional model, analyze the distribution rule of the axial data of each tooth in the CT three-dimensional model, and determine the axial data of the implant;

[0043] S402. Determine the implant point coordinate data of the implant according to the tooth CT three-dimensional model and the jawbone CT three-dimensional model, ensure that the distance between the implant and the roots of the missing adjacent teeth is greater than 1.5 mm, and ensure that the bone thickness of the upper labial buccal side and the lingual palatal side of the implant is not less than 1.5 mm;

[0044] The step S5 specifically includes:

[0045] S501. Obtain the implant safety distance data based on the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model, the jawbone CT three-dimensional model, and the missing adjacent tooth coordinate data, ensure that the distance between the implant and any one of the neural tube and the blood vessel is ≥2 mm, ensure that the distance between the implant and the roots of the missing adjacent teeth is ≥1.5 mm, and ensure that the implant depth does not exceed 80% of the available height of the jawbone;

[0046] S502. Generate an implant cylindrical wrapper in combination with the implant point coordinate data, the axial data of the implant, and the implant safety distance data;

[0047] S503. Obtain the coordinate data of the implant cylindrical wrapper, thereby obtaining the safety boundary area coordinate data of the implant.

[0048] The beneficial effect of the above scheme is that the axial data of the implant is determined according to the axial distribution law of the teeth of the patient, the implant point coordinate data is determined according to the operation requirements of the implant tooth, the axis and the orientation of the implant are determined through the axial data and the implant point coordinate data, and the safe boundary region coordinate data of the implant is further analyzed and obtained in combination with the implant region coordinate data and the axial data of the implant.

[0049] The technical scheme adopted by the present application to solve the technical problems is as follows:

[0050] An oral implant surgery planning system comprises:

[0051] A learning model construction module is configured to construct an oral model segmentation deep learning model, and segment a dental CT three-dimensional model, a jawbone CT three-dimensional model, a neural tube CT three-dimensional model and a blood vessel CT three-dimensional model from an oral CT three-dimensional model according to the oral model segmentation deep learning model.

[0052] A dental arch curve generation module is configured to generate a dental arch curve based on the dental CT three-dimensional model.

[0053] A missing tooth positioning module is configured to analyze and obtain missing tooth coordinate data, missing adjacent tooth coordinate data and implant region coordinate data from the dental CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model and the dental arch curve.

[0054] An implant planning module is configured to analyze and obtain axial data of the missing adjacent tooth based on the dental CT three-dimensional model, and analyze and obtain axial data of the implant and implant point coordinate data from the axial data of the missing adjacent tooth.

[0055] A wrapping body design module is configured to analyze and obtain safe boundary region coordinate data of the implant in combination with the implant region coordinate data and the axial data of the implant.

[0056] An implant body screening module is configured to screen and obtain an implant model from an implant body database according to the axial data of the implant and the safe boundary region coordinate data.

[0057] Further, the learning model construction module specifically comprises:

[0058] A grayscale processing unit is configured to obtain a plurality of sets of oral CBCT basic data, and perform grayscale normalization processing on the plurality of sets of oral CBCT basic data.

[0059] A classification and labeling unit is configured to label the oral CBCT basic data subjected to the grayscale normalization processing to obtain corresponding tooth label models, jawbone label models, neural tube label models and blood vessel label models.

[0060] a data set classification unit configured to divide the gray scale normalized oral CBCT base data into a training set, a verification set and a test set for oral model segmentation;

[0061] a function definition unit configured to define a combined loss function wherein, is a total loss coefficient, is a dice loss coefficient, is a cross-entropy loss coefficient, is a boundary loss coefficient, is a loss function parameter;

[0062] a segmentation training unit configured to input the oral CBCT base data in the training set for oral model segmentation into the SA-UNet model based on the SA-UNet model as a basic network architecture, analyze the probability of each three-dimensional voxel in the oral CBCT base data belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model, and output the highest probability as a judgment result;

[0063] a parameter updating unit configured to compare the judgment result with the tooth label model, the jawbone label model, the neural tube label model and the blood vessel label model obtained by labeling the oral CBCT base data, calculate the total loss coefficient corresponding to the judgment result through the combined loss function, and update the loss function parameter according to the total loss coefficient;

[0064] a traversal training unit configured to input the training set for oral model segmentation into the SA-UNet model after updating the loss function parameter, and iteratively operate until to obtain the oral model segmentation deep learning model;

[0065] an automatic segmentation unit configured to obtain an oral CT three-dimensional model, input the oral CT three-dimensional model into the oral model segmentation deep learning model, analyze the oral CT three-dimensional model through the oral model segmentation deep learning model, and calculate the probability of each three-dimensional voxel in the oral CT three-dimensional model belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model; when the probability of the three-dimensional voxel belonging to the tooth label model is the highest, the three-dimensional voxel is determined as a tooth CT three-dimensional model; when the probability of the three-dimensional voxel belonging to the jawbone label model is the highest, the three-dimensional voxel is determined as a jawbone CT three-dimensional model; when the probability of the three-dimensional voxel belonging to the neural tube label model is the highest, the three-dimensional voxel is determined as a neural tube CT three-dimensional model; and when the probability of the three-dimensional voxel belonging to the blood vessel label model is the highest, the three-dimensional voxel is determined as a blood vessel CT three-dimensional model;

[0066] a three-dimensional reconstruction unit configured to divide each three-dimensional voxel in the oral cavity CT three-dimensional model into a tooth CT three-dimensional model, a jawbone CT three-dimensional model, a nerve duct CT three-dimensional model, and a blood vessel CT three-dimensional model;

[0067] a fine processing unit configured to perform smoothing of a partition boundary and removal of an isolated region on the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the nerve duct CT three-dimensional model, and the blood vessel CT three-dimensional model to obtain an optimized tooth CT three-dimensional model, an optimized jawbone CT three-dimensional model, an optimized nerve duct CT three-dimensional model, and an optimized blood vessel CT three-dimensional model.

[0068] Further, the dental arch curve generation module specifically comprises:

[0069] a coordinate calculation unit configured to obtain a center point coordinate of each tooth according to the tooth CT three-dimensional model , calculate an average coordinate of all teeth ;

[0070] a centering processing unit configured to take the average coordinate of all teeth as an origin, calculate a coordinate distance between the center point coordinate of each tooth and the origin, and obtain a centered data point set;

[0071] a covariance calculation unit configured to calculate a covariance matrix of the centered data point set, take a direction with a maximum data variance as a dental arch width direction, and take a direction perpendicular to the dental arch width direction and with a second largest data variance as a dental arch depth direction;

[0072] a fitting unit configured to fit a dental arch curve according to the dental arch width direction, the dental arch depth direction, and the center point coordinate of each tooth by using a weighted least square method;

[0073] a smoothing processing unit configured to extract points on the fitted dental arch curve as control points at equal intervals, generate a spline curve through the control points, and perform smoothing processing on the fitted curve through spline interpolation to obtain a smoothed dental arch curve.

[0074] Further, the missing tooth positioning module specifically comprises:

[0075] a missing tooth identification unit configured to obtain a standard dentition model, and identify missing tooth coordinate data by comparing the tooth CT three-dimensional model with the standard dentition model;

[0076] a neighboring tooth coordinate acquisition unit configured to obtain missing neighboring tooth coordinate data according to the tooth CT three-dimensional model and the missing tooth coordinate data;

[0077] a horizontal range positioning unit configured to calculate a horizontal direction range of the implant according to the tooth CT three-dimensional model, the missing tooth coordinate data, and the missing neighboring tooth coordinate data; ​​

[0078] a vertical range positioning unit configured to calculate a vertical direction range of the implant according to the CT three-dimensional model of the jawbone;

[0079] an implant region positioning unit configured to determine an implant region range of the implant according to the CT three-dimensional model of the nerve canal and the CT three-dimensional model of the blood vessel, in combination with the horizontal direction range and the vertical direction range of the implant.

[0080] Further, the implant planning module specifically comprises:

[0081] an axial data acquisition unit configured to extract axial data of each tooth in the CT three-dimensional model of the tooth, analyze the distribution rule of the axial data of each tooth in the CT three-dimensional model of the tooth, and determine axial data of the implant;

[0082] an implant point determination unit configured to determine implant point coordinate data of the implant according to the CT three-dimensional model of the tooth and the CT three-dimensional model of the jawbone, to ensure that the distance between the implant and the tooth root of the missing adjacent tooth is greater than 1.5 mm, and to ensure that the bone thickness of the upper labial buccal side and the lingual palatal side of the implant is not less than 1.5 mm;

[0083] The implant body design module specifically comprises:

[0084] a safety distance acquisition unit configured to acquire implant safety distance data based on the CT three-dimensional model of the nerve canal, the CT three-dimensional model of the blood vessel, the CT three-dimensional model of the jawbone, and the coordinate data of the missing adjacent tooth, to ensure that the distance between the implant and any one of the nerve canal and the blood vessel is greater than or equal to 2 mm, to ensure that the distance between the implant and the tooth root of the missing adjacent tooth is greater than or equal to 1.5 mm, and to ensure that the implant depth does not exceed 80% of the available height of the jawbone;

[0085] an implant body generation unit configured to generate an implant cylindrical implant body in combination with the implant point coordinate data, the axial data of the implant, and the implant safety distance data, wherein the implant cylindrical implant body meets the requirements of the implant safety distance data;

[0086] a safety boundary acquisition unit configured to acquire coordinate data of the implant cylindrical implant body, thereby obtaining safety boundary region coordinate data of the implant.

[0087] Correspondingly, a storage medium storing a computer program, the computer program comprising program instructions, when the program instructions are executed by a processor, the processor executes the oral implant surgery planning method as described above. BRIEF DESCRIPTION OF DRAWINGS

[0088] Figure 1 is a flowchart of the oral implant surgery planning method of the present application.

[0089] Figure 2is a schematic diagram of an oral implant surgery planning system of the present application.

[0090] In the drawings, the components represented by the reference numbers are listed as follows:

[0091] The learning model construction module 1, the dental arch curve generation module 2, the edentulous positioning module 3, the implant planning module 4, the wrapping body design module 5, and the implant body screening module 6. DETAILED DESCRIPTION

[0092] To make the objectives, technical solutions and advantages of the present application clearer and more apparent, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are merely used to explain the present application and should not be used to limit the present application.

[0093] In the description of the present application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", and the like indicate the orientation or positional relationship shown in the drawings, and are merely for the convenience of describing the present application and simplifying the description, and are not intended to indicate or imply that the devices or components referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore should not be construed as limiting the present application. In addition, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0094] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or detachably connected, or integrally connected; it can be mechanically connected, or electrically connected; it can be directly connected, or indirectly connected through an intermediate medium, or it can be connected inside two components. When a component is referred to as "fixed to" or "provided on" another element, it can be directly on another component or there can be a middle component. When a component is considered to be "connected" to another element, it can be directly connected to another element or there can be a middle element. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0095] Oral implant surgery is one of the most ideal methods for repairing tooth loss at present, which can effectively restore the mastication function and facial aesthetics of patients. The success of implant surgery depends largely on the precise design of the implant. Precise implant design can ensure that after the implant tooth surgery is implemented, the implant tooth can normally perform the mastication function, bear the occlusal force and the force distribution is reasonable, avoid excessive load, at the same time, maximize the preservation of the original tissue, promote the combination of bone and implant, and ensure that the implant tooth can exist stably in the oral cavity for a long time.

[0096] In the implant design process, the operator needs to design the implantation point, implantation depth and axial direction of the implant according to the patient's alveolar bone, dental arch line and missing teeth. In the prior art, the operator needs to first obtain the patient's oral CT three-dimensional model through CT scanning, and then obtain the implantation scheme by manually analyzing the oral CT three-dimensional model to obtain the implantation scheme. Manual analysis of the oral CT three-dimensional model pays more attention to the state of the teeth and the jaw, and cannot comprehensively consider the combination of the two from multiple angles, and pays less attention to the nerve canal and blood vessels. Therefore, the existing implant design technical solution has problems such as low efficiency, strong subjectivity, large anatomical structure recognition error, etc.

[0097] The industry urgently needs to propose a new oral implant surgery planning scheme to improve the recognition accuracy of teeth, jaw and other soft tissues in the oral cavity, so as to better improve the success rate of implant surgery.

[0098] As shown in Figure 1 To solve the above problems, the present application provides an oral implant surgery planning method, comprising the following steps:

[0099] S1. Construct an oral model segmentation deep learning model, and segment the oral CT three-dimensional model to obtain a tooth CT three-dimensional model, a jaw CT three-dimensional model, a nerve canal CT three-dimensional model and a blood vessel CT three-dimensional model according to the oral model segmentation deep learning model. For the process of segmenting the oral CT three-dimensional model to obtain the tooth CT three-dimensional model, the jaw CT three-dimensional model, the nerve canal CT three-dimensional model and the blood vessel CT three-dimensional model in the prior art, the operator needs to identify the bone and soft tissue morphology, and then segment according to the identification result. There are problems such as low efficiency, strong subjectivity, large anatomical structure recognition error, etc. In step S1, the oral model segmentation deep learning model is constructed, which specifically includes establishing the oral model segmentation deep learning model and training the oral model segmentation deep learning model, and obtaining a perfect oral model segmentation deep learning model after training. After obtaining the perfect oral model segmentation deep learning model, input the oral CT three-dimensional model obtained by CT scanning into the oral model segmentation deep learning model, and segment the oral CT three-dimensional model by the oral model segmentation deep learning model to obtain the tooth CT three-dimensional model, the jaw CT three-dimensional model, the nerve canal CT three-dimensional model and the blood vessel CT three-dimensional model.

[0100] S2. Generate a dental arch curve based on the tooth CT three-dimensional model. The dental arch curve describes the geometric profile formed by the arrangement of teeth on the jaw. In step S2, the position information of each tooth can be obtained from the tooth CT three-dimensional model, and the dental arch curve can be directly generated according to the position information of each tooth.

[0101] S3. Obtain missing tooth coordinate data, missing adjacent tooth coordinate data, and implant region coordinate data according to the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model, and the dental arch curve analysis. The missing tooth coordinate data is used to refer to the position of the missing tooth. The missing adjacent tooth coordinate data is used to refer to the position of the adjacent tooth of the missing tooth. A missing tooth usually has two adjacent teeth. If it is the most marginal tooth as the missing tooth, the missing tooth has only one adjacent tooth. The implant region coordinate data is used to refer to the position that can be used to arrange the implant. In step S3, on the basis of the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model, and the dental arch curve, the missing tooth coordinate data, the missing adjacent tooth coordinate data, and the implant region coordinate data can be obtained.

[0102] S4. Obtain the axial data of the missing adjacent tooth based on the tooth CT three-dimensional model analysis, and obtain the axial data of the implant and the implant point coordinate data according to the axial data of the missing adjacent tooth. In step S3, the orientation data of each tooth can be obtained according to the tooth CT three-dimensional model. In the human oral cavity, the axial direction of adjacent teeth gradually changes and has a traceable rule. Therefore, the axial data of the missing adjacent tooth can be obtained according to the tooth CT three-dimensional model. On the basis of the known axial data of the missing adjacent tooth, the implant point coordinate data can be obtained in combination with the operation requirements.

[0103] S5. In combination with the implant region coordinate data and the axial data of the implant, the safe boundary region coordinate data of the implant is obtained. The relationship between the safe boundary region coordinate data and the implant region coordinate data is that the safe boundary region coordinate data is contained in the implant region coordinate data. In the implant region coordinate data, the implant is wrapped based on the axial data of the implant to obtain the safe boundary region coordinate data.

[0104] S6. According to the axial data of the implant and the safe boundary region coordinate data, the implant model is screened in the implant database. After step S6 is completed, the implant model, the safe boundary region coordinate data, the axial data, and the implant point coordinate data can be obtained, and the design stage of the implant is completed.

[0105] Based on the above technical solution, the key anatomical structures in the oral CT three-dimensional model are automatically recognized and segmented by the oral model segmentation deep learning model, the safe boundary region of the implant is calculated, and the optimal implant scheme is screened based on the implant library, thereby providing decision support for the clinician. Compared with the prior art, the above technical solution comprehensively and accurately analyzes the patient's oral structure through the oral model segmentation deep learning model, provides data support for implant design, and provides decision support for clinicians, thereby effectively improving the success rate of implant surgery.

[0106] Preferably, step S1 specifically comprises:

[0107] S101. Obtain a plurality of sets of oral CBCT basic data, and perform grayscale normalization processing on the plurality of sets of oral CBCT basic data. In step S101, due to the large difference in the grayscale value range of the oral CBCT basic data of different devices and different patients, by using truncated grayscale normalization, such as reserving 1%-99% quantile grayscale values and mapping them to 0-255, the oral CBCT basic data is uniformly processed in the grayscale scale, which can enhance the model generalization capability.

[0108] S102. Label the grayscale normalized oral CBCT basic data to obtain corresponding tooth label model, jawbone label model, neural tube label model and blood vessel label model. In the operation process, professional labeling tools (such as 3D Slicer) can be used to label the oral CBCT basic data layer by layer, identify and label key anatomical structures such as teeth, jawbones, neural tubes, blood vessels and maxillary sinuses, and generate a three-dimensional label map. For example, if a certain tissue is identified as a tooth, the corresponding data is labeled to obtain a tooth label model.

[0109] S103. Divide the grayscale normalized oral CBCT basic data into data sets to obtain an oral model segmentation training set, an oral model segmentation validation set and an oral model segmentation test set. In all oral CBCT basic data, the oral model segmentation training set accounts for 70%, and the oral model segmentation validation set and the oral model segmentation test set each account for 15%.

[0110] S104. Define a combined loss function wherein, is a total loss coefficient, is a dice loss coefficient, is a cross-entropy loss coefficient, is a boundary loss coefficient, is a loss function parameter. For multi-class segmentation, the method of calculating the Dice loss for each class in the teeth, jawbones, neural tubes and blood vessels and then averaging is usually used, which can avoid the model being dominated by the class with a large number of voxels (such as the jawbone). A standard multi-class cross-entropy loss is used to compare the probability distribution output by the model with the one-hot encoding of the true label pixel by pixel. S105. Based on the SA-UNet model as the basic network architecture, the dental model segmentation training set of the dental CBCT basic data is input into the SA-UNet model, the probability of each three-dimensional voxel in the dental CBCT basic data belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model is analyzed, and the highest probability is output as the judgment result. In step S105, the SA-UNet model is the foundation of medical image segmentation, and the encoder-decoder structure and skip connection of the SA-UNet model can capture the adjacent information of the image at the same time to identify different dental tissues and accurately locate the different tissue and tooth boundary.

[0111] S106. The judgment result is compared with the tooth label model, the jawbone label model, the neural tube label model and the blood vessel label model obtained by labeling the dental CBCT basic data, the total loss coefficient corresponding to the judgment result is calculated by combining the loss function, and the loss function parameters are updated according to the total loss coefficient. Through parameter updating by training, the target is to reduce the total loss .

[0112] S107. The dental model segmentation training set is input into the SA-UNet model after updating the loss function parameters, and iterative operation is performed until , and the dental model segmentation deep learning model is obtained. After reaching the training target of , the training can be stopped, and the trained dental model segmentation deep learning model is obtained. At this time, automatic identification can be performed based on the dental model segmentation deep learning model.

[0113] S108. The dental CT three-dimensional model is obtained and input into the dental model segmentation deep learning model, and the dental CT three-dimensional model is analyzed by the dental model segmentation deep learning model to calculate the probability of each three-dimensional voxel in the dental CT three-dimensional model belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model. Specifically, when the probability of the three-dimensional voxel belonging to the tooth label model is the highest, it is judged as the three-dimensional voxel of the tooth CT three-dimensional model, when the probability of the three-dimensional voxel belonging to the jawbone label model is the highest, it is judged as the three-dimensional voxel of the jawbone CT three-dimensional model, when the probability of the three-dimensional voxel belonging to the neural tube label model is the highest, it is judged as the three-dimensional voxel of the neural tube CT three-dimensional model, and when the probability of the three-dimensional voxel belonging to the blood vessel label model is the highest, it is judged as the three-dimensional voxel of the blood vessel CT three-dimensional model.

[0114] S109. Form a corresponding tooth CT three-dimensional model, a jawbone CT three-dimensional model, a nerve duct CT three-dimensional model and a blood vessel CT three-dimensional model from each three-dimensional voxel in the oral cavity CT three-dimensional model. Step S109 is a three-dimensional reconstruction process, and three-dimensional reconstruction is performed according to the category corresponding to the three-dimensional voxel to obtain the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the nerve duct CT three-dimensional model and the blood vessel CT three-dimensional model.

[0115] S110. Perform smoothing of the segmentation boundary and removal of isolated regions on the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the nerve duct CT three-dimensional model and the blood vessel CT three-dimensional model to obtain an optimized tooth CT three-dimensional model, an optimized jawbone CT three-dimensional model, an optimized nerve duct CT three-dimensional model and an optimized blood vessel CT three-dimensional model. In step S110, small noise regions are removed and the segmentation boundary is smoothed through operations such as dilation and erosion; the largest connected region is retained as the target structure through removal of small isolated regions for three-dimensional connected component processing. Through the above processing, the optimized tooth CT three-dimensional model, the optimized jawbone CT three-dimensional model, the optimized nerve duct CT three-dimensional model and the optimized blood vessel CT three-dimensional model are obtained.

[0116] Based on the above technical solution, the gray scale is unified through gray scale normalization processing, which can effectively improve the model generalization ability of the oral cavity CBCT basic data, facilitate the labeling operation of the operator on multiple sets of oral cavity CBCT basic data, define different label models in the oral cavity CBCT basic data, and then optimize the oral model segmentation deep learning model to an ideal state through a combination loss function, and then input the oral CT three-dimensional model obtained through CT scanning into the oral model segmentation deep learning model for automatic recognition, so as to obtain the corresponding tooth CT three-dimensional model, jawbone CT three-dimensional model, nerve duct CT three-dimensional model and blood vessel CT three-dimensional model, and improve the efficiency and accuracy of data analysis.

[0117] Preferably, step S2 specifically comprises:

[0118] S201. Obtain the center point coordinates of each tooth according to the tooth CT three-dimensional model , and calculate the average coordinates of all teeth . The three-dimensional data corresponding to each tooth in the tooth CT three-dimensional model is known, and accordingly, the center point coordinates of each tooth can be calculated with the center of gravity of the tooth as the center point . After obtaining the center point coordinates of all teeth, the average coordinates of all teeth can be obtained by averaging, which prepares for subsequent center removal processing.

[0119] S202. Calculate the coordinate distance between the center point coordinates of each tooth and the origin with the average coordinates of all teeth as the origin to obtain a set of centralized data points. The position offset is eliminated through the centralization processing.

[0120] S203. Calculate the covariance matrix of the centralized data point set, and take the direction with the largest data variance as the arch width direction, and take the direction perpendicular to the arch width direction and with the second largest data variance as the arch depth direction. The width direction of the dental arch is the direction from the left posterior tooth to the right posterior tooth. The depth direction of the dental arch is the direction from the incisor at the front end of the dental arch to the molar at the rear end.

[0121] S204. According to the arch width direction, the arch depth direction, and the coordinates of the center points of each tooth, a dental arch curve is fitted by using the weighted least squares method. The weighted least squares method can make the generated dental arch curve not only conform to the tooth arrangement as a whole, but also have high accuracy in the aesthetic and functional key areas.

[0122] S205. The points on the fitted dental arch curve are extracted as control points at equal intervals, a spline curve is generated through the control points, and the fitted curve is smoothed by spline interpolation to obtain a smoothed dental arch curve. The dental arch curve is optimized into a smooth curve that is beautiful, natural, and consistent with the biomechanical characteristics through smoothing, and the noise and unnatural fluctuations of the fitted curve are eliminated.

[0123] Based on the above scheme, after finding the average coordinates of all teeth, the coordinates of each tooth are subtracted from the average coordinates to obtain a new point set with the average coordinate origin as the center, eliminating the influence of position deviation, and fitting the dental arch curve by using the weighted least squares method can make the dental arch curve conform to the tooth arrangement as a whole.

[0124] Preferably, step S3 specifically comprises:

[0125] S301. Obtain a standard dentition model, and identify the missing tooth coordinate data by comparing the tooth CT three-dimensional model with the standard dentition model. In an ideal state, a human body has 28-32 teeth, and the standard dentition model reflects the ideal situation of the human oral cavity, while the tooth CT three-dimensional model reflects the actual situation of the patient's oral cavity. By comparing the two, it can be known where the missing tooth is, and then the missing tooth coordinate data is obtained.

[0126] S302. Obtain the missing adjacent tooth coordinate data according to the tooth CT three-dimensional model and the missing tooth coordinate data. In the case where the missing tooth coordinate data has been determined, the missing adjacent tooth coordinate data can be further obtained.

[0127] S303. Calculate the horizontal direction range of the implant according to the tooth CT three-dimensional model, the missing tooth coordinate data, and the missing adjacent tooth coordinate data. In the tooth CT three-dimensional model, the center position of the implant can be known according to the missing tooth coordinate data, and the edge position of the implant can be known according to the missing adjacent tooth coordinate data, and the horizontal direction range of the implant can be determined according to the center position and the edge position of the implant.

[0128] S304. Calculate the vertical direction range of the implant according to the jawbone CT three-dimensional model. The alveolar bone height can be obtained from the jawbone CT three-dimensional model, and the vertical direction range of the implant is calculated accordingly.

[0129] S305. Determine the implant area range of the implant according to the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model, combined with the horizontal direction range and the vertical direction range of the implant.

[0130] Based on the above technical solution, the missing tooth coordinate data is identified by comparing the tooth CT three-dimensional model with the standard dentition model, the missing adjacent tooth coordinate data is further obtained from the tooth CT three-dimensional model, and the implant area coordinate data is analyzed from the jawbone CT three-dimensional model, the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model, so as to determine the implant area range for implanting the tooth.

[0131] Preferably, the step S4 specifically comprises:

[0132] S401. Extract the axial data of each tooth in the tooth CT three-dimensional model, analyze the distribution rule of the axial data of each tooth in the CT three-dimensional model, and determine the axial data of the implant. As described above, the axis of adjacent teeth in the human oral cavity gradually changes and has a traceable rule, so the axial data of each tooth in the tooth CT three-dimensional model is extracted, and the axial data of the implant is determined according to the distribution rule.

[0133] S402. Determine the implant point coordinate data of the implant according to the tooth CT three-dimensional model and the jawbone CT three-dimensional model, ensure that the distance between the implant and the root of the missing adjacent tooth is greater than 1.5mm, and ensure that the bone thickness of the upper lip buccal side and the tongue palatal side of the implant is not less than 1.5mm. The distance between the implant and the root of the missing adjacent tooth is greater than 1.5mm, and the bone thickness of the upper lip buccal side and the tongue palatal side of the implant is not less than 1.5mm, which are the basic requirements for implanting the tooth. Based on the determination of the axial data of the implant, the above requirements are met by selecting the characteristic position, and the implant point coordinate data of the implant is determined.

[0134] Preferably, the step S5 specifically comprises:

[0135] S501. Obtain the implant safety distance data based on the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model, the jawbone CT three-dimensional model and the missing adjacent tooth coordinate data, ensure that the distance between the implant and any one of the neural tube and the blood vessel is greater than or equal to 2mm, the distance between the implant and the root of the missing adjacent tooth is greater than or equal to 1.5mm, and the implant depth is not more than 80% of the available height of the jawbone. The above implant safety distance data is also a basic requirement for implanting the tooth.

[0136] S502. Generate a planting cylinder wrapper based on the implant point coordinate data, the axial data of the implant, and the implant safety distance data, wherein the planting cylinder wrapper meets the requirements of the implant safety distance data.

[0137] S503. Obtain coordinate data of the planting cylinder wrapper to obtain safety boundary region coordinate data of the implant. A cylindrical safety boundary region is generated as the planting cylinder wrapper based on the implant point coordinate data of the implant, wherein the diameter of the planting cylinder wrapper is determined according to the alveolar bone width and the implant size, and the height of the planting cylinder wrapper is determined according to the alveolar bone height and the safety distance requirement.

[0138] Based on the above technical solution, the axial data of the implant is determined according to the axial distribution rule of the teeth of the patient, and the implant point coordinate data is determined according to the implant tooth operation requirement. The axial center and the orientation of the implant can be determined based on the axial data and the implant point coordinate data. Further, the safety boundary region coordinate data of the implant is obtained by analyzing the implant region coordinate data and the axial data of the implant.

[0139] As shown in Figure 2 To solve the above problems, the present application provides an oral implant surgery planning system, which comprises a learning model construction module 1, a dental arch curve generation module 2, a missing tooth positioning module 3, an implant planning module 4, a wrapper design module 5, and an implant selection module 6. Specifically:

[0140] The learning model construction module 1 is used to construct an oral model segmentation deep learning model, and the oral model segmentation deep learning model is used to segment the oral CT three-dimensional model to obtain a tooth CT three-dimensional model, a jawbone CT three-dimensional model, a neural tube CT three-dimensional model, and a blood vessel CT three-dimensional model.

[0141] The dental arch curve generation module 2 is used to generate a dental arch curve based on the tooth CT three-dimensional model.

[0142] The missing tooth positioning module 3 is used to analyze the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model, and the dental arch curve to obtain missing tooth coordinate data, missing adjacent tooth coordinate data, and implant region coordinate data.

[0143] The implant planning module 4 is used to analyze the axial data of the missing adjacent tooth based on the tooth CT three-dimensional model to obtain the axial data of the implant and the implant point coordinate data.

[0144] The wrapper design module 5 is used to analyze the safety boundary region coordinate data of the implant based on the implant region coordinate data and the axial data of the implant.

[0145] The implant screening module 6 is used for screening implant models in the implant database according to the axial data of the implant and the coordinate data of the safety boundary region.

[0146] Preferably, the learning model construction module specifically comprises:

[0147] The gray processing unit is configured to acquire a plurality of sets of oral CBCT basic data and perform gray normalization processing on the plurality of sets of oral CBCT basic data.

[0148] The classification and labeling unit is configured to label the gray normalized oral CBCT basic data to obtain corresponding tooth label models, jawbone label models, neural tube label models and blood vessel label models.

[0149] The data set classification unit is configured to divide the gray normalized oral CBCT basic data into a plurality of data sets to obtain an oral model segmentation training set, an oral model segmentation verification set and an oral model segmentation test set.

[0150] The function definition unit is configured to define a combined loss function wherein, is a total loss coefficient, is a dice loss coefficient, is a cross-entropy loss coefficient, is a boundary loss coefficient, is a loss function parameter.

[0151] The segmentation training unit is configured to input the oral CBCT basic data in the oral model segmentation training set into the SA-UNet model based on the SA-UNet model as a basic network architecture, analyze the probability of each three-dimensional voxel in the oral CBCT basic data belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model, and output the highest probability as a judgment result.

[0152] The parameter updating unit is configured to compare the judgment result with the tooth label model, the jawbone label model, the neural tube label model and the blood vessel label model obtained by labeling the oral CBCT basic data, calculate the total loss coefficient corresponding to the judgment result through the combined loss function, and update the loss function parameter according to the total loss coefficient.

[0153] The traversal training unit is configured to input the oral model segmentation training set into the SA-UNet model after updating the loss function parameter, and iteratively operate until to obtain an oral model segmentation deep learning model.

[0154] An automatic segmentation unit is used to acquire a 3D oral CT model and input it into a deep learning model for oral model segmentation. The deep learning model analyzes the 3D oral CT model and calculates the probability that each 3D voxel in the 3D oral CT model belongs to a tooth label model, jawbone label model, neural tube label model, or vascular label model. When the probability of a 3D voxel belonging to the tooth label model is the highest, it is identified as a 3D voxel of the tooth CT model. When the probability of a 3D voxel belonging to the jawbone label model is the highest, it is identified as a 3D voxel of the jawbone CT model. When the probability of a 3D voxel belonging to the neural tube label model is the highest, it is identified as a 3D voxel of the neural tube CT model. When the probability of a 3D voxel belonging to the vascular label model is the highest, it is identified as a 3D voxel of the vascular CT model.

[0155] The three-dimensional reconstruction unit is used to assemble each three-dimensional voxel in the oral CT three-dimensional model into a corresponding tooth CT three-dimensional model, jawbone CT three-dimensional model, nerve tube CT three-dimensional model, and blood vessel CT three-dimensional model.

[0156] The fine processing unit is used to smooth the boundary segmentation and remove isolated regions of the CT 3D models of teeth, jawbone, nerve canal, and blood vessels, resulting in optimized CT 3D models of teeth, jawbone, nerve canal, and blood vessels.

[0157] Preferably, the dental arch curve generation module specifically includes:

[0158] The coordinate calculation unit is used to obtain the center point coordinates of each tooth based on the 3D CT model of the teeth. Calculate the average coordinates of all teeth. ;

[0159] Centralized processing unit, used to calculate the average coordinates of all teeth Using the origin as the coordinates, calculate the center point coordinates of each tooth. The coordinate distance from the origin is used to obtain the centralized data point set;

[0160] The covariance calculation unit is used to calculate the covariance matrix of the centered data point set. The direction with the largest data variance is the dental arch width direction, and the direction perpendicular to the dental arch width direction and with the second largest data variance is the dental arch depth direction.

[0161] The fitting unit is used to fit the dental arch curve using the weighted least squares method based on the width direction, depth direction, and center point coordinates of each tooth.

[0162] The smoothing unit is configured to extract points on the fitted dental arch curve as control points at equal intervals, generate a spline curve based on the control points, and smooth the fitted curve by spline interpolation to obtain a smoothed dental arch curve.

[0163] Preferably, the edentulous positioning module specifically comprises:

[0164] The edentulous recognition unit is configured to obtain a standard dentition model, and recognize missing tooth coordinate data by comparing the tooth CT three-dimensional model with the standard dentition model.

[0165] The adjacent tooth coordinate acquisition unit is configured to obtain missing adjacent tooth coordinate data based on the tooth CT three-dimensional model and the missing tooth coordinate data.

[0166] The horizontal range positioning unit is configured to calculate a horizontal direction range of the implant based on the tooth CT three-dimensional model, the missing tooth coordinate data, and the missing adjacent tooth coordinate data.

[0167] The vertical range positioning unit is configured to calculate a vertical direction range of the implant based on the jawbone CT three-dimensional model.

[0168] The implant region positioning unit is configured to determine an implant region range of the implant based on the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model, and in combination with the horizontal direction range and the vertical direction range of the implant.

[0169] The implant planning module specifically comprises:

[0170] The axial data acquisition unit is configured to extract axial data of each tooth in the tooth CT three-dimensional model, analyze the distribution rule of the axial data of each tooth in the CT three-dimensional model, and determine axial data of the implant.

[0171] The implant point determination unit is configured to determine implant point coordinate data of the implant based on the tooth CT three-dimensional model and the jawbone CT three-dimensional model, to ensure that a distance between the implant and a tooth root of the missing adjacent tooth is greater than 1.5 mm, and to ensure that a bone thickness of a labial buccal side and a lingual palatal side of the implant is not less than 1.5 mm.

[0172] Preferably, the wrapping body design module specifically comprises:

[0173] The safe distance acquisition unit is configured to obtain implant safety distance data based on the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model, the jawbone CT three-dimensional model, and the missing adjacent tooth coordinate data, to ensure that a distance between the implant and any one of the neural tube and the blood vessel is greater than or equal to 2 mm, to ensure that a distance between the implant and the tooth root of the missing adjacent tooth is greater than or equal to 1.5 mm, and to ensure that an implant depth of the implant does not exceed 80% of a usable height of the jawbone.

[0174] The implant body generating unit is configured to generate an implant cylinder envelope by combining the implant point coordinate data, the axial data of the implant body, and the implant safety distance data, wherein the implant cylinder envelope meets the requirements of the implant safety distance data.

[0175] The safety boundary obtaining unit is configured to obtain coordinate data of the implant cylinder envelope, thereby obtaining safety boundary region coordinate data of the implant body.

[0176] To solve the above problems, the application provides a storage medium, which stores a computer program, and the computer program comprises program instructions. When the program instructions are executed by a processor, the processor executes the oral implant surgery planning method.

[0177] It should be understood that the application is not limited to the above examples, and can be improved or changed according to the above description for those skilled in the art, and all these improvements and changes shall belong to the protection scope of the appended claims of the application.

Claims

1. A method of oral implant surgery planning, characterized by, The method comprises the following steps: S1. Constructing an oral model segmentation deep learning model, segmenting an oral CT three-dimensional model according to the oral model segmentation deep learning model to obtain a tooth CT three-dimensional model, a jawbone CT three-dimensional model, a neural tube CT three-dimensional model and a blood vessel CT three-dimensional model; S2. Generating a dental arch curve based on the tooth CT three-dimensional model; S3. Analyzing to obtain missing tooth coordinate data, missing adjacent tooth coordinate data and implant region coordinate data according to the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model and the dental arch curve; S4. Analyzing to obtain the axial data of the missing adjacent tooth based on the tooth CT three-dimensional model, and analyzing to obtain the axial data of the implant and the implant point coordinate data according to the axial data of the missing adjacent tooth; S5. Combining the implant region coordinate data and the axial data of the implant, and analyzing to obtain the safe boundary region coordinate data of the implant; S6. Screening the implant model in the implant database according to the axial data of the implant and the safe boundary region coordinate data.

2. A dental implant surgery planning method according to claim 1, characterized in that, The step S1 specifically comprises: S101. Obtaining a plurality of sets of oral CBCT basic data, and performing gray scale normalization processing on the plurality of sets of oral CBCT basic data; S102. Labeling the gray scale normalized oral CBCT basic data to obtain corresponding tooth label models, jawbone label models, neural tube label models and blood vessel label models; S103. Dividing the gray scale normalized oral CBCT basic data into a data set to obtain an oral model segmentation training set, an oral model segmentation verification set and an oral model segmentation test set; S104. Defining a combined loss function wherein, is a total loss coefficient, is a dice loss coefficient, is a cross-entropy loss coefficient, is a boundary loss coefficient, is a loss function parameter; S105. Taking the SA-UNet model as a basic network architecture, inputting the oral CBCT basic data in the oral model segmentation training set into the SA-UNet model, analyzing the probability of each three-dimensional voxel in the oral CBCT basic data belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model, and outputting the highest one as the judgment result; S106. Comparing the judgment result with the tooth label model, the jawbone label model, the neural tube label model and the blood vessel label model labeled by the oral CBCT basic data, calculating the total loss coefficient corresponding to the judgment result through a combination loss function, and updating the loss function parameters according to the total loss coefficient; S107. The SA-UNet model after the oral model segmentation training set is traversed to input the updated loss function parameters, and iterative operation is performed until an oral model segmentation deep learning model is obtained. S108. Obtain the oral cavity CT three-dimensional model and input the oral cavity CT three-dimensional model into the oral cavity model segmentation deep learning model, analyze the oral cavity CT three-dimensional model through the oral cavity model segmentation deep learning model, calculate the probability of each three-dimensional voxel in the oral cavity CT three-dimensional model belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model; when the probability of the three-dimensional voxel belonging to the tooth label model is the highest, it is judged as the three-dimensional voxel of the tooth CT three-dimensional model, when the probability of the three-dimensional voxel belonging to the jawbone label model is the highest, it is judged as the three-dimensional voxel of the jawbone CT three-dimensional model, when the probability of the three-dimensional voxel belonging to the neural tube label model is the highest, it is judged as the three-dimensional voxel of the neural tube CT three-dimensional model, and when the probability of the three-dimensional voxel belonging to the blood vessel label model is the highest, it is judged as the three-dimensional voxel of the blood vessel CT three-dimensional model; S109. Each three-dimensional voxel in the oral cavity CT three-dimensional model constitutes a corresponding tooth CT three-dimensional model, a jawbone CT three-dimensional model, a neural tube CT three-dimensional model and a blood vessel CT three-dimensional model; S110. The tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model are subjected to smooth segmentation boundary and isolated region removal processing to obtain optimized tooth CT three-dimensional model, jawbone CT three-dimensional model, neural tube CT three-dimensional model and blood vessel CT three-dimensional model.

3. The method according to claim 1, wherein, The step S2 specifically comprises: S201. Obtain the center point coordinates of each tooth according to the tooth CT three-dimensional model , calculate the average coordinates of all teeth ; S202. Calculate the center point coordinate of each tooth with the average coordinate of all teeth as the origin S204. Calculate the coordinate distance of each data point from the origin to obtain the centered data point set S204. Calculate the coordinate distance of each data point from the origin to obtain the centered data point set S203. Calculate the covariance matrix of the centralized data point set, and take the direction with the maximum data variance as the dental arch width direction, and take the direction perpendicular to the dental arch width direction and with the second largest data variance as the dental arch depth direction; S204. According to the dental arch width direction, the dental arch depth direction and the center point coordinates of each tooth, a dental arch curve is fitted by using a weighted least squares method; S205. The points on the fitted dental arch curve are extracted as control points at equal intervals, a spline curve is generated through the control points, and the fitted curve is smoothed by spline interpolation to obtain a smoothed dental arch curve.

4. The method of planning a dental implant surgery according to claim 1, wherein, The step S3 specifically comprises: S301. Obtain a standard dentition model, and identify missing tooth coordinate data by comparing the tooth CT three-dimensional model with the standard dentition model; S302. Obtain missing adjacent tooth coordinate data according to the tooth CT three-dimensional model and the missing tooth coordinate data; S303. Calculate the horizontal direction range of the implant according to the tooth CT three-dimensional model, the missing tooth coordinate data and the missing adjacent tooth coordinate data; S304. Calculate the vertical direction range of the implant according to the jawbone CT three-dimensional model; S305. Determine the implant area range of the implant according to the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model, in combination with the horizontal direction range and the vertical direction range of the implant.

5. The method of planning a dental implant surgery according to claim 1, wherein, The step S4 specifically comprises: S401. Extract the axial data of each tooth in the tooth CT three-dimensional model, analyze the distribution rule of the axial data of each tooth in the CT three-dimensional model, and determine the axial data of the implant; S402. determining implant point coordinate data of the implant according to the tooth CT three-dimensional model and the jawbone CT three-dimensional model, ensuring that the distance between the implant and the roots of the missing adjacent teeth is greater than 1.5 mm, and ensuring that the bone thickness on the labial and buccal side and the lingual and palatal side of the implant is not less than 1.5 mm; The step S5 specifically comprises: S501. obtaining implant safety distance data based on the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model, the jawbone CT three-dimensional model and the missing adjacent tooth coordinate data, ensuring that the distance between the implant and any one of the neural tube and the blood vessel is greater than or equal to 2 mm, ensuring that the distance between the implant and the roots of the missing adjacent teeth is greater than or equal to 1.5 mm, and ensuring that the implant depth does not exceed 80% of the available height of the jawbone; S502. generating an implant cylindrical envelope by combining the implant point coordinate data, the axial data of the implant and the implant safety distance data, wherein the implant cylindrical envelope meets the requirements of the implant safety distance data; S503. obtaining coordinate data of the implant cylindrical envelope, thereby obtaining safety boundary region coordinate data of the implant.

6. An oral implant surgery planning system, characterized by, Comprise: A learning model construction module configured to construct an oral cavity model segmentation deep learning model, and segment an oral cavity CT three-dimensional model based on the oral cavity model segmentation deep learning model to obtain a tooth CT three-dimensional model, a jawbone CT three-dimensional model, a neural tube CT three-dimensional model and a blood vessel CT three-dimensional model; A dental arch curve generation module configured to generate a dental arch curve based on the tooth CT three-dimensional model; A missing tooth positioning module configured to analyze the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model and the dental arch curve to obtain missing tooth coordinate data, missing adjacent tooth coordinate data and implant region coordinate data; An implant planning module configured to analyze the tooth CT three-dimensional model to obtain axial data of the missing adjacent teeth, and analyze the axial data of the missing adjacent teeth to obtain axial data of the implant and implant point coordinate data; An envelope design module configured to combine the implant region coordinate data and the axial data of the implant to analyze and obtain safety boundary region coordinate data of the implant; An implant body screening module configured to screen an implant model from an implant body database based on the axial data of the implant and the safety boundary region coordinate data.

7. A dental implant surgery planning system according to claim 6, wherein, The learning model construction module specifically comprises: A grayscale processing unit configured to obtain a plurality of sets of oral cavity CBCT basic data, and perform grayscale normalization processing on the plurality of sets of oral cavity CBCT basic data; A classification and labeling unit configured to label the oral cavity CBCT basic data subjected to the grayscale normalization processing to obtain corresponding tooth label models, jawbone label models, neural tube label models and blood vessel label models; A data set classification unit configured to divide the oral cavity CBCT basic data subjected to the grayscale normalization processing into a training set, a verification set and a test set for oral cavity model segmentation. A function definition unit is configured to define a combined loss function wherein, is a total loss coefficient, is a dice loss coefficient, is a cross-entropy loss coefficient, is a boundary loss coefficient, is a loss function parameter; The segmentation training unit is configured to input the dental model segmentation training set of the dental CBCT basic data into the SA-UNet model based on the SA-UNet model as a basic network architecture, analyze the probability of each three-dimensional voxel in the dental CBCT basic data belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model, and output the highest probability as a judgment result; The parameter updating unit is configured to compare the judgment result with the tooth label model, the jawbone label model, the neural tube label model and the blood vessel label model obtained by labeling the dental CBCT basic data, calculate the total loss coefficient corresponding to the judgment result by using a combination loss function, and update the loss function parameters according to the total loss coefficient; traversing the training unit to traverse the SA-UNet model after the input update loss function parameter of the dental model segmentation training set, and iteratively operating until a dental model segmentation deep learning model is obtained. The automatic segmentation unit is configured to obtain the dental CT three-dimensional model and input the dental CT three-dimensional model into the dental model segmentation deep learning model, analyze the dental CT three-dimensional model by using the dental model segmentation deep learning model, and calculate the probability of each three-dimensional voxel in the dental CT three-dimensional model belonging to the tooth label model, the jawbone label model, the neural tube label model or the blood vessel label model; when the probability of the three-dimensional voxel belonging to the tooth label model is the highest, the three-dimensional voxel is determined as a tooth CT three-dimensional model; when the probability of the three-dimensional voxel belonging to the jawbone label model is the highest, the three-dimensional voxel is determined as a jawbone CT three-dimensional model; when the probability of the three-dimensional voxel belonging to the neural tube label model is the highest, the three-dimensional voxel is determined as a neural tube CT three-dimensional model; and when the probability of the three-dimensional voxel belonging to the blood vessel label model is the highest, the three-dimensional voxel is determined as a blood vessel CT three-dimensional model; The three-dimensional reconstruction unit is configured to form the corresponding tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model from each three-dimensional voxel in the dental CT three-dimensional model; The fine processing unit is configured to perform smooth segmentation boundary and isolated region removal processing on the tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model, and obtain the optimized tooth CT three-dimensional model, the jawbone CT three-dimensional model, the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model.

8. The dental implant surgery planning system of claim 6, wherein, The dental arch curve generation module specifically comprises: A coordinate calculation unit is configured to obtain a center point coordinate of each tooth according to a tooth CT three-dimensional model , and calculate average coordinates of all teeth ; a centralizing processing unit for calculating the average coordinates of all teeth with the origin as the center point, the coordinates of each tooth are calculated the distance between the coordinates of each tooth and the origin, and the centralizing data point set is obtained The covariance calculation unit is configured to calculate the covariance matrix of the centralized data point set, take the direction with the maximum data variance as the dental arch width direction, and take the direction perpendicular to the dental arch width direction and with the second largest data variance as the dental arch depth direction; The fitting unit is configured to fit the dental arch curve by using the weighted least squares method according to the dental arch width direction, the dental arch depth direction and the center point coordinates of each tooth; The smoothing processing unit is configured to extract the points on the fitted dental arch curve as control points at equal intervals, generate a spline curve through the control points, and perform smoothing processing on the fitted curve through spline interpolation to obtain a smoothed dental arch curve.

9. The dental implant surgery planning system of claim 6, wherein, The missing tooth positioning module specifically comprises: The missing tooth identification unit is configured to obtain a standard dentition model, and identify the missing tooth coordinate data by comparing the tooth CT three-dimensional model with the standard dentition model; The adjacent tooth coordinate acquisition unit is configured to obtain the missing adjacent tooth coordinate data according to the tooth CT three-dimensional model and the missing tooth coordinate data. A horizontal range positioning unit is configured to calculate a horizontal range of the implant according to the tooth CT three-dimensional model, the missing tooth coordinate data, and the missing adjacent tooth coordinate data; A vertical range positioning unit is configured to calculate a vertical range of the implant according to the jawbone CT three-dimensional model; An implant region positioning unit is configured to determine an implant region range of the implant according to the neural tube CT three-dimensional model and the blood vessel CT three-dimensional model, in combination with the horizontal range and the vertical range of the implant; The implant planning module specifically includes: An axial data acquisition unit is configured to extract axial data of each tooth in the tooth CT three-dimensional model, analyze the distribution of the axial data of each tooth in the CT three-dimensional model, and determine axial data of the implant; An implant point determination unit is configured to determine implant point coordinate data of the implant according to the tooth CT three-dimensional model and the jawbone CT three-dimensional model, to ensure that a distance between the implant and the roots of the missing adjacent teeth is greater than 1.5 mm, and to ensure that a bone thickness of a labial and buccal side and a lingual and palatal side of the implant is not less than 1.5 mm; The implant body design module specifically includes: A safety distance acquisition unit is configured to acquire implant safety distance data based on the neural tube CT three-dimensional model, the blood vessel CT three-dimensional model, the jawbone CT three-dimensional model, and the missing adjacent tooth coordinate data, to ensure that a distance between the implant and any one of the neural tube and the blood vessel is greater than or equal to 2 mm, to ensure that a distance between the implant and the roots of the missing adjacent teeth is greater than or equal to 1.5 mm, and to ensure that an implant depth of the implant does not exceed 80% of a usable height of the jawbone; An implant body generation unit is configured to generate an implant cylindrical implant body in combination with the implant point coordinate data, the axial data of the implant, and the implant safety distance data, the implant cylindrical implant body satisfying requirements of the implant safety distance data; A safety boundary acquisition unit is configured to acquire coordinate data of the implant cylindrical implant body, to obtain safety boundary region coordinate data of the implant.

10. A storage medium, characterized by The storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by a processor, the processor performs the oral implant surgery planning method in any one of claims 1-5.

Images