Oral implant surgery real-time navigation method, system and storage medium

Abstract

The application provides an oral implant surgery real-time navigation method, system and storage medium, comprising: acquiring oral CT three-dimensional data and oral optical two-dimensional data; performing segmentation processing based on an oral segmentation data processing model; selecting a target operation area; identifying three-dimensional coordinate data of an attached identification point in a CT three-dimensional coordinate system and three-dimensional coordinate data of the attached identification point in an optical three-dimensional coordinate system; obtaining a coordinate conversion relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system through rigid body registration operation; calibrating three-dimensional coordinate data of a drill needle tip in a tracker three-dimensional coordinate system; collecting real-time image data of a mobile phone tracker and the attached identification point, and analyzing to obtain real-time relative positions of a patient's oral cavity and the drill needle tip in the CT three-dimensional coordinate system. The coordinate conversion relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system is obtained based on direct registration of the attached identification point, and real-time navigation can be performed based on the real-time image data of the mobile phone tracker and the attached identification point.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and to a real-time navigation method, system, and storage medium for dental implant surgery that provides auxiliary navigation during oral treatment procedures. Background Technology

[0002] The procedure for dental implant surgery is as follows: an implant hole is drilled in the alveolar bone at the location of the missing tooth, the implant is inserted into the implant hole, and then a fixed crown is installed on the implant to form the overall structure of the dental implant.

[0003] Drilling the implant hole is a crucial step in the entire procedure. Ideally, the operator needs to drill an implant hole of a specific orientation and depth, taking into account factors such as the patient's alveolar bone condition and the orientation of adjacent teeth. Since damage to the alveolar bone is irreparable, accurately drilling an implant hole suitable for specific oral conditions is extremely important. A binocular navigation system is used to collect data from an oral reference board and a handpiece tracker to determine the real-time position of the patient's mouth and the implant handpiece. Based on this real-time relative position, real-time navigation can be provided for the implant handpiece operation.

[0004] Current navigation methods require first fixing a registration device with ceramic balls to the tooth, then installing a dental reference plate with markers on the registration device. Real-time navigation is then achieved by establishing a connection between the ceramic balls, the dental reference plate, and the binocular navigation system via CT scans. However, this method has the following problems: First, the dental reference plate is indirectly fixed through the registration device, which can easily loosen during operation, leading to navigation and positioning errors and affecting implant accuracy. Second, the dental reference plate is large, and when it protrudes outside the mouth, surgical instruments may touch it, interfering with the operation and affecting navigation accuracy. Third, the dual marker design of the ceramic balls and the reference plate not only increases registration errors but also makes the registration process cumbersome. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a real-time navigation method, system and storage medium for oral implant surgery, which can simultaneously serve as CT markers and optical markers through attached markers, thereby simplifying the navigation registration process and improving navigation accuracy.

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

[0007] A real-time navigation method for oral implant surgery, the method comprising the following steps:

[0008] S1. Control the CT scanning equipment to perform a CT scan on the patient's oral cavity and obtain three-dimensional CT data of the oral cavity;

[0009] S2. Control the binocular navigation device to perform image scanning of the patient's oral cavity and acquire two-dimensional optical data of the oral cavity;

[0010] S3. Construct an oral cavity segmentation data processing model and train and optimize the oral cavity segmentation data processing model. Based on the trained and optimized oral cavity segmentation data processing model, segment the oral cavity CT three-dimensional data and oral cavity optical two-dimensional data. Segment the CT three-dimensional model into CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data, and segment the oral cavity optical two-dimensional data into optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

[0011] S4. Based on the segmentation processing results of the oral cavity segmentation data processing model, select the target operation area, select CT maxillary tooth 3D data or CT mandibular tooth 3D data as CT target operation data, and select optical maxillary tooth 2D data or optical mandibular tooth 2D data as optical target operation data.

[0012] S5. Identify the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system from the CT target operation data;

[0013] S6. Based on the two-dimensional data of oral optics, identify the three-dimensional coordinate data of the attached marker points in the three-dimensional optical coordinate system from the optical target operation data;

[0014] S7. Perform rigid body registration operation based on the three-dimensional coordinate data of the attached markers in the CT three-dimensional coordinate system and the three-dimensional coordinate data in the optical three-dimensional coordinate system to obtain the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0015] S8. The planting mobile phone is calibrated, the image data of the mobile phone tracker is obtained, the direction vector of the drill bit rotation axis on the planting mobile phone is analyzed and the three-dimensional coordinate data of the tracker in the three-dimensional coordinate system is obtained. Combined with the structural parameters of the drill bit, the three-dimensional coordinate data of the drill bit tip in the three-dimensional coordinate system of the tracker is calculated.

[0016] S9. Control the binocular navigator to acquire real-time image data from the mobile phone tracker and the attached markers. Based on the real-time image data from the mobile phone tracker and the attached markers, and combined with the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system, analyze the real-time relative position of the patient's oral cavity and the drill tip in the CT three-dimensional coordinate system, thereby performing real-time navigation.

[0017] Compared with existing technologies, the beneficial effects of this technical solution are as follows: After attaching the markers to the patient's teeth, CT scans and image scans are performed on the patient's oral cavity. Based on the attaching markers, the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system is directly registered. Then, real-time navigation can be performed by collecting real-time image data of the mobile phone tracker and the attaching markers through a binocular navigator.

[0018] Furthermore, step S3 specifically includes the following steps:

[0019] S301. Obtain multiple sets of basic data of oral CT models and multiple sets of basic data of oral optical models. Label the basic data of oral CT models to obtain CT maxillary tooth label models and CT mandibular tooth label models. Label the basic data of oral optical models to obtain optical maxillary tooth label models and optical mandibular tooth label models.

[0020] S302. Divide the basic data of the oral CT model into a dataset to obtain the oral CT model segmentation training set, oral CT model segmentation validation set, and oral CT model segmentation test set; divide the basic data of the oral optical model into a dataset to obtain the oral optical model segmentation training set, oral optical model segmentation validation set, and oral optical model segmentation test set.

[0021] S303. Define the combination loss function ,in, This is the total loss coefficient. The Dice loss coefficient, The cross-entropy loss coefficient is... For boundary loss coefficients, These are the parameters of the loss function;

[0022] S304. Using the SA-UNet model as the basic network architecture, input the basic data of the oral CT model from the training set of the oral CT model segmentation into the SA-UNet model, analyze the probability that each three-dimensional voxel in the basic data of the oral CT model belongs to the CT maxillary tooth label model and the CT mandibular tooth label model, and output the one with the highest probability as the CT model judgment result; using the YOLO model as the basic network architecture, input the basic data of the oral optical model from the training set of the oral optical model segmentation into the YOLO model, analyze the probability that each two-dimensional pixel in the basic data of the oral optical model belongs to the optical maxillary tooth label model and the optical mandibular tooth label model, and output the one with the highest probability as the optical model judgment result;

[0023] S305. Compare the CT model judgment results with the labeled CT maxillary tooth label models and CT mandibular tooth label models, calculate the total loss coefficient corresponding to the CT model judgment results through a combined loss function, and update the corresponding loss function parameters based on the total loss coefficient; compare the optical model judgment results with the labeled optical maxillary tooth label models and optical mandibular tooth label models, calculate the total loss coefficient corresponding to the optical model judgment results through a combined loss function, and update the corresponding loss function parameters based on the total loss coefficient;

[0024] S306. Iterate through the segmented training set of the oral CT model and input it into the SA-UNet model after updating the loss function parameters, until... This process yields a deep learning model for oral CT model segmentation; the oral optical model segmentation training set is then iterated through and input into the updated loss function parameters of the YOLO model, and the computation continues until... This yields a deep learning model for oral optical model segmentation.

[0025] S307. Acquire oral CT 3D data and input it into an oral CT model segmentation deep learning model. Analyze the oral CT 3D data using the oral CT model segmentation deep learning model, calculating the probability that each 3D voxel in the oral CT 3D data belongs to the CT maxillary tooth label model and the CT mandibular tooth label model. When the probability of a 3D voxel belonging to the CT maxillary tooth label model is the highest, it is determined to be a 3D voxel of the CT maxillary tooth 3D data. When the probability of a 3D voxel belonging to the CT mandibular tooth label model is the highest, it is determined to be a 3D voxel of the CT mandibular tooth 3D data. Acquire oral optical 2D data and input it into an oral optical model segmentation deep learning model. Analyze the oral optical 2D data using the oral optical model segmentation deep learning model, calculating the probability that each 2D pixel in the oral optical 2D data belongs to the optical maxillary tooth label model and the optical mandibular tooth label model. When the probability of a 2D pixel belonging to the optical maxillary tooth label model is the highest, it is determined to be a 2D pixel of the optical maxillary tooth 2D data. When the probability of a 2D pixel belonging to the optical mandibular tooth label model is the highest, it is determined to be a 2D pixel of the optical mandibular tooth 2D data.

[0026] S308. Combine each three-dimensional voxel in the oral CT three-dimensional data into corresponding CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data; combine each two-dimensional pixel in the oral optical two-dimensional data into corresponding optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

[0027] The beneficial effects of adopting the above scheme are: constructing an oral cavity segmentation data processing model and training and optimizing the oral cavity segmentation data processing model, and segmenting oral CT three-dimensional data and oral optical two-dimensional data based on the trained and optimized oral cavity segmentation data processing model, can effectively eliminate interference and quickly determine the target area.

[0028] Furthermore, step S5 specifically includes the following steps:

[0029] S501. Obtain the three-dimensional voxel values ​​based on the three-dimensional data analysis of oral CT scans. Among them, the three-dimensional voxel values This is the CT value of the voxel. and This represents the two-dimensional coordinates of the voxel on a CT two-dimensional slice. The slice number of the CT 2D slice;

[0030] S502. Determine whether the three-dimensional voxel value is a valid voxel value based on the preset CT threshold, filter out invalid voxel values ​​based on the determination result and output the valid voxel value;

[0031] S503. Perform 3D connected component labeling on the output effective voxel values ​​to obtain multiple candidate 3D marker point regions. ;

[0032] S504. Perform orthogonal cutting on the candidate three-dimensional marker point region to obtain axial two-dimensional section, sagittal two-dimensional section and coronal two-dimensional section;

[0033] S505. Extract the two-dimensional voxel regions from the axial, sagittal, and coronal two-dimensional sections of the candidate three-dimensional marker point region. ;

[0034] S506. Obtaining a 2D voxel region Given the corresponding area A and perimeter P, calculate the two-dimensional voxel region. roundness ,in, ;

[0035] S507. Filtering Two-dimensional voxel region with a size not less than 0.9 S508. Obtain the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system on the two-dimensional voxel region corresponding to the attached marker. ,in, ;

[0036] S509. Based on the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system, calculate the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system. ,in, , , .

[0037] The beneficial effects of adopting the above scheme are: analyzing the three-dimensional voxel values ​​of CT two-dimensional slices one by one, marking the effective voxel values ​​with three-dimensional connected components and then orthogonally cutting them, and accurately locating the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system by the roundness of the two-dimensional voxel region.

[0038] Furthermore, step S7 specifically includes:

[0039] S701. Obtain the CT 3D point set of the attached markers based on the 3D coordinate data of the attached markers in the CT 3D coordinate system. The optical three-dimensional point set of the attached markers is obtained based on the three-dimensional coordinate data of the attached markers in the optical three-dimensional coordinate system. ,in, This refers to the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system. This refers to the three-dimensional coordinate data of the attached marker points in the optical three-dimensional coordinate system.

[0040] S702. CT 3D point set for attached markers Perform decentralized computation to obtain the centroid of the CT 3D point set. Optical three-dimensional point set for attached markers Perform decentralized computation to obtain the centroid of the optical 3D point set. ;

[0041] S703. Calculate the CT 3D decentering point set of attached markers. Calculate the CT 3D decenter point set of attached markers. ,in, , ;

[0042] S704. Constructing the covariance matrix ,in, By analyzing the covariance matrix Perform singular value decomposition to calculate the rotation matrix ;

[0043] S705. Based on the rotation matrix Calculate the translation vector ,in, ;

[0044] S706. Based on the rotation matrix Translation vector To obtain the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0045] The beneficial effects of adopting the above scheme are: based on the same attached marker point in different three-dimensional coordinate systems such as the CT three-dimensional coordinate system and the optical three-dimensional coordinate system, the rotation matrix and translation vector are calculated through decentralization operation and construction of covariance matrix, thereby obtaining the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0046] Furthermore, step S8 specifically includes:

[0047] S801. When the drill bit mounting sleeve for planting the mobile phone is fitted onto the calibration rod on the calibration plate, control the binocular navigator to acquire image data of the calibration mark points on the calibration plate and the mobile phone tracker;

[0048] S802. Based on the structural parameters of the calibration plate, obtain the three-dimensional coordinate data of the calibration mark point and the calibration rod in the three-dimensional coordinate system of the calibration plate, and calculate the three-dimensional coordinate data of the top and bottom of the calibration rod in the optical three-dimensional coordinate system by combining the image data of the calibration mark point;

[0049] S803. Based on the image data of the mobile phone tracker, the three-dimensional coordinate data of the mobile phone tracker in the optical three-dimensional coordinate system is obtained. Combined with the three-dimensional coordinate data of the mobile phone tracker in the tracker's three-dimensional coordinate system, the coordinate transformation relationship between the tracker's three-dimensional coordinate system and the optical three-dimensional coordinate system is calculated.

[0050] S804. Based on the three-dimensional coordinate data of the top and bottom of the calibration rod in the optical three-dimensional coordinate system, combined with the coordinate transformation relationship between the tracker's three-dimensional coordinate system and the optical three-dimensional coordinate system, the three-dimensional coordinate data of the top and bottom of the calibration rod in the tracker's three-dimensional coordinate system are analyzed and obtained, and the direction vector of the rotation axis of the planting mobile phone in the tracker's three-dimensional coordinate system is calculated.

[0051] S805. Combining the structural parameters of the drill bit and the direction vector of the rotation axis of the implantation mobile phone in the three-dimensional coordinate system of the tracker, the three-dimensional coordinate data of the drill bit tip in the three-dimensional coordinate system of the tracker are calculated.

[0052] The beneficial effects of adopting the above scheme are: by using an external calibration plate to calibrate the implantation mobile phone, the direction vector of the rotation axis of the implantation mobile phone in the three-dimensional coordinate system of the tracker is obtained, and the three-dimensional coordinate data of the drill tip in the three-dimensional coordinate system of the tracker is calculated, which makes it easier for the binocular navigation device to collect image data of the mobile phone tracker to locate the implantation mobile phone in real time during the implantation process.

[0053] The technical solution adopted by this invention to solve the technical problem is as follows:

[0054] A real-time navigation system for oral implant surgery, the system comprising:

[0055] The CT scanning module is used to control the CT scanning equipment to perform CT scans on the patient's oral cavity and acquire three-dimensional CT data of the oral cavity;

[0056] The optical scanning module is used to control the binocular navigation system to scan the patient's oral cavity and acquire two-dimensional optical data of the oral cavity;

[0057] The automatic segmentation processing module is used to construct an oral cavity segmentation data processing model and train and optimize the oral cavity segmentation data processing model. Based on the trained and optimized oral cavity segmentation data processing model, it performs segmentation processing on oral CT three-dimensional data and oral optical two-dimensional data, segmenting the CT three-dimensional model into CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data, and segmenting the oral cavity optical two-dimensional data into optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

[0058] The target region selection module is used to select the target operation region based on the segmentation processing results of the oral cavity segmentation data processing model. It selects CT maxillary tooth 3D data or CT mandibular tooth 3D data as CT target operation data, and selects optical maxillary tooth 2D data or optical mandibular tooth 2D data as optical target operation data.

[0059] The CT data recognition module is used to identify the three-dimensional coordinate data of attached markers in the CT three-dimensional coordinate system from CT target operation data.

[0060] The optical data recognition module is used to identify the three-dimensional coordinate data of the attached marker points in the optical three-dimensional coordinate system based on the two-dimensional oral optical data;

[0061] The rigid body registration module is used to perform rigid body registration operations based on the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system and the three-dimensional coordinate data in the optical three-dimensional coordinate system, and to obtain the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0062] The calibration processing module is used to calibrate the implantation mobile phone, acquire image data from the mobile phone tracker, analyze and obtain the three-dimensional coordinate data of the direction vector of the drill bit rotation axis on the implantation mobile phone in the three-dimensional coordinate system of the tracker, and calculate the three-dimensional coordinate data of the drill bit tip in the three-dimensional coordinate system of the tracker by combining the structural parameters of the drill bit.

[0063] The real-time navigation module is used to control the binocular navigator to acquire real-time image data from the mobile phone tracker and the attached markers. Based on the real-time image data from the mobile phone tracker and the attached markers, and combined with the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system, the module analyzes and obtains the real-time relative position of the patient's oral cavity and the drill tip in the CT three-dimensional coordinate system, thereby enabling real-time navigation.

[0064] Furthermore, the automatic segmentation processing module includes:

[0065] The basic data acquisition unit is used to acquire multiple sets of basic data of oral CT models and multiple sets of basic data of oral optical models. The basic data of oral CT models are labeled to obtain CT maxillary tooth label models and CT mandibular tooth label models. The basic data of oral optical models are labeled to obtain optical maxillary tooth label models and optical mandibular tooth label models.

[0066] The dataset partitioning unit is used to partition the basic data of the oral CT model to obtain the oral CT model segmentation training set, oral CT model segmentation validation set, and oral CT model segmentation test set; and to partition the basic data of the oral optical model to obtain the oral optical model segmentation training set, oral optical model segmentation validation set, and oral optical model segmentation test set.

[0067] Function building blocks, used to define the combined loss function ,in, This is the total loss coefficient. The Dice loss coefficient, The cross-entropy loss coefficient is... For boundary loss coefficients, These are the parameters of the loss function;

[0068] The training and judgment unit, based on the SA-UNet model as the basic network architecture, inputs the basic data of the oral CT model from the oral CT model segmentation training set into the SA-UNet model, analyzes the probability that each 3D voxel in the basic data belongs to the CT maxillary tooth label model and the CT mandibular tooth label model, and outputs the one with the highest probability as the CT model judgment result. Based on the YOLO model as the basic network architecture, it inputs the basic data of the oral optical model from the oral optical model segmentation training set into the YOLO model, analyzes the probability that each 2D pixel in the basic data belongs to the optical maxillary tooth label model and the optical mandibular tooth label model, and outputs the one with the highest probability as the optical model judgment result. The result comparison unit compares the CT model judgment result with the labeled CT maxillary tooth label model and the CT mandibular tooth label model, calculates the total loss coefficient corresponding to the CT model judgment result using a combined loss function, and updates the corresponding loss function parameters based on the total loss coefficient. It also compares the optical model judgment result with the labeled optical maxillary tooth label model and the optical mandibular tooth label model, calculates the total loss coefficient corresponding to the optical model judgment result using a combined loss function, and updates the corresponding loss function parameters based on the total loss coefficient.

[0069] The iterative computation unit is used to iterate through the SA-UNet model after segmenting the oral CT model training set and updating the loss function parameters, until... This process yields a deep learning model for oral CT model segmentation; the oral optical model segmentation training set is then iterated through and input into the updated loss function parameters of the YOLO model, and the computation continues until... This yields a deep learning model for oral optical model segmentation.

[0070] An automatic segmentation unit is used to acquire oral CT 3D data and input it into an oral CT model segmentation deep learning model. The model analyzes the oral CT 3D data, calculating the probability that each 3D voxel belongs to the CT maxillary tooth label model or the CT mandibular tooth label model. If the probability of a voxel belonging to the maxillary tooth label model is highest, it is identified as a 3D voxel of the CT maxillary tooth 3D data; if the probability of a voxel belonging to the mandibular tooth label model is highest, it is identified as a 3D voxel of the CT mandibular tooth 3D data. Similarly, an automatic segmentation unit acquires oral optical 2D data and inputs it into an oral optical model segmentation deep learning model. This model analyzes the oral optical 2D data, calculating the probability that each 2D pixel belongs to the optical maxillary tooth label model or the optical mandibular tooth label model. If the probability of a 2D pixel belonging to the maxillary tooth label model is highest, it is identified as a 2D pixel of the optical maxillary tooth 2D data; if the probability of a 2D pixel belonging to the mandibular tooth label model is highest, it is identified as a 2D pixel of the optical mandibular tooth 2D data.

[0071] The model building unit is used to combine each three-dimensional voxel in the oral CT three-dimensional data into corresponding CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data; and to combine each two-dimensional pixel in the oral optical two-dimensional data into corresponding optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

[0072] Furthermore, the CT data recognition module includes:

[0073] The voxel value calculation unit is used to obtain three-dimensional voxel values ​​based on the analysis of three-dimensional oral CT data. Among them, the three-dimensional voxel values This is the CT value of the voxel. and This represents the two-dimensional coordinates of the voxel on a CT two-dimensional slice. The slice number of the CT 2D slice;

[0074] The voxel value filtering unit is used to determine whether the three-dimensional voxel value is a valid voxel value based on the preset CT threshold, filter invalid voxel values ​​based on the judgment result and output the valid voxel value;

[0075] Region connectivity units are used to label the output effective voxel values ​​with 3D connectivity components, resulting in multiple candidate 3D marker point regions. ;

[0076] Orthogonal cutting units are used to orthogonally cut the candidate three-dimensional marker point region to obtain axial two-dimensional cut surfaces, sagittal two-dimensional cut surfaces, and coronal two-dimensional cut surfaces;

[0077] The region extraction unit is used to extract two-dimensional voxel regions from the axial two-dimensional cross-sections, sagittal two-dimensional cross-sections, and coronal two-dimensional cross-sections of the candidate three-dimensional marker point regions. ;

[0078] Roundness calculation unit, used to obtain two-dimensional voxel regions Given the corresponding area A and perimeter P, calculate the two-dimensional voxel region. roundness ,in, ;

[0079] Roundness judgment unit, used for filtering Two-dimensional voxel region with a size not less than 0.9 As the two-dimensional voxel region corresponding to the attached marker point;

[0080] The coordinate value acquisition unit acquires the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system on the two-dimensional voxel region of the attached marker point. ,in, ;

[0081] The average coordinate calculation unit is used to calculate the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system based on the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system. ,in, , , .

[0082] Furthermore, the rigid body registration module includes:

[0083] The 3D point set acquisition unit is used to acquire the CT 3D point set of the attached markers based on the 3D coordinate data of the attached markers in the CT 3D coordinate system. The optical three-dimensional point set of the attached markers is obtained based on the three-dimensional coordinate data of the attached markers in the optical three-dimensional coordinate system. ,in, This refers to the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system. This refers to the three-dimensional coordinate data of the attached marker points in the optical three-dimensional coordinate system.

[0084] Decentralized unit, used for CT 3D point sets of attached markers. Perform decentralized computation to obtain the centroid of the CT 3D point set. Optical three-dimensional point set for attached markers Perform decentralized computation to obtain the centroid of the optical 3D point set. ;

[0085] The decentering point set calculation unit is used to calculate the CT 3D decentering point set of attached markers. Calculate the CT 3D decenter point set of attached markers. ,in, , ;

[0086] Rotation parameter calculation unit, used to construct the covariance matrix. ,in, By analyzing the covariance matrix Perform singular value decomposition to calculate the rotation matrix ;

[0087] The translation parameter calculation unit root is used to calculate the rotation matrix. Calculate the translation vector ,in, ;

[0088] Transformation relation calculation unit, used to calculate the transformation relation based on the rotation matrix. Translation vector To obtain the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0089] Correspondingly, a storage medium stores a computer program, the computer program including program instructions, which, when executed by a processor, execute the real-time navigation method for oral implant surgery as described above. Attached Figure Description

[0090] Figure 1 This is a flowchart of the real-time navigation method for oral implant surgery according to the present invention.

[0091] Figure 2 This is a schematic diagram of the real-time navigation system for oral implant surgery of the present invention.

[0092] The components represented by each number in the diagram are listed below:

[0093] CT Scanning Module 1, Optical Scanning Module 2, Automatic Segmentation Processing Module 3, Target Region Selection Module 4, CT Data Recognition Module 5, Optical Data Recognition Module 6, Rigid Body Registration Module 7, Calibration Processing Module 8, Real-time Navigation Module 9. Detailed Implementation

[0094] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0095] In the description of this invention, it should be understood that the terms "center," "upper," "lower," "front," "rear," "left," and "right," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0096] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. When a component is referred to as being "fixed to" or "set on" another element, it can be directly on the other component or there may be an intervening component. When a component is considered to be "connected" to another element, it can be directly connected to the other element or there may be an intervening component. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0097] The procedure for dental implant surgery is as follows: an implant hole is drilled in the alveolar bone at the location of the missing tooth, the implant is inserted into the implant hole, and then a fixed crown is installed on the implant to form the overall structure of the dental implant.

[0098] Drilling the implant hole is a crucial step in the entire procedure. Ideally, the operator needs to drill an implant hole of a specific orientation and depth, taking into account factors such as the patient's alveolar bone condition and the orientation of adjacent teeth. Since damage to the alveolar bone is irreparable, accurately drilling an implant hole suitable for specific oral conditions is extremely important. A binocular navigation system is used to collect data from an oral reference board and a handpiece tracker to determine the real-time position of the patient's mouth and the implant handpiece. Based on this real-time relative position, real-time navigation can be provided for the implant handpiece operation.

[0099] Current navigation methods require first fixing a registration device with ceramic balls to the tooth, then installing a dental reference plate with markers on the registration device. Real-time navigation is then achieved by establishing a connection between the ceramic balls, the dental reference plate, and the binocular navigation system via CT scans. However, this method has the following problems: First, the dental reference plate is indirectly fixed through the registration device, which can easily loosen during operation, leading to navigation and positioning errors and affecting implant accuracy. Second, the dental reference plate is large, and when it protrudes outside the mouth, surgical instruments may touch it, interfering with the operation and affecting navigation accuracy. Third, the dual marker design of the ceramic balls and the reference plate not only increases registration errors but also makes the registration process cumbersome.

[0100] To address the aforementioned problems, this invention provides a real-time navigation method, system, and storage medium for oral implant surgery. The method, system, and storage medium are implemented based on adhesive markers, wherein at least three adhesive markers are provided. The adhesive markers are fixedly disposed on the labial side of the patient's maxillary teeth and / or mandibular teeth, and the adhesive markers are black resin discs containing barium sulfate contrast agent.

[0101] In layman's terms, at least three teeth in the upper jaw or at least three teeth in the lower jaw of the patient's mouth are selected. Adhesive markers with contrast-enhancing properties are attached to the tooth surfaces, avoiding the formation of straight lines between different markers during the application process. On one hand, the high contrast between the markers and the teeth facilitates the acquisition of image data of the markers using a binocular navigation system, determining their relative positions on the oral cavity's two-dimensional optical data. On the other hand, the markers are barium sulfate contrast agents, allowing for high CT values ​​obtained after scanning with a CT scanner, which are then used to determine their relative positions on the oral cavity's three-dimensional CT data.

[0102] like Figure 1 As shown, in order to solve the above problems, the present invention provides a real-time navigation method for oral implant surgery, the method comprising the following steps:

[0103] S1. Control the CT scanning equipment to perform a CT scan on the patient's oral cavity to obtain oral CT three-dimensional data. In step S1, the CT scanning equipment can establish oral CT three-dimensional data by performing a CT scan on the patient's oral cavity; while performing a CT scan on the patient's oral cavity, a CT scan is also performed on the area of ​​the fixed attached marker, thereby determining the position of the attached marker on the oral CT three-dimensional data.

[0104] S2. Control the binocular navigation system to scan the patient's oral cavity and acquire two-dimensional optical data of the oral cavity. In step S2, scanning the patient's oral cavity with the binocular navigation system establishes two-dimensional optical data of the oral cavity; while scanning the patient's oral cavity, the area of ​​the fixed attached marker is also scanned, thereby determining the position of the attached marker on the two-dimensional optical data of the oral cavity. As can be seen from steps S1 and S2, in this technical solution, thanks to the use of black resin discs containing barium sulfate contrast agent as attached markers, it can simultaneously serve as CT markers and optical markers.

[0105] S3. Construct and train an oral cavity segmentation data processing model. Based on the trained and optimized model, segment the oral CT 3D data and oral optical 2D data. The CT 3D model is segmented into CT maxillary tooth 3D data and CT mandibular tooth 3D data, and the oral optical 2D data is segmented into optical maxillary tooth 2D data and optical mandibular tooth 2D data. During oral implant surgery navigation, both CT scanning equipment and binocular navigation systems have very large scanning ranges. The attached markers are 3mm in diameter and 0.5mm thick, requiring precise location over a large area, making them susceptible to interference. To eliminate interference, in step S3, the oral CT 3D data and oral optical 2D data are segmented using the oral cavity segmentation data processing model. AI technology is used to roughly locate the maxillary and mandibular tooth regions, which helps improve data processing efficiency.

[0106] S4. Based on the segmentation results of the oral cavity segmentation data processing model, the target operation region is selected. CT 3D data of the maxillary or mandibular teeth is selected as the CT target operation data, and optical 2D data of the maxillary or mandibular teeth is selected as the optical target operation data. In step S4, the target operation region for this operation is determined on the already segmented CT 3D data of the maxillary and mandibular teeth. If implantation is to be performed in the maxilla, the CT 3D data of the maxilla is selected as the CT target operation data, and correspondingly, the optical 2D data of the maxilla is also selected as the optical target operation data. If implantation is to be performed in the mandible, the CT 3D data of the mandibular teeth is selected as the CT target operation data, and correspondingly, the optical 2D data of the maxilla is also selected as the optical target operation data.

[0107] S5. Identify the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system from the CT target operation data. In step S1, during the scanning process to form the oral CT three-dimensional data, the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system is also obtained. Step S5 then identifies the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system on the CT target operation data corresponding to the oral CT three-dimensional data.

[0108] S6. Based on the two-dimensional oral optical data, identify the three-dimensional coordinates of the attached marker in the optical three-dimensional coordinate system within the optical target operation data. The binocular navigation system calculates the three-dimensional coordinates of the two-dimensional oral optical data in the optical three-dimensional coordinate system using the triangulation principle, based on the data acquired by the left and right cameras. In step S1, during the scanning process to form the two-dimensional oral optical data, the three-dimensional coordinates of the attached marker in the optical three-dimensional coordinate system are also obtained. Step S5 then identifies the three-dimensional coordinates of the attached marker in the optical three-dimensional coordinate system on the optical target operation data corresponding to the two-dimensional oral optical data.

[0109] S7. Perform rigid body registration based on the 3D coordinate data of the attached markers in the CT 3D coordinate system and the 3D coordinate data in the optical 3D coordinate system to obtain the coordinate transformation relationship between the CT 3D coordinate system and the optical 3D coordinate system. In step S7, based on the 3D coordinate data of the attached markers in the CT 3D coordinate system and the 3D coordinate data in the optical 3D coordinate system, the coordinate transformation relationship between the CT 3D coordinate system and the optical 3D coordinate system can be calculated through rigid body registration using the 3D coordinate data of the same object in different 3D coordinate systems.

[0110] S8. The implantation device is calibrated to obtain image data from the device tracker. The 3D coordinates of the drill bit rotation axis direction vector in the tracker's 3D coordinate system are analyzed. Combined with the drill bit's structural parameters, the 3D coordinates of the drill bit tip in the tracker's 3D coordinate system are calculated. In step S8, the implantation device is calibrated using an external calibration device, such as a calibration plate, to determine the 3D coordinates of the drill bit rotation axis direction vector and the drill bit tip in the tracker's 3D coordinate system.

[0111] S9. Control the binocular navigation system to acquire real-time image data from the implant handpiece tracker and the attached markers. Based on the real-time image data from the implant handpiece tracker and the attached markers, and combining the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system, analyze the real-time relative position of the patient's oral cavity and the drill tip in the CT three-dimensional coordinate system, thereby performing real-time navigation. In step S9, the real-time image data from the implant handpiece tracker acquired by the binocular navigation system can determine the real-time pose of the implant handpiece, and the real-time image data from the attached markers acquired by the binocular navigation system can determine the real-time pose of the patient's oral cavity. Combining the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system, the real-time relative position of the implant handpiece and the patient's oral cavity in the CT three-dimensional coordinate system can be obtained, thereby performing real-time navigation.

[0112] Based on the above technical solution, after attaching adhesive markers to the patient's teeth, CT scans and image scans are performed on the patient's oral cavity. The coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system is obtained directly through registration using the adhesive markers. Real-time navigation is then achieved by acquiring real-time image data from the mobile phone tracker and the adhesive markers using a binocular navigation system. The adhesive markers simultaneously serve as CT and optical markers, eliminating the need for indirect fixation of the oral reference plate using a registration device, thus fundamentally solving the potential loosening and positioning errors between the registration device and the oral reference plate. The adhesive markers are attached to the patient's teeth, with negligible volume, avoiding contact that could interfere with the operation. By simultaneously serving as CT and optical markers, the registration operation in existing technologies is eliminated, effectively reducing errors.

[0113] Preferably, step S3 specifically includes the following steps:

[0114] S301. Acquire multiple sets of basic data from oral CT models and multiple sets of basic data from oral optical models. Annotate the basic data from the oral CT models to obtain CT maxillary tooth label models and CT mandibular tooth label models. Annotate the basic data from the oral optical models to obtain optical maxillary tooth label models and optical mandibular tooth label models. In step S301, operators manually identify and annotate the multiple sets of basic data from the oral CT models and multiple sets of basic data from the oral optical models. Annotating the basic data from the oral CT models yields CT maxillary tooth label models and CT mandibular tooth label models, and annotating the basic data from the oral optical models yields optical maxillary tooth label models and optical mandibular tooth label models. This prepares the basic data for subsequent dataset partitioning and model training. During training optimization, a larger quantity of basic data from the oral CT models and oral optical models results in better training optimization performance.

[0115] S302. The basic data of the oral CT model is divided into a training set, a validation set, and a test set for oral CT model segmentation; the basic data of the oral optical model is also divided into a training set, a validation set, and a test set for oral optical model segmentation. In step S302, the basic data is divided according to function, specifically into a training set, a validation set, and a test set according to the model training process.

[0116] S303. Define the combination loss function ,in, This is the total loss coefficient. The Dice loss coefficient, The cross-entropy loss coefficient is... For boundary loss coefficients, These are the parameters of the loss function. For multi-class segmentation, the Dice loss is usually calculated separately for each class of maxillary and mandibular teeth and then averaged. This can avoid the model being dominated by the class with a large number of voxels (such as the jawbone). Using standard multi-class cross-entropy loss, the probability distribution of the model output is compared pixel-by-pixel with the one-hot encoding of the true label. For each category, a boundary distance map is calculated. For each category segmentation region of the ground truth label, the shortest distance from each voxel within that region to the boundary of that category is calculated. The loss is calculated based on the predicted probability map and the boundary distance map. S304. Using the SA-UNet model as the basic network architecture, the basic data of the oral CT model from the oral CT model segmentation training set is input into the SA-UNet model. The probability of each 3D voxel in the basic data of the oral CT model belonging to the CT maxillary tooth label model and the CT mandibular tooth label model is analyzed, and the highest probability is output as the CT model judgment result. Using the YOLO model as the basic network architecture, the basic data of the oral optical model from the oral optical model segmentation training set is input into the YOLO model. The probability of each 2D pixel in the basic data of the oral optical model belonging to the optical maxillary tooth label model and the optical mandibular tooth label model is analyzed, and the highest probability is output as the optical model judgment result. In step 304, the SA-UNet model and the YOLO model are the foundational architectures for medical 3D and 2D image segmentation, respectively. The encoder-decoder structure and skip connections of the SA-UNet model and the YOLO model can simultaneously capture the adjacent information of the image to identify different oral tissues and accurately locate and clearly indicate where the boundaries of different tissues and different teeth are, so as to better distinguish between maxillary and mandibular teeth.

[0117] S305. Compare the CT model judgment results with the labeled CT maxillary and mandibular tooth label models. Calculate the total loss coefficient corresponding to the CT model judgment result using a combined loss function, and update the corresponding loss function parameters based on the total loss coefficient. Compare the optical model judgment results with the labeled optical maxillary and mandibular tooth label models. Calculate the total loss coefficient corresponding to the optical model judgment result using a combined loss function, and update the corresponding loss function parameters based on the total loss coefficient. The purpose of step S305 is to update parameters through training, with the goal of reducing the total loss. .

[0118] S306. Iterate through the segmented training set of the oral CT model and input it into the SA-UNet model after updating the loss function parameters, until... This process yields a deep learning model for oral CT model segmentation; the oral optical model segmentation training set is then iterated through and input into the updated loss function parameters of the YOLO model, and the computation continues until... This yields a deep learning model for oral optical model segmentation. (The goal is to achieve...) Once the training objective is achieved, training can be stopped, resulting in the trained deep learning models for oral CT model segmentation and oral optical model segmentation. At this point, automatic recognition can be performed based on these two models. S307. Acquire oral CT 3D data and input it into an oral CT model segmentation deep learning model. Analyze the oral CT 3D data using the oral CT model segmentation deep learning model, calculating the probability that each 3D voxel in the oral CT 3D data belongs to the CT maxillary tooth label model and the CT mandibular tooth label model. When the probability of a 3D voxel belonging to the CT maxillary tooth label model is the highest, it is determined to be a 3D voxel of the CT maxillary tooth 3D data. When the probability of a 3D voxel belonging to the CT mandibular tooth label model is the highest, it is determined to be a 3D voxel of the CT mandibular tooth 3D data. Acquire oral optical 2D data and input it into an oral optical model segmentation deep learning model. Analyze the oral optical 2D data using the oral optical model segmentation deep learning model, calculating the probability that each 2D pixel in the oral optical 2D data belongs to the optical maxillary tooth label model and the optical mandibular tooth label model. When the probability of a 2D pixel belonging to the optical maxillary tooth label model is the highest, it is determined to be a 2D pixel of the optical maxillary tooth 2D data. When the probability of a 2D pixel belonging to the optical mandibular tooth label model is the highest, it is determined to be a 2D pixel of the optical mandibular tooth 2D data.

[0119] S308. Combine each three-dimensional voxel in the oral CT three-dimensional data into corresponding CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data; combine each two-dimensional pixel in the oral optical two-dimensional data into corresponding optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

[0120] Based on the above data processing steps, an oral cavity segmentation data processing model is constructed and trained and optimized. Based on the trained and optimized oral cavity segmentation data processing model, oral CT three-dimensional data and oral optical two-dimensional data are segmented, which can effectively eliminate interference and quickly determine the target area.

[0121] Preferably, step S5 specifically includes the following steps:

[0122] S501. Obtain the three-dimensional voxel values ​​based on the three-dimensional data analysis of oral CT scans. Among them, the three-dimensional voxel values This is the CT value of the voxel. and This represents the two-dimensional coordinates of the voxel on a CT two-dimensional slice. This refers to the slice number of the CT 2D slice. In step S501, because the CT value of the attached marker is higher than that of the surrounding tissues such as gums, teeth, and bones, the voxel corresponding to the attached marker will be distinguished from the voxels of the surrounding tissues during CT scanning.

[0123] S502. Determine whether the three-dimensional voxel value is a valid voxel value based on a preset CT threshold, filter invalid voxel values ​​based on the determination result, and output the valid voxel value. In step S502, the three-dimensional voxel value is determined to be a valid voxel value by using a preset CT threshold, thereby determining whether the voxel is the voxel corresponding to the attached marker point.

[0124] S503. Perform 3D connected component labeling on the output effective voxel values ​​to obtain multiple candidate 3D marker point regions. In step S503, the interconnected sets of voxels in the oral CT three-dimensional data are identified as independent clumps, and each clump is a candidate three-dimensional marker region.

[0125] S504. The candidate three-dimensional marker point region is orthogonally cut to obtain axial two-dimensional sections, sagittal two-dimensional sections, and coronal two-dimensional sections. In step S505, the candidate three-dimensional marker point region is projected and analyzed on the three orthogonal two-dimensional planes, namely the axial two-dimensional section, the sagittal two-dimensional section, and the coronal two-dimensional section, so as to more comprehensively and three-dimensionally determine the roundness of the candidate three-dimensional marker point region.

[0126] S505. Extract the two-dimensional voxel regions from the axial, sagittal, and coronal two-dimensional sections of the candidate three-dimensional marker point region. .

[0127] S506. Obtaining a 2D voxel region Given the corresponding area A and perimeter P, calculate the two-dimensional voxel region. roundness ,in, .

[0128] S507. Filtering Two-dimensional voxel region with a size not less than 0.9 This refers to the two-dimensional voxel region corresponding to the attached marker. For an attached marker, it needs to meet a high roundness across multiple consecutive slices to be finally confirmed as an attached marker. In this technical solution, roundness is judged using 0.9.

[0129] S508. Obtain the 3D coordinate data of all N voxels in the CT 3D coordinate system on the 2D voxel region of the attached marker. ,in, In step S508, for the two-dimensional voxel region that has been determined to be an attached marker point, the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system are obtained. .

[0130] S509. Based on the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system, calculate the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system. ,in, , , In step S509, the average coordinate values ​​of N voxels are used as the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system.

[0131] Based on the above data processing flow, the three-dimensional voxel values ​​of CT two-dimensional slices are analyzed one by one. The effective voxel values ​​are marked with three-dimensional connected components and then orthogonally cut. By the roundness of the two-dimensional voxel region, the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system can be accurately located.

[0132] Preferably, step S7 specifically includes:

[0133] S701. Obtain the CT 3D point set of the attached markers based on the 3D coordinate data of the attached markers in the CT 3D coordinate system. The optical three-dimensional point set of the attached markers is obtained based on the three-dimensional coordinate data of the attached markers in the optical three-dimensional coordinate system. ,in, This refers to the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system. The three-dimensional coordinate data of the attached marker points in the optical three-dimensional coordinate system are given in steps S5 and S6. Based on this, the corresponding CT three-dimensional point set and optical three-dimensional point set are established.

[0134] S702. CT 3D point set for attached markers Perform decentralized computation to obtain the centroid of the CT 3D point set. Optical three-dimensional point set for attached markers Perform decentralized computation to obtain the centroid of the optical 3D point set. .

[0135] S703. Calculate the CT 3D decentering point set of attached markers. Calculate the CT 3D decenter point set of attached markers. ,in, , .

[0136] S704. Constructing the covariance matrix ,in, By analyzing the covariance matrix Perform singular value decomposition to calculate the rotation matrix . for The transpose of .

[0137] S705. Based on the rotation matrix Calculate the translation vector ,in, .

[0138] S706. Based on the rotation matrix Translation vector To obtain the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0139] Preferably, step S8 specifically includes:

[0140] S801. When the drill bit mounting sleeve of the implantation mobile phone is fitted onto the calibration rod on the calibration plate, the binocular navigator is controlled to acquire image data of the calibration markers on the calibration plate and the mobile phone tracker. When the calibration markers and the mobile phone tracker are within the camera field of view of the binocular navigator, the image data of the calibration markers and the mobile phone tracker on the calibration plate can be acquired through the binocular navigator, preparing for real-time positioning of the calibration plate and the implantation mobile phone.

[0141] S802. Based on the structural parameters of the calibration plate, obtain the three-dimensional coordinate data of the calibration markers and calibration rod in the three-dimensional coordinate system of the calibration plate. Combine this with the image data of the calibration markers to calculate the three-dimensional coordinate data of the top and bottom ends of the calibration rod in the optical three-dimensional coordinate system. When designing the calibration plate, its structural parameters are known. The calibration plate is fabricated based on these parameters. Therefore, the three-dimensional coordinate data of the calibration markers and calibration rod in the three-dimensional coordinate system of the calibration plate can be directly obtained from these parameters. Since the image data of the calibration markers on the calibration plate has already been obtained in step S802, the three-dimensional coordinate data of the calibration markers in the optical three-dimensional coordinate system can be known. This data can then be converted to obtain the three-dimensional coordinate data of the calibration rod in the optical three-dimensional coordinate system, thus yielding the three-dimensional coordinate data of the top and bottom ends of the calibration rod in the optical three-dimensional coordinate system.

[0142] S803. Based on the image data analysis of the mobile phone tracker, the three-dimensional coordinate data of the mobile phone tracker in the optical three-dimensional coordinate system is obtained. Combined with the three-dimensional coordinate data of the mobile phone tracker in the tracker's three-dimensional coordinate system, the coordinate transformation relationship between the tracker's three-dimensional coordinate system and the optical three-dimensional coordinate system is calculated. In step S803, the three-dimensional coordinate data of the mobile phone tracker in the tracker's three-dimensional coordinate system is known by design, which is the same as the known three-dimensional coordinate data of the calibration markers and calibration rods in the calibration plate's three-dimensional coordinate system. Since the image data of the mobile phone tracker has already been acquired through the binocular navigator in step S801, the three-dimensional coordinate data of the mobile phone tracker in the optical three-dimensional coordinate system can be obtained, and the coordinate transformation relationship between the tracker's three-dimensional coordinate system and the optical three-dimensional coordinate system can be further calculated.

[0143] S804. Based on the three-dimensional coordinate data of the top and bottom ends of the calibration rod in the optical three-dimensional coordinate system, and combining the coordinate transformation relationship between the tracker's three-dimensional coordinate system and the optical three-dimensional coordinate system, analyze and obtain the three-dimensional coordinate data of the top and bottom ends of the calibration rod in the tracker's three-dimensional coordinate system, and calculate the direction vector of the rotation axis of the implantation mobile phone in the tracker's three-dimensional coordinate system. Based on the data known in steps S801-S803, the direction vector of the rotation axis of the implantation mobile phone in the tracker's three-dimensional coordinate system can be obtained through a simple matrix transformation.

[0144] S805. Combining the structural parameters of the drill bit and the direction vector of the rotation axis of the implantation mobile phone in the three-dimensional coordinate system of the tracker, the three-dimensional coordinate data of the drill bit tip in the three-dimensional coordinate system of the tracker are calculated.

[0145] Based on the above data processing, the planting device is calibrated with the help of an external calibration plate to obtain the direction vector of the rotation axis of the planting device in the three-dimensional coordinate system of the tracker. The three-dimensional coordinate data of the drill tip in the three-dimensional coordinate system of the tracker are calculated, so that the binocular navigation device can collect image data of the device tracker to locate the planting device in real time during the planting process.

[0146] like Figure 2 As shown, to solve the above problems, the present invention provides a real-time navigation system for oral implant surgery. The system includes a CT scanning module 1, an optical scanning module 2, an automatic segmentation processing module 3, a target region selection module 4, a CT data recognition module 5, an optical data recognition module 6, a rigid body registration module 7, a calibration processing module 8, and a real-time navigation module 9, wherein:

[0147] CT scanning module 1 is used to control the CT scanning equipment to perform CT scans on the patient's oral cavity and acquire three-dimensional CT data of the oral cavity;

[0148] Optical scanning module 2 is used to control the binocular navigation device to scan the patient's oral cavity and acquire two-dimensional optical data of the oral cavity;

[0149] Automatic segmentation processing module 3 is used to construct an oral cavity segmentation data processing model and train and optimize the oral cavity segmentation data processing model. Based on the trained and optimized oral cavity segmentation data processing model, it performs segmentation processing on oral CT three-dimensional data and oral optical two-dimensional data, segmenting the CT three-dimensional model into CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data, and segmenting the oral cavity optical two-dimensional data into optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

[0150] The target region selection module 4 is used to select the target operation region based on the segmentation processing results of the oral cavity segmentation data processing model. It selects CT maxillary tooth 3D data or CT mandibular tooth 3D data as CT target operation data, and selects optical maxillary tooth 2D data or optical mandibular tooth 2D data as optical target operation data.

[0151] CT data recognition module 5 is used to identify the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system in the CT target operation data;

[0152] The optical data recognition module 6 is used to identify the three-dimensional coordinate data of the attached marker point in the optical three-dimensional coordinate system based on the two-dimensional oral optical data;

[0153] Rigid body registration module 7 is used to perform rigid body registration calculations based on the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system and the three-dimensional coordinate data in the optical three-dimensional coordinate system, and to obtain the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0154] The calibration processing module 8 is used to calibrate the implantation mobile phone, acquire image data from the mobile phone tracker, analyze and obtain the three-dimensional coordinate data of the direction vector of the drill bit rotation axis on the implantation mobile phone in the three-dimensional coordinate system of the tracker, and calculate the three-dimensional coordinate data of the drill bit tip in the three-dimensional coordinate system of the tracker by combining the structural parameters of the drill bit.

[0155] The real-time navigation module 9 is used to control the binocular navigator to acquire real-time image data from the mobile phone tracker and the attached markers. Based on the real-time image data from the mobile phone tracker and the attached markers, and combined with the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system, the real-time relative position of the patient's oral cavity and the drill tip in the CT three-dimensional coordinate system is analyzed to obtain the real-time relative position of the patient's oral cavity and the drill tip in the CT three-dimensional coordinate system, thereby performing real-time navigation.

[0156] Preferably, the automatic segmentation processing module includes:

[0157] The basic data acquisition unit is used to acquire multiple sets of basic data of oral CT models and multiple sets of basic data of oral optical models. The basic data of oral CT models are labeled to obtain CT maxillary tooth label models and CT mandibular tooth label models. The basic data of oral optical models are labeled to obtain optical maxillary tooth label models and optical mandibular tooth label models.

[0158] The dataset partitioning unit is used to partition the basic data of the oral CT model to obtain the oral CT model segmentation training set, oral CT model segmentation validation set, and oral CT model segmentation test set; and to partition the basic data of the oral optical model to obtain the oral optical model segmentation training set, oral optical model segmentation validation set, and oral optical model segmentation test set.

[0159] Function building blocks, used to define the combined loss function ,in, This is the total loss coefficient. The Dice loss coefficient, The cross-entropy loss coefficient is... For boundary loss coefficients, These are the parameters of the loss function;

[0160] The training and judgment unit is used to input the basic data of the oral CT model from the training set of the oral CT model segmentation into the SA-UNet model, analyze the probability that each three-dimensional voxel in the basic data of the oral CT model belongs to the CT maxillary tooth label model and the CT mandibular tooth label model, and output the one with the highest probability as the CT model judgment result; and to input the basic data of the oral optical model from the training set of the oral optical model segmentation into the YOLO model, analyze the probability that each two-dimensional pixel in the basic data of the oral optical model belongs to the optical maxillary tooth label model and the optical mandibular tooth label model, and output the one with the highest probability as the optical model judgment result.

[0161] The result comparison unit is used to compare the CT model judgment results with the labeled CT maxillary tooth label models and CT mandibular tooth label models. It calculates the total loss coefficient corresponding to the CT model judgment result through a combined loss function and updates the corresponding loss function parameters based on the total loss coefficient. It also compares the optical model judgment results with the labeled optical maxillary tooth label models and optical mandibular tooth label models. It calculates the total loss coefficient corresponding to the optical model judgment result through a combined loss function and updates the corresponding loss function parameters based on the total loss coefficient.

[0162] The iterative computation unit is used to iterate through the SA-UNet model after segmenting the oral CT model training set and updating the loss function parameters, until... This process yields a deep learning model for oral CT model segmentation; the oral optical model segmentation training set is then iterated through and input into the updated loss function parameters of the YOLO model, and the computation continues until... This yields a deep learning model for oral optical model segmentation.

[0163] An automatic segmentation unit is used to acquire oral CT 3D data and input it into an oral CT model segmentation deep learning model. The model analyzes the oral CT 3D data, calculating the probability that each 3D voxel belongs to the CT maxillary tooth label model or the CT mandibular tooth label model. If the probability of a voxel belonging to the maxillary tooth label model is highest, it is identified as a 3D voxel of the CT maxillary tooth 3D data; if the probability of a voxel belonging to the mandibular tooth label model is highest, it is identified as a 3D voxel of the CT mandibular tooth 3D data. Similarly, an automatic segmentation unit acquires oral optical 2D data and inputs it into an oral optical model segmentation deep learning model. This model analyzes the oral optical 2D data, calculating the probability that each 2D pixel belongs to the optical maxillary tooth label model or the optical mandibular tooth label model. If the probability of a 2D pixel belonging to the maxillary tooth label model is highest, it is identified as a 2D pixel of the optical maxillary tooth 2D data; if the probability of a 2D pixel belonging to the mandibular tooth label model is highest, it is identified as a 2D pixel of the optical mandibular tooth 2D data.

[0164] The model building unit is used to combine each three-dimensional voxel in the oral CT three-dimensional data into corresponding CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data; and to combine each two-dimensional pixel in the oral optical two-dimensional data into corresponding optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

[0165] Preferably, the CT data recognition module includes:

[0166] The voxel value calculation unit is used to obtain three-dimensional voxel values ​​based on the analysis of three-dimensional oral CT data. Among them, the three-dimensional voxel values This is the CT value of the voxel. and This represents the two-dimensional coordinates of the voxel on a CT two-dimensional slice. The slice number of the CT 2D slice;

[0167] The voxel value filtering unit is used to determine whether the three-dimensional voxel value is a valid voxel value based on the preset CT threshold, filter invalid voxel values ​​based on the judgment result and output the valid voxel value;

[0168] Region connectivity units are used to label the output effective voxel values ​​with 3D connectivity components, resulting in multiple candidate 3D marker point regions. ;

[0169] Orthogonal cutting units are used to orthogonally cut the candidate three-dimensional marker point region to obtain axial two-dimensional cut surfaces, sagittal two-dimensional cut surfaces, and coronal two-dimensional cut surfaces;

[0170] The region extraction unit is used to extract two-dimensional voxel regions from the axial two-dimensional cross-sections, sagittal two-dimensional cross-sections, and coronal two-dimensional cross-sections of the candidate three-dimensional marker point regions. ;

[0171] Roundness calculation unit, used to obtain two-dimensional voxel regions Given the corresponding area A and perimeter P, calculate the two-dimensional voxel region. roundness ,in, ;

[0172] Roundness judgment unit, used for filtering Two-dimensional voxel region with a size not less than 0.9 As the two-dimensional voxel region corresponding to the attached marker point;

[0173] The coordinate value acquisition unit acquires the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system on the two-dimensional voxel region of the attached marker point. ,in, ;

[0174] The average coordinate calculation unit is used to calculate the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system based on the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system. ,in, , , .

[0175] Preferably, the rigid body registration module includes:

[0176] The 3D point set acquisition unit is used to acquire the CT 3D point set of the attached markers based on the 3D coordinate data of the attached markers in the CT 3D coordinate system. The optical three-dimensional point set of the attached markers is obtained based on the three-dimensional coordinate data of the attached markers in the optical three-dimensional coordinate system. ,in, This refers to the three-dimensional coordinate data of the attached marker points in the CT three-dimensional coordinate system. This refers to the three-dimensional coordinate data of the attached marker points in the optical three-dimensional coordinate system.

[0177] Decentralized unit, used for CT 3D point sets of attached markers. Perform decentralized computation to obtain the centroid of the CT 3D point set. Optical three-dimensional point set for attached markers Perform decentralized computation to obtain the centroid of the optical 3D point set. The decentering point set calculation unit is used to calculate the CT 3D decentering point set of attached markers. Calculate the CT 3D decenter point set of attached markers. ,in, , ;

[0178] Rotation parameter calculation unit, used to construct the covariance matrix. ,in, By analyzing the covariance matrix Perform singular value decomposition to calculate the rotation matrix ;

[0179] The translation parameter calculation unit root is used to calculate the rotation matrix. Calculate the translation vector ,in, ;

[0180] Transformation relation calculation unit, used to calculate the transformation relation based on the rotation matrix. Translation vector To obtain the coordinate transformation relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0181] Preferably, the calibration processing module includes:

[0182] The calibration image data acquisition unit is used to control the binocular navigator to acquire image data of the calibration mark points on the calibration plate and the mobile phone tracker when the drill bit mounting sleeve of the planting mobile phone is fitted onto the calibration rod on the calibration plate.

[0183] The calibration optical coordinate calculation unit is used to obtain the three-dimensional coordinate data of the calibration mark point and the calibration rod in the three-dimensional coordinate system of the calibration plate according to the structural parameters of the calibration plate, and calculate the three-dimensional coordinate data of the top and bottom of the calibration rod in the optical three-dimensional coordinate system by combining the image data of the calibration mark point.

[0184] The coordinate transformation relationship analysis unit is used to analyze the image data of the mobile phone tracker to obtain the three-dimensional coordinate data of the mobile phone tracker in the optical three-dimensional coordinate system, and combine the three-dimensional coordinate data of the mobile phone tracker in the tracker's three-dimensional coordinate system to calculate the coordinate transformation relationship between the tracker's three-dimensional coordinate system and the optical three-dimensional coordinate system.

[0185] The direction vector calculation unit is used to analyze the three-dimensional coordinate data of the top and bottom of the calibration rod in the optical three-dimensional coordinate system, and combine the coordinate transformation relationship between the tracker's three-dimensional coordinate system and the optical three-dimensional coordinate system to obtain the three-dimensional coordinate data of the top and bottom of the calibration rod in the tracker's three-dimensional coordinate system, and calculate the direction vector of the rotation axis of the planting mobile phone in the tracker's three-dimensional coordinate system.

[0186] The calibration calculation unit is used to combine the structural parameters of the drill bit and the direction vector of the rotation axis of the implantation mobile phone in the three-dimensional coordinate system of the tracker to calculate the three-dimensional coordinate data of the drill bit tip in the three-dimensional coordinate system of the tracker.

[0187] To address the aforementioned problems, the present invention provides a storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, execute the real-time navigation method for oral implant surgery as described above.

[0188] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method of real-time navigation for oral implant surgery, characterized in that, The method is based on the implementation of the attached identification point, at least three attached identification points are provided, the attached identification points are fixedly arranged on the labial side of the upper teeth and / or lower teeth of the patient, the attached identification points are black resin discs containing barium sulfate developer, and the method comprises the following steps: S1. Control the CT scanning device to perform CT scanning on the patient's oral cavity to obtain oral CT three-dimensional data; S2. Control the binocular navigator to perform image scanning on the patient's oral cavity to obtain oral optical two-dimensional data; S3. Construct an oral segmentation data processing model and optimize the oral segmentation data processing model, based on the trained and optimized oral segmentation data processing model, perform segmentation processing on the oral CT three-dimensional data and the oral optical two-dimensional data, and divide the CT three-dimensional model into CT upper tooth three-dimensional data and CT lower tooth three-dimensional data, and divide the oral optical two-dimensional data into optical upper tooth two-dimensional data and optical lower tooth two-dimensional data; S4. Select a target operation area based on the segmentation processing result of the oral segmentation data processing model, select the CT upper tooth three-dimensional data or the CT lower tooth three-dimensional data as the CT target operation data, and select the optical upper tooth two-dimensional data or the optical lower tooth two-dimensional data as the optical target operation data; S5. Identify the three-dimensional coordinate data of the attached identification point in the CT three-dimensional coordinate system in the CT target operation data; S6. According to the oral optical two-dimensional data, identify the three-dimensional coordinate data of the attached identification point in the optical three-dimensional coordinate system in the optical target operation data; S7. Perform rigid registration operation according to the three-dimensional coordinate data of the attached identification point in the CT three-dimensional coordinate system and the three-dimensional coordinate data in the optical three-dimensional coordinate system, and obtain the coordinate conversion relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system; S8. Calibrate the implant handpiece to obtain image data of the handpiece tracker, analyze to obtain the three-dimensional coordinate data of the direction vector of the drill needle rotating shaft on the implant handpiece in the tracker three-dimensional coordinate system, and calculate the three-dimensional coordinate data of the drill needle tip in the tracker three-dimensional coordinate system in combination with the structural parameters of the drill needle; S9. Control the binocular navigator to collect real-time image data of the handpiece tracker and the attached identification point, analyze the real-time relative position of the patient's oral cavity and the drill needle tip in the CT three-dimensional coordinate system in combination with the coordinate conversion relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system according to the real-time image data of the handpiece tracker and the attached identification point, and perform real-time navigation; The step S7 specifically comprises: S701. Obtain a CT three-dimensional point set of the attached identification point according to the three-dimensional coordinate data of the attached identification point in the CT three-dimensional coordinate system , obtain an optical three-dimensional point set of the attached identification point according to the three-dimensional coordinate data of the attached identification point in the optical three-dimensional coordinate system , wherein, the three-dimensional coordinate data of the attached identification point in the CT three-dimensional coordinate system, the three-dimensional coordinate data of the attached identification point in the optical three-dimensional coordinate system; S702. CT three-dimensional point set of the attached identification point Carrying out decentralization operation, obtaining the centroid of the CT three-dimensional point set Optical three-dimensional point set of the attached identification point Carrying out decentralization operation, obtaining the centroid of the optical three-dimensional point set ; S703. Calculate the CT three-dimensional de-center point set of the attached identification point , Calculate the CT three-dimensional de-center point set of the attached identification point wherein, , ; S704. Constructing a covariance matrix wherein, by performing a singular value decomposition of the covariance matrix to compute a rotation matrix ; S705. Compute the translation vector Compute the translation vector where, ; S706. Obtain the coordinate conversion relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system according to the rotation matrix and the translation vector .

2. The real-time navigation method for oral implant surgery according to claim 1, wherein, The step S3 specifically comprises the following steps: S301. Obtain a plurality of sets of oral CT model basic data and a plurality of sets of oral optical model basic data, label the oral CT model basic data to obtain a CT upper tooth label model and a CT lower tooth label model, and label the oral optical model basic data to obtain an optical upper tooth label model and an optical lower tooth label model; S302. Perform dataset division on the oral CT model basic data to obtain an oral CT model segmentation training set, an oral CT model segmentation verification set, and an oral CT model segmentation test set; perform dataset division on the oral optical model basic data to obtain an oral optical model segmentation training set, an oral optical model segmentation verification set, and an oral optical model segmentation test set; S303. 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; S304. Take the SA-UNet model as a basic network architecture, input the oral CT model basic data of the oral CT model segmentation training set into the SA-UNet model, analyze the probability of each three-dimensional voxel in the oral CT model basic data belonging to the CT maxillary tooth label model and the CT mandibular tooth label model, and output the highest probability as the CT model judgment result; take the YOLO model as a basic network architecture, input the oral optical model basic data of the oral optical model segmentation training set into the YOLO model, analyze the probability of each two-dimensional pixel in the oral optical model basic data belonging to the optical maxillary tooth label model and the optical mandibular tooth label model, and output the highest probability as the optical model judgment result; S305. Compare the CT model judgment result with the CT maxillary tooth label model and the CT mandibular tooth label model obtained by labeling, calculate the total loss coefficient corresponding to the CT model judgment result through the combined loss function, and update the corresponding loss function parameters according to the total loss coefficient; compare the optical model judgment result with the optical maxillary tooth label model and the optical mandibular tooth label model obtained by labeling, calculate the total loss coefficient corresponding to the optical model judgment result through the combined loss function, and update the corresponding loss function parameters according to the total loss coefficient; S306. The SA-UNet model is traversed to input the oral CT model segmentation training set after updating the loss function parameters, and iterative operation is performed until , an oral CT model segmentation deep learning model is obtained; the YOLO model is traversed to input the oral optical model segmentation training set after updating the loss function parameters, and iterative operation is performed until , an oral optical model segmentation deep learning model is obtained; S307. Obtain oral CT three-dimensional data and input the oral CT three-dimensional data into the oral CT model segmentation deep learning model, analyze the oral CT three-dimensional data through the oral CT model segmentation deep learning model, calculate the probability of each three-dimensional voxel in the oral CT three-dimensional data belonging to the CT maxillary tooth label model and the CT mandibular tooth label model, and when the probability of the three-dimensional voxel belonging to the CT maxillary tooth label model is the highest, it is judged as a three-dimensional voxel of the CT maxillary tooth three-dimensional data, and when the probability of the three-dimensional voxel belonging to the CT mandibular tooth label model is the highest, it is judged as a three-dimensional voxel of the CT mandibular tooth three-dimensional data; obtain oral optical two-dimensional data and input the oral optical two-dimensional data into the oral optical model segmentation deep learning model, analyze the oral optical two-dimensional data through the oral optical model segmentation deep learning model, calculate the probability of each two-dimensional pixel in the oral optical two-dimensional data belonging to the optical maxillary tooth label model and the optical mandibular tooth label model, and when the probability of the two-dimensional pixel belonging to the optical maxillary tooth label model is the highest, it is judged as a two-dimensional pixel of the optical maxillary tooth two-dimensional data, and when the probability of the two-dimensional pixel belonging to the optical mandibular tooth label model is the highest, it is judged as a two-dimensional pixel of the optical mandibular tooth two-dimensional data; S308. Grouping each three-dimensional voxel in the oral cavity CT three-dimensional data into corresponding CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data; and grouping each two-dimensional pixel in the oral cavity optical two-dimensional data into corresponding optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

3. The real-time navigation method for oral implant surgery according to claim 1, wherein, The step S5 specifically comprises the following steps: S501. obtaining a three-dimensional voxel value according to oral cavity CT three-dimensional data analysis wherein the three-dimensional voxel value is a CT value of the voxel, and is a pixel two-dimensional coordinate value of the voxel on a CT two-dimensional slice, is a slice serial number of the CT two-dimensional slice; S502. Judging whether a three-dimensional voxel value is a valid voxel value according to a preset CT threshold value, filtering invalid voxel values and outputting valid voxel values according to a judgment result; S503. Perform three-dimensional connected component labeling on the output valid voxel values to obtain a plurality of candidate three-dimensional landmark regions ; S504. Orthogonal cutting the candidate three-dimensional mark point region to obtain axial two-dimensional sections, sagittal two-dimensional sections and coronal two-dimensional sections; S505. Extracting the two-dimensional voxel regions on the axial two-dimensional cut surface, the sagittal two-dimensional cut surface and the coronal two-dimensional cut surface of the axis candidate three-dimensional landmark point region ; S506. Acquire a two-dimensional voxel region The roundness of the two-dimensional voxel region is calculated corresponding to the area A and the perimeter P wherein, ; S507. screening a two-dimensional voxel region of not less than 0.9 a two-dimensional voxel region corresponding to the attached marker S508. Obtain three-dimensional coordinate data of all N voxels on the two-dimensional voxel region of the attached marker point in the CT three-dimensional coordinate system wherein, ; S509. Calculate the three-dimensional coordinate data of the attached identification point in the CT three-dimensional coordinate system according to the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system wherein, , , .

4. The real-time navigation method for oral implant surgery according to claim 1, wherein, The step S8 specifically comprises: S801. When the drill bit mounting sleeve of the implant handset is sleeved on the calibration rod on the calibration plate, controlling the binocular navigator to acquire image data of the calibration mark points on the calibration plate and the handset tracker; S802. According to the structural parameters of the calibration plate, acquiring three-dimensional coordinate data of the calibration mark points and the calibration rod in the calibration plate three-dimensional coordinate system, and combining the image data of the calibration mark points to calculate three-dimensional coordinate data of the top end and the bottom end of the calibration rod in the optical three-dimensional coordinate system; S803. According to the image data of the handset tracker, analyzing to obtain three-dimensional coordinate data of the handset tracker in the optical three-dimensional coordinate system, combining three-dimensional coordinate data of the handset tracker in the tracker three-dimensional coordinate system to calculate a coordinate conversion relationship between the tracker three-dimensional coordinate system and the optical three-dimensional coordinate system; S804. According to the three-dimensional coordinate data of the top end and the bottom end of the calibration rod in the optical three-dimensional coordinate system, combining the coordinate conversion relationship between the tracker three-dimensional coordinate system and the optical three-dimensional coordinate system to analyze to obtain three-dimensional coordinate data of the top end and the bottom end of the calibration rod in the tracker three-dimensional coordinate system, and calculating a directional vector of the rotation axis of the implant handset in the tracker three-dimensional coordinate system; S805. Combining the structural parameters of the drill needle and the directional vector of the rotation axis of the implant handset in the tracker three-dimensional coordinate system to calculate three-dimensional coordinate data of the tip of the drill needle in the tracker three-dimensional coordinate system.

5. An oral implant surgery real-time navigation system, characterized by, The system is implemented based on attached mark points, at least three attached mark points are provided, the attached mark points are fixedly arranged on the labial side of the maxillary teeth and / or the mandibular teeth of a patient, the attached mark points are black resin round pieces containing barium sulfate developer, and the system comprises: A CT scanning module configured to control a CT scanning device to perform CT scanning on the oral cavity of a patient to obtain oral cavity CT three-dimensional data; An optical scanning module configured to control a binocular navigator to perform image scanning on the oral cavity of a patient to obtain oral cavity optical two-dimensional data; An automatic segmentation processing module configured to construct an oral cavity segmentation data processing model, train and optimize the oral cavity segmentation data processing model, segment the oral cavity CT three-dimensional data and the oral cavity optical two-dimensional data based on the trained and optimized oral cavity segmentation data processing model, segment the CT three-dimensional model into CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data, and segment the oral cavity optical two-dimensional data into optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data; The target region selection module is configured to select a target operation region based on a segmentation processing result of the oral cavity segmentation data processing model, select CT maxillary tooth three-dimensional data or CT mandibular tooth three-dimensional data as CT target operation data, and select optical maxillary tooth two-dimensional data or optical mandibular tooth two-dimensional data as optical target operation data; The CT data recognition module is configured to recognize three-dimensional coordinate data of the attached identification point in a CT three-dimensional coordinate system in the CT target operation data; The optical data recognition module is configured to recognize three-dimensional coordinate data of the attached identification point in an optical three-dimensional coordinate system in the optical target operation data according to the optical two-dimensional data of the oral cavity; The rigid body registration module is configured to perform rigid body registration operation according to the three-dimensional coordinate data of the attached identification point in the CT three-dimensional coordinate system and the three-dimensional coordinate data of the attached identification point in the optical three-dimensional coordinate system, and obtain a coordinate conversion relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system; The calibration processing module is configured to perform calibration processing on the implant handset, obtain image data of the handset tracker, analyze to obtain three-dimensional coordinate data of a direction vector of a drill needle rotating shaft on the handset tracker in a tracker three-dimensional coordinate system, and calculate three-dimensional coordinate data of a drill needle tip in the tracker three-dimensional coordinate system in combination with a structural parameter of the drill needle; The real-time navigation module is configured to control the binocular navigator to collect real-time image data of the handset tracker and the attached identification point, analyze to obtain real-time relative positions of the patient's oral cavity and the drill needle tip in the CT three-dimensional coordinate system in combination with the coordinate conversion relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system, and thus perform real-time navigation; The rigid body registration module includes: The three-dimensional point set acquisition unit is configured to acquire a CT three-dimensional point set of the attached marker point according to three-dimensional coordinate data of the attached marker point in a CT three-dimensional coordinate system acquire an optical three-dimensional point set of the attached marker point according to three-dimensional coordinate data of the attached marker point in an optical three-dimensional coordinate system wherein, the three-dimensional coordinate data of the attached marker point in the CT three-dimensional coordinate system, the three-dimensional coordinate data of the attached marker point in the optical three-dimensional coordinate system; Decentralized unit for obtaining the centroid of a CT three-dimensional point set of attached identification points Performing a decentralization operation to obtain the centroid of a CT three-dimensional point set Optical three-dimensional point set of attached identification points Performing a decentralization operation to obtain the centroid of an optical three-dimensional point set ; The center point set calculation unit is configured to calculate a CT three-dimensional center point set of the attached identification point The center point set calculation unit is configured to calculate a CT three-dimensional center point set of the attached identification point Wherein, , ; a rotation parameter calculation unit for constructing a covariance matrix wherein a rotation matrix is calculated by singular value decomposition of the covariance matrix ;​ A translation parameter calculation unit root for calculating a translation vector from a rotation matrix calculating a translation vector wherein ; The conversion relationship calculation unit is configured to obtain a coordinate conversion relationship between the CT three-dimensional coordinate system and the optical three-dimensional coordinate system according to the rotation matrix and the translation vector .

6. The real-time navigation system for oral implant surgery according to claim 5, wherein The automatic segmentation processing module includes: The basic data acquisition unit is configured to acquire a plurality of sets of oral cavity CT model basic data and a plurality of sets of oral cavity optical model basic data, label the oral cavity CT model basic data to obtain a CT maxillary tooth label model and a CT mandibular tooth label model, and label the oral cavity optical model basic data to obtain an optical maxillary tooth label model and an optical mandibular tooth label model; The data set division unit is configured to divide the oral cavity CT model basic data into a data set to obtain an oral cavity CT model segmentation training set, an oral cavity CT model segmentation verification set, and an oral cavity CT model segmentation test set, and divide the oral cavity optical model basic data into a data set to obtain an oral cavity optical model segmentation training set, an oral cavity optical model segmentation verification set, and an oral cavity optical model segmentation test set. a function construction 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; The training judgment unit is configured to take the SA-UNet model as a basic network architecture, input the oral CT model basic data of the oral CT model segmentation training set into the SA-UNet model, analyze the probability of each three-dimensional voxel in the oral CT model basic data belonging to the CT maxillary tooth label model and the CT mandibular tooth label model, and output the one with the highest probability as the CT model judgment result; take the YOLO model as a basic network architecture, input the oral optical model basic data of the oral optical model segmentation training set into the YOLO model, analyze the probability of each two-dimensional pixel in the oral optical model basic data belonging to the optical maxillary tooth label model and the optical mandibular tooth label model, and output the one with the highest probability as the optical model judgment result; The result comparison unit is configured to compare the CT model judgment result with the CT maxillary tooth label model and the CT mandibular tooth label model obtained by labeling, calculate the total loss coefficient corresponding to the CT model judgment result by using a combined loss function, and update the loss function parameter corresponding to the total loss coefficient; compare the optical model judgment result with the optical maxillary tooth label model and the optical mandibular tooth label model obtained by labeling, calculate the total loss coefficient corresponding to the optical model judgment result by using a combined loss function, and update the loss function parameter corresponding to the total loss coefficient; An iterative operation unit is configured to iterate the SA-UNet model after the updated loss function parameters are input to the oral CT model segmentation training set, and iterate until an oral CT model segmentation deep learning model is obtained; and the YOLO model after the updated loss function parameters are input to the oral optical model segmentation training set is iteratively operated until an oral optical model segmentation deep learning model is obtained. The automatic segmentation unit is configured to obtain oral CT three-dimensional data, input the oral CT three-dimensional data into the oral CT model segmentation deep learning model, analyze the oral CT three-dimensional data by using the oral CT model segmentation deep learning model, calculate the probability of each three-dimensional voxel in the oral CT three-dimensional data belonging to the CT maxillary tooth label model and the CT mandibular tooth label model, and determine that the three-dimensional voxel is CT maxillary tooth three-dimensional data when the probability of the three-dimensional voxel belonging to the CT maxillary tooth label model is the highest, and determine that the three-dimensional voxel is CT mandibular tooth three-dimensional data when the probability of the three-dimensional voxel belonging to the CT mandibular tooth label model is the highest; obtain oral optical two-dimensional data, input the oral optical two-dimensional data into the oral optical model segmentation deep learning model, analyze the oral optical two-dimensional data by using the oral optical model segmentation deep learning model, calculate the probability of each two-dimensional pixel in the oral optical two-dimensional data belonging to the optical maxillary tooth label model and the optical mandibular tooth label model, and determine that the two-dimensional pixel is optical maxillary tooth two-dimensional data when the probability of the two-dimensional pixel belonging to the optical maxillary tooth label model is the highest, and determine that the two-dimensional pixel is optical mandibular tooth two-dimensional data when the probability of the two-dimensional pixel belonging to the optical mandibular tooth label model is the highest; The model construction unit is configured to group each three-dimensional voxel in the oral CT three-dimensional data into corresponding CT maxillary tooth three-dimensional data and CT mandibular tooth three-dimensional data, and group each two-dimensional pixel in the oral optical two-dimensional data into corresponding optical maxillary tooth two-dimensional data and optical mandibular tooth two-dimensional data.

7. The real-time navigation system for oral implant surgery according to claim 5, wherein The CT data recognition module comprises: a voxel value calculation unit configured to obtain a three-dimensional voxel value according to oral cavity CT three-dimensional data analysis wherein the three-dimensional voxel value is a CT value of the voxel, and is a pixel two-dimensional coordinate value of the voxel on a CT two-dimensional slice, is a slice serial number of the CT two-dimensional slice. The voxel value screening unit is configured to determine whether a three-dimensional voxel value is a valid voxel value according to a preset CT threshold value, filter invalid voxel values according to a determination result, and output valid voxel values. The regional connection unit is configured to perform three-dimensional connected component labeling on the output effective voxel value to obtain a plurality of candidate three-dimensional marker point regions ; The orthogonal cutting unit is used for orthogonal cutting of the candidate three-dimensional mark point region to obtain an axial two-dimensional section, a sagittal two-dimensional section and a coronal two-dimensional section. The region extraction unit is configured to extract a two-dimensional voxel region on an axial two-dimensional section, a sagittal two-dimensional section, and a coronal two-dimensional section of the axis candidate three-dimensional landmark region ; A roundness calculation unit for obtaining a two-dimensional voxel region corresponding area A and perimeter P, calculating the roundness of a two-dimensional voxel region wherein ;​ A roundness judging unit is configured to screen A two-dimensional voxel region when the roundness is not less than 0.9 A two-dimensional voxel region corresponding to the attached marker The coordinate value acquisition unit acquires three-dimensional coordinate data of all N voxels on a two-dimensional voxel region of the attached identification point in a CT three-dimensional coordinate system wherein ; An average coordinate calculation unit is configured to calculate the three-dimensional coordinate data of the attached identification point in the CT three-dimensional coordinate system according to the three-dimensional coordinate data of all N voxels in the CT three-dimensional coordinate system wherein, , , .

8. 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 the processor, the processor executes the real-time navigation method for oral implant surgery in any one of claims 1-4.

Images