Intelligent alveolar bone increment design method and system based on bone defect analysis
Through intelligent design methods, the weighted model and dynamic adjustment of resolution coefficients are used to fuse multiple data sources to build an alveolar bone incremental design model, which solves the problems of prediction inaccurate and surgical risks in traditional design, and achieves higher accuracy and robustness.
Patent Information
- Application Number
- CN202410994788.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-15
AI Technical Summary
Traditional alveolar bone increment design relies on physician experience and manual measurements, and lacks comprehensive analysis of multiple data sources, resulting in inaccurate prediction results, increased surgical risks, and failed to dynamically adjust the resolution coefficient, affecting the flexibility and accuracy of the model.
Using an intelligent design method based on bone defect analysis, bone defect parameters, adjacent dental environment and three-dimensional image data are collected, and the resolution coefficient is adjusted dynamically, and the implant position error prediction model and three-dimensional spatial model are constructed, and multiple data sources are fused. Bone incremental scheme.
It improves the accuracy and reliability of predictions, reduces surgical errors, enhances the robustness of the model, can better adapt to the characteristics of different data sets, and provides more accurate surgical solutions.
Smart Images

Figure CN118967471A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of alveolar bone regeneration, and more specifically, to an intelligent alveolar bone increment design method and system based on bone defect analysis. Background Art
[0002] The development of modern oral restoration technology has provided people with more possibilities to keep their own teeth. However, due to various diseases, trauma and other reasons, patients' alveolar bones often suffer from varying degrees of defects, which brings challenges to subsequent dental implant treatment. Traditional evaluation methods are time-consuming and rely on high-level experts, which limits their good application in a wide range of clinical settings.
[0003] The patent with application publication number CN113520638A discloses a design method of titanium mesh for alveolar bone augmentation, including: collecting the original data of the user's jaw position, importing the original data into medical image data processing software to reconstruct the jaw model; importing the jaw model into forward and reverse hybrid design software, and performing corresponding preliminary processing on the jaw model; importing the jaw model after preliminary processing into finite element analysis software for force analysis to obtain the force deformation diagram of the jaw model; importing the force deformation diagram into engineering design software, and designing a hyperbolic paraboloid titanium mesh corresponding to the position with the largest force deformation, designing retention nail holes in the area with small force deformation, and obtaining the corresponding titanium mesh design diagram. The present invention also discloses a manufacturing method of titanium mesh for alveolar bone augmentation. The present invention can use software to design titanium mesh so that the titanium mesh has a better shape.
[0004] Traditional alveolar bone augmentation design mainly relies on the doctor's experience and manual measurement, and has the following main problems:
[0005] Without integrating information from multiple data sources, the prediction model may not be able to capture all factors that affect implant position errors; this will cause the prediction results to deviate from the actual situation and reduce the prediction accuracy; if predictions are made based on a single data source, the robustness of the model will be seriously challenged; in the medical field, such deviations may increase surgical risks and affect patient recovery outcomes;
[0006] Directly using a fixed resolution coefficient for correlation calculation will exaggerate or reduce the impact of certain data sets on the results, thus affecting the accuracy and reliability of the model; failure to dynamically adjust the resolution coefficient means that the model lacks flexibility when facing different data sets; inaccurate correlation calculation and reduced model robustness will directly affect the accuracy of the prediction; large deviations from the actual situation will increase surgical risks and may lead to adverse consequences;
[0007] Without the support of a three-dimensional spatial model, the prediction model may not be able to fully consider the size, shape, position of the bone defect area and the impact of the adjacent tooth environment on the implant position. This will lead to inaccurate prediction results and fail to provide doctors with a reliable reference basis; due to inaccurate prediction results, doctors face greater errors and uncertainties during surgery; and may increase the risk of surgery, such as implant position deviation, adjacent tooth damage, etc., which in turn affects the surgical effect and patient recovery.
[0008] In view of this, the present invention proposes an intelligent alveolar bone augmentation design method and system based on bone defect analysis to solve the above problems. Summary of the invention
[0009] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: an intelligent alveolar bone augmentation design method based on bone defect analysis, comprising:
[0010] S1, collecting bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data;
[0011] S2, preprocessing the bone defect parameter data, the adjacent tooth environment data and the alveolar bone three-dimensional image data to obtain a bone defect feature data set, an adjacent tooth environment feature data set and a three-dimensional image feature data set; correlating and fusing the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set to obtain a comprehensive feature data set;
[0012] S3, constructing an implant position error prediction model, inputting the comprehensive feature data set into the implant position error prediction model, and predicting the implant position error;
[0013] S4, comparing the predicted implant position error with a preset implant position error threshold to determine whether the implant position is an optimal position;
[0014] S5. If the implant position is the optimal position, construct a three-dimensional space model between the implant, the bone defect feature data set, and the adjacent tooth environment feature data set, and process the three-dimensional space model to obtain three-dimensional space feature data; construct a defect type diagnosis model, which is used to input the three-dimensional space feature data into the defect type diagnosis model to predict the bone defect type data;
[0015] S6. The target bone augmentation morphology is designed according to the defect type data through the bone augmentation design algorithm, and the bone augmentation plan is intelligently recommended based on the designed target bone augmentation morphology.
[0016] Furthermore, the bone defect parameter data includes the length, width, depth, two-dimensional projection area of the defect area, three-dimensional volume of the defect area and bone defect position data; the adjacent tooth environment data includes the thickness, height, density, adjacent tooth position data and adjacent tooth health status data of the adjacent tooth bone; the alveolar bone three-dimensional image data includes alveolar bone anatomical structure image data, alveolar bone and tooth position relationship data; the alveolar bone anatomical structure image data includes alveolar bone height, depth and density distribution data; the alveolar bone and tooth position relationship data includes the embedding depth and inclination angle of the teeth in the alveolar bone.
[0017] Furthermore, the method of preprocessing the bone defect parameter data, the adjacent tooth environment data and the alveolar bone three-dimensional image data to obtain the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set includes:
[0018] The box plot method is used to detect and eliminate abnormal data in the bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data to obtain the processed bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set. The bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set are normalized and converted into standard normal distribution to obtain the normalized bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set.
[0019] Furthermore, the method of performing correlation processing on the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set to obtain a comprehensive feature data set includes:
[0020] The bone defect feature dataset, the adjacent tooth environment feature dataset and the three-dimensional image feature dataset are associated and fused through a weighted model to form a comprehensive feature dataset; the bone defect feature dataset is recorded as Q1, the adjacent tooth environment feature dataset is recorded as Q2, and the three-dimensional image feature dataset is recorded as Q3;
[0021] The weighted model is: GW=Q1·ω1+Q2·ω2+Q3·ω3; wherein ω1 is the weight coefficient of the bone defect feature data set; ω2 is the weight coefficient of the adjacent tooth environment feature data set; ω3 is the weight coefficient of the three-dimensional image feature data set;
[0022] The weight coefficient of each feature data set in the weighted model is designed: the reference data set is preset as Q0, the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set are used as samples, and the samples are normalized by the normalization formula; the normalization formula is: Among them, Q i′ is the i′th feature data set; Q′ i′ is the normalized i′th feature data set; min(Qi′ ) is the minimum value in the i′th feature data set; max(Q i′ ) is the maximum value in the i′th feature data set;
[0023] For each feature data set, calculate its correlation with the reference data set, and determine the weight coefficient of each feature data set based on the correlation; the calculation formula of the weight coefficient is: Among them, ω i′ is the weight coefficient of the i′th feature data set; i′ is the correlation degree of the i′th feature data set;
[0024] The correlation degree of the feature data set is: Where n is the number of samples; ξ i′ (k) is the correlation coefficient of the i′th feature data set in the kth sample; k is the type of feature data set;
[0025] The correlation coefficient between each feature data set and the reference data set is calculated using the correlation coefficient formula; the correlation coefficient formula is:
[0026] Where Q0(k) is the value of the kth sample in the reference data set Q0, which is usually 1; Q i ″(k) is the normalized value of the i′th feature data set at the kth sample; ρ is the resolution coefficient, which usually takes a value between 0 and 1; It is the smallest absolute difference among all feature data sets; It is the resolution coefficient multiplied by the maximum absolute difference among all feature data sets; Q0(k)-Q i ″(k) is the absolute difference between the current feature dataset and the reference dataset;
[0027] The resolution coefficient in the correlation coefficient formula is restricted by a dynamic adjustment model, and the dynamic adjustment model is: Among them, SR is the type of all feature data sets; is the mean of the standard deviations of all feature data sets, indicating the average degree of dispersion of the overall feature data set; σ max is the maximum standard deviation of all feature data sets, indicating the most discrete feature in the feature data set.
[0028] Furthermore, the training method of the implant position error prediction model includes:
[0029] The data set is divided into a training set, a validation set and a test set; an implant position error prediction model is constructed, and the implant position error prediction model includes input data and output labels; the input data of the model is a historical comprehensive feature data set, and the output label of the model is an implant position error; RBF is selected as the kernel function of the model, and the hyperparameters of the model are initialized, and the hyperparameters of the model include a regularization parameter C and a γ value of the RBF kernel function; the implant position error prediction model is a support vector machine model;
[0030] The ε-insensitive loss function is used as the loss function of the model to measure the difference between the predicted value and the true value of the model; the ε-insensitive loss function is:
[0031] Where y is the true value of the implant position error; f(x) is the predicted value of the implant position error; ε is the width of the insensitive interval, indicating that errors within this range are not counted as losses;
[0032] The model is trained on the data set, the SMO algorithm is used to solve the optimization problem of ε-SVR, and the hyperparameter combination is adjusted through cross-validation based on the validation set data to find the hyperparameter that minimizes the loss function on the validation set;
[0033] The model is evaluated using the test data set. The model performance is evaluated by observing the performance indicators of the model on the prediction task. The test is stopped when the model performance reaches the preset performance threshold to obtain the implant position error prediction model. The trained implant position error prediction model is used to predict the current feature data set to obtain the implant position error.
[0034] Furthermore, the method of comparing the predicted implant position error with a preset implant position error threshold to determine whether the implant position is the optimal position includes:
[0035] If the predicted implant position error is less than a preset implant position error threshold, the implant position is determined to be the optimal position;
[0036] If the predicted implant placement error is greater than or equal to a preset implant placement error threshold, it is determined that the implant placement is not the optimal placement.
[0037] Furthermore, the method of constructing a three-dimensional space model between the implant, bone defect feature data set and the adjacent tooth environment feature data set, and processing the three-dimensional space model to obtain the three-dimensional space feature data includes:
[0038] The image segmentation algorithm is used to identify the bone defect area in the three-dimensional image data of the alveolar bone, and the boundary of the bone defect area is extracted by threshold segmentation and edge detection algorithm; the segmentation threshold is preset to L, and each pixel in the image is compared with the threshold L: Among them, g′(a,b) is the segmented image; g(a,b) is each pixel in the image; 255 is the bone defect area; 0 is the background area;
[0039] The segmented image is used to extract the boundary of the bone defect area using the Canny edge detector; the pixel set in the defect area is preset as {(a i″ ,b i″ )}, the number of pixels in the bone defect area is: Among them, N p is the number of pixels in the bone defect area; g′(a i″ ,b i″ ) is the set of image pixels after segmentation; R is a set of real numbers;
[0040] The actual area of each pixel is assumed to be A. p , the number of pixels in the bone defect area N p Multiply by the actual area of each pixel A p , and obtain the two-dimensional projection area A of the bone defect area 2D ; Use the 3D reconstruction algorithm to build a 3D space, select each voxel in the bone defect area in the 3D space, and the voxel is the smallest unit in the 3D space; multiply the number of voxels in all bone defect areas by the volume of each voxel to obtain the total volume V of the bone defect area 3D ;
[0041] Fusion of bone defect feature dataset and adjacent tooth environment feature dataset into a unified 3D coordinate system in, is the spatial position coordinate of the feature data point; is the characteristic data of the position coordinates;
[0042] Build a 3D spatial model and map all feature data points into a grid; subdivide the grid into voxels, each of which contains all feature data information at a specific location; for sparse data in the 3D spatial model, use an interpolation algorithm to fill in missing data and standardize different types of data;
[0043] Extract geometric features from the three-dimensional model; the geometric features include the two-dimensional projection area A of the bone defect area 2D and the total volume of the bone defect area V 3D ; Calculate the statistical characteristics that describe the distribution of all feature data points; statistical characteristics include mean and standard deviation;
[0044] The spatial relationship features between the bone defect feature data set and the adjacent tooth environment feature data set are calculated by the nearest neighbor search algorithm; the spatial relationship features include the minimum distance between all points on the surface of the bone defect area and all points on the surface of the adjacent teeth; the relative position relationship between the bone defect area and the adjacent teeth in three-dimensional space;
[0045] Calculate the point u(x u ,y u ,z u ) and the point w(x w ,y w ,z w ) between the relative position vector: r = (x u -x w ,y u -y w ,z u -z w ), where r is the relative position vector between each point on the surface of the bone defect area and the point on the surface of the adjacent tooth;
[0046] The extracted geometric features, statistical features and spatial relationship features are integrated into a feature vector. Each feature data point can be represented as a vector with multi-dimensional features to obtain three-dimensional spatial feature data.
[0047] Furthermore, the training method of the defect type diagnosis model includes:
[0048] The data set is divided into a training set, a validation set and a test set; a defect type diagnosis model is constructed, and the defect type diagnosis model includes an input layer, a convolution layer, a pooling layer, a batch normalization layer, a fully connected layer and an output layer; the input layer is the historical three-dimensional spatial feature data, and the output layer is the bone defect type data; softmax is used as the activation function; the defect type diagnosis model is a 3D convolutional neural network model;
[0049] Multi-classification cross entropy is used as the loss function of the model to measure the difference between the predicted value and the actual value of the model; the multi-classification cross entropy loss function is: Where L is the average loss of the data set; N is the total number of samples in the data set; B is the number of bone defect type data; y ib is the true label of the i-th sample for the b-th type; y ib When it is 0, it means that the sample does not belong to the bth category, y ib When it is 1, it means that the sample belongs to the bth category; ib The probability that the i-th sample belongs to the b-th type predicted by the model;
[0050] Use the training set data to train the model and use the Adam optimizer to minimize the loss function. Use the validation set to evaluate the performance of the defect type diagnosis model, calculate the accuracy index to measure the performance of the model, and tune the model's hyperparameters to improve the model performance.
[0051] The performance of the model in the prediction task is evaluated through the test set. When the performance of the model in the prediction task reaches the preset performance threshold, the test is stopped to obtain the defect type diagnosis model. The trained defect type diagnosis model is used to predict the current three-dimensional spatial feature data to obtain the bone defect type data.
[0052] Furthermore, the method of designing the target bone augmentation morphology according to the defect type data by using the bone augmentation design algorithm and intelligently recommending the bone augmentation scheme according to the designed target bone augmentation morphology includes:
[0053] The obtained three-dimensional spatial feature data is recorded as J, and the three-dimensional spatial feature data J is standardized: Among them, J OR is the standardized three-dimensional spatial feature data; μ′ is the mean of the feature data; σ′ is the standard deviation of the feature data;
[0054] The processed three-dimensional spatial feature data is input into the bone augmentation design algorithm; the defect type data is divided into m types to form a defect feature data set {H1, H2, ..., H m}; where m is the type of defect; for each defect type H q , use the generative adversarial network to generate the target bone augmentation morphology, predefine the corresponding bone augmentation morphology rules, the bone augmentation morphology rules are: in, The target bone augmentation morphology for design; Designing rule functions for bone augmentation morphology;
[0055] Genetic algorithm is used to optimize the design of target bone augmentation morphology, and the objective function F that reflects the applicability of bone augmentation morphology is defined; the target bone augmentation morphology is optimized according to the defect type data:
[0056] in, is the optimized target bone augmentation shape; argmax is the value that maximizes the objective function F value;
[0057] A database containing various bone augmentation schemes was established, each of which included a detailed description and applicable conditions. The similarity between the designed target bone augmentation morphology and each scheme in the database was calculated using the cosine similarity function, and the scheme with the highest similarity was selected as the recommended scheme. Based on the designed target bone augmentation morphology, the optimal bone augmentation scheme was recommended from the database.
[0058] An intelligent alveolar bone augmentation design system based on bone defect analysis, comprising:
[0059] A data acquisition module, used to collect bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data;
[0060] A data processing module is used to pre-process the bone defect parameter data, the adjacent tooth environment data and the alveolar bone three-dimensional image data to obtain a bone defect feature data set, an adjacent tooth environment feature data set and a three-dimensional image feature data set; and to associate and fuse the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set to obtain a comprehensive feature data set;
[0061] An implant error prediction module is used to construct an implant position error prediction model, input the comprehensive feature data set into the implant position error prediction model, and predict the implant position error;
[0062] An implant position evaluation module is used to compare the predicted implant position error with a preset implant position error threshold to determine whether the implant position is the optimal position;
[0063] The defect type judgment module, if the implant position is the optimal position, constructs a three-dimensional space model between the implant, the bone defect feature data set and the adjacent tooth environment feature data set, processes the three-dimensional space model to obtain three-dimensional space feature data; constructs a defect type diagnosis model, which is used to input the three-dimensional space feature data into the defect type diagnosis model to predict the bone defect type data;
[0064] The bone augmentation intelligent recommendation module is used to design the target bone augmentation morphology according to the defect type characteristic data through the bone augmentation design algorithm, and intelligently recommend the bone augmentation plan according to the designed target bone augmentation morphology; each module is connected by wired and / or wireless means.
[0065] Technical effects and advantages of the intelligent alveolar bone augmentation design method and system based on bone defect analysis of the present invention:
[0066] Through the weighted model, the bone defect feature dataset, the adjacent tooth environment feature dataset and the three-dimensional image feature dataset are effectively integrated together. The weight coefficient of each dataset is determined based on its correlation with the reference dataset, ensuring that the comprehensive feature dataset can more comprehensively reflect the actual situation and improve the accuracy and reliability of subsequent analysis or prediction.
[0067] Dynamically adjust the resolution coefficient in the model's restricted correlation coefficient formula, taking into account the discreteness of all feature data sets; by adjusting the resolution coefficient, the differences between different feature data sets can be balanced, making the calculation of correlation more reasonable and stable; improve the robustness of the model so that it can better adapt to the characteristics of different data sets; by fusing feature data sets from multiple sources, the implant position error prediction model can obtain more comprehensive and richer information; help the model more accurately capture the key factors affecting implant position error, thereby improving the accuracy of prediction;
[0068] By constructing a three-dimensional spatial model, the spatial relationship between the implant, bone defect feature data set and the adjacent tooth environment feature data set can be accurately represented, and the size, shape, position of the bone defect area and the influence of the adjacent tooth environment can be comprehensively considered to make more accurate predictions; it helps to reduce errors and uncertainties in surgery; this visualization capability helps doctors better understand the condition and develop more accurate surgical plans. BRIEF DESCRIPTION OF THE DRAWINGS
[0069] Figure 1 A schematic diagram of a flow chart of an intelligent alveolar bone augmentation design method based on bone defect analysis of the present invention;
[0070] Figure 2 The figure is a schematic diagram of the structure of an intelligent alveolar bone augmentation design system based on bone defect analysis of the present invention. DETAILED DESCRIPTION
[0071] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0072] Example 1
[0073] See also Figure 1 As shown, this embodiment provides an intelligent alveolar bone augmentation design method based on bone defect analysis, including:
[0074] S1, collecting bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data;
[0075] S2, preprocessing the bone defect parameter data, the adjacent tooth environment data and the alveolar bone three-dimensional image data to obtain a bone defect feature data set, an adjacent tooth environment feature data set and a three-dimensional image feature data set; correlating and fusing the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set to obtain a comprehensive feature data set;
[0076] S3, constructing an implant position error prediction model, inputting the comprehensive feature data set into the implant position error prediction model, and predicting the implant position error;
[0077] S4, comparing the predicted implant position error with a preset implant position error threshold to determine whether the implant position is an optimal position;
[0078] S5. If the implant position is the optimal position, construct a three-dimensional space model between the implant, the bone defect feature data set, and the adjacent tooth environment feature data set, and process the three-dimensional space model to obtain three-dimensional space feature data; construct a defect type diagnosis model, which is used to input the three-dimensional space feature data into the defect type diagnosis model to predict the bone defect type data;
[0079] S6. The target bone augmentation morphology is designed according to the defect type data through the bone augmentation design algorithm, and the bone augmentation plan is intelligently recommended based on the designed target bone augmentation morphology.
[0080] The bone defect parameter data include the length, width, depth, two-dimensional projection area of the defect area, three-dimensional volume of the defect area and bone defect position data; the adjacent tooth environment data include the thickness, height, density, adjacent tooth position data and adjacent tooth health status data of the adjacent tooth bone; the alveolar bone three-dimensional image data include the alveolar bone anatomical structure image data and the alveolar bone and tooth position relationship data; the alveolar bone anatomical structure image data include the alveolar bone height, depth and density distribution data; the alveolar bone and tooth position relationship data include the embedding depth and inclination angle of the teeth in the alveolar bone.
[0081] The length, width and depth of the bone defect are obtained through CT images; the two-dimensional projection area and three-dimensional volume of the defect area are obtained through three-dimensional reconstruction software; the bone defect position data is obtained through CT images; the thickness, height, density, adjacent tooth position data and adjacent tooth health status data of the adjacent tooth are obtained through CT scanning; the alveolar bone anatomical structure image data and the alveolar bone and tooth position relationship data are obtained through CBCT scanning and three-dimensional reconstruction software;
[0082] The method of preprocessing bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data to obtain a bone defect feature data set, an adjacent tooth environment feature data set and a three-dimensional image feature data set includes:
[0083] The box plot method is used to detect and eliminate abnormal data in the bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data to obtain the processed bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set. The bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set are normalized and converted into standard normal distribution to obtain the normalized bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set.
[0084] The method of performing correlation processing on the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set to obtain a comprehensive feature data set includes:
[0085] The bone defect feature dataset, the adjacent tooth environment feature dataset and the three-dimensional image feature dataset are associated and fused through a weighted model to form a comprehensive feature dataset; the bone defect feature dataset is recorded as Q1, the adjacent tooth environment feature dataset is recorded as Q2, and the three-dimensional image feature dataset is recorded as Q3;
[0086] The weighted model is: GW=Q1·ω1+Q2·ω2+Q3·ω3; wherein ω1 is the weight coefficient of the bone defect feature data set; ω2 is the weight coefficient of the adjacent tooth environment feature data set; ω3 is the weight coefficient of the three-dimensional image feature data set;
[0087] The weight coefficient of each feature data set in the weighted model is designed: the reference data set is preset as Q0, the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set are used as samples, and the samples are normalized by the normalization formula; the normalization formula is: Among them, Q i′ is the i′th feature data set; Q′ i′ is the normalized i′th feature data set; min(Q i′ ) is the minimum value in the i′th feature data set; max(Q i′ ) is the maximum value in the i′th feature data set;
[0088] For each feature data set, calculate its correlation with the reference data set, and determine the weight coefficient of each feature data set based on the correlation; the calculation formula of the weight coefficient is: Among them, ω i′ is the weight coefficient of the i′th feature data set; i′ is the correlation degree of the i′th feature data set;
[0089] The correlation degree of the feature data set is: Where n is the number of samples; ξ i′ (k) is the correlation coefficient of the i′th feature data set in the kth sample; k is the type of feature data set;
[0090] The correlation coefficient between each feature data set and the reference data set is calculated using the correlation coefficient formula; the correlation coefficient formula is:
[0091] Where Q0(k) is the value of the kth sample in the reference data set Q0, which is usually 1; Q i ″(k) is the normalized value of the i′th feature data set at the kth sample; ρ is the resolution coefficient, which usually takes a value between 0 and 1; It is the smallest absolute difference among all feature data sets; It is the resolution coefficient multiplied by the maximum absolute difference among all feature data sets; |Q0(k)-Q′ i′ (k)| is the absolute difference between the current feature dataset and the reference dataset;
[0092] The resolution coefficient in the correlation coefficient formula is restricted by a dynamic adjustment model, and the dynamic adjustment model is: Among them, SR is the type of all feature data sets; is the mean of the standard deviations of all feature data sets, indicating the average degree of dispersion of the overall feature data set; σ max is the maximum standard deviation of all feature data sets, indicating the most discrete feature in the feature data set;
[0093] By dynamically adjusting the model's restrictions on the resolution coefficient, if the discreteness of all feature data sets is smaller, the resolution coefficient is smaller; the larger the discreteness of all feature data sets, the larger the resolution coefficient; the resolution coefficient will be dynamically adjusted according to the diversity of the feature data sets, thereby adjusting the correlation coefficient between each feature data set and the reference data set, further adjusting the correlation between each feature data set and the reference data set, and dynamically adjusting the weight coefficient of each feature data set according to the correlation.
[0094] For example, suppose you have the following data:
[0095] Reference data: 1,1,1; first feature data: 0.8,0.9,0.7; second feature data: 0.6,0.7,0.6; calculate the correlation coefficient of the first feature data in the first sample:
[0096] |1-0.8|=0.2; Calculate the minimum and maximum values of all absolute differences: min(0.2,0.1,0.3,0.4,0.3,0.4)=0.1; max(0.2,0.1,0.3,0.4,0.3,0.4)=0.4; Assume ρ=0.5, then:
[0097] The training method of the implant position error prediction model includes:
[0098] The data set is divided into a training set, a validation set and a test set; an implant position error prediction model is constructed, and the implant position error prediction model includes input data and output labels; the input data of the model is a historical comprehensive feature data set, and the output label of the model is an implant position error; RBF is selected as the kernel function of the model, and the hyperparameters of the model are initialized, and the hyperparameters of the model include a regularization parameter C and a γ value of the RBF kernel function; the implant position error prediction model is a support vector machine model;
[0099] The ε-insensitive loss function is used as the loss function of the model to measure the difference between the predicted value and the true value of the model; the ε-insensitive loss function is:
[0100] Where y is the true value of the implant position error; f(x) is the predicted value of the implant position error; ε is the width of the insensitive interval, indicating that errors within this range are not counted as losses;
[0101] The model is trained on the data set, the SMO algorithm is used to solve the optimization problem of ε-SVR, and the hyperparameter combination is adjusted through cross-validation based on the validation set data to find the hyperparameter that minimizes the loss function on the validation set;
[0102] The model is evaluated using the test data set. The model performance is evaluated by observing the performance indicators of the model on the prediction task. The test is stopped when the model performance reaches the preset performance threshold to obtain the implant position error prediction model. The trained implant position error prediction model is used to predict the current feature data set to obtain the implant position error.
[0103] The method of comparing the predicted implant placement error with a preset implant placement error threshold to determine whether the implant placement is an optimal placement includes:
[0104] If the predicted implant position error is less than a preset implant position error threshold, the implant position is determined to be the optimal position;
[0105] If the predicted implant placement error is greater than or equal to a preset implant placement error threshold, it is determined that the implant placement is not the optimal placement.
[0106] A method for constructing a three-dimensional space model between an implant, a bone defect feature data set and an adjacent tooth environment feature data set, and processing the three-dimensional space model to obtain three-dimensional space feature data includes:
[0107] The image segmentation algorithm is used to identify the bone defect area in the three-dimensional image data of the alveolar bone, and the boundary of the bone defect area is extracted by threshold segmentation and edge detection algorithm; the segmentation threshold is preset to L, and each pixel in the image is compared with the threshold L: Among them, g′(a,b) is the segmented image; g(a,b) is each pixel in the image; 255 is the bone defect area; 0 is the background area;
[0108] The segmented image is used to extract the boundary of the bone defect area using the Canny edge detector; the pixel set in the defect area is preset as {(a i″ ,b i″ )}, the number of pixels in the bone defect area is: Among them, N p is the number of pixels in the bone defect area; g′(a i″ ,b i″ ) is the set of image pixels after segmentation; R is a set of real numbers;
[0109] The actual area of each pixel is assumed to be A. p , the number of pixels in the bone defect area N p Multiply by the actual area of each pixel A p , and obtain the two-dimensional projection area A of the bone defect area 2D ; Use the 3D reconstruction algorithm to build a 3D space, select each voxel in the bone defect area in the 3D space, and the voxel is the smallest unit in the 3D space; multiply the number of voxels in all bone defect areas by the volume of each voxel to obtain the total volume V of the bone defect area 3D ;
[0110] Fusion of bone defect feature dataset and adjacent tooth environment feature dataset into a unified 3D coordinate system in, is the spatial position coordinate of the feature data point; is the characteristic data of the position coordinates;
[0111] Build a 3D spatial model and map all feature data points into a grid; subdivide the grid into voxels, each of which contains all feature data information at a specific location; for sparse data in the 3D spatial model, use an interpolation algorithm to fill in missing data and standardize different types of data;
[0112] Extract geometric features from the three-dimensional model; the geometric features include the two-dimensional projection area A of the bone defect area 2D and the total volume of the bone defect area V 3D ; Calculate the statistical characteristics that describe the distribution of all feature data points; statistical characteristics include mean and standard deviation;
[0113] The spatial relationship features between the bone defect feature data set and the adjacent tooth environment feature data set are calculated by the nearest neighbor search algorithm; the spatial relationship features include the minimum distance between all points on the surface of the bone defect area and all points on the surface of the adjacent teeth; the relative position relationship between the bone defect area and the adjacent teeth in three-dimensional space;
[0114] Calculate the point u(x u ,y u ,z u ) and the point w(x w ,y w ,z w ) between the relative position vector: r = (x u -x w ,y u -y w ,z u -z w ), where r is the relative position vector between each point on the surface of the bone defect area and the point on the surface of the adjacent tooth;
[0115] The extracted geometric features, statistical features and spatial relationship features are integrated into a feature vector. Each feature data point can be represented as a vector with multi-dimensional features to obtain three-dimensional spatial feature data.
[0116] For example, suppose you have the following data:
[0117] Bone defect position: (1,2,3); adjacent tooth position: (4,5,6); bone defect length: 10mm; bone defect width: 5mm; bone defect depth: 3mm; two-dimensional projection area of defect area: 50mm 2 ; Three-dimensional volume of defect area: 150mm 3 ; Adjacent bone thickness: 2mm; Adjacent bone height: 8mm; Adjacent bone density: 1200HU; Through the above steps and feature extraction, we can obtain the following feature vectors: [10,5,3,50,150,2,8,1200,…].
[0118] The training method of the defect type diagnosis model includes:
[0119] The data set is divided into a training set, a validation set and a test set; a defect type diagnosis model is constructed, and the defect type diagnosis model includes an input layer, a convolution layer, a pooling layer, a batch normalization layer, a fully connected layer and an output layer; the input layer is the historical three-dimensional spatial feature data, and the output layer is the bone defect type data; softmax is used as the activation function; the defect type diagnosis model is a 3D convolutional neural network model;
[0120] Multi-classification cross entropy is used as the loss function of the model to measure the difference between the predicted value and the actual value of the model; the multi-classification cross entropy loss function is: Where L is the average loss of the data set; N is the total number of samples in the data set; B is the number of bone defect type data; y ib is the true label of the i-th sample for the b-th type; y ib When it is 0, it means that the sample does not belong to the bth category, y ib When it is 1, it means that the sample belongs to the bth category; ib The probability that the i-th sample belongs to the b-th type predicted by the model;
[0121] Use the training set data to train the model and use the Adam optimizer to minimize the loss function. Use the validation set to evaluate the performance of the defect type diagnosis model, calculate the accuracy index to measure the performance of the model, and tune the model's hyperparameters to improve the model performance.
[0122] The performance of the model in the prediction task is evaluated through the test set. When the performance of the model in the prediction task reaches the preset performance threshold, the test is stopped to obtain the defect type diagnosis model. The trained defect type diagnosis model is used to predict the current three-dimensional spatial feature data to obtain the bone defect type data.
[0123] The target bone augmentation morphology is designed according to the defect type data through the bone augmentation design algorithm. According to the designed target bone augmentation morphology, the method of intelligently recommending the bone augmentation plan includes:
[0124] The obtained three-dimensional spatial feature data is recorded as J, and the three-dimensional spatial feature data J is standardized: Among them, J OR is the standardized three-dimensional spatial feature data; μ′ is the mean of the feature data; σ′ is the standard deviation of the feature data;
[0125] The processed three-dimensional spatial feature data is input into the bone augmentation design algorithm; the defect type data is divided into m types to form a defect feature data set {H1, H2, ..., H m}; where m is the type of defect; for each defect type H q , use the generative adversarial network to generate the target bone augmentation morphology, predefine the corresponding bone augmentation morphology rules, the bone augmentation morphology rules are: in, The target bone augmentation morphology for design; Designing rule functions for bone augmentation morphology;
[0126] Genetic algorithm is used to optimize the design of target bone augmentation morphology, and the objective function F that reflects the applicability of bone augmentation morphology is defined; the target bone augmentation morphology is optimized according to the defect type data:
[0127] in, is the optimized target bone augmentation shape; argmax is the value that maximizes the objective function F value;
[0128] A database containing various bone augmentation schemes was established, each of which included a detailed description and applicable conditions. The similarity between the designed target bone augmentation morphology and each scheme in the database was calculated using the cosine similarity function, and the scheme with the highest similarity was selected as the recommended scheme. Based on the designed target bone augmentation morphology, the optimal bone augmentation scheme was recommended from the database.
[0129] The preset implant position error threshold is set by the staff, and different implant position errors are collected through the alveolar bone intelligent management terminal, and the average value of multiple implant position errors is taken as the preset implant position error threshold; the preset performance threshold and performance threshold are set similarly;
[0130] In this embodiment, the bone defect feature dataset, the adjacent tooth environment feature dataset and the three-dimensional image feature dataset are effectively integrated together through a weighted model, and the weight coefficient of each dataset is determined based on its correlation with the reference dataset, ensuring that the comprehensive feature dataset can more comprehensively reflect the actual situation and improve the accuracy and reliability of subsequent analysis or prediction;
[0131] Dynamically adjust the resolution coefficient in the model's restricted correlation coefficient formula, taking into account the discreteness of all feature data sets; by adjusting the resolution coefficient, the differences between different feature data sets can be balanced, making the calculation of correlation more reasonable and stable; improve the robustness of the model so that it can better adapt to the characteristics of different data sets; by fusing feature data sets from multiple sources, the implant position error prediction model can obtain more comprehensive and richer information; help the model more accurately capture the key factors affecting implant position error, thereby improving the accuracy of prediction;
[0132] By constructing a three-dimensional spatial model, the spatial relationship between the implant, bone defect feature data set and the adjacent tooth environment feature data set can be accurately represented, and the size, shape, position of the bone defect area and the influence of the adjacent tooth environment can be comprehensively considered to make more accurate predictions; it helps to reduce errors and uncertainties in surgery; this visualization capability helps doctors better understand the condition and develop more accurate surgical plans.
[0133] Example 2
[0134] See also Figure 2 As shown, this embodiment provides an intelligent alveolar bone augmentation design system based on bone defect analysis, including:
[0135] A data acquisition module, used to collect bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data;
[0136] A data processing module is used to pre-process the bone defect parameter data, the adjacent tooth environment data and the alveolar bone three-dimensional image data to obtain a bone defect feature data set, an adjacent tooth environment feature data set and a three-dimensional image feature data set; and to associate and fuse the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set to obtain a comprehensive feature data set;
[0137] An implant error prediction module is used to construct an implant position error prediction model, input the comprehensive feature data set into the implant position error prediction model, and predict the implant position error;
[0138] An implant position evaluation module is used to compare the predicted implant position error with a preset implant position error threshold to determine whether the implant position is the optimal position;
[0139] The defect type judgment module, if the implant position is the optimal position, constructs a three-dimensional space model between the implant, the bone defect feature data set and the adjacent tooth environment feature data set, processes the three-dimensional space model to obtain three-dimensional space feature data; constructs a defect type diagnosis model, which is used to input the three-dimensional space feature data into the defect type diagnosis model to predict the bone defect type data;
[0140] The bone augmentation intelligent recommendation module is used to design the target bone augmentation morphology according to the defect type characteristic data through the bone augmentation design algorithm, and intelligently recommend the bone augmentation plan according to the designed target bone augmentation morphology.
[0141] Since the electronic device introduced in this embodiment is an electronic device used to implement the intelligent alveolar bone augmentation design method and system based on bone defect analysis in the embodiment of this application, based on the intelligent alveolar bone augmentation design method and system based on bone defect analysis introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used by the intelligent alveolar bone augmentation design method and system based on bone defect analysis in the embodiment of this application, it belongs to the scope of protection of this application.
[0142] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.
[0143] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.
Claims
1. An intelligent alveolar bone augmentation design method based on bone defect analysis, characterized in that: include: S1, collecting bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data; S2, preprocessing the bone defect parameter data, the adjacent tooth environment data and the alveolar bone three-dimensional image data to obtain a bone defect feature data set, an adjacent tooth environment feature data set and a three-dimensional image feature data set; correlating and fusing the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set to obtain a comprehensive feature data set; S3, constructing an implant position error prediction model, inputting the comprehensive feature data set into the implant position error prediction model, and predicting the implant position error; S4, comparing the predicted implant position error with a preset implant position error threshold to determine whether the implant position is an optimal position; S5. If the implant position is the optimal position, construct a three-dimensional space model between the implant, the bone defect feature data set, and the adjacent tooth environment feature data set, and process the three-dimensional space model to obtain three-dimensional space feature data; construct a defect type diagnosis model, which is used to input the three-dimensional space feature data into the defect type diagnosis model to predict the bone defect type data; S6. The target bone augmentation morphology is designed according to the defect type data through the bone augmentation design algorithm, and the bone augmentation plan is intelligently recommended based on the designed target bone augmentation morphology.
2. The intelligent alveolar bone augmentation design method based on bone defect analysis according to claim 1, characterized in that: The bone defect parameter data includes the length, width, depth, two-dimensional projection area of the defect area, three-dimensional volume of the defect area and bone defect position data; the adjacent tooth environment data includes the thickness, height, density of the adjacent tooth bone, adjacent tooth position data and adjacent tooth health status data; The three-dimensional image data of the alveolar bone includes the image data of the anatomical structure of the alveolar bone and the position relationship data of the alveolar bone and the teeth; the image data of the anatomical structure of the alveolar bone includes the height, depth and density distribution data of the alveolar bone; the position relationship data of the alveolar bone and the teeth includes the embedding depth and inclination angle of the teeth in the alveolar bone.
3. The intelligent alveolar bone augmentation design method based on bone defect analysis according to claim 2, characterized in that: The method of preprocessing the bone defect parameter data, the adjacent tooth environment data and the alveolar bone three-dimensional image data to obtain the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set comprises: The box plot method is used to detect and eliminate abnormal data in the bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data to obtain the processed bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set. The bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set are normalized and converted into standard normal distribution to obtain the normalized bone defect feature data set, adjacent tooth environment feature data set and three-dimensional image feature data set.
4. The intelligent alveolar bone augmentation design method based on bone defect analysis according to claim 3, characterized in that: The method for obtaining a comprehensive feature data set by associating a bone defect feature data set, an adjacent tooth environment feature data set, and a three-dimensional image feature data set comprises: The bone defect feature dataset, the adjacent tooth environment feature dataset and the three-dimensional image feature dataset are associated and fused through a weighted model to form a comprehensive feature dataset; the bone defect feature dataset is recorded as Q1, the adjacent tooth environment feature dataset is recorded as Q2, and the three-dimensional image feature dataset is recorded as Q3; The weighted model is: GW=Q1·ω1+Q2·ω2+Q3·ω3; wherein ω1 is the weight coefficient of the bone defect feature data set; ω2 is the weight coefficient of the adjacent tooth environment feature data set; ω3 is the weight coefficient of the three-dimensional image feature data set; The weight coefficient of each feature data set in the weighted model is designed: the reference data set is preset as Q0, the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set are used as samples, and the samples are normalized by the normalization formula; the normalization formula is: Among them, Q i′ is the i′th feature data set; Q i ″ is the normalized i′th feature data set; min(Q i′ ) is the minimum value in the i′th feature data set; max(Q i′ ) is the maximum value in the i′th feature data set; For each feature data set, calculate its correlation with the reference data set, and determine the weight coefficient of each feature data set based on the correlation; the calculation formula of the weight coefficient is: Among them, ω i′ is the weight coefficient of the i′th feature data set; i′ is the correlation degree of the i′th feature data set; The correlation degree of the feature data set is: Where n is the number of samples; ξ i′ (k) is the correlation coefficient of the i′th feature data set in the kth sample; k is the type of feature data set; The correlation coefficient between each feature data set and the reference data set is calculated using the correlation coefficient formula; the correlation coefficient formula is: Where Q0(k) is the value of the kth sample in the reference data set Q0; Q′ i′ (k) is the normalized value of the i′th feature data set at the kth sample; ρ is the resolution coefficient; It is the smallest absolute difference among all feature data sets; It is the resolution coefficient multiplied by the maximum absolute difference among all feature data sets; |Q0(k)-Q′ i′ (k)| is the absolute difference between the current feature dataset and the reference dataset; The resolution coefficient in the correlation coefficient formula is restricted by a dynamic adjustment model, and the dynamic adjustment model is: Among them, SR is the type of all feature data sets; is the mean of the standard deviations of all feature data sets, indicating the average dispersion of the overall feature data set; σ max is the maximum standard deviation of all feature data sets, indicating the most discrete feature in the feature data set.
5. The intelligent alveolar bone augmentation design method based on bone defect analysis according to claim 4, characterized in that: The training method of the implant position error prediction model comprises: The data set is divided into a training set, a validation set and a test set; an implant position error prediction model is constructed, and the implant position error prediction model includes input data and output labels; the input data of the model is a historical comprehensive feature data set, and the output label of the model is an implant position error; RBF is selected as the kernel function of the model, and the hyperparameters of the model are initialized, and the hyperparameters of the model include a regularization parameter C and a γ value of the RBF kernel function; the implant position error prediction model is a support vector machine model; The ε-insensitive loss function is used as the loss function of the model to measure the difference between the predicted value and the true value of the model; the ε-insensitive loss function is: Where y is the true value of the implant position error; f(x) is the predicted value of the implant position error; ε is the width of the insensitive interval; The model is trained on the data set, the SMO algorithm is used to solve the optimization problem of ε-SVR, and the hyperparameter combination is adjusted through cross-validation based on the validation set data to find the hyperparameter that minimizes the loss function on the validation set; The model is evaluated using the test data set. The model performance is evaluated by observing the performance indicators of the model on the prediction task. The test is stopped when the model performance reaches the preset performance threshold to obtain the implant position error prediction model. The trained implant position error prediction model is used to predict the current feature data set to obtain the implant position error.
6. The intelligent alveolar bone augmentation design method based on bone defect analysis according to claim 5, characterized in that: The method of comparing the predicted implant position error with a preset implant position error threshold to determine whether the implant position is the optimal position includes: If the predicted implant position error is less than a preset implant position error threshold, the implant position is determined to be the optimal position; If the predicted implant placement error is greater than or equal to a preset implant placement error threshold, it is determined that the implant placement is not the optimal placement.
7. The intelligent alveolar bone augmentation design method based on bone defect analysis according to claim 6, characterized in that: The method of constructing a three-dimensional space model between the implant, bone defect feature data set and the adjacent tooth environment feature data set, and processing the three-dimensional space model to obtain the three-dimensional space feature data comprises: The image segmentation algorithm is used to identify the bone defect area in the three-dimensional image data of the alveolar bone, and the boundary of the bone defect area is extracted by threshold segmentation and edge detection algorithm; the segmentation threshold is preset to L, and each pixel in the image is compared with the threshold L: Among them, g′(a,b) is the segmented image; g(a,b) is each pixel in the image; 255 is the bone defect area; 0 is the background area; The segmented image is used to extract the boundary of the bone defect area using the Canny edge detector; the pixel set in the defect area is preset as {(a i″ ,b i″ )}, the number of pixels in the bone defect area is: Among them, N p is the number of pixels in the bone defect area; g′(a i″ ,b i″ ) is the set of image pixels after segmentation; R is a set of real numbers; The actual area of each pixel is assumed to be A. p , the number of pixels in the bone defect area N p Multiply by the actual area of each pixel A p , and obtain the two-dimensional projection area A of the bone defect area 2D ; Use the 3D reconstruction algorithm to build a 3D space, select each voxel in the bone defect area in the 3D space; multiply the number of voxels in all bone defect areas by the volume of each voxel to obtain the total volume V of the bone defect area 3D ; Fusion of bone defect feature dataset and adjacent tooth environment feature dataset into a unified 3D coordinate system in, is the spatial position coordinate of the feature data point; is the characteristic data of the position coordinates; Build a 3D spatial model and map all feature data points into a grid. For sparse data in the 3D spatial model, use interpolation algorithms to fill in missing data and standardize different types of data. Extract geometric features from the three-dimensional model; the geometric features include the two-dimensional projection area A of the bone defect area 2D and the total volume of the bone defect area V 3D ; Calculate the statistical characteristics that describe the distribution of all feature data points; statistical characteristics include mean and standard deviation; The spatial relationship features between the bone defect feature data set and the adjacent tooth environment feature data set are calculated by the nearest neighbor search algorithm; the spatial relationship features include the minimum distance between all points on the surface of the bone defect area and all points on the surface of the adjacent teeth; the relative position relationship between the bone defect area and the adjacent teeth in three-dimensional space; Calculate the point u(x u ,y u ,z u ) and the point w(x w ,y w ,z w ) between the relative position vector: r = (x u -x w ,y u -y w ,z u -z w ), where r is the relative position vector between each point on the surface of the bone defect area and the point on the surface of the adjacent tooth; The extracted geometric features, statistical features and spatial relationship features are integrated into a feature vector to obtain three-dimensional spatial feature data.
8. The intelligent alveolar bone augmentation design method based on bone defect analysis according to claim 7, characterized in that: The training method of the defect type diagnosis model includes: The data set is divided into a training set, a validation set and a test set; a defect type diagnosis model is constructed, and the defect type diagnosis model includes an input layer, a convolution layer, a pooling layer, a batch normalization layer, a fully connected layer and an output layer; the input layer is the historical three-dimensional spatial feature data, and the output layer is the bone defect type data; softmax is used as the activation function; the defect type diagnosis model is a 3D convolutional neural network model; Multi-classification cross entropy is used as the loss function of the model to measure the difference between the predicted value and the actual value of the model; the multi-classification cross entropy loss function is: Where L is the average loss of the data set; N is the total number of samples in the data set; B is the number of bone defect type data; y ib is the true label of the i-th sample for the b-th type; p ib The probability that the i-th sample belongs to the b-th type predicted by the model; Use the training set data to train the model and use the Adam optimizer to minimize the loss function. Use the validation set to evaluate the performance of the defect type diagnosis model, calculate the accuracy index to measure the performance of the model, and tune the model's hyperparameters to improve the model performance. The performance of the model in the prediction task is evaluated through the test set. When the performance of the model in the prediction task reaches the preset performance threshold, the test is stopped to obtain the defect type diagnosis model. The trained defect type diagnosis model is used to predict the current three-dimensional spatial feature data to obtain the bone defect type data.
9. The intelligent alveolar bone augmentation design method based on bone defect analysis according to claim 8, characterized in that: The method of designing the target bone augmentation morphology according to the defect type data by using the bone augmentation design algorithm and intelligently recommending the bone augmentation scheme according to the designed target bone augmentation morphology includes: The obtained three-dimensional spatial feature data is recorded as J, and the three-dimensional spatial feature data J is standardized: Among them, J OR is the standardized three-dimensional spatial feature data; μ′ is the mean of the feature data; σ′ is the standard deviation of the feature data; The processed three-dimensional spatial feature data is input into the bone augmentation design algorithm; the defect type data is divided into m types to form a defect feature data set {H1, H2, ..., H m }; where m is the type of defect; for each defect type H q , use the generative adversarial network to generate the target bone augmentation morphology, predefine the corresponding bone augmentation morphology rules, the bone augmentation morphology rules are: in, The target bone augmentation morphology for design; Designing rule functions for bone augmentation morphology; Genetic algorithm is used to optimize the design of target bone augmentation morphology, and the objective function F that reflects the applicability of bone augmentation morphology is defined; the target bone augmentation morphology is optimized according to the defect type data: in, is the optimized target bone augmentation shape; argmax is the value that maximizes the objective function F value; A database containing various bone augmentation schemes was established, each of which included a detailed description and applicable conditions. The similarity between the designed target bone augmentation morphology and each scheme in the database was calculated using the cosine similarity function, and the scheme with the highest similarity was selected as the recommended scheme. Based on the designed target bone augmentation morphology, the optimal bone augmentation scheme was recommended from the database.
10. An intelligent alveolar bone augmentation design system based on bone defect analysis, used to implement the intelligent alveolar bone augmentation design method based on bone defect analysis according to any one of claims 1 to 9, characterized in that: include: A data acquisition module, used to collect bone defect parameter data, adjacent tooth environment data and alveolar bone three-dimensional image data; A data processing module is used to pre-process the bone defect parameter data, the adjacent tooth environment data and the alveolar bone three-dimensional image data to obtain a bone defect feature data set, an adjacent tooth environment feature data set and a three-dimensional image feature data set; and to associate and fuse the bone defect feature data set, the adjacent tooth environment feature data set and the three-dimensional image feature data set to obtain a comprehensive feature data set; An implant error prediction module is used to construct an implant position error prediction model, input the comprehensive feature data set into the implant position error prediction model, and predict the implant position error; An implant position evaluation module is used to compare the predicted implant position error with a preset implant position error threshold to determine whether the implant position is the optimal position; The defect type judgment module, if the implant position is the optimal position, constructs a three-dimensional space model between the implant, the bone defect feature data set and the adjacent tooth environment feature data set, processes the three-dimensional space model to obtain three-dimensional space feature data; constructs a defect type diagnosis model, which is used to input the three-dimensional space feature data into the defect type diagnosis model to predict the bone defect type data; The bone augmentation intelligent recommendation module is used to design the target bone augmentation morphology according to the defect type characteristic data through the bone augmentation design algorithm, and intelligently recommend the bone augmentation plan according to the designed target bone augmentation morphology; each module is connected by wired and / or wireless means.
Citation Information
Cited By
Titanium mesh intelligent positioning method, apparatus and device, and storage medium
CN119152039A
Titanium network intelligent positioning method, device, equipment and storage medium
CN119152039B
