Anterior tooth immediate implantation measurement optimization analysis method based on artificial intelligence

Abstract

The invention discloses an anterior tooth immediate implantation measurement optimization analysis method based on artificial intelligence, which belongs to the technical field of medical information processing and comprises the steps of data acquisition, data preprocessing, abnormal data processing, anterior tooth immediate implantation data clustering and anterior tooth immediate implantation measurement optimization analysis. According to the scheme, the optimal neighbor value is determined through the local density index, the optimal neighbor set is determined through the multi-metric comprehensive distance, the weighted local reachable density is obtained by combining the neighbor weight and the reachable distance, the data are layered based on the local density index, the layered comprehensive abnormal factors are obtained, and the abnormal data are screened. The importance weight of the feature is obtained according to the variable coefficient and the redundancy, a weighted similarity graph is constructed in combination with the self-adaptive connection radius, the structure entropy of the child nodes is calculated, the tree structure is updated, clustering is carried out based on the optimal clustering parameter, the reliability of an implantation effect prediction result is improved, and a precise analysis basis is provided for immediate anterior tooth implantation measurement optimization.

Description

Technical Field

[0001] This invention belongs to the field of medical information processing technology, specifically referring to an artificial intelligence-based method for immediate anterior tooth implant measurement optimization analysis. Background Technology

[0002] The anterior immediate implant measurement and optimization analysis method is a technique that integrates data mining and artificial intelligence technologies. By analyzing anterior immediate implant data, it can uncover potential correlations in historical data and predict the implant outcome level. This provides a scientific basis for developing personalized, high-success-rate anterior immediate implant plans, thereby improving patients' implant treatment experience and prognosis.

[0003] However, existing methods for immediate anterior implant measurement and optimization analysis suffer from several drawbacks. Firstly, the data features of immediate anterior implants are numerous and their meanings differ significantly. These methods cannot comprehensively and accurately measure the similarity between data points, leading to inaccurate anomaly detection results and inaccurate implant outcome predictions. Secondly, the varying weights of different features in immediate anterior implant data on implant outcomes, coupled with the complex interrelationships and diverse data types, result in low reliability of immediate anterior implant outcome predictions, failing to provide targeted analytical basis for immediate anterior implant measurement and optimization. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an artificial intelligence-based method for optimizing and analyzing immediate anterior implant measurements. Existing methods for optimizing and analyzing immediate anterior implant measurements suffer from several problems. Due to the numerous features and significant differences in the meanings of these features, existing methods cannot comprehensively and accurately measure the similarity between data, leading to inaccurate anomaly detection results and inaccurate implant outcome predictions. This invention determines the optimal nearest neighbor value through a local density index, establishes the optimal nearest neighbor set through multi-metric comprehensive distance, and obtains a weighted local reachability density by combining nearest neighbor weights and reachability distance. Based on the local density index, the data is stratified, and local and global anomaly factors are calculated separately to obtain a stratified comprehensive anomaly factor. Finally, anomaly thresholds are determined based on kernel density estimation to filter out anomaly data, preventing it from interfering with subsequent implant outcome predictions and ultimately ensuring the accuracy of anterior implant measurements. Immediate implant measurement optimization analysis improves accuracy. Addressing the issues of low reliability in existing anterior immediate implant measurement optimization methods due to varying weights of features in the data, complex data types, and intricate relationships, this approach fails to provide targeted analytical support for anterior immediate implant measurement optimization. This new method derives feature importance weights based on the coefficient of variation and redundancy, constructs a weighted similarity graph using adaptive connection radii, builds a tree based on initial vertex set partitioning, calculates the structural entropy of child nodes using external penalty terms, local fluctuation coefficients, and feature cohesion, and obtains the total structural entropy of the tree. The tree structure is updated through merging operations to obtain the optimal tree. Finally, density peak clustering is performed based on the optimal clustering parameters to obtain the final clustering result. This achieves accurate grouping of anterior immediate implant data, improves the reliability of implant effect prediction results, and provides precise analytical support for anterior immediate implant measurement optimization.

[0005] The technical solution adopted by this invention is as follows: This invention provides an artificial intelligence-based method for immediate anterior tooth implant measurement and optimization analysis, which includes the following steps:

[0006] Step S1: Data Acquisition;

[0007] Step S2: Data preprocessing;

[0008] Step S3: Abnormal data processing;

[0009] Step S4: Clustering of anterior tooth immediate implant data;

[0010] Step S5: Measurement and optimization analysis of immediate anterior implantation.

[0011] Furthermore, in step S1, the data acquisition involves collecting historical anterior immediate implantation data and real-time anterior immediate implantation data;

[0012] Both the historical immediate anterior implant data and the real-time immediate anterior implant data include the patient's anterior implant measurement data and implantation plan data. The historical immediate anterior implant data also includes the implantation effect level.

[0013] The anterior tooth implant measurement data of the patient includes alveolar bone data, adjacent tooth data, gingival soft tissue data, and infection data;

[0014] The implantation plan data includes implant selection data, implantation parameter data, auxiliary treatment data, and repair design data;

[0015] The planting effect level includes excellent, good, medium and poor. The planting effect level is used as a data label in the clustering process and is used to assign cluster labels to the clusters in the future.

[0016] Furthermore, in step S2, the data preprocessing involves sequentially cleaning, encoding, and normalizing the historical immediate anterior implant data and the real-time immediate anterior implant data.

[0017] Furthermore, in step S3, the abnormal data processing specifically includes the following steps:

[0018] Step S31: Dynamic nearest neighbor determination; set the candidate nearest neighbor value interval as... An anomaly detection dataset was constructed based on all historical immediate anterior implant data after data preprocessing. For each data point in the anomaly detection dataset, its optimal nearest neighbor value was determined by calculating its local density index. To avoid the optimal nearest neighbor value exceeding a reasonable range, if the optimal nearest neighbor value of a data point is less than a certain value... Then the best nearest neighbor value of the data is forced to be equal to If the best nearest neighbor value of the data is greater than Then the best nearest neighbor value of the data is forced to be equal to ; where k min and k max These are the lower and upper limits of the nearest neighbor candidate interval, respectively;

[0019] Step S32: Multi-distance fusion; Calculate the three basic distances between each data point in the anomaly detection dataset and all other data points. The three basic distances are Euclidean distance, Manhattan distance, and cosine distance; For each basic distance, calculate the distance between the current data point and all other data points by multiplying the distances by k. min The standard deviation of all data in the nearest neighbor set determined by the number of nearest neighbors is used to determine the stability of the nearest neighbor distance of this basic distance. Then, each basic distance is normalized, and the normalized distance values ​​are weighted and fused according to their nearest neighbor distance stability to obtain the multi-metric comprehensive distance.

[0020] Step S33: Reachability distance calculation; Determine the optimal nearest neighbor set for each data based on the multi-metric comprehensive distance, and calculate the reachability distance from each data to any data in its optimal nearest neighbor set;

[0021] Step S34: Calculate the weighted local reachability density; calculate the nearest neighbor weights of each data point and each data point in its best nearest neighbor set, and combine the nearest neighbor weights with the reachability distance to obtain the weighted local reachability density of each data point;

[0022] Step S35: Stratified anomaly scoring; sort all data in ascending order of local density indices and calculate the 25th percentile P after sorting. 25% and 75th percentile P 75% ; All local density indices in the anomaly detection dataset that are greater than or equal to P 75% The data is divided into high-density layers, with all local density indices less than or equal to P. 25% The data is divided into a low-density layer and the remaining data is a medium-density layer. Layer weights are set for the high-density layer, medium-density layer and low-density layer respectively. The local anomaly factor of the data is obtained based on the average weighted local reachability density of the data's best nearest neighbor set. The global anomaly factor of the data is then obtained based on the average weighted local reachability density of the density layer to which it belongs. The local anomaly factor and the global anomaly factor are weighted and fused through the layer weights to obtain the layered comprehensive anomaly factor of the data.

[0023] Step S36: Adaptive anomaly screening; perform kernel density estimation on the hierarchical comprehensive anomaly factor of all data in the anomaly detection dataset to obtain the kernel density function. By calculating the second derivative of the kernel density function, find the first valley point of the kernel density function curve. Use the value corresponding to this point as the anomaly threshold. Mark data with hierarchical comprehensive anomaly factor greater than the anomaly threshold as abnormal data, and mark data with hierarchical comprehensive anomaly factor less than or equal to the anomaly threshold as normal data.

[0024] Furthermore, in step S4, the clustering of the anterior tooth immediate implant data specifically includes the following steps:

[0025] Step S41: Weighted Feature Graph Construction; Based on all historical and real-time anterior immediate implant data labeled as normal data, construct an anterior immediate implant dataset; Calculate the coefficient of variation for each feature based on the mean and standard deviation, calculate the redundancy of each feature based on the Pearson correlation coefficient, combine the coefficient of variation and redundancy to obtain the importance weight of each feature, and calculate the adaptive connection radius of each data point; Treat each data point in the anterior immediate implant dataset as a vertex of a graph. If the weighted Euclidean distance between two vertices is less than the minimum of their adaptive connection radii, add an edge between the two vertices, and calculate the edge weight based on the weighted Euclidean distance to construct a weighted similarity graph;

[0026] Step S42: Tree structure optimization clustering; including the following steps:

[0027] Step S421: Initial partitioning and tree construction; Using a stratified random sampling method, the vertex set of the weighted similarity graph is randomly divided into R non-overlapping subsets. The vertex set is used as the root node of the tree, and each subset is used as the direct child node of the root node to construct the initial tree structure.

[0028] Step S422: Calculate the structural entropy of child nodes; calculate the structural entropy of each child node based on the external penalty term, local fluctuation coefficient, and characteristic cohesion;

[0029] Step S423: Calculate the total structural entropy; use the weighted sum of the structural entropies of all child nodes as the total structural entropy of the entire tree;

[0030] Step S424: Merge optimization; Under the same parent node in the current tree, select any two child nodes and simulate merging them into a new child node. Recalculate the new total structural entropy of the merged tree. If the new total structural entropy is less than the total structural entropy of the tree before merging, retain the merge operation and update the tree structure; otherwise, abandon the merge and maintain the original tree structure.

[0031] Step S425: Determine the optimal structure; Repeat step S424 until the total structural entropy of the tree does not decrease after three consecutive iterations. At this point, the tree structure is optimal, and the optimal tree and the subset of data corresponding to each child node in the optimal tree are obtained.

[0032] Step S43: Density peak clustering optimization; using the average silhouette coefficient as the fitness function, the optimal clustering parameters corresponding to each subset of the optimal tree are searched using the bacterial foraging algorithm. Density peak clustering is then performed on each subset based on its optimal clustering parameters. The clustering results of all subsets are then merged to obtain the final clustering result of the anterior immediate implant dataset.

[0033] Further, in step S5, the anterior immediate implant measurement optimization analysis involves calculating the frequency of different data labels in all historical anterior immediate implant data within each cluster of the final clustering result of the anterior immediate implant dataset, taking the data label with the highest frequency as the cluster label of that cluster, and taking the cluster label of the cluster to which the real-time anterior immediate implant data belongs as its corresponding predicted implant effect level, and generating an optimization analysis report.

[0034] The beneficial effects achieved by the present invention using the above solution are as follows:

[0035] (1) To address the problem that existing methods for optimizing and analyzing immediate anterior implant measurements are inaccurate due to the large number of features and significant differences in the meaning of each feature in immediate anterior implant data, which makes it impossible to comprehensively and accurately measure the similarity between data, resulting in inaccurate abnormal data detection results and inaccurate implant effect prediction, this solution determines the optimal nearest neighbor value through local density index to avoid irrelevant data interfering with similarity judgment and prevent subsequent abnormal detection from deviating from the true features of the data due to inappropriate nearest neighbor range; determines the optimal nearest neighbor set through multi-metric comprehensive distance to avoid distortion of the nearest neighbor set due to one-sided similarity measurement, and provides a reliable data comparison benchmark for subsequent abnormal data detection; combines nearest neighbor weight and reachability distance to obtain weighted local reachability density, solves the problem of density calculation distortion, and reduces the misjudgment of abnormal data; based on the local density index, the data is stratified, and local abnormal factors and global abnormal factors are calculated separately to obtain the stratified comprehensive abnormal factors of the data, accurately capturing local and global feature anomalies of the data; and based on kernel density estimation, anomaly thresholds are determined to screen abnormal data, avoid abnormal data interfering with subsequent implant effect prediction, and ultimately ensure the accuracy of immediate anterior implant measurement and optimization analysis.

[0036] (2) To address the problems in existing methods for optimizing anterior immediate implant measurement, such as the varying weights of different features in anterior immediate implant data on implant outcomes, and the complex correlations among multiple data types, leading to low reliability in predicting anterior immediate implant outcomes and the inability to provide targeted analytical basis for optimizing anterior immediate implant measurement, this solution uses the coefficient of variation and redundancy to determine the importance weights of features, ensuring accurate capture of the correlation between key features and implant outcomes. Furthermore, it combines adaptive connection radii to construct a weighted similarity graph, providing a precise similarity framework for subsequent tree-structured clustering, thus avoiding discrepancies between clustering and implant outcomes. The process involves several steps: First, a tree is constructed based on the initial partitioning of the vertex set. Then, the structural entropy of the child nodes is calculated according to the external penalty term, local fluctuation coefficient, and feature cohesion to obtain the total structural entropy of the tree. This provides a quantitative evaluation and iterative basis for subsequent tree structure optimization. The tree structure is updated through a merging operation to obtain the optimal tree, which provides a subset of data with clear boundaries and uniform types for subsequent density peak clustering. Finally, density peak clustering is performed based on the optimal clustering parameters to obtain the final clustering result. This achieves accurate grouping of immediate anterior implant data, improves the reliability of implant effect prediction results, and provides a precise analytical basis for the optimization of immediate anterior implant measurement. Attached Figure Description

[0037] Figure 1 A flowchart illustrating an artificial intelligence-based measurement optimization analysis method for immediate anterior tooth implantation provided by this invention;

[0038] Figure 2 This is a flowchart illustrating step S3;

[0039] Figure 3 This is a flowchart illustrating step S4.

[0040] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0042] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0043] Example 1, see Figure 1 This invention provides an artificial intelligence-based method for immediate anterior tooth implant measurement optimization analysis, which includes the following steps:

[0044] Step S1: Data Acquisition; Acquire historical and real-time anterior immediate implantation data;

[0045] Step S2: Data preprocessing; Data cleaning, data encoding, and data normalization are performed sequentially.

[0046] Step S3: Anomaly data processing; Determine the optimal nearest neighbor value through local density index, determine the optimal nearest neighbor set through multi-metric comprehensive distance, combine nearest neighbor weight and reachability distance to obtain weighted local reachability density, stratify the data based on local density index, calculate local anomaly factor and global anomaly factor respectively to obtain the stratified comprehensive anomaly factor of the data, and determine the anomaly threshold based on kernel density estimation to screen out anomaly data.

[0047] Step S4: Clustering of anterior tooth immediate implantation data; The importance weights of features are obtained based on the coefficient of variation and redundancy, and a weighted similarity graph is constructed by combining the adaptive connection radius. A tree is built based on the initial partition of the vertex set. The structural entropy of the child nodes is calculated based on the external penalty term, local fluctuation coefficient and feature cohesion to obtain the total structural entropy of the tree. The tree structure is updated through merging operations to obtain the optimal tree. Density peak clustering is then performed based on the optimal clustering parameters to obtain the final clustering result.

[0048] Step S5: Anterior immediate implant measurement optimization analysis; Based on the cluster label of the cluster to which the real-time anterior immediate implant data belongs, obtain the corresponding predicted implant effect level and generate an optimization analysis report.

[0049] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the data acquisition is to collect historical anterior tooth immediate implantation data and real-time anterior tooth immediate implantation data.

[0050] Both the historical immediate anterior implant data and the real-time immediate anterior implant data include the patient's anterior implant measurement data and implantation plan data. The historical immediate anterior implant data also includes the implantation effect level.

[0051] The anterior tooth implant measurement data of the patient includes alveolar bone data, adjacent tooth data, gingival soft tissue data, and infection data;

[0052] The alveolar bone data includes alveolar bone width, alveolar bone height, alveolar bone density, labial bone plate thickness, and bone defect type;

[0053] The adjacent tooth data includes the distance between adjacent teeth, the inclination of adjacent teeth, the periodontal condition of adjacent teeth, and the root morphology of adjacent teeth;

[0054] The gingival soft tissue data include gingival thickness, gingival attachment level, gingival biotype, and gingival index;

[0055] The infection data refers to periapical inflammation and soft tissue inflammation in the implantation area.

[0056] The implantation plan data includes implant selection data, implantation parameter data, auxiliary treatment data, and repair design data;

[0057] The implant selection data includes implant brand, implant model, implant diameter, implant length, and implant surface treatment type;

[0058] The implantation parameter data includes implantation angle, implantation depth, and implantation torque;

[0059] The adjunctive treatment data includes whether bone grafting was performed, the type of bone graft material, and whether guided bone regeneration was used.

[0060] The repair design data includes the implantation surgical method, repair timing, healing abutment type, and temporary prosthesis type;

[0061] The planting effect level includes excellent, good, medium and poor. The planting effect level is used as a data label in the clustering process and is used to assign cluster labels to the clusters in the future.

[0062] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the data preprocessing is to perform data cleaning, data encoding and data normalization on historical anterior immediate implantation data and real-time anterior immediate implantation data in sequence.

[0063] The data cleaning process involves removing erroneous and missing values ​​from the data.

[0064] The data encoding uses One-Hot encoding to convert categorical data into numerical data.

[0065] The data normalization method uses a max-min scaling approach to unify numerical data to the same range.

[0066] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, the abnormal data processing specifically includes the following steps:

[0067] Step S31: Dynamic Nearest Neighbor Determination; Anterior tooth immediate implant data exhibits significant local density differences. Using fixed nearest neighbor values ​​for anomaly detection would result in overly broad nearest neighbor criteria for clustered data and overly narrow criteria for dispersed data, failing to adapt to local feature differences. Determining the optimal nearest neighbor value through local density indices provides a precise nearest neighbor range for subsequent reachable distance and local reachable density calculations, ensuring that implant data with different local features can be anomaly detected based on their own nearest neighbor environment, reducing misjudgments due to inappropriate nearest neighbor ranges; The candidate nearest neighbor value interval is set as follows: An anomaly detection dataset was constructed based on all historical immediate anterior implant data after data preprocessing. For each data point in the anomaly detection dataset, its optimal nearest neighbor value was determined by calculating its local density index. To avoid the optimal nearest neighbor value exceeding a reasonable range, if the optimal nearest neighbor value of a data point is less than a certain value... Then the best nearest neighbor value of the data is forced to be equal to If the best nearest neighbor value of the data is greater than Then the best nearest neighbor value of the data is forced to be equal to The formula used is as follows:

[0068] ;

[0069] ;

[0070] In the formula, x i It is the i-th data point in the anomaly detection dataset. It is x i Local density index, It is k min x is determined by the number of nearest neighbors i The nearest neighbor set, x j yes The j-th data in the array, where i and j are data indices, and k is the index. min and k max These are the lower and upper bounds of the nearest neighbor candidate interval, respectively, k min =3,k max =20, It is x i and x j The Euclidean distance between them, P min and P max These are the minimum and maximum values ​​of the local density index for all data in the anomaly detection dataset, respectively. It is x i The optimal nearest neighbor value, It is the floor function. It is a smoothing term. , It is a function that takes the maximum value;

[0071] Step S32: Multi-distance fusion; Anterior tooth immediate implantation data has multiple dimensions and different feature meanings. A single distance cannot comprehensively measure the similarity between data, easily leading to biased data similarity judgments and affecting the accuracy of anomaly detection. Fusion of three distances provides a more accurate distance basis for subsequent anomaly data identification; Calculate three basic distances between each data point in the anomaly detection dataset and all other data points: Euclidean distance, Manhattan distance, and cosine distance; For each basic distance, calculate the distance between the current data and its k-th... min The standard deviation of all data within the nearest neighbor set, determined by the number of nearest neighbors, is used to determine the stability of the nearest neighbor distance. Then, each basic distance is normalized, and the normalized distance values ​​are weighted and fused based on their nearest neighbor distance stability to obtain the multi-metric comprehensive distance. The formula used is as follows:

[0072] ;

[0073] ;

[0074] In the formula, This represents the nearest neighbor distance stability of the v-th type of basic distance, where v is the type index of the basic distance. and They are xi and x j The v-th basic distance and the multi-metric integrated distance between them It is the standard deviation function. It is the sum of the nearest neighbor distance stability of the three basic distances. It is x i arrive The maximum value of the v-th basic distance among all data in the dataset;

[0075] Step S33: Reachability distance calculation; In immediate anterior implant data, some marginally normal data are easily misjudged as abnormal. If the original distance is used directly, it may be marked as abnormal because it is far from most data. At the same time, the distance features of some truly abnormal data are not prominent, making it difficult to distinguish from normal data. By using reachability distance, we can avoid marginally normal data being misjudged due to their large absolute distance, improve the accuracy of anomaly detection, and eliminate invalid data interference for subsequent cluster analysis. Based on the multi-metric comprehensive distance, we determine the optimal nearest neighbor set for each data point and calculate the reachability distance from each data point to any data point in its optimal nearest neighbor set. The formula used is as follows:

[0076] ;

[0077] ;

[0078] In the formula, It is x i The optimal nearest neighbor set, x s It is the anomaly detection dataset D excluding x i The s-th data point outside of x w yes The w-th data in the array, where s and w are data indices. It is x i and x s Multi-metric integrated distance between them It is x i and x w Multi-metric integrated distance between them It is for x i After sorting the multi-metric composite distances with all other data in D in ascending order The corresponding ranking It is x i x to its best nearest neighbor set w The reachable distance, It is x w In the optimal nearest neighbor set with x w Multi-metric composite distance of the data with the furthest distance;

[0079] Step S34: Weighted Local Reachability Density Calculation; In immediate anterior implant data, the influence of data within the same nearest neighbor set on the central data varies. However, traditional local reachability density calculation treats all nearest neighbor data equally, failing to reflect this difference and leading to inaccurate density calculation, which affects subsequent anomaly scoring. By introducing neighbor weights to differentiate and measure the contribution of nearest neighbor data, the accuracy of density calculation is improved. The nearest neighbor weights of each data point and each data point within its optimal nearest neighbor set are calculated. Combining the nearest neighbor weights with the reachability distance, the weighted local reachability density of each data point is obtained. The formula used is as follows:

[0080] ;

[0081] ;

[0082] In the formula, It is x i With x w nearest neighbor weights, It is x i In the optimal nearest neighbor set with x i Multi-metric composite distance of the data with the furthest distance It is x i to its best nearest neighbor set The reachable distance, yes The first in One data point, It is a data index. It is x i Weighted locally reachable density;

[0083] Step S35: Stratified anomaly scoring; In the immediate anterior implant data, the criteria for judging anomalies differ across different density layers. Using a uniform scoring standard would lead to missed detection of abnormal data in high-density layers and misjudgment of normal data in low-density layers, affecting the accuracy of anomaly screening. By establishing stratified anomaly judgment criteria, the accuracy of anomaly scoring is improved. The local density indicators of all data are sorted in ascending order, and the 25th percentile P after sorting is calculated. 25% and 75th percentile P 75% ; All local density indices in the anomaly detection dataset that are greater than or equal to P 75% The data is divided into high-density layers, with all local density indices less than or equal to P. 25%The data is divided into a low-density layer and the remaining data into a medium-density layer. The layer weights for the high-density, medium-density, and low-density layers are set to 0.3, 0.5, and 0.7, respectively. The local anomaly factor is obtained based on the average weighted local reachability density of the data's best nearest neighbor set. The global anomaly factor is then obtained based on the average weighted local reachability density of its respective density layer. The local and global anomaly factors are weighted and fused using the layer weights to obtain the comprehensive layered anomaly factor. The formula used is as follows:

[0084] ;

[0085] ;

[0086] ;

[0087] In the formula, , and They are x i Local anomaly factors, global anomaly factors, and hierarchical comprehensive anomaly factors, It is x i The density layer to which it belongs, yes The amount of data in yes Hierarchical weights, x z yes The z-th data in the data, where z is the data index. and They are x w and x z Weighted locally reachable density;

[0088] Step S36: Adaptive Anomaly Screening; The anomaly threshold for immediate anterior implant data is difficult to set manually. If a fixed threshold is used, it will lead to a mismatch between the threshold and the data distribution, resulting in missed detection of abnormal data or misjudgment of normal data. It cannot adapt to dynamic changes in data. By dynamically matching the anomaly threshold to the data distribution, it is ensured that different batches and types of immediate anterior implant data can obtain objective and consistent anomaly screening results, providing clean and reliable normal data for subsequent cluster analysis. Kernel density estimation is performed on the hierarchical comprehensive anomaly factor of all data in the anomaly detection dataset to obtain the kernel density function. By calculating the second derivative of the kernel density function, the first valley point of the kernel density function curve is found. The value corresponding to this point is used as the anomaly threshold. Data with a hierarchical comprehensive anomaly factor greater than the anomaly threshold are marked as abnormal data, and data with a hierarchical comprehensive anomaly factor less than or equal to the anomaly threshold are marked as normal data.

[0089] By performing the above operations, this solution addresses the problem in existing methods for optimizing anterior immediate implant measurement. Existing methods for anterior immediate implant data have many feature dimensions and significantly different meanings, making it impossible to comprehensively and accurately measure the similarity between data. This leads to inaccurate anomaly detection results, resulting in inaccurate implant outcome predictions and ultimately, inaccurate anterior immediate implant measurement optimization analysis results. This solution uses a local density index to determine the optimal nearest neighbor value, avoiding interference from irrelevant data in similarity judgment and preventing subsequent anomaly detection from deviating from the true characteristics of the data due to inappropriate nearest neighbor ranges. Furthermore, it uses a multi-metric comprehensive distance to determine the optimal nearest neighbor set, avoiding the influence of... The one-sided similarity measurement leads to distortion of the nearest neighbor set, providing a reliable data comparison benchmark for subsequent anomaly detection. By combining nearest neighbor weights and reachability distance, a weighted local reachability density is obtained, solving the problem of density calculation distortion and reducing misjudgment of anomaly data. Based on the local density index, the data is stratified, and local and global anomaly factors are calculated separately to obtain the stratified comprehensive anomaly factor of the data, accurately capturing local and global feature anomalies. Furthermore, based on kernel density estimation, anomaly thresholds are determined to screen out anomaly data, avoiding interference from anomaly data in subsequent implantation effect prediction, and ultimately ensuring the accuracy of the measurement optimization analysis of immediate anterior implantation.

[0090] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S4, the clustering of anterior tooth immediate implant data specifically includes the following steps:

[0091] Step S41: Weighted Feature Mapping; In immediate anterior implant data, the influence weight of each feature on the implantation effect is different. If all features are treated equally, secondary features will interfere with the clustering results. At the same time, the connection radius between data is fixed, which can easily lead to excessive dense data connections or insufficient sparse data connections. By accurately allocating feature weights and adaptively adjusting the data connection range, a weighted similarity map that better matches the distribution of immediate anterior implant data is constructed, providing accurate similarity topology for subsequent tree structure clustering; based on all historical immediate anterior implant data marked as normal data... An immediate anterior implant dataset was constructed using real-time anterior tooth implantation data. The coefficient of variation for each feature was calculated based on the mean and standard deviation, and the redundancy of each feature was calculated based on the Pearson correlation coefficient. The coefficient of variation and redundancy were combined to obtain the importance weight of each feature, and the adaptive connection radius for each data point was calculated. Each data point in the immediate anterior implant dataset was treated as a vertex in a graph. If the weighted Euclidean distance between two vertices was less than the minimum of their adaptive connection radii, an edge was added between the two vertices, and the edge weight was calculated based on the weighted Euclidean distance to construct a weighted similarity graph. The formula used is as follows:

[0092] ;

[0093] ;

[0094] ;

[0095] ;

[0096] ;

[0097] ;

[0098] In the formula, q a and q b These are the a-th and b-th features, respectively, CV a RE a and K a They are q a The coefficient of variation, redundancy, and importance weights, O m and O n These are the adaptive connection radii for the m-th and n-th data points, respectively. It is the weighted Euclidean distance between the m-th and n-th data points in the immediate anterior implant dataset, where m and n are data indices. and Z represents the values ​​of the m-th and n-th data points on the a-th feature, respectively. def This is the baseline distance, which is the median of the weighted Euclidean distances among all data points in the immediate anterior implant dataset. It is the weighted Euclidean distance average between the m-th data point and all data points in its best nearest neighbor set. and These are the central q data sets for immediate anterior implantation. a The standard deviation and mean of the dataset are given, where A is the number of features in the immediate anterior implant dataset, and a, b, and c are the feature indices. It is q a and q b Pearson correlation coefficient between them, CV c and RE c These are the coefficient of variation and redundancy of the c-th feature, respectively. It represents the edge weight between the m-th and n-th data vertices, where V, E, and W are the vertex set, edge set, and edge weight set, respectively. It is a function that takes the minimum value;

[0099] Step S42: Tree Structure Optimization Clustering; Anterior immediate implant data exhibits characteristics of multi-type mixing and complex data correlation. Directly performing single clustering on the entire dataset makes it difficult to accurately distinguish data groups under different scenarios, leading to a disconnect between clustering results and actual needs. Furthermore, the clustering process lacks quantitative quality assessment standards, relying solely on distance similarity judgment, resulting in low reliability of clustering results and failing to provide an effective clustering foundation for implantation scheme optimization. By constructing an optimal tree structure, accurate dataset partitioning is provided for subsequent density peak clustering, ensuring that clustering results directly correspond to data groups with different needs; including the following steps:

[0100] Step S421: Initial partitioning and tree construction; using a stratified random sampling method, the vertex set of the weighted similarity graph is randomly partitioned into... Given N non-overlapping subsets, use these subsets as the root node of the tree, and each subset as a direct child of the root node to construct the initial tree structure; where N V It is the number of vertices contained in V;

[0101] Step S422: Calculate the structural entropy of child nodes; calculate the structural entropy of each child node based on the external penalty term, local fluctuation coefficient, and characteristic cohesion; the formula used is as follows:

[0102] ;

[0103] ;

[0104] ;

[0105] ;

[0106] In the formula, WL r JM r TN r and U r These are the outer penalty term, local fluctuation coefficient, feature cohesion, and structural entropy of the r-th child node, respectively, where r is the child node index, and B... r It is a subset of the vertices corresponding to the r-th child node. N is the edge weight between vertex g and vertex p, where g and p are vertex indices. r It is B r The number of vertices contained in It is B r The set of vertices in the interval that are connected to g by edges. It is B r All values ​​of the a-th feature, It is B r The coefficient of variation of the a-th feature. It is B r The mean of the a-th feature, It is B r The eigenvalues ​​of the a-th feature and The maximum absolute difference;

[0107] Step S423: Calculate the total structural entropy; the weighted sum of the structural entropies of all child nodes is taken as the total structural entropy U of the entire tree. total The formula used is as follows:

[0108] ;

[0109] Step S424: Merge optimization; Under the same parent node in the current tree, select any two child nodes and simulate merging them into a new child node. Recalculate the new total structural entropy of the merged tree. If the new total structural entropy is less than the total structural entropy of the tree before merging, retain the merge operation and update the tree structure; otherwise, abandon the merge and maintain the original tree structure.

[0110] Step S425: Determine the optimal structure; Repeat step S424 until the total structural entropy of the tree does not decrease after three consecutive iterations. At this point, the tree structure is optimal, and the optimal tree and the subset of data corresponding to each child node in the optimal tree are obtained.

[0111] Step S43: Density peak clustering optimization; After tree structure optimization clustering of the immediate anterior implant data, the resulting sub-node datasets exhibit significant heterogeneity, with sparse data distribution and large feature differences. Using a uniform density peak clustering parameter for all sub-node datasets would distort the clustering results and fail to provide accurate cluster category support for subsequent implantation effect prediction. By achieving accurate clustering of the sub-node datasets, matching the data heterogeneity characteristics, and improving clustering standardization and practicality, accurate implantation effect prediction can be supported. Using the average silhouette coefficient as the fitness function, the optimal clustering parameter [d] corresponding to each sub-dataset in the optimal tree when using the density peak clustering algorithm is searched using a bacterial foraging algorithm. c Based on their respective optimal clustering parameters, density peak clustering is performed on each subset of data, and the clustering results of all subsets are merged to obtain the final clustering result of the anterior immediate implant dataset; where d c γ and γ are the cutoff distance and the cluster center screening threshold, respectively.

[0112] By performing the above operations, this solution addresses the problems in existing methods for optimizing anterior immediate implant measurement. These problems stem from the varying weights of different features in anterior immediate implant data on implant outcomes, as well as the complex interrelationships and diverse data types, leading to low reliability in predicting anterior immediate implant outcomes and an inability to provide targeted analytical support for optimization. This new approach derives the importance weights of features based on the coefficient of variation and redundancy, ensuring accurate capture of the correlation between key features and implant outcomes. Furthermore, it constructs a weighted similarity graph using an adaptive connection radius, providing a precise similarity framework for subsequent tree-structured clustering and avoiding misalignment between clustering and seeding. The implantation results were not properly aligned with the target data. A tree was initially constructed based on the vertex set. The structural entropy of the child nodes was calculated using the external penalty term, local fluctuation coefficient, and feature cohesion to obtain the total structural entropy of the tree. This provides a quantitative assessment and iterative basis for subsequent tree structure optimization. The tree structure was updated through a merging operation to obtain the optimal tree, providing a subset of data with clear boundaries and uniform types for subsequent density peak clustering. Density peak clustering was then performed based on the optimal clustering parameters to obtain the final clustering result. This achieved accurate grouping of immediate anterior implantation data, improved the reliability of implantation result prediction, and provided a precise analytical basis for optimizing immediate anterior implantation measurements.

[0113] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the anterior immediate implant measurement optimization analysis is to count the frequency of different data labels in all historical anterior immediate implant data in each cluster of the final clustering result of the anterior immediate implant dataset, take the data label with the highest frequency as the cluster label of the cluster, take the cluster label of the cluster to which the real-time anterior immediate implant data belongs as its corresponding predicted implant effect level, and generate an optimization analysis report.

[0114] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0115] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0116] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A measurement optimization analysis method for immediate anterior tooth implantation based on artificial intelligence, characterized in that: The method includes the following steps: Step S1: Data Acquisition; Acquire historical and real-time anterior immediate implantation data; Step S2: Data preprocessing; Data cleaning, data encoding, and data normalization are performed sequentially. Step S3: Anomaly data processing; Determine the optimal nearest neighbor value through local density index, determine the optimal nearest neighbor set through multi-metric comprehensive distance, combine nearest neighbor weight and reachability distance to obtain weighted local reachability density, stratify the data based on local density index, calculate local anomaly factor and global anomaly factor respectively to obtain the stratified comprehensive anomaly factor of the data, and determine the anomaly threshold based on kernel density estimation to screen out anomaly data. Step S4: Clustering of anterior tooth immediate implantation data; The importance weights of features are obtained based on the coefficient of variation and redundancy, and a weighted similarity graph is constructed by combining the adaptive connection radius. A tree is built based on the initial partition of the vertex set. The structural entropy of the child nodes is calculated based on the external penalty term, local fluctuation coefficient and feature cohesion to obtain the total structural entropy of the tree. The tree structure is updated through merging operations to obtain the optimal tree. Density peak clustering is then performed based on the optimal clustering parameters to obtain the final clustering result. Step S5: Anterior immediate implant measurement optimization analysis; Based on the cluster label of the cluster to which the real-time anterior immediate implant data belongs, obtain the corresponding predicted implant effect level and generate an optimization analysis report.

2. The method for measurement and optimization analysis of immediate anterior tooth implantation based on artificial intelligence according to claim 1, characterized in that: In step S3, the abnormal data processing specifically includes the following steps: Step S31: Dynamic nearest neighbor determination; Set the candidate range of nearest neighbor values, construct an anomaly detection dataset based on all historical anterior tooth immediate implantation data after data preprocessing, and determine the optimal nearest neighbor value for each data in the anomaly detection dataset by calculating its local density index. Step S32: Multi-distance fusion; Step S33: Reachability distance calculation; Determine the optimal nearest neighbor set for each data based on the multi-metric comprehensive distance, and calculate the reachability distance from each data to any data in its optimal nearest neighbor set; Step S34: Calculate the weighted local reachability density; calculate the nearest neighbor weights of each data point and each data point in its best nearest neighbor set, and combine the nearest neighbor weights with the reachability distance to obtain the weighted local reachability density of each data point; Step S35: Stratified anomaly scoring; Step S36: Adaptive anomaly screening; perform kernel density estimation on the hierarchical comprehensive anomaly factor of all data in the anomaly detection dataset to obtain the kernel density function. By calculating the second derivative of the kernel density function, find the first valley point of the kernel density function curve. Use the value corresponding to this point as the anomaly threshold. Mark data with hierarchical comprehensive anomaly factor greater than the anomaly threshold as abnormal data, and mark data with hierarchical comprehensive anomaly factor less than or equal to the anomaly threshold as normal data.

3. The method for measurement and optimization analysis of immediate anterior tooth implantation based on artificial intelligence according to claim 2, characterized in that: In step S32, the multi-distance fusion calculates three basic distances between each data point in the anomaly detection dataset and all other data points: Euclidean distance, Manhattan distance, and cosine distance. For each basic distance, the stability of the nearest neighbor distance is determined by calculating the standard deviation of the current data point and all data points in its nearest neighbor set. Then, each basic distance is normalized, and the normalized distance values ​​are weighted and fused based on their nearest neighbor distance stability to obtain the multi-metric comprehensive distance.

4. The method for measurement and optimization analysis of immediate anterior tooth implantation based on artificial intelligence according to claim 2, characterized in that: In step S35, the stratified anomaly scoring involves sorting all data in ascending order by their local density indices and calculating the 25th percentile P after sorting. 25% and 75th percentile P 75% ; All local density indices in the anomaly detection dataset that are greater than or equal to P 75% The data is divided into high-density layers, with all local density indices less than or equal to P. 25% The data is divided into a low-density layer and the remaining data is a medium-density layer. Layer weights are set for the high-density layer, medium-density layer and low-density layer respectively. The local anomaly factor of the data is obtained based on the average weighted local reachability density of the data's best nearest neighbor set. The global anomaly factor of the data is then obtained based on the average weighted local reachability density of the density layer to which it belongs. The local anomaly factor and the global anomaly factor are weighted and fused through the layer weights to obtain the layered comprehensive anomaly factor of the data.

5. The method for measurement and optimization analysis of immediate anterior tooth implantation based on artificial intelligence according to claim 1, characterized in that: In step S4, the clustering of anterior tooth immediate implant data specifically includes the following steps: Step S41: Weighted Feature Graph Construction; Based on all historical and real-time anterior immediate implant data labeled as normal data, construct an anterior immediate implant dataset; Calculate the coefficient of variation for each feature based on the mean and standard deviation, calculate the redundancy of each feature based on the Pearson correlation coefficient, combine the coefficient of variation and redundancy to obtain the importance weight of each feature, and calculate the adaptive connection radius of each data point; Treat each data point in the anterior immediate implant dataset as a vertex of a graph. If the weighted Euclidean distance between two vertices is less than the minimum of their adaptive connection radii, add an edge between the two vertices, and calculate the edge weight based on the weighted Euclidean distance to construct a weighted similarity graph; Step S42: Optimize clustering using tree structure; Step S43: Density peak clustering optimization; using the average silhouette coefficient as the fitness function, the optimal clustering parameters corresponding to each subset of the optimal tree are searched using the bacterial foraging algorithm. Density peak clustering is then performed on each subset based on its optimal clustering parameters. The clustering results of all subsets are then merged to obtain the final clustering result of the anterior immediate implant dataset.

6. The method for measurement and optimization analysis of immediate anterior tooth implantation based on artificial intelligence according to claim 5, characterized in that: In step S42, the tree structure optimization clustering specifically includes the following steps: Step S421: Initial partitioning and tree construction; Using a stratified random sampling method, the vertex set of the weighted similarity graph is randomly divided into non-overlapping subsets. The vertex set is used as the root node of the tree, and each subset is used as the direct child node of the root node to construct the initial tree structure. Step S422: Calculate the structural entropy of child nodes; calculate the structural entropy of each child node based on the external penalty term, local fluctuation coefficient, and characteristic cohesion; Step S423: Calculate the total structural entropy; use the weighted sum of the structural entropies of all child nodes as the total structural entropy of the entire tree; Step S424: Merge optimization; Under the same parent node in the current tree, select any two child nodes and simulate merging them into a new child node. Recalculate the new total structural entropy of the merged tree. If the new total structural entropy is less than the total structural entropy of the tree before merging, retain the merge operation and update the tree structure; otherwise, abandon the merge and maintain the original tree structure. Step S425: Determine the optimal structure; Repeat step S424 until the total structural entropy of the tree does not decrease after three consecutive iterations. At this point, the tree structure is optimal, and the optimal tree and the subset of data corresponding to each child node in the optimal tree are obtained.

7. The method for measurement and optimization analysis of immediate anterior tooth implantation based on artificial intelligence according to claim 1, characterized in that: In step S1, the data acquisition involves collecting historical anterior immediate implantation data and real-time anterior immediate implantation data. Both the historical immediate anterior implant data and the real-time immediate anterior implant data include the patient's anterior implant measurement data and implantation plan data. The historical immediate anterior implant data also includes the implantation effect level. The anterior tooth implant measurement data of the patient includes alveolar bone data, adjacent tooth data, gingival soft tissue data, and infection data; The implantation plan data includes implant selection data, implantation parameter data, auxiliary treatment data, and repair design data; The planting effect level includes excellent, good, medium and poor. The planting effect level is used as a data label in the clustering process and is used to assign cluster labels to the clusters in the future.

8. The method for measurement and optimization analysis of immediate anterior tooth implantation based on artificial intelligence according to claim 1, characterized in that: In step S5, the anterior immediate implant measurement optimization analysis involves calculating the frequency of different data labels in all historical anterior immediate implant data for each cluster in the final clustering result of the anterior immediate implant dataset, taking the data label with the highest frequency as the cluster label of the cluster, and taking the cluster label of the cluster to which the real-time anterior immediate implant data belongs as its corresponding predicted implant effect level, and generating an optimization analysis report.

Images