Multi-modal missing tooth positioning method based on random forest

Abstract

The invention relates to a multi-modal missing tooth positioning method based on a random forest. The method comprises the following steps: S1, acquiring oral cavity CBCT image data; s2, obtaining a tooth number set and a missing tooth number set of the actual dentition; s3, determining an adjacent tooth of one missing tooth in the missing tooth number set through a preset first mode and a preset second mode, if the adjacent tooth meets a prediction condition, executing S4, otherwise, returning to S3 until all elements in the missing tooth number set are traversed; s4, calculating actual centroid coordinates of adjacent teeth; s5, obtaining a missing tooth centroid prediction model; s6, obtaining an actual centroid coordinate of the missing tooth; and S7, traversing all elements in the missing tooth number set. Compared with the prior art, the method has the advantages of coping with complex challenges caused by deletion of multiple teeth and the like.

Description

Technical Field

[0001] This invention relates to the field of missing tooth localization technology, and in particular to a multimodal missing tooth localization method based on random forest. Background Technology

[0002] Precise positioning of missing teeth is the core foundation of dental restoration and orthodontic treatment, directly affecting the functional fit and long-term stability of restorations. Determining the centroid of the missing tooth and its relative relationship to adjacent teeth using three-dimensional spatial coordinates allows for the restoration of the physiological curvature of the dental arch and the distribution of occlusal biomechanics, avoiding problems such as decreased chewing efficiency, abnormal wear of adjacent teeth, or temporomandibular joint disorder caused by positional deviations. Clinical studies have shown that precise spatial positioning can improve implant success rates and provide crucial anatomical basis for the design of personalized restorative solutions (such as all-ceramic crowns and clear aligners).

[0003] However, locating missing teeth presents multi-dimensional challenges. Irregular tooth alignment, loosening or loss of teeth due to periodontitis may obscure the original missing position; congenital tooth agenesis patients may have hidden tooth germs or variations in dental arch morphology due to abnormal dental lamina development, increasing the difficulty of location. Long-term unrepaired missing teeth can lead to the displacement of adjacent teeth, changes in occlusal relationships, and the original tooth position may be covered by newly formed gaps or tilted teeth.

[0004] Currently, in the field of missing tooth localization technology, traditional methods and emerging technologies coexist and develop. Traditional manual localization relies on the doctor's extensive experience, such as determining the implantation site by creating oral plaster models or manually marking on CT 3D reconstructed images. This process is extremely time-consuming and labor-intensive, and is prone to localization errors due to factors such as doctor fatigue and differences in experience. Emerging technologies, based on deep learning algorithms, have achieved certain results in 2D image analysis, processing 2D images such as X-rays and panoramic radiographs to detect the location and number of missing teeth. However, current AI technology mainly focuses on 2D images and processes single teeth. When processing 3D images that provide more comprehensive oral structural information, the algorithm's maturity is insufficient, and when faced with complex oral conditions, such as multiple missing teeth or interference from dental instruments or restorations, its recognition accuracy is difficult to meet the stringent clinical requirements for high-precision localization. Summary of the Invention

[0005] The purpose of this invention is to provide a multimodal missing tooth localization method based on random forest to address the complex challenges posed by multiple missing teeth.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A multimodal missing tooth localization method based on random forest, comprising the following steps:

[0008] S1. Acquire oral CBCT image data;

[0009] S2. Segment the teeth in the oral CBCT image data, standardize and label them, and number the teeth in the oral cavity to obtain the set of tooth numbers S of the actual dentition. Combine this with the set of tooth numbers of the standard dentition as T = {t1, t2, ..., t...} 28 If the set of missing tooth numbers N = TS is not empty, then execute S3; otherwise, return to S1.

[0010] S3. Using the preset first and second modes, determine the adjacent teeth of a missing tooth in the set of missing tooth numbers. If the adjacent teeth meet the prediction conditions, execute S4; otherwise, return to S3, until all elements in the set of missing tooth numbers are traversed.

[0011] S4. Calculate the actual centroid coordinates of adjacent teeth;

[0012] S5. Obtain training data and train the random forest model to obtain a centroid prediction model for missing teeth.

[0013] S6. Input the actual centroid coordinates of the adjacent teeth into the centroid prediction model of the missing teeth to obtain the actual centroid coordinates of the missing teeth.

[0014] S7. Delete the missing tooth from the set of missing tooth numbers M. Determine whether the set of missing tooth numbers after deletion is empty. If so, end the localization process and use the actual centroid coordinates of all missing teeth as the localization result of the missing tooth.

[0015] Conversely, return to S3 and continue until all elements in the set of missing tooth numbers have been traversed.

[0016] Furthermore, the specific steps for determining the adjacent teeth of a missing tooth in the set of missing tooth numbers, using the preset first and second modes, are as follows:

[0017] Let set T = {t1, t2, ..., t} 28}, where {t1,t2,…,t 14} represents the upper row of teeth, {t 15 ,t 16 …,t 28} represents the lower row of teeth, arranged from left to right. The first pattern is defined as follows: for a tooth, if the tooth belongs to {t1, t2, ..., t...} 12 The first pattern of the tooth is adjacent to the first tooth a to its right, the lower tooth b corresponding to tooth a, and the first tooth c to the right of tooth b.

[0018] For teeth t 13 Its first mode of adjacent teeth is tooth t 12t 27 and t 28 ;

[0019] For teeth t 14 Its first mode of adjacent teeth is tooth t 13 t 26 and t 27 ;

[0020] If the tooth belongs to {t} 17 ,…,t 28 The first pattern of this tooth is adjacent to the first tooth to its left, d, the upper tooth e corresponding to tooth d, and the first tooth f at the coordinate of tooth e.

[0021] For teeth t 15 Its first mode of adjacent teeth is tooth t 16 t2 and t3;

[0022] For teeth t 16 Its first mode of adjacent teeth is tooth t 17 t2 and t1;

[0023] The second mode is defined as follows:

[0024] If the tooth belongs to {t1, t2, ..., t 13 The first pattern adjacent tooth of this tooth is the first tooth g on its right, the lower tooth h corresponding to tooth g, and the lower tooth i corresponding to this tooth.

[0025] For teeth t 14 Its first mode of adjacent teeth is tooth t 13 t 27 and t 28 ;

[0026] If the tooth belongs to {t} 16 ,t2,…,t 27 The first pattern adjacent tooth of this tooth is the first tooth j to its right, the lower tooth k corresponding to tooth j, and the lower tooth l corresponding to this tooth.

[0027] For teeth t 38 Its first mode of adjacent teeth is tooth t 27 t 13 and t 14 ;

[0028] Let A be the set of adjacent teeth of a tooth. m .

[0029] Furthermore, the specific conditions for satisfying the prediction are as follows:

[0030] For a missing tooth, if the set A of adjacent teeth of the missing tooth is... m Belongs to set S, and |A m If |=3, then the prediction condition is met.

[0031] Furthermore, the specific steps of S5 are as follows:

[0032] The centroid coordinates of the missing tooth and the centroid coordinates of the three adjacent teeth are obtained as training data. The training data is then input into a random forest model for training to obtain a centroid prediction model for the missing tooth.

[0033] Furthermore, the centroid coordinates of the missing tooth and the centroid coordinates of its three adjacent teeth are obtained as training data. The specific steps for inputting the training data into the random forest model for training are as follows:

[0034] The centroid coordinates of three adjacent teeth are combined to form a feature vector. This feature vector is then input into a random forest model, which outputs the predicted centroid coordinates of the missing tooth. These centroid coordinates are used as labels to optimize the parameters of the random forest model.

[0035] Furthermore, S5 also includes:

[0036] After obtaining the centroid coordinates of the missing tooth and the centroid coordinates of the three adjacent teeth as training data, the training data is divided into a training set and a test set. The training set is then input into the random forest model for training.

[0037] Furthermore, the training data was divided using a random sampling method.

[0038] Furthermore, the centroid coordinates are three-dimensional coordinates.

[0039] Furthermore, the specific steps of S1 are as follows:

[0040] We acquire initial oral CBCT image data of patients of different ages, genders, races, and various dental health conditions. We then perform anomaly removal and data transformation on the initial data to obtain oral CBCT image data.

[0041] Furthermore, the data is converted into an image with the same resolution and grayscale values ​​in the range [0,1], which is the same as the initial data.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] This invention effectively addresses the complex challenges posed by the loss of multiple teeth. Utilizing the positional information of adjacent teeth, a framework adaptable to different loss patterns is designed, accurately handling the localization of single and multiple missing teeth. This invention employs a first mode and a second mode, considering the positional relationship between upper and lower rows of teeth. This three-dimensional localization logic more closely resembles the actual oral structure, reducing errors in single-row localization. The refined classification of teeth in different positions accommodates individual differences in tooth arrangement. Furthermore, the first and second modes define adjacency relationships from different dimensions. This dual-mode approach allows for correction of results using an alternative mode when some tooth data is missing or adjacency relationships are ambiguous, reducing the limitations of a single rule. Attached Figure Description

[0044] Figure 1 This is a diagram of the multi-modal missing tooth localization framework of the present invention;

[0045] Figure 2 This is a diagram showing the missing pattern partitioning of the present invention. Detailed Implementation

[0046] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0047] This invention proposes a multimodal missing tooth localization method based on random forest, the flowchart of which is shown below. Figure 1 As shown, the method includes the following steps:

[0048] S1. Acquire oral CBCT image data;

[0049] S2. Segment the teeth in the oral CBCT image data, standardize and label them, and number the teeth in the oral cavity to obtain the set of tooth numbers S of the actual dentition. Combine this with the set of tooth numbers of the standard dentition as T = {t1, t2, ..., t...} 28 If the set of missing tooth numbers N = TS is not empty, then execute S3; otherwise, return to S1.

[0050] S3. Using the preset first and second modes, determine the adjacent teeth of a missing tooth in the set of missing tooth numbers. If the adjacent teeth meet the prediction conditions, execute S4; otherwise, return to S3, until all elements in the set of missing tooth numbers are traversed.

[0051] S4. Calculate the actual centroid coordinates of adjacent teeth;

[0052] S5. Obtain training data and train the random forest model to obtain a centroid prediction model for missing teeth.

[0053] S6. Input the actual centroid coordinates of the adjacent teeth into the centroid prediction model of the missing teeth to obtain the actual centroid coordinates of the missing teeth.

[0054] S7. Delete the missing tooth from the set of missing tooth numbers M. Determine whether the set of missing tooth numbers after deletion is empty. If so, end the localization process and use the actual centroid coordinates of all missing teeth as the localization result of the missing tooth.

[0055] Conversely, return to S3 and continue until all elements in the set of missing tooth numbers have been traversed.

[0056] A fully automated tooth segmentation method based on deep learning was used to segment teeth from three-dimensional cone-beam computed tomography (CBCT) images. The segmented teeth were then standardized and labeled according to the standard tooth position coding FDI table established by the International Dental Federation (FDI), and finally, a dataset of segmented tooth images with clear tooth position labels was output.

[0057] The system calculates the numbers of three adjacent teeth based on the missing tooth numbering for different missing tooth patterns. It iterates through the tooth numbering sequence, identifies the labels of existing teeth, compares them with the standard FDI table, and calculates the label of the missing tooth. Centered on the missing tooth, it detects and analyzes the area formed by its adjacent teeth, calculates the numbers and spatial location information of three key adjacent teeth, and dynamically selects adjacent teeth for different missing tooth patterns, providing precise three-dimensional positioning data support for dental restoration or implantation plans.

[0058] For each non-zero unique label (tooth label) in the image, find the voxel coordinates of the corresponding label, calculate the minimum and maximum coordinates of the voxel in three-dimensional space, and then calculate the center coordinates and length of the tooth bounding box. Normalize the data according to the dimensions of the image data.

[0059] This invention uses the labeled 3D centroid positions of missing teeth and adjacent teeth, and employs a random sampling method to divide the training and test sets. This ensures statistical consistency between the training and test sets in dimensions such as tooth loss type and adjacent structure complexity. Root mean square error (RMSE) and mean absolute error (MAE) are used to evaluate model performance, ensuring that the trained random forest model has good generalization ability.

[0060] This invention uses regression to calculate the location of a missing single tooth. The processed feature vectors of adjacent tooth locations are input into a trained and optimized random forest model. Based on the learned relationship between the locations of adjacent teeth and the missing tooth, the model outputs the predicted coordinates of the missing tooth's location in three-dimensional space.

[0061] By utilizing the three-dimensional centroid features of adjacent teeth, the model can effectively adapt to the problem of unknown missing tooth locations in different missing tooth modalities, achieving high-precision missing tooth localization.

[0062] The specific steps of this invention are as follows:

[0063] Step 1: Data preprocessing.

[0064] Raw oral CBCT image data were collected from various medical institutions and research centers, covering patients of different ages, genders, ethnicities, and various dental health conditions. Data with obvious errors or noise interference were removed. Data from different sources were uniformly converted to the same resolution (0.4mm × 0.4mm × 0.4mm), and the grayscale values ​​of the images were mapped to [0,1].

[0065] Step two, tooth segmentation marking.

[0066] Teeth are segmented from 3D cone-beam computed tomography (CBCT) images. The segmented teeth are then standardized and labeled according to the FDI (French Dental Association) standard tooth position coding table, resulting in a dataset of segmented tooth images with clearly defined tooth positions. The FDI coding system is an internationally recognized tooth numbering method that divides the oral cavity into four quadrants (upper, lower, left, and right), with teeth within each quadrant numbered sequentially.

[0067] Step 3: Select adjacent teeth based on the different missing teeth.

[0068] Calculate the numbers of three adjacent teeth based on the missing tooth numbers for different missing tooth situations. Traverse the tooth number sequence, identify the labels of existing teeth, compare them with the standard FDI table, and calculate the missing tooth number. Using the missing tooth as the center, analyze the area formed by its adjacent teeth according to our pre-defined criteria. Figure 2 The missing pattern shown is used to calculate the numbers and spatial location information of three key adjacent teeth. Adjacent teeth and missing teeth are categorized into two missing patterns, based on... Figure 2 If the position of the other tooth is known within the box in the image, the position of the third tooth can be calculated using a regression algorithm, and the position information of the third tooth can be updated into the existing tooth information.

[0069] Calculate the numbers of three adjacent teeth based on the missing tooth numbers for different missing tooth situations. Traverse the tooth number sequence, identify the labels of existing teeth, compare them with the standard FDI table, and calculate the missing tooth number. Using the missing tooth as the center, analyze the area formed by its adjacent teeth according to our pre-defined criteria. Figure 2The missing pattern shown is used to calculate the numbers and spatial location information of three key adjacent teeth. Adjacent teeth and missing teeth are categorized into two missing patterns, based on... Figure 2 If the position of the other tooth is known within the box in the image, the position of the third tooth can be calculated using a regression algorithm, and the position information of the third tooth can be updated into the existing tooth information.

[0070] Calculate the numbers of three adjacent teeth based on the missing tooth numbers for different missing tooth situations. Traverse the tooth number sequence, identify the labels of existing teeth, compare them with the standard FDI table, and calculate the missing tooth number. Using the missing tooth as the center, analyze the area formed by its adjacent teeth according to our pre-defined criteria. Figure 2 The missing pattern shown is used to calculate the numbers and spatial location information of three key adjacent teeth. Adjacent teeth and missing teeth are categorized into two missing patterns, based on... Figure 2 If the position of the other tooth is known within the box in the image, the position of the third tooth can be calculated using a regression algorithm, and the position information of the third tooth can be updated into the existing tooth information.

[0071] Let the set of tooth numbers in a standard dentition be T = {t1, t2, ..., t3}. 28 (Total 28 teeth) are used to identify teeth in different positions. The set of tooth numbers for the actual dentition is obtained through methods such as oral scanning. The set M of missing tooth numbers is determined by the difference between the standard set and the actual set, i.e., M = T\S. Define the spatial location mapping of teeth. P(t i )=(x i ,y i ,z i ) indicates that the number is t i The coordinates of the centroid of the tooth in three-dimensional space.

[0072] Set T = {t1, t2, ..., t} 28 In the table, t1, t2, ..., t are labeled according to the standard FDI table. 28 Corresponding to the labels respectively, that is Figure 1 and Figure 2 17, 16, 15, 14, 13, 12, 11, 21, 22, 23, 24, 25, 26, 27 shown;

[0073] 47, 46, 45, 44, 43, 42, 41, 31, 32, 33, 34, 35, 36, 37.

[0074] For two preset missing patterns (Table 1 and Table 2), rules for locating adjacent teeth and missing teeth are constructed: For any missing tooth m∈M, based on its position in the dentition, the corresponding set of adjacent teeth is selected from Table 1 or Table 2. Satisfy | A m |=3 and (That is, adjacent teeth are all present in the actual dentition).

[0075] Table 1 Pattern: When the missing tooth m satisfies the topological features of Pattern 1, the set of adjacent teeth A m Determined according to the rules in Table 1;

[0076] Table 2 Pattern: When the missing tooth m satisfies the topological features of Pattern 2, the set of adjacent teeth A m Determine according to the rules in Table 2.

[0077] Table 1. Pattern 1

[0078]

[0079]

[0080] Table 2 Mode 2

[0081] Missing tooth number Corresponding adjacent tooth number Missing tooth number Corresponding adjacent tooth number 11 [21,41,31] 31 [32,22,21] 12 [11,42,41] 32 [33,23,22] 13 [12,43,42] 33 [34,24,23] 14 [13,44,43] 34 [35,25,24] 15 [14,45,44] 35 [36,25,26] 16 [15,46,45] 36 [26,27,37] 17 [16,47,46] 37 [26,27,36] 21 [22,31,32] 41 [31,21,11] 22 [23,32,33] 42 [41,11,12] 23 [24,33,34] 43 [42,12,13] 24 [25,34,35] 44 [43,13,14] 25 [26,35,36] 45 [44,15,14] 26 [27,36,37] 46 [45,15,16] 27 [26,36,37] 47 [17,16,46]

[0082] Table 3 shows the tooth arrangements corresponding to the subscripts of set T.

[0083] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

[0084] The tooth arrangements corresponding to the subscripts of set T are shown in Table 3.

[0085] Step four: Process the centroid features of the teeth.

[0086] For each non-zero unique label (tooth label) in the image, find the voxel coordinates of the corresponding label, calculate the minimum and maximum coordinates of the voxel in three-dimensional space, and then calculate the center coordinates and length of the tooth bounding box. Normalize the data according to the dimensions of the image data.

[0087] Step 5: Construct a random forest model.

[0088] Based on the geometric relationships between adjacent teeth in three-dimensional space, a model is established using a regression algorithm:

[0089] P(m)=F(A m )=F(P(a1),P(a2),P(a3))

[0090] Where a1, a2, a3∈A m For adjacent teeth, function F learns from a large amount of standard dental data to determine regression parameters, so that when the coordinates of three adjacent teeth are known, the centroid position of the missing tooth m can be calculated.

[0091] Assuming the centroid coordinates of the missing tooth are (x) 1i ,y1i ,z 1i Let $\mathbf{i}$ represent the geometric center position of the missing tooth in the $i$-th sample, and use it as the label. The coordinates of the three adjacent teeth are denoted as $\mathbf{i}$. 21 ,y 21 ,z 21 ),(x 22 ,y 22 ,z 22 ),(x 23 ,y 23 ,z 23 All features are combined to form a feature vector. A random sampling method is used, dividing the set into 80% training set and 20% test set to ensure statistical consistency between the two sets in dimensions such as tooth loss type and adjacency structure complexity. Root mean square error (RMSE) and mean absolute error (MAE) are used to evaluate model performance, ensuring the trained random forest model has good generalization ability.

[0092] Step 6: Calculate the location of the missing single tooth using regression analysis.

[0093] The processed feature vectors of adjacent tooth positions are input into a trained and optimized random forest model. Based on the learned relationship between the positions of adjacent teeth and the position of the missing tooth, the model outputs the predicted position coordinates of the missing tooth in three-dimensional space.

[0094] Once the coordinates of the missing tooth m are obtained, it is added to the actual tooth set S′=S∪{m}, and the missing set M′=M\{m} is updated. For each element in the new missing set M′, its adjacency relationship is re-examined. If there exists a combination of adjacent teeth that meets the conditions... Then the positioning algorithm is called again to gradually improve the spatial location information of the dental arch.

[0095] This invention discloses a multimodal missing tooth localization method based on random forest suitable for precision dental implantation, including: establishing a framework that includes existing tooth identification, adjacent tooth selection, adjacent tooth centroid position calculation, random forest model construction, and regression solution for missing tooth location.

[0096] The multimodal missing tooth localization method based on this invention accurately predicts the location of missing teeth based on the positional information of adjacent teeth for different modes of missing teeth, thus assisting in dental treatment planning. This invention sets different adjacent teeth for teeth with different missing modes, extracts the three-dimensional centroid position information of the adjacent teeth, and normalizes the image data according to its dimensions as feature input. The three-dimensional spatial position information of the centroid of the missing tooth is then calculated using a random forest algorithm. This design provides flexibility and adaptability for different clinical situations, improves the accuracy of tooth localization during implant placement, and becomes an effective tool for dentists to handle various tooth loss cases.

[0097] According to the present invention, by utilizing the three-dimensional spatial position information of adjacent teeth, it is possible to effectively address the problem of different missing tooth locations, accurately locate missing teeth, improve the accuracy and stability of calculations, and provide a reliable basis for the formulation of oral medical treatment plans.

[0098] The present invention has the following beneficial effects:

[0099] 1. This invention introduces a method for calculating the location of missing teeth, which can accurately predict the three-dimensional spatial location of missing teeth, rather than just their location in a two-dimensional image, providing a precise basis for digital implant planning. This method constructs a tooth database by collecting tooth data from intelligent tooth recognition and uses a random forest algorithm for regression analysis to determine the location of missing teeth.

[0100] 2. This invention effectively addresses the complex challenges posed by the loss of multiple teeth. By utilizing the positional information of adjacent teeth, a framework adaptable to different loss patterns is designed, enabling accurate localization of both single and multiple missing teeth.

[0101] 3. This invention employs an efficient training set generation method to solve the data annotation problem. By simulating tooth loss, data on the missing tooth and its adjacent teeth are extracted from the segmentation results, a process that significantly saves time and labor costs. The versatility and practicality of this framework make it suitable for various dental implant scenarios.

[0102] 4. This invention provides an efficient method for predicting the location of missing teeth, offering new ideas and methods for research on missing tooth restoration, reducing the time and effort required for treatment planning, lowering the cost burden on doctors and patients, and promoting the development of digital dental technology.

[0103] 5. This invention has the ability to predict the location and number of missing teeth, which allows for more personalized treatment plans. Based on each patient's specific needs and unique dental characteristics, a more precise treatment plan can be developed, thereby improving treatment outcomes and patient satisfaction.

[0104] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A multimodal missing tooth localization method based on random forest, characterized in that, The method includes the following steps: S1. Acquire oral CBCT image data; S2. Segment the teeth in the oral CBCT image data, standardize and label them, and number the teeth in the oral cavity to obtain the set of tooth numbers S of the actual dentition. Combine this with the set of tooth numbers of the standard dentition as T = {t1, t2, ..., t...} 28 If the set of missing tooth numbers M = T\S is not empty, then execute S3; otherwise, return to S1. S3. Using the preset first and second modes, determine the adjacent teeth of a missing tooth in the set of missing tooth numbers. If the adjacent teeth meet the prediction conditions, execute S4; otherwise, return to S3, until all elements in the set of missing tooth numbers are traversed. S4. Calculate the actual centroid coordinates of adjacent teeth; S5. Obtain training data and train the random forest model to obtain a centroid prediction model for missing teeth. S6. Input the actual centroid coordinates of the adjacent teeth into the centroid prediction model of the missing teeth to obtain the actual centroid coordinates of the missing teeth. S7. Delete the missing tooth from the set of missing tooth numbers M. Determine whether the set of missing tooth numbers after deletion is empty. If so, end the localization process and use the actual centroid coordinates of all missing teeth as the localization result of the missing tooth. Conversely, return to S3 and continue until all elements in the set of missing tooth numbers have been traversed.

2. The multimodal missing tooth localization method based on random forest according to claim 1, characterized in that, The specific steps for determining the adjacent teeth of a missing tooth in the set of missing tooth numbers, using the preset first and second modes, are as follows: Let set T = {t1, t2, ..., t} 28 }, where {t1,t2,…,t 14 } represents the upper row of teeth, {t 15 ,t 16 …,t 28 } represents the lower row of teeth, arranged from left to right. The first pattern is defined as follows: for a tooth, if the tooth belongs to {t1, t2, ..., t...} 12 The first pattern of the tooth is adjacent to the first tooth a to its right, the lower tooth b corresponding to tooth a, and the first tooth c to the right of tooth b. For teeth t 13 Its first mode of adjacent teeth is tooth t 12 t 27 and t 28 ; For teeth t 14 Its first mode of adjacent teeth is tooth t 13 t 26 and t 27 ; If the tooth belongs to {t} 17 ,…,t 28 The first pattern of this tooth is adjacent to the first tooth to its left, d, the upper tooth e corresponding to tooth d, and the first tooth f at the coordinate of tooth e. For teeth t 15 Its first mode of adjacent teeth is tooth t 16 t2 and t3; For teeth t 16 Its first mode of adjacent teeth is tooth t 17 t2 and t1; The second mode is defined as follows: If the tooth belongs to {t1, t2, ..., t 13 The first pattern adjacent tooth of this tooth is the first tooth g on its right, the lower tooth h corresponding to tooth g, and the lower tooth i corresponding to this tooth. For teeth t 14 Its first mode of adjacent teeth is tooth t 13 t 27 and t 28 ; If the tooth belongs to {t} 16 ,t2,…,t 27 The first pattern adjacent tooth of this tooth is the first tooth j to its right, the lower tooth k corresponding to tooth j, and the lower tooth l corresponding to this tooth. For teeth t 28 Its first mode of adjacent teeth is tooth t 27 t 13 and t 14 ; Let A be the set of adjacent teeth of a tooth. m .

3. The multimodal missing tooth localization method based on random forest according to claim 2, characterized in that, The specific conditions for satisfying the prediction are: For a missing tooth, if the set A of adjacent teeth of the missing tooth is... m Belongs to set S, and |A m If |=3, then the prediction condition is met.

4. The multimodal missing tooth localization method based on random forest according to claim 1, characterized in that, The specific steps of S5 are as follows: The centroid coordinates of the missing tooth and the centroid coordinates of the three adjacent teeth are obtained as training data. The training data is then input into a random forest model for training to obtain a centroid prediction model for the missing tooth.

5. The multimodal missing tooth localization method based on random forest according to claim 4, characterized in that, The specific steps for obtaining the centroid coordinates of the missing tooth and the centroid coordinates of its three adjacent teeth as training data and inputting this training data into the random forest model are as follows: The centroid coordinates of three adjacent teeth are combined to form a feature vector. This feature vector is then input into a random forest model, which outputs the predicted centroid coordinates of the missing tooth. These centroid coordinates are used as labels to optimize the parameters of the random forest model.

6. The multimodal missing tooth localization method based on random forest according to claim 5, characterized in that, S5 also includes: After obtaining the centroid coordinates of the missing tooth and the centroid coordinates of the three adjacent teeth as training data, the training data is divided into a training set and a test set. The training set is then input into the random forest model for training.

7. The multimodal missing tooth localization method based on random forest according to claim 6, characterized in that, The training data was divided using a random sampling method.

8. The multimodal missing tooth localization method based on random forest according to claim 1, characterized in that, The centroid coordinates are three-dimensional coordinates.

9. The multimodal missing tooth localization method based on random forest according to claim 1, characterized in that, The specific steps of S1 are as follows: We acquire initial oral CBCT image data of patients of different ages, genders, races, and various dental health conditions. We then perform anomaly removal and data transformation on the initial data to obtain oral CBCT image data.

10. The multimodal missing tooth localization method based on random forest according to claim 9, characterized in that, The data is converted into an image with the same resolution and grayscale values ​​in the range [0,1], replacing the initial data.

Images