Intraoral scanning interactive segmentation method based on adaptive curvature

Abstract

The invention relates to the technical field of computer graphic processing, in particular to an intraoral scanning interactive segmentation method based on self-adaptive curvature, which comprises the following steps: importing a three-dimensional model into a system, calculating the curvature value of each vertex by the system by adopting a self-adaptive curvature analysis method, and identifying and extracting a gingival margin line part; correcting an inaccurate part through an interactive interface to ensure the continuity and accuracy of the gingival margin line; gradually expanding adjacent vertexes and patches which conform to a curvature threshold based on the seed points, and extracting a complete gingival margin line; accurate control on the form of the gingival margin line is realized through a self-adaptive curve fitting morphological processing method; the system uses a connected component detection algorithm to identify and segment the independent dental crown area, and each connected component is classified into an independent dental crown or gingival area to realize segmentation. According to the method, curvature analysis and a small amount of manual interaction are combined, so that the method can adapt to diversified tooth forms and maintain relatively high segmentation precision.

Description

Technical Field

[0001] The present invention relates to the field of computer graphics processing technology, particularly to the segmentation and curvature analysis of three-dimensional models of intraoral scans in oral medicine. Specifically, it relates to an interactive segmentation method for intraoral scans based on adaptive curvature, which is used for model processing in digital dental diagnosis and treatment, including applications such as crown restoration design and surgical simulation. Background Art

[0002] In the application of computer vision in oral medicine, intraoral scan segmentation is of great significance. By accurately segmenting the dental crowns to generate high-precision three-dimensional models, it can not only be used for digital tooth restoration but also help dentists understand the structure of diseased teeth in detail and formulate detailed treatment plans. Currently, the processing of three-dimensional models of intraoral scans faces many challenges, mainly manifested in the significant differences in tooth shapes among different individuals, and it is impossible to describe and predict them through simple parametric models. Especially in the case of complex tooth shapes, the segmentation results are prone to errors. In clinical cases, many patients have tooth abnormalities, missing teeth, or crowded and misaligned teeth. In particular, when the patient's teeth are severely crowded or misaligned, the shape of the dental crown becomes more complex, resulting in the inability of conventional segmentation algorithms to accurately distinguish and extract individual dental crowns. For these abnormal situations, most of the existing methods rely on a large amount of manual operations, with low efficiency and difficult to guarantee accuracy.

[0003] Although deep learning techniques have been widely used in tooth segmentation in recent years (such as MeshSegNet, TSegNet methods), these methods usually rely on the training of large-scale data and perform poorly in dealing with tooth abnormalities and misalignment situations. In addition, deep learning methods often lack sufficient robustness in dealing with the noise introduced by scanning devices, resulting in unsatisfactory segmentation results. Therefore, there is an urgent need for an interactive segmentation method for intraoral scans based on adaptive curvature, which is used for model processing in digital dental diagnosis and treatment, including applications in fields such as crown restoration design and surgical simulation, so as to solve the deficiencies of the above-mentioned existing technologies. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an interactive segmentation method for intraoral scans based on adaptive curvature for the segmentation problems caused by the diversity of tooth shapes, tooth abnormalities, and the limitations of scanning devices in the prior art.

[0005] The technical problem to be solved by the present invention is achieved through the following technical solutions. The present invention discloses an interactive segmentation method for intraoral scans based on adaptive curvature, including the following steps. Step S1, import the three-dimensional model into the system. The system calculates using the adaptive curvature analysis method, calculates the curvature value of each vertex, and enhances the visual contrast of the curvature through custom color mapping. According to the curvature threshold, the system automatically identifies and extracts the gingival margin line part;

[0006] Step S2: For the discontinuous or misjudged areas that may occur during the segmentation process in Step S1, the system provides a manual interactive correction method editing tool. Users can correct inaccurate parts through the interface to ensure the continuity and accuracy of the gingival margin line.

[0007] Step S3: The user selects an arbitrary seed point on the gingival margin line of the model using the key area segmentation and extraction method. Based on this seed point, the system starts the region growing algorithm, and extracts the complete gingival margin line by gradually expanding adjacent vertices and patches that meet the curvature threshold.

[0008] Step S4: The system uses an adaptive curve fitting morphological processing method to achieve precise control of the gingival margin line morphology and further optimize the gingival margin line morphology. The morphological processing includes dilation, erosion, opening operation, and closing operation based on introducing the shape and size of the adaptive structure element.

[0009] Step S5: Remove the optimized gingival margin line patches from the three-dimensional model to obtain clear and independent crown and gingiva regions. After removing the gingival margin line, the system uses a connected component detection algorithm to identify and segment independent crown regions, assign different colors to each crown region for differentiation, and save them separately as needed.

[0010] Step S6: By analyzing the morphological features and positions of the connected components, classify each connected component into an independent crown or gingiva region to complete the segmentation, and save them as different three-dimensional model files respectively.

[0011] The technical problem to be solved by the present invention can also be further realized by the following technical solution. The adaptive curvature analysis method in Step S1 includes the principal curvature analysis method in surface analysis, and includes the following steps.

[0012] First, through statistical analysis of the curvature of the entire model, obtain the mean value μ of the curvature k and the standard deviation σ k , and set a preliminary basic threshold T s : T s =μ k +λ s σ k

[0013] where λ s is an adjustment parameter;

[0014] Then, in combination with the curvature gradient, further analyze the local area of the model. By calculating the curvature gradient of each vertex evaluate the curvature change rate, and calculate the mean value and the standard deviation of the gradient. Based on the dynamic threshold T of the gradientg Expressed as:

[0015] Finally, the dynamic curvature threshold T d is obtained by combining curvature statistics and curvature gradient calculation: T d = T s + αT g , where α is a balance factor used to adjust the contributions of statistics and gradient in the threshold.

[0016] The technical problem to be solved by the present invention can also be further realized by the following technical solution. The mean value μ k and standard deviation σ k of the curvature are calculated as The mean value of the gradient and standard deviation are calculated as

[0017] The technical problem to be solved by the present invention can also be further realized by the following technical solution. In the manual interactive correction method in step S2, after the system initially automatically identifies the gingival margin line area based on the adaptive curvature algorithm, the curvature values of the vertices related to the gingival margin line are dynamically adjusted by means of mouse interaction ray picking during the manual editing process.

[0018] The technical problem to be solved by the present invention can also be further realized by the following technical solution. The region growing algorithm in step S3 adopts the depth-first search (DFS) method, traverses the vertices through a stack structure, and is set to only select the vertices that meet the curvature conditions for expansion.

[0019] The technical problem to be solved by the present invention can also be further realized by the following technical solution. The depth-first search (DFS) method includes adding a vertex to the "similar vertex" set when a certain vertex and its neighboring vertices meet the curvature threshold; correspondingly, the patches where these vertices are located will also be added to the "similar patch" set, and finally the vertices and corresponding patches that meet the conditions are automatically selected.

[0020] The technical problem to be solved by the present invention can also be further realized by the following technical solution. The morphological processing in step S4 includes, for each local area, using an adaptively generated structural element to perform dilation first and then erosion to fill the holes and breaks in the gingival margin line, and then after realizing adaptive morphological closing, performing B-spline curve fitting on the extracted vertices of the gingival margin line.

[0021] The technical problem to be solved by the present invention can also be further realized through the following technical solutions. The connected component detection algorithm in step S5 includes traversing each vertex of the model in the graph by constructing an adjacency graph. On the constructed adjacency graph, the breadth-first search (BFS) algorithm is used to traverse all vertices and identify the vertices and patches within the same connected component.

[0022] The present invention has the following advantages compared with the prior art:

[0023] (1) By combining curvature analysis and a small amount of manual interaction, the present invention can not only adapt to diverse tooth shapes but also maintain a high segmentation accuracy in the face of anomalies or noise effects, featuring high robustness and adaptability. Thus, it provides a more efficient and accurate model data processing method for the field of oral digital medicine.

[0024] (2) The present invention does not rely on a large amount of labeled data for training. Only through curvature analysis, geometric shape processing, and user interaction, it can accurately identify and segment the crown area. Finally, the intraoral scan model after segmentation can be independently saved in the OBJ or PLY format according to the three parts of the crown, gingival margin line, and gingiva, supporting further digital restoration and surgical planning.

[0025] (3) The interactive segmentation method based on adaptive curvature of the present invention realizes precise control of the gingival margin line morphology and can accurately segment and independently save the crown and gingiva areas, with high segmentation accuracy and stability, especially suitable for complex tooth shapes and scan data. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is the overall functional schematic diagram of the intraoral scan interactive segmentation system of the present invention;

[0027] Figure 2 is the visualization schematic diagram of the adaptive curvature processing of the present invention;

[0028] Figure 3 is the schematic diagram of the growth of the gingival margin line area with adaptive curvature of the present invention;

[0029] Figure 4 is the schematic diagram of the optimization of the intraoral scan gingival margin line morphology of the present invention;

[0030] Figure 5 is the schematic diagram of the result comparison of the defects between this method and the traditional deep learning method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] The technical solutions of the present invention will be further described below with reference to the accompanying drawings.

[0032] In the specific implementation process, in order to verify the robustness and applicability of the method of the present invention, the embodiments of the present invention specifically selected an intraoral scanning model of edentulous jaws for experiments. Compared with a healthy intraoral scanning model, this model has more general applicability and challenges, and can more significantly show the effect differences of different methods when dealing with complex shapes, so as to better evaluate the performance of the present invention in extreme cases.

[0033] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the field to which the present invention belongs.

[0034] The overall structure of the system is as Figure 1 shown. After importing the three-dimensional model, the system adopts the principal curvature analysis method to calculate the curvature value of each vertex, and enhances the visual contrast of the curvature through a custom color mapping. This process lays the foundation for subsequent gingival margin line extraction and crown segmentation. According to the curvature threshold, the system automatically identifies and extracts the gingival margin line part. For the discontinuous or misjudged areas that may occur during the segmentation process, the system provides a manual editing tool, and the user can correct the inaccurate parts through the interaction interface to ensure the continuity and accuracy of the gingival margin line. After removing the gingival margin line, the system uses the connected component detection algorithm to identify and segment the independent crown areas. Each crown area is assigned a different color for distinction and can be saved separately as needed. After the final segmentation is completed, the user can choose to save the result as an independent three-dimensional model file (such as OBJ, PLY format) for subsequent digital restoration, surgical simulation or other application scenarios. The above process involves processing method steps, mainly including the adaptive curvature analysis method, the manual interactive correction method, the segmentation key area extraction method, the segmentation key area extraction method, the shape optimization based on curve fitting, and the connected area extraction and model saving.

[0035] 1. Adaptive curvature analysis method. In surface analysis, the principal curvature is used to describe the local curvature of the surface. In the intraoral scanning model, key areas such as the gingival margin line are usually accompanied by significant changes in curvature. The minimum principal curvature can effectively highlight these areas and better reflect the subtle shape changes on the crown surface. Therefore, for the convenience of subsequent segmentation and processing, the present invention selects the minimum principal curvature of each vertex. First, through the statistical analysis of the curvature of the entire model, the mean value μ k and the standard deviation σ k of the curvature are obtained.

[0036]

[0037] Based on this, a preliminary basic threshold T s is set:

[0038] T s = μ k+λ s σ k (3)

[0039] where λ s is an adjustment parameter.

[0040] Then, in combination with the curvature gradient, further analysis is performed on the local area of the model. By calculating the curvature gradient of each vertex the rate of change of curvature is evaluated, and the mean value of the gradient and the standard deviation

[0041]

[0042] The dynamic threshold T g based on the gradient is expressed as:

[0043]

[0044] The final dynamic curvature threshold T d is obtained by combining curvature statistics and curvature gradient calculation, where α is a balance factor used to adjust the contributions of statistics and gradient to the threshold.

[0045] T d = T s + αT g (7)

[0046] This adaptive dynamic adjustment method ensures that the segmentation threshold can be automatically adjusted under diverse tooth morphologies and complex surface structures, optimizing the segmentation accuracy and stability.

[0047] Figure 2 The visualization effect of the dynamic curvature threshold is shown.

[0048] 2. Manual interactive correction method. Using an algorithm based on adaptive curvature, the gingival margin line area is initially automatically recognized. The manual editing process dynamically adjusts the curvature values of the vertices related to the gingival margin line by means of mouse interaction ray picking. This interaction mode is similar to the brush function. Through click and drag operations, the user can correct the boundary of the gingival margin line. Here, only the vertices related to the gingival margin line need to be corrected, and other non-related vertices (such as the edges of the grooves formed at the soft tissues of the upper and lower jaws) do not need to be corrected. This real-time visualization operation not only improves the user experience but also prepares for subsequent vertex and patch filtering based on the curvature threshold.

[0049] 3. Segmentation key area extraction method. The user arbitrarily selects a seed point on the gingival margin line of the model. Based on the seed point, the system starts the region growing algorithm to extract the complete gingival margin line by gradually expanding the adjacent vertices and patches that meet the curvature threshold. The region growing algorithm uses the depth-first search (DFS) method to traverse the vertices through the stack structure and ensure that only vertices that meet the curvature conditions are selected for expansion. Specifically, when a vertex and its adjacent vertices meet the curvature threshold, these vertices will be added to the "similar vertices" set; accordingly, the patches where these vertices are located will also be added to the "similar patches" set, and finally the vertices and corresponding patches that meet the conditions will be automatically screened out. Define the seed vertex selected by the user as v s , the initial vertex set is V0, which contains the seed vertex: V0 = {v s}, the initial patch set is empty:

[0050] For vertex v i Each neighbor node v j , if v j Satisfying the curvature condition k(v j )≤T(where T d is the dynamically adjusted curvature threshold), then add it to the growth region:

[0051] V1=V0∪{v j ∣k(v j )≤T d} (8)

[0052] Accordingly, if v j With v i Adjacent patches f j If the conditions are met, add the face set:

[0053] F1=F0∪{f j ∣v j ∈V1} (9)

[0054] The search is iterated until no new vertices or faces meet the conditions. The final set of vertices and faces is represented by V f and F f .

[0055] Figure 3 Shown is the effect of curvature-corrected gingival margin area growth.

[0056] 4. Morphological optimization based on curve fitting. This step is the final step of extracting the key segmentation area. Since the gingival margin line is the key information for independent segmentation of the crown, we need to further optimize the gingival margin line morphology. The present invention adopts an adaptive curve fitting morphological processing method to achieve precise control of the gingival margin line morphology.

[0057] Morphological processing mainly includes dilation, erosion, opening, and closing. These operations depend on the shape and size of the structuring element (SE). Here, an adaptive structuring element is introduced to dynamically adjust the parameters of the structuring element according to local curvature information, improving the flexibility and adaptability of morphological operations. Let the scale s of the structuring element i be related to the vertex curvature k(v i ):

[0058]

[0059] where the scale s of the structuring element i is represented by the number of adjacent triangles at that vertex, β is the proportionality coefficient, and ∈ is a small constant to prevent division by zero. Specifically, the larger the curvature, the smaller the scale of the structuring element to retain details; the smaller the curvature, the larger the scale of the structuring element to achieve a smoothing effect.

[0060] For each local region, use the adaptively generated structuring element to perform dilation or erosion operations to smooth the boundary of the gingival margin line and fill in breaks.

[0061] Dilation operation:

[0062]

[0063] Erosion operation:

[0064]

[0065] First perform dilation and then erosion to fill in small holes and breaks in the gingival margin line:

[0066]

[0067] After implementing adaptive morphological closing, perform Bezier curve fitting on the extracted vertices of the gingival margin line, and use the least squares method to optimize and adjust the positions of the control points to make the Bezier curve fit the actual gingival margin line as closely as possible. The optimization objective function can be defined as:

[0068]

[0069] Adaptive morphological processing uses local curvature information to dynamically adjust the structuring element, achieving local smoothing and detail retention; Bezier curve fitting ensures the global curve is smooth and accurate through control point optimization.

[0070] Such as Figure 4As shown, during the gingival margin line extraction process of the present invention, high-precision control of the crown morphology is achieved, ensuring the accuracy and visual consistency of the segmentation results.

[0071] 5. Connectivity region extraction and model saving. First, the optimized gingival margin line patches are removed from the 3D model, and clear and independent crown and gingiva regions can be obtained. By constructing an adjacency graph and traversing each vertex in the graph (i.e., the vertices of the model), on the constructed adjacency graph, using the breadth-first search (BFS) algorithm to traverse all vertices and identify the vertices and patches within the same connected component. These connected components represent independent crowns or gingiva. By analyzing the morphological features and positions of the connected components, each connected component is classified as an independent crown or gingiva region and saved as different 3D model files respectively for subsequent use.

[0072] In summary, the above-mentioned interactive segmentation method based on adaptive curvature provided by the present invention mainly reflects its technical effects through the following aspects, including:

[0073] (1) Global curvature calculation and depth analysis: By performing global curvature calculation and analysis on the entire 3D model, the subtle features on the tooth surface are deeply identified and visualized as a curvature map, laying a foundation for subsequent precise segmentation;

[0074] (2) Precise gingival margin line region identification: Combining global curvature analysis and morphological processing techniques, an adaptive method is used to accurately identify the key segmentation regions. Even in tooth models with complex or abnormal shapes, high-precision segmentation can still be achieved;

[0075] (3) User interaction optimization and correction: Provide highly intuitive interaction tools, and users can manually fine-tune the boundaries of the gingival margin line and the crown in the visualization interface. Through a small amount of semi-automatic interaction, noise or complex topological structures are effectively processed to ensure the continuity and accuracy of the segmentation results;

[0076] (4) Output of segmentation results: Through the analysis of morphological and position features, the independent crown and gingiva regions are saved in multiple 3D formats respectively for digital restoration, surgical simulation and other clinical applications.

[0077] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, various changes and modifications can be made to the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intraoral scanning interactive segmentation method based on adaptive curvature, characterized in that: It includes the following steps: Step S1: Import the 3D model into the system. The system calculates using the adaptive curvature analysis method, calculates the curvature value of each vertex, enhances the visual contrast of the curvature through a custom color mapping, and automatically identifies and extracts the gingival margin line part according to the curvature threshold. Step S2: For the discontinuous or misjudged areas that may occur during the segmentation process in Step S1, the system provides a manual interactive correction method editing tool. The user can correct the inaccurate parts through the interface to ensure the continuity and accuracy of the gingival margin line. Step S3: The user uses the segmentation key area extraction method to arbitrarily select a seed point on the gingival margin line of the model. Based on this seed point, the system starts the region growing algorithm, and extracts the complete gingival margin line by gradually expanding the adjacent vertices and patches that meet the curvature threshold. Step S4: The system achieves precise control of the gingival margin line morphology through an adaptive curve fitting morphological processing method, and further optimizes the gingival margin line morphology. The morphological processing includes dilation, erosion, opening operation, and closing operation based on introducing the shape and size of the adaptive structure element. Step S5: Remove the optimized gingival margin line patches from the 3D model, and a clear and independent crown and gingiva area can be obtained. After removing the gingival margin line, the system uses the connected component detection algorithm to identify and segment the independent crown areas, assigns different colors to each crown area for distinction, and can be saved separately as needed. Step S6: By analyzing the morphological characteristics and positions of the connected components, classify each connected component as an independent crown or gingiva area to complete the segmentation, and save them as different 3D model files respectively.

2. The intraoral scanning interactive segmentation method based on adaptive curvature according to claim 1, wherein: The adaptive curvature analysis method in Step S1 includes the principal curvature analysis method in surface analysis, and includes the following steps: First, by statistically analyzing the curvature of the entire model, the mean value μ of the curvature is obtained. k and standard deviation σ k , set the initial basic threshold T s : T s = μ k + λ s σ k where λ s is an adjustment parameter; Then, in combination with the curvature gradient, further analyze the local area of the model by calculating the curvature gradient of each vertex Evaluate the rate of change of curvature and calculate the mean value of the gradient and the standard deviation The dynamic threshold T based on the gradient g is expressed as: Finally, the dynamic curvature threshold T d is obtained by combining curvature statistics and curvature gradient calculation: T d = T s + αT g , where α is a balance factor used to adjust the contributions of statistics and gradients to the threshold.

3. The intraoral scanning interactive segmentation method based on adaptive curvature according to claim 2, wherein: The mean value μ of the curvature k and the standard deviation σ k are calculated as The mean value of the gradient and the standard deviation are calculated as 4. An intraoral scanning interactive segmentation method based on adaptive curvature according to claim 1, characterized in that: The manual interactive correction method in Step S2 dynamically adjusts the curvature value of the vertices related to the gingival margin line by means of manual editing and mouse interaction ray picking after the system initially automatically identifies the gingival margin line area using the adaptive curvature algorithm.

5. The intraoral scanning interactive segmentation method based on adaptive curvature according to claim 1, wherein: The region growing algorithm in Step S3 adopts the depth-first search (DFS) method, traverses the vertices through a stack structure, and is set to only select the vertices that meet the curvature condition for expansion.

6. The intraoral scanning interactive segmentation method based on adaptive curvature according to claim 5, wherein: The depth-first search (DFS) method includes adding the vertex to the "similar vertex" set when a certain vertex and its adjacent vertices meet the curvature threshold; correspondingly, the patches where these vertices are located will also be added to the "similar patch" set, and finally the vertices and corresponding patches that meet the conditions are automatically selected.

7. A method for intraoral scanning interactive segmentation based on adaptive curvature according to claim 1, characterized in that: The morphological processing in Step S4 includes, for each local area, using the adaptively generated structure element to perform dilation first and then erosion to fill the holes and breaks in the gingival margin line, and then after realizing the adaptive morphological closing, performing Bezier curve fitting on the vertices of the extracted gingival margin line.

8. An intraoral scanning interactive segmentation method based on adaptive curvature according to claim 1, characterized in that: The connected component detection algorithm in Step S5 includes traversing each vertex of the model in the graph by constructing an adjacency graph, and using the breadth-first search (BFS) algorithm on the constructed adjacency graph to traverse all vertices and identify the vertices and patches within the same connected component.

Images