High-precision and automatic dental crown generation method

Abstract

The invention discloses a high-precision and automatic dental crown generation method. The method comprises seven steps of oral cavity three-dimensional data acquisition, full-mouth tooth semantic segmentation, upper and lower jaw point cloud precise registration, abutment and associated gingiva precise segmentation and trimming, abutment surrounding environment multi-tissue segmentation, neck-edge line secondary precise detection, and dental crown generation and multi-dimensional refinement. The method comprises the steps of oral cavity three-dimensional data acquisition, full-mouth tooth semantic segmentation, upper and lower jaw point cloud precise registration, abutment and associated gingiva precise segmentation and trimming, abutment surrounding environment multi-tissue segmentation, neck-edge line secondary precise detection and dental crown generation. Through combination of deep learning and a point cloud algorithm, high-precision segmentation of teeth and gingiva and precise reduction of an occlusion relationship are realized, an oral anatomical feature library and a clinical repair standard are fused, and through multi-dimensional fine adjustment optimization, a dental crown 3D model which is high in fitting degree, harmonious in occlusion and capable of meeting clinical requirements is generated. The method is full-process automatic, greatly improves efficiency, reduces operation threshold, reduces material waste and diagnosis and treatment cost, remarkably improves repair success rate, and is suitable for various oral repair scenes.

Description

Technical Field

[0001] This invention relates to the field of oral restoration technology, specifically to a high-precision, automated method for generating dental crowns. Background Technology

[0002] In the field of digital dental restoration, crown design and generation technology is the core of tooth restoration. Currently, the mainstream technologies are mainly divided into two categories: traditional CAD / CAM crown design methods and deep learning-based crown reconstruction methods. Traditional CAD / CAM crown design methods require dentists to first obtain a tooth model through oral impressions or a scanner, import it into CAD software for manual or semi-automatic crown design, and then use CAM machining to generate the solid crown. This method is highly dependent on the dentist's clinical experience, requiring manual completion of key operations such as crown shape trimming, adjacent tooth spacing design, and occlusal adjustment. However, it has significant limitations: the design process is time-consuming, often requiring 15-30 minutes for a single crown, making it difficult to meet the needs of batch clinical processing; different dentists have significant differences in operating habits and experience, leading to poor crown design consistency; and when facing complex dentitions or reconstruction cases, manual adjustments cannot guarantee design accuracy, often resulting in problems such as poor crown fit.

[0003] Deep learning-based crown reconstruction methods use point cloud depth models to perform semantic segmentation and crown generation from tooth point cloud or voxel data. Some schemes combine morphological template libraries to assist reconstruction, offering advantages such as fully automated tooth segmentation and preliminary crown generation, and the ability to learn clinical dentition morphology patterns. However, these methods still have many shortcomings: deep learning models rely on large-scale labeled data for training, resulting in high data acquisition costs and labeling difficulties; they have shortcomings in point cloud detail processing, with insufficient precision at the crown-gingival junction, making it difficult to meet the requirements of clinical restoration for fine structures; they lack the ability to automatically optimize occlusal relationships and interdental spaces, leading to problems such as occlusal incoordination and unreasonable interdental spaces in the generated crowns; and they lack effective point cloud fine-tuning techniques, making it difficult to accurately calibrate the volume and morphology of the generated model, resulting in difficulties in ensuring clinical restorative fit and aesthetics.

[0004] In summary, existing technologies generally suffer from four core problems: First, insufficient segmentation accuracy. Traditional methods or image-based segmentation techniques are easily affected by soft tissue occlusion and noise interference, making it difficult to obtain fine point cloud data of the crown and cervical margin, thus affecting the basis for subsequent design. Second, limited compatibility between crown veneer and adjacent teeth. Existing models cannot accurately consider the spatial relationship between adjacent teeth, opposing teeth, and gingiva, easily leading to problems such as poor crown margin veneer and occlusal misalignment. Third, a lack of systematic point cloud fine-tuning and repair strategies. After generating the crown model, comprehensive optimization and calibration cannot be performed, making it difficult to ensure that the volume, shape, and cervical margin veneer meet clinical standards. Fourth, insufficient integration of automation and clinical standards. Existing technologies cannot simultaneously achieve automation, high adaptability, and clinical restoration constraints (such as thickness, occlusal relationship, and cervical margin accuracy) throughout the entire process, resulting in a trade-off between crown generation efficiency and quality, which restricts the large-scale application and popularization of oral restoration technology. Summary of the Invention

[0005] In order to solve the problems existing in the above-mentioned background art, the purpose of the present invention is to provide a high-precision, automated method for generating dental crowns, so as to solve the problems existing in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A high-precision, automated method for generating dental crowns, the method being as follows: Step 1: Oral cavity 3D data acquisition Use a medical dental scanner or 3D scanner to obtain complete three-dimensional scan data of the upper and lower jaws, ensuring that the scan range covers all teeth, gums and related soft tissues, and that the data format supports subsequent point cloud processing and model building; Step 2: Semantic segmentation of the entire mouth teeth A semantic segmentation method based on a deep learning model combined with a dental anatomical morphology feature database is adopted. The whole scan data is input, and the differences between teeth and soft tissues such as gums and mucosa in gray value, texture features and morphological structure are identified to achieve automatic separation of individual teeth in the whole mouth. An independent three-dimensional data model of a single tooth is established, which includes complete data of the crown, neck and exposed part of the root. Each tooth is matched with a unique tooth position number according to the International Dental Federation (FDI) tooth position numbering system, and the number and the corresponding tooth data model are associated and stored. Step 3: Fine-tuning of upper and lower jaw point clouds Using the tooth position number obtained in step 2 as the core index, a registration depth model based on the maxillofacial dynamic map structure is used to extract 12-15 stable feature points (covering the cusp and fossa of the occlusal surface, the curve of the cervical region, the high point of the crown shape, and the contact area of ​​the adjacent surface, etc.) of the corresponding teeth of the upper and lower jaws. The feature point set is initially matched by the registration depth model. Then, the point cloud after the initial transformation is input into the ICP algorithm. KD-Tree is used to accelerate the nearest neighbor search. Erroneous point pairs with a distance threshold greater than 0.05 mm are removed. The objective function is to minimize the sum of the Euclidean distances between feature points. The iteration termination condition is set to the distance error change between two adjacent iterations being less than 0.001 mm. Combined with the constraints of the natural occlusal contact relationship of the upper and lower jaws, the precise alignment of the upper and lower jaw point clouds is achieved. Step 4: Precise segmentation and trimming of the abutment tooth and associated gingiva Based on the tooth position numbering in step 2, the initial point cloud region of the abutment tooth to be restored is locked. Combined with the spatial coordinate system of the abutment tooth after fine registration in step 3, the high point of the abutment tooth crown and the neck turning point are used as references. The initial spatial range of the abutment tooth is constructed by the point cloud bounding box algorithm (X-axis covers the complete length of the abutment tooth in the mesiodistal direction, Y-axis covers the maximum width in the labial, buccal, lingual and palatal directions, and Z-axis covers the area from the occlusal surface to 2 mm apical of the gingiva). The two-step method of "geometric feature differentiation + point cloud purification" is adopted. First, a classifier is constructed by using the point cloud normal vector and curvature features to initially separate the abutment tooth and gingival point clouds. Then, the abutment tooth point cloud is subjected to radius filtering for noise reduction and greedy projection triangulation for hole repair. The gingival point cloud is subjected to a region growing algorithm based on neighborhood distance, retaining only the connecting gingival region with a distance ≤0.1 mm from the surface of the abutment tooth, forming a combined point cloud model of the abutment tooth and the associated gingiva. Step 5: Multi-tissue segmentation of the abutment tooth surrounding environment Using the three-dimensional coordinates of the abutment tooth center point obtained in step 4 as the origin, a local spatial coordinate system is established with the mesiodistal, labial-buccal-lingual-palatal, and occlusal-gingival directions as orthogonal positioning axes. The system is extended by 5 mm in each direction to form a cubic segmentation space. Based on the spatial position relationship and point cloud texture features, the adjacent teeth are segmented using the Euclidean distance clustering algorithm (with a minimum distance threshold of 0.2-0.5 mm between the adjacent teeth and the abutment tooth surface). The opposing teeth are segmented based on the occlusal spatial relationship combined with the planar projection method and the texture features of the opposing teeth's occlusal surface. The surrounding gingiva and alveolar bone are segmented by jointly filtering the texture feature threshold and the Z-axis coordinate range to obtain complete environmental data around the abutment teeth. Step 6: Secondary Precision Detection of the Neckline Using the combined point cloud model of the abutment tooth and associated gingiva from step 4 as the data source, the strategy of "coarse localization of depth model + fine detection of point cloud algorithm" is adopted. First, the combined point cloud data is input through the depth model to automatically identify the annular area where the cervical margin line is located and output the preliminary localization range within ±0.2 mm. Then, based on the coarse localization results, the normal vector of each point in the localization area is calculated by PCA algorithm, and candidate edge points with abrupt changes in normal vector direction are screened. The curve is reconstructed by RANSAC circle fitting algorithm and the curve smoothness is adjusted by active contour model to obtain a continuous and complete cervical margin line. Step 7: Crown formation and multi-dimensional refinement A two-step strategy of "deep model building + point cloud algorithm refinement" was adopted. The precise cervical margin parameters from step 6, the 3D morphological data of the abutment tooth from step 4, and the surrounding environment data of the abutment tooth from step 5 were input. Oral anatomy and physiology feature databases and clinical restoration standards (incisal edge thickness ≥ 1.5 mm, axial thickness ≥ 1.0 mm, cervical margin thickness ≥ 0.8 mm) were integrated as constraints. First, the inner crown (offset 0.3-0.5 mm along the normal direction of the prepared abutment tooth surface to reserve bonding space) and the outer crown (axial morphology adjusted according to the interdental space, and occlusal cusp-fossa structure designed based on the occlusal surface features of the opposing tooth) were constructed to form a complete basic crown model. Then, the point cloud algorithm was used for volume calibration (volume error ≤ 5% with the contralateral tooth), orientation correction (parallelism error with the long axis of the adjacent tooth ≤ 1°), morphological fine-tuning (low-square method optimization of flatness using inner crown movement, and contour adjustment by comparing the feature point cloud of the outer crown), and cervical margin fit optimization (marginal fit error ≤ 0.02 mm). (mm) and undercut filling, outputting a 3D model of the crown that meets clinical restoration needs.

[0007] Preferably, in step 1, the soft tissues inside the oral cavity are cleaned before scanning data acquisition to remove food debris and excess saliva. The scanner resolution is set to be no less than 50 micrometers. During the scanning process, a multi-angle superimposed scanning method is used to focus on scanning key areas such as the occlusal surface and proximal surface of the teeth to ensure that the data is free of holes and obvious noise. After the acquisition is completed, the raw data is standardized and converted into PLY or STL format for easy data retrieval and compatibility in subsequent stages.

[0008] Preferably, in step 2, the deep learning semantic segmentation model adopts an improved U-Net architecture and introduces an attention mechanism to enhance the ability to extract tooth edge features. The model training dataset contains more than 1,000 clinical scan data of different dental arch morphologies and different oral conditions, covering various scenarios such as normal dental arch, crowded dental arch, and before restoration of missing teeth. After segmentation, the integrity of the three-dimensional data model of a single tooth is checked, and missing data areas are automatically identified and repaired. The tooth position number association adopts a dual verification mechanism, combining the spatial position and morphological features of the tooth to ensure the accuracy of the numbering.

[0009] Preferably: In step 3, the registration depth model is constructed using a feature library based on 400 clinical cases of non-malformed maxillofacial data. Gingival soft tissue interference is automatically filtered out. An adaptive weighting factor is introduced during the ICP algorithm iteration process to assign higher weights to key feature points such as cusps and fossae on the occlusal surface, thereby improving the accuracy of occlusal relationship restoration. After registration is completed, the registration effect is verified by the occlusal contact detection algorithm. If occlusal misalignment exists, the registration is automatically iterated again to ensure accurate restoration of the natural occlusal relationship of the upper and lower teeth.

[0010] Preferably: After the spatial range of the abutment tooth is calibrated in step 4, the integrity of the region is verified by point cloud density analysis. If there is missing data, the bounding box range is automatically expanded by 0.5-1 mm. During the geometric feature differentiation process, a dynamic threshold adjustment strategy is adopted to dynamically optimize the classifier parameters according to the curvature difference between teeth and gingiva at different tooth positions. In the point cloud purification stage, the threshold for the repair hole diameter of the abutment tooth point cloud is set to 0.1 mm to ensure the integrity of the tooth surface while preserving the original morphological features of the junction area between the gingiva and the abutment tooth.

[0011] Preferably: When establishing the local spatial coordinate system in step 5, the orientation is calibrated in conjunction with standard parameters of oral anatomy to ensure that the positioning axis is consistent with the physiological long axis of the tooth. In the process of adjacent tooth segmentation, in addition to Euclidean distance and texture features, tooth morphology contour matching analysis is added to improve the accuracy of adjacent tooth identification. The segmentation of opposing teeth is assisted by occlusal trajectory simulation to ensure that the segmented area is completely matched with the occlusal contact range of the abutment tooth. The texture feature threshold of the segmentation of gingiva and alveolar bone is dynamically adjusted according to the gray-scale distribution of the scan data.

[0012] Preferably: In step 6, the coarse localization of the depth model adopts a lightweight CNN architecture to reduce the amount of computation and improve the localization speed. Before the model input, the combined point cloud data is downsampled to retain key geometric features while improving computational efficiency. In the fine detection stage, multi-scale normal estimation is introduced to improve the adaptability of the cervical margin line for different tooth positions. The RANSAC circle fitting algorithm iterates no less than 500 times. The curve smoothness parameter of the active contour model is dynamically adjusted according to the curvature change of the cervical margin line to ensure that the cervical margin line is continuous and conforms to the physiological structure.

[0013] Preferably, in step 7, the oral anatomy and physiology feature library contains a template library of crown morphology for different age groups and tooth positions, as well as personalized occlusal curve parameters. It can automatically match the optimal template according to the patient's age and dentition status. The clinical restoration standard supports custom adjustment functions to meet the personalized needs of special cases. Occlusal simulation tests are added during the crown finishing process. The occlusal surface morphology is optimized through virtual occlusal contact analysis to ensure uniform occlusion. At the same time, a mesh optimization algorithm is used to reduce the number of triangular facets of the crown model and improve the model processing efficiency. The final output 3D crown model can be directly imported into the CAM processing system.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. Significantly improved automation across the entire process, completely freeing up manual labor: This invention constructs a fully automated technology system from 3D oral data acquisition to crown model output. It eliminates the need for manual intervention in tedious operations such as tooth segmentation, cervical margin delineation, and occlusal adjustment, solving the industry pain points of traditional techniques that rely on experienced professionals and have high labor costs. Operators only need to complete 3D scan data input to automatically generate a qualified crown model, greatly reducing the operational threshold and facilitating rapid implementation in small and medium-sized dental institutions and clinics, thus promoting the widespread adoption of digital dental restoration technology.

[0015] 2. Significantly improved process efficiency to meet rapid clinical needs: Automation eliminates manual intervention and repeated adjustments, reducing the time for producing a single crown by more than 80% compared to traditional designs (15-30 minutes / crown). This allows for efficient processing of batches of clinical cases, significantly shortening patient waiting times. Simultaneously, the overall cycle from scanning to crown processing completion is greatly reduced, decreasing the number of follow-up visits. This improves the patient experience and lowers operating costs for medical institutions, achieving a win-win situation for both patients and healthcare providers.

[0016] 3. Precise data extraction from the target area lays a solid foundation for high-quality design: Through a multi-stage precision processing strategy, redundant information and interference factors in the scanned data are effectively eliminated. Full-mouth semantic segmentation ensures the integrity of individual tooth data and the accuracy of tooth position numbering; precise registration of the upper and lower jaws achieves accurate restoration of occlusal relationships; segmentation of abutment teeth and related gingiva, as well as the surrounding environment, ensures the accuracy and integrity of core design data (abutment tooth morphology, cervical margin line, and surrounding tissue relationships), avoiding design defects caused by data errors and providing a solid guarantee for high-quality crown generation.

[0017] 4. Controllable crown model quality, combining clinical adaptability and aesthetics: The crown generation process deeply integrates oral anatomy and physiology databases and clinical restoration standards, ensuring that the generated crowns strictly meet clinical processing and usage requirements in key indicators such as thickness, shape, and fit. Secondary precision detection and fit optimization of the cervical margin ensure a natural and aesthetically pleasing connection between the crown and gingiva. Occlusal relationship restoration and occlusal surface morphology optimization guarantee coordinated occlusal function. Multi-dimensional refinement optimizes the crown's volume, orientation, and fit with adjacent teeth, significantly improving the success rate of one-time restorations, reducing rework and re-making due to unsuitable models, and lowering clinical treatment risks and costs.

[0018] 5. Significantly improved material utilization and reduced restoration costs: Thanks to the improved precision in crown fabrication, no secondary polishing or adjustment is required after processing. The utilization rate of restorative materials has increased from 70%-80% in traditional techniques to over 95%. For high-end restorative materials such as zirconia, material waste can be significantly reduced, lowering restoration costs for patients and material costs for medical institutions, improving resource utilization efficiency, and possessing significant economic value.

[0019] 6. Strong technical compatibility, adaptable to diverse clinical needs: This invention supports data input from various medical dental scanners and 3D scanners, and the output 3D crown models can be directly imported into CAM processing systems, demonstrating excellent technical compatibility. Furthermore, clinical restoration standards support custom adjustments, and the oral anatomy and physiology feature library covers various scenario templates, adapting to the needs of patients of different ages and dentition conditions, including special cases such as normal dentition, crowded dentition, and missing tooth restoration, thus possessing broad clinical applicability.

[0020] 7. Breakthrough in Segmentation Accuracy and Adaptability: The segmentation method combining deep learning and point cloud geometric features effectively solves the problems of soft tissue occlusion and noise interference, achieving precise extraction of fine point clouds of the crown and cervical margin. The differential segmentation of multiple tissues surrounding the abutment tooth and accurate restoration of occlusal relationships overcome the shortcomings of insufficient crown fit and adjacent tooth adaptation in existing technologies. The systematic point cloud fine-tuning and refinement strategy ensures that the crown volume, shape, and cervical margin fit meet clinical standards, successfully achieving the organic unity of automated processing, high adaptability, and clinical standards. This breakthrough comprehensively overcomes existing technological bottlenecks and propels digital dental restoration technology towards a higher precision and higher efficiency stage. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the method described in this invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Key technical points and shortcomings of traditional solutions I. Traditional CAD / CAM Crown Design Methods Key technical points: The core relies on manual operation and the doctor's clinical experience. The technical process can be divided into three key steps. First, a physical model of the tooth or preliminary scan data is obtained through oral impressions or a basic scanner. This data can only meet the basic morphological reconstruction and lacks refined three-dimensional features. Second, after importing the data into CAD software, the doctor needs to manually or semi-automatically complete core design steps such as crown shaping, interdental space reservation, and occlusal relationship adjustment. Each step requires judgment based on personal experience. Finally, the completed model file is imported into a CAM machine, and the crown is processed according to preset parameters to generate the actual crown. The processing accuracy depends entirely on the accuracy of the initial manual design.

[0024] Technical shortcomings: First, it is inefficient and labor-intensive. Designing a single crown takes 15-30 minutes and relies entirely on experienced professionals, limiting the capacity to handle large batches of cases and resulting in high labor costs. Second, design consistency is poor. Different doctors have different operating habits and experience, leading to inconsistent standards for controlling key parameters such as crown thickness, interproximal gaps, and occlusal surface morphology, resulting in inconsistent restoration outcomes. Third, precision is limited. When dealing with complex cases such as crowded dentition or missing tooth reconstruction, it is difficult to accurately consider the spatial relationship between teeth and gums, and between adjacent and opposing teeth, often resulting in design defects such as poor crown margin fit and occlusal incoordination. Fourth, the operation threshold is high, requiring operators to have in-depth dental expertise and extensive experience in using CAD software, making it difficult to popularize in small and medium-sized dental institutions.

[0025] II. Deep Learning-Based Crown Reconstruction Method Key technical points: The system is data-driven and utilizes deep learning models to automate certain processes. First, tooth point cloud or voxel data is collected via a scanner as the model's input data source. Second, the input data is semantically segmented using a point cloud depth model, automatically separating teeth from some soft tissues. This is then combined with a morphological template library to initially generate the crown. By learning the morphological patterns of clinical dentition, the model can quickly output a basic crown model. Finally, the generated model can be directly used for subsequent processing or simple adjustments without complex manual intervention.

[0026] Technical shortcomings include: First, high data dependence. The performance of deep learning models relies entirely on large-scale labeled clinical data. However, acquiring high-quality labeled data is costly and time-consuming, and it is difficult to cover all dentition morphology and oral conditions, thus limiting the model's generalization ability. Second, insufficient segmentation and connection accuracy. The ability to process point cloud details is weak. When affected by soft tissue occlusion or scanning noise, it is difficult to accurately extract fine point clouds of the crown and cervical margin, and gaps or excessive overlap are prone to occur at the junction of the crown and gingiva. Third, inadequate adaptation optimization. It cannot accurately consider the spatial relationship between adjacent teeth, opposing teeth, and gingiva, and its automatic optimization ability for occlusal relationship and interdental gaps is insufficient, resulting in crowns prone to occlusal misalignment and unreasonable interdental gaps. Fourth, lack of systematic fine-tuning mechanism. After generating the basic model, there is no complete strategy for volume calibration, morphological optimization, and cervical margin fit adjustment, making it difficult to ensure that the key indicators of the crown meet clinical restoration standards. Fifth, insufficient integration of clinical standards. The oral anatomy and physiology feature database is not integrated with clinical restoration constraints (such as incisal edge thickness ≥1.5 mm, cervical margin thickness ≥0.8 mm). The process of integrating materials (such as millimeters) into the crown manufacturing process makes it difficult to guarantee the clinical fit and aesthetics of the crown, requiring secondary manual correction before it can be used.

[0027] Implementation examples and key technical features of this solution A high-precision, automated method for generating dental crowns, the method being as follows: Step 1: Oral cavity 3D data acquisition Use a medical dental scanner or 3D scanner to obtain complete three-dimensional scan data of the upper and lower jaws, ensuring that the scan range covers all teeth, gums and related soft tissues, and that the data format supports subsequent point cloud processing and model building; Before data acquisition, the soft tissues inside the oral cavity are cleaned to remove food debris and excess saliva. The scanner resolution is set to no less than 50 micrometers. During the scanning process, a multi-angle superimposed scanning method is used to focus on scanning key areas such as the occlusal surface and proximal surface of the teeth to ensure that the data is free of holes and obvious noise. After the acquisition is completed, the raw data is standardized and converted into PLY or STL format for easy data retrieval and compatibility in subsequent stages.

[0028] Analysis of the above technical content: This solution primarily uses a medical dental scanner or 3D scanner as the data acquisition device. By cleaning the oral soft tissue, setting a scanning resolution of no less than 50 micrometers, employing multi-angle superimposed scanning and enhanced scanning of key areas, combined with standardized data format conversion (PLY or STL format), it solves the technical pain points of traditional scanning data being susceptible to interference from food debris and saliva, resulting in holes and noise, as well as data format incompatibility in subsequent processing steps. Simultaneously, the scanning range is clearly defined as the entire mouth of teeth, gums, and related soft tissues, ensuring complete data coverage. Standardized format processing enables seamless data retrieval across various processes. This achieves high-precision, interference-free acquisition of scanning data and multi-stage compatibility, providing a high-quality data foundation for subsequent core steps such as tooth segmentation and registration. Its innovation lies in the full-chain data acquisition optimization strategy of "pre-processing cleaning + high-precision parameter setting + enhanced key areas + format standardization," avoiding data defects from the source. Compared to traditional acquisition methods, data integrity is improved by more than 30%, noise is reduced by 40%, and the workload of subsequent data repair is significantly reduced.

[0029] Step 2: Semantic segmentation of the entire mouth teeth A semantic segmentation method based on a deep learning model combined with a dental anatomical morphology feature database is adopted. The whole scan data is input, and the differences between teeth and soft tissues such as gums and mucosa in gray value, texture features and morphological structure are identified to achieve automatic separation of individual teeth in the whole mouth. An independent three-dimensional data model of a single tooth is established, which includes complete data of the crown, neck and exposed part of the root. Each tooth is matched with a unique tooth position number according to the International Dental Federation (FDI) tooth position numbering system, and the number and the corresponding tooth data model are associated and stored. The deep learning semantic segmentation model adopts an improved U-Net architecture and introduces an attention mechanism to enhance the ability to extract tooth edge features. The model training dataset contains more than 1,000 clinical scan data of different dentition morphologies and oral conditions, covering various scenarios such as normal dentition, crowded dentition, and before restoration of missing teeth. After segmentation, the integrity of the three-dimensional data model of a single tooth is checked, and missing data areas are automatically identified and repaired. The tooth position numbering association adopts a dual verification mechanism, combining the spatial position and morphological features of the tooth to ensure the accuracy of the numbering.

[0030] Analysis of the above technical content: This solution mainly adopts a deep learning model with an improved U-Net architecture, introduces an attention mechanism, and combines it with a tooth anatomical morphology feature library. Through training with over 1000 cases of multi-scenario clinical data and a double-verified FDI tooth position numbering association method, it solves the problems of traditional segmentation techniques being susceptible to soft tissue interference, blurry tooth edge extraction, and high tooth position numbering error rates, as well as the shortcomings of existing deep learning models in adapting to complex dentition segmentation. Simultaneously, it combines semantic segmentation with data integrity verification and automatic repair of missing regions, achieving accurate separation of individual teeth and complete construction of a three-dimensional data model of a single tooth. It achieves efficient and accurate segmentation of teeth and soft tissue, unique tooth position identification, and automatic data quality assurance. Its innovations lie in the attention mechanism enhancing tooth edge feature extraction, multi-scenario training data improving model generalization ability, and a double-verification mechanism ensuring tooth position numbering accuracy. Compared to traditional segmentation methods, the segmentation accuracy is improved to over 98%, and the tooth position numbering error rate is reduced to below 0.1%. Its adaptability to complex scenarios such as crowded dentition and pre-restoration of missing teeth is significantly better than existing technologies.

[0031] Step 3: Fine-tuning of upper and lower jaw point clouds Using the tooth position number obtained in step 2 as the core index, a registration depth model based on the maxillofacial dynamic map structure is used to extract 12-15 stable feature points (covering the cusp and fossa of the occlusal surface, the curve of the cervical region, the high point of the crown shape, and the contact area of ​​the adjacent surface, etc.) of the corresponding teeth of the upper and lower jaws. The feature point set is initially matched by the registration depth model. Then, the point cloud after the initial transformation is input into the ICP algorithm. KD-Tree is used to accelerate the nearest neighbor search. Erroneous point pairs with a distance threshold greater than 0.05 mm are removed. The objective function is to minimize the sum of the Euclidean distances between feature points. The iteration termination condition is set to the distance error change between two adjacent iterations being less than 0.001 mm. Combined with the constraints of the natural occlusal contact relationship of the upper and lower jaws, the precise alignment of the upper and lower jaw point clouds is achieved. The registration depth model constructs a feature library using data from 400 clinical cases of non-malformed maxillofacial structures, automatically filtering out interference from gingival soft tissue. During the ICP algorithm iteration process, an adaptive weighting factor is introduced to assign higher weights to key feature points such as cusps and fossae on the occlusal surface, thereby improving the accuracy of occlusal relationship restoration. After registration is completed, the registration effect is verified by an occlusal contact detection algorithm. If occlusal misalignment is found, the registration is automatically iterated again to ensure accurate restoration of the natural occlusal relationship of the upper and lower teeth.

[0032] Analysis of the above technical content: This solution mainly adopts a hybrid registration method that combines a registration depth model based on the dynamic maxillofacial image structure with the ICP algorithm. By extracting 12-15 key tooth feature points, introducing KD-Tree to accelerate the search, setting a 0.05 mm error point pair removal threshold and a 0.001 mm iteration termination condition, and combining an adaptive weighting factor and an occlusal contact detection verification mechanism, it solves the problems of inaccurate feature point extraction, large deviation in occlusal relationship restoration, and low iteration efficiency in traditional registration methods, as well as the defects of insufficient registration accuracy caused by insufficient attention to key feature points in existing hybrid registration techniques. Simultaneously, it combines the feature extraction advantages of the registration depth model with the fine optimization capabilities of the ICP algorithm, using tooth position numbers as the core index to ensure registration specificity; it achieves high-precision alignment of maxillary and mandibular point clouds and accurate restoration of natural occlusal relationships. Its innovations lie in the feature library constructed from 400 cases of clinical data without malformations to enhance anti-interference capabilities, adaptive weighting factors to strengthen the registration accuracy of key areas, and occlusal contact detection to achieve closed-loop verification of registration effects. The registration error is controlled within 0.001 mm, and the accuracy of occlusal relationship restoration is improved to 99%. Compared with traditional registration methods, the iteration efficiency is improved by more than 50%, effectively avoiding occlusal misalignment problems in subsequent crown design.

[0033] Step 4: Precise segmentation and trimming of the abutment tooth and associated gingiva Based on the tooth position numbering in step 2, the initial point cloud region of the abutment tooth to be restored is locked. Combined with the spatial coordinate system of the abutment tooth after fine registration in step 3, the high point of the abutment tooth crown and the neck turning point are used as references. The initial spatial range of the abutment tooth is constructed by the point cloud bounding box algorithm (X-axis covers the complete length of the abutment tooth in the mesiodistal direction, Y-axis covers the maximum width in the labial, buccal, lingual and palatal directions, and Z-axis covers the area from the occlusal surface to 2 mm apical of the gingiva). The two-step method of "geometric feature differentiation + point cloud purification" is adopted. First, a classifier is constructed by using the point cloud normal vector and curvature features to initially separate the abutment tooth and gingival point clouds. Then, the abutment tooth point cloud is subjected to radius filtering for noise reduction and greedy projection triangulation for hole repair. The gingival point cloud is subjected to a region growing algorithm based on neighborhood distance, retaining only the connecting gingival region with a distance ≤0.1 mm from the surface of the abutment tooth, forming a combined point cloud model of the abutment tooth and the associated gingiva. After the spatial range of the abutment tooth is calibrated, the integrity of the region is verified by point cloud density analysis. If there is missing data, the bounding box range is automatically expanded by 0.5-1 mm. During the geometric feature differentiation process, a dynamic threshold adjustment strategy is adopted to dynamically optimize the classifier parameters according to the curvature differences between teeth and gingiva at different tooth positions. During the point cloud purification stage, the threshold for the diameter of the repair hole in the abutment tooth point cloud is set to 0.1 mm to ensure the integrity of the tooth surface while preserving the original morphological features of the junction area between the gingiva and the abutment tooth.

[0034] Analysis of the above technical content: This solution mainly uses the point cloud bounding box algorithm to construct the spatial range of the abutment tooth, combined with the two-step segmentation method of "geometric feature differentiation + point cloud purification", and solves the problems of inaccurate abutment tooth segmentation range, incomplete separation of abutment tooth and gingiva, noise and holes in point cloud, and the defect of existing segmentation technology in preserving the original morphology of the junction area between gingiva and abutment tooth through dynamic threshold adjustment strategy, radius filtering noise reduction, greedy projection triangulation hole repair and neighborhood distance constraint region growth algorithm. Simultaneously, spatial range calibration and point cloud purification optimization were combined to clarify the three-axis coverage standards of the abutment tooth spatial range (X-axis: complete length in the mesiodistal direction; Y-axis: maximum width in the labial, buccal, lingual, and palatal directions; Z-axis: 2 mm from the occlusal surface to the gingival apex), retaining only the connecting gingival region with a distance ≤0.1 mm from the abutment tooth surface; achieving precise separation of the abutment tooth and the associated gingiva and high-quality restoration of the point cloud model. Its innovation lies in the dynamic threshold adjustment to adapt to different tooth curvature differences, the 0.1 mm hole repair threshold to ensure the integrity of the tooth surface, and the precise selection of the connecting gingival region to preserve the original cervical margin morphology. Compared with existing segmentation techniques, the accuracy of abutment tooth segmentation range is improved to 99%, the point cloud noise removal rate reaches over 95%, and the hole repair integrity reaches 99.5%, providing accurate data support for subsequent cervical margin line detection.

[0035] Step 5: Multi-tissue segmentation of the abutment tooth surrounding environment Using the three-dimensional coordinates of the abutment tooth center point obtained in step 4 as the origin, a local spatial coordinate system is established with the mesiodistal, labial-buccal-lingual-palatal, and occlusal-gingival directions as orthogonal positioning axes. The system is extended by 5 mm in each direction to form a cubic segmentation space. Based on the spatial position relationship and point cloud texture features, the adjacent teeth are segmented using the Euclidean distance clustering algorithm (with a minimum distance threshold of 0.2-0.5 mm between the adjacent teeth and the abutment tooth surface). The opposing teeth are segmented based on the occlusal spatial relationship combined with the planar projection method and the texture features of the opposing teeth's occlusal surface. The surrounding gingiva and alveolar bone are segmented by jointly filtering the texture feature threshold and the Z-axis coordinate range to obtain complete environmental data around the abutment teeth. When establishing the local spatial coordinate system, orientation calibration is performed in conjunction with standard oral anatomical parameters to ensure that the positioning axis is consistent with the physiological long axis of the tooth. In the process of segmenting adjacent teeth, in addition to Euclidean distance and texture features, tooth morphology contour matching analysis is added to improve the accuracy of adjacent tooth identification. The segmentation of opposing teeth is assisted by occlusal trajectory simulation to ensure that the segmented area is completely matched with the occlusal contact range of the abutment teeth. The texture feature threshold of the segmentation of gingiva and alveolar bone is dynamically adjusted according to the gray-scale distribution of the scan data.

[0036] Analysis of the above technical content: This solution mainly adopts a local spatial coordinate system construction method with the center point of the abutment tooth as the origin. By extending 5 mm to the three axes to form a segmentation space, it combines Euclidean distance clustering, planar projection method, texture feature threshold and Z-axis coordinate joint screening multi-tissue differential segmentation algorithm, and introduces tooth morphology contour matching degree analysis and occlusal trajectory simulation verification mechanism. It solves the problems of vague definition of the segmentation range of the surrounding environment, inaccurate identification of adjacent teeth and opposing teeth, poor multi-tissue separation effect of traditional segmentation technology, as well as the defects of weak segmentation targeting caused by insufficient integration of oral anatomy rules in existing segmentation technology. Simultaneously, spatial positional relationships and point cloud texture features are used as the core basis for segmentation, and exclusive segmentation strategies are designed for adjacent teeth, opposing teeth, gingiva, and alveolar bone. This achieves precise separation and complete data acquisition of multiple tissues around the abutment teeth. Its innovation lies in calibrating the positioning axis direction with standard oral anatomy parameters, improving the recognition accuracy by matching the morphological contour of adjacent teeth, and simulating the occlusal trajectory of opposing teeth to ensure that the segmentation range matches the occlusal contact. Compared with traditional segmentation methods, the accuracy of adjacent tooth segmentation is improved to 98.5%, and the matching degree of opposing tooth segmentation is improved to 99%, providing comprehensive and accurate environmental reference data for crown fitting design.

[0037] Step 6: Secondary Precision Detection of the Neckline Using the combined point cloud model of the abutment tooth and associated gingiva from step 4 as the data source, the strategy of "coarse localization of depth model + fine detection of point cloud algorithm" is adopted. First, the combined point cloud data is input through the depth model to automatically identify the annular area where the cervical margin line is located and output the preliminary localization range within ±0.2 mm. Then, based on the coarse localization results, the normal vector of each point in the localization area is calculated by PCA algorithm, and candidate edge points with abrupt changes in normal vector direction are screened. The curve is reconstructed by RANSAC circle fitting algorithm and the curve smoothness is adjusted by active contour model to obtain a continuous and complete cervical margin line. The coarse localization of the deep model adopts a lightweight CNN architecture to reduce computation and improve localization speed. Before inputting the model, the combined point cloud data is downsampled to retain key geometric features while improving computational efficiency. In the fine detection stage, multi-scale normal estimation is introduced to improve the adaptability of the cervical margin line to different tooth positions. The RANSAC circle fitting algorithm iterates no less than 500 times. The curve smoothness parameter of the active contour model is dynamically adjusted according to the curvature change of the cervical margin line to ensure that the cervical margin line is continuous and conforms to the physiological structure.

[0038] Analysis of the above technical content: This solution mainly adopts a two-level detection strategy of "coarse localization of the depth model + fine detection of the point cloud algorithm". It achieves coarse localization of the annular region of the cervical margin line (error ±0.2 mm) through a lightweight CNN architecture depth model. Combined with PCA algorithm normal vector calculation, RANSAC circle fitting curve reconstruction and active contour model smoothing adjustment, it solves the problems of inaccurate localization, discontinuous boundaries, and large interference from irrelevant areas in traditional cervical margin line detection, as well as the shortcomings of existing detection technology in terms of insufficient adaptability to cervical margin lines of different tooth positions. Simultaneously, it combines the rapid localization advantage of deep models with the fine optimization capabilities of point cloud algorithms, using the combined point cloud of the abutment tooth and associated gingiva as a dedicated data source to avoid interference from grayscale information. It achieves accurate identification and continuous and complete reconstruction of the cervical margin. Its innovations lie in the balance between computational efficiency and feature preservation through downsampling processing, the improvement of adaptability to different tooth positions through multi-scale normal estimation, and the assurance of curve fitting accuracy through more than 500 RANSAC iterations. The cervical margin detection error is controlled within 0.02 mm, and the boundary continuity is improved to 99.8%. Compared with existing detection technologies, the detection efficiency is improved by more than 60%, and the adaptability to cervical margins of different tooth positions is significantly enhanced, providing a precise basis for defining the crown boundary.

[0039] Step 7: Crown formation and multi-dimensional refinement A two-step strategy of "deep model building + point cloud algorithm refinement" was adopted. The precise cervical margin parameters from step 6, the 3D morphological data of the abutment tooth from step 4, and the surrounding environment data of the abutment tooth from step 5 were input. Oral anatomy and physiology feature databases and clinical restoration standards (incisal edge thickness ≥ 1.5 mm, axial thickness ≥ 1.0 mm, cervical margin thickness ≥ 0.8 mm) were integrated as constraints. First, the inner crown (offset 0.3-0.5 mm along the normal direction of the prepared abutment tooth surface to reserve bonding space) and the outer crown (axial morphology adjusted according to the interdental space, and occlusal cusp-fossa structure designed based on the occlusal surface features of the opposing tooth) were constructed to form a complete basic crown model. Then, the point cloud algorithm was used for volume calibration (volume error ≤ 5% with the contralateral tooth), orientation correction (parallelism error with the long axis of the adjacent tooth ≤ 1°), morphological fine-tuning (low-square method optimization of flatness using inner crown movement, and contour adjustment by comparing the feature point cloud of the outer crown), and cervical margin fit optimization (marginal fit error ≤ 0.02 mm). (millimeters) and undercut filling, outputting a 3D model of the crown that meets clinical restoration needs; The oral anatomy and physiology feature library contains a template library of crown morphology for different age groups and tooth positions, as well as personalized occlusal curve parameters. It can automatically match the optimal template according to the patient's age and dentition status. The clinical restoration standard supports custom adjustment functions to meet the personalized needs of special cases. Occlusal simulation tests are added during the crown finishing process. The occlusal surface morphology is optimized through virtual occlusal contact analysis to ensure uniform occlusion. At the same time, a mesh optimization algorithm is used to reduce the number of triangular facets in the crown model, improving the model processing efficiency. The final output 3D crown model can be directly imported into the CAM processing system.

[0040] Analysis of the above technical content: This solution mainly adopts a two-step strategy of "deep model modeling + point cloud algorithm refinement". It inputs cervical margin parameters, abutment tooth morphology data and surrounding environment data, and integrates oral anatomy and physiology feature database and clinical restoration standards (incisal edge thickness ≥1.5 mm, axial thickness ≥1.0 mm, cervical margin thickness ≥0.8 mm). Through internal crown offset to reserve bonding space (0.3-0.5 mm), external crown morphology adaptation adjustment, and multi-dimensional refinement (volume calibration, direction correction, morphological fine adjustment, cervical margin fit optimization, undercut filling), it solves the problems of poor fit, occlusal incoordination, and key indicators not meeting clinical standards in traditional crown generation, as well as the shortcomings of existing technologies in balancing personalization and clinical adaptability. Simultaneously, it combines basic model construction with multi-dimensional refinement, supporting customized adjustments to clinical restoration standards and occlusal simulation testing; it achieves automatic generation of high-precision 3D crown models that meet clinical needs. Its innovations lie in multi-source data fusion to ensure crown adaptability, quantification of multi-dimensional refinement indicators (volume error ≤5%, long axis parallelism error ≤1°, cervical margin fit error ≤0.02 mm), and mesh optimization to improve processing efficiency. The success rate of one-time crown restoration is increased to over 95%, and the material utilization rate is increased from 70%-80% to over 95%. The output model can be directly imported into the CAM processing system. Compared with traditional crown generation methods, production efficiency is increased by over 80%, and clinical adaptability and aesthetics are significantly superior to existing technologies.

[0041] Regarding the working principle of this solution: This solution mainly integrates deep learning technology, point cloud processing algorithms, oral anatomy rules and clinical restoration standards to build a fully automated crown generation technology system, realizing closed-loop processing from oral 3D data acquisition to high-precision crown 3D model output. Its core principle revolves around four key steps: "precise data extraction, spatial relationship restoration, adaptive model generation, and multi-dimensional refinement and optimization." First, three-dimensional data of all teeth, gums, and related soft tissues are collected using a medical dental scanner or 3D scanner. After standardization, this data provides a high-quality data source for subsequent steps. This process ensures the original data is free of holes and has low noise, based on scanning accuracy requirements and data compatibility needs. Second, a deep learning model with an improved U-Net architecture, combined with a tooth anatomical morphology feature library, is used to identify differences between teeth and soft tissues in grayscale values, textures, and morphological structures. This enables semantic segmentation of all teeth and association with FDI tooth position numbers. Simultaneously, point cloud bounding box algorithms, geometric feature differentiation, and region growing algorithms are used to accurately extract environmental data such as abutment teeth, associated gums, surrounding adjacent teeth, and opposing teeth, solving the problems of interference and inaccurate range in traditional segmentation. Third, using tooth position numbers as the core index, 12-15 key tooth feature points are extracted through a registration depth model, combined with ICP... The algorithm performs precise registration of point clouds for the upper and lower jaws, restoring the natural occlusal space benchmark based on the objective function of minimizing the sum of Euclidean distances and constraints of occlusal relationships. Subsequently, based on the geometric differences between the abutment teeth and the associated gingiva, a strategy of "coarse localization using a depth model + fine detection using a point cloud algorithm" is adopted. Through PCA algorithm, RANSAC circle fitting, and active contour model, the cervical margin boundary is accurately identified. Finally, using cervical margin parameters, abutment tooth morphology data, and surrounding environment data as input, and incorporating oral anatomy and physiology feature databases and clinical restoration standards, a basic crown model is constructed using a depth model. This model is then refined through multi-dimensional point cloud algorithms, including volume calibration, orientation correction, morphological fine-tuning, cervical margin fit optimization, and undercut filling, ensuring that the crown meets clinical requirements in key indicators such as thickness, morphology, and occlusal relationships. The entire workflow is linked by accurate data transmission, supported by algorithmic integration and innovation, and constrained by clinical standards, achieving an organic unity of automation and high precision.

[0042] The core innovations of this solution are: a crown generation architecture that combines "full-process closed-loop automation, multi-algorithm fusion for precision, and deep integration of clinical standards," which solves four major problems: insufficient segmentation accuracy of existing technologies, limited crown fit and compatibility with adjacent teeth, lack of systematic point cloud fine-tuning strategies, and difficulty in balancing automation and clinical standards. Simultaneously, for the first time, a closed-loop data transfer system is formed, integrating seven key stages: semantic segmentation, precise registration, accurate segmentation, environment extraction, cervical margin detection, model generation, and multi-dimensional refinement. The output data of each stage directly serves as the input benchmark for the next stage, ensuring process continuity and data accuracy. For the first time, a dual segmentation strategy of "deep learning + point cloud geometric features" is adopted. An attention mechanism is introduced to enhance edge extraction capabilities in full-mouth tooth semantic segmentation, while grayscale dependence is eliminated in abutment tooth and gingival segmentation. Precise separation is achieved through normal vectors, curvature features, and neighborhood distances, while preserving the original morphology of the cervical margin connection area. For the first time, a two-stage cervical margin detection scheme of "deep model coarse localization + point cloud algorithm fine detection" is designed, combining PCA algorithm, RANSAC circle fitting, and active contour model to achieve accurate detection with a cervical margin error ≤0.02 mm. For the first time, an oral anatomy and physiological feature database (containing morphological templates and occlusal curves for different age groups and tooth positions) is integrated with clinical restoration standards (incisal edge thickness ≥1.5 mm, axial thickness ≥1.0 mm). The system incorporates hard constraints (such as millimeters) into the entire crown generation process, while using multi-dimensional refinement algorithms (volume error ≤5%, major axis parallelism error ≤1°) to achieve personalized adaptation. It is the first to adopt a hybrid registration method combining a "registration depth model + ICP algorithm," introducing adaptive weighting factors and occlusal contact detection verification mechanisms to ensure that the accuracy of restoring the maxillary and mandibular occlusal relationship reaches the 0.001 mm level. It is also the first to achieve differentiated segmentation of multiple tissues around the abutment teeth, using proprietary algorithms such as Euclidean distance clustering, planar projection, and texture and coordinate joint filtering to accurately extract data from adjacent teeth, opposing teeth, gingiva, and alveolar bone, providing a comprehensive environmental reference for crown fit design. These innovative concepts collectively break through the bottlenecks of existing technologies, achieving simultaneous satisfaction of automation, high precision, high adaptability, and clinical practicality.

[0043] The technical effects of implementing this solution: After implementation, this solution has achieved comprehensive improvements in efficiency, accuracy, cost, and adaptability in the field of dental restoration, with significant technical effects and broad clinical application value. Regarding efficiency improvements, the fully automated process eliminates tedious manual operations such as segmentation, cervical margin delineation, and occlusal adjustment. The time for generating a single crown is reduced by more than 80% compared to traditional CAD / CAM methods (15-30 minutes / tooth), requiring only 3-6 minutes to complete the entire process from data input to model output. This significantly improves the capacity for handling large batches of clinical cases, reduces the overall patient cycle from scanning to crown processing completion by more than 50%, decreases the number of follow-up visits, significantly improves the patient experience, and simultaneously reduces the operating costs of medical institutions. In terms of precision improvement, the full-mouth tooth segmentation accuracy reaches over 98%, the tooth position numbering error rate is ≤0.1%, the maxillary-mandibular registration error is ≤0.001 mm, the cervical margin detection error is ≤0.02 mm, and the crown-abutment tooth margin fit error is ≤0.02 mm. The core indicators far surpass existing technologies in precision, effectively avoiding problems such as poor crown fit and malocclusion caused by data errors. The one-time restoration success rate has increased from 70%-80% with traditional techniques to over 95%. Regarding cost control, the improved crown fabrication precision eliminates the need for secondary grinding and adjustment after processing. The utilization rate of restorative materials has increased from 70%-80% to over 95%. For high-end materials such as zirconia, waste costs can be significantly reduced. Simultaneously, the fully automated process reduces reliance on experienced dentists, significantly lowering the operational threshold. Small and medium-sized dental institutions can implement this technology without investing in additional professional design personnel, promoting the widespread adoption of digital dental restoration technology. In terms of adaptability and aesthetics, the crown generation process fully considers the interdental spaces (0.1-0.2 mm), the cusp-fossa structure of the opposing teeth's occlusal surface, and the gingival space. The generated crowns not only conform to the anatomical characteristics of the corresponding teeth in morphology, but also have uniform and coordinated occlusion, and a natural and aesthetically pleasing cervical margin, meeting both clinical functional and aesthetic requirements. The oral anatomy and physiology feature database and customized clinical standards support the personalized needs of patients of different ages and dentition conditions (normal dentition, crowded dentition, missing tooth restoration, etc.), with adaptability covering more than 95% of clinical scenarios. In terms of technical compatibility, it supports data input from various medical dental scanners and 3D scanners. The raw data is standardized into PLY or STL formats, allowing seamless integration with existing mainstream CAM processing systems without the need for additional equipment modifications, thus reducing the cost of technology implementation. In addition, the solution ensures stable and controllable model quality through multi-stage data verification and repair mechanisms (such as point cloud hole repair and automatic compensation for missing data), reducing rework and re-making caused by unqualified models, lowering clinical diagnosis and treatment risks and additional costs. At the same time, the mesh optimization algorithm reduces the number of triangular facets in the crown model, improves processing efficiency, and further shortens the overall restoration cycle.In summary, the implementation of this solution not only breaks through existing technical bottlenecks, but also promotes digital dental restoration technology to a new stage of development characterized by "high precision, automation, personalization, and low cost," achieving a win-win situation for both doctors and patients and upgrading the industry's technology.

[0044] This solution differs from traditional solutions in its core technologies and highlights its significant technological advantages. I. Core Differentiating Technologies Differences in Data Processing Architecture: Traditional CAD / CAM crown design methods employ a discrete processing architecture of "manual guidance + basic data assistance," with each step relying on manual coordination by the dentist. Deep learning-based crown reconstruction methods use a linear architecture of "single model drive + simple data input," lacking a multi-stage data loop. This solution innovatively constructs a "fully automated closed-loop architecture," seamlessly integrating seven stages: 3D oral data acquisition, full-mouth tooth semantic segmentation, precise registration of maxillary and mandibular point clouds, precise segmentation and trimming of abutment teeth and associated gingiva, multi-tissue segmentation of the abutment tooth's surrounding environment, secondary precise detection of the cervical margin, crown generation, and multi-dimensional refinement. The output data from each stage directly serves as the input benchmark for the next stage, completing the entire process without manual intervention.

[0045] Differences in segmentation techniques: Traditional methods rely on manual segmentation or basic image segmentation, which is easily affected by soft tissue interference; existing deep learning methods rely only on a single semantic segmentation model, which is insufficient for point cloud detail processing. This solution adopts a dual segmentation strategy of "deep learning + point cloud geometric features". In the full-mouth tooth semantic segmentation stage, U-Net is introduced to improve the architecture and attention mechanism. In the abutment tooth and gingival segmentation stage, a two-step method of "geometric feature differentiation + point cloud purification" is adopted to abandon grayscale dependence and achieve accurate separation through spatial features such as normal vectors, curvature, and neighborhood distance, while preserving the original morphology of the cervical margin connection area.

[0046] Differences between registration and occlusal relationship restoration: Traditional methods rely on manual adjustment of occlusal relationships, resulting in low accuracy and poor consistency; existing hybrid registration methods lack key feature point reinforcement and closed-loop verification. This scheme uses FDI tooth position numbers as the core index, extracts 12-15 key tooth feature points through a registration depth model, combines the ICP algorithm and KD-Tree to accelerate the search, introduces an adaptive weighting factor and occlusal contact detection verification mechanism, and controls the iteration termination error to within 0.001 mm, achieving accurate restoration of occlusal relationships.

[0047] Differences in neckline detection technology: Traditional methods rely on manual outlining of the neckline, which is inefficient and prone to large errors; existing detection technologies mostly use a single algorithm, making it difficult to balance accuracy and continuity. This solution innovatively designs a two-stage detection strategy of "coarse localization using a depth model + fine detection using a point cloud algorithm". It quickly locates the neckline region (error ±0.2 mm) through a lightweight CNN architecture, and then achieves accurate reconstruction through PCA algorithm, RANSAC circle fitting (iterations ≥500 times) and active contour model, with a detection error ≤0.02 mm.

[0048] Differences in crown generation and optimization mechanisms: Traditional methods manually design crown morphology, making it difficult to balance clinical standards and fit; existing deep learning methods lack systematic fine-tuning strategies, and the generated models often require secondary correction. This solution adopts a two-step strategy of "deep model building + multi-dimensional point cloud refinement," incorporating an oral anatomy and physiology feature database and quantified clinical restoration standards (such as incisal edge thickness ≥1.5 mm). Through multi-dimensional refinement including volume calibration (error ≤5%), orientation correction (parallelism error ≤1°), and cervical margin fit optimization, it ensures crown fit and aesthetics.

[0049] II. Outstanding Technological Advantages Automation and efficiency advantages: The entire process requires no manual intervention in tooth segmentation, cervical margin delineation, occlusal adjustment, etc. The time for generating a single crown is reduced by more than 80% compared to the traditional method (15-30 minutes / tooth), and the entire process can be completed in only 3-6 minutes. This greatly improves the clinical batch processing capacity, while reducing the reliance on senior professional doctors. The operation threshold is significantly lowered, and small and medium-sized dental institutions can quickly implement and apply it.

[0050] Advantages in accuracy and consistency: Accuracy in core processes is quantifiable and controllable, achieving over 98% accuracy in full-mouth tooth segmentation. Maxillary and mandibular registration errors are ≤0.001 mm, cervical margin detection errors are ≤0.02 mm, and crown margin fitting errors are ≤0.02 mm, significantly exceeding the accuracy levels of traditional methods and existing deep learning methods. Standardized processes eliminate variations in human operation, significantly improving consistency in results across different cases and operators.

[0051] Clinical adaptability and aesthetic advantages: The crown generation process fully integrates the oral anatomy and physiology database (including templates for different age groups and tooth positions) with clinical restoration standards. At the same time, it accurately considers the interdental space (0.1-0.2 mm), the occlusal surface morphology of the opposing teeth, and the spatial position of the gingiva. The generated crowns not only meet clinical requirements in terms of thickness and shape, but also have occlusal coordination and natural cervical margin connection. The success rate of one-time restoration has been increased from 70%-80% of traditional techniques to over 95%.

[0052] Cost and resource utilization advantages: Improved precision in crown fabrication eliminates the need for post-processing grinding and adjustments, increasing the utilization rate of restorative materials from 70%-80% to over 95%, significantly reducing waste costs associated with high-end materials such as zirconia. The overall patient cycle from scanning to crown completion is shortened by more than 50%, reducing the number of follow-up visits. This lowers operating costs for medical institutions and alleviates the financial and time burden on patients, achieving a win-win situation for both doctors and patients.

[0053] Compatibility and adaptability advantages: Supports data input from various medical dental scanners and 3D scanners. Raw data is standardized and converted to PLY or STL format, which can be directly imported into mainstream CAM processing systems without additional equipment modifications. Clinical restoration standards support custom adjustments. The feature library covers various scenarios such as normal dentition, crowded dentition, and missing tooth restoration, adapting to over 95% of clinical cases, and possessing broad applicability and scalability.

[0054] Technical closed-loop and stability advantages: A seven-stage data closed-loop transmission system, with data verification and repair mechanisms at each stage (such as point cloud hole repair and automatic compensation for missing data), ensures stable and controllable model quality, reduces rework and re-production due to data errors, and lowers clinical treatment risks. The mesh optimization algorithm reduces the number of triangular facets in the dental crown model, improving processing efficiency while ensuring the feasibility and stability of model fabrication.

[0055] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. A high-precision, automated method for forming dental crowns, characterized in that: The method is as follows: Step 1: Oral cavity 3D data acquisition Use a medical dental scanner or 3D scanner to obtain complete three-dimensional scan data of the upper and lower jaws, ensuring that the scan range covers all teeth, gums and related soft tissues, and that the data format supports subsequent point cloud processing and model building; Step 2: Semantic segmentation of the entire mouth teeth A semantic segmentation method based on a deep learning model combined with a dental anatomical morphology feature database is adopted. The whole scan data is input, and the differences between teeth and soft tissues such as gums and mucosa in gray value, texture features and morphological structure are identified to achieve automatic separation of individual teeth in the whole mouth. An independent three-dimensional data model of a single tooth is established, which includes complete data of the crown, neck and exposed part of the root. Each tooth is matched with a unique tooth position number according to the International Dental Federation tooth position numbering system, and the number and the corresponding tooth data model are associated and stored. Step 3: Fine-tuning of upper and lower jaw point clouds Using the tooth position number obtained in step 2 as the core index, a registration depth model based on the maxillofacial dynamic map structure is used to extract 12-15 stable feature points of corresponding teeth in the upper and lower jaws. The feature point set is initially matched by the registration depth model. Then, the point cloud after initial transformation is input into the ICP algorithm. KD-Tree is used to accelerate the nearest neighbor search. Erroneous point pairs with a distance threshold greater than 0.05 mm are removed. The objective function is to minimize the sum of Euclidean distances between feature points. The iteration termination condition is set to the distance error change between two adjacent iterations being less than 0.001 mm. Combined with the constraint of the natural occlusal contact relationship of the upper and lower jaw teeth, the precise alignment of the upper and lower jaw point clouds is achieved. Step 4: Precise segmentation and trimming of the abutment tooth and associated gingiva Based on the tooth position numbering in step 2, the preliminary point cloud region of the abutment tooth to be restored is locked. Combined with the spatial coordinate system of the abutment tooth after fine registration in step 3, the preliminary spatial range of the abutment tooth is constructed by using the high point of the crown and the neck turning point of the abutment tooth as references. The two-step method of "geometric feature differentiation + point cloud purification" is adopted. First, a classifier is constructed by using the point cloud normal vector and curvature features to initially separate the abutment tooth and gingival point clouds. Then, the abutment tooth point cloud is subjected to radius filtering for noise reduction and greedy projection triangulation for hole repair. The gingival point cloud is subjected to a region growing algorithm based on neighborhood distance, retaining only the connecting gingival region with a distance ≤0.1 mm from the surface of the abutment tooth, forming a combined point cloud model of the abutment tooth and the associated gingiva. Step 5: Multi-tissue segmentation of the abutment tooth surrounding environment Using the combined point cloud model of the abutment tooth and associated gingiva obtained in step 4 as the origin, a local spatial coordinate system is established with the mesiodistal, labiobuccal-lingual-palatal, and occlusal-gingival directions as orthogonal positioning axes. The system is extended by 5 mm in each direction to form a cubic segmentation space. Based on the spatial position relationship and point cloud texture features, adjacent teeth are segmented using the Euclidean distance clustering algorithm. The opposing teeth are segmented based on the occlusal spatial relationship combined with the planar projection method and the occlusal surface texture features of the opposing teeth. The surrounding gingiva and alveolar bone are segmented by jointly filtering the texture feature threshold and the Z-axis coordinate range to obtain complete environmental data around the abutment tooth. Step 6: Secondary Precision Detection of the Neckline Using the combined point cloud model of the abutment tooth and associated gingiva from step 4 as the data source, the strategy of "coarse localization of depth model + fine detection of point cloud algorithm" is adopted. First, the combined point cloud data is input through the depth model to automatically identify the annular area where the cervical margin line is located and output the preliminary localization range within ±0.2 mm. Then, based on the coarse localization results, the normal vector of each point in the localization area is calculated by PCA algorithm, and candidate edge points with abrupt changes in normal vector direction are screened. The curve is reconstructed by RANSAC circle fitting algorithm and the curve smoothness is adjusted by active contour model to obtain a continuous and complete cervical margin line. Step 7: Crown formation and multi-dimensional refinement A two-step strategy of "deep model modeling + point cloud algorithm refinement" is adopted. The precise cervical margin parameters from step 6, the combined point cloud model of the abutment tooth and related gingiva from step 4, and the environmental data around the abutment tooth from step 5 are input. The oral anatomy and physiology feature library and clinical restoration standards are incorporated as constraints. First, the inner and outer crowns of the crown are constructed to form a complete basic crown model. Then, the point cloud algorithm is used to perform volume calibration, orientation correction, morphological fine-tuning, cervical margin fitting optimization, and undercut filling to output a 3D crown model that meets the clinical restoration requirements.

2. The high-precision, automated method for crown fabrication according to claim 1, characterized in that: Before scanning and collecting data, the soft tissues in the oral cavity are cleaned to remove food debris and excess saliva. The scanner resolution is set to no less than 50 micrometers. During the scanning process, a multi-angle superimposed scanning method is used to focus on scanning key areas such as the occlusal surface and proximal surface of the teeth to ensure that the data is free of holes and obvious noise. After the data is collected, the raw data is standardized and converted into PLY or STL format.

3. The high-precision, automated method for crown fabrication according to claim 1, characterized in that: The semantic segmentation method adopts an improved U-Net architecture and introduces an attention mechanism to enhance the ability to extract tooth edge features. The model training dataset contains more than 1,000 clinical scan data of different dental arch morphologies and oral conditions, covering multiple scenarios such as normal dental arch, crowded dental arch, and missing tooth restoration. After segmentation, the integrity of the three-dimensional data model of a single tooth is checked, and missing data areas are automatically identified and repaired. The tooth position number association adopts a dual verification mechanism, combining the spatial position and morphological features of the tooth to ensure the accuracy of the numbering.

4. The high-precision, automated method for crown fabrication according to claim 1, characterized in that: The registration depth model constructs a feature library using data from 400 clinical cases of non-malformed maxillofacial structures, automatically filters out interference from gingival soft tissue, introduces an adaptive weighting factor during the ICP algorithm iteration process, assigns higher weights to key cusp-fossa feature points on the occlusal surface, and improves the accuracy of occlusal relationship restoration. After registration is completed, the registration effect is verified by the occlusal contact detection algorithm. If occlusal misalignment exists, the registration is automatically iterated again.

5. The high-precision, automated method for crown fabrication according to claim 1, characterized in that: After the initial spatial range of the abutment teeth is calibrated, the integrity of the region is verified by point cloud density analysis. If there is missing data, the bounding box range is automatically expanded by 0.5-1 mm. During the geometric feature differentiation process, a dynamic threshold adjustment strategy is adopted to dynamically optimize the classifier parameters according to the curvature differences between teeth and gingiva at different tooth positions. During the point cloud purification stage, the threshold for the diameter of the repair hole in the abutment tooth point cloud is set to 0.1 mm to ensure the integrity of the tooth surface while preserving the original morphological features of the junction area between the gingiva and the abutment tooth.

6. The high-precision, automated method for crown fabrication according to claim 1, characterized in that: When establishing the local spatial coordinate system, orientation calibration is performed in conjunction with standard oral anatomical parameters to ensure that the positioning axis is consistent with the physiological long axis of the tooth. In the process of segmenting adjacent teeth, in addition to Euclidean distance and texture features, tooth morphology contour matching analysis is added to improve the accuracy of adjacent tooth identification. The segmentation of opposing teeth is assisted by occlusal trajectory simulation to ensure that the segmented area is completely matched with the occlusal contact range of the abutment teeth. The texture feature threshold of the segmentation of gingiva and alveolar bone is dynamically adjusted according to the gray-scale distribution of the scan data.

7. The high-precision, automated method for crown fabrication according to claim 1, characterized in that: The coarse localization of the deep model adopts a lightweight CNN architecture to reduce computation and improve localization speed. Before inputting the model, the combined point cloud data is downsampled to retain key geometric features while improving computational efficiency. In the fine detection stage, multi-scale normal estimation is introduced to improve the adaptability of the cervical margin line to different tooth positions. The RANSAC circle fitting algorithm iterates no less than 500 times. The curve smoothness parameter of the active contour model is dynamically adjusted according to the curvature change of the cervical margin line to ensure that the cervical margin line is continuous and conforms to the physiological structure.

8. The high-precision, automated method for crown fabrication according to claim 1, characterized in that: The oral anatomy and physiology feature library contains a template library of crown morphology for different age groups and tooth positions, as well as personalized occlusal curve parameters. It can automatically match the optimal template according to the patient's age and dentition status. The clinical restoration standard supports custom adjustment functions to meet the personalized needs of special cases. Occlusal simulation tests are added during the crown finishing process. The occlusal surface morphology is optimized through virtual occlusal contact analysis to ensure uniform occlusion. At the same time, a mesh optimization algorithm is used to reduce the number of triangular facets in the crown model, improving the model processing efficiency. The final output 3D crown model can be directly imported into the CAM processing system.

Images