An implant navigation system for oral medicine image AI maxillofacial reconstruction
By using an improved nnU-Net network to automatically segment and 3D reconstruct oral CBCT images, the problem of discontinuous segmentation of the inferior alveolar nerve canal was solved, enabling automatic optimization of implant pose and guide design, thus improving the efficiency and safety of implant planning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUILIN UNIV OF ELECTRONIC TECH
- Filing Date
- 2026-04-13
- Publication Date
- 2026-05-29
AI Technical Summary
In existing technologies, the inferior alveolar nerve canal is not segmented continuously during dental implant planning, the reconstruction of key anatomical structures is not accurate enough, and the segmentation results are difficult to directly serve implant planning and guide design, resulting in low planning efficiency and poor safety.
An improved nnU-Net network, combined with a fracture-aware continuity constraint module and a boundary-semantic collaborative fusion module, is used to automatically segment and 3D reconstruct key anatomical structures such as the dentition, maxilla, mandible, inferior alveolar nerve canal, and maxillary sinus, generating implant pose and guide model. By pre-setting safety constraints, the implant position and angle are optimized to achieve automated planning.
It improves the efficiency of automatic extraction of key anatomical structures and the stability of segmentation results, reduces the risk of nerve injury, enhances the automatic generation capability and clinical safety of implant planning, and improves the level of intelligence in preoperative planning.
Smart Images
Figure CN122096969A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of oral medical image processing, artificial intelligence medical image analysis, preoperative planning for oral implant surgery, and digital guide design, specifically to an implant navigation system for AI maxillofacial reconstruction based on oral medical images. Background Technology
[0002] With the development of dental implant technology, digital preoperative planning and implant guide design based on imaging data have become important means to improve implant accuracy and safety. Current dental implant planning usually relies on cone-beam computed tomography (CBCT) to obtain three-dimensional images of the patient's maxillofacial region, and combines this with the doctor's experience to complete bone volume assessment of the edentulous area, identification of dangerous anatomical structures, implant positioning design, and guide fabrication.
[0003] However, existing technologies still have the following problems. First, CBCT images often show blurred boundaries and low grayscale contrast for slender anatomical structures such as the inferior alveolar nerve canal, and local areas are prone to breaks, discontinuities, or missed detections, leading to insufficient stability of automatic segmentation results. Especially in pre-implantation planning, inaccurate identification of the inferior alveolar nerve canal can easily affect the assessment of implant safety distances, increasing the risk of nerve damage. Second, although anatomical structures such as the maxillary sinus floor, maxilla, mandible, and dentition can be extracted using conventional segmentation methods, current segmentation results often lack effective linkage with subsequent implant positioning planning and guide design, making it difficult to form a complete closed loop from image analysis to guide output. Third, some existing methods focus on single-structure segmentation or 3D reconstruction, failing to directly transform the spatial constraints of dangerous anatomical structures into quantitative evaluation criteria for implantable areas, resulting in insufficient automated planning capabilities.
[0004] Furthermore, existing deep learning-based oral CBCT segmentation methods primarily focus on improving overall segmentation accuracy, while lacking specific designs for implant safety planning scenarios, addressing issues such as insufficient ability to restore the continuity of the inferior alveolar nerve canal and weak boundary recognition capabilities. Current implant planning procedures also generally rely on dentists manually determining the location of the inferior alveolar nerve canal and maxillary sinus, and manually adjusting the implant position, angle, and length, resulting in limited planning efficiency and significant variations among different operators.
[0005] Therefore, it is necessary to propose an implant navigation system that can intelligently reconstruct key anatomical structures of the oral cavity using CBCT and automatically optimize implant position and guide plate output based on the safe distance of dangerous anatomical structures in the context of preoperative planning for oral implant surgery, so as to improve the efficiency of preoperative planning, the degree of automation of guide plate design, and the safety of clinical implantation. Summary of the Invention
[0006] This invention aims to provide an implant navigation method, system, electronic device, and storage medium for AI-based maxillofacial reconstruction in oral medical imaging, in order to solve the problems of discontinuous segmentation of the inferior alveolar nerve canal, insufficient accuracy in reconstruction of key anatomical structures, and difficulty in directly using segmentation results to serve implant planning and guide design in the prior art.
[0007] This invention also aims to provide an intelligent implant navigation solution based on oral CBCT images. By automatically segmenting and three-dimensionally reconstructing key anatomical structures such as the dentition, maxilla, mandible, inferior alveolar nerve canal, and maxillary sinus, candidate implant areas are determined, and implant pose and implant guide model are automatically generated under preset safety constraints, thereby reducing the risk of nerve damage and penetration into the maxillary sinus during the implantation process.
[0008] To achieve the above objectives, the present invention adopts the following technical solution: An implant navigation system for AI-based maxillofacial reconstruction based on oral medical imaging includes: The data acquisition module is used to acquire CBCT image data of the patient's oral and maxillofacial region; The preprocessing module is used to perform one or more preprocessing operations on the CBCT image data, including standardization, resampling, region of interest extraction, and format conversion. An automatic segmentation module is used to automatically segment the target anatomical structures in the CBCT image data based on a segmentation model. The target anatomical structures include at least one or more of the following: dentition, maxilla, mandible, inferior alveolar nerve canal, and maxillary sinus. In some embodiments, the segmentation model is an improved nnU-Net network. The three-dimensional reconstruction module is used to generate a corresponding three-dimensional model based on the segmentation results, and to convert the target anatomical structure into a three-dimensional mesh model that can be used for planting planning and guide plate design. The assessment module is used to generate planting assessment results based on the spatial relationship between candidate planting areas and dangerous anatomical structures, so as to characterize the planting feasibility of different areas; The planting planning module is used to optimize the position, angle, and length information of the implants according to the planting assessment results and preset safety constraints, and generate implant planning results. The guide plate design module is used to generate implant guide plate design parameters and / or implant guide plate three-dimensional models based on the implant planning results and the dentition three-dimensional model. The output module is used to output and / or visualize the three-dimensional reconstruction results of the target anatomical structure, the implant planning results, and the guide plate model.
[0009] Furthermore, in some embodiments, the segmentation model includes a break-aware continuity constraint module to enhance the connectivity representation of the slender structures of the inferior alveolar nerve tube within adjacent slices, adjacent voxels, or local spatial ranges, thereby reducing the probability of nerve tube breakage, missed segmentation, or local discontinuity.
[0010] Furthermore, in some embodiments, the segmentation model includes a boundary-semantic co-fusion module, which is used to fuse boundary information in shallow features with semantic information in deep features to improve the boundary localization accuracy and structural recognition capability of the inferior alveolar nerve canal and other key anatomical structures.
[0011] Furthermore, the output of the 3D reconstruction module includes a segmentation label file and a corresponding 3D surface mesh model. The segmentation label file is an NIfTI format file, and the 3D surface mesh model is an STL format file.
[0012] Furthermore, the implant assessment result is an implantation area scoring map, and the scoring value is related to at least one or more of the minimum distance from the candidate location to the inferior alveolar nerve canal and the minimum distance from the candidate location to the maxillary sinus floor.
[0013] Furthermore, the planting planning module performs planning in the following manner: it extracts candidate sites that meet preset conditions from the candidate planting area, performs safety screening on each candidate site, optimizes the position, angle and length of the implant through preset safety constraints, and outputs the implant planning result that meets the planting safety requirements.
[0014] Furthermore, the preset safety constraints include at least one or more of the following: maintaining a safe distance from the inferior alveolar nerve canal and maintaining a safe distance from the floor of the maxillary sinus.
[0015] Furthermore, the guide plate design output module preferentially generates tooth-supported guide plates, and can be expanded to generate mucosa-supported guide plates or bone-supported guide plates.
[0016] Furthermore, the guide plate design output module determines the guide plate base fitting area based on the three-dimensional surface information of the dental arch, determines the position and direction of the guide hole based on the implant pose, and generates a three-dimensional model of the guide plate.
[0017] Compared with the prior art, the present invention has the following beneficial effects: First, this invention improves the efficiency of automatic extraction of key structures in preoperative planning of oral implant surgery by automatically segmenting and reconstructing key anatomical structures in oral CBCT images, and provides a structural basis for subsequent implant planning.
[0018] Secondly, in a preferred embodiment, the present invention enhances the continuity recovery capability of the inferior alveolar nerve canal by setting a fracture sensing continuity constraint module, which is designed to address the characteristics of the inferior alveolar nerve canal being slender, easily broken, and discontinuous, thereby improving the integrity and stability of the inferior alveolar nerve canal reconstruction results.
[0019] Third, in a preferred embodiment, the present invention can improve the boundary recognition accuracy of the inferior alveolar nerve canal and adjacent bony structures by setting a boundary-semantic collaborative fusion module, thereby reducing missegmentation and missed segmentation caused by boundary ambiguity.
[0020] Fourth, this invention further transforms the segmentation results of key anatomical structures into planting assessment results based on the spatial relationships of dangerous anatomical structures, realizing the linkage from structure identification to planning decision-making and improving the ability to automatically generate planting plans.
[0021] Fifth, this invention automatically generates the implant position, angle, and length under preset safety constraints, and outputs guide plate design parameters and guide plate model, which can improve preoperative planning efficiency and reduce the burden of manual adjustment by doctors.
[0022] Sixth, this invention constructs a complete technical chain for clinical applications of oral implantology, from image input, automatic segmentation, risk assessment, pose planning to guide plate output, which helps to improve the intelligence level of preoperative implant planning and surgical safety. Attached Figure Description
[0023] Figure 1 This is a block diagram of the overall structure of an implant navigation system for AI-based maxillofacial reconstruction based on oral medical imaging, according to the present invention. Figure 2 This is a flowchart of an implant navigation method for AI-based maxillofacial reconstruction based on oral medical imaging, according to the present invention. Detailed Implementation
[0024] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments, but the scope of protection of the present invention is not limited thereto.
[0025] Example 1: System Overall Structure This embodiment provides an implant navigation system for AI-based maxillofacial reconstruction in oral medical imaging, applicable to preoperative planning scenarios for oral implant surgery. The system mainly includes a data acquisition module, a preprocessing module, an automatic segmentation module, a 3D reconstruction module, an evaluation module, an implant planning module, a guide plate design module, and an output module.
[0026] The data acquisition module is used to acquire the patient's preoperative CBCT image data. This CBCT image data can originate from clinical dental CBCT equipment, and the original format can be DICOM, or it can be further converted to NIfTI format for network training and inference.
[0027] The preprocessing module is used to preprocess the acquired CBCT image data, including one or more of the following: grayscale normalization, voxel spacing unification, region of interest cropping, and data format unification, in order to improve the stability of subsequent automatic segmentation.
[0028] The automatic segmentation module segments key anatomical structures in the oral cavity based on a segmentation model. These key anatomical structures include the dentition, maxilla, mandible, inferior alveolar nerve canal, and maxillary sinus. The segmentation model can be a deep learning-based encoder-decoder network, preferably an improved nnU-Net network. In some implementations, the segmentation model introduces a break-aware continuity constraint module and a boundary-semantic co-fusion module on top of the original network framework. The former enhances the continuity representation ability of the slender structure of the inferior alveolar nerve canal, while the latter fuses shallow boundary features and deep semantic features to improve boundary recognition accuracy.
[0029] The 3D reconstruction module generates 3D models of each key structure based on the segmentation labels output by the AI segmentation module. Preferably, the segmentation results are saved in NIfTI format, and the dentition, maxilla, mandible, inferior alveolar nerve canal, and maxillary sinus are converted into STL format 3D models through surface meshing for subsequent planning and guide design.
[0030] The assessment module constructs an implant assessment result for the edentulous area based on the distribution information of dangerous anatomical structures such as the inferior alveolar nerve canal and the maxillary sinus floor in three-dimensional space. Preferably, the implant assessment result is a three-dimensional voxel scoring map. The higher the score, the higher the degree to which the area meets the implant safety requirements. The scoring criteria include at least one or more of the following: the safe distance from the candidate location to the inferior alveolar nerve canal and the safe distance from the candidate location to the maxillary sinus floor.
[0031] The planting planning module automatically optimizes the implant pose based on the planting assessment results and preset safety constraints. The pose includes at least the implant position, angle, and length. Preferably, the final implant planning result is determined by traversing or filtering candidate regions with higher scores and applying safety constraints to these regions.
[0032] The implant guide design output module automatically generates implant guide design parameters based on the optimized implant pose and 3D model of the dental arch, and establishes a 3D model of the guide. The guide is preferably a tooth-supported guide, but can be expanded to support mucosa-supported or bone-supported guides. The guide model is finally output in STL format for subsequent 3D printing manufacturing of the implant guide.
[0033] The output module is used to output and / or visualize the three-dimensional reconstruction results of key anatomical structures, implant assessment results, implant planning results, and guide plate models for doctors to review and confirm.
[0034] Example 2: Fault Detection Continuity Constraint Module In this embodiment, the fracture-sensing continuity constraint module is set in the decoding stage of the improved nnU-Net network, or in the feature processing stage after the fusion of decoding features and skip connection features. This module mainly optimizes the inferior alveolar nerve canal for issues such as local fractures, blurred elongated structures, and poor connectivity in CBCT images.
[0035] Specifically, this module can model the continuity consistency of neural tube candidate regions within adjacent slices and local neighborhoods, and perform continuity enhancement processing on suspected broken regions, thereby improving the network's ability to predict the continuity of slender tubular structures. During training, a continuity constraint loss can be applied to the connected regions of the neural tube, making the segmentation results more spatially smooth and coherent. During inference, this module helps reduce discontinuities in the neural tube caused by local grayscale changes, noise interference, or differences in imaging conditions.
[0036] Example 3: Boundary-Semantic Co-fusion Module In this embodiment, the boundary-semantic co-fusion module is positioned at the skip connection fusion location of the improved nnU-Net network, or at the feature recovery location of each level of the decoder. This module is used to simultaneously utilize boundary information in shallow features and semantic information in deep features to enhance the boundary recognition capability of the neural tube and other key anatomical structures.
[0037] Specifically, the boundary-semantic co-fusion module extracts structural boundary responses from shallow features and structural category semantic responses from deep features, and performs weighted or gated fusion of the two types of features to generate fused features that combine boundary sensitivity and semantic discriminative ability. These fused features are used to guide the localization of neural tube edges, making the segmentation results more consistent with real anatomical boundaries.
[0038] Example 4: Construction of Planting Assessment Results and Plant Pose Planning In this embodiment, the system constructs the implant assessment result based on the spatial relationship between the inferior alveolar nerve canal and the maxillary sinus floor. Preferably, for any candidate voxel point within the edentulous area, the minimum distance to the inferior alveolar nerve canal and the minimum distance to the maxillary sinus floor are calculated, and a corresponding score is assigned based on the distance. The larger the distance, the higher the implant safety at that location, and the higher the score.
[0039] After obtaining the implant assessment results, the implant planning module searches for candidate implant positions that meet the implantation conditions within the candidate region that satisfy preset criteria, and determines an optimization plan based on preset safety constraints. Preferably, the safety constraints include at least maintaining a safe distance from the inferior alveolar nerve canal and a safe distance from the maxillary sinus floor. The output results include the implant's position, angle, length, and corresponding safety margin.
[0040] Example 5: Guide Plate Design Output In this embodiment, the guide plate design output module automatically builds a guide plate model based on the optimized implant pose and 3D surface information of the dental arch. The system first determines the contact area of the guide plate on the dental arch, and then automatically generates the position and orientation of the guide holes according to the axial position of the implant, further constructing the overall 3D geometry of the guide plate. The guide plate model can be output as an STL format file for use in 3D printing equipment to manufacture implant guide plates.
[0041] Concluding remarks The above embodiments are merely preferred embodiments of the present invention. Any equivalent substitutions, modifications, or improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit and substance of the present invention shall fall within the protection scope of the present invention.
Claims
1. An implant navigation method for AI-based maxillofacial reconstruction based on oral medical imaging, characterized in that, Includes the following steps: S1. Acquire cone-beam computed tomography (CBCT) image data of the patient's oral and maxillofacial region, and preprocess the CBCT image data. S2. Input the preprocessed CBCT image data into the segmentation model to automatically segment the target anatomical structure and obtain the segmentation result; S3. Based on the segmentation results, the target anatomical structure is reconstructed in three dimensions to generate a corresponding three-dimensional model; S4. Based on the three-dimensional model, determine the candidate planting areas, and construct the planting assessment results according to the spatial relationship between the candidate planting areas and the dangerous anatomical structures; S5. Based on the planting assessment results, optimize the implant pose under preset safety constraints to obtain implant planning results; S6. Generate planting guide design parameters and / or a three-dimensional model of the planting guide based on the implant planning results; S7. Output the three-dimensional reconstruction results, implant planning results, and implant guide plate design results.
2. The planting navigation method according to claim 1, characterized in that, The target anatomical structures mentioned in step S2 include at least one or more of the following: the dentition, maxilla, mandible, inferior alveolar nerve canal, and maxillary sinus.
3. The planting navigation method according to claim 1, characterized in that, The segmentation model mentioned in step S2 is a segmentation network based on deep learning, preferably an encoder-decoder structure network; in some embodiments, the segmentation model is an improved nnU-Net network.
4. The planting navigation method according to claim 3, characterized in that, The segmentation model includes a fracture-aware continuity constraint module and / or a boundary-semantic co-fusion module to improve the continuity representation capability of slender structures and the boundary recognition capability of key anatomical structures.
5. The planting navigation method according to claim 1, characterized in that, The three-dimensional reconstruction in step S3 includes: generating a segmentation label file based on the segmentation results, and generating a three-dimensional surface mesh model through surface meshing processing; the segmentation label file is preferably an NIfTI format file, and the three-dimensional surface mesh model is preferably an STL format file.
6. The planting navigation method according to claim 1, characterized in that, The implant assessment result in step S4 is an implantation area scoring map, which is preferably a three-dimensional voxel scoring map. The score of any candidate location is at least related to its minimum distance to the inferior alveolar nerve canal and / or its minimum distance to the floor of the maxillary sinus.
7. The planting navigation method according to claim 1, characterized in that, Step S5 includes: selecting candidate sites that meet preset conditions from the candidate implantation area, and optimizing the position, angle and length of the implant based on the preset safety constraints; the preset safety constraints include at least maintaining a safe distance from the inferior alveolar nerve canal and / or maintaining a safe distance from the floor of the maxillary sinus.
8. The planting navigation method according to claim 1, characterized in that, Step S6 includes: determining the fitting area of the guide plate base based on the three-dimensional model of the dental arch, determining the position and direction of the guide hole based on the implant pose, establishing a three-dimensional model of the implant guide plate, and outputting the guide plate STL file.
9. An implant navigation system for AI-based maxillofacial reconstruction based on oral medical imaging, characterized in that, include: The data acquisition module is used to acquire CBCT image data of the patient's oral and maxillofacial region; The preprocessing module is used to preprocess the CBCT image data; The automatic segmentation module is used to automatically segment the target anatomical structures in the preprocessed CBCT image data; the 3D reconstruction module is used to generate a 3D model of the target anatomical structure based on the segmentation results; and the evaluation module is used to generate a planting evaluation result based on the spatial relationship between the candidate planting area and the dangerous anatomical structure. The planting planning module is used to generate planting planning results based on the planting assessment results and under preset safety constraints. The guide plate design module is used to generate planting guide plate design parameters and / or a three-dimensional model of the planting guide plate based on the implant planning results. The output module is used to output the 3D reconstruction results, implant planning results, and implant guide plate design results.