Measurement method of buccal-lingual torque of maxillary molars based on improved ResNET

Abstract

The embodiment of the present application discloses a method for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET; the method comprises: obtaining and preprocessing three-dimensional images of teeth and alveolar bones, and performing three-dimensional reconstruction of the preprocessed images; positioning the long axis of the first tooth based on the coupling relationship between the crown and the root; positioning the long axis of the second tooth based on the coupling relationship between the tooth and the alveolar bone; calculating the angle between the long axis of the first tooth and the long axis of the second tooth and generating a new long axis; calculating and verifying the buccal-lingual torque value of the maxillary molar; and achieving accurate measurement and calculation of the torque value of the maxillary first molar through an automated process, thereby improving effectiveness and accuracy.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of torque value measurement, and in particular to a method for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET. Background Art

[0002] With the rapid development of modern oral medicine and artificial intelligence, the precise measurement of tooth posture values ​​is becoming increasingly important in orthodontic treatment. Posture values ​​describe the spatial angle of the tooth's long axis relative to a reference plane (such as the maxillary plane), and their precise measurement directly impacts the design of orthodontic appliances. Accurate assessment of posture values ​​allows doctors to more scientifically design orthodontic treatment plans. However, existing technologies still have limitations in measurement efficiency and accuracy. Summary of the Invention

[0003] The embodiment of the present application provides a method for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET, which solves the problems of feature extraction limitations, high data requirements, high computational complexity and insufficient model generalization ability of existing methods.

[0004] In a first aspect, an embodiment of the present application provides a method for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET, the method comprising the following steps:

[0005] Acquire and pre-process three-dimensional images of teeth and alveolar bones, and perform three-dimensional reconstruction on the pre-processed images;

[0006] Based on the coupling relationship between the crown and the root, the long axis of the first tooth is located;

[0007] Based on the coupling relationship between the tooth and the alveolar bone, the long axis of the second tooth is located;

[0008] Calculate the angle between the long axis of the first tooth and the long axis of the second tooth and generate a new long axis;

[0009] The buccolingual torque values ​​of maxillary molars were calculated and verified.

[0010] Furthermore, the obtaining and preprocessing of three-dimensional images of teeth and alveolar bones, and the three-dimensional reconstruction of the pre-processed images, include:

[0011] Use cone beam computed tomography (CBCT) equipment to obtain three-dimensional images of teeth and alveolar bones;

[0012] For 3D images, median filtering or Gaussian filtering techniques are applied to remove noise introduced during the scanning process, and adaptive histogram equalization (CLAHE) technology is used to enhance the contrast between teeth and alveolar bone, and to extract the region of interest (ROI).

[0013] Open source tools are used to perform three-dimensional reconstruction on the pre-processed images to generate a visual model of the teeth and alveolar bone; the three-dimensional reconstruction includes image segmentation, feature extraction and three-dimensional modeling.

[0014] Furthermore, positioning the long axis of the first tooth based on the coupling relationship between the crown and the root includes:

[0015] Region growing or threshold segmentation algorithms are used to separate the crown and root. Region growing algorithms start from a seed point and gradually add neighboring pixels similar to the seed point until the entire tooth area is covered. Threshold segmentation distinguishes the tooth from the background based on the range of pixel values. A histogram analysis is usually performed first to determine the optimal threshold.

[0016] Use U-Net or its variants to automatically segment the tooth area;

[0017] Extracting point cloud data from the segmented crown and root regions, wherein the point cloud data includes three-dimensional coordinate information of the tooth;

[0018] Principal component analysis (PCA) was applied to calculate the principal axis directions of the crown and root. PCA projects the data into a new coordinate system through linear transformation so that the variance on the principal component of the first coordinate axis is maximized to determine the principal axis direction.

[0019] The midlines of the crown and root are fitted using the least squares method. The best fitting line is found by minimizing the sum of squared errors, and the midlines of the crown and root are connected to form the first tooth long axis.

[0020] Display the fitting results through a visual interface;

[0021] If the long axis is found to be seriously deviated, the position of the connection point between the crown and the root is adjusted.

[0022] Furthermore, positioning the second tooth long axis based on the coupling relationship between the tooth and the alveolar bone includes:

[0023] Perform three-dimensional curvature analysis on the alveolar bone area and identify the main boundaries of the alveolar bone by calculating the curvature of the alveolar bone surface;

[0024] Select the intersection of the extended lines so that the intersection is inside the alveolar bone;

[0025] Use the direction vector of the extended line to calculate its bisection direction. The formula is:

[0026] ;

[0027] Among them, vector L1 and vector L2 are the direction vectors of two extension lines passing through the contact point between the tooth and the alveolar bone; the direction vectors are bisected to determine a direction that takes both the tooth and the alveolar bone into consideration; the bisector is used as the second tooth long axis, and its spatial coordinates are recorded.

[0028] Furthermore, the calculating the angle between the first tooth long axis and the second tooth long axis and generating a new long axis includes:

[0029] Starting from the direction vectors of the long axis of the two teeth, the vectors are obtained by fitting the long axis data and define the direction of the long axis in three-dimensional space;

[0030] The cosine of the angle between two vectors is calculated using the dot product formula:

[0031] ;

[0032] Among them, a and b are two vectors, and is the magnitude of the vector, and θ is the angle between the vectors;

[0033] The included angle θ is obtained by the arccosine function. This angle represents the spatial relationship between the two long axes and is an important parameter for evaluating tooth arrangement and alveolar bone structure.

[0034] The bisector of the angle is used to generate a new major axis. The direction vector of the bisector can be obtained by taking the average of the normalized sum of the two original major axis direction vectors, i.e.:

[0035] ;

[0036] The new long axis is recorded and visualized to be consistent with the actual anatomical structure; wherein, during the visualization process, the position and direction of the new long axis are adjusted.

[0037] Furthermore, the calculation and verification of the buccal-lingual torque value of the maxillary molars includes:

[0038] Determine the angle between the new long axis and the alveolar bone plane, and calculate the angle by vector dot product. The dot product of the direction vector of the new long axis and the normal vector of the alveolar bone plane is used to calculate the cosine of the angle, and then the angle is obtained.

[0039] Calculate the horizontal and vertical offsets between the crown top and the root bottom along the new long axis;

[0040] Determine the precise position of the crown and root and measure the offset relative to the new long axis;

[0041] The calculation formula of the buccal-lingual torque value of maxillary molars is expressed as:

[0042] ;

[0043] Where θ' is the angle between the new long axis and the alveolar bone plane, Δy and Δx are the horizontal and vertical offsets of the crown top and root bottom along the new long axis, respectively;

[0044] Compare the calculated torque value with the normal anatomical reference value or clinical correction target value;

[0045] If the measurement error is large, the CBCT image quality or the accuracy of the segmentation model should be checked. Image quality inspection includes evaluation of resolution, noise level, and contrast parameters, and segmentation model accuracy inspection involves retraining or adjusting model parameters.

[0046] In a second aspect, the present invention also provides a device for measuring the buccal-lingual torque of maxillary molars based on the improved ResNET, comprising:

[0047] An image acquisition module is used to acquire and pre-process three-dimensional images of teeth and alveolar bones, and to perform three-dimensional reconstruction of the pre-processed images;

[0048] A first positioning module is used to locate the long axis of the first tooth based on the coupling relationship between the crown and the root;

[0049] A second positioning module is used to locate the long axis of the second tooth based on the coupling relationship between the tooth and the alveolar bone;

[0050] An angle calculation module, used to calculate the angle between the long axis of the first tooth and the long axis of the second tooth and generate a new long axis;

[0051] The torque calculation module is used to calculate and verify the buccal and lingual torque values ​​of maxillary molars.

[0052] In a third aspect, an embodiment of the present application further provides a computer device, comprising: a memory and one or more processors;

[0053] The memory is used to store one or more programs;

[0054] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET.

[0055] In a fourth aspect, an embodiment of the present application further provides a storage medium comprising computer-executable instructions, which, when executed by a computer processor, are used to execute a method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET as described above.

[0056] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes instructions. When the instructions are executed by a computer, the computer implements the method as described above.

[0057] The embodiment of the present application obtains a three-dimensional image of the teeth and alveolar bone and preprocesses it, and reconstructs the preprocessed image in three dimensions; locates the long axis of the first tooth based on the coupling relationship between the crown and the root; locates the long axis of the second tooth based on the coupling relationship between the tooth and the alveolar bone; calculates the angle between the long axis of the first tooth and the long axis of the second tooth and generates a new long axis; calculates the buccal-lingual torque value of the maxillary molar and verifies it; and through an automated process, achieves accurate measurement and calculation of the torque value, thereby improving effectiveness and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a flow chart of a method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in an embodiment of the present application;

[0059] Figure 2 1 is a schematic structural diagram of a device for measuring buccal-lingual torque of maxillary molars based on an improved ResNET, provided in an embodiment of the present application;

[0060] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present application;

[0061] Figure 4 1 is a torque measurement diagram of a method for measuring buccal-lingual torque of maxillary molars based on an improved ResNET provided in an embodiment of the present application;

[0062] Figure 5 1 is a schematic diagram of a calibration baseline of a method for measuring buccal-lingual torque of maxillary molars based on an improved ResNET provided in an embodiment of the present application;

[0063] Figure 6 This is a schematic diagram of the MLP neural network structure of a method for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] To further clarify the objectives, technical solutions, and advantages of this application, specific embodiments of the present application are described in further detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are intended only to illustrate this application and are not intended to limit it. It should also be noted that, for ease of description, the drawings only illustrate portions relevant to this application, not all of them. Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts depict the various operations (or steps) as sequential processes, many of the operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process may terminate upon completion of its operations, but may also include additional steps not shown in the accompanying drawings. The process may correspond to a method, function, procedure, subroutine, subprogram, or the like.

[0065] 1. Traditional technology

[0066] a) Manual measurement and analysis

[0067] Traditional manual torque measurement and analysis methods are primarily based on two-dimensional images or physical models. These methods typically involve observing and measuring the side of a tooth's clinical crown to determine the angle between its long axis and a reference plane (such as the maxillary plane or true horizontal plane). This involves manually drawing the long axis of the tooth using an intraoral scanner, plaster models, or X-rays, and then calculating the torque using an angle measurement tool.

[0068] This method is characterized by its reliance on the operator's experience and subjective judgment, and measurement accuracy may be limited by model angles, image distortion, or measurement tools. In addition, due to the stereoscopic nature of the tooth's long axis in different directions, two-dimensional images may cause the measurement results to be inconsistent with the actual three-dimensional position. Although manual measurement methods are relatively simple and easy to implement, they lack accuracy and repeatability in complex case analysis and have gradually been replaced by more advanced methods such as CBCT combined with AI technology.

[0069] b) 3D model

[0070] To overcome the limitations of two-dimensional projection measurement, three-dimensional modeling technology has gradually been introduced to the measurement of tooth gravity values. By reconstructing three-dimensional tooth models using CBCT (cone-beam computed tomography) technology, doctors can more intuitively analyze teeth. However, most current applications of three-dimensional models still require doctors to manually calibrate key points of the tooth, such as the cusp and root center, and finally calculate the long axis angle of the tooth. While this method improves the observability of the measurement, the operation steps are cumbersome and repetitive, and it is still susceptible to human influence on the workpiece. Therefore, while three-dimensional model-assisted technology has improved measurement accuracy to a certain extent, its efficiency and consistency issues urgently need to be further optimized.

[0071] 2. Improvement methods based on digital technology

[0072] a) Digital tooth model analysis

[0073] With the widespread adoption of digital technology, digital tooth models generated using CBCT or optical scanning have become widely used in orthodontic treatment evaluation. With the help of specialized software (such as OrthoAnalyzer, InVivo, and Dental), clinicians can extract the long axis of teeth and calculate their inclination angles from digital models. These software programs typically provide a degree of automated analysis, enabling rapid completion of basic measurements by setting key parameters. However, these methods have limitations in complex cases. For example, the accuracy of analysis results is often difficult to guarantee when dealing with abnormal tooth morphology or missing data. Furthermore, operators require a high level of technical expertise, particularly when setting parameters and understanding measurement logic, as errors can easily affect measurement results. These requirements have, to a certain extent, limited the popularity and scope of application of this technology.

[0074] b) Reference plane calculation

[0075] When measuring posture values, a standard reference plane, such as the maxillary plane (PP) or the usual occlusal plane, must be selected. Adjusting the posture between the model and the reference plane is a critical step in the measurement process, but this process often requires manual adjustment of the model's posture. Differences in operator skill and adjustment accuracy can lead to inconsistent planning results, impacting measurement reliability. Therefore, achieving automated planning and maintaining accuracy in complex scenarios is a key area of ​​improvement for digital technology.

[0076] c) Tooth long axis extraction

[0077] The long axis of a tooth is a basic parameter for evaluating the action value of a tooth, and its extraction process is directly related to the accuracy of the measurement results. Existing methods usually determine the long axis direction by determining the geometric center line of the tooth surface. This method performs well when dealing with teeth with regular morphology, but the accuracy of the center line may drop significantly when dealing with complex or abnormal tooth morphology. Especially in the presence of data noise or tooth defects, the deviation of the extraction results will be further increased, affecting the reliability of subsequent measurements. Therefore, long axis extraction technology needs to be more robust to adapt to complex and changing clinical scenarios.

[0078] 3. Preliminary application of artificial intelligence technology

[0079] a) Image Search

[0080] Recent advances in deep learning technology have brought new opportunities to dentistry. Image segmentation techniques based on models such as U-Net have become widely applicable for processing CBCT data, significantly improving the accuracy of tooth and bone segmentation. These techniques train image data using topological neural networks, enabling rapid identification of tooth boundaries and morphology. However, most existing models are limited to the overall distinction between teeth, making it difficult to achieve precise segmentation of tooth morphology, especially during extraction. Further optimizing segmentation models to enhance their fine-grained segmentation capabilities is an important area of ​​artificial intelligence technology in this field.

[0081] b) 3D feature display

[0082] In addition to image segmentation, deep learning techniques have also been used to process 3D point cloud data of teeth, automatically extracting geometric features through models such as PointNet or PointNet++. These models can analyze the spatial morphology of teeth and quickly extract the long axis of the tooth, providing a basis for torque calculation. However, the generalization ability of these methods in complex scenarios still needs to be improved, especially when dealing with data noise and outliers. Their robustness and stability require further verification.

[0083] c) Automatic torque value calculation

[0084] Some studies have attempted to develop automated torque measurement systems by combining CBCT images with artificial intelligence (AI) technology. These systems utilize deep learning models to process data and directly output tooth posture inclination angles and torque values. However, as research in this area is still in its infancy, existing systems still have significant room for improvement in algorithm performance and clinical applicability. Furthermore, challenges with standardizing data from diverse sources hinder widespread application.

[0085] The disadvantages of the above solutions are:

[0086] 1. Manual operation steps are cumbersome and inefficient

[0087] Existing technologies still rely on a significant amount of manual work when measuring tooth torque values, such as model construction, calibration of key gear points, and extraction of the tooth's long axis. Since these steps require high levels of operator expertise, the entire process is not only slow but also prone to low task completion rates. This cumbersome process becomes drastically less efficient when handling large volumes of cases, increasing the workload and potentially leading to lengthy operations that impact the timeliness of diagnostic results. Therefore, simplifying the process and improving operational efficiency have become pressing issues.

[0088] 2. Accuracy and consistency are difficult to guarantee

[0089] Whether using two-dimensional measurement or three-dimensional model-assisted technology, the accuracy and consistency of the results are easily affected by various factors. For example, the layout of the reference plane can lead to different measurement results depending on the operator's skill level and experience. Furthermore, during the extraction of the tooth's long axis, existing algorithms often fail to adapt to the complexity and anomalies of tooth morphology (such as missing or misaligned teeth), resulting in extraction results that often fail to meet clinical requirements. These issues make it difficult to ensure the repeatability and reliability of torque measurement results, potentially impacting the development of treatment cycle plans.

[0090] 3. Insufficient automation

[0091] While some AI technologies have achieved image segmentation and three-dimensional structure, fully automated sequential value calculation systems are completely incompatible with clinical practice. This is partly due to the limited generalization capabilities of existing models when handling complex scenarios; it also reflects the incomplete integration of different technologies. For example, the entire process from CBCT data acquisition to model reconstruction, calculation, tooth long axis elevation, and torque values ​​requires the development of a separate system, resulting in a low level of automation. This not only increases the difficulty of use but also limits the promotion and widespread adoption of the technology.

[0092] Accurate measurement of tooth torque values ​​is of great significance in orthodontic treatment. Although existing technologies have made certain progress in efficiency and accuracy, they still face problems such as cumbersome manual operation, difficulty in ensuring accuracy, insufficient automation, and insufficient applicability. This embodiment of the present application provides a method for measuring the labial (buccal) and lingual torque values ​​of maxillary molars based on an improved ResNET, aiming to address the existing problems of insufficient accuracy, low efficiency, and low automation. To achieve this goal, this embodiment of the present application addresses three aspects: improving measurement accuracy and efficiency and optimizing clinical treatment decision support. Regarding improving measurement accuracy and efficiency, a measurement model of posterior tooth torque values ​​based on the crown-alveolar bone coupling relationship is used to address the issues of insufficient accuracy and low efficiency in traditional technologies, ensuring the accuracy and consistency of measurement results while reducing artifacts and improving work efficiency. Regarding optimizing clinical treatment decision support, a tooth long axis positioning system based on the improved ResNET model and a modified model of the tooth long axis positioning system for head position calibration provide orthodontists with real-time, personalized treatment recommendations and treatment results, promoting the digitalization and standardization of oral treatment.

[0093] The method for measuring the buccal-lingual torque of a maxillary molar based on the improved ResNET provided in the embodiments can be performed by a device for measuring the buccal-lingual torque of a maxillary molar based on the improved ResNET. The device for measuring the buccal-lingual torque of a maxillary molar based on the improved ResNET can be implemented by software and / or hardware and integrated into a device for measuring the buccal-lingual torque of a maxillary molar based on the improved ResNET. The device for measuring the buccal-lingual torque of a maxillary molar based on the improved ResNET can be a computer or other device.

[0094] Figure 1 This is a flow chart of a method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in an embodiment of the present application. Figure 1 , the method comprises the following steps:

[0095] 100. Acquire and pre-process three-dimensional images of teeth and alveolar bones, and reconstruct the pre-processed images in three dimensions.

[0096] Specifically, a cone-beam computed tomography (CBCT) device is used to obtain three-dimensional images of teeth and alveolar bones. For the three-dimensional images, median filtering or Gaussian filtering techniques are applied to remove noise introduced during the scanning process. Adaptive histogram equalization (CLAHE) technology is used to enhance the contrast between teeth and alveolar bones, and regions of interest (ROIs) are extracted. Open-source tools are used to perform three-dimensional reconstruction of the preprocessed images to generate a visual model of the teeth and alveolar bones. The three-dimensional reconstruction includes image segmentation, feature extraction, and three-dimensional modeling.

[0097] 200. Based on the coupling relationship between the crown and the root, the long axis of the first tooth is positioned.

[0098] Specifically, a region growing or threshold segmentation algorithm is used to separate the crown and root; the region growing algorithm starts from the seed point and gradually adds neighboring pixels similar to the seed point until the entire tooth area is covered. The threshold segmentation is used to distinguish the teeth from the background based on the range of pixel values. Usually, a histogram analysis is required to determine the optimal threshold. The U-Net or its variant network is used to automatically segment the tooth area and extract point cloud data from the segmented crown and root areas, where the point cloud data includes the three-dimensional coordinate information of the tooth.

[0099] Principal component analysis (PCA) is applied to calculate the principal axis direction of the crown and root. PCA projects the data into a new coordinate system through linear transformation so that the variance on the principal component of the first coordinate axis is maximized to determine the principal axis direction; the least squares method is used to fit the midline of the crown and root; the best fitting line is found by minimizing the sum of squares of the errors, and the midline of the crown and root are connected to form the first long axis of the tooth; the fitting results are displayed through a visual interface; if the long axis deviation is found to be serious, the position of the connection point between the crown and the root is adjusted.

[0100] 300. Based on the coupling relationship between the tooth and the alveolar bone, the long axis of the second tooth is positioned.

[0101] Specifically, a three-dimensional curvature analysis is performed on the alveolar bone area. The main boundary of the alveolar bone is identified by calculating the curvature of the alveolar bone surface. The intersection point of the extension line is selected so that the intersection point is located inside the alveolar bone. The direction vector of the extension line is used to calculate its bisection direction. The formula is:

[0102] ;

[0103] Vectors L1 and L2 are the direction vectors of two extension lines passing through the contact point between the tooth and the alveolar bone. The direction vectors are bisected to determine a direction that takes both the tooth and the alveolar bone into account. The bisector is used as the second tooth long axis, and its spatial coordinates are recorded. Vectors L1 and L2 are generated by vector addition and normalization to generate the bisecting direction of the two extension lines. This is used to correct the positioning of the tooth long axis, improving the robustness of long axis extraction, reducing manual intervention, and adapting to complex anatomical structures.

[0104] 400. Calculate the angle between the long axis of the first tooth and the long axis of the second tooth and generate a new long axis.

[0105] Specifically, we start with the direction vectors of the two tooth long axes, where the vectors are obtained by fitting the long axis data and define the direction of the long axis in three-dimensional space. We use the dot product formula to calculate the cosine of the angle between the two vectors. The dot product formula is:

[0106] ;

[0107] Among them, a and b are two vectors, and is the modulus of the vector, and θ is the angle between the vectors; the angle θ is obtained by the inverse cosine function. This angle represents the spatial relationship between the two long axes and is an important parameter for evaluating tooth arrangement and alveolar bone structure.

[0108] The bisector of the angle is used to generate a new major axis. The direction vector of the bisector can be obtained by taking the average of the normalized sum of the two original major axis direction vectors, i.e.:

[0109] ;

[0110] The new long axis is recorded and visualized to be consistent with the actual anatomical structure; wherein, during the visualization process, the position and direction of the new long axis are adjusted.

[0111] 500. Calculate and verify the buccolingual torque value of maxillary molars.

[0112] Specifically, the angle between the new long axis and the alveolar bone plane is determined and calculated by vector dot product. The direction vector of the new long axis is dot-producted with the normal vector of the alveolar bone plane, and the result is used to calculate the cosine of the angle, and then the angle is obtained. The horizontal and vertical offsets of the crown top and the root bottom along the new long axis are calculated. The precise position of the crown and root is determined, and the offset relative to the new long axis is measured. The formula for calculating the buccal-lingual torque value of the maxillary molar is expressed as:

[0113] ;

[0114] Where θ' is the angle between the new long axis and the alveolar bone plane, and Δy and Δx are the horizontal and vertical offsets of the crown top and root bottom along the new long axis, respectively.

[0115] The calculated torque value is compared with the normal anatomical reference value or the clinical correction target value. If the measurement error is large, the CBCT image quality or the accuracy of the segmentation model is checked. The image quality check includes the evaluation of resolution, noise level and contrast parameters, and the accuracy check of the segmentation model involves retraining or adjusting the model parameters.

[0116] As described above, the embodiment of the present application obtains a three-dimensional image of the teeth and alveolar bone and performs preprocessing, and performs three-dimensional reconstruction on the preprocessed image; based on the coupling relationship between the crown and the root, the long axis of the first tooth is positioned; based on the coupling relationship between the tooth and the alveolar bone, the long axis of the second tooth is positioned; the angle between the long axis of the first tooth and the long axis of the second tooth is calculated and a new long axis is generated; the buccal-lingual torque value of the maxillary molar is calculated and verified; through an automated process, accurate measurement and calculation of the torque value is achieved, thereby improving effectiveness and accuracy.

[0117] For example, based on the crown-root coupling relationship, the CBCT image is used to obtain Figure 4 The true vertical line shown forms the long axis positioning of the first tooth. This method takes into account the relationship between the teeth themselves and believes that the formation of the torque value is mainly the deflection formed by the long-term stress of the posterior teeth. Figure 4The Lnew shown forms the second tooth long axis position. This method considers the relationship between the tooth and alveolar bone, assuming that torque is primarily caused by the deflection of the tooth by the opposing force of the alveolar bone. The angle between the two tooth long axes is simultaneously calculated. While ensuring ipsilateral alignment, the bisector of this new angle is used as the new tooth long axis. This deviation-balancing method ensures that the forces acting on the tooth and alveolar bone have equal impact.

[0118] The specific implementation process can be divided into the following parts:

[0119] Acquisition and preprocessing of CBCT image data

[0120] The data acquisition phase begins in the dental 3D image analysis process. Using cone-beam computed tomography (CBCT) equipment, detailed 3D images of the patient's teeth and alveolar bone can be obtained. To ensure image clarity and low noise, the precise settings of scanning parameters are crucial. These include setting the resolution to 0.3 mm or higher to capture fine structures, adjusting the voltage between 70 and 120 kVp to achieve adequate X-ray penetration, and controlling the current between 5 and 10 mA to balance image quality and user radiation dose. Accurately setting these parameters is crucial for subsequent diagnosis and treatment planning.

[0121] The second stage is image preprocessing. First, to remove noise introduced during the scanning process, median filtering or Gaussian filtering techniques are applied. These methods can effectively reduce noise while preserving edge information. Then, to enhance the contrast between the teeth and the alveolar bone, adaptive histogram equalization (CLAHE) technology is used. This is a local contrast enhancement method that can significantly improve the clarity of the boundary area. Finally, the region of interest (ROI) is extracted. This step manually or automatically marks the tooth area and eliminates irrelevant background to reduce interference for subsequent analysis.

[0122] Finally, the 3D reconstruction phase begins. Using open-source tools such as ITK-SNAP or 3D Slicer, CBCT images are reconstructed in 3D to generate a visual model of the teeth and alveolar bone. This step involves complex image processing and computer vision techniques, including image segmentation, feature extraction, and 3D modeling. The resulting 3D model intuitively displays the structure of the teeth and alveolar bone, providing strong visual support for clinical diagnosis and treatment. These techniques achieve detailed extraction of precise 3D information from the original CBCT images, laying a solid foundation for subsequent dental analysis and treatment planning.

[0123] Positioning of the long axis of the first tooth: Based on the crown-root coupling relationship, tooth segmentation:

[0124] Traditional methods use region growing or threshold segmentation algorithms to separate the crown and root. Region growing algorithms start with a seed point and gradually add neighboring pixels similar to the seed point until the entire tooth area is covered. Threshold segmentation, on the other hand, distinguishes the tooth from the background based on the range of pixel values, often requiring a histogram analysis to determine the optimal threshold.

[0125] Deep learning methods use U-Net or its variants, which are trained on a large amount of labeled dental image data to automatically segment tooth regions. The U-Net structure consists of a contraction path to capture contextual information and a symmetrical expansion path to accurately locate tooth boundaries, thus achieving high-precision automatic segmentation.

[0126] Centerline Fitting:

[0127] First, point cloud data are extracted from the segmented crown and root areas, which contain the three-dimensional coordinate information of the teeth.

[0128] Next, principal component analysis (PCA) was applied to calculate the principal axis directions of the crown and root. PCA projects the data into a new coordinate system through linear transformation so that the variance on the first coordinate axis (principal component) is maximized, thereby determining the principal axis directions.

[0129] Finally, the midline of the crown and root is fitted using the least squares method. This method finds the best-fit line by minimizing the sum of squared errors and connects the midline of the crown and root to form the first tooth long axis.

[0130] Inspection and correction:

[0131] The fitting results are displayed through a visual interface, and professionals can check whether the long axis deviates from the actual anatomical shape to ensure the accuracy of the fitting.

[0132] If the long axis is found to be seriously deviated, the system allows manual adjustment of the crown-root connection point. This interactive correction process ensures the accuracy of the final result, so that the fitting of the tooth long axis can truly reflect the anatomical structure of the tooth.

[0133] Positioning of the long axis of the second tooth: Based on the coupling relationship between the tooth and the alveolar bone, the alveolar bone extension line is extracted:

[0134] First, a three-dimensional curvature analysis is performed on the alveolar bone area. This step identifies the main boundaries of the alveolar bone by calculating the curvature of the alveolar bone surface. Curvature analysis can reveal the geometric properties of the alveolar bone surface and help determine the edges of the alveolar bone. Next, the extension line of the alveolar bone is fitted using the contact points between the teeth and the alveolar bone. This usually involves linear regression or least squares method to determine the best straight line passing through the contact points. Finally, the intersection point of the extension line is selected, ensuring that the intersection point is inside the alveolar bone. This step requires precise geometric calculations to ensure that the intersection point is not only geometrically feasible but also anatomically meaningful.

[0135] Bisector calculation:

[0136] Use the direction vector of the extended line to calculate its bisection direction. The formula is:

[0137] ;

[0138] The direction vector is bisected to determine an orientation that takes both the tooth and the alveolar bone into account. The bisector is used as the second tooth's long axis, and its spatial coordinates are recorded. This step involves parameterizing the bisector in three-dimensional space, typically using parametric equations to describe the line's position and orientation in space.

[0139] Calculate the angle between the two tooth long axes and generate a new long axis. Angle calculation:

[0140] First, we need to start with two vectors that define the direction of the long axis of the tooth. These vectors can be obtained by fitting the long axis data, and they define the direction of the long axis in three-dimensional space. Next, we use the dot product formula to calculate the cosine of the angle between the two vectors. The dot product formula is

[0141] , where a and b are two vectors, and is the magnitude of the vector, and θ is the angle between the vectors. Finally, the angle θ is calculated using the arccosine function. This angle represents the spatial relationship between the two long axes and is an important parameter for evaluating tooth alignment and alveolar bone structure.

[0142] New major axis generation:

[0143] The bisector of the angle is used to generate a new major axis. The direction vector of the bisector can be obtained by taking the average of the normalized sum of the two original major axis direction vectors, i.e.:

[0144] ;

[0145] The new long axis is recorded and visualized to ensure that it is consistent with the actual anatomy. This step usually involves drawing the new long axis in 3D visualization software and comparing it with the original image data to verify the accuracy and practicality of the new long axis. During the visualization process, the position and orientation of the new long axis can be adjusted to ensure that it matches the actual anatomy of the teeth and alveolar bone. Measurement and verification of torque values, torque value calculation:

[0146] First, determine the angle between the new long axis and the alveolar bone plane. This angle can be calculated by a vector dot product, where the direction vector of the new long axis is dot-producted with the normal vector of the alveolar bone plane, and the result is used to calculate the cosine of the angle, and then the angle is obtained. Next, calculate the horizontal and vertical offsets (Δy and Δx) of the top of the crown and the bottom of the root along the new long axis. This usually involves positioning the tooth model in three-dimensional space, determining the precise position of the crown and root, and measuring their offset relative to the new long axis. The formula for calculating the torque value can be expressed as:

[0147] ;

[0148] where θ' is the angle between the new long axis and the alveolar bone plane, and Δy and Δx are the horizontal and vertical offsets of the crown top and root bottom along the new long axis, respectively.

[0149] Verification results:

[0150] The calculated torque values ​​are compared with normal anatomical reference values ​​or clinical correction targets. This step verifies that the measured values ​​are within acceptable limits and meet the clinical correction targets. If the measurement error is large, the CBCT image quality or the accuracy of the segmentation model needs to be checked. Image quality checks may include assessments of resolution, noise levels, and contrast parameters. Checking the accuracy of the segmentation model may involve retraining or adjusting model parameters to ensure accurate segmentation of the tooth area.

[0151] Furthermore, a tooth long-axis positioning system based on an improved ResNet model utilizes an improved ResNet18 architecture (ResNet18-s) during CBCT (Cone Beam Computed Tomography) image training to meet the real-time computing requirements of field deployment. The SE (Squeeze-and-Excitation) mechanism optimizes computation time. The SE module adaptively assigns weights to each feature channel, increasing the network's sensitivity to key features while reducing reliance on irrelevant information. Embedded in the network's feature extraction phase, it enhances the model's expressive power.

[0152] The SE module of the input feature map first performs global average pooling (GAP) on the spatial information of each channel and calculates the global statistics of each channel:

[0153] ;

[0154] Then, two fully connected layers and nonlinear activation functions are used to model the dependencies between channels and generate the weight ratio s for each channel:

[0155] ;

[0156] in, It is the weight matrix of the first fully connected layer, which is used to compress the channel dimension and reduce the computational cost. is the weight matrix of the second fully connected layer, used to restore the channel dimension. is the Sigmoid activation function, which is used to normalize the channel weights.

[0157] Finally, the original feature map is reweighted using the channel weight ratio s to obtain the output feature map:

[0158] ;

[0159] The residual block of ResNet consists of two parts: the main branch and the bypass connection. The main branch contains two convolutional layers and their corresponding batch normalization (BN) and activation function (ReLU). The bypass connection directly adds the input feature map to the output of the main branch:

[0160] ;

[0161] In ResNet18-s, the SE module directly replaces the convolutional layer and its activation process of the main branch while retaining the bypass connection. The expression after replacement is:

[0162] ;

[0163] in is the feature weighted by the SE module.

[0164] Because the SE module's computation involves only distributing weights between channels, its spatial complexity is much lower than that of the original convolution operation. Therefore, in ResNet18, SE modules are used to replace specific layers, reducing computational effort while improving the model's ability to focus on key features. Specifically, for ResNet18-s: layers 3, 4, and 5 are combined into a single SE module; layers 8, 9, 10, and 11 are combined into a single SE module; and layers 12, 13, 14, and 15 are combined into a single SE module. By eliminating redundant convolution operations, computational overhead is significantly reduced.

[0165] After training is completed, the torque value of each person is calculated using the following formula during inference:

[0166] ;in,

[0167] are the torque values ​​output by the anterior teeth model, canine teeth model, and posterior teeth model respectively; is the torque value of the global model.

[0168] Modified model of tooth long axis positioning system combined with head position calibration: The key to tooth long axis positioning is the head position level. Based on CBCT images, the palatal suture, anterior nasal ridge point-posterior nasal ridge point (ANS-PNS) and nasal septum are determined as three calibration baselines, such as Figure 5 As shown, the tooth long axis positioning is pre-corrected. The correction model uses the Unet framework, marking the left and right segments of the three calibration baselines. The imbalance ratio is determined based on the area ratio. The three proportional coefficients are input into an MLP to calculate the 3D rotation angle. The CBCT rotation angle is adjusted according to the calculated value until the imbalance ratio is 1. Head position calibration is complete.

[0169] The specific implementation process can be divided into the following parts: data acquisition and preliminary preprocessing, CBCT image acquisition:

[0170] The use of cone-beam computed tomography (CBCT) is a key step in acquiring 3D imaging data of the user's head, enabling the capture of detailed structures including the skull and teeth. To ensure clear visibility of fine structures such as the nasal septum, mid-palatal suture, and ANS-PNS baseline in the images, the scan resolution must be set to 0.3 mm or higher. This high resolution setting is crucial for subsequent diagnosis and treatment planning, as it provides sufficient detail for professional analysis.

[0171] Image preprocessing:

[0172] When processing CBCT data, Gaussian filtering is first used to reduce image noise, eliminating artifacts and noise that could affect image quality. This step is crucial for improving the accuracy of subsequent analysis. Next, image normalization is performed to adjust the grayscale values ​​of all image data to the standard interval [0, 1]. This enhances data consistency across users and facilitates comparative analysis. Finally, regions of interest (ROIs) are manually or automatically annotated, such as key structures such as the nasal cavity, mid-palatal suture, and ANS-PNS. This cropping step not only reduces data processing volume but also significantly improves the efficiency of subsequent processing and analysis. These three steps together constitute the core process of image preprocessing, providing a high-quality imaging data foundation for subsequent diagnosis and treatment planning.

[0173] Extraction of calibration baseline, extraction of palatal suture:

[0174] In the initial localization phase, a deep learning segmentation model, such as the U-Net architecture, is first trained on previously labeled data to accurately extract the palatal suture region from CBCT images. The U-Net model is widely used in medical image processing for its excellent image segmentation capabilities, capable of identifying and segmenting complex biological structures. After segmentation, the segmentation results are further processed using a centerline extraction algorithm to obtain the central axis of the palatal suture. This step is critical, as it provides a precise anatomical reference line for subsequent analysis. Subsequently, during the optimization and smoothing phase, the central axis is smoothed using polynomial fitting or spline interpolation techniques. These mathematical tools effectively interpolate points on the central axis, reducing fluctuations caused by segmentation errors or noise, thereby ensuring the continuity and smoothness of the central axis and providing a more accurate and reliable data foundation for subsequent measurement and analysis. These technical implementation details collectively ensure the accuracy and usability of the central axis of the palatal suture extracted from CBCT data, providing strong technical support for clinical applications.

[0175] ANS-PNS line extraction:

[0176] In oral and maxillofacial imaging analysis, accurate identification of the anterior nasal spine (ANS) and posterior nasal spine (PNS) is crucial. First, the palatal bone region is extracted using a deep learning segmentation model, and shape-based algorithms, such as curvature analysis, are used to detect the ANS at the anterior end of the palatal bone. The PNS is then determined at the posterior end of the palatal bone using an edge detection algorithm, which may involve extreme point search and gradient analysis to ensure its accuracy. Finally, a straight line vector between the ANS and PNS is calculated and linear interpolation is used to connect these two points to form a calibration baseline, providing a precise reference for maxillofacial structure analysis. The combination of these techniques ensures the accuracy of key anatomical points extracted from the imaging data and the reliability of the calibration baseline, providing strong technical support for clinical diagnosis and treatment planning.

[0177] Septum extraction:

[0178] During the automatic segmentation phase, the U-Net deep learning model is first applied to accurately segment the nasal cavity region in CBCT images. This model identifies and extracts the center plane of the nasal septum. Widely used for its outstanding performance in image segmentation tasks, U-Net's architecture consists of a contraction path to capture contextual information and a symmetric expansion path for precise localization. After segmenting the center plane of the nasal septum, a 3D skeletonization operation, also known as a morphological thinning process, is then used to extract the septum's centerline. This step highlights the central axis of the 3D structure, providing a precise geometric reference for subsequent analysis. During the symmetry analysis phase, the system automatically compares the volumes of the two sides of the nasal cavity to verify the accuracy of the nasal septum segmentation. This typically involves calculating the difference in volume between the two sides and comparing it to a preset threshold. If significant deviations are detected, the system automatically corrects them or prompts the operator through the user interface for manual review to ensure the accuracy of the analysis results. This process may involve machine learning algorithms to identify and correct deviations in the segmentation, or geometric algorithms to adjust the centerline position to ensure the geometric centerline of the nasal septum, thereby providing reliable clinical analysis results.

[0179] Baseline segmentation and imbalance ratio calculation

[0180] In the baseline segmentation process, the output of the deep learning segmentation model is first used to split the three baselines of the palatal suture, ANS-PNS, and nasal septum into two along their center lines, and the left and right areas are extracted respectively. Then, by calculating the number of pixels or volume of each area, the segmented area A of the left and right sides is obtained. L and A R Specifically, for each baseline, the imbalance ratio is calculated as:

[0181] ;in,

[0182] represents the absolute difference between the left and right areas, The value of this proportional coefficient R ranges from 0 to 1. The closer the value is to 0, the more balanced the left and right sides are, while the closer the value is to 1, the more unbalanced the left and right sides are.

[0183] Next, the imbalance ratios are calculated for the three baselines: the palatal suture, the ANS-PNS, and the nasal septum, resulting in three proportional coefficients, R1, R2, and R3. These coefficients can provide clinicians with a quantitative reference for the degree of septal deviation, helping them assess the necessity and plan for surgical correction. For example, if any of the values ​​of R1, R2, or R3 exceeds a preset threshold, it may indicate significant septal deviation and require further medical intervention.

[0184] Implementation of the head position calibration model, model selection:

[0185] like Figure 6 As shown, a calibration correction model based on a multi-layer perceptron (MLP) is used to input the imbalance ratios of the three baselines and output the 3D rotation angle required for calibration.

[0186] MLP structure design:

[0187] When designing a multilayer perceptron (MLP) architecture to predict 3D rotation angles, we first construct an input layer that receives the three scaling coefficients R1, R2, and R3 obtained from the baseline segmentation and imbalance scale calculation steps. These coefficients serve as input features for the network to predict the required rotation angle for septal correction. For 3D rotation, rotation matrices can be used to represent rotations around the X, Y, and Z axes. For example, the rotation matrix for a rotation angle θ around the Z axis is:

[0188] ;

[0189] Next, the MLP structure includes two hidden layers, each of which uses a fully connected network. Fully connected layers map input features to a higher-dimensional space, enabling the network to learn complex relationships between input data. In both hidden layers, ReLU (Rectified Linear Unit) activation functions introduce nonlinearity, enhancing the network's expressive power while also preventing the vanishing gradient problem and making network training more efficient.

[0190] Finally, the MLP output layer is designed to output three values, corresponding to the rotation angles α, β, and γ about the X, Y, and Z axes, respectively. These rotation angles are predicted based on the input imbalance coefficients and are used to guide the actual septum correction surgery. By minimizing the error between the predicted rotation angles and the actual required rotation angles, the MLP model can be trained to accurately predict the 3D rotation angles required to correct septal deviation.

[0191] The entire MLP architecture is designed to automatically provide precise guidance for septum correction surgery by learning the relationship between the imbalance ratio of baseline segmentation and the corrective rotation angle. This machine learning-based approach can improve the accuracy and efficiency of surgery and reduce reliance on manual experience.

[0192] Model training:

[0193] During the model training phase, the dataset is prepared. This involves collecting a series of labeled CBCT images and calculating the required rotation angles (α, β, and γ) for each sample. These angles serve as labels for model training. This labeled data is typically manually annotated by experts or obtained through precise medical measurements, ensuring the accuracy and reliability of the training data.

[0194] Next, we define the model's loss function as the mean squared error (MSE), which measures the average of the squares of the differences between the model's predictions and the true values. Using MSE as the loss function helps optimize the model parameters so that the predicted rotation angles more accurately match the actual rotation angles.

[0195] The Adam optimizer was chosen as the optimizer. This is an adaptive learning rate optimization algorithm widely used for its high efficiency in solving non-convex optimization problems. The initial learning rate was set to 1e-4, a common starting value that can be adjusted based on actual training conditions.

[0196] During training, the dataset is divided into multiple batches. For each batch, forward propagation is performed to calculate the error between the predicted rotation angle and the true angle. This error is then propagated back through the backpropagation algorithm to update the model's weights and biases. This process is repeated until the model's performance on the validation set stops improving or the preset number of iterations is reached.

[0197] Through this training process, the model learns how to accurately predict the required 3D rotation angle based on the input scale factors R1, R2, and R3 to calibrate the nasal septum. This deep learning-based method can automatically extract features from complex imaging data and predict precise rotation angles, providing powerful support for clinical surgery.

[0198] 5) Head position calibration:

[0199] During head alignment, a trained multi-layer perceptron (MLP) model is first used to infer the input imbalance coefficients R1, R2, and R3. Specifically, these coefficients are used as input to the model, and the model output is calculated through forward propagation: the predicted 3D rotation angles (α, β, and γ). These angles correspond to the rotations around the X, Y, and Z axes, respectively. This step involves applying the model's weights and biases to the input data, and calculating the final output through activation functions and connections between layers.

[0200] After obtaining the predicted rotation angles, the next step is to adjust the pose of the CBCT image based on these angles. Technically, this typically involves applying a rotation transformation to the image data using a 3D graphics library such as VTK or ITK. In practice, a rotation matrix is ​​constructed, defined based on the predicted rotation angles α, β, and γ. This rotation matrix is ​​then applied to each voxel or pixel in the CBCT image to achieve spatial rotation of the image.

[0201] After rotation adjustment, the imbalance ratios of the three baselines are recalculated to evaluate the effectiveness of the calibration. If the imbalance ratios have not yet reached the target of 1, meaning that the volumetric difference between the left and right sides is still significant, the above steps are repeated, using the MLP model for inference again and adjusting the image pose based on the new rotation angle. This process is iterated until the imbalance ratios of all three baselines are close to 1, indicating that head alignment is complete and the CBCT image has achieved the desired left-right symmetry.

[0202] The key to this process lies in the accuracy of the MLP model and the fine control of iterative calibration, which ensures gradual adjustment from the initial pose to the final pose, ultimately achieving precise head alignment. By automating this process, the efficiency and accuracy of head alignment can be significantly improved, providing high-quality imaging data for subsequent clinical analysis and treatment planning.

[0203] 6) Optimization of alignment of posterior teeth long axis

[0204] In the calibrated CBCT image, the long axis of the tooth is first re-extracted. Specifically, the first step is to use image processing and pattern recognition techniques, such as Hough transform or deep learning segmentation model, to identify and extract the coupling part of the crown and root to obtain the first long axis of the tooth. The key to this step is to accurately identify the boundary and direction of the tooth in order to accurately determine the position of the long axis. Next, by analyzing the coupling area between the tooth and the alveolar bone, similar image analysis technology is used to recalculate the second long axis of the tooth, which requires an in-depth understanding of the anatomical structure of the tooth and the alveolar bone.

[0205] During the revised long axis angle calculation phase, the angle between the two long axes is recalculated based on the calibrated tooth long axis data. This typically involves vector operations, such as dot or cross products, to determine the spatial relationship between the two axes. Once the angle is calculated, the angle bisector is determined, which can be achieved through vector decomposition and rotation operations. This results in a more accurate new long axis that represents the average orientation of the tooth.

[0206] During the error verification phase, the tooth long axis angles before and after calibration are compared to verify the effect of head position calibration on tooth long axis positioning. This step evaluates the effect of calibration by comparing the angle changes before and after calibration. If the angle change is within the expected range, the calibration is considered successful. If the error exceeds the expected range, the output of the calibration model will be checked, including the accuracy of the predicted rotation angle and the correctness of the proportional coefficients R1, R2, and R3 input into the model. This may involve retraining or adjusting the MLP model, or rechecking the input data to ensure that all data is accurate.

[0207] Through this series of technical implementation steps, the head position can be accurately calibrated and the long axis of the tooth can be accurately positioned and measured.

[0208] As mentioned above, existing traditional motion control methods, such as two-dimensional cephalometric measurement and three-dimensional model analysis, often rely on manual operation and experience judgment, and the measurement accuracy is increasingly affected by the operator's skills and additional factors. The embodiment of the present application uses deep learning technology to automatically process the three-dimensional data of the teeth, and uses AI algorithms to accurately obtain the long axis of the teeth and calculate the displacement value, which greatly reduces manual work and provides higher accuracy and consistency. Through the automated process, the gravity value measurements of all users will remain highly consistent, avoiding deviations caused by operational deviations.

[0209] Traditional techniques typically require orthodontists to manually calibrate key points, adjust reference layouts, extract long axes, and perform other tedious tasks. However, the fully automated process in this embodiment, from CBCT data input to action value calculation, is automatically completed by the AI ​​system, significantly improving work efficiency. Data upload and treatment plan confirmation are now possible, reducing tedious manual operations, saving significant time, and enabling the processing of more data.

[0210] Tooth posture measurement methods typically rely on professional experience, but can result in tables in complex cases. For example, in cases with complex tooth morphology or uneven dentition, traditional methods struggle to ensure accuracy. However, the present embodiment utilizes deep learning and artificial intelligence technologies to achieve automated tooth segmentation and long axis removal, eliminating the significant impact and potential errors associated with manual operation, reducing operational risks, and ensuring the accuracy and reliability of measurement results.

[0211] Existing technologies often require manual processing when dealing with complex tooth morphologies, malocclusion, and other special cases, which increases workload and complexity. However, the system in the embodiment of the present application uses a deep learning model to automatically identify and process complex tooth morphologies, reducing doctor intervention, enabling efficient processing of various complex cases, and improving the system's suitability and versatility.

[0212] Existing technologies often suffer from data incompatibility or poor information transfer between multiple software tools, resulting in cumbersome operations and difficulty in analysis. The system in the embodiment of the present application seamlessly integrates modules such as data input, 3D model reconstruction, long axis extraction, and motion calculation, providing a unified platform that reduces the complexity of data processing and greatly improves operational convenience. The user-friendly interface makes it easy for doctors to get started and quickly complete the entire measurement process.

[0213] Based on the above examples, please refer to Figure 2 , an embodiment of the present application provides a device for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET, and the device for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET specifically includes: an image acquisition module 201, a first positioning module 202, a second positioning module 203, an angle calculation module 204 and a torque calculation module 205.

[0214] Among them, the image acquisition module 201 is used to acquire three-dimensional images of teeth and alveolar bones and perform preprocessing, and perform three-dimensional reconstruction of the preprocessed images; the first positioning module 202 is used to locate the long axis of the first tooth based on the coupling relationship between the crown and the root; the second positioning module 203 is used to locate the long axis of the second tooth based on the coupling relationship between the teeth and the alveolar bone; the angle calculation module 204 is used to calculate the angle between the long axis of the first tooth and the long axis of the second tooth and generate a new long axis; the torque calculation module 205 is used to calculate the buccal-lingual torque value of the maxillary molar and verify it.

[0215] As described above, the embodiment of the present application obtains a three-dimensional image of the teeth and alveolar bone and performs preprocessing, and performs three-dimensional reconstruction on the preprocessed image; based on the coupling relationship between the crown and the root, the long axis of the first tooth is positioned; based on the coupling relationship between the tooth and the alveolar bone, the long axis of the second tooth is positioned; the angle between the long axis of the first tooth and the long axis of the second tooth is calculated and a new long axis is generated; the buccal-lingual torque value of the maxillary molar is calculated and verified; through an automated process, accurate measurement and calculation of the torque value is achieved, thereby improving effectiveness and accuracy.

[0216] The device for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in the embodiment of the present application can be used to execute the method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in the above embodiment, and has corresponding functions and beneficial effects.

[0217] An embodiment of the present application also provides a computer device that can integrate the device for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in an embodiment of the present application. Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present application. Figure 3 The computer device includes: an input device 33, an output device 34, a memory 32, and one or more processors 31; the memory 32 is used to store one or more programs; when the one or more programs are executed by the one or more processors 31, the one or more processors 31 implement the method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET as provided in the above embodiment. The input device 33, the output device 34, the memory 32, and the processor 31 can be connected by a bus or other means. Figure 3 The bus connection is taken as an example.

[0218] The processor 31 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 32, that is, realizes the above-mentioned method for measuring the buccal-lingual torque value of the maxillary molar based on the improved ResNET.

[0219] The computer device provided above can be used to execute the method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in the above embodiment, and has corresponding functions and beneficial effects.

[0220] An embodiment of the present application also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to execute a method for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET. The method for measuring the buccal-lingual torque value of maxillary molars based on an improved ResNET includes: adding scheduling information through the scheduling system background, wherein the scheduling information includes point positions, paths, and action information; based on the starting point and target point of the robot, automatically generating an operation path and dispatching it to the robot; according to the real-time task status and real-time working status of the elevator, combined with the real-time action information of the robot, adjusting the operation path in real time, and controlling the robot to operate.

[0221] Storage medium - any of various types of memory devices or storage devices. The term "storage medium" is intended to include: installation media, such as CD-ROMs, floppy disks, or tape devices; computer device memory or random access memory, such as DRAM, DDRRAM, SRAM, EDORAM, Rambus RAM, etc.; non-volatile memory, such as flash memory, magnetic media (such as hard disks or optical storage); registers or other similar types of memory elements, etc. Storage media may also include other types of memory or combinations thereof. In addition, the storage medium may be located in the first computer device where the program is executed, or it may be located in a different second computer device that is connected to the first computer device via a network (such as the Internet). The second computer device can provide program instructions to the first computer for execution. The term "storage medium" can include two or more storage media that can reside in different locations (e.g., in different computer devices connected via a network). The storage medium can store program instructions (e.g., embodied as a computer program) that can be executed by one or more processors.

[0222] Of course, the storage medium containing computer-executable instructions provided in an embodiment of the present application is not limited to the computer-executable instructions of the method for measuring the buccal-lingual torque value of the maxillary molar based on the improved ResNET as described above, and can also execute the relevant operations in the method for measuring the buccal-lingual torque value of the maxillary molar based on the improved ResNET provided in any embodiment of the present application.

[0223] The measuring device, storage medium and computer equipment for the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in the above embodiments can execute the measuring method for the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in any embodiment of the present application. For technical details not described in detail in the above embodiments, please refer to the measuring method for the buccal-lingual torque value of maxillary molars based on the improved ResNET provided in any embodiment of the present application.

[0224] The embodiments of the present application also provide a computer program product. The methods described in the various embodiments of the present application can be implemented in whole or in part through software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the various embodiments of the present application are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, a core network device, an OAM (Open Application Model), or other programmable device.

[0225] The above are only preferred embodiments of the present application and the technical principles employed. The present application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions that are possible for those skilled in the art will not depart from the scope of protection of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments and may include more other equivalent embodiments without departing from the concept of the present application. The scope of the present application is determined by the scope of the claims.

Claims

1. A method for measuring the buccal-lingual torque of maxillary molars based on an improved ResNET, characterized in that: The method comprises the following steps: Acquire and pre-process three-dimensional images of teeth and alveolar bones, and perform three-dimensional reconstruction on the pre-processed images; Based on the coupling relationship between the crown and the root, the long axis of the first tooth is located; Based on the coupling relationship between the tooth and the alveolar bone, the long axis of the second tooth is located; Calculate the angle between the long axis of the first tooth and the long axis of the second tooth and generate a new long axis; Calculate and verify the buccal-lingual torque value of maxillary molars; Positioning the second tooth long axis based on the coupling relationship between the tooth and the alveolar bone includes: Perform three-dimensional curvature analysis on the alveolar bone area and identify the main boundaries of the alveolar bone by calculating the curvature of the alveolar bone surface; Using the contact points between the teeth and the alveolar bone, fit the extension line of the alveolar bone, and select the intersection point of the extension line so that the intersection point is inside the alveolar bone; Use the direction vector of the extended line to calculate its bisection direction. The formula is: ； Wherein, vector L1 and vector L2 are the direction vectors of two extension lines passing through the contact point between the tooth and the alveolar bone; the direction vectors are bisected to determine a direction that takes both the tooth and the alveolar bone into consideration; the bisector is taken as the second tooth long axis, and its spatial coordinates are recorded; The calculating the angle between the first tooth long axis and the second tooth long axis and generating a new long axis includes: Starting from the direction vectors of the long axis of the two teeth, the vectors are obtained by fitting the long axis data and define the direction of the long axis in three-dimensional space; The cosine of the angle between two vectors is calculated using the dot product formula: ； Among them, a and b are two vectors, and is the magnitude of the vector, and θ is the angle between the vectors; The included angle θ is obtained by the arccosine function. This angle represents the spatial relationship between the two long axes and is an important parameter for evaluating tooth arrangement and alveolar bone structure. The bisector of the angle is used to generate a new major axis. The direction vector of the bisector can be obtained by taking the average of the normalized sum of the two original major axis direction vectors, i.e.: ； The new long axis is recorded and visualized to be consistent with the actual anatomical structure; wherein, during the visualization process, the position and direction of the new long axis are adjusted.

2. The method for measuring the buccal-lingual torque of maxillary molars based on the improved ResNET according to claim 1, characterized in that: The obtaining and preprocessing of three-dimensional images of teeth and alveolar bones, and three-dimensional reconstruction of the pre-processed images, include: Use cone beam computed tomography (CBCT) equipment to obtain three-dimensional images of teeth and alveolar bones; For 3D images, median filtering or Gaussian filtering techniques are applied to remove noise introduced during the scanning process, and adaptive histogram equalization (CLAHE) technology is used to enhance the contrast between teeth and alveolar bone, and to extract the region of interest (ROI). Open source tools are used to perform three-dimensional reconstruction on the pre-processed images to generate a visual model of the teeth and alveolar bone; the three-dimensional reconstruction includes image segmentation, feature extraction and three-dimensional modeling.

3. The method for measuring the buccal-lingual torque of maxillary molars based on the improved ResNET according to claim 1, characterized in that: Positioning the long axis of the first tooth based on the coupling relationship between the crown and the root includes: Region growing or threshold segmentation algorithms are used to separate the crown and root. Region growing algorithms start from a seed point and gradually add neighboring pixels similar to the seed point until the entire tooth area is covered. Threshold segmentation distinguishes the tooth from the background based on the range of pixel values. A histogram analysis is usually performed first to determine the optimal threshold. Use U-Net or its variants to automatically segment the tooth area; Extracting point cloud data from the segmented crown and root regions, wherein the point cloud data includes three-dimensional coordinate information of the tooth; Principal component analysis (PCA) was applied to calculate the principal axis directions of the crown and root. PCA projects the data into a new coordinate system through linear transformation so that the variance on the principal component of the first coordinate axis is maximized to determine the principal axis direction. The midlines of the crown and root are fitted using the least squares method. The best fitting line is found by minimizing the sum of squared errors, and the midlines of the crown and root are connected to form the first tooth long axis. Display the fitting results through a visual interface; If the long axis is found to be seriously deviated, the position of the connection point between the crown and the root is adjusted.

4. The method for measuring the buccal-lingual torque of maxillary molars based on the improved ResNET according to claim 1, characterized in that: The calculation and verification of the buccal-lingual torque value of the maxillary molars includes: Determine the angle between the new long axis and the alveolar bone plane, and calculate the angle by vector dot product. The dot product of the direction vector of the new long axis and the normal vector of the alveolar bone plane is used to calculate the cosine of the angle, and then the angle is obtained. Calculate the horizontal and vertical offsets between the crown top and the root bottom along the new long axis; Determine the precise position of the crown and root and measure the offset relative to the new long axis; The calculation formula of the buccal-lingual torque value of maxillary molars is expressed as: ； Where θ' is the angle between the new long axis and the alveolar bone plane, Δy and Δx are the horizontal and vertical offsets of the crown top and root bottom along the new long axis, respectively; Compare the calculated torque value with the normal anatomical reference value or clinical correction target value; If the measurement error is large, the CBCT image quality or the accuracy of the segmentation model should be checked. Image quality inspection includes evaluation of resolution, noise level, and contrast parameters, and segmentation model accuracy inspection involves retraining or adjusting model parameters.

5. A device for measuring buccal-lingual torque of maxillary molars based on improved ResNET, characterized in that: include: An image acquisition module is used to acquire and pre-process three-dimensional images of teeth and alveolar bones, and to perform three-dimensional reconstruction of the pre-processed images; A first positioning module is used to locate the long axis of the first tooth based on the coupling relationship between the crown and the root; A second positioning module is used to locate the long axis of the second tooth based on the coupling relationship between the tooth and the alveolar bone; An angle calculation module, used to calculate the angle between the long axis of the first tooth and the long axis of the second tooth and generate a new long axis; Torque calculation module, used to calculate and verify the buccal and lingual torque values ​​of maxillary molars; Positioning the second tooth long axis based on the coupling relationship between the tooth and the alveolar bone includes: Perform three-dimensional curvature analysis on the alveolar bone area and identify the main boundaries of the alveolar bone by calculating the curvature of the alveolar bone surface; Using the contact points between the teeth and the alveolar bone, fit the extension line of the alveolar bone, and select the intersection point of the extension line so that the intersection point is inside the alveolar bone; Use the direction vector of the extended line to calculate its bisection direction. The formula is: ； Wherein, vector L1 and vector L2 are the direction vectors of two extension lines passing through the contact point between the tooth and the alveolar bone; the direction vectors are bisected to determine a direction that takes both the tooth and the alveolar bone into consideration; the bisector is taken as the second tooth long axis, and its spatial coordinates are recorded; The calculating the angle between the first tooth long axis and the second tooth long axis and generating a new long axis includes: Starting from the direction vectors of the long axis of the two teeth, the vectors are obtained by fitting the long axis data and define the direction of the long axis in three-dimensional space; The cosine of the angle between two vectors is calculated using the dot product formula: ； Among them, a and b are two vectors, and is the magnitude of the vector, and θ is the angle between the vectors; The included angle θ is obtained by the arccosine function. This angle represents the spatial relationship between the two long axes and is an important parameter for evaluating tooth arrangement and alveolar bone structure. The bisector of the angle is used to generate a new major axis. The direction vector of the bisector can be obtained by taking the average of the normalized sum of the two original major axis direction vectors, i.e.: ； The new long axis is recorded and visualized to be consistent with the actual anatomical structure; wherein, during the visualization process, the position and direction of the new long axis are adjusted.

6. A computer device, characterized in that: include: memory and one or more processors; The memory is used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement a method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET as described in any one of claims 1-4.

7. A storage medium containing computer-executable instructions, characterized in that: When executed by a computer processor, the computer executable instructions are used to execute a method for measuring the buccal-lingual torque value of maxillary molars based on the improved ResNET as described in any one of claims 1 to 4.

8. A computer program product, characterized in that The computer program product comprises instructions which, when executed by a computer, cause the computer to implement the method according to any one of claims 1 to 4.

Images