A Bandit algorithm-based bone block selection method for alveolar bone defects

Abstract

The present invention discloses a method for selecting bone blocks for alveolar bone defects using a Bandit algorithm, including S1, obtaining three-dimensional image data of a patient's jaw; S2, determining multiple groups of candidate areas for safe bone removal; S3, analyzing the bone surface contour features of each candidate area; S4, applying the Bandit algorithm, wherein each candidate area is an arm, based on the consistency data evaluated in step S3, applying an attenuation factor to adjust the weight of the historical data of each candidate area, so that the Bandit algorithm is more sensitive to the latest clinical feedback results, and iterating the selection process; S5, using computer simulation in preclinical trials; S6, adjusting the Bandit algorithm parameters in step S4; S7, performing bone removal surgery in the selected best candidate area. The present invention ensures the best match between the selected bone block and the defect area in terms of morphology and biomechanical properties by analyzing the consistency of the bone surface contour features of each candidate area with the surface features of the tissue fitting surface of the bone block required to be implanted in the alveolar bone defect area.

Description

Technical Field

[0001] The invention relates to the technical field of dental analysis, and in particular to a method for selecting alveolar bone defect bone blocks using a Bandit algorithm. Background Art

[0002] In oral and maxillofacial surgery, repairing alveolar bone defects is a common and complex operation. In order to repair the defect area, bone blocks need to be taken from other areas of the patient's jaw. In existing technologies, the selection of bone harvesting areas usually depends on the doctor's experience and intuition, considering that there are many bone harvesting areas available in the jaw, such as the external oblique line of the mandibular ramus, chin, maxillary tuberosity, hard palate, etc. Each area has its own specific advantages and disadvantages, such as the ease of bone harvesting, the quality, density, morphology of the bone blocks, and the relative position to important anatomical structures.

[0003] In traditional surgical planning, although doctors can judge the potential value and risks of different bone harvesting areas through clinical experience, the lack of accurate data support often makes it difficult for doctors to accurately assess the best suitability of each area for specific cases. In addition, preoperative evaluation often cannot fully reveal the internal characteristics and potential risks of bone, such as proximity to tooth roots and nerves and blood vessels. These factors may affect the safety and effectiveness of the operation, increase the complexity of the operation and the patient's recovery time.

[0004] In addition, the prior art generally lacks a systematic evaluation of the consistency between the bone surface contour features and the surface features of the fitting surface of the implanted bone block when matching the bone harvesting area and the defect area. This inaccurate matching may lead to incompatibility between the implanted bone block and the defect area in terms of biomechanical properties, affecting the long-term success rate of the operation. Therefore, how to provide a method for selecting bone blocks for alveolar bone defects using a Bandit algorithm is an urgent problem that those skilled in the art need to solve. Summary of the invention

[0005] One purpose of the present invention is to propose a method for selecting bone blocks for alveolar bone defects based on a Bandit algorithm. The present invention ensures the best match between the selected bone block and the defect area in terms of morphology and biomechanical properties by analyzing the consistency between the bone surface contour characteristics of each candidate area and the surface characteristics of the tissue fitting surface of the bone block required to be implanted in the alveolar bone defect area.

[0006] A method for selecting bone blocks for alveolar bone defects using a Bandit algorithm according to an embodiment of the present invention comprises the following steps:

[0007] S1. Obtaining three-dimensional image data of the patient's jaw through CT imaging technology;

[0008] S2. In the three-dimensional image data, identify the potential dangerous bone harvesting areas adjacent to the tooth root and neurovascular anatomical structures, exclude them, and determine multiple groups of candidate areas for safe bone harvesting;

[0009] S3. Analyze the bone surface contour features of each candidate area to evaluate its consistency with the surface features of the tissue fitting surface of the bone block to be implanted in the alveolar bone defect area of ​​the patient;

[0010] S4, applying the Bandit algorithm, where each candidate region is an arm, and based on the consistency data evaluated in step S3, applying an attenuation factor to adjust the weight of the historical data of each candidate region, so that the Bandit algorithm is more sensitive to the latest clinical feedback results, and iterating the selection process;

[0011] S5. Use computer simulations in preclinical trials to simulate defects of different bone qualities, shapes, and sizes to test the performance of each candidate region in the intended clinical application;

[0012] S6, adjusting the Bandit algorithm parameters in step S4 according to the computer simulation results in step S5, and selecting the best candidate region of the bone block that best suits the specific alveolar bone defect conditions of the patient;

[0013] S7. Perform bone harvesting surgery in the selected best candidate area to extract bone blocks for repairing alveolar bone defects.

[0014] Optionally, the S1 includes the following steps:

[0015] S11. Use CT scanning equipment to position and fix the patient's head;

[0016] S12. Adjust the parameter settings of the CT scanning equipment, including the scanning layer thickness, voltage and current, corresponding to the jaw bone density and structure of different patients:

[0017]

[0018] Among them, V opt and I opt They are the optimized voltage and current settings, kV and mA are the default voltage and current parameters of the device, and D bone is the estimated jaw bone density, D0 and T0 are the standard jaw bone density and standard exposure time, t exp It is the exposure time adjusted according to the patient’s specific condition;

[0019] S13, starting CT scanning to acquire continuous cross-sectional images of the jaw, wherein each cross-sectional image is acquired at a preset interval;

[0020] S14. Using image reconstruction software, the collected cross-sectional images are synthesized into a complete three-dimensional jaw model, which reflects the anatomical structure and biological markers of the jaw:

[0021]

[0022] Among them, M(x, y, z) is the density value of a point on the 3D jaw model, α i is the weight coefficient corresponding to the i-th cross-sectional image, f i (x, y, z) is the local model function reconstructed based on the i-th cross-sectional image, and n is the number of cross-sectional images;

[0023] S15. Store the generated three-dimensional jaw model in a medical image database and perform standardization processing.

[0024] Optionally, S2 includes the following steps:

[0025] S21. Using 3D image processing software, the areas adjacent to the tooth roots and neurovascular anatomical structures are automatically identified in the 3D jaw model. The areas are determined by analyzing the density threshold of specific markers in the image data, and the dangerous areas are identified based on the similarity of local image features and anatomical structures:

[0026]

[0027] Wherein, τ(x, y, z) is the density threshold at the coordinate (x, y, z), δ is an adjustment factor used to quantitatively analyze the sensitivity of the adjacent area, D(x, y, z) is the jaw density at the coordinate (x, y, z), D0 is the preset reference value of jaw density, k is a gain constant used to adjust the sensitivity of the density threshold function, and θ is a minimum threshold;

[0028] S22. In the three-dimensional jaw model, based on the threshold value defined in step S21, the potential dangerous areas adjacent to the key anatomical structures are excluded by a spatial filter:

[0029]

[0030] Where R is the area of ​​the entire jaw, λ is the threshold used to determine whether to exclude, and B((x, y, z), r) is a spherical area with the center at (x, y, z) and a radius of r;

[0031] S23. Screen the remaining areas to identify candidate areas for bone harvesting. The screening is based on the quality and volume standards of the jawbone and the specific needs of the alveolar bone defect area:

[0032]

[0033] Where χ is a characteristic function, when ρ(x, y, z) is greater than the minimum density ρmi n The value is 1 when , otherwise it is 0, p(x, y, z) is the jaw bone density at coordinates (x, y, z);

[0034] S24, adding the selected candidate areas into the database.

[0035] Optionally, the S24 includes the following steps:

[0036] A i represents the i-th candidate region, S(A i ) represents the priority score of the candidate region, reflecting the suitability and safety of the region for the treatment of alveolar bone defects. The formula is as follows:

[0037]

[0038] Among them, τ safe is a safety threshold used to assess the safety boundary of the region from important anatomical structures. safe is a parameter that defines the safe bone harvesting distance, and κ is a gain constant used to adjust the sensitivity of surgical risk.

[0039] Optionally, S3 includes the following steps:

[0040] S31. For each candidate area selected, three-dimensional image processing technology is used to extract bone surface contour features, including contour line shape, surface roughness and density distribution, in combination with the importance of density gradient and curvature to surface features:

[0041]

[0042] Among them, C(v) represents the comprehensive measure of the curvature of the bone surface at point v, which is used to evaluate the geometric characteristics of the bone surface, θV represents the boundary surface of the candidate area, that is, the outer contour of the bone surface, Represents the density gradient at point v, indicating the direction and intensity of the density change at that point. represents the second-order derivative of the density at point v, reflecting the curvature and local shape of the density change, and n(v) represents the external normal vector at point v, representing the direction of the surface at that point;

[0043] S32. Use similarity measurement in high-dimensional space to analyze the bone surface contour features, match them with the surface features of the tissue fitting surface of the bone block to be implanted in the alveolar bone defect area stored in the database, and evaluate the consistency:

[0044]

[0045] Among them, Sim(V, W) represents the similarity measure between the candidate region v and the fitting surface W of the bone block required for implantation in the alveolar bone defect area, σ represents the parameter for adjusting the similarity sensitivity, which determines the sensitivity of the similarity function to the density gradient difference, and K(v, w) represents the kernel function, which is used to quantify the similarity between two points v and w;

[0046] S33. According to the analysis result of step S32, the degree of fit between each candidate region and the alveolar bone defect region is calculated:

[0047]

[0048] Among them, Fit(V) represents the overall fitness score of the entire candidate region v, which is used to select the bone block configuration that best suits the patient's needs, ω i represents the weight coefficient of each scoring component, α i represents the attenuation constant, which is used to adjust the influence of density gradient differences within each candidate region, λ; represents the weight of the similarity score in the overall fitness, represents the minimum density gradient threshold of the jaw, which is used to determine the minimum standard of bone quality, A; represents the i-th candidate area, a potential safe bone harvesting area screened from the entire jaw.

[0049] Optionally, S4 includes the following steps:

[0050] S41. Define each screened candidate region as an arm in the Bandit algorithm, and combine the evaluated consistency data with the historical selection data to initialize the performance evaluation of each arm:

[0051]

[0052] Among them, Cons(a) is the fit score between the candidate area a and the alveolar bone defect area, R a，i is the historical reward for selecting candidate region a, and N is the number of historical selections;

[0053] S42. Apply the upper confidence limit algorithm to select the best candidate region for each iteration. The calculation formula is as follows:

[0054]

[0055] Among them, a t is the arm selected at iteration number t, is the estimated mean return of arm a, n a is the number of times arm a is selected;

[0056] S43, introduce the attenuation factor γ to adjust the weight of the historical data of each candidate region, so that the algorithm is more sensitive to the latest clinical feedback results, and calculate the adjusted average return:

[0057]

[0058] in, is an adjustment function based on the latest return r new and the number of selections n a A nonlinear function that adjusts the rate at which new information is added:

[0059] τ(r new , n a )=r new ·(1+α·ln(1+n a )).

[0060] Optionally, S5 includes the following steps:

[0061] S51. Configure a computer simulation environment to simulate defects of different bone qualities, shapes and sizes using density and mechanical parameters;

[0062] S52, construct multiple defect models M (d, s, q) for testing, the model parameters include depth d, shape s and quality q, the parameters are passed through the function f d， s(x, y, z) describes the geometric characteristics of the defect:

[0063]

[0064] Among them, f d，s is a function describing the defect shape, f ref is the standard shape function of reference healthy bone, S is the surface area considered, d, s, and q represent the depth, shape, and quality parameters, respectively;

[0065] S53. Use the above-mentioned defect model to perform mechanical simulation and evaluate the stress distribution σ(x, y, z) and displacement u(x, y, z) of each candidate area under the simulation environment:

[0066]

[0067] Among them, σ0 is the reference stress value, ||xx′|| represents the distance between points in space;

[0068] S54. Evaluate the performance of each candidate area based on the simulation test results, simulating durability and safety under chewing force.

[0069] Optionally, the S6 includes the following steps:

[0070] S61, collecting computer simulation results, including stress distribution and displacement data of each candidate region under various simulated loading conditions;

[0071] S62. Evaluate the performance of each candidate region based on the simulation results to determine the potential success rate in actual clinical applications. The evaluation formula is:

[0072] P(a)=ω p ·exp(-∫ M(d，s，q) ||σ(x, y, z)-σ ideal ||dV);

[0073] Where P(a) is the performance index of candidate region a, ω p is the weight parameter of the performance index, σ(x, y, z) is the simulated stress distribution, and σ ideal is the ideal stress distribution;

[0074] S63. Update the parameters of the Bandit algorithm according to the fit score and the performance index to optimize the selection probability of the arm. The update formula is as follows:

[0075]

[0076] S64. Apply the updated Bandit algorithm parameters in the subsequent selection process to optimize and determine the candidate area that best suits the patient's specific alveolar bone defect conditions.

[0077] The beneficial effects of the present invention are:

[0078] (1) The present invention utilizes the detailed three-dimensional images obtained by CT imaging technology and the decision framework of the Bandit algorithm, so that doctors can accurately select the most suitable area from multiple potential bone harvesting areas, reducing the uncertainty of relying on the doctor's subjective experience. By analyzing the consistency of the bone surface contour characteristics of each candidate area with the surface characteristics of the tissue fitting surface of the bone block required to be implanted in the alveolar bone defect area, the optimal match between the selected bone block and the defect area in terms of morphology and biomechanical properties is ensured.

[0079] (2) The present invention significantly improves the safety of surgery by automatically identifying and eliminating potentially dangerous bone harvesting areas adjacent to important anatomical structures such as tooth roots and nerves and blood vessels. This not only reduces the risk of possible injury during surgery, but also reduces postoperative pain and recovery time.

[0080] (3) The present invention uses computer simulation to test defects of different bone qualities, shapes and sizes in preclinical trials, and dynamically adjusts the parameters of the Bandit algorithm based on the simulation results, so that doctors can foresee and optimize surgical plans. By considering the specific alveolar bone defect conditions and bone characteristics of each patient, different treatment plans can be supported. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0082] Figure 1 The present invention provides a flow chart of a method for selecting bone blocks for alveolar bone defects using a Bandit algorithm. DETAILED DESCRIPTION

[0083] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0084] refer to Figure 1 A method for selecting bone blocks for alveolar bone defects based on a Bandit algorithm comprises the following steps:

[0085] S1. Obtaining three-dimensional image data of the patient's jaw through CT imaging technology;

[0086] In this implementation, S1 includes the following steps:

[0087] S11. Use CT scanning equipment to position and fix the patient's head;

[0088] S12. Adjust the parameter settings of the CT scanning equipment, including the scanning layer thickness, voltage and current, corresponding to the jaw bone density and structure of different patients:

[0089]

[0090] Among them, V opt and I opt They are the optimized voltage and current settings, kV and mA are the default voltage and current parameters of the device, and D bone is the estimated jaw bone density, D0 and T0 are the standard jaw bone density and standard exposure time, t exp It is the exposure time adjusted according to the patient’s specific condition;

[0091] S13, starting CT scanning to acquire continuous cross-sectional images of the jaw, wherein each cross-sectional image is acquired at a preset interval;

[0092] S14. Using image reconstruction software, the collected cross-sectional images are synthesized into a complete three-dimensional jaw model, which reflects the anatomical structure and biological markers of the jaw:

[0093]

[0094] Among them, M(x, y, z) is the density value of a point on the 3D jaw model, α i is the weight coefficient corresponding to the i-th cross-sectional image, f i (x, y, z) is the local model function reconstructed based on the i-th cross-sectional image, and n is the number of cross-sectional images;

[0095] S15. Store the generated three-dimensional jaw model in a medical image database and perform standardization processing.

[0096] This embodiment uses CT imaging technology to accurately obtain three-dimensional image data of the patient's jaw, which can fully display the structural details of the jaw, including bone density, bone distribution and the location of key anatomical structures, providing basic information for bone block selection.

[0097] S2. In the three-dimensional image data, identify the potential dangerous bone harvesting areas adjacent to the tooth root and neurovascular anatomical structures, exclude them, and determine multiple groups of candidate areas for safe bone harvesting;

[0098] In this implementation, S2 includes the following steps:

[0099] S21. Using 3D image processing software, the areas adjacent to the tooth roots and neurovascular anatomical structures are automatically identified in the 3D jaw model. The areas are determined by analyzing the density threshold of specific markers in the image data, and the dangerous areas are identified based on the similarity of local image features and anatomical structures:

[0100]

[0101] Wherein, τ(x, y, z) is the density threshold at the coordinate (x, y, z), δ is an adjustment factor used to quantitatively analyze the sensitivity of the adjacent area, D(x, y, z) is the jaw density at the coordinate (x, y, z), D0 is the preset reference value of jaw density, k is a gain constant used to adjust the sensitivity of the density threshold function, and θ is a minimum threshold;

[0102] S22. In the three-dimensional jaw model, based on the threshold value defined in step S21, the potential dangerous areas adjacent to the key anatomical structures are excluded by a spatial filter:

[0103]

[0104] Where R is the area of ​​the entire jaw, λ is the threshold used to determine whether to exclude, and B((x, y, z), r) is a spherical area with the center at (x, y, z) and a radius of r;

[0105] S23. Screen the remaining areas to identify candidate areas for bone harvesting. The screening is based on the quality and volume standards of the jawbone and the specific needs of the alveolar bone defect area:

[0106]

[0107] Where χ is a characteristic function, when ρ(x, y, z) is greater than the minimum density ρmi n The value is 1 when , otherwise it is 0, p(x, y, z) is the jaw bone density at coordinates (x, y, z);

[0108] S24, adding the selected candidate areas into the database.

[0109] In this implementation, S24 includes the following steps:

[0110] A i represents the i-th candidate region, S(A i ) represents the priority score of the candidate region, reflecting the suitability and safety of the region for the treatment of alveolar bone defects. The formula is as follows:

[0111]

[0112] Among them, τ safe is a safety threshold used to assess the safety boundary of the region from important anatomical structures. safe is a parameter that defines the safe bone harvesting distance, and κ is a gain constant used to adjust the sensitivity of surgical risk.

[0113] This embodiment uses image processing software to identify dangerous areas adjacent to important anatomical structures such as tooth roots, nerves and blood vessels through an automated process, effectively avoiding possible damage to these important structures during surgery, improving the safety of the surgery and reducing the risk of postoperative complications.

[0114] S3. Analyze the bone surface contour features of each candidate area to evaluate its consistency with the surface features of the tissue fitting surface of the bone block to be implanted in the alveolar bone defect area of ​​the patient;

[0115] In this implementation, S3 includes the following steps:

[0116] S31. For each candidate area selected, three-dimensional image processing technology is used to extract bone surface contour features, including contour line shape, surface roughness and density distribution, in combination with the importance of density gradient and curvature to surface features:

[0117]

[0118] Where C(v) represents the comprehensive measure of the curvature of the bone surface at point v, which is used to evaluate the geometric characteristics of the bone surface. represents the boundary surface of the candidate region, i.e., the outer contour of the bone surface, Represents the density gradient at point v, indicating the direction and intensity of the density change at that point. represents the second-order derivative of the density at point v, reflecting the curvature and local shape of the density change, and n(v) represents the external normal vector at point v, representing the direction of the surface at that point;

[0119] S32. Use similarity measurement in high-dimensional space to analyze the bone surface contour features, match them with the surface features of the tissue fitting surface of the bone block to be implanted in the alveolar bone defect area stored in the database, and evaluate the consistency:

[0120]

[0121] Among them, Sim(V, W) represents the similarity measure between the candidate region v and the fitting surface W of the bone block required for implantation in the alveolar bone defect area, σ represents the parameter for adjusting the similarity sensitivity, which determines the sensitivity of the similarity function to the density gradient difference, and K(v, w) represents the kernel function, which is used to quantify the similarity between two points v and w;

[0122] S33. According to the analysis result of step S32, the degree of fit between each candidate region and the alveolar bone defect region is calculated:

[0123]

[0124] Among them, Fit(V) represents the overall fitness score of the entire candidate region v, which is used to select the bone block configuration that best suits the patient's needs, ω i represents the weight coefficient of each scoring component, α i represents the attenuation constant, which is used to adjust the influence of density gradient differences within each candidate region, λ; represents the weight of the similarity score in the overall fitness, represents the minimum density gradient threshold of the jaw, which is used to determine the minimum standard of bone quality, A; represents the i-th candidate area, a potential safe bone harvesting area screened from the entire jaw.

[0125] S4, applying the Bandit algorithm, where each candidate region is an arm, and based on the consistency data evaluated in step S3, applying an attenuation factor to adjust the weight of the historical data of each candidate region, so that the Bandit algorithm is more sensitive to the latest clinical feedback results, and iterating the selection process;

[0126] This embodiment selects the best bone harvesting site by carefully analyzing the bone surface contour features of the candidate area and comparing these features with the fitting surface of the bone block required to be implanted in the defect area, ensuring that the implanted bone block is highly consistent with the defect area in morphology and function, thereby improving the repair success rate and structural stability after surgery.

[0127] In this implementation, S4 includes the following steps:

[0128] S41. Define each screened candidate region as an arm in the Bandit algorithm, and combine the evaluated consistency data with the historical selection data to initialize the performance evaluation of each arm:

[0129]

[0130] Among them, Cons(a) is the fit score between the candidate area a and the alveolar bone defect area, R a，i is the historical reward for selecting candidate region a, and N is the number of historical selections;

[0131] S42. Apply the upper confidence limit algorithm to select the best candidate region for each iteration. The calculation formula is as follows:

[0132]

[0133] Among them, a t is the arm selected at iteration number t, is the estimated mean return of arm a, n a is the number of times arm a is selected;

[0134] S43, introduce the attenuation factor γ to adjust the weight of the historical data of each candidate region, so that the algorithm is more sensitive to the latest clinical feedback results, and calculate the adjusted average return:

[0135]

[0136] in, is an adjustment function based on the latest return r new and the number of selections n a A nonlinear function that adjusts the rate at which new information is added:

[0137] τ(r new , n a )=r new ·(1+α·ln(1+n a )).

[0138] S5. Use computer simulations in preclinical trials to simulate defects of different bone qualities, shapes, and sizes to test the performance of each candidate region in the intended clinical application;

[0139] In this implementation, S5 includes the following steps:

[0140] S51. Configure a computer simulation environment to simulate defects of different bone qualities, shapes and sizes using density and mechanical parameters;

[0141] S52, construct multiple defect models M (d, s, q) for testing, the model parameters include depth d, shape s and quality q, the parameters are passed through the function f d，s ( x , y, z) describe the geometric characteristics of the defect:

[0142]

[0143] Among them, f d，s is a function describing the defect shape, f ref is the standard shape function of reference healthy bone, S is the surface area considered, d, s, and q represent the depth, shape, and quality parameters, respectively;

[0144] S53. Use the above-mentioned defect model to perform mechanical simulation and evaluate the stress distribution σ(x, y, z) and displacement u(x, y, z) of each candidate area under the simulation environment:

[0145]

[0146] Among them, σ0 is the reference stress value, ||xx′|| represents the distance between points in space;

[0147] S54. Evaluate the performance of each candidate area based on the simulation test results, simulating durability and safety under chewing force.

[0148] This embodiment can predict the performance of each candidate area under actual load conditions by testing defects of different bone qualities, shapes and sizes through computer simulation before clinical practice, helping doctors understand the mechanical performance of each candidate area, optimize bone block selection and surgical planning, and reduce risks in actual surgery.

[0149] S6, adjusting the Bandit algorithm parameters in step S4 according to the computer simulation results in step S5, and selecting the best candidate region of the bone block that best suits the specific alveolar bone defect conditions of the patient;

[0150] In this implementation, S6 includes the following steps:

[0151] S61, collecting computer simulation results, including stress distribution and displacement data of each candidate region under various simulated loading conditions;

[0152] S62. Evaluate the performance of each candidate region based on the simulation results to determine the potential success rate in actual clinical applications. The evaluation formula is:

[0153] P(a)=ω p ·exp(-∫ M(d，s，q) ||σ(x, y, z)-σ ideal ||dV);

[0154] Where P(a) is the performance index of candidate region a, ω p is the weight parameter of the performance index, σ(x, y, z) is the simulated stress distribution, and σ ideal is the ideal stress distribution;

[0155] S63. Update the parameters of the Bandit algorithm according to the fit score and the performance index to optimize the selection probability of the arm. The update formula is as follows:

[0156]

[0157] S64. Apply the updated Bandit algorithm parameters in the subsequent selection process to optimize and determine the candidate area that best suits the patient's specific alveolar bone defect conditions.

[0158] S7. Perform bone harvesting surgery in the selected best candidate area to extract bone blocks for repairing alveolar bone defects.

[0159] The present invention utilizes the detailed three-dimensional images obtained by CT imaging technology and the decision framework of the Bandit algorithm, so that doctors can accurately select the most suitable area from multiple potential bone harvesting areas, reducing the uncertainty of relying on the doctor's subjective experience. By analyzing the consistency of the bone surface contour characteristics of each candidate area with the surface characteristics of the tissue fitting surface of the bone block required to be implanted in the alveolar bone defect area, the optimal match between the selected bone block and the defect area in terms of morphology and biomechanical properties is ensured.

[0160] The present invention significantly improves the safety of surgery by automatically identifying and excluding potentially dangerous bone harvesting areas adjacent to important anatomical structures such as tooth roots and nerves and blood vessels. This not only reduces the risk of possible injury during surgery, but also reduces postoperative pain and recovery time.

[0161] The present invention uses computer simulation to test defects of different bone qualities, shapes and sizes in preclinical trials, and dynamically adjusts the parameters of the Bandit algorithm based on the simulation results, so that doctors can foresee and optimize surgical plans. Different treatment plans are supported by considering each patient's specific alveolar bone defect conditions and bone quality characteristics.

[0162] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for selecting alveolar bone defect blocks based on Bandit algorithm, characterized in that: The method comprises the following steps: S1. Obtaining three-dimensional image data of the patient's jaw through CT imaging technology; S2. In the three-dimensional image data, identify the potential dangerous bone harvesting areas adjacent to the tooth root and neurovascular anatomical structures, exclude them, and determine multiple groups of candidate areas for safe bone harvesting; S3. Analyze the bone surface contour features of each candidate area to evaluate its consistency with the surface features of the tissue fitting surface of the bone block to be implanted in the alveolar bone defect area of ​​the patient; S4, applying the Bandit algorithm, where each candidate region is an arm, and based on the consistency data evaluated in step S3, applying an attenuation factor to adjust the weight of the historical data of each candidate region, so that the Bandit algorithm is more sensitive to the latest clinical feedback results, and iterating the selection process; S5. Use computer simulations in preclinical trials to simulate defects of different bone qualities, shapes, and sizes to test the performance of each candidate region in the intended clinical application; S6. Adjust the Bandit algorithm parameters in step S4 according to the computer simulation results in step S5, and select the best candidate area suitable for the patient's specific alveolar bone defect conditions.

2. The method for selecting alveolar bone defect blocks based on Bandit algorithm according to claim 1, characterized in that: The S1 comprises the following steps: S11. Use CT scanning equipment to position and fix the patient's head; S12. Adjust the parameter settings of the CT scanning equipment, including the scanning layer thickness, voltage and current, corresponding to the jaw bone density and structure of different patients: Among them, V opt and I opt They are the optimized voltage and current settings, kV and mA are the default voltage and current parameters of the device, and D bone is the estimated jaw bone density, D0 and T0 are the standard jaw bone density and standard exposure time, and t exp It is the exposure time adjusted according to the patient’s specific condition; S13, starting CT scanning to acquire continuous cross-sectional images of the jaw, wherein each cross-sectional image is acquired at a preset interval; S14. Using image reconstruction software, the collected cross-sectional images are synthesized into a complete three-dimensional jaw model, which reflects the anatomical structure and biological markers of the jaw: Among them, M(x,y,z) is the density value of a point on the three-dimensional jaw model, α i is the weight coefficient corresponding to the i-th cross-sectional image, f i (x, y, z) is the local model function reconstructed based on the i-th cross-sectional image, and n is the number of cross-sectional images; S15. Store the generated three-dimensional jaw model in a medical image database and perform standardization processing.

3. The method for selecting alveolar bone defect blocks based on Bandit algorithm according to claim 2, characterized in that: The S2 comprises the following steps: S21. Using 3D image processing software, the areas adjacent to the tooth roots and neurovascular anatomical structures are automatically identified in the 3D jaw model. The areas are determined by analyzing the density threshold of specific markers in the image data, and the dangerous areas are identified based on the similarity of local image features and anatomical structures: Where τ(x,y,z) is the density threshold at the coordinate (x,y,z), δ is an adjustment factor used to quantitatively analyze the sensitivity of the adjacent area, D(x,y,z) is the jaw density at the coordinate (x,y,z), D0 is the preset reference value of jaw density, k is a gain constant used to adjust the sensitivity of the density threshold function, and θ is a minimum threshold; S22. In the three-dimensional jaw model, based on the threshold value defined in step S21, the potential dangerous areas adjacent to the key anatomical structures are excluded by a spatial filter: Where R is the area of ​​the entire jaw, λ is the threshold used to determine whether to exclude, and B((x, y, z), r) is a spherical area with the center at (x, y, z) and a radius of r; S23. Screen the remaining areas to identify candidate areas for bone harvesting. The screening is based on the quality and volume standards of the jawbone and the specific needs of the alveolar bone defect area: Among them, χ is a characteristic function, when ρ(x,y,z) is greater than the minimum density ρ min The value is 1 when , otherwise it is 0. ρ(x,y,z) is the jaw bone density at coordinate (x,y,z); S24, adding the selected candidate areas into the database.

4. The method for selecting alveolar bone defect blocks based on Bandit algorithm according to claim 3, characterized in that: The S24 comprises the following steps: A i represents the i-th candidate region, S(A i ) represents the priority score of the candidate region, reflecting the suitability and safety of the region for the treatment of alveolar bone defects. The formula is as follows: Among them, τ safe is a safety threshold used to assess the safety boundary of the region from important anatomical structures. safe is a parameter that defines the safe bone harvesting distance, and κ is a gain constant used to adjust the sensitivity of surgical risk.

5. The method for selecting alveolar bone defect blocks based on Bandit algorithm according to claim 4, characterized in that: The S3 comprises the following steps: S31. For each candidate area selected, three-dimensional image processing technology is used to extract bone surface contour features, including contour line shape, surface roughness and density distribution, in combination with the importance of density gradient and curvature to surface features: Where C(v) represents the comprehensive measure of the curvature of the bone surface at point v, which is used to evaluate the geometric characteristics of the bone surface. represents the boundary surface of the candidate region, i.e., the outer contour of the bone surface, Represents the density gradient of point v in the candidate region, indicating the direction and intensity of the density change at the point. represents the second-order derivative of the density at point v in the candidate region, reflecting the curvature and local shape of the density change, and n(v) represents the external normal vector at point v in the candidate region, representing the direction of the surface of the point; S32. Use similarity measurement in high-dimensional space to analyze the bone surface contour features, match them with the surface features of the tissue fitting surface of the bone block to be implanted in the alveolar bone defect area stored in the database, and evaluate the consistency: Among them, Sim(V,W) represents the similarity measure between the candidate region V and the fitting surface W of the bone block required for implantation in the alveolar bone defect area, σ represents the parameter for adjusting the similarity sensitivity, which determines the sensitivity of the similarity function to the density gradient difference, and K(v,w) represents the kernel function, which is used to quantify the similarity between two candidate region points v and candidate region points w; S33. According to the analysis result of step S32, the degree of fit between each candidate region and the alveolar bone defect region is calculated: Among them, Fit(V) represents the overall fitness score of the entire candidate region point v, which is used to select the bone block configuration that best suits the patient's needs, ω1 i represents the weight coefficient of each scoring component, α i represents the attenuation constant, which is used to adjust the influence of density gradient differences within each candidate region, λ i Represents the weight of the similarity score in the overall fitness, A represents the minimum density gradient threshold of jaw bone, which is used to determine the minimum standard of bone quality. i represents the i-th candidate region, which is a potential safe bone harvesting area screened from the entire mandible.

6. The method for selecting alveolar bone defect blocks based on Bandit algorithm according to claim 5, characterized in that: The S4 comprises the following steps: S41. Define each screened candidate region as an arm in the Bandit algorithm, and combine the evaluated consistency data with the historical selection data to initialize the performance evaluation of each arm: Among them, R a,i is the historical reward for selecting candidate region a, and N is the number of historical selections; S42. Apply the upper confidence limit algorithm to select the best candidate region for each iteration. The calculation formula is as follows: Among them, a t is the arm selected at iteration number t, is the estimated mean return of arm a, n a is the number of times arm a is selected; S43, introduce the attenuation factor γ to adjust the weight of the historical data of each candidate region, so that the algorithm is more sensitive to the latest clinical feedback results, and calculate the adjusted average return: in, is an adjustment function based on the latest return r new and the number of selections n a A nonlinear function that adjusts the rate at which new information is added: τ(r new ,n a )=r new ·(1+α·ln(1+n a ))。 7. The method for selecting alveolar bone defect blocks based on Bandit algorithm according to claim 6, characterized in that: The S5 comprises the following steps: S51. Configure a computer simulation environment to simulate defects of different bone qualities, shapes and sizes using density and mechanical parameters; S52. Construct various defect models M(d,s,q) for testing. The model parameters include depth d, shape s and quality q. The parameters are calculated by function f d,s (x,y,z) describes the geometric properties of the defect: Among them, f d,s is a function describing the defect shape, f ref is the standard shape function of reference healthy bone, S is the surface area considered, d, s, and q represent the depth, shape, and quality parameters, respectively; S53. Use the above defect model to perform mechanical simulation to evaluate the stress distribution σ(x, y, z) and displacement u(x, y, z) of each candidate area under the simulation environment: Among them, σ0 is the reference stress value, ||xx′|| represents the distance between points in space; S54. Evaluate the performance of each candidate area based on the simulation test results, simulating durability and safety under chewing force.

8. The method for selecting bone blocks for alveolar bone defects based on Bandit algorithm according to claim 7, characterized in that: The S6 comprises the following steps: S61, collecting computer simulation results, including stress distribution and displacement data of each candidate region under various simulated loading conditions; S62. Evaluate the performance of each candidate region based on the simulation results to determine the potential success rate in actual clinical applications. The evaluation formula is: P(a)=ω p ·exp(-∫ M(d,s,q) ||σ(x,y,z)-σ ideal ||dV); Where P(a) is the performance index of candidate region a, ω p is the weight parameter of the performance index, σ(x,y,z) is the simulated stress distribution, and σ ideal is the ideal stress distribution; S63. Update the parameters of the Bandit algorithm according to the fit score and the performance index to optimize the selection probability of the arm. The update formula is as follows: S64. Apply the updated Bandit algorithm parameters in the subsequent selection process to optimize and determine the candidate area that best suits the patient's specific alveolar bone defect conditions.

Images