Accurate generation method of single dental implant based on enhanced game learning

Abstract

The invention discloses a method for accurately generating a single dental implant based on enhanced game learning. The method comprises the following steps: S1, generating a three-dimensional digital model of the oral cavity of a patient; s2, constructing a multi-objective optimization environment required by implant design based on the three-dimensional digital model of the patient; s3, in a multi-objective optimization environment of implant design, utilizing an enhanced game learning algorithm to simulate an interaction relationship between the implant and peripheral tissues; s4, dynamically adjusting the weight of the reward function according to the digital model of the patient; s5, generating a single dental implant three-dimensional model matched with the bone tissue and soft tissue characteristics of the patient in the three-dimensional digital model by utilizing the optimized implant parameter combination; and S6, performing mechanical simulation analysis on the generated three-dimensional model of the single dental implant. The invention ensures that the implant can simultaneously meet the requirements of biomechanical stability and aesthetics after being implanted.

Description

Technical Field

[0001] The present invention relates to the field of dental technology, and particularly to a precise generation method of a single - tooth implant based on enhanced game learning. Background Art

[0002] With the rapid development of oral implant technology, single - tooth implant restoration has gradually become an important means for missing - tooth restoration. Its core lies in the precise design and personalized matching of implants. However, there are still many technical bottlenecks in existing implant design methods, mainly reflected in the fact that the shape, size, and mechanical properties of implants are difficult to precisely match the individual anatomical structures of patients, thus affecting the long - term stability and biomechanical adaptability of implants.

[0003] Currently, the design of implants mainly relies on doctors' experience and a standardized implant library. Doctors usually analyze bone mass, bone density, and soft - tissue characteristics manually based on the patient's CBCT images, and then select the closest implant model from the existing implant library for implantation. This method has the following defects in practical applications:

[0004] Firstly, the adaptability of standardized implants is limited. Due to the large individual differences in the bone structure, soft - tissue thickness, and biomechanical properties of each patient, standardized implants cannot fully meet the needs of all patients, resulting in problems such as bone resorption, insufficient initial stability, or soft - tissue mismatch after implantation.

[0005] Secondly, traditional implant design methods lack the ability of intelligent optimization. During the design process, it is difficult for doctors to precisely quantify and optimize the multi - objective performance of implants, and they can only rely on personal experience for adjustment, which leads to the uncertainty of implant design and increases the risk of implant failure.

[0006] In addition, existing biomechanical simulation analysis techniques are mainly used for postoperative evaluation rather than implant design optimization. Existing methods generally use finite - element analysis to evaluate the stress distribution and stability of implanted implants, but they cannot dynamically optimize implant parameters during the design stage, resulting in poor biomechanical adaptability after implantation and affecting the repair effect.

[0007] In summary, there are significant deficiencies in the existing technology for precise implant design, mainly reflected in low personalized adaptability, lack of intelligent optimization means, and lagging biomechanical evaluation. Therefore, there is an urgent need for a new technical method to improve the personalized matching ability of implants and achieve multi - objective optimization during the design stage, thereby improving the biomechanical stability and repair effect of implants. Summary of the Invention

[0008] An object of the present invention is to propose a precise generation method for a single tooth implant based on enhanced game learning, and the present invention ensures that the implant can simultaneously meet the biomechanical stability and aesthetic requirements after implantation.

[0009] A precise generation method for a single tooth implant based on enhanced game learning according to an embodiment of the present invention includes the following steps:

[0010] S1. Collect the oral CBCT image data of the patient, and preprocess the oral CBCT image data to generate a three-dimensional digital model of the patient's oral cavity;

[0011] S2. Construct a multi-objective optimization environment required for implant design based on the three-dimensional digital model of the patient;

[0012] S3. In the multi-objective optimization environment for implant design, use the enhanced game learning algorithm to simulate the interaction relationship between the implant and the surrounding tissues, and the enhanced game learning algorithm includes a reinforcement learning module and a population game strategy module;

[0013] S4. Iteratively optimize the implant parameter combinations output by the population game strategy module, and gradually converge the implant parameter configuration to the parameter combination that meets the multi-objective optimization requirements through multiple rounds of game interaction and reinforcement learning update. At the same time, dynamically adjust the weight of the reward function according to the digital model of the patient;

[0014] S5. Use the optimized implant parameter combination to generate a three-dimensional model of a single tooth implant in the three-dimensional digital model that is adapted to the bone tissue and soft tissue characteristics of the patient;

[0015] S6. Perform mechanical simulation analysis on the generated three-dimensional model of the single tooth implant, simulate the stress distribution of the implant under the action of biting force and its stress transfer effect with the surrounding bone tissue, and at the same time evaluate the biomechanical stability and aesthetic layout of the implant. If the simulation analysis does not meet the preset threshold conditions, feedback the analysis results to the enhanced game learning algorithm for parameter adjustment and re-optimization.

[0016] Optionally, the S1 step specifically includes:

[0017] S11. Collect the cone beam computed tomography image data of the patient's oral cavity, and the cone beam computed tomography image data is composed of multiple consecutive slices I i (x, y), where I i (x, y) represents the gray value of the i-th image slice at the spatial coordinates (x, y);

[0018] S12. Perform noise reduction processing on the cone beam computed tomography image data;

[0019] S13. Perform tissue segmentation on the denoised cone beam computed tomography (CBCT) image data. Using a combined method based on threshold segmentation and region growing algorithm, segment the image data into a bone tissue region B(x, y, z) and a soft tissue region S(x, y, z), where:

[0020]

[0021] where H(x, y, z) is the Hounsfield unit of the CBCT image data at the three-dimensional coordinates (x, y, z), and T B is the threshold segmentation parameter for bone tissue;

[0022] S14. Perform bone density analysis on the segmented bone tissue region, and calculate the bone density D(x, y, z) of each voxel, where:

[0023]

[0024] where H air and H water represent the standard HU values of air and water respectively, and D(x, y, z) reflects the relative density of the bone tissue;

[0025] S15. Generate a three-dimensional digital model M oral of the patient's oral cavity based on the bone tissue region, soft tissue region, and the bone density of each voxel:

[0026] M oral = {B(x, y, z), D(x, y, z), S(x, y, z)}.

[0027] Optionally, the S2 step specifically includes:

[0028] S21. Establish an objective function f oral for the biomechanical stability of the implant based on the three-dimensional digital model M bio . The objective function for the biomechanical stability of the implant is calculated based on the local stress distribution σ(x, y, z), the corresponding voxel volume V(x, y, z), and the bone density D(x, y, z) within the bone tissue region B(x, y, z):

[0029]

[0030] where ∈ is a small positive number set to prevent division by zero;

[0031] S22. Establish an objective function f aesth for the aesthetics of the implant, and the objective function for the aesthetics of the implant is evaluated based on the morphological difference between the soft tissue region S(x, y, z) and the target soft tissue morphology function T s (x, y, z):

[0032]

[0033] Among them, Ω S represents the integration domain of the soft tissue region;

[0034] S23. Establish the implant bite force transfer objective function f occl , and the implant bite force transfer objective function is evaluated within the bone tissue region B(x, y, z) according to the difference between the local stress F(x, y, z) and the average stress within the bone tissue region :

[0035]

[0036] Among them, F(x, y, z) represents the local stress within the bone tissue region under the action of the bite force, represents the average stress within the bone tissue region;

[0037] S24. Establish the implant soft and hard tissue matching degree objective function f match , and the implant soft and hard tissue matching degree objective function is evaluated by combining the bone density information D(x, y, z) and the matching situation of the soft tissue region S(x, y, z) at the implant implantation site:

[0038]

[0039] Among them, represents the junction region between the bone tissue and the soft tissue, D opt is the ideal bone density, and S opt (x, y, z) is the target soft tissue distribution function at the implant implantation site;

[0040] S25. Integrate the objective functions of S21 - S24 to construct the multi-objective optimization environment F opt ;

[0041] F opt ={f bio , f aesth , f occl , f match}.

[0042] Optionally, the specific steps of the S3 step include:

[0043] S31. Based on the patient's oral three-dimensional digital model M oral and the multi-objective optimization environment, construct a multi-agent interaction model with enhanced game learning. The multi-agent interaction model conducts exploratory reinforcement learning with multiple hypothetical implants as game agents. The parameter combination of the enhanced game agent is defined as:

[0044] θ = {d, l, p};

[0045] Wherein, d is the diameter of the implant, l is the length of the implant, and p is the surface treatment process of the implant;

[0046] S32. Construct the dynamic reward function R(θ, t) of the reinforcement learning module, and the dynamic reward function is dynamically updated according to the multi-objective function fed back by the population game strategy module:

[0047]

[0048] Wherein, w j (t) is the weight coefficient of the j-th objective function in the t-th iteration, and f j (θ) is the multi-objective function, which respectively corresponds to the implant biomechanical stability objective function f bio , the implant aesthetics objective function f aesth , the implant bite force transmission objective function f occl and the implant hard and soft tissue matching degree objective function f match ;

[0049] S33. Construct the game interaction revenue function U bio (θ) of the implant and the bone tissue area to evaluate the interaction mechanical performance of the implant and the patient's bone tissue area:

[0050]

[0051] Wherein, σ θ (x, y, z) is the stress generated in the bone tissue area after the hypothetical implant corresponding to the parameter combination θ is implanted, and |B| is the total number of voxels in the bone tissue area;

[0052] S34. The population game strategy module calculates the comprehensive performance score function U com (θ):

[0053]

[0054] Wherein, F θ (x, y, z) represents the local stress of the implant acting on the bone tissue area under the parameter combination θ, represents the average stress in the bone tissue area, is the bone tissue and soft tissue junction area, D opt is the ideal bone density, and α, β, γ, δ are weight coefficients.

[0055] Optionally, the specific steps of the S4 step include:

[0056] S41. The population game strategy module outputs multiple implant parameter combinations θ k={d k , l k , p k}, where d k is the diameter of the k-th group of implants, l k is the length of the k-th group of implants, p k is the surface treatment process of the k-th group of implants. All parameter combinations constitute the initial candidate solution set Θ = {θ k | k = 1, 2, …, K};

[0057] S42. For the implant parameter combination θ k , the reinforcement learning module uses the comprehensive performance score function U com (θ k ) to dynamically adjust the weight coefficient w j (t) of the reward function:

[0058]

[0059] where λ is the step factor for dynamic weight adjustment, U com (θ k ) is the value of the comprehensive performance score function corresponding to the k-th group of parameter combinations, and w j (t) is the weight coefficient of the j-th objective function in the t-th iteration;

[0060] S43. The reinforcement learning module calculates the dynamic reward value R(θ j , t) of the implant parameter combination based on the dynamically adjusted weight coefficient w k (t + 1):

[0061]

[0062] S43. The reinforcement learning module updates the state value function V(θ k , t) according to the dynamic reward function R(θ k , t), and the calculation formula is:

[0063]

[0064] where η is the learning rate, γ rl is the discount factor of reinforcement learning, and θ′ k is the new parameter combination state that can be selected in the next iteration step;

[0065] S44. Through multiple rounds of interaction and iterative loop of reinforcement learning and population game, until the following convergence judgment condition is met:

[0066] |V(θ k , t + 1) - V(θ k , t)| < δthr ;

[0067] Among them, δ thr is a preset convergence threshold, and the optimal parameter combination θ * ={d * , l * , p *} that finally meets the requirements of multi-objective optimization is determined.

[0068] Optionally, the specific steps of S5 include:

[0069] S51. Based on the optimized implant parameter combination θ * ={d * , l * , p *} and the three-dimensional digital model M of the patient's oral cavity oral , construct a personalized implant geometric model M implant in the CAD / CAM module;

[0070] S52. Based on the relationship between the bone tissue region B(x, y, z) and the implant diameter d * , implant length l * , calculate the implant trajectory T implant (x, y, z):

[0071]

[0072] Among them, D opt is the ideal bone density, S opt is the ideal soft tissue morphology, and W B and W S respectively represent the bone tissue and soft tissue weight coefficients;

[0073] S53. Based on the surface treatment parameter p * of the implant and the bone density information D(x, y, z), optimize the interface contact characteristics between the implant and the bone tissue, and calculate the implant surface roughness distribution function R s (x, y, z):

[0074]

[0075] Among them, R s (x, y, z) reflects the roughness characteristics of different regions of the implant surface to optimize the bonding strength between the implant and the bone tissue;

[0076] S54. According to the implant trajectory T implant (x, y, z) and the optimized implant geometric parameters θ * , generate the final personalized implant three-dimensional model M in the CAD / CAM moduleimplant :

[0077] M implant ={T implant (x, y, z), R s (x, y, z), d * , l * , p *};

[0078] The three - dimensional model M implant ensures that the geometric structure of the implant matches the characteristics of the patient's bone tissue and soft tissue, meeting the requirements of biomechanical stability and aesthetics.

[0079] The beneficial effects of the present invention are as follows:

[0080] (1) By constructing an enhanced game - learning model, combining a reinforcement - learning module and a population - game strategy module, the present invention dynamically optimizes the design parameters of the implant. Using the population - game strategy to explore parameter combinations in a multi - objective optimization environment, and continuously updating the reward function through reinforcement learning, the implant parameters gradually converge to the optimal solution during the iterative process, and can adaptively adjust the implant design according to the bone density, soft - tissue morphology, and bite - force distribution of different patients, improving the personalized matching degree and implantation success rate.

[0081] (2) The present invention breaks through the limitation of traditional single - optimization objectives, constructs four key objective functions of biomechanical stability, aesthetics, bite - force transmission, and hard - and - soft - tissue matching degree, and adopts a multi - objective optimization strategy for joint solution. Through the multi - objective optimization environment, different objectives are dynamically balanced under the weight - adjustment mechanism, so as to ensure that the implant can meet both the biomechanical stability and aesthetic requirements after implantation.

[0082] (3) Based on the patient's CBCT image data, the present invention generates a high - precision three - dimensional digital oral model, and combines mechanical simulation analysis to predict and optimize the force distribution of the implant during the design stage. During the implant design process, the stress - transmission situation is calculated in real time, and the parameters are automatically adjusted when the simulation analysis does not meet the preset threshold, ensuring the biomechanical adaptability of the final implant. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] The 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, but do not constitute a limitation to the present invention. In the drawings:

[0084] Figure 1 is a flowchart of a method for accurately generating a single - tooth implant based on enhanced game learning proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0085] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, and thus only showing the components related to the present invention.

[0086] Reference Figure 1 , a precise single-tooth implant generation method based on enhanced game learning, comprising the following steps:

[0087] S1. Collect the oral CBCT image data of the patient, and preprocess the oral CBCT image data to generate a three-dimensional digital model of the patient's oral cavity;

[0088] S2. Construct a multi-objective optimization environment required for implant design based on the three-dimensional digital model of the patient;

[0089] S3. In the multi-objective optimization environment for implant design, use the enhanced game learning algorithm to simulate the interaction relationship between the implant and the surrounding tissues. The enhanced game learning algorithm includes a reinforcement learning module and a population game strategy module;

[0090] S4. Iteratively optimize the implant parameter combinations output by the population game strategy module, and gradually converge the implant parameter configuration to the parameter combination that meets the multi-objective optimization requirements through multiple rounds of game interaction and reinforcement learning updates. At the same time, dynamically adjust the weight of the reward function according to the digital model of the patient;

[0091] S5. Use the optimized implant parameter combination to generate a three-dimensional model of a single-tooth implant in the three-dimensional digital model that is adapted to the bone tissue and soft tissue characteristics of the patient;

[0092] S6. Perform mechanical simulation analysis on the generated three-dimensional model of the single-tooth implant, simulate the stress distribution of the implant under the action of biting force and its stress transfer effect with the surrounding bone tissue, and at the same time evaluate the biomechanical stability and aesthetic layout of the implant. If the simulation analysis does not meet the preset threshold conditions, feedback the analysis results to the enhanced game learning algorithm for parameter adjustment and re-optimization.

[0093] In this embodiment, step S1 specifically includes:

[0094] S11. Collect the cone beam computed tomography image data of the patient's oral cavity. The cone beam computed tomography image data is composed of multiple consecutive slices I i (x, y), where I i (x, y) represents the gray value of the i-th image slice at the spatial coordinates (x, y);

[0095] S12. Perform noise reduction processing on the cone beam computed tomography image data;

[0096] S13. Perform tissue segmentation on the denoised cone beam computed tomography (CBCT) image data. Using a combined method based on threshold segmentation and region growing algorithm, segment the image data into a bone tissue region B(x, y, z) and a soft tissue region S(x, y, z), where:

[0097]

[0098] where H(x, y, z) is the Hounsfield unit of the CBCT image data at the three-dimensional coordinates (x, y, z), and T B is the threshold segmentation parameter for bone tissue;

[0099] S14. Perform bone density analysis on the segmented bone tissue region, and calculate the bone density D(x, y, z) of each voxel, where:

[0100]

[0101] where H air and H water represent the standard HU values of air and water respectively, and D(x, y, z) reflects the relative density of bone tissue;

[0102] S15. Generate a three-dimensional digital model M oral of the patient's oral cavity based on the bone tissue region, soft tissue region, and bone density of each voxel:

[0103] M oral = {B(x, y, z), D(x, y, z), S(x, y, z)}.

[0104] In this embodiment, step S2 specifically includes:

[0105] S21. Establish an objective function f oral for the biomechanical stability of the implant based on the three-dimensional digital model M bio . The objective function for the biomechanical stability of the implant is calculated based on the local stress distribution σ(x, y, z), the corresponding voxel volume V(x, y, z), and the bone density D(x, y, z) within the bone tissue region B(x, y, z):

[0106]

[0107] where ∈ is a small positive number set to prevent division by zero;

[0108] S22. Establish an objective function f aesth for the aesthetics of the implant, and evaluate the objective function for the aesthetics of the implant based on the morphological difference between the soft tissue region S(x, y, z) and the target soft tissue morphology function T s (x, y, z):

[0109]

[0110] Among them, Ω S represents the integration domain of the soft tissue region;

[0111] S23. Establish the implant bite force transfer objective function f occl , and the implant bite force transfer objective function is evaluated according to the difference between the local stress F(x, y, z) and the average stress in the bone tissue region B(x, y, z) within the bone tissue region: :

[0112]

[0113] Among them, F(x, y, z) represents the local stress in the bone tissue region under the action of the bite force, represents the average stress in the bone tissue region;

[0114] S24. Establish the implant hard and soft tissue matching degree objective function f match , and the implant hard and soft tissue matching degree objective function is evaluated by combining the bone density information D(x, y, z) and the matching situation of the soft tissue region S(x, y, z) at the implant implantation site:

[0115]

[0116] Among them, represents the junction region between the bone tissue and the soft tissue, D opt is the ideal bone density, and S opt (x, y, z) is the target soft tissue distribution function at the implant implantation site;

[0117] S25. Integrate the objective functions of S21 - S24 to construct a multi-objective optimization environment F opt ;

[0118] F opt ={f bio , f aesth , f occl , f match}.

[0119] In this embodiment, the S3 step specifically includes:

[0120] S31. Based on the patient's oral three-dimensional digital model M oral and the multi-objective optimization environment, construct a multi-agent interaction model with enhanced game learning. The multi-agent interaction model conducts exploratory reinforcement learning with multiple hypothetical implants as game agents. The parameter combinations of the enhanced game agents are defined as:

[0121] θ = {d, l, p};

[0122] Wherein, d is the diameter of the implant, l is the length of the implant, and p is the surface treatment process of the implant;

[0123] S32. Construct the dynamic reward function R(θ, t) of the reinforcement learning module, and the dynamic reward function is dynamically updated according to the multi-objective function feedback by the population game strategy module:

[0124]

[0125] Wherein, w j (t) is the weight coefficient of the j-th objective function in the t-th iteration, and f j (θ) is the multi-objective function, which respectively corresponds to the implant biomechanical stability objective function f bio , the implant aesthetics objective function f aesth , the implant bite force transmission objective function f occl and the implant hard and soft tissue matching degree objective function f match ;

[0126] S33. Construct the game interaction revenue function U bio (θ) of the implant and the bone tissue region to evaluate the interactive mechanical performance of the implant and the patient's bone tissue region:

[0127]

[0128] Wherein, σ θ (x, y, z) is the stress generated in the bone tissue region after the hypothetical implant corresponding to the parameter combination θ is implanted, and |B| is the total number of voxels in the bone tissue region;

[0129] S34. The population game strategy module calculates the comprehensive performance score function U com (θ) of each game player under multi-objective conditions:

[0130]

[0131] Wherein, F θ (x, y, z) represents the local stress of the implant acting on the bone tissue region under the parameter combination θ, represents the average stress in the bone tissue region, is the bone tissue and soft tissue junction region, D opt is the ideal bone density, and α, β, γ, δ are weight coefficients.

[0132] In this embodiment, the S4 step specifically includes:

[0133] S41. The population game strategy module outputs multiple implant parameter combinations θ k={d k , l k , p k}, where d k is the diameter of the k-th group of implants, l k is the length of the k-th group of implants, p k is the surface treatment process of the k-th group of implants, and all parameter combinations constitute the initial candidate solution set Θ = {θ k | k = 1, 2, …, K};

[0134] S42. For the implant parameter combination θ k , the reinforcement learning module uses the comprehensive performance score function U com (θ k ) to dynamically adjust the weight coefficient w j (t) of the reward function:

[0135]

[0136] where λ is the step factor for dynamic weight adjustment, U com (θ k ) is the value of the comprehensive performance score function corresponding to the k-th group of parameter combinations, and w j (t) is the weight coefficient of the j-th objective function in the t-th iteration;

[0137] S43. The reinforcement learning module calculates the dynamic reward value R(θ j , t) of the implant parameter combination based on the dynamically adjusted weight coefficient w k (t + 1):

[0138]

[0139] S43. The reinforcement learning module updates the state value function V(θ k , t) according to the dynamic reward function R(θ k , t), and the calculation formula is:

[0140]

[0141] where η is the learning rate, γ rl is the discount factor of reinforcement learning, and θ′ k is the new parameter combination state that can be selected in the next iteration step;

[0142] S44. Through multiple rounds of interaction and iterative cycles of reinforcement learning and population game, until the following convergence determination condition is met:

[0143] |V(θ k , t + 1) - V(θ k , t)| < δthr ;

[0144] Among them, δ thr is a preset convergence threshold, and the optimal parameter combination θ * ={d * , l * , p *} that finally meets the requirements of multi-objective optimization is determined.

[0145] In this embodiment, step S5 specifically includes:

[0146] S51. Based on the optimized implant parameter combination θ * ={d * , l * , p *} and the three-dimensional digital model M of the patient's oral cavity, oral construct a personalized implant geometric model M implant in the CAD / CAM module;

[0147] S52. Based on the relationship between the bone tissue region B(x, y, z) and the implant diameter d * , implant length l * , calculate the implant trajectory T implant (x, y, z):

[0148]

[0149] Among them, D opt is the ideal bone density, S opt is the ideal soft tissue morphology, and W B and W S respectively represent the bone tissue and soft tissue weight coefficients;

[0150] S53. Based on the surface treatment parameter p * of the implant and the bone density information D(x, y, z), optimize the interface contact characteristics between the implant and the bone tissue, and calculate the implant surface roughness distribution function R s (x, y, z):

[0151]

[0152] Among them, R s (x, y, z) reflects the roughness characteristics of different regions of the implant surface to optimize the bonding strength between the implant and the bone tissue;

[0153] S54. According to the implant trajectory T implant (x, y, z) and the optimized implant geometric parameters θ * , generate the final personalized implant three-dimensional model M in the CAD / CAM moduleimplant :

[0154] M implant = {T implant (x, y, z), R s (x, y, z), d * , l * , p *};

[0155] Three - dimensional model M implant Ensures that the geometric structure of the implant matches the characteristics of the patient's bone and soft tissues, meeting the requirements of biomechanical stability and aesthetics.

[0156] Example 1:

[0157] At 9:00 am on March 15, 2024, Mr. Wang, a 35 - year - old male patient, came to the implant center of a dental hospital in City A. He fell in a cycling accident, resulting in the complete loss of the left maxillary first premolar (tooth No. 24). After preliminary examination, the doctor found that Mr. Wang's alveolar bone condition was rather special. The bone height was only 8.5 mm, the width was 4.2 mm, the bone density was uneven, and in some areas it was lower than 650 HU. Conventional implants had led to implantation failure. In order to improve the long - term stability of the implant, the doctor decided to use the method for precisely generating single - tooth implants based on enhanced game learning of the present invention to design a personalized implant for Mr. Wang.

[0158] At 9:30, Mr. Wang underwent a CBCT examination and complete three - dimensional oral imaging data was obtained. The system automatically analyzed the image, generated a three - dimensional digital model, and performed the following data analysis:

[0159] Alveolar bone height: 8.5 mm; alveolar bone width: 4.2 mm; local bone density distribution: the highest is 780 HU, the lowest is 560 HU; bite force distribution analysis: the chewing force on the left side is stronger, accounting for 58%; soft tissue thickness: 2.3 mm

[0160] Under the traditional method, doctors would directly select existing standardized implants on the market, usually choosing implants with a diameter of 3.8 mm and a length of 8 mm. However, due to Mr. Wang's low bone density, implants under the traditional method would result in insufficient initial stability after surgery, affecting the bone - bonding effect.

[0161] At 9:45, the system automatically established a multi - objective optimization environment for the implant and started to execute the enhanced game learning algorithm to optimize the key parameters (diameter, length, surface treatment process) of the implant. In the initial stage, the system generated 100 groups of implant parameter combinations and optimized them based on the following key factors:

[0162] Biomechanical stability objective: Maximize the bonding force between the implant and the bone tissue and reduce stress concentration.

[0163] Aesthetic goal: Optimize the shape of the implant to make it more conform to the shape of natural teeth.

[0164] Occlusal force transmission goal: Optimize the force-bearing situation of the implant to avoid loosening of the implant caused by uneven force during long-term use.

[0165] Hard and soft tissue matching goal: Ensure the matching degree of the implant at the junction of hard and soft tissues, and reduce the risk of soft tissue inflammation around the implant.

[0166] After 3500 iterations of reinforcement learning, the system finally output the optimal implant design:

[0167] Diameter: 4.1 mm; Length: 9 mm; Surface treatment process: Micro-nano coating treatment; The implant matching degree is increased by 15%, and the soft tissue matching error is reduced by 12%; The bone density adaptation degree is increased from 72% of the traditional plan to 86%; Mechanical simulation analysis shows that the maximum stress of the implant is reduced by 13%, and the stress uniformity is increased by 22%.

[0168] The doctor checked the optimization results and confirmed that the implant design was reasonable. Then, a 3D printing technology was used to quickly produce an implant model for pre-implantation testing. Mr. Wang entered the operating room. After local anesthesia, the doctor began the implant implantation. Using the implant optimized by the method of the present invention, the operation process was very smooth: the implant socket was precisely matched with the implant, reducing additional bone excavation; After the implant was implanted, the torque value reached 45 N·cm, and the initial stability was increased by 18% compared with the traditional method; The implant remained stable in the area with low bone density and there was no loosening phenomenon. After the operation, Mr. Wang entered the recovery room, and the doctor arranged regular follow-up visits to observe the postoperative bone integration situation.

[0169] 1 month after surgery: The CBCT examination results showed that the bone tissue density around the implant increased and there was no obvious bone resorption phenomenon. Mr. Wang did not report any discomfort.

[0170] 3 months after surgery: The implant had good bone integration and the chewing function recovered to the normal level.

[0171] 6 months after surgery: The patient had good adaptability, the bone absorption rate was reduced by 67%, and the bone absorption range was reduced by about 0.7 mm compared with the traditional method.

[0172] 12 months after surgery: The implant was completely stable, and the patient reported that the chewing comfort score increased by 24%.

[0173] In this experiment, we selected 100 patients for comparison. Among them, 50 patients used the traditional method and 50 patients used the method of the present invention. The data of implant stability and patient satisfaction are as follows:

[0174]

[0175]

[0176] It can be seen from the comparison of data that the method of the present invention has significant advantages in improving the implant matching degree, reducing postoperative bone resorption and enhancing the patient experience, especially in patients with lower bone density. The traditional method mainly relies on the doctor's experience and lacks precise optimization for individual conditions. However, the present invention realizes the intelligent adjustment of implant parameters through enhanced game learning and mechanical simulation analysis, greatly improving the biomechanical adaptability and long-term stability of implants.

[0177] The present invention constructs an enhanced game learning model, combines a reinforcement learning module and a population game strategy module to dynamically optimize the design parameters of implants. The population game strategy is used to explore parameter combinations in a multi-objective optimization environment, and the reward function is continuously updated through reinforcement learning so that the implant parameters gradually converge to the optimal solution during the iteration process, and can adaptively adjust the implant design according to the bone density, soft tissue morphology and bite force distribution of different patients, improving the personalized matching degree and implantation success rate.

[0178] The present invention breaks through the limitation of traditional single optimization objectives, constructs four key objective functions of biomechanical stability, aesthetics, bite force transmission, and hard and soft tissue matching degree, and adopts a multi-objective optimization strategy for joint solution. Through the multi-objective optimization environment, different objectives are dynamically balanced under the weight adjustment mechanism, so as to ensure that the implant can meet the biomechanical stability and aesthetic requirements simultaneously after implantation.

[0179] The present invention generates a high-precision three-dimensional digital oral model based on the patient's CBCT image data, and combines mechanical simulation analysis to predict and optimize the stress distribution of the implant in the design stage. During the implant design process, the stress transmission situation is calculated in real time, and the parameters are automatically adjusted when the simulation analysis does not meet the preset threshold to ensure the biomechanical adaptability of the final implant.

[0180] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A method for accurately generating a single dental implant based on enhanced game learning, characterized in that: The steps include: S1, collecting the patient's oral CBCT image data, and preprocessing the oral CBCT image data to generate a three-dimensional digital model of the patient's oral cavity; S2, constructing a multi-objective optimization environment for implant design based on the patient's 3D digital model; S3. In a multi-objective optimization environment for implant design, a reinforcement game learning algorithm is used to simulate the interaction between the implant and the surrounding tissue, wherein the reinforcement game learning algorithm includes a reinforcement learning module and a group game strategy module; S4. Iteratively optimize the implant parameter combination output by the group game strategy module, and gradually converge the implant parameter configuration to a parameter combination that meets the multi-objective optimization requirements through multiple rounds of game interaction and reinforcement learning updates. At the same time, dynamically adjust the weight of the reward function according to the patient's digital model; S5, using the optimized implant parameter combination to generate a single tooth implant three-dimensional model that matches the patient's bone tissue and soft tissue characteristics in the three-dimensional digital model; S6. Perform mechanical simulation analysis on the generated three-dimensional model of a single tooth implant to simulate the force distribution of the implant under the action of occlusal force and the stress transfer effect with the surrounding bone tissue, and evaluate the biomechanical stability and aesthetic layout of the implant. If the simulation analysis does not meet the preset threshold conditions, the analysis results are fed back to the enhanced game learning algorithm for parameter adjustment and re-optimization.

2. The method for accurately generating a single dental implant based on enhanced game learning according to claim 1, characterized in that: The S1 step specifically includes: S11, collecting cone beam computer tomography image data of the patient's oral cavity, wherein the cone beam computer tomography image data consists of a plurality of continuous slices I i (x,y), where I i (x,y) represents the gray value of the i-th image slice at the spatial coordinate (x,y); S12, performing noise reduction processing on the cone beam computer tomography image data; S13, performing tissue segmentation on the cone beam computed tomography image data after noise reduction, using a combination of threshold segmentation and region growing algorithm to segment the image data into a bone tissue region B (x, y, z) and a soft tissue region S (x, y, z), wherein: Where H(x,y,z) is the Hounsfield unit of the cone-beam computed tomography image data at the three-dimensional coordinate (x,y,z), T B is the threshold segmentation parameter of bone tissue; S14. Perform bone density analysis on the segmented bone tissue area and calculate the bone density D(x, y, z) of each voxel, where: Among them, H air and H water represent the standard HU values ​​of air and water, respectively, and D(x, y, z) reflects the relative density of bone tissue; S15. Generate a three-dimensional digital model M of the patient's oral cavity based on the bone tissue area, soft tissue area and bone density of each voxel oral : M oral ={B(x,y,z),D(x,y,z),S(x,y,z)}。 3. The method for accurately generating a single dental implant based on enhanced game learning according to claim 1, characterized in that: The S2 step specifically includes: S21, based on the three-dimensional digital model M oral Establishing implant biomechanical stability objective function f bio The implant biomechanical stability objective function is calculated based on the local stress distribution σ(x,y,z) in the bone tissue area B(x,y,z), the corresponding voxel volume V(x,y,z) and the bone density D(x,y,z): Among them, ∈ is a small positive number set to prevent division by zero; S22. Establishing the implant aesthetic objective function f aesth The implant aesthetic objective function is based on the soft tissue area S (x, y, z) and the target soft tissue morphology function T s The morphological differences between (x,y,z) are evaluated: Among them, Ω S The integration domain represents the soft tissue region; S23. Establishing implant occlusal force transmission objective function f occl The implant occlusal force transmission objective function is in the bone tissue area B(x,y,z), according to the local stress F(x,y,z) and the average stress in the bone tissue area The difference is evaluated: Among them, F(x,y,z) represents the local stress in the bone tissue area under the bite force, Represents the average stress within the bone tissue area; S24. Establishing the objective function f of the matching degree between implant soft and hard tissues match The objective function of the matching degree of the soft and hard tissues of the implant is combined with the bone density information D(x, y, z) and the matching of the soft tissue area S(x, y, z) at the implant placement site to evaluate: in, Indicates the junction area between bone tissue and soft tissue, D opt For ideal bone density, S opt (x, y, z) is the target soft tissue distribution function of the implant placement site; S25. Integrate the objective functions of S21-S24 to construct a multi-objective optimization environment for implant design F opt ; F opt ={f bio ,f aesth ,f occl ,f match }。 4. The method for accurately generating a single dental implant based on enhanced game learning according to claim 1, characterized in that: The S3 step specifically includes: S31, based on the patient's oral 3D digital model M oral A multi-agent interaction model of enhanced game learning is constructed in a multi-objective optimization environment. The multi-agent interaction model uses multiple hypothetical implants as game subjects to conduct reinforcement learning experiments. The parameter combination of the enhanced game subject is defined as: θ={d,l,p}; Wherein, d is the diameter of the implant, l is the length of the implant, and p is the surface treatment process of the implant; S32. Construct a dynamic reward function R(θ,t) of the reinforcement learning module. The dynamic reward function is dynamically updated according to the multi-objective function fed back by the group game strategy module: Among them, w j (t) is the weight coefficient of the jth objective function in the tth iteration, f j (θ) is a multi-objective function, corresponding to the implant biomechanical stability objective function f bio , implant aesthetics objective function f aesth , implant occlusal force transmission objective function f occl The objective function f of the matching degree between implant soft and hard tissues match ; S33. Constructing the interactive profit function U between the implant and the bone tissue area bio (θ) to evaluate the mechanical interaction between the implant and the patient’s bone tissue area: Among them, σ θ (x, y, z) is the stress generated in the bone tissue region after the hypothetical implant is implanted corresponding to the parameter combination θ, and |B| is the total number of voxels in the bone tissue region; S34, the group game strategy module calculates the comprehensive performance score function U of each game subject under multi-objective conditions com (θ): Among them, F θ (x, y, z) represents the local stress of the implant acting on the bone tissue area under the parameter combination θ, represents the average stress in the bone tissue area, D is the junction area between bone tissue and soft tissue. opt is the ideal bone density, and α, β, γ, and δ are weight coefficients.

5. The method for accurately generating a single dental implant based on enhanced game learning according to claim 4, characterized in that: The S4 step specifically includes: S41, the group game strategy module outputs multiple implant parameter combinations θ k ={d k ,l k ,p k }, where d k is the diameter of the implant in group k, l k is the length of the kth group of implants, p k is the surface treatment process of the kth group of implants, and all parameter combinations constitute the initial candidate solution set Θ = {θ k |k=1,2,…,K}; S42, for implant parameter combination θ k The reinforcement learning module uses the comprehensive performance score function U fed back by the group game strategy module com (θ k ) Dynamically adjust the weight coefficient w of the reward function j (t): Among them, λ is the step size factor of dynamic weight adjustment, U com (θ k ) is the comprehensive performance score function value corresponding to the kth group of parameter combinations, w j (t) is the weight coefficient of the jth objective function in the tth iteration; S43, reinforcement learning module based on dynamically adjusted weight coefficient w j (t+1), calculate the dynamic reward value R(θ k ,t): S43, the reinforcement learning module is based on the dynamic reward function R(θ k ,t) Update the state value function V(θ k ,t), the calculation formula is: Among them, η is the learning rate, γ rl is the discount factor for reinforcement learning, θ′ k The state of the new parameter combination that can be selected in the next iteration step; S44, through multiple rounds of reinforcement learning and group game interaction and iteration, until the following convergence judgment conditions are met: |V(θ k ,t+1)-V(θ k ,t)|<δ thr ; Among them, δ thr is the preset convergence threshold, and determines the optimal parameter combination θ that ultimately meets the multi-objective optimization requirements. * ={d * ,l * ,p * }.

6. The method for accurately generating a single dental implant based on enhanced game learning according to claim 5, characterized in that: The S5 step specifically includes: S51, based on the optimized implant parameter combination θ * ={d * ,l * ,p * } and the patient's oral 3D digital model M oral Build personalized implant geometry models in the CAD / CAM module implant ; S52, based on the bone tissue area B (x, y, z) and the implant diameter d * , implant length l * The relationship between the implant and the implant trajectory T is calculated. implant (x,y,z): Among them, D opt For ideal bone density, S opt is the ideal soft tissue morphology, W B and W S Represent the weight coefficients of bone tissue and soft tissue respectively; S53, based on the implant surface treatment parameters p * and bone density information D(x,y,z), optimize the interface contact characteristics between the implant and bone tissue, and calculate the implant surface roughness distribution function R s (x,y,z): Among them, R s (x, y, z) reflects the roughness characteristics of the implant surface in different areas to optimize the bonding strength between the implant and bone tissue; S54, according to the implantation trajectory T implant (x,y,z) and optimized implant geometry θ * , generate the final personalized implant 3D model in the CAD / CAM module implant : M implant ={T implant (x,y,z),R s (x,y,z),d * ,l * ,p * }; The three-dimensional model M implant Ensure that the implant geometry matches the patient's bone and soft tissue characteristics to meet biomechanical stability and aesthetic requirements.

Images