A method for predicting the bone repair performance of oxidized chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterial
By using neural network models to predict the bone repair performance of hydrogels, this approach solves the problems of long R&D cycles and difficulty in determining parameter combinations in existing hydrogel materials. It enables rapid and accurate optimization design of hydrogel materials, thereby improving bone repair effects.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGSU DEVICELAND MEDICAL INSTR CORP LTD
- Filing Date
- 2026-05-25
- Publication Date
- 2026-07-31
AI Technical Summary
The development of existing hydrogel bone repair materials relies on repeated experiments, which is costly and time-consuming. It is difficult to quickly obtain the optimal parameter combination and there is a lack of systematic prediction methods for the multi-parameter-multi-performance mapping relationship of hydrogels using neural network models. In particular, the addition ratio and dispersion state of hydroxyapatite nanomaterials in hydrogels have complex effects, making it difficult to accurately predict bone repair performance.
By employing a neural network model in conjunction with data acquisition, preprocessing, model building, and training, the system outputs predictions of the bone repair performance of hydrogels by inputting concrete material parameters and preparation process parameters. Furthermore, it generates the optimal component parameter ratio through reverse feedback adjustment and utilizes the neural network model to handle nonlinear and multivariate coupling relationships, thereby achieving accurate predictions.
This method enables accurate prediction of the bone repair performance of hydrogel materials while reducing the number of experiments, providing a basis for rapid design optimization and improving the bone repair effect of hydrogel materials.
Smart Images
Figure CN122494048A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the interdisciplinary field of biomaterials and intelligent algorithms, specifically a method for predicting the bone repair performance of oxychloride sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials. Background Technology
[0002] Bone defect repair is a crucial clinical issue in orthopedics, dentistry, and trauma repair. For different types of bone defects (such as irregular bone defects and minimally invasive injection repair scenarios), hydrogels have gradually become a hot research topic in bone repair materials due to their excellent biocompatibility, injectability, and tunable three-dimensional network structure. However, hydrogels for bone repair must simultaneously meet multiple requirements in practical applications: possessing mechanical properties (elastic modulus, compressive strength, etc.) that match bone tissue, exhibiting good osteogenic properties (promoting osteoblast adhesion, proliferation, and differentiation), and possessing suitable degradation behavior to achieve gradual replacement of bone.
[0003] The development of current hydrogel bone repair materials mainly relies on repeated experiments, testing their mechanical properties, swelling properties, and degradation behavior by adjusting parameters such as monomer types and crosslinking densities. This process involves gradually screening for optimal performance combinations. However, this approach suffers from high trial-and-error costs, long development cycles, and difficulty in quickly obtaining the optimal parameter combinations. While neural networks have significant advantages in nonlinear relationship modeling with the advancement of artificial intelligence, there is currently a lack of systematic prediction methods for the multi-parameter-multi-performance mapping relationships of hydrogels, particularly in the deep integration of neural network models with hydrogel material engineering parameters, where a technological gap remains.
[0004] Hydroxyapatite (HA) is the main inorganic component of natural bone tissue, and its nanoscale form (nHA) exhibits excellent biocompatibility and osteogenic induction capabilities. Introducing hydroxyapatite nanomaterials into hydrogel systems helps improve the mechanical properties of the material and promotes osteoblast adhesion and differentiation. However, the proportion of nHA added to the hydrogel, its dispersion state, and its interaction with the matrix material significantly affect the bone repair performance of the composite hydrogel. The mechanisms of this influence are complex and difficult to accurately predict using traditional empirical methods. Therefore, there is an urgent need to provide a method for predicting the bone repair performance of oxy-hydroxyapatite nanomaterial-containing chondroitin sulfate / carboxymethyl chitosan hydrogels to overcome the shortcomings in current practical applications. Summary of the Invention
[0005] To achieve the above objectives, the present invention employs the following technical solution: a method for predicting the bone repair performance of oxychloride sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials, comprising the following steps: S1. Data Acquisition: Collect experimental data, which includes hydrogel material parameters, preparation process parameters and bone repair-related performance data of chondroitin sulfate, carboxymethyl chitosan and hydroxyapatite nanomaterials. S2. Data preprocessing: The experimental data is preprocessed, including normalization and encoding. S3. Model Building: Construct a neural network model based on the preprocessed data; S4. Model Training: Train the neural network model using the preprocessed data; S5. Performance Prediction and Application: Input the parameters of the hydrogel material to be predicted and the preparation process parameters into the trained neural network model, and output the prediction results of the bone repair performance of the hydrogel. S6. Obtain the medical scenario bias factors corresponding to different clinical scenarios, and perform a weighted fusion evaluation calculation based on the bone repair performance prediction results to obtain a comprehensive repair efficacy score. S7. When the comprehensive repair efficiency score does not meet the preset evaluation criteria, calculate the parameter variation matrix of the current model, and perform reverse feedback adjustment on the parameters of the hydrogel material to be predicted based on the extracted parameter variation length to generate the optimal component parameter ratio.
[0006] As a further aspect of the present invention, in S2, the data preprocessing further includes: Data cleaning and dataset partitioning: Data cleaning involves handling missing values, outliers, and obviously erroneous data; dataset partitioning involves randomly dividing the preprocessed dataset into training, validation, and test sets. In S4, the constructed neural network model is trained using training set data, with material parameters and preparation process parameters as inputs and corresponding bone repair performance data as the expected output. The network weights and biases are optimized through backpropagation algorithm, and validation set data is used to monitor model performance and prevent overfitting.
[0007] As a further aspect of the present invention, the parameters of the hydrogel material include: The oxidation degree of chondroitin sulfate, the substitution degree of carboxymethyl chitosan, the ratio parameters of chondroitin sulfate to carboxymethyl chitosan, the amount of hydroxyapatite nanomaterials added, the particle size range and dispersion mode of hydroxyapatite nanomaterials. The neural network model is used to establish a nonlinear mapping relationship between hydrogel material parameters and bone repair performance; The oxidation degree of the chondroitin sulfate is 40-80%, the substitution degree of the carboxymethyl chitosan is 65-90%, the concentration range of the chondroitin sulfate is 1-10%, the concentration range of the carboxymethyl chitosan is 0.5-5%, the mass concentration ratio of the chondroitin sulfate to the carboxymethyl chitosan is 1:(0.1-10), and the amount of hydroxyapatite nanomaterial added is 5-15%. The hydroxyapatite nanomaterials have a particle size range of 10-200 nm and are dispersed by one of the following methods: ultrasonic physical dispersion, surface modification dispersion, and in-situ acoustic field dispersion.
[0008] As a further aspect of the present invention, the preparation process parameters include: One or more of the following parameters: reaction environment parameters, crosslinking condition parameters, dispersion condition parameters, post-treatment parameters, and hydrogel system state parameters.
[0009] As a further aspect of the present invention, the reaction environment parameters include one or more of the following: reaction temperature, reaction time, and pH value of the reaction system. Crosslinking condition parameters include one or more of the following parameters: crosslinking mode and degree of crosslinking during hydrogel formation, and mixing order and mixing method of components. Dispersion condition parameters include one or more of the following: dispersion mode, time, and intensity; The state parameters of a hydrogel system include one or more of the following: concentration, solid content, and viscosity-related parameters.
[0010] As a further aspect of the present invention, the bone repair-related performance parameters include one or more of the following: mechanical performance data, osteogenic performance data, mineralization capacity, biocompatibility-related parameters, and degradation and stability-related parameters.
[0011] As a further aspect of the present invention, the mechanical performance parameters include one or more of elastic modulus and compressive modulus; Osteogenic performance parameters include osteoblast adhesion ability and osteogenic differentiation-related indicators; Mineralization capacity parameters are used to characterize the mineralization behavior of hydrogels during bone repair. Biocompatibility-related parameters include cell viability and cell proliferation capacity; Degradation and stability-related parameters include degradation rate and changes in structural stability.
[0012] As a further aspect of the present invention, the neural network model includes an input layer, at least one hidden layer, and an output layer; The number of neurons in the input layer corresponds to the number of preprocessed material parameters and process parameters; The number of neurons in the output layer corresponds to the number of bone repair-related performance indicators to be predicted.
[0013] As a further aspect of the present invention, step S6 specifically comprises: S61. Based on the system's preset evaluation rules, extract the predicted value of elastic modulus representing mechanical strength, the osteoblast attachment rate representing biological characteristics, and the three-dimensional network degradation cycle representing the consumption progress from the bone repair performance prediction results. S62. Match the specific stress state and tissue healing time window characteristics of the current implantation environment, and obtain the mechanical bias factor, osteogenic bias factor and degradation bias factor corresponding to mechanical, biological and degradation properties respectively, which together constitute the medical scenario bias factor. S63. Using the mechanical bias factor, the osteogenic bias factor, and the degradation bias factor, perform weighted product operations on the predicted elastic modulus value, the osteoblast attachment rate, and the three-dimensional network degradation cycle, respectively, and summarize all the weighted product calculation results to obtain the overall numerical form of the comprehensive repair efficacy score.
[0014] As a further aspect of the present invention, step S7 specifically comprises: S71. When it is determined that the comprehensive repair efficacy score does not meet the implantation standard, perform multivariate partial derivative calculation on the hydrogel material parameters to be predicted that are involved in the calculation to obtain the parameter derivative matrix that reflects the parameter bias characteristics. S72. Scan and extract the extreme value nodes in the parameter derivative matrix that represent the weight of performance influence, and calculate the parameter derivative length corresponding to different material compositions in combination with the parameter adjustment range limit. S73. Using the parameter variation and asynchrony length as an adjustment benchmark, perform multiple rounds of inverse numerical shift and model verification on the current parameters of the hydrogel material to be predicted. Record the proportion data that meets the qualification requirements and generate the final optimal component parameter proportion. A method for predicting the bone repair performance of hydroxyapatite nanomaterial-containing chondroitin sulfate / carboxymethyl chitosan hydrogel. Compared with the prior art, the beneficial effects of the present invention are as follows: This invention establishes an input-output relationship, where concrete material parameters and preparation process parameters are input, and the output is the mechanical properties, osteogenic properties, and mineralization capacity to be predicted by the neural network. This ensures that each set of experiments has a corresponding material and property relationship, ultimately enabling bone repair performance to correspond to a specific range when material parameters appear in a certain combination. Furthermore, the neural network model excels at handling nonlinear, multivariate coupling relationships like those related to bone repair, achieving accurate predictions with fewer experiments, thus providing a reliable basis for the parameter optimization design of bone repair hydrogel materials. Attached Figure Description
[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0016] Figure 1 This is a logical diagram illustrating the prediction and experimental verification of bone repair performance parameters.
[0017] Figure 2 This is a schematic diagram of the predictive control principle.
[0018] Figure 3 This is a schematic diagram of a linear SVR classifier. Detailed Implementation
[0019] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] The present invention will be further explained below with reference to specific embodiments.
[0021] Please see Figures 1-3 This invention provides a method for predicting the bone repair performance of oxychloride sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials. S1. Data Acquisition: Collect experimental data, which includes hydrogel material parameters, preparation process parameters and bone repair-related performance data of chondroitin sulfate, carboxymethyl chitosan and hydroxyapatite nanomaterials. S2. Data preprocessing: The experimental data is preprocessed, including normalization and encoding. S3. Model Building: Construct a neural network model based on the preprocessed data; S4. Model Training: Train the neural network model using the data; S5. Performance Prediction and Application: Input the parameters of the hydrogel material to be predicted and the preparation process parameters into the trained neural network model, and output the predicted results of the bone repair performance of the hydrogel.
[0022] In this embodiment, by establishing an "input-output" relationship, concrete material parameters and preparation process parameters are input, and the mechanical properties, osteogenic properties and mineralization capacity to be predicted by the neural network are output. This ensures that each set of experiments has a corresponding material-property correspondence. Ultimately, when material parameters appear in a certain combination, the bone-repairing performance can correspond to the corresponding range. At the same time, the neural network model is good at handling nonlinear and multivariate coupling relationships such as bone-repairing. Based on the prediction results, the bone repair performance of hydrogels under different parameter combinations was compared and analyzed, and parameter combinations with better bone repair performance were screened to guide the parameter optimization design of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogels containing hydroxyapatite nanomaterials.
[0023] Further, in S1, the oxidation degree of chondroitin sulfate is 40-80%, the substitution degree of carboxymethyl chitosan is 65-90%, the concentration range of chondroitin sulfate is 1-10%, the concentration range of carboxymethyl chitosan is 0.5-5%, the mass concentration ratio of chondroitin sulfate to carboxymethyl chitosan is 1:(0.1-10), and the amount of hydroxyapatite nanomaterial added is 5-15%.
[0024] Furthermore, in S2, the data preprocessing specifically includes:
[0025] Data cleaning: handling missing values, outliers, and obviously erroneous data; Data normalization / standardization: Normalize each input parameter (material and process parameter) and output parameter (performance data) separately (such as Min-Max normalization or Z-score standardization) to eliminate the influence of dimensions and accelerate model convergence; Dataset partitioning: The preprocessed dataset is randomly divided into training, validation and test sets, for example, in a ratio of 70%:15%:15%.
[0026] The neural network model includes an input layer, at least one hidden layer, and an output layer; The number of neurons in the input layer corresponds to the number of preprocessed material parameters and process parameters; The number of neurons in the output layer corresponds to the number of bone repair-related performance indicators to be predicted; The number of hidden layers and the number of neurons in each layer are determined through optimization based on the complexity of the data.
[0027] The model training involves training the constructed neural network model using training set data, with the aforementioned material parameters and process parameters as inputs and the corresponding bone repair performance data as the expected output. The network weights and biases are optimized using the backpropagation algorithm. During the training process, validation set data is used to monitor the model performance to prevent overfitting, and the error between the predicted value and the true value is evaluated using a loss function (such as mean squared error, MSE).
[0028] Optimization and hyperparameter tuning: Adjust the learning rate, number of hidden layers, number of neurons, activation function (such as ReLU, Sigmoid), optimizer (such as Adam, SGD), and other hyperparameters of the neural network, and select the optimal model structure using the performance on the validation set.
[0029] Furthermore, the bone repair performance prediction results are used to guide the parameter optimization design of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials.
[0030] Furthermore, in S3, the neural network model is a model used to establish a nonlinear mapping relationship between material parameters and bone repair performance; The nonlinear mapping relationship between material parameters and bone repair performance was investigated. A training sample set was constructed by using a multi-parameter tiered combination method. Within a reasonable range, material composition parameters and process parameters were systematically combined to obtain representative hydrogel samples. Their correlation with bone repair was then tested, thus forming a sample dataset for neural network training. Support Vector Machines (SVMs) are used to train nonlinear data. A supervised learning model is built by using the regression and classification functions of machine learning. Then, the concept of soft margin is introduced into Support Vector Regression (SVR) to transform inequality constraints into equality constraints, thereby obtaining the Support Vector Regression model.
[0031] Furthermore, in S4, the experimental data is obtained from experimental measurements.
[0032] Furthermore, in S5, the material parameters include: the degree of oxidation of chondroitin sulfate, the degree of substitution of carboxymethyl chitosan, the ratio of chondroitin sulfate to carboxymethyl chitosan, the amount of hydroxyapatite nanomaterials added, and the particle size range and dispersion mode of the hydroxyapatite nanomaterials.
[0033] Furthermore, in S5, the preparation process parameters include one or more of the following: reaction environment parameters, crosslinking condition parameters, dispersion condition parameters, post-treatment parameters, and hydrogel system state parameters.
[0034] Furthermore, in S5, the environmental parameters include one or more of the following: reaction temperature, reaction time, and pH value of the reaction system. Crosslinking condition parameters include one or more of the following parameters: crosslinking mode and degree of crosslinking during hydrogel formation, and mixing order and mixing method of components. The dispersion condition parameters for hydroxyapatite nanomaterials include one or more of the following: dispersion mode, time, and intensity. The state parameters of a hydrogel system include one or more of the following: concentration, solid content, and viscosity-related parameters.
[0035] Furthermore, the bone repair-related performance parameters include one or more of the following: mechanical property data, osteogenic performance data, mineralization capacity, biocompatibility-related parameters, and degradation and stability-related parameters.
[0036] Furthermore, the mechanical performance parameters include one or more of the elastic modulus and the compressive modulus; Osteogenic performance parameters include osteoblast adhesion ability and osteogenic differentiation-related indicators; Mineralization capacity parameters are used to characterize the mineralization behavior of hydrogels during bone repair. Biocompatibility-related parameters include cell viability and cell proliferation capacity; Degradation and stability-related parameters include degradation rate and changes in structural stability.
[0037] Furthermore, the mechanical support properties of hydrogels determine whether they provide a stable osteogenic mechanical microenvironment, their osteogenic biological properties determine whether they promote osteoblast differentiation, and their mineralization capacity determines the formation of bone-like apatite structures.
[0038] Preferably, a compression modulus of 0.1-500 kPa, more preferably 1-100 kPa, provides better elastic recovery and a more stable osteogenic microenvironment.
[0039] Furthermore, the particle size range of the hydroxyapatite nanomaterial is 10-200 nm, and the dispersion method includes one of ultrasonic physical dispersion, surface modification dispersion, and in-situ acoustic field dispersion.
[0040] Furthermore, the bone-repairing performance of the hydrogel was demonstrated in a rat model with critical skull bone defects, using a circular model with a diameter of 8 mm. The advantages of this model include its non-load-bearing nature, controllable shape, and high reproducibility. It is ideally suited for comparing the effects of different formulations (OCS / CMCS ratio, nHA content, dispersion state) on bone regeneration.
[0041] This invention utilizes a neural network model and support vector regression to classify and train parameters such as those of hydrogel materials containing chondroitin sulfate, carboxymethyl chitosan, and hydroxyapatite nanomaterials, including their preparation processes. This yields a stable training set with ideal performance. The control law of the controller is then deduced from the training set. To improve the accuracy of the prediction model and reduce the error between the ideal bone repair performance and the actual results, the original linear relationship model is optimized by introducing a predictive control method. The optimal control input is calculated using a given objective function. Predictive control can achieve continuous rolling optimization within a finite time region, and the rolling optimization process is online and iterative.
[0042] The basic principle of predictive control is to provide the system with a certain reference trajectory, observe the difference between the output of the predictive model and the output of the controlled object, then reduce the difference in the feedback correction stage, and finally, the rolling optimization controller calculates the control law and continues to predict the actual output of the controlled object. The principle of predictive control is as follows: Figure 2 As shown, z(x) of the control system is used as the reference trajectory, z(x) is the control transfer function of the control system before optimization, y(k) is the actual output of the controlled object, and u(k) is the control signal at time k. p (k+1) represents the output after feedback correction, y m (k+1) represents the output of the prediction model. The core of predictive control is to establish a prediction model that best approximates the true output. Historical output information y(k), y(k-1), ..., y(k-n+1) and historical input information u(k), u(k-1), ..., u(k-m+1) are used as inputs to the SVR. m (k+1) is used as the output of SVR. The regression model is trained using historical data to make SVR highly approximate the nonlinear mapping function. (x), then (x) can be used as a predictive mathematical model for the control system.
[0043] Using SVR, a hyperplane is found in the n-dimensional data space as a classification condition function to classify the training sample set and select sample points with good control performance. The following equation is obtained: Equation (2-1) In the formula These are different categories, represented by 1 or -1. 1 represents the degree of oxidation of chondroitin sulfate, and -1 represents the degree of substitution of carboxymethyl chitosan. This represents the weight vector, where b is the bias constant term. Let i be the i-th data point in the data sample, and its structure is as follows: Figure 3 As shown.
[0044] refer to Figure 3The sampling points on the interval lines become support vectors. The training set is classified using these support vectors, selecting sample points with excellent control performance. The dashed line in the diagram represents the hyperplane of the SVR classifier, which is theoretically the optimal point for control performance. The distance from the support vectors to the hyperplane in the diagram is denoted by d (i.e., the optimization objective function):
[0045] By combining and transforming the system of equations, we obtain:
[0046] From the above It can be obtained Therefore, distance Written as:
[0047] At this point, in order to obtain the boundary range with the maximum support vectors, this problem is transformed into solving for the maximum value of distance d, where: Given two tangent lines in the dataset, i.e., the maximum distance from a sampling point to the tangent line is 2, the maximum distance from a support vector to the hyperplane in the graph is converted to:
[0048] To simplify calculations, let's change the format: ; in, It is a weight vector. It is sample data.
[0049] The Lagrange function has significant applications in solving constrained extremum problems. It can transform the problem of solving a constrained control system with an unknown transfer function into an unconstrained problem, allowing the control parameters of the control system to be solved by finding the Lagrange multipliers. To address this problem, the Lagrange function is constructed as follows:
[0050] In the formula For Lagrange coefficients, According to Lagrange's duality theory, the above equation can be transformed into its dual form:
[0051] To improve the generalization ability of the function, all sample points are added as slack variables. The objective function is rewritten as follows:
[0052] Where, constant It is a regularization parameter (also known as a penalty coefficient). The larger the value, the higher the loss function value for misclassified data, resulting in more accurate predictions. However, compared to data with lower values... Compared to the model with the given values, this model will be more complex and may have weaker generalization performance. The rewritten optimization objective function is transformed into a Lagrange dual form:
[0053] in yes The Lagrange multipliers were used, and the result was finally obtained using the sequence minimum optimization algorithm.
[0054] Where S is the set of support vectors, and This represents a support vector sample. With the weight vector and bias constant, the final regression model is:
[0055] The obtained regression model is used as the prediction model for the control system. In order to achieve better control performance, the prediction model should be optimized. (x) The prediction model perfectly fits the true output y(k) of the controlled object. However, in actual motion control, the prediction model will inevitably deviate from the actual controlled object to some extent. Furthermore, in the actual production environment, the predicted value may differ from the actual value due to noise and interference. Therefore, a feedback correction loop is introduced into the control system to obtain this error value and feed it back to the prediction model and the rolling optimization controller for correction. The control quantity sequence u(k) obtained by the rolling optimization loop directly affects the control effect of the controlled object. Therefore, introducing the rolling optimization method into the control system to optimize the prediction model can reduce the prediction model's performance. The error between (x) and the true output y(k) allows the predictive model to more closely approximate the controlled object in future moments, predicting more accurate values and thus achieving optimal control of the controlled object. The objective function of a rolling optimization controller typically chooses a weighted quadratic optimization function of the prediction error and the change in control quantity as its performance function.
[0056] in, For reference trajectory; For feedback correction output; Here, represents the weighting coefficients for the control input; k is the output time, i represents the step number before output time k, and u is the control signal. Feedback correction is introduced into the prediction model, typically in the following form:
[0057] y(k) is the actual output of the controlled object at time k, h(i) is the error correction coefficient at the i-th prediction step, and e is the control output error.
[0058] By calculating the extrema of the objective function of the rolling optimization controller, the solution is obtained through inverse problem-solving. At this point, all controller parameters and the control model of the predictive control model optimized based on the SVR method have been obtained, and the corresponding control output effect can be obtained by changing the output variable.
[0059] Example 1: Preparation method of oxy-hydroxyapatite nanomaterial-containing chondroitin sulfate / carboxymethyl chitosan hydrogel: 1) Materials and reagents Chondroitin sulfate (OCS, containing aldehyde group); Carboxymethyl chitosan (CMCS); Hydroxyapatite nanomaterials (nHA); Solvent: Deionized water or PBS buffer; pH adjuster: acid / base adjusting solution (used to adjust the reaction system to a suitable pH).
[0060] In particular, the crosslinking of OCS / CMCS is achieved by forming dynamic Schiff base bonds between aldehyde groups and amino groups, without the need for additional crosslinking agents.
[0061] 2) Preparation of OCS solution
[0062] Dissolve OCS in PBS buffer to prepare an OCS solution of a specific concentration with a pH of 7.4. Set aside for later use.
[0063] 3) Preparation of CMCS solution
[0064] Dissolve CMCS in deionized water to prepare a CMCS solution of a certain concentration; pH 7. Set aside for later use.
[0065] 4) Predispersion of nHA
[0066] nHA is added to a solvent to form an nHA dispersion; after ultrasonic dispersion for a period of time, it is allowed to stand, and then ultrasonically dispersed again to obtain a stable nHA dispersion.
[0067] 5) Mixing and in-situ gelation of composite systems
[0068] The nHA dispersion obtained in step 4) is mixed with the CMCS solution to allow nHA to uniformly enter the CMCS matrix solution. Then, the OCS solution is added to the above mixture and thoroughly mixed at a set temperature to form a precursor solution. As the aldehyde groups of OCS and the amino groups of CMCS undergo cross-linking reactions, the system gradually forms a three-dimensional network structure to obtain the OCS / CMCS / nHA composite hydrogel.
[0069] The gelation time and degree of cross-linking can be controlled by adjusting the OCS / CMCS ratio, solid content, pH, temperature, and mixing method.
[0070] For bone repair applications, formulations with injectability and in-situ gelation properties are preferred.
[0071] 6) Molding and Post-processing
[0072] Choose the molding method based on the application scenario: In vitro molding: The precursor solution is injected into the mold and molded, and removed after complete gelation; In-situ gelation: The precursor solution is directly injected into the defective area or target container for in-situ gelation.
[0073] 7) Sample characterization (for subsequent performance data acquisition and model training)
[0074] The properties of the prepared hydrogel samples are characterized, including at least the following: Mechanical properties: elastic modulus, compressive modulus, etc.; Structure and dispersion state (optional but recommended): Observe whether nHA is uniformly dispersed and whether there is obvious aggregation; Osteogenic properties (e.g., for use in cell experiments): cell adhesion, ALP activity, mineralization-related indicators, etc. Mineralization capacity: Indicators related to in vitro mineralization capacity.
[0075] Example 2: Sample construction scheme for training neural networks: Multiple hydrogel samples prepared under different oxidation, substitution and ratio conditions were collected in the experiment. The elastic modulus, compressive modulus and osteogenic properties of the hydrogel samples were tested. The osteogenic properties included osteoblast adhesion rate and alkaline phosphatase (ALP) activity.
[0076] The aforementioned material composition parameters, preparation process parameters, and performance test results were input into the neural network model, and the data were normalized and encoded. A multi-layer feedforward neural network model containing an input layer, hidden layers, and an output layer was constructed, and the model was trained using the backpropagation algorithm until the model's prediction error converged.
[0077] After training, the parameters of the OCS / CMCS hydrogel material to be designed are input into the neural network model, which outputs the corresponding predicted mechanical properties and osteogenic properties. The predicted results show good consistency with the actual experimental test results, verifying the feasibility of the method of this invention in predicting the performance of bone repair hydrogels. The results show that the model can distinguish the influence trends of different parameter combinations on bone repair performance, providing guidance for the optimization of hydrogel parameters.
[0078] S6: Obtain the medical scenario bias factors corresponding to different clinical scenarios, and perform a weighted fusion evaluation to calculate the comprehensive repair efficacy score by combining the bone repair performance prediction results; The specific steps are as follows: S61. Based on the system's preset evaluation rules, extract the predicted value of elastic modulus representing mechanical strength, the osteoblast attachment rate representing biological characteristics, and the three-dimensional network degradation cycle representing the consumption progress from the bone repair performance prediction results. S62. Match the specific stress state and tissue healing time window characteristics of the current implantation environment, and obtain the mechanical bias factor, osteogenic bias factor and degradation bias factor corresponding to mechanical, biological and degradation properties respectively, which together constitute the medical scenario bias factor. S63. Using mechanical bias factor, osteogenic bias factor and degradation bias factor, perform weighted product operation on the predicted value of elastic modulus, osteoblast attachment rate and three-dimensional network degradation cycle respectively, and summarize all weighted product calculation results to obtain the overall numerical form of comprehensive repair efficacy score.
[0079] The CPU obtains medical scenario bias factors corresponding to different clinical scenarios. It retrieves the bone repair performance prediction matrix from the system memory via the data bus. The data extraction module scans the matrix based on a preset 16-bit field identifier, extracting the predicted elastic modulus (45.5 MPa) representing mechanical strength, the osteoblast attachment rate (88.2%) representing biological characteristics, and the three-dimensional network degradation cycle (120 days) representing the consumption progress. The CPU receives implantation environment tags input from the clinical terminal. When the implantation environment tag is identified as a load-bearing bone defect scenario, a scenario matching command is triggered. The CPU retrieves the preset bias factor mapping table within the system, extracting the mechanical bias factor (0.5) under specific stress conditions of the load-bearing bone, the osteogenic bias factor (0.3) corresponding to the tissue healing time window, and the degradation bias factor (0.2). These three specific values together constitute the medical scenario bias factor prediction matrix. The bias factor matrix for the treatment scenario is processed by a computation module that receives the predicted elastic modulus, osteoblast attachment rate, and three-dimensional network degradation cycle. The module then uses a division operator to divide these values by preset physical dimension benchmarks of 100 MPa, 100%, and 180 days, respectively, yielding dimensionless mechanical normalization parameters of 0.455, biological normalization parameters of 0.882, and degradation normalization parameters of 0.667. The computation module then performs point-to-point multiplication of the extracted mechanical bias factor, osteogenic bias factor, and degradation bias factor with their corresponding mechanical, biological, and degradation normalization parameters, resulting in individual weighted values of 0.2275, 0.2646, and 0.1334, respectively. The summation unit then sums these three individual weighted values, summing them to obtain an overall numerical value of 0.6255. Finally, a weighted fusion evaluation is performed based on the bone repair performance prediction results to calculate the comprehensive repair efficacy score.
[0080] S7: When the comprehensive repair efficiency score does not meet the preset evaluation criteria, calculate the parameter variation matrix of the current model, and perform reverse feedback adjustment on the parameters of the hydrogel material to be predicted based on the extracted parameter variation length to generate the optimal component parameter ratio. The specific steps are as follows: S71. When the comprehensive repair efficacy score fails to meet the implantation standard, perform multivariate partial derivative calculation on the hydrogel material parameters to be predicted in the calculation to obtain the parameter derivative matrix that reflects the parameter bias characteristics. S72. Scan and extract the extreme nodes in the parameter derivative matrix that represent the weights of performance influence, and calculate the parameter derivative length corresponding to different material compositions by combining the parameter adjustment range limit. S73. Using the variable asynchronous length of parameters as the adjustment benchmark, perform multiple rounds of inverse numerical migration and model verification on the current parameters of the hydrogel material to be predicted, record the proportion data that meets the qualification requirements, and generate the final optimal component parameter proportion.
[0081] When the comprehensive repair efficacy score fails to meet the preset evaluation criteria, the parameter derivative matrix of the current model is calculated. The comparator module receives the comprehensive repair efficacy score of 0.6255 generated in the previous calculation stage and compares it with the preset implantation evaluation standard threshold of 0.75 in the memory. When the current score is determined to be less than 0.75, the start command of the feedback adjustment module is triggered. The derivative calculator performs partial derivative calculations with respect to the objective function of the comprehensive repair efficacy score for all parameters of the hydrogel material to be predicted participating in the forward propagation calculation in the current neural network model, generating a 1-row, 5-column parameter derivative matrix with gradient information of each material parameter. The central processing unit scans the parameter derivative matrix and extracts the node data with an absolute value greater than the weight recognition threshold of 0.1 as extreme value nodes. Specifically, the partial derivative value representing the amount of hydroxyapatite nanomaterial added is 0.45, and the partial derivative value representing the concentration of chondroitin sulfate is -0.25. The central processing unit retrieves the parameter adjustment range limit for each material component, which specifies that the hydroxyapatite nanomaterial... The maximum single change in the material is 3%. The processor multiplies the extracted partial derivative value by the preset base learning rate constant 0.05 to calculate the parameter variation length of hydroxyapatite addition as 0.0225 and the parameter variation length of chondroitin sulfate concentration as -0.0125. The numerical regulator uses the calculated parameter variation length as the adjustment benchmark, adds the corresponding variation length to the current initial hydroxyapatite addition of 10% and the initial chondroitin sulfate concentration of 5% to perform the first round of reverse numerical offset, generating an updated parameter set. The central processing unit re-inputs the updated parameter set into the neural network model to perform the second round of prediction, repeating the above partial derivative calculation and parameter offset process until the comprehensive repair efficacy score output by multiple rounds of verification reaches 0.761. This score is greater than the evaluation standard threshold of 0.75. The data recording unit writes the ratio data that meets the threshold condition into the qualified list area of the system memory. Based on the extracted parameter variation length, the reverse feedback adjustment is performed on the parameters of the hydrogel material to be predicted to generate the optimal component parameter ratio.
[0082] Table 1. Data table of parameter reverse feedback control process
[0083] As shown in Table 1, this table lists the measured data of the inverse numerical offset of the system execution parameters and the model verification process. It records in detail the adjustment range of the hydrogel material ratio parameters and the corresponding dynamic growth of the comprehensive repair efficiency score in each iteration.
[0084] Example 3: Parameter optimization design example of bone repair hydrogel based on prediction results: 1) Purpose of the Implementation Example Based on Examples 1 and 2, the bone repair performance of hydroxyapatite nanomaterials containing chondroitin sulfate / carboxymethyl chitosan hydrogels under different parameter combinations was predicted using a trained neural network model. Based on the prediction results, the parameters of the hydrogel materials were optimized to obtain a hydrogel formulation with excellent bone repair performance.
[0085] 2) Setting the parameters to be optimized
[0086] The following parameters are selected as optimization targets: The ratio parameters of chondroitin sulfate and carboxymethyl chitosan; Parameters for the amount of hydroxyapatite nanomaterials added; The pH value or reaction time parameter in the preparation process parameters.
[0087] By constructing multiple sets of parameter combinations to be evaluated within a reasonable range of the above parameters, candidate hydrogel design schemes are formed.
[0088] 3) Performance prediction based on neural network models
[0089] The candidate parameter combinations are input into the trained neural network model to obtain the bone repair performance prediction results corresponding to each parameter combination.
[0090] The bone repair performance prediction results include one or more of the following indicators: Mechanical property prediction results; Osteogenesis-related performance prediction results; Mineralization capacity prediction results.
[0091] By comparing and analyzing the prediction results under different parameter combinations, parameter combinations that perform better in multiple bone repair performance indicators were selected.
[0092] 4) Parameter optimization and selection of the best solution
[0093] Based on the prediction results, the candidate parameter combinations are sorted or graded, and the parameter combination with the best overall performance in terms of mechanical properties, osteogenic properties and mineralization capacity is selected as the preferred design scheme.
[0094] The preferred solution does not only pursue the maximization of a single performance index, but also achieves synergistic optimization among multiple performance aspects related to bone repair.
[0095] 5) Verification of the optimization scheme
[0096] The selected optimal parameter combination was used to prepare oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel samples containing hydroxyapatite nanomaterials, and its bone repair-related properties were experimentally tested.
[0097] The experimental results were compared with the predictions of the neural network model to verify the effectiveness of the model in guiding the optimization design of hydrogel parameters.
[0098] 6) Closed-loop optimization design process
[0099] Through the above steps, the following closed-loop optimization design process is formed: 1. Construct multi-parameter candidate design schemes; 2. Predicting bone repair performance based on neural network models; 3. Select the optimal parameter combination; 4. Prepare and validate the optimized hydrogel samples; 5. The validation results will be fed back to further improve the training dataset and model.
[0100] This closed-loop design process allows for the continuous optimization of the bone repair properties of oxy-hydroxyapatite nanomaterial-containing chondroitin sulfate / carboxymethyl chitosan hydrogel.
[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting the bone repair performance of oxidized chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials, characterized by, The method includes the following steps: S1. Data Acquisition: Collect experimental data, which includes hydrogel material parameters, preparation process parameters and bone repair-related performance data of chondroitin sulfate, carboxymethyl chitosan and hydroxyapatite nanomaterials. S2. Data preprocessing: The experimental data is preprocessed, including normalization and encoding. S3. Model Building: Construct a neural network model based on the preprocessed data; S4. Model Training: Train the neural network model using the preprocessed data; S5. Performance Prediction and Application: Input the parameters of the hydrogel material to be predicted and the preparation process parameters into the trained neural network model, and output the prediction results of the bone repair performance of the hydrogel. S6. Obtain the medical scenario bias factors corresponding to different clinical scenarios, and perform a weighted fusion evaluation calculation based on the bone repair performance prediction results to obtain a comprehensive repair efficacy score. S7. When the comprehensive repair efficiency score does not meet the preset evaluation criteria, calculate the parameter variation matrix of the current model, and perform reverse feedback adjustment on the parameters of the hydrogel material to be predicted based on the extracted parameter variation length to generate the optimal component parameter ratio.
2. The method for predicting the bone repair performance of hydroxyapatite- containing nanomaterials of oxidized chondroitin sulfate / carboxymethyl chitosan hydrogel according to claim 1, characterized in that, In S2, the data preprocessing further includes: Data cleaning and dataset partitioning: Data cleaning involves handling missing values, outliers, and obviously erroneous data; dataset partitioning involves randomly dividing the preprocessed dataset into training, validation, and test sets. In S4, the constructed neural network model is trained using training set data, with material parameters and preparation process parameters as inputs and corresponding bone repair performance data as the expected output. The network weights and biases are optimized through backpropagation algorithm, and validation set data is used to monitor model performance and prevent overfitting.
3. The method for predicting the bone repair performance of hydroxyapatite- containing nanomaterials of oxidized chondroitin sulfate / carboxymethyl chitosan hydrogel according to claim 1, characterized in that, The parameters of the hydrogel material include: The oxidation degree of chondroitin sulfate, the substitution degree of carboxymethyl chitosan, the ratio parameters of chondroitin sulfate to carboxymethyl chitosan, the amount of hydroxyapatite nanomaterials added, the particle size range and dispersion mode of hydroxyapatite nanomaterials. The neural network model is used to establish a nonlinear mapping relationship between hydrogel material parameters and bone repair performance; The oxidation degree of the chondroitin sulfate is 40-80%, the substitution degree of the carboxymethyl chitosan is 65-90%, the concentration range of the chondroitin sulfate is 1-10%, the concentration range of the carboxymethyl chitosan is 0.5-5%, the mass concentration ratio of the chondroitin sulfate to the carboxymethyl chitosan is 1:(0.1-10), and the amount of hydroxyapatite nanomaterial added is 5-15%. The hydroxyapatite nanomaterials have a particle size range of 10-200 nm and are dispersed by one of the following methods: ultrasonic physical dispersion, surface modification dispersion, and in-situ acoustic field dispersion.
4. The method for predicting the bone repair performance of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials according to claim 1, characterized in that, The preparation process parameters include: One or more of the following parameters: reaction environment parameters, crosslinking condition parameters, dispersion condition parameters, post-treatment parameters, and hydrogel system state parameters.
5. The method for predicting the bone repair performance of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials according to claim 4, characterized in that, The reaction environment parameters include one or more of the following: reaction temperature, reaction time, and pH value of the reaction system. Crosslinking condition parameters include one or more of the following parameters: crosslinking mode and degree of crosslinking during hydrogel formation, and mixing order and mixing method of components. Dispersion condition parameters include one or more of the following: dispersion mode, time, and intensity; The state parameters of a hydrogel system include one or more of the following: concentration, solid content, and viscosity-related parameters.
6. The method for predicting the bone repair performance of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials according to claim 1, characterized in that, The bone repair-related performance parameters include one or more of the following: mechanical property data, osteogenic performance data, mineralization capacity, biocompatibility-related parameters, and degradation and stability-related parameters.
7. The method for predicting the bone repair performance of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials according to claim 6, characterized in that, The mechanical performance parameters include one or more of the elastic modulus and the compressive modulus; Osteogenic performance parameters include osteoblast adhesion ability and osteogenic differentiation-related indicators; Mineralization capacity parameters are used to characterize the mineralization behavior of hydrogels during bone repair; biocompatibility parameters include cell viability and cell proliferation capacity. Degradation and stability-related parameters include degradation rate and changes in structural stability.
8. The method for predicting the bone repair performance of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials according to claim 1, characterized in that, The neural network model includes an input layer, at least one hidden layer, and an output layer; The number of neurons in the input layer corresponds to the number of preprocessed material parameters and process parameters; The number of neurons in the output layer corresponds to the number of bone repair-related performance indicators to be predicted.
9. The method for predicting the bone repair performance of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials according to claim 1, characterized in that, The specific steps of S6 are as follows: S61. Based on the system's preset evaluation rules, extract the predicted value of elastic modulus representing mechanical strength, the osteoblast attachment rate representing biological characteristics, and the three-dimensional network degradation cycle representing the consumption progress from the bone repair performance prediction results. S62. Match the specific stress state and tissue healing time window characteristics of the current implantation environment, and obtain the mechanical bias factor, osteogenic bias factor and degradation bias factor corresponding to mechanical, biological and degradation properties respectively, which together constitute the medical scenario bias factor. S63. Using the mechanical bias factor, the osteogenic bias factor, and the degradation bias factor, perform weighted product operations on the predicted elastic modulus value, the osteoblast attachment rate, and the three-dimensional network degradation cycle, respectively, and summarize all the weighted product calculation results to obtain the overall numerical form of the comprehensive repair efficacy score.
10. The method for predicting the bone repair performance of oxychloride chondroitin sulfate / carboxymethyl chitosan hydrogel containing hydroxyapatite nanomaterials according to claim 1, characterized in that, The specific steps in S7 are as follows: S71. When it is determined that the comprehensive repair efficacy score does not meet the implantation standard, perform multivariate partial derivative calculation on the hydrogel material parameters to be predicted that are involved in the calculation to obtain the parameter derivative matrix that reflects the parameter bias characteristics. S72. Scan and extract the extreme value nodes in the parameter derivative matrix that represent the weight of performance influence, and calculate the parameter derivative length corresponding to different material compositions in combination with the parameter adjustment range limit. S73. Using the parameter variation and asynchrony length as an adjustment benchmark, perform multiple rounds of inverse numerical offset and model verification on the current parameters of the hydrogel material to be predicted, record the proportion data that meets the qualification requirements, and generate the final optimal component parameter proportion.