A machine learning assisted optimized multifunctional composite hydrogel wound dressing and its preparation and optimization method

Abstract

The application relates to a machine learning assisted optimization multifunctional composite hydrogel wound dressing and a preparation and optimization method thereof, and belongs to the hydrogel field. The hydrogel is constructed by ion crosslinking of sodium alginate and calcium ions and chitosan quaternary ammonium salt, and a chemical crosslinking network formed by polyacrylamide is interpenetrated with the network. The hydrogel has high water absorption, strong antibacterial property, reliable adhesion, sensitive strain sensing property and excellent mechanical property, and provides ideal healing for wounds. The method innovatively integrates machine learning intelligent optimization: by establishing a formula-process-performance database, training a Gaussian process regression prediction model, and adopting a Bayesian optimization algorithm for global optimization, optimal preparation parameters for specific performance requirements are efficiently and accurately determined. The method breaks through the traditional trial-and-error development mode depending on experience, realizes high performance, customization and rapid development of the hydrogel dressing, and provides a new paradigm for design and preparation of a new generation of wound care products.

Description

Technical Field

[0001] This invention relates to the field of hydrogel preparation, and more particularly to a multifunctional composite hydrogel wound dressing optimized with machine learning assistance, and its preparation and optimization methods. Background Technology

[0002] Sodium alginate (SA) is a natural anionic polysaccharide with carboxyl groups in its molecular chain. It can form hydrogels through mild "egg-box" ionic crosslinking with divalent cations, exhibiting excellent hydrophilicity, biocompatibility, and moisture absorption and retention capabilities, making it an ideal substrate for wound dressings. Chitosan quaternary ammonium salt (CQAS), as a quaternized derivative of chitosan, not only retains good biocompatibility and biodegradability, but its enhanced positive charge and water solubility also endow the material with highly efficient and broad-spectrum antibacterial activity. Combining anionic SA with cationic CQAS allows them to form a stable polyelectrolyte complex through electrostatic interactions, synergistically integrating SA's excellent moisture management capabilities with CQAS's powerful antibacterial and healing-promoting functions, laying a solid material foundation for constructing multifunctional integrated wound dressings.

[0003] Despite the promising prospects of the SA / CQAS composite system, hydrogels composed of natural polymers generally suffer from inherent defects such as insufficient mechanical properties and limited functionality, making it difficult to meet the clinical needs of dynamic wounds for dressing mechanical strength and multifunctional integration. More importantly, the final performance of composite hydrogels is profoundly affected by the complex nonlinear coupling between parameters such as the concentration of various components and the degree of crosslinking. The traditional experience-based "trial and error" R&D model is inefficient and costly, and it is almost impossible to efficiently and accurately find the optimal performance formulation for specific wound needs in the vast high-dimensional parameter space. This seriously restricts the development process of high-performance, customizable dressings. Summary of the Invention

[0004] The purpose of this invention is to solve the above-mentioned problems in the prior art and provide a multifunctional composite hydrogel wound dressing with machine learning-assisted optimization and its preparation and optimization method. By applying Bayesian optimization to the formulation development of multi-component composite hydrogels, the research and development mode is transformed from "experience-driven and labor-intensive" trial and error to "data-driven and intelligently guided" precise design. This allows for the systematic exploration of a broad parameter space with very few experimental iterations, significantly shortening the research and development cycle, reducing costs, and potentially discovering high-performance formulations that surpass traditional understanding, thus achieving efficient and customized material development.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A multifunctional composite hydrogel wound dressing, wherein the dressing is a hydrogel with an interpenetrating network structure, the network structure comprising:

[0007] The first physical cross-linking network is formed by sodium alginate and divalent metal ions through ionic cross-linking, and jointly constructed with chitosan quaternary ammonium salt through electrostatic interaction.

[0008] And a second chemical crosslinking network formed by covalent crosslinking of polyacrylamide;

[0009] The first physical cross-linking network and the second chemical cross-linking network are interconnected.

[0010] The hydrogel has a three-dimensional porous structure.

[0011] The divalent metal ion is a calcium ion.

[0012] The hydrogel possesses the following properties: 1) excellent tensile toughness, with a tensile strength of not less than 100 kPa and a fracture strain of not less than 500%; 2) strain sensing characteristics based on ion conductivity; 3) reliable tissue adhesion; 4) highly efficient antibacterial activity against common wound pathogens such as Staphylococcus aureus or Escherichia coli, with an antibacterial rate of not less than 94%.

[0013] A method for preparing the multifunctional composite hydrogel wound dressing includes the following steps:

[0014] 1) Preparation of precursor solution: containing sodium alginate, divalent metal ion source, chitosan quaternary ammonium salt, acrylamide monomer, crosslinking agent and initiator;

[0015] 2) Preparation of cross-linking accelerator solution: including slow-release agent and initiator;

[0016] 3) The crosslinking accelerator solution is added dropwise to the precursor solution, and the reaction is carried out under ultraviolet light irradiation, so that the acrylamide monomer polymerizes to form the second chemical crosslinking network of polyacrylamide. At the same time, the sodium alginate undergoes ionic crosslinking under the action of divalent metal ions, and combines with the chitosan quaternary ammonium salt through electrostatic interaction to form the first physical crosslinking network, thereby obtaining the composite hydrogel.

[0017] The concentration of sodium alginate is 0.5 wt.% to 2 wt.% of the total weight of the precursor solution, the concentration of chitosan quaternary ammonium salt is 0.3 wt.% to 1.5 wt.% of the total weight of the precursor solution, and the concentration of acrylamide monomer is 5 wt.% to 47 wt.% of the total weight of the precursor solution.

[0018] The ultraviolet light irradiation has an irradiation power of 200~300W and an irradiation duration of 8~15 minutes, preferably an irradiation power of 250W and 10 minutes.

[0019] A machine learning-assisted optimization method for optimizing process parameters in the above method includes the following steps:

[0020] S1: Construct a training dataset, which includes multiple sets of input parameters and corresponding output performance indicators. The input parameters include at least the concentration of each component, light intensity, and reaction time. The output performance indicators include at least the mechanical properties, sensing properties, adhesion properties, and antibacterial properties of the hydrogel.

[0021] S2: Use the training dataset to train a machine learning model to establish a predictive model from input parameters to output performance metrics;

[0022] S3: Set the objective function and use the optimization algorithm to call the prediction model to perform iterative search to obtain the optimal combination of input parameters for the objective function, which will be used as the optimized process parameters.

[0023] S4: Prepare the multifunctional composite hydrogel wound dressing according to the optimized process parameters.

[0024] The machine learning model is a Bayesian optimization model based on Gaussian process regression (GPR).

[0025] The acquisition function of the optimization algorithm is at least one of the expected improvement (EI), maximum improvement probability (MPI), and lower confidence limit (LCB) functions.

[0026] The objective function is a comprehensive scoring function obtained by normalizing and weighting multiple output performance indicators.

[0027] In this invention, the first physical cross-linked network and the second chemical cross-linked network interpenetrate each other at the microscopic level, forming an interpenetrating polymer network structure. This is not a simple physical blend, but rather two independent and continuous networks intertwined and entangled in three-dimensional space, without covalent bonds connecting them, yet forming a single, complete material at the macroscopic level. This structure is typically prepared using a sequential method, first forming an ionic cross-link of sodium alginate-chitosan quaternary ammonium salt as the first network, followed by acrylamide monomers penetrating into its pores, polymerizing in situ and covalently cross-linking to form a second chemical network penetrating the interior of the physical network, thereby achieving spatial interlocking.

[0028] The fundamental advantage of this structure lies in the synergistic enhancement effect of the networks. The dynamically reversible first physical network (relying on ionic bonds, hydrogen bonds, and electrostatic interactions) can preferentially break as "sacrificial bonds" under external forces, effectively dissipating energy; while the permanently cross-linked second chemical network acts as a rigid framework, maintaining the integrity and stability of the overall structure. This synergistic mechanism of "rigidity and flexibility" is the key to the composite hydrogel of this invention being able to overcome the shortcomings of the mechanical properties of single-network hydrogels and achieve high strength, high toughness, and multifunctional integration.

[0029] Compared with existing technologies, this invention achieves a dual breakthrough in product performance and development paradigm, with the following beneficial effects:

[0030] 1. This invention constructs a unique structure in which the physical network of sodium alginate-chitosan quaternary ammonium salt and the chemical network of polyacrylamide interpenetrate, resulting in a significant synergistic effect. This not only synergistically enhances mechanical properties but also integrates three major functions: reliable wet adhesion, efficient and rapid synergistic antibacterial effect, and sensitive strain sensing. Thus, intelligent wound management can be achieved using a single material.

[0031] 2. For multi-component, high-dimensional formulation spaces, this invention integrates a machine learning-driven Bayesian optimization method. This method, through active learning and iterative feedback, can efficiently lock in the globally optimal formulation with a minimal number of experiments, and performance can be customized to meet clinical needs by adjusting weights. This transforms research and development from an experience-based trial-and-error model to a data-driven intelligent model, offering advantages such as high efficiency, controllable costs, and ease of implementation. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating the hydrogel preparation process of the present invention;

[0033] Figure 2 Closed-loop flowchart for machine learning-assisted optimization of the overall performance of hydrogels;

[0034] Figure 3 The curve representing the change in the overall performance score Y during the optimization process;

[0035] Figure 4 The stress-strain curves, GF curves, and shear-adhesion curves before and after optimization were obtained.

[0036] Figure 5 Representative images and inhibition rates of Escherichia coli grown on different hydrogel samples (the subscripts of the names indicate the mass fraction of the corresponding substances). Detailed Implementation

[0037] To make the technical problems, technical solutions and beneficial effects of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. These embodiments should not be construed as limiting the scope of protection of the present invention.

[0038] Example 1: Preparation of PAM / SA-CQAS interpenetrating network hydrogel dressing

[0039] This embodiment illustrates the preparation method of hydrogels, demonstrating the process of simultaneously constructing a "physical-chemical" interpenetrating network through photo-initiated polymerization and ionic crosslinking. Figure 1 ).

[0040] 1. Preparation of precursor solution: Add 6514 mg of deionized water to a beaker, and while stirring continuously, add 60 mg of chitosan quaternary ammonium salt (CQAS), 300 mg of sodium chloride (NaCl), 1.1 mg of calcium ethylenediaminetetraacetate (EDTA-Ca), 1.0 mg of N,N'-methylenebisacrylamide (MBAA), 13.7 mg of ammonium persulfate (APS), and 100 mg of sodium alginate (SA). Continue stirring for 30 minutes until completely dissolved and mixed evenly. Then, slowly add 2000 mg of acrylamide (AM) to the solution, stirring until completely dissolved, and then degassed by sonication.

[0041] 2. Preparation of cross-linking accelerator solution: Add 1000 mg of deionized water to another beaker, and dissolve 1.1 mg of gluconolactone (GDL) and 9.2 μL of tetramethylethylenediamine (TEMD) in sequence, and stir well for later use.

[0042] 3. Photoinitiated polymerization and network formation: The crosslinking accelerator solution is slowly added dropwise to the precursor mixture and stirred until homogeneous. The final solution is cast into a polytetrafluoroethylene mold, covered with a 1 mm thick glass plate, and irradiated under a 250 W UV lamp for 10 min to form a PAM / SA-CQAS hydrogel.

[0043] Example 2: Intelligent Screening and Validation of Hydrogel Formulations Based on Bayesian Optimization

[0044] This embodiment demonstrates how to apply intelligent optimization methods to efficiently obtain PAM / SA-CQAS hydrogel formulations with excellent overall performance.

[0045] 1. Optimize the design of the scheme

[0046] To efficiently find the optimal solution within a high-dimensional parameter space composed of multiple component concentrations, this embodiment employs a Bayesian optimization framework based on Gaussian processes to construct a closed-loop system of "prediction-recommendation-verification" (see [link]). Figure 2 The specific steps include:

[0047] (1) Define the optimization objective: integrate multiple key performance indicators of hydrogel into a single comprehensive performance scoring function Y by normalization and weighted summation, and minimize the value of Y to be the optimization objective.

[0048] (2) Setting the parameter space: The nine key preparation parameters, including sodium alginate concentration, chitosan quaternary ammonium salt concentration, and acrylamide monomer concentration, and their feasible value ranges, are defined as the nine-dimensional optimization parameter space X.

[0049] Nine parameters and their ranges:

[0050]

[0051] (3) Iterative optimization process: A dataset {X, Y} is constructed through 40 initial experiments; a Gaussian process regression model is trained based on the dataset to predict the performance of the new formulation; then, collection functions such as EI, MPI and LCB are used to actively recommend the next most promising experimental point X_next; after completing the experimental verification, the new data pair {X_next, Y_next} is added to the dataset and the model is updated, and the process is iterated until the target converges.

[0052] 2. Optimization process and results analysis

[0053] After multiple iterations, the overall performance score Y showed a continuous downward trend with the increase of the number of iterations. Figure 3 This indicates that the optimization system can effectively guide the search direction and quickly approach the global optimal region.

[0054] The optimal formula selected from the final iteration results:

[0055]

[0056] Its overall performance score Y-value is significantly lower than all recipes in the initial dataset. Comparing the optimized best recipe with the best-performing initial recipe before optimization, its performance across all categories is comprehensively improved (corresponding to...). Figure 4 ):

[0057] This invention achieves a breakthrough in the mechanical properties of hydrogels by constructing an interpenetrating network structure. The introduction of flexible SA chains and the physical penetration of the PAM network to form hydrogen bonds synergistically enhance the strength and toughness of the material. Furthermore, the introduction of CQAS allows the dynamic ionic bonds to preferentially break as "sacrificial bonds" to dissipate energy, thereby achieving ultra-high toughness (fracture strain reaching 3228% and fracture energy increased to 3747 kJ / m³) while maintaining moderate strength (approximately 168.3 kPa), significantly enhancing the material's ductility and damage tolerance.

[0058] Sensing performance: Strain sensing sensitivity (GF value) has been comprehensively improved. It has been increased to 1.885 in the low strain region (0~50%) and to 5.758 in the high strain region (400%~500%). The optimized material exhibits a more significant resistive response under the same deformation, enhancing signal resolution.

[0059] Adhesion performance: Shear adhesion strength is improved to approximately 6.638 kPa. This is attributed to the algorithm's synergistic optimization of interfacial adhesive components (such as CQAS) and network cohesive strength parameters (such as AM concentration), enabling the material to achieve stronger and more reliable tissue adhesion in a wet state.

[0060] Antibacterial Properties: The hydrogel exhibits highly efficient and rapid antibacterial properties, reaching a peak inhibition rate of 94% within 120 minutes. This is mainly attributed to the fact that chitosan quaternary ammonium salt (CQAS), as a polycation, exerts its contact bactericidal effect by disrupting the bacterial cell membrane, while sodium alginate (SA) inhibits bacterial metabolism by chelating essential metal ions. Together, they constitute a "multi-target" antibacterial mechanism. See details... Figure 5 The results of evaluating the antibacterial properties of different hydrogel samples against *Escherichia coli* were presented. The control group consisted of 400 µL sterile PBS + 200 µL bacterial suspension, while the experimental group consisted of 400 µL hydrogel sample + 200 µL bacterial suspension. Specifically, sample AM... 20 The sample is a polyacrylamide hydrogel with an acrylamide monomer concentration of 20 wt.% (excluding sodium alginate and chitosan quaternary ammonium salt). 20 CQAS 0.6 For AM 20 Based on this, a hydrogel with a concentration of 0.6 wt.% chitosan quaternary ammonium salt (CQAS) was added; sample AM 20 SA1 is in AM 20 Based on this, a hydrogel with a concentration of 1 wt.% sodium alginate (SA) was added; sample AM 20 SA1CQAS 0.6 For AM 20 Based on this, a hydrogel containing 1 wt.% sodium alginate and 0.6 wt.% chitosan quaternary ammonium salt was added to form the composite hydrogel described in this invention, which was used to verify the antibacterial performance of SA and CQAS under synergistic effects.

[0061] This embodiment demonstrates through specific data that the Bayesian optimization-driven method can quickly and purposefully discover the optimal hydrogel formulation with significantly improved overall performance with very few experiments, verifying the efficiency and practicality of this intelligent design strategy and providing an efficient tool for the performance-customized development of wound dressings.

[0062] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multifunctional composite hydrogel wound dressing, characterized in that, The dressing is a hydrogel with an interpenetrating network structure, the network structure of which includes: The first physical cross-linking network is formed by sodium alginate and divalent metal ions through ionic cross-linking, and jointly constructed with chitosan quaternary ammonium salt through electrostatic interaction. And a second chemical crosslinking network formed by covalent crosslinking of polyacrylamide; The first physical cross-linking network and the second chemical cross-linking network are interconnected.

2. The multifunctional composite hydrogel wound dressing as described in claim 1, characterized in that: The divalent metal ion is a calcium ion.

3. The multifunctional composite hydrogel wound dressing as described in claim 1, characterized in that, The hydrogel possesses the following properties: tensile strength of not less than 100 kPa, fracture strain of not less than 500%, antibacterial rate against Staphylococcus aureus or Escherichia coli of not less than 94%, and strain sensing and tissue adhesion properties.

4. A method for preparing a multifunctional composite hydrogel wound dressing as described in any one of claims 1 to 3, characterized in that, Includes the following steps: 1) Preparation of precursor solution: containing sodium alginate, divalent metal ion source, chitosan quaternary ammonium salt, acrylamide monomer, crosslinking agent and initiator; 2) Preparation of cross-linking accelerator solution: including slow-release agent and initiator; 3) The crosslinking accelerator solution is added dropwise to the precursor solution, and the reaction is carried out under ultraviolet light irradiation, so that the acrylamide monomer polymerizes to form the second chemical crosslinking network of polyacrylamide. At the same time, the sodium alginate undergoes ionic crosslinking under the action of divalent metal ions, and combines with the chitosan quaternary ammonium salt through electrostatic interaction to form the first physical crosslinking network, thereby obtaining the composite hydrogel.

5. The method as described in claim 4, characterized in that: The concentration of sodium alginate is 0.5 wt.% to 2 wt.% of the total weight of the precursor solution, the concentration of chitosan quaternary ammonium salt is 0.3 wt.% to 1.5 wt.% of the total weight of the precursor solution, and the concentration of acrylamide monomer is 5 wt.% to 47 wt.% of the total weight of the precursor solution.

6. A machine learning-assisted optimization method for optimizing process parameters in the method of claim 4, characterized in that, Includes the following steps: S1: Construct a training dataset, which includes multiple sets of input parameters and corresponding output performance indicators. The input parameters include at least the concentration of each component, and the output performance indicators include at least the mechanical properties, sensing properties, adhesion properties, and antibacterial properties of the hydrogel. S2: Use the training dataset to train a machine learning model to establish a predictive model from input parameters to output performance metrics; S3: Set the objective function and use the optimization algorithm to call the prediction model to perform iterative search to obtain the optimal combination of input parameters for the objective function, which will be used as the optimized process parameters.

7. The machine learning-assisted optimization method as described in claim 6, characterized in that: The machine learning model is a Bayesian optimization model based on Gaussian process regression.

8. The machine learning-assisted optimization method as described in claim 6, characterized in that: The acquisition function of the optimization algorithm is at least one of the expected improvement, maximum improvement probability, and confidence lower limit function.

9. The machine learning-assisted optimization method as described in claim 6, characterized in that: The objective function is a comprehensive scoring function obtained by normalizing and weighting multiple output performance indicators.

Images