Wound flora monitoring method and system

Abstract

The invention relates to the technical field of wound nursing, and discloses a wound flora monitoring method and system. The method comprises the following steps: acquiring wound physiological characteristic parameters, real-time flora monitoring data and multispectral imaging data, and constructing a three-dimensional flora distribution model and an initial flora dynamic model after preprocessing; applying metabolism and environmental conditions to obtain a comprehensive model and flora dynamic coupling data; training a flora classification model to obtain abnormal characteristic parameters; simulating abnormal evolution of flora to obtain information such as diffusion trend; constructing an optimization model to determine the optimal configuration of the monitoring device; and adjusting a monitoring strategy based on the risk grading model. The system comprises a data acquisition module, a model construction and analysis module, a flora classification module, a simulation analysis module, a configuration optimization module and a risk regulation module. Through multispectral imaging, dynamic modeling and an intelligent algorithm, accurate monitoring and risk regulation and control of wound flora are achieved, infection early warning timeliness and monitoring efficiency are improved, and the method is suitable for clinical wound nursing.

Description

Technical Field

[0001] The present invention relates to the technical field of wound care, and in particular to a method and system for monitoring wound flora. Background Art

[0002] In the field of wound care, microbiome monitoring and infection prevention and control are key links that affect healing outcomes. Traditional wound microbiome monitoring methods have significant limitations and are unable to meet the needs of modern precision medicine.

[0003] From the perspective of detection methods, traditional bacterial culture methods are time-consuming (usually 48-72 hours), cannot achieve real-time monitoring, and rely on manual operations, which are easily interfered by contamination, resulting in delayed results and insufficient accuracy. For example, for complex wounds such as diabetic foot, if abnormal proliferation of the bacterial flora is not discovered in time, the best time for intervention may be missed, increasing the risk of amputation. Detection technology based on single spectral imaging can only obtain two-dimensional planar information and cannot fully present the distribution characteristics of the bacterial flora in the three-dimensional space of the wound, such as the difference in metabolic activity of the bacterial flora in the deep infection area. This makes it difficult for doctors to accurately judge the scope and severity of the infection.

[0004] In terms of monitoring strategies, traditional systems lack dynamic optimization capabilities. The sampling interval, spectral resolution and other parameters of existing devices are usually fixed and cannot be adaptively adjusted according to the dynamic changes of the microbial community. For example, when the wound is in the inflammatory stage, the microbial community metabolism is active and requires high-frequency, high-resolution monitoring, but fixed parameters may lead to data redundancy or omission of key information, which increases monitoring costs and reduces early warning efficiency. In addition, traditional methods do not combine physiological characteristic parameters (such as wound temperature and exudate pH) with microbial community data for analysis, ignore the interaction between the host's physiological state and the microbial community ecology, and make it difficult to establish a comprehensive infection risk assessment model.

[0005] From the perspective of risk prevention and control, existing technologies lack the ability to simulate the abnormal evolution of bacterial flora in a forward-looking manner. Most systems can only realize bacterial flora classification and current status warning, and cannot predict dynamic processes such as bacterial flora diffusion trends and changes in metabolic activity. For example, when a certain type of high-risk bacterial flora (such as methicillin-resistant Staphylococcus aureus) appears, traditional methods cannot quickly simulate its diffusion path and speed under different environmental conditions, resulting in lagging prevention and control strategies and the possibility of causing systemic infection. At the same time, the lack of a risk grading model makes the adjustment of monitoring strategies lack a scientific basis, making it difficult to achieve the optimal balance between monitoring accuracy and cost.

[0006] With the advent of an aging society, the number of patients with chronic wounds (such as pressure ulcers and venous ulcers) has increased sharply, and the deficiencies of traditional monitoring technologies have become increasingly prominent. In the context of precision medicine, there is an urgent need for a wound flora monitoring technology that integrates multi-dimensional data and has the capabilities of dynamic modeling and intelligent optimization to achieve early warning of infection risks, adaptive adjustment of monitoring plans, and precise assessment of treatment effects. It is precisely in such a technical background that this invention breaks through the bottleneck of traditional technologies through the combination of multi-spectral imaging, three-dimensional modeling, dynamic simulation, and intelligent optimization algorithms, providing a more efficient and precise solution for wound care. Summary of the Invention

[0007] The purpose of the present invention is to provide a method and system for monitoring wound flora to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solution: A method for monitoring wound flora,

[0009] including the following steps:

[0010] Obtain the physiological characteristic parameters of the wound and real-time flora monitoring data;

[0011] Obtain the multi-spectral imaging data of the wound and perform preprocessing, then construct a three-dimensional flora distribution model, perform spectral feature analysis on the three-dimensional flora distribution model and combine the physiological characteristic parameters to obtain the initial flora dynamic model of the wound;

[0012] Apply local metabolic conditions and external environmental conditions to the initial flora dynamic model to obtain a comprehensive flora dynamic model, and combine the physiological characteristic parameters to obtain flora dynamic coupling data;

[0013] Construct a flora classification pre-training model and train it with the flora dynamic coupling data to obtain a flora classification model, and then obtain flora abnormal characteristic parameters;

[0014] Through the flora abnormal characteristic parameters, the comprehensive flora dynamic model and the physiological characteristic parameters, obtain flora abnormal evolution simulation information;

[0015] Based on the flora abnormal evolution simulation information and the monitoring device parameters, construct an optimization model and optimize the monitoring device parameters to obtain the optimal monitoring configuration of the monitoring device;

[0016] Based on the risk grading model, infer the current period monitoring configuration to obtain the current period risk grading prediction value. Based on the current period risk grading prediction value, the current period optimal monitoring configuration and the current period monitoring configuration, adjust the current period actual monitoring strategy to achieve the monitoring of the wound flora state.

[0017] Preferably, the dynamic coupling data of the bacterial community at least includes the bacterial community density distribution field, the metabolic activity field and the high-risk bacterial community location set; the abnormal bacterial community characteristic parameters at least include the abnormal bacterial community location, abnormal bacterial community type and abnormal bacterial community morphological parameters; the abnormal bacterial community evolution simulation information at least includes the bacterial community diffusion trend, maximum metabolic activity and maximum density change.

[0018] Preferably, the method of obtaining multispectral imaging data of the wound and performing preprocessing to construct a three-dimensional flora distribution model, performing spectral feature analysis on the three-dimensional flora distribution model and combining it with physiological characteristic parameters to obtain an initial flora dynamic model of the wound comprises the following steps:

[0019] Perform multi-band scanning on the wound surface to obtain a multi-spectral imaging data sequence of the wound in different bands;

[0020] Preprocessing the multispectral imaging data sequence to obtain a preprocessed multispectral imaging data sequence, wherein the preprocessing includes one or more of denoising, standardization, band fusion, abnormal signal removal, data interpolation, and spectral smoothing;

[0021] Based on the preprocessed multispectral imaging data sequence and combined with three-dimensional reconstruction technology, the corresponding three-dimensional bacterial community distribution model is obtained;

[0022] The spectral feature analysis of the three-dimensional bacterial community distribution model is performed to obtain the initial bacterial community dynamic model of the wound, wherein the spectral feature analysis includes at least regional segmentation and characteristic parameter mapping. After regional segmentation, different metabolically active sub-regions are formed, and corresponding physiological characteristic parameters are assigned to different sub-regions.

[0023] Preferably, applying local metabolic conditions and external environmental conditions to the initial microbial community dynamic model to obtain a comprehensive microbial community dynamic model, and combining physiological characteristic parameters to obtain microbial community dynamic coupling data, comprises the following steps:

[0024] Applying local metabolic conditions to the initial bacterial population dynamic model to achieve metabolic boundary constraints, thereby obtaining a first bacterial population dynamic model;

[0025] Applying external environmental conditions to the first bacterial community dynamic model to obtain a comprehensive bacterial community dynamic model, wherein the external environmental conditions include temperature conditions and humidity conditions;

[0026] Based on the comprehensive microbial community dynamic model and physiological characteristic parameters, the microbial community dynamic coupling data is obtained. The specific process includes:

[0027] A dynamic equilibrium equation is constructed, which includes at least a metabolic rate equation, a bacterial diffusion equation, and an activity correlation equation. Combined with a comprehensive bacterial community dynamic model, the dynamic equilibrium equation is solved by a discretization method to obtain the bacterial community density distribution field, the metabolic activity field, and the set of high-risk bacterial community locations.

[0028] Preferably, obtaining the set of high-risk flora positions includes the following steps:

[0029] Based on the metabolic activity field, obtain the average metabolic activity value of each sub-region;

[0030] Identify the sub-regions in the comprehensive flora dynamic model whose average metabolic activity value is not less than the activity threshold, and obtain the set of high-risk flora positions.

[0031] Preferably, constructing the flora classification pre-training model and training it with the flora dynamic coupling data to obtain the flora classification model, and then obtaining the flora abnormal feature parameters, includes the following steps:

[0032] Construct a flora classification pre-training model based on the spectral feature classification network;

[0033] Train and validate the flora classification pre-training model with the flora dynamic coupling data to obtain the flora classification model;

[0034] Input the flora dynamic coupling data to be evaluated into the flora classification model to obtain the predicted set of high-risk flora positions;

[0035] Based on the predicted set of high-risk flora positions, obtain the flora abnormal feature parameters, where the flora abnormal feature parameters at least include the abnormal flora position, abnormal flora type, and abnormal flora morphological parameters.

[0036] Preferably, the step of obtaining the flora abnormal feature parameters based on the predicted set of high-risk flora positions includes the following steps:

[0037] Calculate the centroid coordinates of the predicted set of high-risk flora positions to obtain the abnormal flora position;

[0038] Fit the abnormal flora contour according to the distribution pattern of the predicted set of high-risk flora positions to obtain the abnormal flora morphological parameters, where the abnormal flora morphological parameters include the abnormal flora shape and the abnormal flora coverage range;

[0039] Calculate the metabolic activity direction of the predicted set of high-risk flora positions to obtain the weighted average of the metabolic activity direction, and convert the weighted average into a standard metabolic vector, and the standard metabolic vector is the abnormal flora type.

[0040] Preferably, the step of obtaining the flora abnormal evolution simulation information through the flora abnormal feature parameters, the comprehensive flora dynamic model, and the physiological feature parameters includes the following steps:

[0041] The abnormal bacterial colony location is used as the initial diffusion point, and the abnormal bacterial colony morphological parameters are mapped into the integrated bacterial colony dynamic model. The diffusion direction of the integrated bacterial colony dynamic model is then adjusted to the abnormal bacterial colony type and the abnormal area is meshed to achieve the update of the integrated bacterial colony dynamic model. The diffusion trigger conditions, diffusion path and diffusion rate are defined to obtain the dynamic model of abnormal bacterial colony evolution.

[0042] Based on the dynamic model of abnormal bacterial population evolution, the dynamic equilibrium equation is simulated and analyzed through a recursive calculation method. Based on the results of each recursion in the analysis process, the simulation information of abnormal bacterial population evolution is obtained. The analysis process is the simulation process of bacterial population diffusion, which specifically includes:

[0043] When the diffusion trigger conditions are met, the bacterial population diffuses, and the current bacterial population diffusion trend, diffusion rate, diffusion path, metabolic activity field and density distribution field are obtained, thereby updating the dynamic model of abnormal bacterial population evolution;

[0044] Based on the updated dynamic model of abnormal bacterial population evolution, the dynamic equilibrium equation is re-solved until the termination condition is met, and then the bacterial population diffusion simulation process is terminated;

[0045] Based on the metabolic activity field and density distribution field of each recursion, the maximum metabolic activity and maximum density changes are obtained;

[0046] The diffusion triggering condition includes: if the current metabolic accumulation value is not less than the metabolic threshold, then the bacterial colony is diffused;

[0047] The diffusion path is determined based on the historical diffusion direction and the current metabolic activity direction;

[0048] The bacterial colony diffusion trend and the diffusion rate are determined based on bacterial colony characteristic parameters and environmental change parameters, respectively.

[0049] Preferably, the construction of an optimization model based on the abnormal bacterial flora evolution simulation information and the monitoring device parameters and the optimization of the monitoring device parameters to obtain the optimal monitoring configuration of the monitoring device includes the following steps:

[0050] Constructing an optimization model, wherein the optimization model includes at least an optimization variable model, an optimization target model, and a constraint condition model. The optimization variable model is constructed based on monitoring device parameters. The monitoring device parameters, i.e., optimization variables, include at least sampling interval, spectral resolution, monitoring area density, and data transmission mode. The optimization target model is constructed based on a first risk classification value and a monitoring cost value. The first risk classification value is obtained based on abnormal bacterial flora evolution simulation information. The monitoring cost value includes at least equipment power consumption cost value, calibration cost value, and storage cost value. The constraint condition model includes a monitoring device parameter constraint condition model and a monitoring sensitivity constraint condition model. The monitoring sensitivity constraint condition model is constructed based on data acquisition integrity, data real-timeness, and equipment compatibility.

[0051] The optimization model is preliminarily solved by genetic algorithm to obtain the preliminary optimized monitoring configuration;

[0052] The optimization model is globally solved by the particle swarm algorithm, and the optimal solution set of the particle swarm is the optimal monitoring configuration of the monitoring device.

[0053] Preferably, the present invention further includes a wound flora monitoring system, the system comprising:

[0054] Data acquisition module, used to obtain physiological characteristic parameters of wounds, real-time bacterial flora monitoring data and multispectral imaging data;

[0055] The model construction and analysis module is used to preprocess the multispectral imaging data, construct a three-dimensional bacterial community distribution model, perform spectral feature analysis on the three-dimensional bacterial community distribution model and combine it with physiological characteristic parameters to obtain an initial bacterial community dynamic model of the wound. Local metabolic conditions and external environmental conditions are applied to the initial bacterial community dynamic model to obtain a comprehensive bacterial community dynamic model, which is combined with physiological characteristic parameters to obtain bacterial community dynamic coupling data;

[0056] The bacterial community classification module is used to build a pre-trained bacterial community classification model and train it through the bacterial community dynamic coupling data to obtain the bacterial community classification model and then obtain the bacterial community abnormality characteristic parameters;

[0057] The simulation analysis module is used to obtain abnormal bacterial population evolution simulation information through abnormal bacterial population characteristic parameters, comprehensive bacterial population dynamic model and physiological characteristic parameters;

[0058] Configuration optimization module, which is used to build an optimization model and optimize the monitoring device parameters based on the abnormal bacterial evolution simulation information and monitoring device parameters to obtain the optimal monitoring configuration of the monitoring device;

[0059] The risk control module is used to infer the monitoring configuration of the current period based on the risk grading model, obtain the risk grading prediction value of the current period, and adjust the actual monitoring strategy of the current period based on the risk grading prediction value of the current period, the optimal monitoring configuration of the current period and the monitoring configuration of the current period to realize the monitoring of the wound flora status.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] In terms of data acquisition and modeling, by obtaining physiological characteristic parameters of the wound, real-time microbiota monitoring data, and multispectral imaging data, combining with three-dimensional reconstruction technology to construct a three-dimensional microbiota distribution model, and conducting spectral feature analysis, it is possible to accurately depict the spatial distribution and metabolic activity of the wound microbiota. A variety of techniques such as denoising, standardization, and band fusion used in the preprocessing process effectively improve the data quality and ensure the accuracy of model construction. By applying local metabolic conditions and external environmental conditions (such as temperature and humidity), a comprehensive microbiota dynamic model is constructed, and the dynamic balance equations (including metabolic rate equations, microbiota diffusion equations, and activity correlation equations) are used to solve the microbiota density distribution field, metabolic activity field, and the set of positions of high-risk microbiota, fully considering the internal and external environmental factors of microbiota growth, making the model closer to the actual physiological scenario and being able to truly reflect the dynamic change law of the microbiota.

[0062] In terms of microbiota classification and anomaly recognition, a pre-training model for microbiota classification is constructed based on a spectral feature classification network, and the microbiota dynamic coupling data is used for training and verification, realizing the accurate classification and positioning of high-risk microbiota. By calculating the centroid coordinates of the set of positions of high-risk microbiota, fitting the contour of abnormal microbiota, and converting the metabolic activity direction and other steps, it is possible to accurately obtain the position, type, and morphological parameters of abnormal microbiota, providing a key basis for the early detection of microbiota anomalies. This classification method based on machine learning has higher accuracy and efficiency compared with traditional morphological and biochemical classification methods, and can quickly identify potential infection risks.

[0063] In terms of simulating the abnormal evolution of microbiota, taking the position of abnormal microbiota as the initial diffusion point, by adjusting the diffusion direction, refining the grid, and defining the diffusion trigger conditions, paths, and rates, etc., a dynamic model for the abnormal evolution of microbiota is constructed. Using the recursive calculation method to simulate and analyze the dynamic balance equations, it is possible to track the diffusion trend, metabolic activity changes, and density distribution of the microbiota in real time, and predict the maximum metabolic activity and maximum density changes. This process not only realizes the dynamic simulation of the abnormal evolution process of the microbiota, but also provides a detailed evolution path and key parameters for formulating targeted intervention strategies, helping to take timely measures in the early stage of infection to contain the spread of the microbiota.

[0064] In terms of optimizing the configuration of the monitoring device, an optimization model including an optimization variable model, an objective model, and a constraint condition model is constructed. Taking the sampling interval, spectral resolution, etc. as optimization variables and the risk grading value and monitoring cost as optimization objectives, it is solved by combining the genetic algorithm and the particle swarm algorithm, realizing the dynamic optimization of the monitoring device parameters. This optimization method not only considers the accuracy and sensitivity of monitoring (ensured by constraint conditions such as data acquisition integrity, real-time performance, and device compatibility), but also takes into account the monitoring cost (including device power consumption, calibration, and storage costs). It can adaptively adjust the monitoring strategy according to the abnormal evolution of the flora, improve the utilization efficiency of monitoring resources, avoid resource waste, and ensure the monitoring effect at the same time.

[0065] In terms of risk regulation, based on the risk grading model, the monitoring configuration in the current period is inferred, and the monitoring strategy is adjusted by combining the optimal monitoring configuration and the actual monitoring configuration, realizing the closed-loop monitoring from data acquisition, modeling analysis to strategy adjustment. This dynamic regulation mechanism can timely adjust the monitoring frequency, area, and method according to the real-time monitoring data and the model prediction results, ensuring the effective monitoring of the wound flora status at different risk levels, providing real-time and reliable decision support for clinical treatment, helping to improve the success rate of wound healing, and reducing the incidence of infection complications. Brief Description of the Drawings

[0066] Figure 1 It is the working principle diagram of the monitoring method for the wound flora described in the present invention;

[0067] Figure 2 It is the flow chart for constructing the comprehensive flora dynamic model and generating the flora dynamic coupling data;

[0068] Figure 3 It is the flow chart for constructing the flora classification model and generating the flora abnormal characteristic parameters;

[0069] Figure 4 It is the flow chart for generating the flora abnormal evolution simulation information;

[0070] Figure 5 It is the flow chart for optimizing the optimal monitoring configuration of the monitoring device. Detailed Embodiments

[0071] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0072] Please refer to Figure 1-Figure 5, the present invention relates to a method for monitoring wound flora, and the specific implementation steps are as follows:

[0073] Obtain the physiological characteristic parameters of the wound and real-time flora monitoring data. The physiological characteristic parameters can be collected by devices such as sensors, including but not limited to indicators such as wound temperature, humidity, pH value, etc. that characterize the physiological state of the wound; the real-time flora monitoring data can be obtained through existing flora monitoring technologies and is used to reflect the real-time situation of the wound flora.

[0074] Obtain the multi-spectral imaging data of the wound and perform preprocessing, then construct a three-dimensional flora distribution model, perform spectral feature analysis on the three-dimensional flora distribution model and combine the physiological characteristic parameters to obtain the initial flora dynamic model of the wound.

[0075] Apply local metabolic conditions and external environmental conditions to the initial flora dynamic model to obtain a comprehensive flora dynamic model, and combine the physiological characteristic parameters to obtain flora dynamic coupling data.

[0076] Construct a flora classification pre-training model and train it with the flora dynamic coupling data to obtain a flora classification model, and then obtain flora abnormal characteristic parameters.

[0077] Through the flora abnormal characteristic parameters, the comprehensive flora dynamic model and the physiological characteristic parameters, obtain the flora abnormal evolution simulation information.

[0078] Based on the flora abnormal evolution simulation information and the monitoring device parameters, construct an optimization model and optimize the monitoring device parameters to obtain the optimal monitoring configuration of the monitoring device.

[0079] Based on the risk grading model, reason about the current period monitoring configuration to obtain the risk grading prediction value of the current period. Based on the risk grading prediction value of the current period, the optimal monitoring configuration of the current period and the current period monitoring configuration, adjust the actual monitoring strategy of the current period to achieve the monitoring of the wound flora state.

[0080] Example 1:

[0081] In the process of obtaining the multi-spectral imaging data of the wound and performing preprocessing to construct a three-dimensional flora distribution model and an initial flora dynamic model, perform multi-band scanning on the wound surface. A multi-spectral imaging device can be used to sequentially scan the wound surface according to preset different bands, so as to obtain a multi-spectral imaging data sequence of different bands of the wound. These multi-spectral imaging data sequences contain information on the wound under different spectral bands and can reflect the performance of the flora under different spectral characteristics.

[0082] Preprocess the multi-spectral imaging data sequence. During the preprocessing, one or more processing methods such as denoising, normalization, band fusion, abnormal signal elimination, data interpolation, and spectral smoothing can be selected according to the actual data situation. For example, for denoising, a filtering algorithm can be used to remove the noise interference in the data and improve the signal-to-noise ratio of the data; for normalization, the data in different bands can be unified to the same scale range for subsequent analysis; for band fusion, the information in different bands can be integrated to form more comprehensive image information. Through preprocessing, the preprocessed multi-spectral imaging data sequence is obtained, making it more suitable for subsequent model construction.

[0083] Based on the preprocessed multi-spectral imaging data sequence, combined with three-dimensional reconstruction techniques such as three-dimensional modeling algorithms, convert the two-dimensional multi-spectral imaging data into a distribution model in three-dimensional space to obtain the corresponding three-dimensional flora distribution model. This three-dimensional flora distribution model can intuitively display the spatial distribution of the flora on the wound surface and in the depth direction.

[0084] Perform spectral feature analysis on the three-dimensional flora distribution model. The spectral feature analysis at least includes region segmentation and feature parameter mapping. Through the region segmentation algorithm, the three-dimensional flora distribution model is divided into different sub-regions with metabolic activity, and each sub-region has similar metabolic activity characteristics. Then, corresponding physiological characteristic parameters are assigned to different sub-regions. For example, according to the high or low metabolic activity of the sub-region, different physiological characteristic parameters such as temperature and humidity are assigned, so as to obtain the initial flora dynamic model of the wound, which can reflect the relationship between the metabolic activity and physiological characteristics of the flora in different regions.

[0085] In the process of obtaining the multi-spectral imaging data of the wound and constructing the initial flora dynamic model, take the wound of a diabetic foot ulcer patient as an example for specific implementation:

[0086] Use a multi-spectral imager equipped with ultraviolet (365nm), visible light (450nm, 550nm, 650nm), and near-infrared (850nm) bands to scan the wound surface. During the scanning, the imager is kept 10 cm away from the wound, and images are collected in the order of ultraviolet → visible light → near-infrared. One band is scanned per second, and a total of 5 bands of original image sequences are obtained. Each band contains 10 consecutive images, forming a multi-spectral imaging data sequence. For example, the ultraviolet band image can show the distribution of fluorescent substances in the wound exudate, and the near-infrared band image can penetrate the superficial epidermis to reflect the flora metabolic state of the subcutaneous tissue.

[0087] Subsequently, preprocess the original data sequence:

[0088] Denoising process: For the salt-and-pepper noise generated by environmental light interference in the ultraviolet band image, use the median filtering algorithm to process the 8-neighborhood of each pixel point to eliminate isolated noise points;

[0089] Normalization: For visible light band images, normalize each pixel value to the [0, 1] range to unify the grayscale range of different bands and avoid data deviation caused by differences in device sensitivity;

[0090] Band fusion: The UV, visible, and near-infrared bands are fused into a single image through principal component analysis (PCA) to highlight the spectral differences of metabolites from different bacterial communities. For example, blue areas in the fused image represent areas of high metabolic activity containing Pseudomonas fluorescens, while red areas represent areas where Staphylococcus aureus aggregates.

[0091] Abnormal signal rejection: By setting a grayscale threshold (e.g., pixels below 20 or above 230 are considered abnormal), abnormal pixels caused by device jitter or wound surface contamination are rejected;

[0092] Data interpolation: For the local data missing areas caused by wound folds in the visible light 450nm band, the average value of the adjacent pixels is used to fill them;

[0093] Spectral smoothing: Savitzky-Golay filtering is applied to near-infrared band images to reduce random fluctuations in the spectral curve and highlight the characteristic peaks of bacterial metabolism.

[0094] After preprocessing, a 3D bacterial distribution model was constructed using 3D reconstruction technology based on the principle of optical tomography. The specific steps are as follows: the fused 2D image sequence is used as input, and the coordinates (x, y, z) of each pixel in 3D space are calculated using a triangulation method based on stereo vision, where the z axis is the depth direction perpendicular to the wound surface. The spatial position of each layer of the image is adjusted through an iterative optimization algorithm, and the final resolution is 0.1mm. 3 The 3D model visually displays the distribution of bacteria on the wound surface (xy plane) and at depth (z axis). For example, the model shows dense cocci (approximately 0.5 μm in diameter) at the shallow wound edge (z = 0.2-0.5 mm), while rod-shaped bacteria are distributed in a linear pattern deep in the wound center (z = 1.0-1.5 mm).

[0095] Next, perform spectral feature analysis on the three-dimensional model:

[0096] Region segmentation: The density-based clustering algorithm (DBSCAN) is used to cluster the voxel points in the 3D model, and the density threshold is set to 50 voxels / mm 3 , the neighborhood radius is 0.3mm, and the model is divided into three sub-areas:

[0097] Region A (superficial layer at the wound edge): High voxel density, with spectral characteristics mainly showing high reflectivity in the 450 nm band, determined as a region of high metabolic activity, presumably dominated by aerobic bacteria (such as Staphylococcus aureus);

[0098] Region B (superficial layer in the center of the wound): Medium voxel density, relatively high reflectivity in the 550 nm band, determined as a region of medium metabolic activity, presumably colonized by a mixture of facultative anaerobic bacteria (such as Escherichia coli);

[0099] Region C (deep layer in the center of the wound): Low voxel density, but significant absorbance in the 850 nm band, determined as a region of low metabolic activity, presumably a latent region for anaerobic bacteria (such as Bacteroides).

[0100] Characteristic parameter mapping: Assign physiological characteristic parameters to each sub-region:

[0101] Region A: Considering the local temperature of 38.5 °C (3 °C higher than the normal skin temperature) and humidity of 85% measured by the sensor, it is determined as an inflammation-active region;

[0102] Region B: With a local pH value of 7.2 (close to neutral) and an oxygen partial pressure of 40 mmHg, it is determined as a region of transitional bacterial colonization;

[0103] Region C: With an oxygen partial pressure < 10 mmHg and a carbon dioxide partial pressure of 60 mmHg, it is determined as an anaerobic microenvironment region.

[0104] Through the above processing, an initial bacterial community dynamic model of the wound is finally formed. This model correlates the spatial distribution of the bacterial community, spectral characteristics, and physiological environment parameters, providing a basis for subsequent imposition of metabolic conditions and environmental constraints. For example, the model can show that the aerobic bacteria in Region A have an increased metabolic rate in a high-humidity and high-temperature environment and may spread to Region B, while the anaerobic bacteria in Region C remain relatively stationary due to the low-oxygen environment.

[0105] Example 2:

[0106] In the process of imposing local metabolic conditions and external environmental conditions on the initial bacterial community dynamic model to obtain a comprehensive bacterial community dynamic model and bacterial community dynamic coupling data, local metabolic conditions are imposed on the initial bacterial community dynamic model to achieve metabolic boundary constraints. The local metabolic conditions can be set according to the metabolic environmental characteristics of different regions of the wound. For example, in a specific region of the wound, due to different degrees of tissue damage, there may be different nutrient supplies and metabolite accumulations, which will have a constraining effect on the metabolic activities of the bacterial community. By imposing local metabolic conditions, a first bacterial community dynamic model is obtained, which takes into account the influence of the local metabolic environment on the bacterial community dynamics.

[0107] External environmental conditions, including temperature and humidity, are applied to the first microbial dynamics model. Temperature and humidity are important environmental factors that influence microbial growth and reproduction. Based on the actual wound care environment or the patient's surroundings, appropriate temperature and humidity ranges can be set and applied as external environmental conditions to the first microbial dynamics model, resulting in a comprehensive microbial dynamics model that comprehensively considers the impact of local metabolic conditions and external environmental conditions on microbial dynamics.

[0108] Based on the integrated microbial dynamic model and physiological characteristic parameters, a dynamic equilibrium equation is constructed. This dynamic equilibrium equation includes at least the metabolic rate equation, the microbial diffusion equation, and the activity correlation equation. These equations describe the relationship between the metabolic rate, diffusion behavior, and activity of the microbial community under comprehensive environmental conditions. The dynamic equilibrium equation is solved using discretization methods, such as the finite difference method and the finite element method, to convert the continuous model into a discrete numerical form. This results in the microbial density distribution field, the metabolic activity field, and the high-risk microbial location set. The microbial density distribution field reflects the spatial density distribution of the microbial community; the metabolic activity field shows the metabolic activity level of the microbial community in different regions; and the high-risk microbial location set is determined by analyzing the metabolic activity field. Specifically, based on the metabolic activity field, the average metabolic activity value of each subregion is calculated. Subregions in the integrated microbial dynamic model whose average metabolic activity value is not less than the activity threshold are identified. These subregions are considered high-risk microbial locations, indicating that the microbial community in these areas has high metabolic activity and a higher risk of infection or worsening.

[0109] Taking the wound of a diabetic foot ulcer patient in Example 1 as an example, the process of applying conditions to the initial flora dynamic model and obtaining flora dynamic coupling data is further described:

[0110] Apply local metabolic conditions to the initial bacterial community dynamic model to achieve metabolic boundary constraints. According to the physiological characteristic parameters of different wound areas, set:

[0111] Region A (area of high metabolically active aerobic bacteria): A glucose concentration boundary constraint (5 mmol / L, simulating the nutrient level in wound exudate) was imposed to limit the accumulation of metabolites caused by excessive nutrient consumption by the bacterial community;

[0112] Region C (anaerobic bacteria with low metabolic activity): An oxidation-reduction potential (ORP) boundary constraint (-150 mV, simulating the anaerobic environment of deep tissue) is applied to inhibit the invasion of facultative anaerobic bacteria into this region.

[0113] By using the finite element analysis method, the metabolic boundary conditions are discretized into constraint equations for each voxel point in the model, and the first flora dynamic model is obtained. For example, the metabolic rate of Staphylococcus aureus in region A is limited by the glucose concentration, and its acid production rate decreases from 0.1 mmol / (h·10^6 CFU) to 0.05 mmol / (h·10^6 CFU), and the model shows that the pH value in this region decreases at a slower rate.

[0114] External environmental conditions are applied to the first flora dynamic model, and the temperature of the ward where the patient is located is set to 25°C ± 1°C and the humidity is set to 60% ± 5%. The environmental conditions are coupled to the model through heat and mass transfer equations:

[0115] Influence of temperature field: The temperature on the wound surface is calculated by the heat conduction equation, and the temperature of the deep tissue is corrected based on the blood perfusion model. Finally, the temperature in region A of the model stabilizes at 37.8°C (close to the normal human body temperature), and the temperature in region C stabilizes at 37.2°C;

[0116] Influence of humidity field: The humidity on the wound surface is calculated by the evaporation-condensation equilibrium equation, and the epidermal permeability is set to 0.01 g / (h·cm 2 ), the humidity in region A of the model is maintained at 80%, and the humidity in region C reaches 90% due to the accumulation of exudate.

[0117] After applying the external conditions, a comprehensive flora dynamic model that comprehensively considers local metabolism and environmental factors is obtained. For example, in region B (facultative anaerobe area) under the 25°C environment, the division cycle of Escherichia coli is extended from 30 minutes to 45 minutes, and the growth rate of the flora density decreases by 30%.

[0118] Based on the comprehensive flora dynamic model and physiological characteristic parameters, a dynamic equilibrium equation is constructed and solved:

[0119] Metabolic rate equation: The Monod equation is used to describe the relationship between the growth of the flora and nutrients. For example, the growth rate μ of Staphylococcus aureus in region A is μ = μ_max × (S / (K_s + S)), where μ_max = 1.5 / h (maximum specific growth rate), S = glucose concentration, and K_s = 0.5 mmol / L (half-saturation constant);

[0120] Flora diffusion equation: Fick's second law is used to describe the diffusion behavior of the flora, and the diffusion coefficients are D = 10^-10 m 2 / s (aerobic bacteria) and D = 5 × 10^-11 m 2 / s (anaerobic bacteria), reflecting the differences in the motility of different bacterial species;

[0121] Activity correlation equation: A coupling relationship between metabolites (such as lactic acid and hydrogen peroxide) and the activity of the flora is established. For example, when the lactic acid concentration increases by 1 mmol / L, the activity of Staphylococcus aureus decreases by 5%.

[0122] The dynamic equilibrium equation is discretized and solved by the finite difference method. The three-dimensional model is divided into grid cells of 0.1 mm×0.1 mm×0.1 mm, and each cell is iteratively calculated at a time step of Δt = 10 minutes. After 12 hours of simulation, the following results are obtained:

[0123] Bacterial density distribution field: The bacterial density in region A increases from 10^8 CFU / cm 3 to 1.5×10^8 CFU / cm 3 (aerobic environment promotes growth), and the bacterial density in region C increases from 10^7 CFU / cm 3 to 1.2×10^7 CFU / cm 3 (hypoxic environment restricts growth);

[0124] Metabolic activity field: The metabolic activity is characterized by the ATP fluorescence intensity. The average fluorescence intensity in region A is 800 RLU (relative light unit), in region B is 500 RLU, and in region C is 200 RLU;

[0125] Set of positions of high-risk bacteria: Set the activity threshold to 600 RLU, and identify the sub-regions in region A with a fluorescence intensity ≥ 600 RLU, specifically the shallow layer region 0.5 cm×0.3 cm extending outward from the wound edge (volume about 0.15 cm 3 ), where the aerobic bacteria have active metabolism and there is a risk of infection spread.

[0126] By applying metabolic and environmental conditions and solving the dynamic equilibrium equation, the quantitative analysis of the bacterial dynamics is realized, providing key data support for subsequent bacterial classification and abnormal evolution simulation. For example, the determination of the set of positions of high-risk bacteria can guide clinicians to increase the sampling frequency in this region and monitor the changes of bacteria in a timely manner.

[0127] Example 3:

[0128] In the process of constructing a pre-trained model for bacterial classification and training it with bacterial dynamics coupling data to obtain a bacterial classification model and bacterial abnormal feature parameters, a pre-trained model for bacterial classification is constructed based on a spectral feature classification network. The spectral feature classification network can adopt existing deep learning network architectures, such as convolutional neural network (CNN), etc., which can automatically extract and classify spectral feature data.

[0129] The pre-trained microbial classification model is trained and validated using microbial dynamic coupling data. During training, the microbial dynamic coupling data is fed into the pre-trained model. By adjusting the model's parameters, the model accurately classifies the type and state of the microbial community. The validation process uses a subset of data not used in training to evaluate the performance of the trained model and ensure its accuracy and generalization. After training and validation, a microbial classification model is obtained that accurately classifies the input microbial dynamic coupling data.

[0130] The dynamic coupling data of the microbial community to be evaluated is input into the microbial community classification model. The model analyzes and processes the input data and outputs a predicted set of high-risk microbial community locations. This set predicts areas where high-risk microbial communities may exist.

[0131] Based on the predicted set of high-risk bacterial colony locations, its centroid coordinates are calculated to obtain the abnormal bacterial colony location, which represents the concentrated distribution area of the high-risk bacterial colony. According to the distribution morphology of the predicted set of high-risk bacterial colony locations, a suitable contour fitting algorithm is used to fit the abnormal bacterial colony contour to obtain the abnormal bacterial colony morphological parameters, wherein the abnormal bacterial colony morphological parameters include the abnormal bacterial colony shape and the abnormal bacterial colony coverage, which can describe the spatial morphological characteristics of the high-risk bacterial colony. At the same time, the metabolic activity direction of the predicted set of high-risk bacterial colony locations is calculated, and the weighted average of the metabolic activity direction is obtained by weighted averaging the metabolic activity directions of each location. The weighted average is converted into a standard metabolic vector. The standard metabolic vector is the abnormal bacterial colony type, which is used to characterize the metabolic characteristics and type of the high-risk bacterial colony.

[0132] Taking the wound of a diabetic foot ulcer patient in Example 2 as an example, it is assumed that the dynamic coupling data of the bacterial community is obtained by the integrated bacterial community dynamic model, including the bacterial community density distribution field, metabolic activity field and high-risk bacterial community location set (volume of about 0.15 cm) in area A (shallow layer of wound edge). 3 , containing a mixed flora of Staphylococcus aureus and Staphylococcus epidermidis). The following is the specific process of building a flora classification model and deriving abnormal characteristic parameters:

[0133] 1. Construction of pre-training model for bacterial community classification

[0134] A pre-trained model for bacterial classification is constructed based on a convolutional neural network (CNN). The network structure includes:

[0135] Input layer: Receives the spectral feature data of the three-dimensional bacterial community distribution model (the dimension is 5×10×10, corresponding to 5 bands and 10×10 pixel two-dimensional slices);

[0136] Convolution layer: uses a 3×3 convolution kernel to extract texture features in spectral images (such as the spectral absorption peak morphology of different bacterial communities);

[0137] Pooling layer: Max pooling is used to reduce the dimension and retain key features;

[0138] Fully connected layer: The number of output layer nodes is 3, corresponding to the probability values of aerobic bacteria, facultative anaerobic bacteria, and anaerobic bacteria.

[0139] II. Model Training and Validation

[0140] Training data preparation: The spectral feature data of aerobic bacteria (such as Staphylococcus aureus), facultative anaerobic bacteria (such as Escherichia coli), and anaerobic bacteria (such as Bacteroides) marked in historical monitoring data are used as the training set, with a total of 10,000 groups of samples. Each group of samples contains spectral reflectance data of 5 bands (such as reflectance R365 in the ultraviolet band, reflectance R450 at visible light 450nm, etc.) and the corresponding bacterial community category labels (one-hot encoding).

[0141] Training process: The cross-entropy loss function and Adam optimizer are used. The learning rate is set to 0.001, the batch size is 32, and 20 epochs are trained. During the training process, the model extracts features from the input spectral data. For example:

[0142] Staphylococcus aureus has a relatively high reflectance (>0.6) in the R450 band and a relatively low reflectance (<0.3) in the R850 band;

[0143] Escherichia coli has a characteristic absorption valley (reflectance <0.4) in the R550 band.

[0144] Verification results

[0145] Using 2,000 groups of validation samples to test the model, the classification accuracy rate reaches 92%. Among them, the recognition accuracy rate of aerobic bacteria is 95%, facultative anaerobic bacteria is 88%, and anaerobic bacteria is 90%, meeting the clinical monitoring accuracy requirements.

[0146] III. Derivation of Abnormal Feature Parameters of Bacterial Communities

[0147] Input the currently monitored dynamic coupling data of bacterial communities (spectral feature data of area A) into the trained bacterial community classification model, and the output prediction results are: the probability of aerobic bacteria is 90%, facultative anaerobic bacteria is 8%, and anaerobic bacteria is 2%. It is determined that the high-risk bacterial community is mainly aerobic bacteria dominated by Staphylococcus aureus, and the position set of the predicted high-risk bacterial community is obtained (three-dimensional coordinate range: x = 2.0 - 3.0 cm, y = 1.5 - 2.1 cm, z = 0.2 - 0.4 cm).

[0148] Calculation of abnormal bacterial community location

[0149] The centroid coordinate formula is used to calculate the centroid of the high-risk bacterial community location set:

[0150]

[0151] Where n is the number of voxel points in the set (n = 1500 in this example), (x i ,y i ,z i ) are the three-dimensional coordinates of the i-th voxel point. The calculated coordinates of the center of gravity are G(2.5cm, 1.8cm, 0.3cm), which is the location of the abnormal bacterial colony.

[0152] Abnormal bacterial flora morphological parameter fitting

[0153] Shape fitting: The distribution shape of the high-risk bacterial colony location set was fitted using an ellipse fitting algorithm, resulting in a major axis length of 0.8 cm (along the wound edge) and a minor axis length of 0.6 cm (perpendicular to the edge), and the abnormal bacterial colony shape was determined to be an ellipse.

[0154] Coverage: The three-dimensional volume occupied by the calculated set is 0.15 cm 3 , corresponding to covering 20% of the wound surface area (approximately 3cm 2 ).

[0155] Determination of abnormal bacterial flora type

[0156] Calculate the metabolic activity direction vector of each voxel where a i 、b i 、c i They are the activity values of the three metabolic dimensions of glucose consumption, lactate production, and oxygen consumption at this point (obtained by inferring spectral characteristics);

[0157] Find the weighted mean: Among them, w i is the metabolic activity weight of the voxel point (in this case, the ATP fluorescence intensity value is taken, ranging from 0 to 1000 RLU);

[0158] The weighted average vector Normalized to unit vector to obtain standard metabolic vector Corresponding to the dominance of aerobic metabolism (high proportion of glucose consumption and oxygen consumption), the abnormal bacterial flora type is determined to be a high metabolic activity aerobic type.

[0159] IV. Clinical Application Examples

[0160] Through the above process, it was determined that the patient's wound area A had an abnormal proliferation of highly metabolizing aerobic bacteria with an elliptical distribution, with clear location, morphology, and metabolic type. Based on this, the clinician can take targeted measures:

[0161] Precise sampling: Deep sampling is performed using a sterile swab at the centroid position G(2.5 cm, 1.8 cm, 0.3 cm) to improve the detection rate of pathogenic bacteria;

[0162] Targeted therapy: According to the flora type (aerobic bacteria), β-lactam antibiotics (such as penicillin) are selected, and glucose competitive inhibitors are used in combination with the metabolic characteristics (high glucose consumption) to inhibit the growth of the flora;

[0163] Dynamic monitoring: An encrypted monitoring area is set for the elliptical coverage range, the scanning frequency of spectral imaging is increased, and the morphological changes of the flora are tracked in real time.

[0164] This process realizes the precise identification and characterization of abnormal flora through a data-driven classification model and quantitative analysis, providing a scientific basis for infection diagnosis and treatment plan formulation.

[0165] Example 4:

[0166] In the process of obtaining the simulated information on the abnormal evolution of the flora through the abnormal flora characteristic parameters, the comprehensive flora dynamic model, and the physiological characteristic parameters, first, the position of the abnormal flora is used as the initial diffusion point, and the morphological parameters of the abnormal flora are mapped into the comprehensive flora dynamic model. By adjusting the diffusion direction of the comprehensive flora dynamic model to the abnormal flora type, the model can reflect the influence of the metabolic characteristics of the abnormal flora on the diffusion direction. At the same time, the abnormal area is refined into grids to improve the model resolution of the abnormal area, so as to more accurately simulate the evolution process of the flora in the abnormal area, thereby realizing the update of the comprehensive flora dynamic model. In the updated model, the diffusion trigger condition, diffusion path, and diffusion rate are defined. Among them, the diffusion trigger condition includes that if the current metabolic accumulation value is not less than the metabolic threshold, the flora diffusion will occur, and the metabolic accumulation value can be calculated according to the metabolic activity and time of the flora; the diffusion path is determined based on the historical diffusion direction and the current metabolic activity direction. By analyzing the diffusion direction of the flora at the historical moment and the current metabolic activity direction, the possible diffusion path of the flora is determined; the diffusion rate and the flora diffusion trend are determined based on the flora characteristic parameters and the environmental change parameters respectively. The flora characteristic parameters include the type and growth rate of the flora, and the environmental change parameters include the changes in external environmental conditions such as temperature and humidity.

[0167] Based on the updated dynamic model of abnormal bacterial evolution, the dynamic equilibrium equation is simulated and analyzed using a recursive calculation method. During the analysis, when the diffusion trigger condition is met, bacterial diffusion is carried out, and the current bacterial diffusion trend, diffusion rate, diffusion path, metabolic activity field, and density distribution field are calculated to update the dynamic model of abnormal bacterial evolution. Then, based on the updated dynamic model of abnormal bacterial evolution, the dynamic equilibrium equation is re-solved and the above process is repeated until the termination condition is met, such as reaching the preset simulation time or the bacterial diffusion reaches a stable state, etc., then the bacterial diffusion simulation process is terminated. During each recursive process, the changes in the metabolic activity field and density distribution field are recorded. Based on these records, the maximum metabolic activity and maximum density changes are obtained. This information can reflect the key characteristics and trends in the abnormal bacterial evolution process and provide a basis for subsequent adjustments to the monitoring strategy.

[0168] Taking the wound of a diabetic foot ulcer patient in Example 2 as an example, assume that the flora classification model identifies the presence of an abnormal flora (overgrowth of Staphylococcus aureus) in region A (the shallow aerobic bacterial area at the wound edge). The abnormal flora location is the coordinates of the center of gravity of region A (x = 2.5 cm, y = 1.8 cm, z = 0.3 cm), the abnormal flora morphological parameters are the circular coverage area with a radius of 0.4 cm, and the abnormal flora type is an aerobic type with high metabolic activity (the standard metabolic vector points in the direction of glucose consumption). The following is the specific derivation process of the abnormal flora evolution simulation information:

[0169] 1. Model update and parameter definition

[0170] Initial diffusion points and morphological mapping

[0171] The abnormal bacterial colony position (x = 2.5cm, y = 1.8cm, z = 0.3cm) is used as the initial diffusion point, and this point is marked as the diffusion source in the integrated bacterial colony dynamic model. The abnormal bacterial colony morphological parameters (circle with a radius of 0.4cm) are mapped to a spherical diffusion area in the three-dimensional model. The grid resolution in this area is refined from 0.1mm to 0.05mm to improve the accuracy of the diffusion simulation. For example, the spherical area contains about 2000 grid cells (250 cells at the original resolution), which can capture the micro-diffusion behavior of the bacterial colony in more detail.

[0172] Diffusion direction and trigger condition setting

[0173] Diffusion direction: Based on the abnormal bacterial flora type (aerobic), the diffusion direction of the integrated bacterial flora dynamic model was adjusted to the direction of the glucose concentration gradient (i.e., diffusion from high concentration to low concentration). The glucose concentration in the exudate at the wound edge is high (5 mmol / L), while the central region has a lower concentration (3 mmol / L) due to bacterial consumption, so the diffusion direction points to the center of the wound.

[0174] Diffusion Trigger Condition: Set the metabolic accumulation value threshold to 500 RLU·h (RLU is the unit of ATP fluorescence intensity). When the metabolic accumulation value (fluorescence intensity × time) of a certain grid cell ≥ the threshold, the bacterial colony diffusion is triggered. For example, the metabolic activity of the initial diffusion point is 800 RLU, and after accumulating for 0.625 hours (37.5 minutes), the trigger condition is satisfied.

[0175] Definition of Diffusion Path and Rate

[0176] Diffusion Path: Based on the historical diffusion direction (the diffusion direction in the previous time period is perpendicular to the wound edge towards the center) and the current metabolic activity direction (the glucose concentration gradient direction), the real-time diffusion path is determined through vector synthesis. For example, the initial diffusion path is the angular bisector direction of the angle between the two, that is, extending towards the center at a 45° angle to the wound edge.

[0177] Diffusion Rate: According to the bacterial colony characteristic parameters (the swimming speed of Staphylococcus aureus is 10 μm / s) and the environmental change parameters (the motility decreases by 20% at 25°C), the actual diffusion rate is calculated to be 8 μm / s, that is, the diffusion distance per hour is about 28.8 mm.

[0178] II. Recursive Simulation Process and Results

[0179] Set the total simulation duration to 24 hours, and the time step Δt = 1 hour. Solve the dynamic equilibrium equation through recursive calculation:

[0180] The 1st hour: Diffusion is triggered

[0181] Condition Judgment: The metabolic accumulation value of the initial diffusion point = 800 RLU × 1h = 800 RLU·h ≥ 500 RLU·h, meeting the diffusion trigger condition.

[0182] Diffusion Execution: The bacterial colony diffuses along the set path (at a 45° angle towards the center), and the diffusion distance = 8 μm / s × 3600 s = 28.8 mm, forming a cylindrical diffusion front with a length of 28.8 mm and a diameter of 0.4 cm in the model.

[0183] Parameter Update:

[0184] The bacterial colony density in the diffusion front region increases by 20% (from 10^8 CFU / cm 3 to 1.2×10^8 CFU / cm 3 );

[0185] The metabolic activity field shows that the ATP fluorescence intensity at the diffusion front is 750 RLU (slightly decreased due to nutrient consumption);

[0186] The density distribution field is updated to the superposition form of a spherical source area and a cylindrical diffusion area.

[0187] Hour 6: Path Adjustment and Second Triggering

[0188] Path correction: As the pH value in the central area of the wound dropped to 6.8 (acidic environment), the metabolic activity of Staphylococcus aureus decreased, and the diffusion direction shifted to area B (facultative anaerobic bacteria area) with a more neutral pH value, and the path deviated by 30° from the initial direction.

[0189] Secondary triggering: The metabolic accumulation value of a grid cell in the diffusion front = 750RLU×5h=3750RLU·h≥500RLU·h, triggering secondary diffusion and forming a branched diffusion path.

[0190] Result output:

[0191] The bacterial community diffusion trend showed that the main path extended to the edge of area B, and the branch path spread to the deep tissue below area A;

[0192] The maximum metabolic activity was still located in the initial source region (800 RLU), and the maximum density change was in the diffusion front region (+30%).

[0193] Hour 24: Simulation terminated

[0194] Termination condition: The bacterial colony spreads to healthy tissue beyond the wound edge (beyond the model boundary), triggering the termination mechanism.

[0195] Final result:

[0196] Diffusion trend: The bacterial community formed a tree-root-like diffusion network, with the main path penetrating into the deep layer of area B (z = 0.8 mm) and the branch path invading the shallow layer of area C (z = 0.6 mm);

[0197] Maximum metabolic activity: Due to the accumulation of metabolites around the source area, the activity dropped to 600 RLU, while the activity in the newly colonized area at the diffusion front rose to 700 RLU (sufficient nutrition);

[0198] Maximum density change: The density in the center of the diffusion front reaches 2×10^8 CFU / cm 3 (100% increase compared to the initial level), indicating explosive growth of the bacterial population.

[0199] 3. Key Parameters and Clinical Significance

[0200] Verification of diffusion trigger conditions: During actual monitoring, if the metabolic accumulation value in a certain area rises rapidly to the threshold, it indicates that immediate intervention (such as local debridement) is required.

[0201] Clinical application of diffusion paths and rates: Based on the simulation results, the risk of bacterial flora spreading to areas B and C can be predicted, guiding the adjustment of dressing type (such as using antibacterial dressings containing silver ions on the diffusion path).

[0202] The warning value of maximum metabolic activity and density changes: When the maximum density changes by more than 50%, it indicates that the infection has entered the progressive stage and requires upgraded antibiotic treatment.

[0203] Through this simulation process, the spatiotemporal characteristics of abnormal microbial flora evolution can be dynamically predicted, providing a quantitative basis for developing personalized clinical monitoring and intervention strategies. For example, in this case, the simulation showed that abnormal microbial flora could invade deep tissues within 24 hours, necessitating enhanced real-time monitoring of areas B and C and shortening the sampling interval to every two hours.

[0204] Embodiment 5:

[0205] In the process of constructing an optimization model based on abnormal bacterial flora evolution simulation information and monitoring device parameters and optimizing the monitoring device parameters to obtain the optimal monitoring configuration, the optimization model is first constructed. The optimization model includes at least an optimization variable model, an optimization target model, and a constraint condition model. The optimization variable model is constructed based on the monitoring device parameters. The monitoring device parameters are the optimization variables and include at least the sampling interval, spectral resolution, monitoring area density, and data transmission mode. The sampling interval determines the frequency of data collection, the spectral resolution affects the accuracy of identifying the spectral characteristics of the bacterial flora, the monitoring area density is related to the degree of monitoring coverage of different areas of the wound, and the data transmission mode affects the efficiency and stability of data transmission.

[0206] An optimization target model is constructed based on the first risk classification value and the monitoring cost value. The first risk classification value is obtained based on the simulation information of abnormal bacterial flora evolution and is used to assess the risk level of the current bacterial flora status. The monitoring cost value includes at least the equipment power consumption cost value, calibration cost value, and storage cost value, aiming to minimize the monitoring cost while ensuring the monitoring effect. The constraint model includes the monitoring device parameter constraint model and the monitoring sensitivity constraint model. The monitoring sensitivity constraint model is constructed based on data collection integrity, data real-time, and equipment compatibility to ensure that the optimized monitoring device parameters can meet the actual monitoring needs and guarantee the accuracy and reliability of the data.

[0207] The optimization model is initially solved using a genetic algorithm. A genetic algorithm is a random search algorithm that simulates the natural evolution process. Through operations such as selection, crossover, and mutation, it searches for a better solution in the solution space of the optimization variables and obtains a preliminary optimized monitoring configuration.

[0208] The optimization model is globally solved using a particle swarm optimization algorithm. This algorithm, based on swarm intelligence, simulates the foraging behavior of bird flocks to search for the optimal solution within the solution space. The resulting optimal solution set represents the optimal monitoring configuration for the monitoring device. This optimal monitoring configuration balances monitoring risk and cost while satisfying monitoring sensitivity and constraints, enabling efficient monitoring of wound microbial status.

[0209] Taking the abnormal evolution of the flora of patients with diabetic foot ulcers in Example 4 as the background (the flora spreads to the deep layer, and the risk level is medium risk), the following is the specific process of optimizing the configuration of the monitoring device:

[0210] 1. Clarify optimization goals and parameters

[0211] Example scenario: The patient's wound area is approximately 5cm 2 The high-risk areas are concentrated in the superficial layer of the wound edge (area A) and the diffusion path (extending to the deep center). The monitoring device parameters that need to be optimized include:

[0212] Sampling interval: The original configuration is 60 minutes, and it is necessary to determine whether to shorten it to capture rapidly spreading bacterial colonies;

[0213] Spectral resolution: The original configuration is 20nm, which may not be able to distinguish the spectral differences of similar bacterial groups;

[0214] Monitoring area density: original configuration is 5 points / cm 2 , high-risk areas need to be encrypted;

[0215] Data transmission mode: The original method used wired transmission, which caused the problem of cable entanglement affecting nursing care.

[0216] Optimization goal: Reduce the risk of bacterial spread monitoring while controlling monitoring costs and ensure data integrity and real-time performance.

[0217] 2. Build an optimization model (simplify formula logic)

[0218] Risk grading and cost association

[0219] Risk value: According to simulation results, if the sampling interval exceeds 30 minutes, the risk of missed detection of bacterial spread increases by 20%; when the spectral resolution is lower than 10nm, the misjudgment rate of Staphylococcus aureus and Staphylococcus epidermidis reaches 15%.

[0220] Cost items:

[0221] For every 15 minutes the sampling interval is shortened, the device power consumption increases by 10% (battery life drops from 24 hours to 18 hours);

[0222] The spectral resolution is increased to 10nm, and the calibration cost increases from 50 yuan to 80 yuan per calibration.

[0223] Monitoring area density from 5 points / cm 2 Increased to 10 points / cm 2 , data storage costs increase by 15 yuan per day;

[0224] The power consumption of wireless transmission module is 3 times higher than that of wired one, but it can achieve real-time data synchronization.

[0225] Constraints

[0226] The monitoring coverage of key areas (such as Area A) must be ≥95%;

[0227] The delay from data collection to display on the medical end must be ≤10 minutes;

[0228] The total power consumption of the device must be ≤5W (to avoid battery overheating).

[0229] 3. Optimize configuration through algorithms (simplify the calculation process)

[0230] Genetic algorithm preliminary screening

[0231] Generate candidate solutions:

[0232] Scheme 1: Sampling interval 30 minutes, spectral resolution 10 nm, monitoring density 10 points / cm 2 , wireless transmission;

[0233] Scheme 2: Sampling interval 60 minutes, spectral resolution 5 nm, monitoring density 15 points / cm 2 , wired transmission;

[0234] Scheme 3: Sampling interval 15 minutes, spectral resolution 20 nm, monitoring density 5 points / cm 2 , wireless transmission.

[0235] Evaluate the pros and cons:

[0236] Option 1: Low risk (missed detection rate 5%), moderate cost (power consumption 3W, calibration + storage 95 yuan per day);

[0237] Option 2: Low risk but high cost (calibration costs 80 yuan per day, and wired transmission is inconvenient);

[0238] Option 3: Low cost but high risk (15% misjudgment rate).

[0239] Preliminary optimization: retain Option 1 and Option 2 and proceed to the next step.

[0240] Particle swarm optimization global optimization

[0241] Simulation iteration: By adjusting the parameter combination, it was found that "sampling interval 30 minutes + spectral resolution 10nm + monitoring density 10 points / cm 2 + wireless transmission” combination has always balanced risks and costs in multiple iterations.

[0242] Final configuration:

[0243] Sampling interval of 30 minutes: taking into account both data real-time performance and device battery life (power consumption 2.5W);

[0244] Spectral resolution of 10 nm: Can distinguish the spectral characteristics of more than 90% of aerobic bacteria, with a calibration cost of 80 yuan per day;

[0245] Monitoring density of 10 points / cm 2 (In high-risk areas, it is encrypted to 15 points / cm 2 ): The key area coverage rate is 100%, and the storage cost is 15 yuan per day;

[0246] Transmission mode: Wireless transmission, with a delay of 3 - 5 minutes, meeting the requirements of real-time early warning.

[0247] IV. Configuration verification and clinical effects

[0248] Actual application test

[0249] Set monitoring points of 15 points / cm at the patient's wound area A and the diffusion path 2 Automatically collect multispectral images every 30 minutes and transmit them wirelessly to the medical staff workstation.

[0250] During a certain monitoring, the system found that the bacterial density in area A increased by 40% within 2 hours (exceeding the normal threshold by 20%), immediately triggering an early warning, and the medical staff promptly adjusted the antibiotic dosage.

[0251] Effect comparison

[0252]

[0253]

[0254] Dynamic adjustment mechanism: If the subsequent bacterial risk is upgraded to high risk (such as spreading to healthy tissues), the sampling interval can be temporarily shortened to 15 minutes, and the monitoring density can be encrypted to 20 points / cm 2 , and return to the regular configuration after the risk is downgraded to achieve flexible allocation of resources.

[0255] V. Summary

[0256] Through optimizing the model and algorithm solution, in this example, a combined solution of "medium sampling frequency + relatively high spectral accuracy + regional differential monitoring + wireless transmission" was determined, which not only meets the accurate monitoring requirements of the bacterial dynamics of medium-risk wounds but also avoids cost waste caused by over-configuration. This method can be extended to other types of wounds (such as postoperative infection wounds, chronic ulcers), and by adjusting the parameter weights (such as increasing the risk weight α of patients with high infection risks), personalized monitoring scheme design can be achieved.

[0257] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0258] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring wound flora, characterized in that, It includes the following steps: Obtain the physiological characteristic parameters of the wound and real-time flora monitoring data; Obtain the multispectral imaging data of the wound and perform preprocessing, then construct a three-dimensional flora distribution model, perform spectral feature analysis on the three-dimensional flora distribution model and combine the physiological characteristic parameters to obtain the initial flora dynamic model of the wound; Apply local metabolic conditions and external environmental conditions to the initial flora dynamic model to obtain a comprehensive flora dynamic model, and combine the physiological characteristic parameters to obtain flora dynamic coupling data; Construct a flora classification pre-training model and train it with the flora dynamic coupling data to obtain a flora classification model, and then obtain flora abnormal characteristic parameters; Obtain flora abnormal evolution simulation information through the flora abnormal characteristic parameters, the comprehensive flora dynamic model and the physiological characteristic parameters; Based on the flora abnormal evolution simulation information and the monitoring device parameters, construct an optimization model and optimize the monitoring device parameters to obtain the optimal monitoring configuration of the monitoring device; Infer the current period monitoring configuration based on the risk grading model to obtain the current period risk grading prediction value, and adjust the current period actual monitoring strategy based on the current period risk grading prediction value, the current period optimal monitoring configuration and the current period monitoring configuration to realize the monitoring of the wound flora state.

2. The method for monitoring the wound flora according to claim 1, wherein The flora dynamic coupling data at least includes a flora density distribution field, a metabolic activity field and a set of high-risk flora positions; the flora abnormal characteristic parameters at least include abnormal flora positions, abnormal flora types and abnormal flora morphological parameters; the flora abnormal evolution simulation information at least includes flora diffusion trends, maximum metabolic activities and maximum density changes.

3. The method for monitoring wound flora according to claim 1, characterized in that, The step of obtaining the multispectral imaging data of the wound and performing preprocessing, then constructing a three-dimensional flora distribution model, performing spectral feature analysis on the three-dimensional flora distribution model and combining the physiological characteristic parameters to obtain the initial flora dynamic model of the wound includes the following steps: Perform multi-band scanning on the wound surface to obtain a sequence of multispectral imaging data of different bands of the wound; Perform preprocessing on the multispectral imaging data sequence to obtain a preprocessed multispectral imaging data sequence, where the preprocessing includes one or more of denoising, standardization, band fusion, abnormal signal elimination, data interpolation and spectral smoothing; Based on the preprocessed multispectral imaging data sequence, combine three-dimensional reconstruction technology to obtain a corresponding three-dimensional flora distribution model; Perform spectral feature analysis on the three-dimensional flora distribution model to obtain the initial flora dynamic model of the wound, where the spectral feature analysis at least includes region segmentation and feature parameter mapping, and different metabolic activity sub-regions are formed through region segmentation, and corresponding physiological characteristic parameters are assigned to different sub-regions.

4. The method for monitoring wound flora according to claim 1, characterized in that, The step of applying local metabolic conditions and external environmental conditions to the initial flora dynamic model to obtain a comprehensive flora dynamic model, and combining the physiological characteristic parameters to obtain flora dynamic coupling data includes the following steps: Apply local metabolic conditions to the initial flora dynamic model to achieve metabolic boundary constraints and obtain a first flora dynamic model; Apply external environmental conditions to the first flora dynamic model to obtain a comprehensive flora dynamic model, where the external environmental conditions include temperature conditions and humidity conditions; Based on the comprehensive microbial community dynamic model and physiological characteristic parameters, microbial community dynamic coupling data is obtained. The specific process includes: Construct a dynamic equilibrium equation. The dynamic equilibrium equation at least includes a metabolic rate equation, a microbial community diffusion equation, and an activity correlation equation. Combining with the comprehensive microbial community dynamic model, the dynamic equilibrium equation is solved by a discretization method to obtain a microbial community density distribution field, a metabolic activity field, and a set of positions of high-risk microbial communities.

5. The method for monitoring wound flora according to claim 4, wherein, Obtaining the set of positions of the high-risk microbial communities includes the following steps: Based on the metabolic activity field, obtain the average metabolic activity value of each sub-region; Identify the sub-regions in the comprehensive microbial community dynamic model where the average metabolic activity value is not less than the activity threshold to obtain a set of positions of high-risk microbial communities.

6. The method for monitoring the wound flora according to claim 1, characterized in that, Constructing a microbial community classification pre-training model and training it with the microbial community dynamic coupling data to obtain a microbial community classification model, and then obtaining microbial community abnormal characteristic parameters, including the following steps: Construct a microbial community classification pre-training model based on a spectral feature classification network; Train and validate the microbial community classification pre-training model with the microbial community dynamic coupling data to obtain a microbial community classification model; Input the microbial community dynamic coupling data to be evaluated into the microbial community classification model to obtain a predicted set of positions of high-risk microbial communities; Based on the predicted set of positions of high-risk microbial communities, obtain microbial community abnormal characteristic parameters. Among them, the microbial community abnormal characteristic parameters at least include the position of abnormal microbial communities, the type of abnormal microbial communities, and the morphological parameters of abnormal microbial communities.

7. The method for monitoring wound flora according to claim 6, wherein Based on the predicted set of positions of high-risk microbial communities, obtaining the microbial community abnormal characteristic parameters includes the following steps: Calculate the centroid coordinates of the predicted set of positions of high-risk microbial communities to obtain the position of abnormal microbial communities; Fit the contour of abnormal microbial communities according to the distribution pattern of the predicted set of positions of high-risk microbial communities to obtain the morphological parameters of abnormal microbial communities. Among them, the morphological parameters of abnormal microbial communities include the shape of abnormal microbial communities and the coverage range of abnormal microbial communities; Calculate the metabolic activity direction of the predicted set of positions of high-risk microbial communities to obtain the weighted average value of the metabolic activity direction, and convert the weighted average value into a standard metabolic vector, and the standard metabolic vector is the type of abnormal microbial communities.

8. The method for monitoring wound flora according to claim 1, characterized in that, Through the microbial community abnormal characteristic parameters, the comprehensive microbial community dynamic model, and the physiological characteristic parameters, obtaining microbial community abnormal evolution simulation information includes the following steps: Use the position of abnormal microbial communities as the initial diffusion point, map the morphological parameters of abnormal microbial communities into the comprehensive microbial community dynamic model, and then adjust the diffusion direction of the comprehensive microbial community dynamic model to the type of abnormal microbial communities and refine the grid of the abnormal region to update the comprehensive microbial community dynamic model, and define the diffusion trigger condition, diffusion path, and diffusion rate to obtain a microbial community abnormal evolution dynamic model; Based on the microbial community abnormal evolution dynamic model, perform simulation analysis on the dynamic equilibrium equation through a recursive calculation method. Based on the results of each recursion in the analysis process, obtain microbial community abnormal evolution simulation information. The analysis process is the microbial community diffusion simulation process, specifically including: When the diffusion trigger condition is met, microbial community diffusion occurs to obtain the current microbial community diffusion trend, diffusion rate, diffusion path, metabolic activity field, and density distribution field, and then update the microbial community abnormal evolution dynamic model; Based on the updated dynamic model of abnormal microbiota evolution, the dynamic equilibrium equation is re-solved until the termination condition is satisfied, and then the microbiota diffusion simulation process is terminated; Based on the metabolic activity field and density distribution field of each recursion, the maximum metabolic activity and the maximum density change are obtained; The diffusion trigger condition includes that microbiota diffusion is carried out if the current metabolic accumulation value is not less than the metabolic threshold; The diffusion path is determined based on the historical diffusion direction and the current metabolic activity direction; The microbiota diffusion trend and the diffusion rate are determined based on the microbiota characteristic parameters and the environmental change parameters respectively.

9. The method for monitoring wound flora according to claim 1, characterized in that, Based on the microbiota abnormal evolution simulation information and the monitoring device parameters, an optimization model is constructed and the monitoring device parameters are optimized to obtain the optimal monitoring configuration of the monitoring device, including the following steps: Construct an optimization model, where the optimization model at least includes an optimization variable model, an optimization objective model and a constraint condition model. The optimization variable model is constructed based on the monitoring device parameters, and the monitoring device parameters, which are the optimization variables, at least include the sampling interval, spectral resolution, monitoring area density and data transmission mode. The optimization objective model is constructed based on the first risk classification value and the monitoring cost value. The first risk classification value is obtained based on the microbiota abnormal evolution simulation information, and the monitoring cost value at least includes the device power consumption cost value, calibration cost value and storage cost value. The constraint condition model includes the monitoring device parameter constraint condition model and the monitoring sensitivity constraint condition model. The monitoring sensitivity constraint condition model is constructed based on data acquisition integrity, data real-time and device compatibility; The optimization model is preliminarily solved by the genetic algorithm to obtain a preliminary optimized monitoring configuration; The optimization model is globally solved by the particle swarm algorithm, and the optimal solution set of the particle swarm is the optimal monitoring configuration of the monitoring device.

10. A monitoring system for wound flora, characterized in that, Including: A data acquisition module for obtaining the physiological characteristic parameters of the wound, real-time microbiota monitoring data and multi-spectral imaging data; A model construction and analysis module for preprocessing the multi-spectral imaging data, constructing a three-dimensional microbiota distribution model, performing spectral feature analysis on the three-dimensional microbiota distribution model and combining the physiological characteristic parameters to obtain an initial microbiota dynamic model of the wound, applying local metabolic conditions and external environmental conditions to the initial microbiota dynamic model to obtain a comprehensive microbiota dynamic model, and combining the physiological characteristic parameters to obtain microbiota dynamic coupling data; A microbiota classification module for constructing a pre-trained microbiota classification model and training it with the microbiota dynamic coupling data to obtain a microbiota classification model, and further obtaining microbiota abnormal characteristic parameters; A simulation analysis module for obtaining microbiota abnormal evolution simulation information through the microbiota abnormal characteristic parameters, the comprehensive microbiota dynamic model and the physiological characteristic parameters; A configuration optimization module for constructing an optimization model based on the microbiota abnormal evolution simulation information and the monitoring device parameters and optimizing the monitoring device parameters to obtain the optimal monitoring configuration of the monitoring device; A risk regulation module is used to infer the current period's monitoring configuration based on a risk classification model, obtain the risk classification prediction value for the current period, and adjust the actual monitoring strategy for the current period based on the risk classification prediction value for the current period, the optimal monitoring configuration for the current period, and the monitoring configuration for the current period, so as to monitor the state of the wound flora.

Images