Early infection prediction method and system for child burn wound

Abstract

The invention discloses a method and system for predicting early infection of a child burn wound, and the method comprises the steps: S1, obtaining the information of the child burn wound, and monitoring the local temperature, pH value, specific inflammatory marker level and original data of pathogen nucleic acid of the burn wound; s2, generating a trend change comprehensive model by using a time sequence analysis algorithm; s3, key features are extracted through an adaptive feature selection technology trend change comprehensive model, and data dimensions are reduced; s4, adopting a transfer learning method to perform pre-training in the burn case as a risk assessment model, and utilizing the risk assessment model to generate an infection risk score; and S5, setting a risk threshold value, and when the infection risk score exceeds a preset threshold value, performing early warning, marking child wound infection, generating a detailed infection report, and performing targeted treatment. According to the method, the problem that a high-performance prediction model is difficult to directly construct due to relatively small child medical data volume and complexity is solved, and the ability of early infection recognition of children is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of children's burns, and in particular to a method and system for predicting early infection of children's burn wounds. Background Art

[0002] Infection is an important factor leading to complications and prolonged healing time during the treatment of pediatric burns. Since children's immune systems are more fragile than those of adults, early and accurate infection monitoring of burn wounds is particularly important. Traditionally, doctors rely mainly on clinical symptom observation, laboratory tests, and visual inspections during regular dressing changes to determine whether there is infection. This method has certain limitations, such as relying on the doctor's experience and subjective bias; and the time interval between the appearance of signs of infection and diagnosis is long, resulting in missing the best time for treatment.

[0003] With the development of technology, some existing technical solutions attempt to improve the effect of infection monitoring through non-invasive means. Although the monitoring efficiency can be improved to a certain extent by observing body temperature and physical condition of children, the sensitivity and specificity are limited when used, making it difficult to fully and accurately predict the risk of infection. In addition, existing monitoring methods often require frequent manual intervention, which increases the workload of medical staff and is prone to misdiagnosis or monitoring errors for child patients.

[0004] Therefore, the present invention proposes an innovative solution to the current problems in monitoring infection in children's burn wounds. By comprehensively using a variety of sensing technologies and intelligent algorithms, the ability to identify infection at an early stage is significantly improved, providing strong support for timely and effective intervention measures. This not only helps to reduce the incidence of severe infection, but also speeds up the recovery process. Summary of the invention

[0005] The present invention provides a method for predicting early infection of burn wounds in children, comprising:

[0006] S1. Obtain images and spectral information of children's burn wounds, and monitor the local temperature, pH value, levels of specific inflammatory markers, and raw data of pathogen nucleic acids of burn wounds in real time;

[0007] S2. Preprocess the raw data and use the time series analysis algorithm to generate a comprehensive model of temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend and pathogen nucleic acid change trend;

[0008] S3, extract key features from the comprehensive model of temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend and pathogen nucleic acid change trend through adaptive feature selection technology to reduce data dimension;

[0009] S4. Adopt the transfer learning method to pre-train in burn cases as a risk assessment model, and generate an infection risk score using the risk assessment model by extracting key features.

[0010] S5. Set a risk threshold. When the infection risk score exceeds the preset threshold, give an alarm, mark the child's wound infection, generate a detailed infection report, and implement targeted treatment.

[0011] Preferably, in S2, the time series analysis algorithm is used to generate the temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend, and pathogen nucleic acid change trend, specifically including:

[0012] Taking the real-time monitored temperature T(t), pH value P(t), inflammatory marker concentration I(t), and pathogen nucleic acid level N(t) as time series data, where t represents time, and establishing a trend model for the time series data through the autoregressive moving average model. The expression of the trend model is X(t) = c + φ1X(t - 1) + φ2X(t - 2) + … + φ p X(t - p) + θ1∈(t - 1) + θ2∈(t - 2) + … + θ q ∈(t - q) + + ∈(t), where X(t) represents the time series data, c is a constant term, φ i and θ j are the autoregressive coefficient and the moving average coefficient respectively, p and q are the orders of autoregression and moving average respectively, and ∈(t) is the white noise error term; through the trend model, predict the temperature change trend T'(t), pH value fluctuation trend P'(t), inflammatory marker concentration change trend I'(t), and pathogen nucleic acid change trend N'(t) at future time points t + 1, t + 2, …, t + n.

[0013] Preferably, in S3, the adaptive feature selection technology is used to extract key features from the temperature change trend T'(t), pH value fluctuation trend P'(t), inflammatory marker concentration change trend I'(t), and pathogen nucleic acid change trend N'(t), reducing the data dimension, specifically including:

[0014] Represent the above time series data in vector form X = [T'(t), P'(t), I'(t), N'(t)], and use the adaptive feature selection technology of the autoencoder to map the high-dimensional input vector X to the low-dimensional feature space Z through the encoder function f(X; θ), and the formula is: Z = f(X; θ), where θ is the encoder parameter; through the decoder function Reconstruct the original data where is the decoder parameter, and optimize the model parameters θ and by minimizing the reconstruction error Extract the low-dimensional feature Z as the key feature G, and the formula is

[0015] Preferably, in S4, a risk assessment model is pre-trained using transfer learning method in a large number of burn cases, specifically including:

[0016] Construct a source domain dataset D with a large number of burn cases s , and the formula is: where is the feature vector of the temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend, and pathogen nucleic acid change trend in the source domain, is the infection label, and N s is the number of samples in the source domain; pre-train a deep neural network model M on the source domain dataset D s , and use the pre-trained model M s as the initial model, and fine-tune it on the target domain dataset s to adapt to a specific pediatric burn wound infection prediction task, where is the feature vector in the target domain, is the corresponding infection label, and N is the number of samples in the target domain, and D t is the target domain dataset; during the fine-tuning process, use the loss function t for optimization, where M is the fine-tuned model, l is the classification loss function, R(M t ) is the regularization term, and λ is the regularization coefficient, and the obtained risk assessment model M t generates an infection risk score for the key feature G t For the key feature G, generate an infection risk score

[0017] Preferably, the infection risk score is generated using the risk assessment model with the extracted key feature G. The key feature vector G = [z1, z2,... z k ,] extracted by the adaptive feature selection technique is used as the input, where z i represents the i-th key feature; then, use the pre-trained and fine-tuned risk assessment model M t to process the key feature and generate an infection risk score S. The formula is S = M t (G), where M t is a deep neural network model pre-trained on a large number of burn cases and fine-tuned on specific pediatric burn wound data. The early infection situation of pediatric burn wounds can be quickly judged through the output infection risk score S

[0018] Preferably, by setting a risk threshold T, when the infection risk score S exceeds the preset threshold, an alarm is triggered and the child's wound infection is marked. The risk threshold T is determined based on clinical experience and historical data. The generated infection risk score S is compared with the preset risk threshold T. If S > T, an alarm is triggered, and an alarm message is sent to the designated burn medical staff. The alarm message includes a detailed wound status report of the child, current monitoring data, historical trend analysis, and the infection risk score level.

[0019] An early infection prediction system for children's burn wounds, including an intelligent wound monitoring patch system;

[0020] The intelligent wound monitoring patch system includes a micro temperature sensor module, a pH sensor module, an inflammatory marker detection microfluidic chip module, a pathogen nucleic acid detection module, a micro spectral imaging module, and an early warning module;

[0021] The micro temperature sensor module monitors the local temperature change of the burn wound in real time and transmits the temperature data to the early warning system through low-power wireless communication technology;

[0022] The pH sensor module monitors the pH value of the wound secretion of the burn wound in real time and transmits the pH value data to the early warning system through low-power wireless communication technology;

[0023] The inflammatory marker detection microfluidic chip module detects specific inflammatory markers in the wound exudate through an electrochemical sensor or an optical sensor in the microfluidic chip, including but not limited to interleukin-6 and procalcitonin, and transmits the data to the early warning system through low-power wireless communication technology;

[0024] The pathogen nucleic acid detection module detects the pathogen nucleic acid in the wound exudate in real time, identifies the pathogen, and transmits the detection result to the early warning system through low-power wireless communication technology;

[0025] The micro spectral imaging module acquires high-resolution images and spectral information of the wound, captures color changes and tissue structure changes of the wound through spectral imaging, and assists in the early diagnosis of infection;

[0026] The early warning module processes the received data, determines whether the infection risk score exceeds the preset threshold, and sends the alarm message to the designated medical staff through wireless communication.

[0027] Preferably, the intelligent wound monitoring patch system further includes a biometric sensor module that monitors the unique heart rate variability HRV and skin conductance response GSR physiological parameters of children. The heart rate variability HRV and skin conductance response GSR physiological parameters reflect the state of the autonomic nervous system and stress response of children, and are used to evaluate the physiological condition and immune response of children.

[0028] Preferably, the early warning system aggregates the age, weight, immune status, and nutritional status of children to generate basic initial information, assisting the prediction system to conduct targeted treatment according to the current situation.

[0029] Preferably, when the infection risk score in the early warning module exceeds a preset threshold, the CRISPR-Cas module built into the patch is automatically triggered to precisely detect the infection factor, prompting the burn medical staff to conduct blood culture or imaging examination; the target sequence in the CRISPR-Cas module detects the pathogens in the children's burn wounds.

[0030] Compared with the prior art, the technical solution of this application has the following technical effects:

[0031] Through the intelligent wound monitoring patch system, the present invention solves the problems of subjective deviation caused by relying on doctors' experience judgment in traditional methods and the long time interval between observation and diagnosis; the system can real-time monitor the changes in the local temperature, pH value, specific inflammatory marker levels, and pathogen nucleic acids of children's burn wounds, significantly improving the speed and accuracy of infection early warning, helping to take intervention measures in a timely manner, thereby reducing the risk of complications caused by delayed diagnosis.

[0032] The present invention uses a time series analysis algorithm to generate a comprehensive model, solving the problem of insufficient sensitivity and specificity when using only one indicator, such as temperature or pH value, for monitoring in the prior art. By modeling the time series change trends of multi-dimensional data, it not only improves the comprehensiveness of infection prediction but also can more accurately identify the subtle changes in the early stage of infection, enhancing the correlation analysis ability between different physiological parameters and providing a more reliable basis for clinical decision-making.

[0033] The present invention uses adaptive feature selection technology to extract key features, effectively addressing the problem of excessive data dimensions faced when dealing with a large amount of complex data. This method reduces the unnecessary computational burden in the data processing process, ensuring that the infection risk assessment task can be efficiently executed even in resource-constrained environments. By screening out the most representative features, it further enhances the model's ability to capture potential infection factors and improves the performance and response speed of the overall monitoring system.

[0034] The present invention pre-trains the risk assessment model through transfer learning method, achieving a breakthrough in solving the problem that it is difficult to directly build a high-performance prediction model due to the relatively small amount of children's medical data. Based on a large number of burn cases, the model is quickly adjusted to meet the special needs of children's burn wounds, not only significantly reducing the sample quantity requirements for establishing a dedicated model and accelerating the development cycle, but also improving the model's generalization ability for new cases, ensuring the reliability and consistency of the infection risk score.

[0035] The above description is only an overview of the technical solution of the present application. In order to better understand the technical means of the present application, it can be implemented according to the content of the specification. Moreover, in order to make the above and other purposes, features, and advantages of the present application more obvious and understandable, the following describes the preferred embodiments of the present application in detail in conjunction with the accompanying drawings.

[0036] Those skilled in the art will better understand the above and other purposes, advantages, and features of the present application according to the following detailed description of the specific embodiments of the present application in conjunction with the accompanying drawings. Description of the Drawings

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or in the prior art, the following briefly introduces the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to actual scale.

[0038] Figure 1 It is a flowchart of a method for early infection prediction of children's burn wounds in the present invention;

[0039] Figure 2 It is a risk assessment flowchart of a method for early infection prediction of children's burn wounds in the present invention;

[0040] Figure 3 It is a risk scoring flowchart of a method for early infection prediction of children's burn wounds in the present invention;

[0041] Figure 4 It is a structural diagram of a system for early infection prediction of children's burn wounds in the present invention. Detailed Embodiments

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. In the following description, specific details such as specific configurations and components are provided only to assist in a comprehensive understanding of the embodiments of the present application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described here without departing from the scope and spirit of the present application. Additionally, descriptions of known functions and structures are omitted in the embodiments for clarity and conciseness.

[0043] It should be understood that the "one embodiment" or "this embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, the "one embodiment" or "this embodiment" that appears throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner.

[0044] In addition, the present application may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not in itself indicate the relationship between the various embodiments and / or arrangements discussed.

[0045] The term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another association object relationship, indicating that there can be two relationships. For example, A / and B can represent: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0046] The term "at least one" in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, at least one of A and B can represent: A exists alone, A and B exist simultaneously, and B exists alone.

[0047] It should also be noted that in this article, 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 such actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion.

[0048] Embodiment 1

[0049] This embodiment mainly describes a method for early infection prediction of children's burn wounds, as Figure 1 shown, including:

[0050] S1. Obtain children's burn wound images and spectral information, and real-time monitor the original data of the local temperature, pH value, specific inflammatory marker level and pathogen nucleic acid of the burn wound;

[0051] S2. Preprocess the original data, and use the time series analysis algorithm to generate a comprehensive model of the temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend and pathogen nucleic acid change trend;

[0052] S3. Extract key features from the comprehensive model of temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend, and pathogen nucleic acid change trend through adaptive feature selection technology to reduce the data dimension;

[0053] S4. Adopt the transfer learning method to pre-train in burn cases as a risk assessment model. Through the extracted key features, use the risk assessment model to generate an infection risk score;

[0054] S5. Set a risk threshold. When the infection risk score exceeds the preset threshold, give an early warning, mark the child's wound infection, generate a detailed infection report, and implement targeted treatment.

[0055] Further, in S2, the time series analysis algorithm is used to generate the temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend, and pathogen nucleic acid change trend. Specifically, it includes:

[0056] Take the real-time monitored temperature T(t), pH value P(t), inflammatory marker concentration I(t), and pathogen nucleic acid level N(t) as time series data, where t represents time. Use the autoregressive moving average model to establish a trend model for the time series data. The expression of the trend model is X(t) = c + φ1X(t - 1) + φ2X(t - 2) + … + φ p X(t - p) + θ1∈(t - 1) + θ2∈(t - 2) + … + θ q ∈(t - q) + + ∈(t), where X(t) represents the time series data, c is the constant term, φ i and θ j are the autoregressive coefficient and the moving average coefficient respectively, p and q are the orders of autoregression and moving average respectively, and ∈(t) is the white noise error term; Through the trend model, predict the temperature change trend T'(t), pH value fluctuation trend P'(t), inflammatory marker concentration change trend I'(t), and pathogen nucleic acid change trend N'(t) at future time points t + 1, t + 2, …, t + n.

[0057] Further, extract key features from the temperature change trend T'(t), pH value fluctuation trend P'(t), inflammatory marker concentration change trend I'(t), and pathogen nucleic acid change trend N'(t) through adaptive feature selection technology to reduce the data dimension. Specifically, it includes:

[0058] Represent the above time series data in vector form X = [T'(t), P'(t), I'(t), N'(t)]. Use the adaptive feature selection technology of the autoencoder. Through the encoder function f(X; θ), map the high-dimensional input vector X to the low-dimensional feature space Z. The formula is: Z = f(X; θ), where θ is the encoder parameter; Through the decoder function Reconstruct the original data Among them is the decoder parameter, by minimizing the reconstruction error Optimize the model parameters θ and Extract the low-dimensional feature Z as the key feature G, and the formula is

[0059] Furthermore, as Figure 2 shown, in S4, a risk assessment model is pre-trained in a large number of burn cases through transfer learning method, specifically including:

[0060] Construct a source domain dataset D through a large number of burn cases s , the formula is: Among them is the feature vector of the temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend and pathogen nucleic acid change trend in the source domain, is the infection label, N s is the number of source domain samples; pre-train a deep neural network model M on the source domain dataset D s , and use the pre-trained model M s as the initial model, and fine-tune it on the target domain dataset s to adapt to a specific child burn wound infection prediction task, where is the feature vector in the target domain, is the corresponding infection label, N is the number of target domain samples, D s is the target domain dataset; during the fine-tuning process, use the loss function t for optimization, where M is the fine-tuned model, l is the classification loss function, R(M t ) is the regularization term, λ is the regularization coefficient, and the obtained risk assessment model M t generates an infection risk score for the key feature G t

[0061] Furthermore, as Figure 3 shown, the infection risk score is generated by using the risk assessment model with the extracted key feature G, and the key feature vector G = [z1, z2,... z k i ,] extracted by the adaptive feature selection technology is used as the input, where z t represents the i-th key feature; then, use the pre-trained and fine-tuned risk assessment model M t to process the key feature and generate an infection risk score S, and the formula is S = M t (G), where M tA deep neural network model pre-trained on a large number of burn cases and fine-tuned on specific pediatric burn wound data can quickly determine the early infection status of pediatric burn wounds through the output infection risk score S.

[0062] Furthermore, by setting a risk threshold T, when the infection risk score S exceeds the preset threshold, an early warning is issued and the pediatric wound infection is marked. The risk threshold T is determined based on clinical experience and historical data. The generated infection risk score S is compared with the preset risk threshold T. If S > T, an alarm is triggered, and an alarm message is sent to the designated burn medical staff; the alarm message includes a detailed wound status report of the child, current monitoring data, historical trend analysis, and the infection risk score level.

[0063] This embodiment realizes a highly accurate prediction of the infection risk of pediatric burn wounds by collecting and analyzing multi-dimensional data including temperature, pH value, specific inflammatory marker levels, and pathogen nucleic acids in real time, using time series analysis and adaptive feature selection, combined with a risk assessment model of transfer learning, improves the speed and accuracy of infection early warning, reduces unnecessary medical operations, and alleviates the pain of patients. In addition, by adopting a miniaturized design and low-power wireless communication technology, the entire monitoring process is made more convenient and efficient, while reducing the cost of long-term monitoring.

[0064] Embodiment 2

[0065] This embodiment is based on Embodiment 1 and details a system for early infection prediction of pediatric burn wounds, as Figure 4 shown, specifically including an intelligent wound monitoring patch system;

[0066] The intelligent wound monitoring patch system includes a micro temperature sensor module, a pH value sensor module, a microfluidic chip module for detecting inflammatory markers, a pathogen nucleic acid detection module, a micro spectral imaging module, and an early warning module;

[0067] The micro temperature sensor module monitors the local temperature change of the burn wound in real time and transmits the temperature data to the early warning system through low-power wireless communication technology;

[0068] The pH value sensor module monitors the pH value of the secretions of the burn wound in real time and transmits the pH value data to the early warning system through low-power wireless communication technology;

[0069] The microfluidic chip module for detecting inflammatory markers detects specific inflammatory markers in the wound exudate through an electrochemical sensor or an optical sensor in the microfluidic chip, including but not limited to interleukin-6 and procalcitonin, and transmits the data to the early warning system through low-power wireless communication technology;

[0070] The pathogen nucleic acid detection module detects the pathogen nucleic acid in the wound exudate in real time, identifies the pathogen, and transmits the detection result to the warning system through low-power wireless communication technology;

[0071] The micro-spectral imaging module acquires high-resolution images and spectral information of the wound surface, captures the color changes and tissue structure changes of the wound surface through spectral imaging, and assists in the early diagnosis of infection;

[0072] The warning module processes the received data, judges whether the infection risk score exceeds the preset threshold, and sends the alarm information to the designated medical staff through wireless communication.

[0073] Furthermore, the intelligent wound monitoring patch system also includes a biometric sensor module, which monitors the physiological parameters of heart rate variability HRV and galvanic skin response GSR unique to children. The physiological parameters of heart rate variability HRV and galvanic skin response GSR reflect the state of the autonomic nervous system and stress response of children, and evaluate the physiological condition and immune response of children.

[0074] Furthermore, the warning system aggregates the age, weight, immune status, and nutritional status of children to generate basic initial information, and assists the prediction system to carry out targeted treatment according to the current situation.

[0075] Furthermore, in the warning module, when the infection risk score exceeds the preset threshold, the built-in CRISPR-Cas module of the patch is automatically triggered to perform precise detection of infection factors, prompting the burn medical staff to perform blood culture or imaging examinations; the target sequence in the CRISPR-Cas module detects the pathogens on the children's burn wounds.

[0076] This embodiment details an early infection prediction system for children's burn wounds. By using collaborative monitoring among multiple modules, it solves the problems of subjective deviation caused by relying on doctors' experience judgment in traditional methods and the long time interval from observation to diagnosis. The system can real-time monitor the changes in local temperature, pH value, specific inflammatory marker levels, and pathogen nucleic acid of children's burn wounds, significantly improving the speed and accuracy of infection warning, helping to take intervention measures in a timely manner, and thus reducing the risk of complications caused by delayed diagnosis.

[0077] Embodiment 3

[0078] Based on Embodiment 1, this embodiment details an optimization scheme for an early infection prediction system for children's burn wounds, specifically including:

[0079] This technical solution provides a wearable Smart Wound Monitoring Patch (SWMP), which is integrated with a micro temperature sensor module, a pH sensor module, an inflammatory marker detection microfluidic chip module, and a pathogen nucleic acid rapid detection module; in addition, the patch also includes a high-resolution micro-spectral imaging module capable of obtaining high-resolution images and spectral information of the wound in real time.

[0080] The patch can monitor the local temperature, pH value, specific inflammatory marker levels (such as interleukin-6 (IL-6) and procalcitonin (PCT)), pathogen nucleic acids, and wound images of the burn wound in real time, and transmit these multi-parameter data to the warning module through low-power wireless communication technology;

[0081] In particular, the design of the patch takes into account the sensitivity and comfort of children's skin, is made of soft, breathable and non-irritating materials, and the size and shape of the patch are suitable for children of different ages; the patch also has a biometric sensor dedicated to children for monitoring children-specific physiological parameters, such as heart rate variability (HRV) and galvanic skin response (GSR), which do not have the same sensitivity and specificity in adults.

[0082] The warning module receives data from the Smart Wound Monitoring Patch and performs real-time analysis using an integrated multi-modal artificial intelligence algorithm. The multi-modal artificial intelligence algorithm preprocesses the monitored data, including noise removal, data smoothing, and normalization, and then uses an adaptive feature selection technique to extract key features from the temperature change trend, pH value fluctuation, inflammatory marker concentration change, presence of pathogen nucleic acids, and spectral information of the wound image;

[0083] In particular, through the transfer learning method, a model trained on a large number of pediatric burn cases is used for feature extraction to improve the accuracy and generalization ability of infection prediction, taking into account children-specific physiological parameters and the characteristics of the immune system development, such as age, weight, immune status, nutritional status, heart rate variability (HRV) and galvanic skin response (GSR), etc., to optimize the prediction results. The performance and significance of the above parameters in adults are different from those in children, so the above technology is particularly applicable to children.

[0084] Based on the extracted key features, a comprehensive infection risk score is generated. The score not only considers the changes in individual parameters but also combines the interrelationships among multiple parameters and their dynamic change trends. Further, the Q-learning method in reinforcement learning is used to optimize the prediction model, continuously adjusting the model parameters according to historical data and current data to achieve more accurate infection prediction. When the infection risk score exceeds the preset threshold, the warning module immediately generates personalized alarm information and sends it to the designated medical staff through a secure encrypted channel; at the same time, a detailed wound status report is displayed in the warning module, including current monitoring data, historical trend analysis, infection risk score, and recommended further examination measures, so that medical staff can respond quickly.

[0085] In particular, the alarm information and recommended further examination measures are customized according to the specific conditions of the child (such as age, immune status, nutritional status, HRV, and GSR, etc.) to ensure the accuracy of diagnosis.

[0086] To further confirm the infection situation, the warning module provides a quick confirmation mechanism. When the infection risk score exceeds the preset threshold, the application will automatically trigger the built-in CRISPR-Cas system for rapid detection of infection factors, which can provide the result of the presence or absence of pathogen nucleic acid within a few minutes, thus achieving rapid and accurate infection confirmation.

[0087] The warning module will also prompt medical staff to conduct necessary further examinations, such as blood culture or imaging examinations, to ensure the comprehensiveness and accuracy of diagnosis.

[0088] In particular, the targeting sequences in the CRISPR-Cas system are specifically designed for common pathogens in children's burn wounds, which are different from those in adult burn wounds, thus ensuring the child-specificity of this technical solution.

[0089] This embodiment details a breakthrough in the problem that it is difficult to directly build a high-performance prediction model due to the relatively small amount of medical data for children. Based on a large number of burn cases, the model is quickly adjusted to meet the special needs of children's burn wounds, not only significantly reducing the sample quantity requirements for building a dedicated model, accelerating the development cycle, but also enhancing the generalization ability of the model for new cases, ensuring the reliability and consistency of the infection risk score.

[0090] The above are only the preferred embodiments of the present invention, and thus do not limit the protection scope of the present invention. For those skilled in the art, the present invention may have various modifications and variations. All changes, modifications, substitutions, integrations, and parameter alterations made to these embodiments without departing from the principle and spirit of the present invention by means of conventional substitutions or capable of achieving the same functions fall within the protection scope of the present invention.

Claims

1. A method for early infection prediction of children's burn wounds, characterized in that, Including: S1. Obtain the images and spectral information of children's burn wounds, and monitor the original data of the local temperature, pH value, specific inflammatory marker levels, and pathogen nucleic acids of the burn wounds in real time; S2. Preprocess the original data, and use time series analysis algorithms to generate comprehensive models of temperature change trends, pH value fluctuation trends, inflammatory marker concentration change trends, and pathogen nucleic acid change trends; S3. Extract key features from the comprehensive models of temperature change trends, pH value fluctuation trends, inflammatory marker concentration change trends, and pathogen nucleic acid change trends through adaptive feature selection technology to reduce the data dimension; S4. Adopt the transfer learning method to pre-train in burn cases as a risk assessment model, and generate an infection risk score using the extracted key features through the risk assessment model; S5. Set a risk threshold. When the infection risk score exceeds the preset threshold, give an alarm, mark the child's wound infection, generate a detailed infection report, and implement targeted treatment.

2. The early infection prediction method for children's burn wounds according to claim 1, wherein In S2, using time series analysis algorithms to generate temperature change trends, pH value fluctuation trends, inflammatory marker concentration change trends, and pathogen nucleic acid change trends specifically includes: Taking the real-time monitored temperature T(t), pH value P(t), concentration of inflammatory markers I(t), and pathogen nucleic acid level N(t) as time series data, where t represents time, a trend model is established for the time series data through an autoregressive moving average model. The expression of the trend model is X(t) = c + φ1X(t - 1) + φ2X(t - 2) + … + φ p X(t - p) + θ1∈(t - 1) + θ2∈(t - 2) + … + θ q ∈(t - q) + + ∈(t), where X(t) represents the time series data, c is the constant term, φ i and θ j are the autoregressive coefficient and the moving average coefficient respectively, p and q are the orders of autoregression and moving average respectively, and ∈(t) is the white noise error term; Through the trend model, predict the temperature change trend T'(t), pH value fluctuation trend P'(t), concentration change trend of inflammatory markers I'(t), and pathogen nucleic acid change trend N'(t) at future time points t + 1, t + 2, ···, t + n.

3. A method for early infection prediction of children's burn wounds according to claim 1 or 2, characterized in that In S3, extracting key features from the temperature change trend T'(t), pH value fluctuation trend P'(t), inflammatory marker concentration change trend I'(t), and pathogen nucleic acid change trend N'(t) through adaptive feature selection technology to reduce the data dimension, specifically includes: Represent the above time series data in vector form X = [T'(t), P'(t), I'(t), N'(t)]. Using the adaptive feature selection technique of the autoencoder, map the high-dimensional input vector X to the low-dimensional feature space Z through the encoder function f(X; θ). The formula is: Z = f(X; θ), where θ is the encoder parameter; through the decoder function Reconstruct the original data where is the decoder parameter. By minimizing the reconstruction error Optimize the model parameters θ and Extract the low-dimensional feature Z as the key feature G. The formula is 4. A method for early infection prediction of children's burn wounds according to claim 1, characterized in that, In S4, generating a risk assessment model through pre-training in a large number of burn cases by the transfer learning method, specifically includes: Construct the source domain dataset D from a large number of burn cases s , the formula is: where is the feature vector of the temperature change trend, pH value fluctuation trend, inflammatory marker concentration change trend, and pathogen nucleic acid change trend in the source domain, is the infection label, and N s is the number of source domain samples; Pretrain a deep neural network model M s on the source domain dataset D s , and use the pre-trained model M s as the initial model to fine-tune on the target domain dataset to adapt to a specific pediatric burn wound infection prediction task, where is the feature vector in the target domain, is the corresponding infection label, and N t is the number of target domain samples, and D t is the target domain dataset; During the fine-tuning process, use the loss function for optimization, where M t is the fine-tuned model, l is the classification loss function, R(M t ) is the regularization term, and λ is the regularization coefficient, and the obtained risk assessment model M t generates an infection risk score for the key feature G.

5. A method for early infection prediction of children's burn wounds according to claim 4, characterized in that The infection risk score is generated by using the key feature G extracted through a risk assessment model. The key feature vector G = [z1, z2, ··· z k ,] obtained by the adaptive feature selection technique is used as the input, where z i represents the i-th key feature; Then, use the pre-trained and fine-tuned risk assessment model M t to process the key features and generate an infection risk score S, with the formula S = M t (G), where M t is a deep neural network model pre-trained on a large number of burn cases and fine-tuned on specific pediatric burn wound data. The early infection situation of pediatric burn wounds can be quickly judged through the output infection risk score S.

6. The method for early infection prediction of children's burn wounds according to claim 5, characterized in that, By setting a risk threshold T, when the infection risk score S exceeds the preset threshold, give an alarm and mark the child's wound infection. The risk threshold T is determined based on clinical experience and historical data. Compare the generated infection risk score S with the preset risk threshold T. If S>T, trigger an alarm and send an alarm message to the designated burn medical staff; the alarm message includes a detailed wound status report of the child, current monitoring data, historical trend analysis, and infection risk score level.

7. An early infection prediction system for children's burn wounds, characterized in that, Including an intelligent wound monitoring patch system; The intelligent wound monitoring patch system includes a micro temperature sensor module, a pH value sensor module, a microfluidic chip module for detecting inflammatory markers, a pathogen nucleic acid detection module, a micro spectral imaging module, and an alarm module; The micro temperature sensor module monitors the local temperature change of the burn wound in real time and transmits the temperature data to the alarm system through low-power wireless communication technology; The pH value sensor module monitors the pH value of the secretions of the burn wound in real time and transmits the pH value data to the alarm system through low-power wireless communication technology; The microfluidic chip module for detecting inflammatory markers detects specific inflammatory markers in the wound exudate through electrochemical sensors or optical sensors in the microfluidic chip, including but not limited to interleukin-6 and procalcitonin, and transmits the data to the alarm system through low-power wireless communication technology; The pathogen nucleic acid detection module detects the pathogen nucleic acid in the wound exudate in real time, identifies the pathogen, and transmits the detection result to the early warning system through low-power wireless communication technology; The micro-spectral imaging module acquires high-resolution images and spectral information of the wound surface, captures the color changes and tissue structure changes of the wound surface through spectral imaging, and assists in the early diagnosis of infection; The early warning module processes the received data, judges whether the infection risk score exceeds the preset threshold, and sends the alarm information to the designated medical staff through wireless communication.

8. A system for early infection prediction of children's burn wounds according to claim 7, characterized in that, The intelligent wound monitoring patch system further includes a biometric sensor module, which monitors the physiological parameters of heart rate variability HRV and galvanic skin response GSR unique to children. The physiological parameters of heart rate variability HRV and galvanic skin response GSR reflect the state of the autonomic nervous system and stress response of children, and evaluate the physiological condition and immune response of children.

9. The early infection prediction system for children's burn wounds according to claim 7, characterized in that, The early warning system aggregates the age, weight, immune status, and nutritional status of children to generate basic initial information, and assists the prediction system to conduct targeted treatment for the current situation.

10. A system for early infection prediction of children's burn wounds according to claim 7, characterized in that, In the early warning module, when the infection risk score exceeds the preset threshold, the built-in CRISPR-Cas module of the patch is automatically triggered to accurately detect the infection factor, prompting the burn medical staff to conduct blood culture or imaging examination; the target sequence in the CRISPR-Cas module detects the pathogens in the children's burn wounds.

Images