Hospital drug-resistant bacterium closed-loop monitoring method and system based on space-time trajectory fusion

Abstract

The invention discloses a spatio-temporal trajectory fusion-based closed-loop monitoring method and system for drug-resistant bacteria in a hospital. The method comprises the following steps: collecting and processing strain drug sensitive test results and patient clinical index data in a hospital information system, and screening out key features through an improved genetic algorithm; wherein the clinical index data of the patient comprises original spatio-temporal trajectory data and nursing operation records; analyzing spatio-temporal trajectory data of the patient according to the key features, combining with drug sensitivity spectrum consistency, and constructing a propagation network model by using a graph convolutional network to identify infection source nodes and corresponding topological influence; and generating an early warning signal based on the topological influence of the infection source node, and continuously optimizing prevention and control measures by dynamically adjusting model parameters and a real-time feedback mechanism. By implementing the method provided by the invention, accurate analysis and real-time prevention and control of a drug-resistant bacterium transmission link can be realized.

Description

Technical Field

[0001] This invention relates to data processing methods, and more specifically to a closed-loop monitoring method and system for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion. Background Technology

[0002] Hospital-acquired infections, also known as healthcare-associated infections, refer to any new infection that develops in a patient during the course of receiving healthcare services. Drug-resistant bacteria are particularly challenging in hospital settings because they are resistant to common antibiotics, making treatment more difficult and expensive.

[0003] Drug-resistant bacteria include carbapenem-resistant Enterobacteriaceae and methicillin-resistant Staphylococcus aureus (MRSA). These bacteria are resistant to multiple antibiotics, including carbapenems, the last line of defense. MRSA is a common drug-resistant bacterium that can spread rapidly within hospitals, causing serious health problems. Transmission routes for drug-resistant bacteria include direct contact, environmental colonization, and horizontal gene transfer. Direct contact refers to transmission through direct contact between patients, healthcare workers, and medical equipment. Environmental colonization refers to the survival of drug-resistant bacteria on surfaces in medical facilities, such as bed rails and ventilator tubing, becoming new sources of infection. Horizontal gene transfer refers to the transfer of drug-resistant genes between different bacteria via plasmid-mediated mechanisms, promoting the spread of resistance.

[0004] Traditional monitoring strategies, such as regular environmental hygiene testing and hand hygiene monitoring, while important, are insufficient in terms of timeliness and comprehensive coverage. These methods struggle to capture all potential transmission routes, especially drug-resistant bacterial transmission chains that are highly concealed and complex.

[0005] Therefore, it is necessary to design a new method to achieve accurate analysis and real-time control of drug-resistant bacterial transmission chains. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a closed-loop monitoring method and system for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a closed-loop monitoring method for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion, comprising:

[0008] Collect and process bacterial strain drug susceptibility test results and patient clinical indicator data from the hospital information system, and screen out key features using an improved genetic algorithm; among them, patient clinical indicator data includes raw spatiotemporal trajectory data and nursing operation records;

[0009] Based on the analysis of the patient's spatiotemporal trajectory data according to the key features and combined with the consistency of drug sensitivity spectrum, a propagation network model is constructed using graph convolutional networks to identify infection source nodes and their corresponding topological influence.

[0010] Early warning signals are generated based on the topological influence of the infection source nodes, and prevention and control measures are continuously optimized by dynamically adjusting model parameters and using a real-time feedback mechanism.

[0011] The further technical solution is as follows: The results of bacterial strain drug susceptibility testing and patient clinical indicator data from the hospital information system are collected and processed, and key features are screened out using an improved genetic algorithm, including:

[0012] The MIC values ​​of bacterial strains, sensitivity classification results, strain typing information, as well as patient invasive operation records, hospitalization process information, raw spatiotemporal trajectory data, and nursing operation records are collected through the hospital laboratory information system, electronic medical record system, and hospital information system to obtain bacterial strain drug susceptibility test results and patient clinical indicator data.

[0013] Missing values ​​in the drug susceptibility test results of the strains and the clinical indicator data of patients were processed and standardized to obtain consistent results;

[0014] When the MIC value in the unified result is less than or equal to the clinical breakpoint, linear normalization is performed; when the MIC value in the unified result is greater than the clinical breakpoint, it is converted into a logarithmic compressed value to obtain the encoding result.

[0015] The sensitivity classification results in the unified results are converted into three-dimensional one-hot encoded vector representations to obtain the conversion results;

[0016] Key features were selected from the patient's clinical indicator data using an improved genetic algorithm combined with a fitness function.

[0017] The further technical solution is as follows: the fitness function is Wherein, λ is the balance factor, and its value ranges from 0.2 to 0.5.

[0018] The further technical solution is as follows: Based on the analysis of patient spatiotemporal trajectory data according to the key features and combined with drug sensitivity spectrum consistency, a propagation network model is constructed using a graph convolutional network to identify infection source nodes and their corresponding topological influence, including:

[0019] Based on the aforementioned key features, events and trajectories are aligned using time windows to obtain alignment results;

[0020] Based on the alignment results, nodes including patients, equipment, wards, and nursing groups are defined, and edge weights are calculated based on spatiotemporal co-occurrence and drug sensitivity consistency to establish a propagation graph and form a propagation network model.

[0021] The adjacency matrix and node features of the propagation graph are converted into embedding vectors using a three-layer GCN model for identifying potential infection source nodes.

[0022] Calculate the topological influence of the nodes in the propagation graph in the network, and adjust the parameters according to the propagation medium to obtain the topological influence corresponding to the source node of infection.

[0023] The further technical solution is as follows: the edge weight is calculated as follows: ω ij =C 共现 (i,j) / (1+α·D(i,j))×PRI(i,j), where, C 共现 (i,j) refers to the number of times two nodes co-occur in the same spatiotemporal unit; α is the distance attenuation coefficient, with a value range of 0.05 to 0.1; ·D(i,j) refers to the physical distance, and PRI(i,j) refers to the similarity of the drug susceptibility spectra of the two strains.

[0024] Its further technical solution is: adopting Calculate the topological influence of the nodes in the propagation graph in the network, where D(u) is the degree of node u, L(u, v) is the shortest path length from node u to v, and β is the path decay factor; if the path includes medical equipment nodes, then β = 0.1; if the path only involves patient and ward nodes, then β = 0.05; if the path includes nursing staff nodes, then β = 0.15.

[0025] Its further technical solution is as follows: The early warning signal is generated based on the topological influence of the infection source node, and the prevention and control measures are continuously optimized through dynamic adjustment of model parameters and a real-time feedback mechanism, including:

[0026] Early warning levels are established based on the topological influence of the infection source nodes to generate early warning signals, and prevention and control measures are continuously optimized through dynamic adjustment of model parameters and real-time feedback mechanisms.

[0027] Its further technical solution is as follows: The early warning classification is based on the topological influence of the infection source node to generate an early warning signal, and the prevention and control measures are continuously optimized through dynamic adjustment of model parameters and a real-time feedback mechanism, including:

[0028] When the topological influence of the infection source node exceeds the first threshold, ward-level enhanced disinfection of equipment and contact isolation measures are immediately implemented; when the topological influence of the infection source node is between the second threshold and the first threshold, nursing team zoning operations are initiated and hand hygiene compliance monitoring is strengthened; when the topological influence of the infection source node is below the second threshold, environmental microbial sampling is conducted and high-risk beds are placed under closed management.

[0029] The weights of the drug sensitivity spectrum consistency index are adjusted based on the rate of decline in new cases after intervention, and the optimized parameters are fed back into the transmission network model to achieve continuous self-updating and improvement of the model.

[0030] The further technical solution is as follows: the first threshold is 0.8, the second threshold is 0.5, and when the hospital infection risk level increases, the threshold T' is reduced proportionally. i =T i ·(1-0.1·ΔL), where ΔL is the number of risk level increases; T i For either the first threshold or the second threshold, T' i This refers to either the adjusted first or second threshold.

[0031] This invention also provides an in-hospital closed-loop monitoring system for drug-resistant bacteria based on spatiotemporal trajectory fusion, comprising:

[0032] The collection and processing unit is used to collect and process the results of bacterial strain drug susceptibility tests and patient clinical indicator data from the hospital information system, and to screen out key features through an improved genetic algorithm; among them, patient clinical indicator data includes raw spatiotemporal trajectory data and nursing operation records.

[0033] The identification unit is used to analyze the patient's spatiotemporal trajectory data based on the key features and combine it with the consistency of the drug sensitivity spectrum, and use a graph convolutional network to construct a propagation network model to identify the infection source node and its corresponding topological influence.

[0034] The generation unit is used to generate early warning signals based on the topological influence of the infection source node, and continuously optimize prevention and control measures by dynamically adjusting model parameters and using a real-time feedback mechanism.

[0035] The advantages of this invention compared to existing technologies are as follows: This invention collects and processes bacterial strain drug susceptibility test results and patient clinical indicator data from hospital information systems, including raw spatiotemporal trajectories and nursing operation records. It uses an improved genetic algorithm to screen key features and, based on these features, constructs a propagation network model using a graph convolutional network. It analyzes the consistency of spatiotemporal trajectories among patients combined with drug susceptibility spectrum analysis to identify infection source nodes and their topological influence. Based on this information, it generates early warning signals and dynamically adjusts model parameters, optimizing prevention and control measures through a real-time feedback mechanism, thereby achieving accurate analysis and real-time control of drug-resistant bacterial transmission chains. This process integrates data analysis, machine learning, and real-time monitoring technologies, improving the accuracy and timeliness of identifying infection sources and implementing effective prevention and control measures.

[0036] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0037] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A schematic diagram illustrating the application scenario of the in-hospital drug-resistant bacteria closed-loop monitoring method based on spatiotemporal trajectory fusion provided in an embodiment of the present invention;

[0039] Figure 2 This is a flowchart illustrating the closed-loop monitoring method for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion provided in an embodiment of the present invention.

[0040] Figure 3 A schematic block diagram of an in-hospital drug-resistant bacteria closed-loop monitoring system based on spatiotemporal trajectory fusion provided in an embodiment of the present invention;

[0041] Figure 4 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0044] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0045] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0046] Please see Figure 1 and Figure 2 , Figure 1 This is a schematic diagram illustrating an application scenario of the in-hospital drug-resistant bacteria closed-loop monitoring method based on spatiotemporal trajectory fusion provided in an embodiment of the present invention. Figure 2 This is a schematic flowchart illustrating a closed-loop monitoring method for nosocomial drug-resistant bacteria based on spatiotemporal trajectory fusion, provided in an embodiment of the present invention. This method is applied to a server that interacts with terminals. By fusing strain drug susceptibility test results and patient clinical indicator data from the hospital information system, a modified genetic algorithm is used to screen key features. Based on these features, a propagation network model is constructed using a graph convolutional network to identify infection source nodes and their topological influence. Early warnings are issued based on the topological influence of infection source nodes, model parameters are dynamically adjusted, and real-time feedback is provided to optimize prevention and control measures. This achieves accurate analysis and real-time control of the nosocomial drug-resistant bacteria transmission chain. Specifically, the method includes steps such as data collection and processing, feature screening, establishing a propagation network model, calculating node topological influence, and generating early warning signals. By continuously iterating and updating model parameters to adapt to different hospital infection risk levels, the effectiveness and targeting of prevention and control measures are ensured.

[0047] Figure 2 This is a schematic flowchart of the closed-loop monitoring method for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion provided in an embodiment of the present invention. Figure 2 As shown, the method includes the following steps S110 to S130.

[0048] S110. Collect and process the results of bacterial strain drug susceptibility tests and patient clinical indicator data from the hospital information system, and screen out key features through an improved genetic algorithm; among them, patient clinical indicator data includes raw spatiotemporal trajectory data and nursing operation records.

[0049] In this embodiment, the results of the bacterial susceptibility test refer to specific data on the bacterial strain's antibiotic susceptibility obtained from the hospital laboratory information system (LIS). This data includes, but is not limited to, minimum inhibitory concentration (MIC), susceptibility classification results (e.g., susceptible (S), intermediate (I), resistant (R),) and bacterial genotyping information (e.g., PFGE or MLST typing). This information is crucial for determining the effectiveness of a specific antibiotic against a particular bacterial strain.

[0050] Patient clinical data includes records of invasive procedures, hospitalization process information, raw spatiotemporal trajectory data, and nursing procedure records. This data, sourced from the Electronic Medical Record (EMR) system and Hospital Information System (HIS), reflects patients' behavioral patterns and treatment history during their hospital stay, aiding in the analysis of infection sources and their transmission routes.

[0051] Key features refer to the set of highly discriminative features selected through preprocessing and encoding of the aforementioned data, and then using an improved genetic algorithm combined with a fitness function. These features not only effectively reduce redundant information but also improve the model's ability to characterize drug resistance phenotypes, providing strong support for subsequent propagation network construction.

[0052] In one embodiment, step S110 may include steps S111 to S115.

[0053] S111. Collect the MIC values ​​of the strains for drug susceptibility testing, sensitivity classification results, strain typing information, as well as the patient's invasive operation records, hospitalization process information, original spatiotemporal trajectory data, and nursing operation records through the hospital laboratory information system, electronic medical record system, and hospital information system to obtain the strain drug susceptibility test results and the patient's clinical indicator data.

[0054] In this embodiment, the following information is collected through the Hospital Laboratory Information System (LIS), Electronic Medical Record System (EMR), and Hospital Information System (HIS):

[0055] Antimicrobial susceptibility test results for strains include minimum inhibitory concentration (MIC), susceptibility classification results (e.g., susceptible (S), intermediate (I), resistant (R),) and strain genotyping information (e.g., PFGE or MLST typing). These data are used to determine the effectiveness of a specific antibiotic against a particular strain.

[0056] Patient clinical indicator data includes: records of invasive procedures (such as timestamps and types of procedures such as central vascular catheter insertion and urinary catheterization), hospitalization process information (basic patient information, admission and discharge times, transfer information, bed distribution information, nursing information of medical staff, etc.), original spatiotemporal trajectory data (patient's movement path and location within the hospital), and nursing operation records (such as the personnel and equipment identification for procedures such as disinfection and catheterization).

[0057] S112. Process missing values ​​in the drug susceptibility test results of the strains and the clinical indicator data of patients, and standardize the format to obtain unified results.

[0058] In this embodiment, the unified result refers to the processed, standardized data, including filling in missing values ​​and format conversion. This stage ensures the consistency and integrity of all data, laying the foundation for subsequent data coding and analysis.

[0059] The collected data was processed as follows:

[0060] Missing value handling: Any missing MIC value or morphology information is marked as "not detected" and encoded separately.

[0061] Standardize the format: Convert text-based data (such as "drug resistance") into structured fields to ensure compatibility with numerical data. The purpose of this is to eliminate format differences and redundant noise in the original data, providing high-quality input for subsequent coding.

[0062] S113. When the MIC value in the unified result is less than or equal to the clinical breakpoint, linear normalization is performed; when the MIC value in the unified result is greater than the clinical breakpoint, it is converted into a logarithmic compressed value to obtain the encoding result.

[0063] In this embodiment, the encoding result refers to the segmented normalized encoding of the MIC value according to the following rules:

[0064] When the MIC value is less than or equal to the clinical breakpoint, a linear mapping to the [0, 1] interval is used;

[0065] When the MIC value is greater than the clinical breakpoint, a logarithmic compression formula is used to suppress nonlinear effects, mapping it to the (1, +∞) interval. When the MIC value exceeds the detection limit, a fixed compression value of 0.9 is set. This preserves the linear relationship at low MIC values ​​while avoiding data distribution skew caused by high MIC values.

[0066] Specifically, the MIC values ​​are normalized to the [0, 1] interval after a log2 transformation: When the MIC value exceeds the detection limit, it is set to a fixed compression value of 0.9.

[0067] S114. The sensitivity classification result in the unified result is converted into a three-dimensional one-hot encoded vector representation to obtain the conversion result.

[0068] In this embodiment, the conversion result refers to converting the sensitivity classification result (S / I / R) into a three-dimensional one-hot encoded vector representation. The specific conversion rules are as follows:

[0069] Sensitivity (S) - [1, 0, 0];

[0070] Intermediate (I) - [0, 1, 0];

[0071] Drug resistance (R) - [0, 0, 1];

[0072] This encoding method eliminates the misleading order relationships that may be caused by numerical encoding, and improves the model's sensitivity to classification features.

[0073] S115. Key features are selected from the patient's clinical indicator data using an improved genetic algorithm combined with a fitness function.

[0074] An improved genetic algorithm was used to select the most representative set of key features from patient clinical indicator data. This process includes:

[0075] Population initialization: Randomly generate N feature subsets (e.g., the initial population contains 50 feature subsets).

[0076] Fitness calculation: The predictive performance (e.g., accuracy in classifying infection events) of each subset is evaluated using a random forest model, and a fitness function is defined to balance performance with the number of features. The fitness function considers λ (a balancing factor, ranging from 0.2 to 0.5) to adjust redundancy and representational ability.

[0077] Selection and crossover: Retain the top 30% of individuals by fitness and generate offspring through single-point crossover.

[0078] Dynamic mutation: The initial mutation probability p = 0.1, which decays exponentially with the number of iterations t. If a feature is retained for 5 consecutive generations, it is forcibly locked.

[0079] Example: Input 50 candidate features (such as MIC value, MLST classification, etc.) and output 12 key features after screening (such as "normalized MIC value" and "ST258 classification") to optimize the redundancy of clinical features and the ability to characterize drug resistance phenotypes.

[0080] Through the above steps, a complete process from multi-source data acquisition, preprocessing, encoding conversion to feature selection was achieved, laying a solid foundation for building an accurate spatiotemporal propagation network model.

[0081] In this embodiment, the fitness function is Wherein, λ is the balance factor, and its value ranges from 0.2 to 0.5.

[0082] Activate the bacterial culture data acquisition module and the clinical data analysis module to connect to the hospital's Laboratory Information System (LIS), Electronic Medical Record System (EMR), and Hospital Information System (HIS) to obtain the following two types of information:

[0083] strain phenotypic data:

[0084] The minimum inhibitory concentration (MIC) in drug susceptibility testing, for example, "MIC = 16 μg / mL".

[0085] Sensitivity classification results (Sensitive, Intermediate, Resistant) are classified based on clinical breakpoints.

[0086] Strain typing information, genotype tags determined by PFGE or MLST technology.

[0087] Patient clinical data:

[0088] Invasive procedure records, including timestamps and types of procedures such as central vascular catheter insertion and urinary catheterization.

[0089] The patient's hospitalization details, such as basic information, admission time, discharge time, transfer to other departments, bed distribution, nursing records of medical staff, microbiological information, test results of infection-related markers, use of antimicrobial drugs, diagnostic information, and surgical details.

[0090] The collected data underwent necessary preprocessing, including:

[0091] Missing value handling: If the MIC value or typing information of a strain is missing, it is marked as "not detected" and coded separately.

[0092] Standardized formatting: Transform text-based data (such as "drug resistance") into structured fields to make it compatible with numerical data. This eliminates format differences and redundant noise in the original data, ensuring the quality of subsequent encoding processes.

[0093] Clinical breakpoints are set for different drug-strain combinations according to the CLSI guidelines. For example, the breakpoint for meropenem against Klebsiella pneumoniae is set at 16 μg / mL.

[0094] Different normalization methods are applied based on the position of the MIC value relative to the clinical breakpoint:

[0095] When the MIC value is less than or equal to the clinical breakpoint, it is mapped to the [0, 1] interval using a linear normalization formula. For example, if the MIC = 8 μg / mL and the breakpoint is 16 μg / mL, then x_norm = 0.5.

[0096] If the MIC value is greater than the clinical breakpoint, a logarithmic compression formula is used to limit nonlinear effects. For example, when the MIC = 32 μg / mL and the breakpoint is 16 μg / mL, x_norm ≈ 1.2. This method preserves the linear relationship at low MIC values ​​while avoiding data distribution bias caused by high MIC values.

[0097] Convert sensitivity classification into a three-dimensional binary vector:

[0098] Sensitivity (S) - [1, 0, 0]

[0099] Intermediate (I) - [0, 1, 0]

[0100] Drug resistance (R) - [0, 0, 1]

[0101] If a strain's drug susceptibility test result shows "resistant (R)", then this result should be encoded as [0, 0, 1]. This encoding method eliminates the misleading sequential associations that numerical encoding may introduce, enhancing the model's ability to identify categorical features.

[0102] The process of screening key clinical features using an improved genetic algorithm includes the following stages:

[0103] Population initialization: Randomly generate a series of feature subsets as the initial population, for example, containing 50 feature subsets.

[0104] Fitness evaluation: The performance of each subset in prediction (e.g., classification accuracy of infection events) is evaluated using a random forest model. The fitness function aims to balance the relationship between performance and the number of features, where λ ranges from 0.2 to 0.5, and redundancy is calculated using the Pearson correlation coefficient.

[0105] Selection and crossover: Select the top 30% of individuals with the highest fitness and generate the next generation of individuals through single-point crossover.

[0106] Dynamic mutation: The initial mutation probability is set to p = 0.1, and decreases exponentially with the number of iterations, p(t) = 0.1 × e -0.01t Furthermore, if a feature is selected for five consecutive generations, it will be locked.

[0107] Input 50 candidate features (such as MIC value, MLST classification, etc.) and output 12 selected key features (such as "normalized MIC value" and "ST258 classification"). This process optimizes the selection of clinical features and improves the accuracy of describing drug resistance phenotypes.

[0108] S120. Based on the key features, analyze the patient's spatiotemporal trajectory data and combine it with the consistency of drug sensitivity spectrum. Use graph convolutional networks to construct a propagation network model to identify infection source nodes and their corresponding topological influence.

[0109] In this embodiment, infection source nodes refer to nodes identified in the network model as potential origins of pathogen transmission. These nodes may include patients, medical devices, ward areas, or groups of caregivers, etc., which are highly correlated with drug-resistant bacterial positive events under specific temporal and spatial conditions, and show a high probability of being sources of infection through graph convolutional network (GCN) model analysis. Specifically, infection source nodes have the following characteristics:

[0110] High co-occurrence: It has a lot of contact or overlap with other nodes in the same spatiotemporal unit.

[0111] Similar phenotype: The drug-resistant bacteria carried have a highly consistent drug susceptibility spectrum (high PRI value) with strains on other nodes.

[0112] Centrality: It occupies a relatively central position in the network structure and has a large degree, meaning that it is directly connected to many other nodes.

[0113] Significant topological influence: According to the calculated topological influence index, this node has a significant impact on the propagation path of the entire network.

[0114] Topological influence refers to the importance of each node relative to other nodes in a network; it reflects the role a node plays and the scope of its influence in the spread of disease. The specific definition is as follows: in:

[0115] D(u) is the degree of node u, that is, the number of edges connected to node u;

[0116] L(u, v) is the shortest path length from node u to node v, representing the "distance" between the two;

[0117] β is the path attenuation factor, which takes different values ​​depending on the path type:

[0118] If the path contains a medical device node, then β = 0.1;

[0119] If the path only involves patient and ward nodes, then β = 0.05;

[0120] If the path contains a caregiver node, then β = 0.15.

[0121] This formula comprehensively considers the number of connections a node has, its location within the network, and the impact of different path types on propagation efficiency. Therefore, nodes with significant topological influence may not only be important propagation points themselves, but can also facilitate the spread of pathogens throughout the network through their extensive connections. This quantification method helps to more accurately pinpoint the true source of infection, thereby enabling the implementation of effective prevention and control measures.

[0122] We will use previously identified key features (such as normalized MIC values ​​and MLST classifications) to align events and trajectories through time windows and combine this with the consistency of drug susceptibility profiles to construct a propagation network model. The ultimate goal is to identify potential source nodes of infection and their topological influence within the network.

[0123] In one embodiment, step S120 described above may include steps S121 to S124.

[0124] S121. Based on the key features, align events and trajectories through a time window to obtain the alignment result.

[0125] In this embodiment, alignment results refer to the process of matching discrete events such as patient transfers, bed relocations, and equipment contact records with continuous trajectory information within time windows (e.g., hourly intervals). The purpose of this is to ensure that each positive event for drug-resistant bacteria can be associated with a specific spatiotemporal location, providing an accurate temporal and spatial reference for subsequent analysis. For example, if a patient is detected carrying drug-resistant bacteria within a specific grid within an ICU ward during a particular time window, that time and location will be marked as a potential infection event.

[0126] S122. Based on the alignment results, define nodes including patients, equipment, wards and nursing groups, and calculate edge weights based on spatiotemporal co-occurrence and drug sensitivity consistency to establish a propagation graph and form a propagation network model.

[0127] In this embodiment, the propagation network model refers to a graph structure that includes patients, medical equipment, ward areas, and nursing staff groups as nodes, and calculates edge weights based on the spatiotemporal co-occurrence and drug sensitivity spectrum consistency among nodes. Specifically:

[0128] Node definitions include patient nodes (carrying attributes such as drug-resistant bacteria status), equipment nodes (carrying attributes such as usage frequency), ward nodes (carrying attributes such as current drug-resistant bacteria positivity rate), and nursing group nodes (carrying attributes such as hand hygiene compliance).

[0129] The edge weight is calculated as follows: ω ij =C 共现 (i,j) / (1+α·D(i,j))×PRI(i,j), where, C 共现 (i,j) refers to the number of times two nodes co-occur in the same spatiotemporal unit; α is the distance attenuation coefficient, with a value range of 0.05 to 0.1; D(i,j) refers to the physical distance; PRI(i,j) refers to the similarity of the drug susceptibility spectra of the two strains.

[0130] Two nodes are considered to co-occur if they have any of the following associations within the same spatiotemporal unit (ward + hourly time window):

[0131] Patients use the same medical device (with identical device identification);

[0132] Patient beds are spaced ≤2 meters apart and overlap in time by ≥1 hour;

[0133] Invasive procedures were performed by the same nursing team within adjacent time windows (interval < 4 hours);

[0134] The spatiotemporal co-occurrence coefficient is the number of associated events that satisfy the above conditions.

[0135] This method can more accurately quantify the magnitude of the propagation risk between different nodes.

[0136] S123. Use a three-layer GCN model to convert the adjacency matrix and node features of the propagation graph into embedding vectors for identifying potential infection source nodes.

[0137] In this step, a Graph Convolutional Network (GCN) model is used to transform the adjacency matrix and node features of the propagation graph into embedding vectors. This transformation helps capture complex propagation patterns (such as chain propagation from device to ward to patient), thereby aiding in the identification of potential infection source nodes. GCN models typically consist of three layers, each with an output dimension of 64, using ReLU as the activation function and cross-entropy loss to measure the difference in probability distributions.

[0138] S124. Calculate the topological influence of the nodes in the propagation graph in the network, and adjust the parameters according to the propagation medium to obtain the topological influence corresponding to the source node of infection.

[0139] In this embodiment, the following is adopted: Calculate the topological influence of the nodes in the propagation graph in the network, where D(u) is the degree of node u, L(u, v) is the shortest path length from node u to v, and β is the path decay factor; if the path includes medical equipment nodes, then β = 0.1; if the path only involves patient and ward nodes, then β = 0.05; if the path includes nursing staff nodes, then β = 0.15.

[0140] This method effectively locates infection source nodes and assesses their importance within the network, enabling the implementation of appropriate preventative measures to reduce hospital-acquired infections. Compared to traditional methods, this approach significantly reduces false alarm rates and improves the efficiency and accuracy of infection control efforts.

[0141] In this embodiment, the first step is to obtain relevant patient information from the Hospital Information System (HIS) and Electronic Medical Record System (EMR). This includes:

[0142] Patient spatiotemporal trajectory: Detailed records of the patient's movement within the hospital, such as the time sequence of department transfers (e.g., "transferred to ICU at 08:00 on 2023-01-01"), bed transfer paths (e.g., "bed A1-bed B3-bed C5"), and records of medical equipment used (e.g., ventilator serial number and its usage timestamp).

[0143] Nursing procedure record: includes the type of procedure (such as disinfection, catheterization, etc.), the identification of the nursing staff performing these procedures, and the specific object to which these procedures are performed (which may be a specific device or a patient), i.e., the person performing the procedure (nursing staff number); the object of the procedure (device identification, patient identification).

[0144] Discrete events (such as the time point of a positive culture for drug-resistant bacteria) are aligned with continuous patient trajectories over hourly time windows. If a patient is located in a specific grid cell (e.g., G205) within a particular time period, and drug-resistant bacteria are detected during this period, that location is marked as a potential infection event. This approach helps establish a direct link between strain phenotypic data and patient behavior, providing a foundation for subsequent transmission chain modeling.

[0145] Node definition: Nodes in a graph represent different entities, and each node carries corresponding attributes. For example:

[0146] Patient node: contains the patient's unique identifier, as well as attributes such as drug resistance status and MLST typing.

[0147] Device Nodes: Using the identifier of the medical device as a node, and recording its usage frequency and usage time.

[0148] Ward Node: Using hospital grid units as nodes, record the current positive rate of drug-resistant bacteria.

[0149] Nursing Group Node: Create a node based on the nursing group number and record attributes such as hand hygiene compliance.

[0150] Edge weight calculation: Edge weights are determined by combining the spatiotemporal co-occurrence coefficient and the drug susceptibility profile consistency index (PRI). The formula is as follows: ω ij =C 共现 (i,j) / (1+α·D(i,j))×PRI(i,j)PRI is the similarity of the drug susceptibility spectra of the two strains (with a value between 0 and 1).

[0151] Assuming that patient A and ventilator R1 are in contact within the same spatiotemporal unit and carry the same MLST typing, the edge weights between them can be calculated according to the above formula, thereby accurately quantifying the risk of transmission.

[0152] The model input includes the adjacency matrix (edge ​​weights) of the propagation graph and the node feature matrix (such as drug resistance status and genotype labels). A three-layer GCN architecture is adopted, with each layer having an output dimension of 64. The ReLU activation function is used, and the cross-entropy loss function is used to measure the difference between the prediction results and the actual classification.

[0153] By aggregating information from multi-hop neighbors through GCN, an embedding vector is generated for each node to identify potential sources of infection.

[0154] For an ICU ward transmission map containing 50 nodes and 200 edges, the model outputs a 64-dimensional node embedding vector, which can be used to determine which patients may be potential sources of infection.

[0155] A formula is defined to calculate the topological influence of each node on the entire propagation network. This formula considers the node's degree, shortest path length, and path decay factor β. The β value is dynamically adjusted according to different types of propagation paths. For example, β is set to 0.1 for paths involving medical equipment, while it is increased to 0.15 for paths involving nursing staff.

[0156] If a patient's topological influence (P1) reaches 2.8 (exceeding the threshold T1 = 1.5), the patient is identified as a super-spreader, triggering a Level 1 alert mechanism. This method can more accurately locate the source of infection and reduce false alarm rates.

[0157] In summary, step S2 encompasses a comprehensive series of data processing and technical application procedures, starting with analyzing the raw spatiotemporal trajectory data, constructing complex spatiotemporal propagation maps, utilizing advanced machine learning techniques for in-depth analysis, and finally accurately locating the infection source node. This series of steps not only helps improve the effectiveness of infection control measures but also provides a solid foundation for further research.

[0158] S130. Generate an early warning signal based on the topological influence of the infection source node, and continuously optimize prevention and control measures by dynamically adjusting model parameters and a real-time feedback mechanism.

[0159] In this embodiment, early warning levels are determined based on the topological influence of the infection source node to generate early warning signals, and prevention and control measures are continuously optimized through dynamic adjustment of model parameters and real-time feedback mechanisms.

[0160] Specifically, when the topological influence of the infection source node exceeds the first threshold, enhanced disinfection of ward-level equipment and contact isolation measures are immediately implemented; when the topological influence of the infection source node is between the second threshold and the first threshold, nursing team zoning operations are initiated and hand hygiene compliance monitoring is strengthened; when the topological influence of the infection source node is below the second threshold, environmental microbial sampling is conducted and closed management is implemented for high-risk beds.

[0161] The weights of the drug sensitivity spectrum consistency index are adjusted based on the rate of decline in new cases after intervention, and the optimized parameters are fed back into the transmission network model to achieve continuous self-updating and improvement of the model.

[0162] The first threshold is 0.8, the second threshold is 0.5, and the threshold T' is reduced proportionally when the hospital infection risk level increases. i =T i ·(1-0.1·ΔL), where ΔL is the number of risk level increases; T i For either the first threshold or the second threshold, T' i This refers to either the adjusted first or second threshold.

[0163] The adjustment range of the weight coefficient of the drug susceptibility spectrum consistency index (PRI) ≤ ±30%;

[0164] The value range of the distance decay coefficient α is always limited to 0.05 - 0.1;

[0165] The model update frequency does not exceed once every 24 hours.

[0166] In this embodiment, two warning thresholds are initially set, T1 = 0.8 (primary warning) and T2 = 0.5 (secondary warning). When the hospital infection risk level increases by ΔL levels, the threshold will decrease proportionally.

[0167] Primary warning (I(v) > T1): Trigger enhanced disinfection of equipment at the ward level (such as using ultraviolet circulating disinfection), and implement contact isolation for patients identified as infection sources (including assigning single rooms and dedicated equipment).

[0168] Secondary warning (T2 ≤ I(v) ≤ T1): Trigger partitioned operation of the nursing group (different nursing groups are responsible for different wards), and simultaneously monitor hand hygiene compliance in real-time (detect handwashing actions through smart bracelets).

[0169] Tertiary warning (I(v) < T2): Initiate high-frequency environmental microbial sampling (such as swab tests three times a day), and close high-risk beds (such as beds adjacent to the infection source).

[0170] Assume that the topological influence I(v) of patient P1 = 1.2, and the hospital infection risk level has increased by 1 level (ΔL = 1), then the adjusted warning threshold T1' becomes 0.72. Therefore, the primary warning is triggered, and enhanced disinfection of ICU equipment and isolation measures for patients are implemented.

[0171] Adjust the weight of PRI in the transmission model according to the new case attenuation rate η after intervention. The η benchmark is set as the historical average attenuation rate (for example, η benchmark = 0.5 / day).

[0172] Optimization of the distance decay coefficient α: If the number of associated cases in the same nursing group exceeds the preset value (such as 3 cases per week), then the α value is automatically increased, and the formula is α ← min(α + 0.02, 0.1).

[0173] The adjustment range of the PRI weight is limited within ±30% to avoid model overfitting.

[0174] The value range of α is always limited between 0.05 and 0.1 to ensure the reasonableness of the distance influence.

[0175] The model update frequency does not exceed once every 24 hours to ensure the stability of the system.

[0176] The optimized parameters (PRI weights, α values) are fed back to the propagation network model in step S2 in real time.

[0177] The node influence is recalculated based on the latest data, resulting in an updated risk assessment report. This step ensures that the model can reflect the latest situation in a timely manner and make more accurate predictions and management of potential infection risks.

[0178] The aforementioned closed-loop monitoring method for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion collects and processes strain drug susceptibility test results and patient clinical indicator data from the hospital information system, including raw spatiotemporal trajectories and nursing operation records. An improved genetic algorithm is used to screen key features, and based on these features, a propagation network model is constructed using a graph convolutional network. The consistency of spatiotemporal trajectories among patients, combined with drug susceptibility spectrum analysis, is analyzed to identify infection source nodes and their topological influence. Based on this information, early warning signals are generated, and model parameters are dynamically adjusted. A real-time feedback mechanism optimizes prevention and control measures, thereby achieving accurate analysis and real-time control of drug-resistant bacteria transmission chains. This process integrates data analysis, machine learning, and real-time monitoring technologies, improving the accuracy and timeliness of identifying infection sources and implementing effective prevention and control measures.

[0179] Figure 3 This is a schematic block diagram of an in-hospital drug-resistant bacteria closed-loop monitoring system 300 based on spatiotemporal trajectory fusion provided in an embodiment of the present invention. Figure 3 As shown, corresponding to the above-described closed-loop monitoring method for hospital-acquired drug-resistant bacteria based on spatiotemporal trajectory fusion, this invention also provides a closed-loop monitoring system 300 for hospital-acquired drug-resistant bacteria based on spatiotemporal trajectory fusion. This closed-loop monitoring system 300 includes a unit for executing the above-described closed-loop monitoring method for hospital-acquired drug-resistant bacteria based on spatiotemporal trajectory fusion, and the system can be configured in a server. Specifically, please refer to... Figure 3 The hospital-based drug-resistant bacteria closed-loop monitoring system 300 based on spatiotemporal trajectory fusion includes a collection and processing unit 301, an identification unit 302, and a generation unit 303.

[0180] The collection and processing unit 301 is used to collect and process the results of bacterial strain drug susceptibility tests and patient clinical indicator data from the hospital information system, and to screen out key features through an improved genetic algorithm. Among them, the patient clinical indicator data includes raw spatiotemporal trajectory data and nursing operation records. The identification unit 302 is used to analyze the patient spatiotemporal trajectory data based on the key features and combine it with the consistency of drug susceptibility spectrum, and to construct a propagation network model using a graph convolutional network to identify infection source nodes and their corresponding topological influence. The generation unit 303 is used to generate early warning signals based on the topological influence of the infection source nodes, and to continuously optimize prevention and control measures through dynamic adjustment of model parameters and real-time feedback mechanisms.

[0181] In one embodiment, the collection and processing unit 301 includes:

[0182] The system comprises the following subunits: a collection subunit, used to collect MIC values ​​of bacterial strains for drug susceptibility testing, sensitivity classification results, strain typing information, and patient invasive procedure records, hospitalization process information, raw spatiotemporal trajectory data, and nursing operation records through the hospital laboratory information system, electronic medical record system, and hospital information system, to obtain bacterial strain drug susceptibility test results and patient clinical indicator data; a unification subunit, used to process missing values ​​in the bacterial strain drug susceptibility test results and patient clinical indicator data, and to unify the format to obtain a unified result; an encoding subunit, used to perform linear normalization when the MIC value in the unified result is less than or equal to the clinical inflection point, and to convert it into a logarithmically compressed value when the MIC value in the unified result is greater than the clinical inflection point, to obtain an encoded result; a transformation subunit, used to transform the sensitivity classification results in the unified result into a three-dimensional one-hot encoded vector representation, to obtain a transformed result; and a screening subunit, used to use an improved genetic algorithm combined with a fitness function to screen out key features from the patient clinical indicator data.

[0183] In one embodiment, the identification unit 302 includes:

[0184] The alignment subunit is used to align events and trajectories based on the key features through a time window to obtain alignment results. The model construction subunit is used to define nodes including patients, equipment, wards, and nursing groups according to the alignment results, and calculate edge weights based on spatiotemporal co-occurrence and drug sensitivity consistency to establish a propagation graph and form a propagation network model. The transformation subunit is used to convert the adjacency matrix and node features of the propagation graph into embedding vectors using a three-layer GCN model to identify potential infection source nodes. The calculation subunit is used to calculate the topological influence of the nodes in the propagation graph in the network, and adjust parameters according to the propagation medium to obtain the topological influence corresponding to the infection source node.

[0185] In one embodiment, the generation unit 303 is used to perform early warning classification based on the topological influence of the infection source node to generate an early warning signal, and to continuously optimize prevention and control measures by dynamically adjusting model parameters and a real-time feedback mechanism.

[0186] In one embodiment, the generation unit 303 is further configured to:

[0187] When the topological influence of the infection source node exceeds the first threshold, ward-level enhanced disinfection and contact isolation measures are immediately implemented; when the topological influence of the infection source node is between the second and first thresholds, nursing team zoning operations are initiated and hand hygiene compliance monitoring is strengthened; when the topological influence of the infection source node is below the second threshold, environmental microbial sampling is conducted and high-risk beds are placed under closed management; the weight of the drug sensitivity spectrum consistency index is adjusted according to the new case attenuation rate after intervention, and the optimized parameters are fed back into the transmission network model to achieve continuous self-updating and improvement of the model.

[0188] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned in-hospital drug-resistant bacteria closed-loop monitoring system 300 based on spatiotemporal trajectory fusion and each unit can be referred to the corresponding description in the aforementioned method embodiments. For the sake of convenience and brevity, it will not be repeated here.

[0189] The aforementioned in-hospital drug-resistant bacteria closed-loop monitoring system 300 based on spatiotemporal trajectory fusion can be implemented as a computer program, which can, for example... Figure 4 It runs on the computer device shown.

[0190] Please see Figure 4 , Figure 4 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.

[0191] See Figure 4 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0192] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a closed-loop monitoring method for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion.

[0193] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0194] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a closed-loop monitoring method for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion.

[0195] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0196] The processor 502 is used to run a computer program 5032 stored in the memory to perform the following steps:

[0197] The study collects and processes strain drug susceptibility test results and patient clinical indicator data from the hospital information system, and uses an improved genetic algorithm to screen out key features. The patient clinical indicator data includes raw spatiotemporal trajectory data and nursing operation records. Based on the key features, the study analyzes the patient spatiotemporal trajectory data and combines it with drug susceptibility spectrum consistency. A propagation network model is constructed using a graph convolutional network to identify infection source nodes and their corresponding topological influence. An early warning signal is generated based on the topological influence of the infection source nodes, and prevention and control measures are continuously optimized through dynamic adjustment of model parameters and a real-time feedback mechanism.

[0198] In one embodiment, when the processor 502 collects and processes the bacterial strain drug susceptibility test results and patient clinical indicator data from the hospital information system, and selects key features using an improved genetic algorithm, it specifically implements the following steps:

[0199] The hospital's laboratory information system, electronic medical record system, and hospital information system were used to collect the MIC values ​​of bacterial strains for drug susceptibility testing, sensitivity classification results, strain typing information, as well as patients' invasive procedure records, hospitalization process information, raw spatiotemporal trajectory data, and nursing operation records to obtain bacterial strain drug susceptibility test results and patient clinical indicator data. Missing values ​​in the bacterial strain drug susceptibility test results and patient clinical indicator data were processed and standardized to obtain unified results. When the MIC value in the unified results was less than or equal to the clinical inflection point, linear normalization was performed; when the MIC value in the unified results was greater than the clinical inflection point, it was converted into a logarithmically compressed value to obtain the encoding result. The sensitivity classification results in the unified results were converted into a three-dimensional one-hot encoded vector representation to obtain the transformation result. A modified genetic algorithm combined with a fitness function was used to screen out key features from the patient clinical indicator data.

[0200] Wherein, the fitness function is Wherein, λ is the balance factor, and its value ranges from 0.2 to 0.5.

[0201] In one embodiment, when the processor 502 implements the step of parsing the patient's spatiotemporal trajectory data based on the key features and combining it with the consistency of the drug sensitivity spectrum, and constructing a propagation network model using a graph convolutional network to identify the infection source node and its corresponding topological influence, the specific implementation steps are as follows:

[0202] Based on the key features, events and trajectories are aligned using time windows to obtain alignment results. Nodes, including patients, equipment, wards, and nursing teams, are defined according to the alignment results, and edge weights are calculated based on spatiotemporal co-occurrence and drug sensitivity consistency to establish a propagation graph and form a propagation network model. A three-layer GCN model is used to convert the adjacency matrix and node features of the propagation graph into embedding vectors for identifying potential infection source nodes. The topological influence of the nodes in the propagation graph within the network is calculated, and parameters are adjusted according to the propagation medium to obtain the topological influence corresponding to the infection source nodes.

[0203] The edge weight is calculated as follows: ω ij =C 共现 (i,j) / (1+α·D(i,j))×PRI(i,j), where, C 共现 (i,j) refers to the number of times two nodes co-occur in the same spatiotemporal unit; α is the distance attenuation coefficient, with a value range of 0.05 to 0.1; ·D(i,j) refers to the physical distance, and PRI(i,j) refers to the similarity of the drug susceptibility spectra of the two strains.

[0204] use Calculate the topological influence of the nodes in the propagation graph in the network, where D(u) is the degree of node u, L(u, v) is the shortest path length from node u to v, and β is the path decay factor; if the path includes medical equipment nodes, then β = 0.1; if the path only involves patient and ward nodes, then β = 0.05; if the path includes nursing staff nodes, then β = 0.15.

[0205] In one embodiment, when the processor 502 implements the step of generating an early warning signal based on the topological influence of the infection source node and continuously optimizing prevention and control measures through dynamic adjustment of model parameters and real-time feedback mechanism, it specifically implements the following steps:

[0206] Early warning levels are established based on the topological influence of the infection source nodes to generate early warning signals, and prevention and control measures are continuously optimized through dynamic adjustment of model parameters and real-time feedback mechanisms.

[0207] In one embodiment, when the processor 502 implements the steps of classifying early warnings based on the topological influence of the infection source node to generate early warning signals, and continuously optimizing prevention and control measures through dynamic adjustment of model parameters and real-time feedback mechanisms, it specifically implements the following steps:

[0208] When the topological influence of the infection source node exceeds the first threshold, ward-level enhanced disinfection of equipment and contact isolation measures are immediately implemented; when the topological influence of the infection source node is between the second threshold and the first threshold, nursing team zoning operations are initiated and hand hygiene compliance monitoring is strengthened; when the topological influence of the infection source node is below the second threshold, environmental microbial sampling is conducted and high-risk beds are placed under closed management.

[0209] The weights of the drug sensitivity spectrum consistency index are adjusted based on the rate of decline in new cases after intervention, and the optimized parameters are fed back into the transmission network model to achieve continuous self-updating and improvement of the model.

[0210] The first threshold is 0.8, the second threshold is 0.5, and the threshold T' is reduced proportionally when the hospital infection risk level increases. i =T i ·(1-0.1·ΔL), where ΔL is the number of risk level increases; T i For either the first threshold or the second threshold, T' i This refers to either the adjusted first or second threshold.

[0211] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0212] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0213] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein when executed by a processor, the computer program causes the processor to perform the following steps:

[0214] The study collects and processes strain drug susceptibility test results and patient clinical indicator data from the hospital information system, and uses an improved genetic algorithm to screen out key features. The patient clinical indicator data includes raw spatiotemporal trajectory data and nursing operation records. Based on the key features, the study analyzes the patient spatiotemporal trajectory data and combines it with drug susceptibility spectrum consistency. A propagation network model is constructed using a graph convolutional network to identify infection source nodes and their corresponding topological influence. An early warning signal is generated based on the topological influence of the infection source nodes, and prevention and control measures are continuously optimized through dynamic adjustment of model parameters and a real-time feedback mechanism.

[0215] In one embodiment, when the processor executes the computer program to collect and process bacterial susceptibility test results and patient clinical indicator data from the hospital information system, and selects key features using an improved genetic algorithm, the processor specifically implements the following steps:

[0216] The hospital's laboratory information system, electronic medical record system, and hospital information system were used to collect the MIC values ​​of bacterial strains for drug susceptibility testing, sensitivity classification results, strain typing information, as well as patients' invasive procedure records, hospitalization process information, raw spatiotemporal trajectory data, and nursing operation records to obtain bacterial strain drug susceptibility test results and patient clinical indicator data. Missing values ​​in the bacterial strain drug susceptibility test results and patient clinical indicator data were processed and standardized to obtain unified results. When the MIC value in the unified results was less than or equal to the clinical inflection point, linear normalization was performed; when the MIC value in the unified results was greater than the clinical inflection point, it was converted into a logarithmically compressed value to obtain the encoding result. The sensitivity classification results in the unified results were converted into a three-dimensional one-hot encoded vector representation to obtain the transformation result. A modified genetic algorithm combined with a fitness function was used to screen out key features from the patient clinical indicator data.

[0217] Wherein, the fitness function is Wherein, λ is the balance factor, and its value ranges from 0.2 to 0.5.

[0218] In one embodiment, when the processor executes the computer program to implement the steps of parsing patient spatiotemporal trajectory data based on the key features and combining drug sensitivity spectrum consistency, and constructing a propagation network model using a graph convolutional network to identify infection source nodes and their corresponding topological influence, the processor specifically implements the following steps:

[0219] Based on the key features, events and trajectories are aligned using time windows to obtain alignment results. Nodes, including patients, equipment, wards, and nursing teams, are defined according to the alignment results, and edge weights are calculated based on spatiotemporal co-occurrence and drug sensitivity consistency to establish a propagation graph and form a propagation network model. A three-layer GCN model is used to convert the adjacency matrix and node features of the propagation graph into embedding vectors for identifying potential infection source nodes. The topological influence of the nodes in the propagation graph within the network is calculated, and parameters are adjusted according to the propagation medium to obtain the topological influence corresponding to the infection source nodes.

[0220] The edge weight is calculated as follows: ω ij =C 共现 (i,j) / (1+α·D(i,j))×PRI(i,j), where, C 共现 (i,j) refers to the number of times two nodes co-occur in the same spatiotemporal unit; α is the distance attenuation coefficient, with a value range of 0.05 to 0.1; ·D(i,j) refers to the physical distance, and PRI(i,j) refers to the similarity of the drug susceptibility spectra of the two strains.

[0221] use Calculate the topological influence of the nodes in the propagation graph in the network, where D(u) is the degree of node u, L(u, v) is the shortest path length from node u to v, and β is the path decay factor; if the path includes medical equipment nodes, then β = 0.1; if the path only involves patient and ward nodes, then β = 0.05; if the path includes nursing staff nodes, then β = 0.15.

[0222] In one embodiment, when the processor executes the computer program to generate an early warning signal based on the topological influence of the infection source node, and continuously optimizes the prevention and control measures by dynamically adjusting model parameters and using a real-time feedback mechanism, the processor specifically implements the following steps:

[0223] Early warning levels are established based on the topological influence of the infection source nodes to generate early warning signals, and prevention and control measures are continuously optimized through dynamic adjustment of model parameters and real-time feedback mechanisms.

[0224] In one embodiment, when the processor executes the computer program to implement the early warning classification based on the topological influence of the infection source node to generate an early warning signal, and continuously optimizes the prevention and control measures by dynamically adjusting model parameters and a real-time feedback mechanism, the processor specifically implements the following steps:

[0225] When the topological influence of the infection source node exceeds the first threshold, ward-level enhanced disinfection of equipment and contact isolation measures are immediately implemented; when the topological influence of the infection source node is between the second threshold and the first threshold, nursing team zoning operations are initiated and hand hygiene compliance monitoring is strengthened; when the topological influence of the infection source node is below the second threshold, environmental microbial sampling is conducted and high-risk beds are placed under closed management.

[0226] The weights of the drug sensitivity spectrum consistency index are adjusted based on the rate of decline in new cases after intervention, and the optimized parameters are fed back into the transmission network model to achieve continuous self-updating and improvement of the model.

[0227] The first threshold is 0.8, the second threshold is 0.5, and the threshold T' is reduced proportionally when the hospital infection risk level increases. i =T i ·(1-0.1·ΔL), where ΔL is the number of risk level increases; T i For either the first threshold or the second threshold, T' i This refers to either the adjusted first or second threshold.

[0228] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0229] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0230] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0231] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the system of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0232] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0233] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A closed-loop monitoring method for drug-resistant bacteria within hospitals based on spatiotemporal trajectory fusion, characterized in that, include: Collect and process bacterial strain drug susceptibility test results and patient clinical indicator data from the hospital information system, and screen out key features using an improved genetic algorithm; among them, patient clinical indicator data includes raw spatiotemporal trajectory data and nursing operation records; Based on the analysis of the patient's spatiotemporal trajectory data according to the key features and combined with the consistency of drug sensitivity spectrum, a propagation network model is constructed using graph convolutional networks to identify infection source nodes and their corresponding topological influence. Early warning signals are generated based on the topological influence of the infection source nodes, and prevention and control measures are continuously optimized by dynamically adjusting model parameters and using a real-time feedback mechanism.

2. The method for closed-loop monitoring of drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion according to claim 1, characterized in that, The process involves collecting and processing bacterial strain susceptibility test results and patient clinical indicator data from the hospital information system, and then using an improved genetic algorithm to screen out key features, including: The MIC values ​​of bacterial strains, sensitivity classification results, strain typing information, as well as patient invasive operation records, hospitalization process information, raw spatiotemporal trajectory data, and nursing operation records are collected through the hospital laboratory information system, electronic medical record system, and hospital information system to obtain bacterial strain drug susceptibility test results and patient clinical indicator data. Missing values ​​in the drug susceptibility test results of the strains and the clinical indicator data of patients were processed and standardized to obtain consistent results; When the MIC value in the unified result is less than or equal to the clinical breakpoint, linear normalization is performed; when the MIC value in the unified result is greater than the clinical breakpoint, it is converted into a logarithmic compressed value to obtain the encoding result. The sensitivity classification results in the unified results are converted into three-dimensional one-hot encoded vector representations to obtain the conversion results; Key features were selected from the patient's clinical indicator data using an improved genetic algorithm combined with a fitness function.

3. The method for closed-loop monitoring of drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion according to claim 2, characterized in that, The fitness function is: Wherein, λ is the balance factor, and its value ranges from 0.2 to 0.

5.

4. The method for closed-loop monitoring of drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion according to claim 1, characterized in that, The process of analyzing patient spatiotemporal trajectory data based on the key features and combining it with drug sensitivity spectrum consistency, and constructing a propagation network model using a graph convolutional network to identify infection source nodes and their corresponding topological influence includes: Based on the aforementioned key features, events and trajectories are aligned using time windows to obtain alignment results; Based on the alignment results, nodes including patients, equipment, wards, and nursing groups are defined, and edge weights are calculated based on spatiotemporal co-occurrence and drug sensitivity consistency to establish a propagation graph and form a propagation network model. The adjacency matrix and node features of the propagation graph are converted into embedding vectors using a three-layer GCN model for identifying potential infection source nodes. Calculate the topological influence of the nodes in the propagation graph in the network, and adjust the parameters according to the propagation medium to obtain the topological influence corresponding to the source node of infection.

5. The method for closed-loop monitoring of drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion according to claim 4, characterized in that, The edge weight is calculated as follows: ω ij =C 共现 (i,j) / (1+α·D(i,j))×PRI(i,j), where, C 共现 (i,j) refers to the number of times two nodes co-occur in the same spatiotemporal unit; α is the distance attenuation coefficient, with a value range of 0.05 to 0.1; ·D(i,j) refers to the physical distance, and PRI(i,j) refers to the similarity of the drug susceptibility spectra of the two strains.

6. The method for closed-loop monitoring of drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion according to claim 4, characterized in that, use Calculate the topological influence of the nodes in the propagation graph in the network, where D(u) is the degree of node u, L(u, v) is the shortest path length from node u to v, and β is the path decay factor; if the path includes medical equipment nodes, then β = 0.1; if the path only involves patient and ward nodes, then β = 0.05; if the path includes nursing staff nodes, then β = 0.

15.

7. The method for closed-loop monitoring of drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion according to claim 1, characterized in that, The method of generating early warning signals based on the topological influence of the infection source nodes, and continuously optimizing prevention and control measures through dynamic adjustment of model parameters and real-time feedback mechanisms, includes: Early warning levels are established based on the topological influence of the infection source nodes to generate early warning signals, and prevention and control measures are continuously optimized through dynamic adjustment of model parameters and real-time feedback mechanisms.

8. The method for closed-loop monitoring of drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion according to claim 7, characterized in that, The method of classifying early warnings based on the topological influence of the infection source node to generate early warning signals, and continuously optimizing prevention and control measures through dynamic adjustment of model parameters and real-time feedback mechanisms, includes: When the topological influence of the infection source node exceeds the first threshold, ward-level enhanced disinfection of equipment and contact isolation measures are immediately implemented; when the topological influence of the infection source node is between the second threshold and the first threshold, nursing team zoning operations are initiated and hand hygiene compliance monitoring is strengthened; when the topological influence of the infection source node is below the second threshold, environmental microbial sampling is conducted and high-risk beds are placed under closed management. The weights of the drug sensitivity spectrum consistency index are adjusted based on the rate of decline in new cases after intervention, and the optimized parameters are fed back into the transmission network model to achieve continuous self-updating and improvement of the model.

9. The method for closed-loop monitoring of drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion according to claim 8, characterized in that, The first threshold is 0.8, the second threshold is 0.5, and the threshold T' is reduced proportionally when the hospital infection risk level increases. i =T i ·(1-0.1·ΔL), where ΔL is the number of risk level increases; T i For either the first threshold or the second threshold, T' i This refers to either the adjusted first or second threshold.

10. A closed-loop monitoring system for drug-resistant bacteria in hospitals based on spatiotemporal trajectory fusion, characterized in that, include: The collection and processing unit is used to collect and process the results of bacterial strain drug susceptibility tests and patient clinical indicator data from the hospital information system, and to screen out key features through an improved genetic algorithm; among them, patient clinical indicator data includes raw spatiotemporal trajectory data and nursing operation records. The identification unit is used to analyze the patient's spatiotemporal trajectory data based on the key features and combine it with the consistency of the drug sensitivity spectrum, and use a graph convolutional network to construct a propagation network model to identify the infection source node and its corresponding topological influence. The generation unit is used to generate early warning signals based on the topological influence of the infection source node, and continuously optimize prevention and control measures by dynamically adjusting model parameters and using a real-time feedback mechanism.

Images