Construction method of auxiliary treatment system based on traumatic patient

Abstract

The invention relates to the technical field of auxiliary treatment of traumatic patients, in particular to a construction method of an auxiliary treatment system based on traumatic patients. The method comprises the following steps: carrying out multi-modal physiological data acquisition and physiological data preprocessing on a trauma patient through hospital examination equipment to obtain multi-modal feature fusion data; carrying out physiological feature vector generation on the multi-modal feature fusion data to obtain a physiological feature vector of the trauma patient; carrying out physiological state coding on the physiological feature vector of the trauma patient to obtain physiological state coding data; acquiring trauma medical knowledge data; and according to the trauma medicine knowledge data and the physiological state coding data, carrying out physiological state and knowledge graph association to obtain physiological state and knowledge graph association data. According to the invention, through real-time adjustment of the patient processing sequence and automatic generation of the personalized resource allocation auxiliary processing scheme according to the patient features, the medical resource utilization efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of auxiliary treatment of trauma patients, and in particular to a method for constructing an auxiliary treatment system based on trauma patients. Background Art

[0002] With the continuous development of medical information technology, scholars at home and abroad have begun to study the use of artificial intelligence, big data analysis and other technologies to assist in the treatment of trauma patients. Early research focused on the collection and storage of trauma patient data and rule-based triage systems. Existing methods for constructing auxiliary treatment systems based on trauma patients attempt to use machine learning algorithms to evaluate and predict patient status, but lack dynamic priority adjustment for dynamic changes in patients and lack auxiliary processing support for medical resource allocation. In general, although the existing auxiliary treatment systems for trauma patients have improved medical efficiency to a certain extent, they cannot adapt well to complex clinical needs.

[0003] 1. Lack of dynamic priority adjustment for patients’ dynamic changes:

[0004] The physiological state of trauma patients changes dynamically, and the existing auxiliary treatment system can usually only provide static evaluation results, lacking continuous monitoring and analysis of the patient's real-time state. At the same time, the existing system is not flexible enough in the dynamic adjustment of patient priority, and it is difficult to update the patient treatment order in time according to the changes in the patient's condition, which may delay the rescue of critically ill patients.

[0005] 2. Lack of auxiliary processing support for medical resource allocation:

[0006] The treatment of patient trauma requires the deployment of a variety of medical resources, such as emergency personnel, equipment, and medicines. However, the existing system often relies on manual experience to allocate medical resources, lacks intelligent optimization methods, and leads to low resource utilization efficiency. In addition, the existing system does not provide sufficient support for auxiliary processing of medical resource allocation, making it difficult to provide personalized and precise patient auxiliary treatment solutions. Summary of the invention

[0007] Based on this, it is necessary to provide a method for constructing an auxiliary treatment system based on trauma patients to solve at least one of the above technical problems.

[0008] To achieve the above purpose, a method for constructing an auxiliary treatment system based on trauma patients includes the following steps:

[0009] Step S1: collecting multimodal physiological data of trauma patients through hospital examination equipment and preprocessing the physiological data to obtain multimodal feature fusion data; generating physiological feature vectors for the multimodal feature fusion data to obtain physiological feature vectors of trauma patients;

[0010] Step S2: Encode the physiological state of the physiological feature vector of the trauma patient to obtain physiological state encoding data; obtain trauma medical knowledge data; associate the physiological state with the knowledge graph according to the trauma medical knowledge data and the physiological state encoding data to obtain physiological state and knowledge graph association data; generate a trauma stage label according to the physiological state and knowledge graph association data to obtain a trauma stage label;

[0011] Step S3: Perform multi-dimensional feature fusion according to the trauma stage label to obtain a multi-dimensional comprehensive feature vector; use the pre-trained emergency level prediction model to predict the emergency level probability of the multi-dimensional comprehensive feature vector and calculate the emergency level score to obtain the emergency level score data of the trauma patient;

[0012] Step S4: dynamically adjusting the priority score according to the emergency level score data of the trauma patient to obtain dynamically adjusted priority score data; dynamically sorting the priorities within and outside the team according to the dynamically adjusted priority score data and performing visualization processing to obtain a visualized dynamic priority queue;

[0013] Step S5: Analyze the patient demand of the visualized dynamic priority queue to obtain patient demand data; generate a patient auxiliary treatment plan according to the patient demand data to obtain a patient auxiliary treatment plan;

[0014] Step S6: Push the emergency treatment team resource data according to the patient auxiliary treatment plan to achieve auxiliary treatment of trauma patients.

[0015] The present invention can comprehensively and accurately obtain the real-time physiological status information of the patient by real-time acquisition of multimodal physiological data and preprocessing, and provide a high-quality data basis for subsequent disease assessment and auxiliary treatment. By using the trauma medical knowledge map to encode and associate the patient's physiological state, the patient's trauma stage can be more accurately identified, and the disease development trend can be predicted, providing a more accurate reference basis for subsequent emergency assessment and personalized medical resource allocation. Through multi-dimensional feature fusion and pre-trained emergency level prediction model, the patient's urgency can be more comprehensively and objectively evaluated, avoiding the error caused by relying on a single indicator, and improving accuracy and efficiency. According to the patient's urgency score and real-time physiological state changes, dynamic priority sorting is performed and presented to medical staff in a visual manner, which can help medical staff to grasp the latest situation of the patient in real time and improve team collaboration efficiency and resource allocation efficiency. According to patient needs and available medical resources, intelligent resource matching and optimization configuration are performed, and personalized patient auxiliary treatment plans are generated, which can assist medical staff in the treatment of trauma patients and improve resource utilization efficiency. Converting the auxiliary treatment plan into a specific medical resource arrangement and automatically pushing it to the corresponding emergency treatment team members can improve the efficiency and execution of teamwork and ensure that patients can receive timely treatment in a timely manner. Therefore, the present invention provides a method for constructing an auxiliary treatment system based on trauma patients, which is used to solve the problem that the existing method for constructing an auxiliary treatment system based on trauma patients lacks dynamic priority adjustment for dynamic changes of patients and lacks auxiliary treatment support for medical resource allocation. Through the real-time collection and analysis of the patient's multimodal physiological data, the patient's condition is dynamically evaluated and comprehensively analyzed using time series algorithms and medical knowledge graphs, a dynamic priority scheduling mechanism based on patient status and medical resources is established, and real-time adjustment of the patient's treatment order is achieved. At the same time, the method comprehensively considers the patient's needs and the real-time status of medical resources, uses optimization algorithms for intelligent resource matching and scheduling, and automatically generates personalized resource allocation auxiliary treatment plans based on patient characteristics, providing medical staff with intelligent auxiliary treatment, and improving the efficiency of medical resource utilization and the accuracy of patient treatment.

[0016] Preferably, step S1 comprises the following steps:

[0017] Step S11: collecting real-time physiological data of trauma patients through hospital examination equipment to obtain a physiological data stream of trauma patients;

[0018] Step S12: performing data stream synchronization and alignment on the physiological data stream of the trauma patient to obtain a synchronized data stream; performing noise filtering and artifact elimination on the synchronized data stream to obtain preprocessed data;

[0019] Step S13: extracting physiological index features from the preprocessed data to obtain a physiological index feature set; performing multimodal feature fusion on the physiological index feature set to obtain multimodal feature fusion data;

[0020] Step S14: Generate a physiological feature vector for the multimodal feature fusion data to obtain a physiological feature vector of the trauma patient.

[0021] The present invention uses hospital examination equipment to collect real-time physiological data of trauma patients, which can comprehensively and dynamically obtain the patient's vital signs information, provide the original data basis for subsequent disease assessment and auxiliary processing, and is conducive to timely discovering changes in the patient's condition; time synchronization and format alignment of multi-source physiological data streams ensure the consistency and comparability of data in time, and eliminate the impact of differences in data collected by different devices. At the same time, noise filtering and artifact elimination effectively improve the accuracy and reliability of the data, providing high-quality data guarantee for subsequent analysis; extracting physiological indicator characteristics related to trauma from pre-processed physiological data can convert the original data into more clinically significant indicators, which is convenient for medical staff to quickly understand the patient's condition. Multimodal feature fusion integrates information from multiple physiological indicators, which can more comprehensively and accurately reflect the patient's physiological state and avoid the one-sidedness of a single indicator; converting multimodal feature fusion data into structured physiological feature vectors is convenient for subsequent computer processing and analysis, such as input for machine learning models. The physiological feature vector quantifies the patient's physiological state, providing data support for subsequent disease assessment, priority sorting, etc.

[0022] Preferably, step S2 comprises the following steps:

[0023] Step S21: dividing the physiological characteristic vector of the trauma patient into time windows to obtain time window sequence data; enhancing the physiological characteristic sequence within the time window to obtain time window characteristic sequence enhanced data;

[0024] Step S22: performing deep feature learning on the time window feature sequence enhancement data to obtain a deep feature embedding vector; inputting the deep feature embedding vector into a preset encoder-decoder network to perform physiological state encoding to obtain physiological state encoding data;

[0025] Step S23: Acquire trauma medical knowledge data; construct a trauma knowledge graph based on the trauma medical knowledge data to obtain a trauma knowledge graph;

[0026] Step S24: Associating the physiological state with the knowledge graph of the trauma and the physiological state encoding data to obtain physiological state and knowledge graph association data;

[0027] Step S25: Generate a time axis probability trajectory according to the physiological state and the knowledge graph association data to obtain time axis probability trajectory data; perform probability trajectory smoothing optimization on the time axis probability trajectory data to obtain time axis smoothed data;

[0028] Step S26: Identify key nodes on the time axis smoothing data to obtain time axis key node data; generate trauma stage labels based on the time axis key node data to obtain trauma stage labels.

[0029] The present invention can divide the patient's continuous physiological state change trend into multiple time periods for analysis through time window division, so as to better capture the dynamic development process of the patient's condition. The enhancement of the physiological feature sequence within the time window can make up for the problem of insufficient information within the time window and improve the accuracy of subsequent analysis; using a deep learning model to learn the time window feature sequence can extract the complex patterns and nonlinear relationships hidden in the data, and more accurately characterize the patient's physiological state change trend. Physiological state coding maps high-dimensional deep feature vectors to low-dimensional coding space, reduces the data dimension, facilitates subsequent processing, and extracts more representative physiological state information. By acquiring trauma medical knowledge data and constructing a knowledge graph, the experience and knowledge of experts in the field can be integrated into the system to provide knowledge support for subsequent disease assessment and auxiliary processing. Knowledge graphs can effectively organize and represent medical knowledge related to trauma, making it easier for computers to understand and reason. By associating the patient's physiological state encoding data with the trauma knowledge graph, the patient's physiological state can be mapped to the medical knowledge system, and the disease stage or development path that is most similar to the current patient's state can be found, providing a basis for subsequent condition analysis and prediction. Based on the association results between the physiological state and the knowledge graph, the probability distribution of the patient in different trauma stages is generated, and a timeline probability trajectory is formed, which can intuitively display the timeline of the patient's disease development and predict possible future changes in the disease. Probabilistic trajectory smoothing optimization can eliminate the impact of noise and fluctuations on the trajectory and improve the accuracy and reliability of the prediction. By identifying the key nodes on the timeline smoothed data, the time point when the patient's condition changes significantly can be determined, such as the time point when it develops from one stage to another. Based on the key nodes and timeline smoothed data, more refined and accurate trauma stage labels can be generated for patients, providing more detailed disease information for subsequent urgency assessment and auxiliary treatment plan generation.

[0030] Preferably, step S24 comprises the following steps:

[0031] Step S241: probabilizing the trauma knowledge graph to obtain a probabilistic trauma knowledge graph; calculating the physiological state probability distribution of the physiological state coding data according to the probabilistic trauma knowledge graph to obtain physiological state probability distribution data;

[0032] Step S242: constructing a Bayesian network according to the probabilistic trauma knowledge graph to obtain a Bayesian network; inputting the physiological state probability distribution data into the Bayesian network, setting observation variables, and obtaining observation variables;

[0033] Step S243: performing latent variable state inference according to the Bayesian network and the observed variables to obtain latent variable state change data;

[0034] Step S244: evaluating the probability of the time nodes of the patient being in different trauma stages according to the latent variable state change data to obtain the time node probability data;

[0035] Step S245: Perform dynamic Bayesian network joint reasoning based on the Bayesian network, time node probability data, and physiological state probability distribution data to obtain physiological state and knowledge graph association data.

[0036] The present invention can integrate the uncertainty and probability relationship in medical knowledge into the system by probabilizing the trauma knowledge graph, so that the system can better handle complex trauma situations in the real world. By calculating the probability distribution of physiological state according to the probabilistic knowledge graph, the possibility of patients being in different stages of trauma can be quantified, providing basic data for subsequent probabilistic reasoning; by converting the probabilistic trauma knowledge graph into a Bayesian network, the powerful probabilistic reasoning ability of the Bayesian network can be used to dynamically analyze and predict the patient's condition. By inputting the physiological state probability distribution data as an observed variable into the Bayesian network, the patient's real-time physiological state information can be integrated into the reasoning process, improving the accuracy and real-time nature of the reasoning results; through Bayesian network reasoning, the physiological state that cannot be directly observed in the patient's body can be estimated, such as the amount of bleeding, the degree of organ damage, etc. These hidden variable states are crucial for disease assessment. Hidden variable state reasoning can help medical staff understand the patient's condition more comprehensively and provide a reference for subsequent adjustment of the treatment plan. According to the hidden variable state change data, the probability of the patient being in different stages of trauma at different time points in the future can be evaluated, such as predicting when the patient may enter a state of shock and when surgery is required. Time node probability assessment can help medical staff prepare in advance and develop more forward-looking treatment plans for patients. Through dynamic Bayesian network joint reasoning, the patient's physiological state information, time information and medical knowledge can be organically combined to conduct a more accurate and comprehensive dynamic assessment of the patient's condition and predict future disease development trends.

[0037] Preferably, step S3 comprises the following steps:

[0038] Step S31: performing trauma stage feature encoding on the trauma stage label to obtain trauma stage feature encoding data;

[0039] Step S32: obtaining individual information of the trauma patient; characterizing the individual information of the trauma patient to obtain individual information characteristic data;

[0040] Step S33: obtaining the injury site information of the trauma patient; quantifying the injury site information of the patient to obtain quantified injury site information data;

[0041] Step S34: performing multi-dimensional feature fusion on the trauma stage feature coding data, the individual information feature data and the injury site information quantification data to obtain a multi-dimensional comprehensive feature vector;

[0042] Step S35: using the pre-trained emergency level prediction model to perform emergency level probability prediction on the multi-dimensional comprehensive feature vector to obtain an emergency level probability value;

[0043] Step S36: Calculate the emergency level score according to the emergency level probability value and the preset emergency weight coefficient to obtain the emergency level score data of the trauma patient.

[0044] The present invention can convert text information that is difficult to be understood by computers into numerical form by encoding the trauma stage labels, which is convenient for subsequent machine learning models to process and analyze, and provide a data basis for automatically evaluating the urgency of patients. By obtaining and characterizing individual information such as the patient's age, gender, and past medical history, these key factors that affect the development of the patient's condition can be taken into consideration, making the urgency assessment more personalized and more in line with the specific circumstances of each patient. By quantifying the patient's injury site information, the site information that is difficult to quantify can be converted into numerical form, which is convenient for the computer to process and analyze, and at the same time, the severity and scope of the injury can be more accurately described, thereby improving the accuracy of the urgency assessment. By integrating multi-dimensional features such as trauma stage, individual information, and injury site, a comprehensive feature vector containing various aspects of the patient's information can be constructed, which comprehensively describes the patient's condition and physical condition, and provides a more comprehensive information basis for the urgency assessment. By using the pre-trained emergency level prediction model to predict the comprehensive feature vector, the probability of the patient being at different emergency levels can be automatically evaluated, providing a fast and objective reference basis for medical staff to help them perform auxiliary processing.

[0045] Preferably, step S35 includes the following steps:

[0046] Step S351: Collecting comprehensive data of historical trauma patients from the medical system to obtain comprehensive data of historical trauma patients; constructing an emergency level sample library based on the comprehensive data of historical trauma patients to obtain an initial emergency level sample library; performing sample category balancing on the initial emergency level sample library to obtain an emergency level sample library;

[0047] Step S352: constructing a multimodal feature graph for the multimodal feature data of patients in the emergency level sample library to obtain a multimodal feature graph;

[0048] Step S353: performing feature correlation calculation on the multimodal feature graph to obtain feature correlation data;

[0049] Step S354: using the feature correlation data to assign graph node weights to the multimodal feature graph to obtain weighted feature graph structure data;

[0050] Step S355: Select a graph neural network model according to the weight feature graph structure data to obtain a graph neural network model;

[0051] Step S356: Train the graph neural network model for emergency level prediction and perform parameter optimization to obtain a pre-trained emergency level prediction model.

[0052] The present invention can provide a data basis for the training of the emergency level prediction model by collecting comprehensive data of historical trauma patients and constructing an emergency level sample library, and convert historical data into a format that can be used for machine learning. Sample category balance can solve the problem of unbalanced number of samples of different emergency levels, avoid deviations in model training, and improve the generalization ability and prediction accuracy of the model. Constructing the multimodal feature data of patients into a graph structure can more effectively express the correlation between different features, such as the correlation between physiological indicators and injury sites, the correlation between individual information and trauma stages, etc. The multimodal feature graph can provide richer information for subsequent feature correlation calculation and graph neural network model training. By calculating the correlation between different features in the multimodal feature graph, the correlation strength between the features can be quantified, and the key features and feature combinations with high contribution to the emergency level prediction can be identified, providing a basis for subsequent feature selection and model optimization. According to the feature correlation calculation results, the weights of the nodes in the multimodal feature graph are assigned, which can highlight the role of important features, improve the sensitivity of the model to key information, and thus improve the accuracy of emergency level prediction. Selecting an appropriate graph neural network model based on the weighted feature graph structure data can make full use of the graph structure information and node weight information, more effectively learn the complex relationship between features, and improve the model's expressiveness and predictive performance. By training and optimizing the parameters of the graph neural network model, the model can learn the patterns and rules in historical data and accurately predict the patient's emergency level based on their multimodal feature data.

[0053] Preferably, step S4 comprises the following steps:

[0054] Step S41: generating an initial priority queue according to the emergency level score data of trauma patients to obtain an initial priority queue;

[0055] Step S42: monitoring physiological characteristic changes of each patient in the trauma patient emergency level score data to obtain physiological characteristic change trend data;

[0056] Step S43: performing indicator abnormality identification on the physiological characteristic change trend data to obtain physiological indicator abnormality record data; performing risk level assessment on the physiological indicator abnormality record data according to the trauma medical knowledge data to obtain abnormal indicator risk data;

[0057] Step S44: Calculate the priority adjustment weight according to the abnormal indicator risk data to obtain the priority adjustment weight data;

[0058] Step S45: dynamically adjusting the priority score of the priority adjustment weight data and the trauma patient emergency level score data to obtain dynamically adjusted priority score data;

[0059] Step S46: updating the order in the queue according to the dynamically adjusted priority score data to obtain updated order data in the queue; performing upgrade and downgrade processing between queues according to the updated order data in the queue to obtain a dynamic priority queue;

[0060] Step S47: Visualize the queue state of the dynamic priority queue to obtain a visualized dynamic priority queue.

[0061] The present invention can preliminarily determine the treatment priority of the patient by generating an initial priority queue according to the emergency level score data, and provide a reference basis for medical staff so that limited medical resources can be allocated to the most urgent patients in the first place. Continuously monitoring the change trend of the patient's physiological characteristics can timely discover the changes in the patient's condition, provide a data basis for dynamically adjusting the priority, and avoid missing the best treatment opportunity by relying solely on the initial assessment results. By identifying abnormal physiological indicators and combining trauma medical knowledge to perform risk assessment, the degree of deterioration of the patient's condition can be more accurately judged, providing a more reliable basis for priority adjustment, and avoiding making wrong judgments based on changes in a single indicator. Priority adjustment weights are calculated according to the risk level of abnormal indicators, the risk levels of different indicators can be quantified, and the priority of the patient can be dynamically adjusted according to the risk level, so that the priority sorting is more in line with the patient's actual condition, and the accuracy and effectiveness of patient treatment are improved. The initial emergency level score is dynamically adjusted according to the priority adjustment weight, which can more accurately reflect the patient's real-time condition changes, avoid the priority sorting lag caused by relying solely on the initial score, and ensure that the most urgent patients are treated first. The priority queue is updated according to the dynamically adjusted priority score, and the upgrade and downgrade processing between queues is performed, which can reflect the changes in the patient's condition in real time, and adjust the patient's treatment priority in time, ensuring that medical resources are always allocated to the most needy patients. Visualizing dynamic priority queues can help medical staff understand the latest patient conditions and priority rankings in real time, improve information transparency and communication efficiency, and make it easier for medical staff to adjust treatment plans and resource allocation in a timely manner, thereby improving team collaboration efficiency and patient treatment efficiency.

[0062] Preferably, step S43 includes the following steps:

[0063] Step S431: extracting physiological index time series from the physiological characteristic change trend data to obtain a physiological index time series; constructing a dynamic baseline according to the physiological index time series to obtain a dynamic baseline;

[0064] Step S432: performing indicator abnormality identification on the physiological indicator time series and dynamic baseline to obtain physiological indicator abnormality record data;

[0065] Step S433: quantifying the abnormality index of the abnormal physiological index record data to obtain quantified abnormality index data;

[0066] Step S434: performing risk factor association on the quantitative abnormal index data and the trauma medical knowledge data to obtain index-risk association data;

[0067] Step S435: using a preset risk scoring rule to evaluate the risk level of the quantitative abnormal indicator data according to the indicator-risk association data to obtain abnormal indicator risk data.

[0068] The present invention can more accurately identify abnormal changes in the patient's physiological indicators by extracting the physiological indicator time series and constructing a dynamic baseline, overcome the limitations of the traditional fixed baseline method, and better adapt to the individual differences of patients and the dynamic changes of the disease. By comparing the physiological indicator time series with the dynamic baseline, the abnormal indicators that exceed the normal physiological range of the patient can be more sensitively identified, and the potential risks of the patient's condition can be discovered in time, providing a basis for subsequent risk assessment. Quantitative processing of indicator abnormalities can convert the abnormal degree and duration that were originally difficult to compare into numerical form, facilitate subsequent analysis and calculation, and provide more refined data support for risk assessment. Associating the quantitative abnormal indicator data with the risk factors in the trauma medicine knowledge base can link the changes in the patient's physiological indicators with the potential disease risks, assist medical staff in more comprehensively assessing the patient's condition, and predicting possible disease development trends. According to the preset risk scoring rules, the quantitative abnormal indicator data is evaluated for risk level, and the risk levels of different indicators can be quantitatively compared, providing a more objective and accurate basis for priority adjustment.

[0069] Preferably, step S5 comprises the following steps:

[0070] Step S51: Real-time monitoring of system medical resources to obtain medical resource status data; resource availability analysis of the medical resource status data to obtain available resource data;

[0071] Step S52: Analyze patient demand on the visualized dynamic priority queue to obtain patient demand data;

[0072] Step S53: Evaluate the matching degree between resources and needs of the patient's needs data and available resource data to obtain resource matching data;

[0073] Step S54: setting constraint conditions according to the matching degree data to obtain resource matching constraint data;

[0074] Step S55: constructing a path planning objective function according to the resource matching degree data and the resource matching constraint data to obtain a path planning objective function;

[0075] Step S56: Utilizing a genetic algorithm to optimize the resource allocation scheme and the patient path according to the path planning objective function, the resource matching degree data, and the resource matching constraint data, and obtaining resource allocation path optimization data;

[0076] Step S57: Generate a patient-assisted treatment plan based on the resource allocation path optimization data to obtain a patient-assisted treatment plan.

[0077] The present invention can help hospitals grasp the real-time resource situation, timely discover resource shortages and uneven distribution problems, and provide data support for subsequent resource optimization configuration and scheduling by monitoring the status of medical resources in real time and analyzing their availability. By analyzing the patient's treatment needs, it can be clear what types of medical resources and services the patient needs, provide a basis for subsequent resource matching and allocation, and avoid resource waste and service loss. By evaluating the degree of match between patient needs and available resources, the most suitable resources and services for patients can be found, resource utilization efficiency and patient satisfaction can be improved, and resource mismatch and service quality decline can be avoided. By setting resource matching constraints, the actual situation of the hospital and the medical resource scheduling rules can be taken into consideration, such as the professional expertise of medical staff, equipment usage restrictions, etc., so that resource allocation is more reasonable, efficient and safe. By constructing a path planning objective function, the goal and optimization direction of resource allocation can be quantified, such as maximizing resource utilization or minimizing patient waiting time, providing target guidance for subsequent optimization algorithms. Using intelligent optimization algorithms such as genetic algorithms, the optimal resource allocation plan and patient path can be automatically searched according to the objective function, resource matching degree and constraints, improving resource allocation efficiency and reducing the inefficiency caused by manual scheduling. Based on the optimized resource allocation plan and patient pathway, specific patient assistance treatment plans are generated to improve the precision and intelligence of medical services.

[0078] Preferably, step S6 comprises the following steps:

[0079] Step S61: extracting parameters of the patient auxiliary treatment plan to obtain treatment plan parameters;

[0080] Step S62: combining the processing solution parameters with the solution content according to the preset solution template to obtain a content combination solution;

[0081] Step S63: Generate emergency processing team resource data and emergency processing team data according to the content combination plan;

[0082] Step S64: Pushing data to the emergency treatment team's resource data to assist in the treatment of trauma patients.

[0083] The present invention can convert complex scheme information into structured data by extracting key scheme parameters from the patient auxiliary treatment scheme, which is convenient for subsequent processing by the computer and provides a basis for automatically generating scheme content. By combining the scheme parameters using a preset scheme template, it is possible to generate scheme content with standardized format, complete content, and easy to understand, thereby avoiding ambiguity or errors in manually written schemes and improving the accuracy and efficiency of scheme delivery. By automatically retrieving and matching the corresponding emergency treatment team resource data according to the scheme content, a medical team that meets the requirements can be quickly formed, delays caused by manual search and coordination can be avoided, and the efficiency of team formation and response speed can be improved. The emergency treatment team resource data is pushed to relevant medical staff in a timely manner to ensure that the relevant personnel receive the scheme data in a timely manner and take action, thereby improving the efficiency and execution of teamwork.

[0084] The present invention can comprehensively and accurately obtain the real-time physiological status information of the patient by real-time acquisition of multimodal physiological data and preprocessing, and provide a high-quality data basis for subsequent disease assessment and auxiliary treatment. By using the trauma medical knowledge map to encode and associate the patient's physiological state, the patient's trauma stage can be more accurately identified, and the disease development trend can be predicted, providing a more accurate reference basis for subsequent emergency assessment and personalized medical resource allocation. Through multi-dimensional feature fusion and pre-trained emergency level prediction model, the patient's urgency can be more comprehensively and objectively evaluated, avoiding the error caused by relying on a single indicator, and improving accuracy and efficiency. According to the patient's urgency score and real-time physiological state changes, dynamic priority sorting is performed and presented to medical staff in a visual manner, which can help medical staff to grasp the latest situation of the patient in real time and improve team collaboration efficiency and resource allocation efficiency. According to patient needs and available medical resources, intelligent resource matching and optimization configuration are performed, and personalized patient auxiliary treatment plans are generated, which can assist medical staff in the treatment of trauma patients and improve resource utilization efficiency. Converting the auxiliary treatment plan into a specific medical resource arrangement and automatically pushing it to the corresponding emergency treatment team members can improve the efficiency and execution of teamwork and ensure that patients can receive timely treatment in a timely manner. Therefore, the present invention provides a method for constructing an auxiliary treatment system based on trauma patients, which is used to solve the problem that the existing method for constructing an auxiliary treatment system based on trauma patients lacks dynamic priority adjustment for dynamic changes of patients and lacks auxiliary treatment support for medical resource allocation. Through the real-time collection and analysis of the patient's multimodal physiological data, the patient's condition is dynamically evaluated and comprehensively analyzed using time series algorithms and medical knowledge graphs, a dynamic priority scheduling mechanism based on patient status and medical resources is established, and real-time adjustment of the patient's treatment order is achieved. At the same time, the method comprehensively considers the patient's needs and the real-time status of medical resources, uses optimization algorithms for intelligent resource matching and scheduling, and automatically generates personalized resource allocation auxiliary treatment plans based on patient characteristics, providing medical staff with intelligent auxiliary treatment, and improving the efficiency of medical resource utilization and the accuracy of patient treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 A schematic diagram of the steps of the method for constructing an auxiliary treatment system for trauma patients;

[0086] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0087] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0088] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0089] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0090] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0091] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0092] To achieve this, please refer to Figures 1 to 3 , a method for constructing an auxiliary treatment system based on trauma patients, comprising the following steps:

[0093] Step S1: collecting multimodal physiological data of trauma patients through hospital examination equipment and preprocessing the physiological data to obtain multimodal feature fusion data; generating physiological feature vectors for the multimodal feature fusion data to obtain physiological feature vectors of trauma patients;

[0094] Step S2: Encode the physiological state of the physiological feature vector of the trauma patient to obtain physiological state encoding data; obtain trauma medical knowledge data; associate the physiological state with the knowledge graph according to the trauma medical knowledge data and the physiological state encoding data to obtain physiological state and knowledge graph association data; generate a trauma stage label according to the physiological state and knowledge graph association data to obtain a trauma stage label;

[0095] Step S3: Perform multi-dimensional feature fusion according to the trauma stage label to obtain a multi-dimensional comprehensive feature vector; use the pre-trained emergency level prediction model to predict the emergency level probability of the multi-dimensional comprehensive feature vector and calculate the emergency level score to obtain the emergency level score data of the trauma patient;

[0096] Step S4: dynamically adjusting the priority score according to the emergency level score data of the trauma patient to obtain dynamically adjusted priority score data; dynamically sorting the priorities within and outside the team according to the dynamically adjusted priority score data and performing visualization processing to obtain a visualized dynamic priority queue;

[0097] Step S5: Analyze the patient demand of the visualized dynamic priority queue to obtain patient demand data; generate a patient auxiliary treatment plan according to the patient demand data to obtain a patient auxiliary treatment plan;

[0098] Step S6: Push the emergency treatment team resource data according to the patient auxiliary treatment plan to achieve auxiliary treatment of trauma patients.

[0099] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of the method for constructing an auxiliary treatment system for trauma patients according to the present invention. In this example, the method for constructing an auxiliary treatment system for trauma patients includes the following steps:

[0100] Step S1: collecting multimodal physiological data of trauma patients through hospital examination equipment and preprocessing the physiological data to obtain multimodal feature fusion data; generating physiological feature vectors for the multimodal feature fusion data to obtain physiological feature vectors of trauma patients;

[0101] In an embodiment of the present invention, after a trauma patient arrives at the hospital, a variety of examination equipment (such as an electrocardiograph, a sphygmomanometer, a blood oxygen saturation meter, etc.) is immediately used to collect real-time physiological data, and the collected multi-source data streams are synchronized in time and aligned in format. Subsequently, the synchronized data streams are subjected to noise filtering and artifact elimination to obtain preprocessed physiological data. Finally, the preprocessed data is subjected to physiological index feature extraction and multimodal feature fusion, and finally a feature vector describing the patient's physiological state is generated.

[0102] Step S2: Encode the physiological state of the physiological feature vector of the trauma patient to obtain physiological state encoding data; obtain trauma medical knowledge data; associate the physiological state with the knowledge graph according to the trauma medical knowledge data and the physiological state encoding data to obtain physiological state and knowledge graph association data; generate a trauma stage label according to the physiological state and knowledge graph association data to obtain a trauma stage label;

[0103] In an embodiment of the present invention, the physiological feature vector sequence is divided according to the time window, and the feature sequence in each time window is enhanced. The enhanced feature sequence is learned using a deep learning model to obtain a deep feature embedding vector, which is input into an encoder-decoder network to obtain physiological state encoding data. At the same time, trauma medical knowledge data is obtained from medical databases, clinical guidelines, medical literature, etc., and a trauma knowledge graph containing entities and relationships is constructed. Finally, the physiological state encoding data is associated with the trauma knowledge graph to generate timeline probability trajectory data, and the final trauma stage label is generated through smoothing optimization and key node identification.

[0104] Step S3: Perform multi-dimensional feature fusion according to the trauma stage label to obtain a multi-dimensional comprehensive feature vector; use the pre-trained emergency level prediction model to predict the emergency level probability of the multi-dimensional comprehensive feature vector and calculate the emergency level score to obtain the emergency level score data of the trauma patient;

[0105] In the embodiment of the present invention, the trauma stage label is feature encoded, and combined with the patient's individual information (age, gender, past medical history, etc.) and injury site information, multi-dimensional feature fusion is performed to generate a comprehensive feature vector containing multiple aspects of the patient's information. The emergency level probability prediction model that has been trained in advance is used to predict the comprehensive feature vector, and the emergency level score of the trauma patient is calculated based on the prediction result and the preset emergency weight coefficient.

[0106] Step S4: dynamically adjusting the priority score according to the emergency level score data of the trauma patient to obtain dynamically adjusted priority score data; dynamically sorting the priorities within and outside the team according to the dynamically adjusted priority score data and performing visualization processing to obtain a visualized dynamic priority queue;

[0107] In the embodiment of the present invention, an initial priority queue is generated according to the emergency level score. At the same time, the physiological characteristics of each patient in the queue are continuously monitored, abnormal indicators are identified, risk assessment is performed, and priority adjustment weights are calculated. Then, the patient's priority score is dynamically adjusted according to the priority adjustment weight, and the ranking of the priority queue is updated, and finally a visual dynamic priority queue is obtained, which is convenient for medical staff to grasp the latest situation and priority ranking of patients in real time.

[0108] Step S5: Analyze the patient demand of the visualized dynamic priority queue to obtain patient demand data; generate a patient auxiliary treatment plan according to the patient demand data to obtain a patient auxiliary treatment plan;

[0109] In the embodiment of the present invention, the needs of patients in the visualized dynamic priority queue are analyzed, such as which examinations are needed, how long the monitoring is required, etc. At the same time, the medical resources in the hospital are monitored in real time and the availability is analyzed. Then, according to the patient's needs and available resources, the matching degree of resources and needs is evaluated, and resource matching constraints are set. Finally, the resource allocation plan and the patient's path in the hospital are optimized using an optimization algorithm, and the final patient auxiliary treatment plan is generated.

[0110] Step S6: Push the emergency treatment team resource data according to the patient auxiliary treatment plan to achieve auxiliary treatment of trauma patients.

[0111] In the embodiment of the present invention, key program parameters are extracted from the patient auxiliary treatment program, and a complete and standardized program content is generated according to a preset program template. Then, the qualified emergency treatment team resource data is retrieved according to the program content, and pushed to the corresponding medical staff and relevant departments to notify the relevant personnel to arrive in time and collaborate to complete the patient's auxiliary treatment work.

[0112] Preferably, step S1 comprises the following steps:

[0113] Step S11: collecting real-time physiological data of trauma patients through hospital examination equipment to obtain a physiological data stream of trauma patients;

[0114] Step S12: performing data stream synchronization and alignment on the physiological data stream of the trauma patient to obtain a synchronized data stream; performing noise filtering and artifact elimination on the synchronized data stream to obtain preprocessed data;

[0115] Step S13: extracting physiological index features from the preprocessed data to obtain a physiological index feature set; performing multimodal feature fusion on the physiological index feature set to obtain multimodal feature fusion data;

[0116] Step S14: Generate a physiological feature vector for the multimodal feature fusion data to obtain a physiological feature vector of the trauma patient.

[0117] In an embodiment of the present invention, after a trauma patient arrives at the hospital, a variety of examination equipment (such as an electrocardiograph, a sphygmomanometer, an oximeter, etc.) of the hospital is immediately used to collect real-time physiological data on the patient. The collected data includes, but is not limited to, electrocardiogram signals, blood pressure values, blood oxygen saturation, respiratory rate, body temperature, etc., forming a continuous physiological data stream of the trauma patient. Since the sampling rates and data formats of different examination equipment may differ, it is first necessary to synchronize the time and align the formats of the collected multi-source physiological data streams to ensure the consistency of the data in time. Then, the synchronized data stream is noise filtered to remove the random noise generated during the acquisition process; at the same time, the pseudo-error data caused by equipment failure or human factors is identified and eliminated to ensure the accuracy and reliability of the data, and finally the pre-processed physiological data is obtained. The pre-processed physiological data is analyzed to extract the physiological index characteristics related to the condition of the trauma patient, such as heart rate variability, blood pressure change trend, respiratory rate fluctuation, etc. Since a single physiological indicator may be one-sided, in order to more comprehensively reflect the patient's physiological state, it is necessary to fuse the extracted multimodal physiological indicator feature set, such as using principal component analysis, canonical correlation analysis, etc., to integrate multiple physiological indicator information into a multimodal feature fusion data. Convert the multimodal feature fusion data into a structured physiological feature vector. Each feature vector represents the patient's physiological state at a certain point in time and contains the fusion information of multiple physiological indicators. For example, a vector can be used to represent the comprehensive situation of the patient's heart rate, blood pressure, blood oxygen saturation and other indicators at a certain moment, thereby quantifying the patient's physiological state.

[0118] Preferably, step S2 comprises the following steps:

[0119] Step S21: dividing the physiological characteristic vector of the trauma patient into time windows to obtain time window sequence data; enhancing the physiological characteristic sequence within the time window to obtain time window characteristic sequence enhanced data;

[0120] Step S22: performing deep feature learning on the time window feature sequence enhancement data to obtain a deep feature embedding vector; inputting the deep feature embedding vector into a preset encoder-decoder network to perform physiological state encoding to obtain physiological state encoding data;

[0121] Step S23: Acquire trauma medical knowledge data; construct a trauma knowledge graph based on the trauma medical knowledge data to obtain a trauma knowledge graph;

[0122] Step S24: Associating the physiological state with the knowledge graph of the trauma and the physiological state encoding data to obtain physiological state and knowledge graph association data;

[0123] Step S25: Generate a time axis probability trajectory according to the physiological state and the knowledge graph association data to obtain time axis probability trajectory data; perform probability trajectory smoothing optimization on the time axis probability trajectory data to obtain time axis smoothed data;

[0124] Step S26: Identify key nodes on the time axis smoothing data to obtain time axis key node data; generate trauma stage labels based on the time axis key node data to obtain trauma stage labels.

[0125] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0126] Step S21: dividing the physiological characteristic vector of the trauma patient into time windows to obtain time window sequence data; enhancing the physiological characteristic sequence within the time window to obtain time window characteristic sequence enhanced data;

[0127] In the embodiment of the present invention, the physiological feature vector sequence of the trauma patient is divided according to a certain time window, for example, every 30 seconds or 1 minute is used as a time window, to form time window sequence data. In order to avoid information loss in the time window, the physiological feature sequence in each time window is enhanced, for example, using a sliding window method, a data interpolation method, etc., to enrich the feature information in the time window, and obtain the time window feature sequence enhanced data.

[0128] Step S22: performing deep feature learning on the time window feature sequence enhancement data to obtain a deep feature embedding vector; inputting the deep feature embedding vector into a preset encoder-decoder network to perform physiological state encoding to obtain physiological state encoding data;

[0129] In an embodiment of the present invention, a deep learning model, such as a recurrent neural network (RNN), a long short-term memory network (LSTM), etc., is used to perform deep feature learning on the time window feature sequence enhancement data, and the time series information is compressed and encoded into a deep feature embedding vector, which can better express the trend of changes in the patient's physiological state within the time window. Then the deep feature embedding vector is input into a pre-trained encoder-decoder network, such as a variational autoencoder (VAE), and the high-dimensional deep feature vector is mapped to low-dimensional physiological state encoding data to achieve an abstract representation of the patient's physiological state.

[0130] Step S23: Acquire trauma medical knowledge data; construct a trauma knowledge graph based on the trauma medical knowledge data to obtain a trauma knowledge graph;

[0131] In the embodiment of the present invention, trauma medical knowledge data, such as definitions, symptoms, and judgment criteria of different trauma types, are obtained from medical databases, clinical guidelines, medical literature, and other channels. The obtained trauma medical knowledge data is structured using natural language processing technology to construct a trauma knowledge graph containing entities and relationships. For example, "hemorrhagic shock" is defined as a trauma type entity, "hypotension" is defined as a symptom entity, and a "cause" relationship is established between the two.

[0132] Step S24: Associating the physiological state with the knowledge graph of the trauma and the physiological state encoding data to obtain physiological state and knowledge graph association data;

[0133] In an embodiment of the present invention, the physiological state encoding data is associated with the trauma knowledge graph to find the knowledge graph node or path that is most similar to the current patient's physiological state. For example, the semantic similarity between the physiological state encoding data and each node in the knowledge graph can be calculated to find the most similar node, or a graph neural network model can be used to learn the mapping relationship between the physiological state encoding data and the knowledge graph node.

[0134] Step S25: Generate a time axis probability trajectory according to the physiological state and the knowledge graph association data to obtain time axis probability trajectory data; perform probability trajectory smoothing optimization on the time axis probability trajectory data to obtain time axis smoothed data;

[0135] In an embodiment of the present invention, the probability distribution of the patient being in different trauma stages at different time points is generated based on the physiological state and the knowledge graph association data to form time axis probability trajectory data. For example, the knowledge graph node that is most similar to the patient's current physiological state can be regarded as the trauma stage that the patient is most likely to be in, and the probability of being in other trauma stages is calculated. In order to eliminate noise and fluctuations in the time axis probability trajectory data, a smoothing algorithm, such as a sliding average method, a Gaussian filter method, etc., is used to smooth and optimize the time axis probability trajectory data to obtain time axis smoothed data.

[0136] Step S26: identifying key nodes of the time axis smoothing data to obtain time axis key node data; generating trauma stage labels according to the time axis key node data to obtain trauma stage labels;

[0137] In the embodiment of the present invention, the time axis smoothed data is analyzed to identify key nodes representing changes in the patient's trauma stage, such as points where the probability changes suddenly. Based on the key nodes and the time axis smoothed data, a final trauma stage label is generated, such as labeling the patient as the "hemorrhagic shock" stage in a certain time period and the "recovery" stage in another time period.

[0138] Preferably, step S24 comprises the following steps:

[0139] Step S241: probabilizing the trauma knowledge graph to obtain a probabilistic trauma knowledge graph; calculating the physiological state probability distribution of the physiological state coding data according to the probabilistic trauma knowledge graph to obtain physiological state probability distribution data;

[0140] Step S242: constructing a Bayesian network according to the probabilistic trauma knowledge graph to obtain a Bayesian network; inputting the physiological state probability distribution data into the Bayesian network, setting observation variables, and obtaining observation variables;

[0141] Step S243: performing latent variable state inference according to the Bayesian network and the observed variables to obtain latent variable state change data;

[0142] Step S244: evaluating the probability of the time nodes of the patient being in different trauma stages according to the latent variable state change data to obtain the time node probability data;

[0143] Step S245: Perform dynamic Bayesian network joint reasoning based on the Bayesian network, time node probability data, and physiological state probability distribution data to obtain physiological state and knowledge graph association data.

[0144] In the embodiment of the present invention, it is first necessary to perform probabilistic processing on it. For example, a probability value can be assigned to each edge in the knowledge graph based on existing data or expert experience to indicate the possibility of the relationship occurring. Then, according to the probabilistic trauma knowledge graph, the probability distribution of the patient's current physiological state encoding data on each node in the knowledge graph is calculated to obtain physiological state probability distribution data, which is used to indicate the possibility that the patient is in different trauma stages. According to the probabilistic trauma knowledge graph, a Bayesian network model is constructed. The entities and relationships in the knowledge graph are mapped to nodes and edges in the Bayesian network, and the conditional probabilities in the Bayesian network are set according to the probability values ​​of the edges in the knowledge graph. The physiological state probability distribution data is input into the Bayesian network as an observed variable, for example, the probability that the patient is in the "hemorrhagic shock" stage is set as the value of the observed variable. In addition to the observed variables, there are some hidden variables in the Bayesian network that cannot be directly observed, such as the specific amount of bleeding and the degree of injury of the patient. Using Bayesian network inference algorithms, such as belief propagation algorithms and importance sampling algorithms, the state of latent variables is inferred based on observed variables (physiological state probability distribution data) to obtain latent variable state change data, such as the probability distribution of patient bleeding over time. According to the latent variable state change data, the probability of the patient being in different trauma stages at different time points is evaluated. For example, the probability of the patient entering the "hemorrhagic shock" stage at a certain time point in the future can be evaluated based on the change trend of the patient's bleeding volume, and the time node probability data can be obtained. Combined with the Bayesian network, time node probability data and physiological state probability distribution data, dynamic Bayesian network joint reasoning is performed to obtain the final physiological state and knowledge graph association data. For example, the probability values ​​of the nodes in the Bayesian network can be updated based on the physiological state probability distribution data and time node probability data of the patient at different time points, and the trauma stage and its probability that the patient is most likely to experience in the future period of time can be inferred based on the updated Bayesian network.

[0145] Preferably, step S3 comprises the following steps:

[0146] Step S31: performing trauma stage feature encoding on the trauma stage label to obtain trauma stage feature encoding data;

[0147] Step S32: obtaining individual information of the trauma patient; characterizing the individual information of the trauma patient to obtain individual information characteristic data;

[0148] Step S33: obtaining the injury site information of the trauma patient; quantifying the injury site information of the patient to obtain quantified injury site information data;

[0149] Step S34: performing multi-dimensional feature fusion on the trauma stage feature coding data, the individual information feature data and the injury site information quantification data to obtain a multi-dimensional comprehensive feature vector;

[0150] Step S35: using the pre-trained emergency level prediction model to perform emergency level probability prediction on the multi-dimensional comprehensive feature vector to obtain an emergency level probability value;

[0151] Step S36: Calculate the emergency level score according to the emergency level probability value and the preset emergency weight coefficient to obtain the emergency level score data of the trauma patient.

[0152] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0153] Step S31: performing trauma stage feature encoding on the trauma stage label to obtain trauma stage feature encoding data;

[0154] In the embodiment of the present invention, the trauma stage label is converted into a numerical form that can be recognized by the machine learning model, that is, the trauma stage feature encoding is performed. For example, one-hot encoding can be used to represent each trauma stage label as a vector, such as [1,0,0] represents the "hemorrhagic shock" stage, [0,1,0] represents the "recovery" stage, and so on.

[0155] Step S32: obtaining individual information of the trauma patient; characterizing the individual information of the trauma patient to obtain individual information characteristic data;

[0156] In the embodiment of the present invention, individual information of trauma patients is obtained, such as age, gender, past medical history, allergy history, etc. These individual information are characterized and converted into numerical or categorical features, such as converting age into continuous numerical features and gender into binary categorical features (0 or 1).

[0157] Step S33: obtaining the injury site information of the trauma patient; quantifying the injury site information of the patient to obtain quantified injury site information data;

[0158] In an embodiment of the present invention, the injury site information of the trauma patient is obtained, such as the head, chest, abdomen, etc. The injury site information is quantified, for example, each injury site can be converted into a specific code using an anatomical site coding system, or multiple injury sites can be represented as a multidimensional vector using one-hot coding.

[0159] Step S34: performing multi-dimensional feature fusion on the trauma stage feature coding data, the individual information feature data and the injury site information quantification data to obtain a multi-dimensional comprehensive feature vector;

[0160] In the embodiment of the present invention, the trauma stage feature coding data, individual information feature data and injury site information quantification data are multi-dimensionally fused to form a comprehensive feature vector containing multiple aspects of patient information. For example, the above three feature data can be concatenated into a long vector, or a more complex feature fusion method can be used, such as feature fusion based on an attention mechanism.

[0161] Step S35: using the pre-trained emergency level prediction model to perform emergency level probability prediction on the multi-dimensional comprehensive feature vector to obtain an emergency level probability value;

[0162] In the embodiment of the present invention, a pre-trained emergency level prediction model is used to predict the probability of emergency level for a multi-dimensional comprehensive feature vector. The model can be any machine learning model suitable for classification tasks, such as logistic regression, support vector machine, random forest, neural network, etc. The input of the model is the multi-dimensional comprehensive feature vector of the patient, and the output is the probability value of the patient being at different emergency levels, for example, the probability of the patient being at "first-level emergency" is 0.8, and the probability of "second-level emergency" is 0.2.

[0163] Step S36: Calculate the emergency level score according to the emergency level probability value and the preset emergency weight coefficient to obtain the emergency level score data of the trauma patient;

[0164] In the embodiment of the present invention, the final emergency level score of the trauma patient is calculated based on the emergency level probability value and the pre-set emergency weight coefficient. For example, different weights can be assigned to different emergency levels, such as a weight of 10 for "first-level emergency" and a weight of 5 for "second-level emergency", and then a weighted average is performed based on the probability values ​​of the patient at different emergency levels and the corresponding weights to obtain the final emergency level score.

[0165] Preferably, step S35 includes the following steps:

[0166] Step S351: Collecting comprehensive data of historical trauma patients from the medical system to obtain comprehensive data of historical trauma patients; constructing an emergency level sample library based on the comprehensive data of historical trauma patients to obtain an initial emergency level sample library; performing sample category balancing on the initial emergency level sample library to obtain an emergency level sample library;

[0167] Step S352: constructing a multimodal feature graph for the multimodal feature data of patients in the emergency level sample library to obtain a multimodal feature graph;

[0168] Step S353: performing feature correlation calculation on the multimodal feature graph to obtain feature correlation data;

[0169] Step S354: using the feature correlation data to assign graph node weights to the multimodal feature graph to obtain weighted feature graph structure data;

[0170] Step S355: Select a graph neural network model according to the weight feature graph structure data to obtain a graph neural network model;

[0171] Step S356: Train the graph neural network model for emergency level prediction and perform parameter optimization to obtain a pre-trained emergency level prediction model.

[0172] In an embodiment of the present invention, comprehensive data of historical trauma patients are collected from the hospital's medical information system, including the patient's physiological index data, trauma stage label, individual information, injury site information, and the final emergency level. The collected historical data is used to construct an initial emergency level sample library. Since the number of samples of different emergency levels may be quite different, for example, the number of patients with first-level emergency is much less than the number of patients with second-level emergency, in order to avoid deviations in model training, it is necessary to balance the sample categories of the initial emergency level sample library, such as using oversampling, undersampling, data enhancement, etc., so that the number of samples of different emergency levels is roughly the same. A multimodal feature graph is constructed for the multimodal feature data of patients in the emergency level sample library (including physiological index features, individual information features, injury site information features, etc.). For example, a graph database can be used to store feature data of different modalities as nodes, and edges can be established according to the association relationship between features to construct a graph structure that can reflect the interdependence between multimodal features. Feature correlation calculation is performed on the multimodal feature graph, for example, the correlation between different features can be calculated using methods such as Pearson correlation coefficient and Spearman correlation coefficient, and feature correlation data is obtained to represent the strength of association between different features. According to the feature correlation data, weights are assigned to the nodes in the multimodal feature graph. For example, a weight can be assigned to each node based on indicators such as the degree centrality and betweenness centrality of the node, as well as the correlation with other nodes, indicating the importance of the node in the entire graph structure. According to the weight feature graph structure data, a suitable graph neural network model is selected, such as graph convolutional network (GCN), graph attention network (GAT), etc. The selection of the model needs to consider factors such as the characteristics of the graph structure, feature dimensions, and computational complexity. The emergency level prediction model is trained using the emergency level sample library and the graph neural network model. During the training process, the node weight is used to optimize the model, such as using the node weight as the weight coefficient of the loss function, or introducing the node weight information in the model structure. Through training and parameter optimization, a pre-trained emergency level prediction model is finally obtained.

[0173] Preferably, step S4 comprises the following steps:

[0174] Step S41: generating an initial priority queue according to the emergency level score data of trauma patients to obtain an initial priority queue;

[0175] Step S42: monitoring physiological characteristic changes of each patient in the trauma patient emergency level score data to obtain physiological characteristic change trend data;

[0176] Step S43: performing indicator abnormality identification on the physiological characteristic change trend data to obtain physiological indicator abnormality record data; performing risk level assessment on the physiological indicator abnormality record data according to the trauma medical knowledge data to obtain abnormal indicator risk data;

[0177] Step S44: Calculate the priority adjustment weight according to the abnormal indicator risk data to obtain the priority adjustment weight data;

[0178] Step S45: dynamically adjusting the priority score of the priority adjustment weight data and the trauma patient emergency level score data to obtain dynamically adjusted priority score data;

[0179] Step S46: updating the order in the queue according to the dynamically adjusted priority score data to obtain updated order data in the queue; performing upgrade and downgrade processing between queues according to the updated order data in the queue to obtain a dynamic priority queue;

[0180] Step S47: Visualize the queue state of the dynamic priority queue to obtain a visualized dynamic priority queue.

[0181] In an embodiment of the present invention, according to the emergency level score data of trauma patients, the patients are sorted in order from high to low scores to generate an initial priority queue. For example, a priority queue data structure can be used to store patient information and corresponding emergency level scores, and the head of the queue is always the patient with the highest score, that is, the patient who needs to be treated with the highest priority. Continuous physiological characteristics monitoring, such as heart rate, blood pressure, respiratory rate, blood oxygen saturation, etc., are performed on each patient in the queue, and the change trend of these physiological characteristics over time is recorded to obtain physiological characteristics change trend data. The physiological characteristics change trend data is analyzed to identify abnormal indicators beyond the normal range, such as a sudden increase in heart rate, a continuous decrease in blood pressure, etc., and the type, degree and duration of the abnormal indicators are recorded to obtain physiological indicator abnormality record data. Then, according to trauma medical knowledge data, such as typical physiological indicator change patterns corresponding to different types of trauma and different severities, the physiological indicator abnormality record data is evaluated for risk level, for example, the risk level of cardiac arrest is evaluated as the highest level, and slight blood pressure fluctuations are evaluated as low risk levels, to obtain abnormal indicator risk data. According to the abnormal indicator risk data, the priority adjustment weight of each patient is calculated. For example, a priority adjustment weight calculation formula can be designed based on factors such as the risk level, number, and duration of abnormal indicators. The higher the risk level, the greater the number, and the longer the duration of the patient, the greater the priority adjustment weight. According to the priority adjustment weight, the emergency level score of the trauma patient is dynamically adjusted to obtain a dynamically adjusted priority score. For example, the priority adjustment weight can be weighted and summed with the initial emergency level score, or the priority adjustment weight can be used as a multiplication factor of the initial emergency level score to obtain a new priority score. According to the dynamically adjusted priority score, the priority queue is updated. For example, the queue can be reordered according to the new score, or only the scores and positions of some patients can be updated. For patients whose scores have changed significantly, inter-queue promotion and demotion can also be performed, such as moving patients with significantly improved scores from a low-level queue to a high-level queue so that they can obtain higher-level medical resources and treatment in a timely manner. The dynamically updated priority queue is presented to medical staff in a visual manner, such as using a graphical interface, an electronic screen, etc. to display the ID, name, emergency level, main physiological indicators, dynamic change trends, and other information of each patient in the queue.

[0182] Preferably, step S43 includes the following steps:

[0183] Step S431: extracting physiological index time series from the physiological characteristic change trend data to obtain a physiological index time series; constructing a dynamic baseline according to the physiological index time series to obtain a dynamic baseline;

[0184] Step S432: performing indicator abnormality identification on the physiological indicator time series and dynamic baseline to obtain physiological indicator abnormality record data;

[0185] Step S433: quantifying the abnormality index of the abnormal physiological index record data to obtain quantified abnormality index data;

[0186] Step S434: performing risk factor association on the quantitative abnormal index data and the trauma medical knowledge data to obtain index-risk association data;

[0187] Step S435: using a preset risk scoring rule to evaluate the risk level of the quantitative abnormal indicator data according to the indicator-risk association data to obtain abnormal indicator risk data.

[0188] In an embodiment of the present invention, time series data of each physiological index, such as heart rate time series, blood pressure time series, etc., are extracted from the physiological characteristic change trend data. Considering the individual differences of patients and the development of the disease, the traditional fixed baseline method is difficult to accurately identify the abnormality of the index. Therefore, it is necessary to use a dynamic baseline construction method, such as a sliding window method, an exponential smoothing method, etc., according to the physiological index time series of each patient, to obtain a dynamic baseline that can reflect the patient's current physiological state. The physiological index time series is compared with the corresponding dynamic baseline to identify the abnormality of the index that exceeds the preset range of the dynamic baseline. For example, a threshold can be set. When the physiological index value exceeds the dynamic baseline plus or minus threshold, it is determined to be an abnormal index, and the time, duration, degree of deviation, and other information of the abnormal occurrence are recorded to obtain the abnormal physiological index record data. The abnormal physiological index record data is quantified, and the degree and duration of the abnormality are converted into numerical form to facilitate subsequent analysis and calculation. For example, the degree of deviation, duration, rate of change, and other indicators can be used to quantify the abnormality to obtain quantitative abnormal index data. According to trauma medical knowledge data, such as typical physiological index change patterns corresponding to different types of trauma and different degrees of severity, the quantitative abnormal index data is associated with potential risk factors. For example, abnormally accelerated heart rate can be associated with the risk of hemorrhagic shock, and a continuous drop in blood pressure can be associated with the risk of intracranial hemorrhage, etc., to obtain indicator-risk association data. According to pre-set risk scoring rules, such as expert scoring methods, machine learning models, etc., combined with indicator-risk association data, the risk level of quantitative abnormal indicator data is assessed. For example, the risk score of each abnormal indicator can be calculated based on factors such as the quantitative value of the abnormal indicator and the severity of the associated risk factors to obtain abnormal indicator risk data for subsequent priority adjustment weight calculations.

[0189] Preferably, step S5 comprises the following steps:

[0190] Step S51: Real-time monitoring of system medical resources to obtain medical resource status data; resource availability analysis of the medical resource status data to obtain available resource data;

[0191] Step S52: Analyze patient demand on the visualized dynamic priority queue to obtain patient demand data;

[0192] Step S53: Evaluate the matching degree between resources and needs of the patient's needs data and available resource data to obtain resource matching data;

[0193] Step S54: setting constraint conditions according to the matching degree data to obtain resource matching constraint data;

[0194] Step S55: constructing a path planning objective function according to the resource matching degree data and the resource matching constraint data to obtain a path planning objective function;

[0195] Step S56: Utilizing a genetic algorithm to optimize the resource allocation scheme and the patient path according to the path planning objective function, the resource matching degree data, and the resource matching constraint data, and obtaining resource allocation path optimization data;

[0196] Step S57: Generate a patient-assisted treatment plan based on the resource allocation path optimization data to obtain a patient-assisted treatment plan.

[0197] In an embodiment of the present invention, medical resources in a hospital, such as beds, operating rooms, medical staff, medical equipment, etc., are monitored in real time, and real-time status information of resources, such as idle / occupied status, geographical location, available time, etc., is collected to obtain medical resource status data. Then, according to preset resource scheduling rules and algorithms, the medical resource status data is analyzed, such as the number of available beds, idle time periods in operating rooms, the number of doctors in different departments, etc., to obtain available resource data that can be used to treat current trauma patients. According to the visualized dynamic priority queue, as well as the electronic medical records, trauma information, physiological indicators and other data of each patient, the patient's treatment needs are analyzed to obtain patient demand data. According to the available resource data and the patient demand data, the resource and demand matching degree is evaluated, such as calculating the matching score between each patient and different resources, considering factors such as resource type, quantity, distance, time, etc., to obtain resource matching degree data for guiding subsequent resource allocation and path optimization. According to the actual situation of the hospital and the medical resource scheduling rules, resource matching constraints are set, such as some special resources can only be allocated to specific types of patients, each doctor can only treat a certain number of patients at the same time, and resource allocation needs to consider time window constraints, etc., to obtain resource matching constraint data, which is used to limit the subsequent optimization algorithm search space. According to the resource matching degree data and resource matching constraint data, a path planning objective function is constructed to evaluate the pros and cons of the resource allocation plan. The objective function can consider multiple optimization goals, such as maximizing resource utilization, minimizing patient waiting time, etc., and set the weights of different goals according to actual needs to obtain a path planning objective function that can comprehensively measure the pros and cons of the resource allocation plan. Using intelligent optimization algorithms such as genetic algorithms, according to the path planning objective function, resource matching degree data, and resource matching constraint data, the resource allocation plan and the patient's path in the hospital are optimized to obtain resource allocation path optimization data. According to the resource allocation path optimization data, combined with the patient's specific situation and treatment needs, the final patient auxiliary treatment plan is generated, such as reminding relevant departments to prepare in advance, etc., and displayed to medical staff in a visual and easy-to-understand way to assist medical staff in patient treatment and improve the efficiency of medical resource utilization.

[0198] Preferably, step S6 comprises the following steps:

[0199] Step S61: extracting parameters of the patient auxiliary treatment plan to obtain treatment plan parameters;

[0200] Step S62: combining the processing solution parameters with the solution content according to the preset solution template to obtain a content combination solution;

[0201] Step S63: Generate emergency processing team resource data and emergency processing team data according to the content combination plan;

[0202] Step S64: Pushing data to the emergency treatment team's resource data to assist in the treatment of trauma patients.

[0203] In an embodiment of the present invention, key instruction parameters are extracted from the patient auxiliary treatment plan, such as what types of medical staff (doctors, nurses, radiologists, etc.) are needed, and what medical equipment (ventilators, monitors, infusion pumps, etc.) are needed, to form structured treatment plan parameters. According to the pre-set instruction template, the treatment plan parameters are filled into the template to generate complete, standardized, and easy-to-understand instruction content. For example, a template of "{time}, please {department} {title} doctor to {location} need {patient name} to perform {operation}" can be used, and specific instruction parameters can be filled into the corresponding placeholders to generate instruction content similar to "10:00, please ask the chief physician of general surgery to go to operating room No. 3 to perform surgery on Zhang San". According to the content combination instruction, the emergency treatment team resource data that meets the requirements of the plan is retrieved from the hospital's personnel information system, equipment management system and other databases, such as finding qualified doctors, nurses, equipment, etc., and organizing them into a structured data format to facilitate subsequent data push. The resource data of the emergency treatment team will be pushed to the corresponding medical staff and relevant departments in a timely manner through text messages, APP push, electronic screen display, etc. For example, surgical parameters will be pushed to doctors and nurses in relevant departments, and examination parameters will be pushed to radiologists and technicians, etc., to notify relevant personnel to arrive in time and collaborate to complete the auxiliary treatment of patients.

[0204] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0205] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for constructing an auxiliary treatment system based on trauma patients, characterized in that: The following steps are involved: Step S1: collecting multimodal physiological data of trauma patients through hospital examination equipment and preprocessing the physiological data to obtain multimodal feature fusion data; Generate physiological feature vectors for multimodal feature fusion data to obtain physiological feature vectors of trauma patients; Step S2: Encoding the physiological state of the physiological feature vector of the trauma patient to obtain physiological state encoding data; Acquire trauma medical knowledge data; associate the physiological state with the knowledge graph based on the trauma medical knowledge data and the physiological state coding data to obtain physiological state and knowledge graph association data; Generate trauma stage labels based on physiological status and knowledge graph association data to obtain trauma stage labels; Step S3: Perform multi-dimensional feature fusion according to the trauma stage label to obtain a multi-dimensional comprehensive feature vector; use the pre-trained emergency level prediction model to predict the emergency level probability of the multi-dimensional comprehensive feature vector and calculate the emergency level score to obtain the emergency level score data of the trauma patient; Step S4: dynamically adjusting the priority score according to the emergency level score data of the trauma patient to obtain dynamically adjusted priority score data; dynamically sorting the priorities within and outside the team according to the dynamically adjusted priority score data and performing visualization processing to obtain a visualized dynamic priority queue; Step S5: Analyze the patient demand of the visualized dynamic priority queue to obtain patient demand data; generate a patient auxiliary treatment plan according to the patient demand data to obtain a patient auxiliary treatment plan; Step S6: Push the emergency treatment team resource data according to the patient auxiliary treatment plan to achieve auxiliary treatment of trauma patients.

2. The method for constructing an auxiliary treatment system based on trauma patients according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: collecting real-time physiological data of trauma patients through hospital examination equipment to obtain a physiological data stream of trauma patients; Step S12: performing data stream synchronization and alignment on the physiological data stream of the trauma patient to obtain a synchronized data stream; performing noise filtering and artifact elimination on the synchronized data stream to obtain preprocessed data; Step S13: extracting physiological index features from the preprocessed data to obtain a physiological index feature set; performing multimodal feature fusion on the physiological index feature set to obtain multimodal feature fusion data; Step S14: Generate a physiological feature vector for the multimodal feature fusion data to obtain a physiological feature vector of the trauma patient.

3. The method for constructing an auxiliary treatment system based on trauma patients according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: dividing the physiological characteristic vector of the trauma patient into time windows to obtain time window sequence data; enhancing the physiological characteristic sequence within the time window to obtain time window characteristic sequence enhanced data; Step S22: performing deep feature learning on the time window feature sequence enhancement data to obtain a deep feature embedding vector; inputting the deep feature embedding vector into a preset encoder-decoder network to perform physiological state encoding to obtain physiological state encoding data; Step S23: Acquire trauma medical knowledge data; construct a trauma knowledge graph based on the trauma medical knowledge data to obtain a trauma knowledge graph; Step S24: Associating the physiological state with the knowledge graph of the trauma and the physiological state encoding data to obtain physiological state and knowledge graph association data; Step S25: Generate a time axis probability trajectory according to the physiological state and the knowledge graph association data to obtain time axis probability trajectory data; perform probability trajectory smoothing optimization on the time axis probability trajectory data to obtain time axis smoothed data; Step S26: Identify key nodes on the time axis smoothing data to obtain time axis key node data; generate trauma stage labels based on the time axis key node data to obtain trauma stage labels.

4. The method for constructing an auxiliary treatment system based on trauma patients according to claim 3, characterized in that: Step S24 includes the following steps: Step S241: probabilizing the trauma knowledge graph to obtain a probabilistic trauma knowledge graph; calculating the physiological state probability distribution of the physiological state coding data according to the probabilistic trauma knowledge graph to obtain physiological state probability distribution data; Step S242: constructing a Bayesian network according to the probabilistic trauma knowledge graph to obtain a Bayesian network; inputting the physiological state probability distribution data into the Bayesian network, setting observation variables, and obtaining observation variables; Step S243: performing latent variable state inference according to the Bayesian network and the observed variables to obtain latent variable state change data; Step S244: evaluating the probability of the time nodes of the patient being in different trauma stages according to the latent variable state change data to obtain the time node probability data; Step S245: Perform dynamic Bayesian network joint reasoning based on the Bayesian network, time node probability data, and physiological state probability distribution data to obtain physiological state and knowledge graph association data.

5. The method for constructing an auxiliary treatment system based on trauma patients according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing trauma stage feature encoding on the trauma stage label to obtain trauma stage feature encoding data; Step S32: obtaining individual information of the trauma patient; characterizing the individual information of the trauma patient to obtain individual information characteristic data; Step S33: obtaining the injury site information of the trauma patient; quantifying the injury site information of the patient to obtain quantified injury site information data; Step S34: performing multi-dimensional feature fusion on the trauma stage feature coding data, the individual information feature data and the injury site information quantification data to obtain a multi-dimensional comprehensive feature vector; Step S35: using the pre-trained emergency level prediction model to perform emergency level probability prediction on the multi-dimensional comprehensive feature vector to obtain an emergency level probability value; Step S36: Calculate the emergency level score according to the emergency level probability value and the preset emergency weight coefficient to obtain the emergency level score data of the trauma patient.

6. The method for constructing an auxiliary treatment system based on trauma patients according to claim 6, characterized in that: Step S35 includes the following steps: Step S351: Collecting comprehensive data of historical trauma patients from the medical system to obtain comprehensive data of historical trauma patients; constructing an emergency level sample library based on the comprehensive data of historical trauma patients to obtain an initial emergency level sample library; performing sample category balancing on the initial emergency level sample library to obtain an emergency level sample library; Step S352: constructing a multimodal feature graph for the multimodal feature data of patients in the emergency level sample library to obtain a multimodal feature graph; Step S353: performing feature correlation calculation on the multimodal feature graph to obtain feature correlation data; Step S354: using the feature correlation data to assign graph node weights to the multimodal feature graph to obtain weighted feature graph structure data; Step S355: Select a graph neural network model according to the weight feature graph structure data to obtain a graph neural network model; Step S356: Train the graph neural network model for emergency level prediction and perform parameter optimization to obtain a pre-trained emergency level prediction model.

7. The method for constructing an auxiliary treatment system based on trauma patients according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: generating an initial priority queue according to the emergency level score data of trauma patients to obtain an initial priority queue; Step S42: monitoring physiological characteristic changes of each patient in the trauma patient emergency level score data to obtain physiological characteristic change trend data; Step S43: performing indicator abnormality identification on the physiological characteristic change trend data to obtain physiological indicator abnormality record data; performing risk level assessment on the physiological indicator abnormality record data according to the trauma medical knowledge data to obtain abnormal indicator risk data; Step S44: Calculate the priority adjustment weight according to the abnormal indicator risk data to obtain the priority adjustment weight data; Step S45: dynamically adjusting the priority score of the priority adjustment weight data and the trauma patient emergency level score data to obtain dynamically adjusted priority score data; Step S46: updating the order in the queue according to the dynamically adjusted priority score data to obtain updated order data in the queue; performing upgrade and downgrade processing between queues according to the updated order data in the queue to obtain a dynamic priority queue; Step S47: Visualize the queue state of the dynamic priority queue to obtain a visualized dynamic priority queue.

8. The method for constructing an auxiliary treatment system based on trauma patients according to claim 1, characterized in that: Step S43 includes the following steps: Step S431: extracting physiological index time series from the physiological characteristic change trend data to obtain a physiological index time series; constructing a dynamic baseline according to the physiological index time series to obtain a dynamic baseline; Step S432: performing indicator abnormality identification on the physiological indicator time series and dynamic baseline to obtain physiological indicator abnormality record data; Step S433: quantifying the abnormality index of the abnormal physiological index record data to obtain quantified abnormality index data; Step S434: performing risk factor association on the quantitative abnormal index data and the trauma medical knowledge data to obtain index-risk association data; Step S435: using a preset risk scoring rule to evaluate the risk level of the quantitative abnormal indicator data according to the indicator-risk association data to obtain abnormal indicator risk data.

9. The method for constructing an auxiliary treatment system based on trauma patients according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: Real-time monitoring of system medical resources to obtain medical resource status data; resource availability analysis of the medical resource status data to obtain available resource data; Step S52: Analyze patient demand on the visualized dynamic priority queue to obtain patient demand data; Step S53: Evaluate the matching degree between resources and needs of the patient's needs data and available resource data to obtain resource matching data; Step S54: setting constraint conditions according to the matching degree data to obtain resource matching constraint data; Step S55: constructing a path planning objective function according to the resource matching degree data and the resource matching constraint data to obtain a path planning objective function; Step S56: Utilizing a genetic algorithm to optimize the resource allocation scheme and the patient path according to the path planning objective function, the resource matching degree data, and the resource matching constraint data, and obtaining resource allocation path optimization data; Step S57: Generate a patient-assisted treatment plan based on the resource allocation path optimization data to obtain a patient-assisted treatment plan.

10. The method for constructing an auxiliary treatment system based on trauma patients according to claim 1, characterized in that: Step S6 includes the following steps: Step S61: extracting parameters of the patient auxiliary treatment plan to obtain treatment plan parameters; Step S62: combining the processing solution parameters with the solution content according to the preset solution template to obtain a content combination solution; Step S63: Generate emergency processing team resource data and emergency processing team data according to the content combination plan; Step S64: Pushing data to the emergency treatment team's resource data to assist in the treatment of trauma patients.

Images