Situalized medical care emergency training method based on intelligent decision tree

Abstract

The invention discloses a situational medical care emergency training method based on an intelligent decision tree, and the method comprises the following steps: S1, collecting and preprocessing medical care emergency historical case data, and extracting an expert path; s2, constructing a medical emergency situation virtual environment based on the structured data and initializing an intelligent decision tree; s3, performing evolution operation on the plurality of decision paths by adopting a genetic algorithm to generate a decision path set; s4, fusing the decision path and the expert path by adopting a belief propagation algorithm to generate an emergency decision path; s5, emergency training is carried out in the virtual environment, recording and evaluation are carried out, and an adjusted path is formed; and S6, redeploying the adjusted path to the decision tree, and carrying out targeted drilling. According to the method, the intelligent decision tree integrating genetic optimization and a belief propagation mechanism is constructed, so that accurate and efficient training and path optimization of medical staff in a situational emergency environment are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of virtual simulation, and in particular to a situational medical emergency training method based on an intelligent decision tree. BACKGROUND

[0002] In the existing medical training system, the cultivation of emergency response ability usually depends on classroom teaching, paper textbook learning and regular centralized drills. This kind of traditional training method has a certain effect on knowledge transmission and basic process familiarization, but its real simulation ability for emergencies is limited, and it cannot fully reproduce the emergency situation of complex scene, rapid decision-making and highly coupled behavior. Therefore, when medical personnel actually face an emergency, there are often problems such as operation delay, improper decision-making or unsmooth team cooperation, which affects the timeliness and accuracy of emergency disposal.

[0003] In recent years, virtual simulation technology and intelligent decision-making system have gradually emerged in medical training, especially in reconstructing real scenes, simulating event evolution and interactive feedback. Some systems have initially introduced situational simulation platforms and rule engines to guide training paths, and some research has tried to automatically generate drill processes based on case libraries. However, these systems generally have fixed training paths, lack individual difference response mechanisms, and have imperfect feedback loops, making it difficult to meet the needs of high complexity, multiple paths and multiple strategies for emergency training. In addition, although some technologies integrate path recommendation or model prediction modules, they lack expert path references based on real cases, making it difficult to ensure the authority and applicability of the recommended path.

[0004] In terms of path optimization, some research attempts to use machine learning methods to model and improve the decision-making process, such as reinforcement learning or decision tree optimization, but these methods often ignore key issues such as multi-strategy integration, semantic consistency between paths, and confidence information transmission, resulting in poor explainability of training paths, unstable optimization effects, and difficulty in guiding medical personnel to make correct responses in complex emergency situations. In addition, the feedback of training results is difficult to effectively integrate into subsequent path optimization, lacking a dynamic updating and continuous evolution mechanism.

[0005] Therefore, how to provide a situational medical emergency training method based on an intelligent decision tree is a problem that needs to be solved by those skilled in the art. SUMMARY

[0006] One purpose of the present application is to propose a contextualized medical emergency training method based on intelligent decision tree. The present application fully integrates various artificial intelligence technologies such as expert path extraction, genetic algorithm evolution optimization, belief propagation path fusion, dynamic path evaluation and artificial fine-tuning, and describes in detail the emergency training process that is intelligent, personalized and sustainable in the medical emergency situation virtual environment. The present application collects and analyzes historical medical emergency case data through structured collection, constructs an authoritative expert path, combines with an intelligent decision tree, uses an evolutionary algorithm to globally optimize multiple decision paths, and then realizes the reliable fusion and dynamic update of the paths through a belief propagation algorithm. Finally, the experience fine-tuning and feedback integration are combined with the actual training performance of medical personnel to continuously improve the emergency decision path, not only improving the response efficiency and decision accuracy of medical personnel in real emergency scenes, but also realizing the intelligent generation of individualized training paths, with the advantages of strong scene adaptability, quantifiable training effect, efficient path optimization iteration and sustainable closed-loop training system.

[0007] The contextualized medical emergency training method based on intelligent decision tree according to the embodiment of the present application comprises the following steps:

[0008] S1, collecting medical emergency historical case data, preprocessing the medical emergency historical case data, generating structured data, extracting paths verified by medical emergency historical cases and conforming to emergency operation specifications, and generating expert paths;

[0009] S2, based on the structured data, constructing a medical emergency situation virtual environment, setting multiple key decision points in the medical emergency situation virtual environment, and initializing a medical emergency intelligent decision tree;

[0010] S3, based on the medical emergency intelligent decision tree, using a genetic algorithm to perform evolutionary operations on multiple decision paths therein, and generating a medical emergency decision path set;

[0011] S4, using a belief propagation algorithm to transfer the confidence of the node information between the medical emergency decision path set and the expert path, realizing the reliable fusion of the decision nodes through the information interaction between the paths, and generating an emergency decision path;

[0012] S5, organizing medical personnel to perform emergency training operations in the medical emergency situation virtual environment according to the emergency decision path, recording key behavior data and training performance, comprehensively evaluating the execution effect of the key nodes of the emergency decision path, and performing artificial fine-tuning and experience summary according to the evaluation results, to form an adjusted emergency decision path;

[0013] S6, redeploying the adjusted emergency decision path to the medical emergency intelligent decision tree, and guiding medical personnel to carry out targeted drills in the medical emergency situation virtual environment.

[0014] Optionally, the medical emergency historical case data specifically includes emergency event types, spatial arrangement information, and task elements.

[0015] Optionally, the preprocessing of the medical emergency historical case data specifically includes data cleaning, denoising, missing data completion, format standardization, and key field extraction.

[0016] Optionally, S2 specifically includes:

[0017] S21, collecting emergency event types, spatial arrangement information, and task elements in structured data, and constructing a scene factor matrix , wherein, represents the semantic matching degree between the i-th event and the j-th environmental factor;

[0018] S22, performing semantic scene mapping according to the scene factor matrix , and calling a three-dimensional modeling engine to construct a medical emergency scenario virtual environment with spatial layout, task flow, and interactive element configuration, the medical emergency scenario virtual environment including operation areas, path nodes, device distribution, and personnel positions;

[0019] S23, introducing an event label-semantic dictionary mapping mechanism to normalize the concept of labels in structured data, forming a standard semantic template set ;

[0020] S24, injecting medical role entities from a pre-set role library according to the scene task type, and performing medical role binding operations to map each type of medical role to executable operation nodes, the operation nodes being matched and selected according to corresponding template items in the standard semantic template set ;

[0021] S25, identifying key operation nodes in the medical emergency scenario virtual environment, constructing a key decision point set , and assigning a corresponding input state set and a behavior set to each decision point ;

[0022] S26, automatically generating a medical emergency intelligent decision tree structure based on the configured nodes and behaviors in the scene , wherein , represents a node state transition path triggered by the behavior set , wherein, represents a node set in the medical emergency intelligent decision tree;

[0023] ​S27, embedding the built medical emergency intelligent decision tree into the medical emergency scenario virtual environment.

[0024] Optionally, the S3 specifically comprises:

[0025] S31, extracting a path set from the medical emergency intelligent decision tree as a to-be-optimized path set;

[0026] S32, constructing an initial path set , wherein, is an expert path, is a path randomly generated based on structured data, and the initial path set is an initial population of a genetic algorithm;

[0027] S33, representing each path in the initial path set as a sequence of state-action pairs and converting it into a chromosome representation

[0028]

[0029] wherein, represents an i-th decision node in an i-th path, represents an action decision taken at the decision node

[0030] S34, calculating a fitness value for each path

[0031] S35, sorting the paths in the initial path set according to fitness function values, performing selection operation by using a roulette method, and generating a path subset for cross operation;

[0032] S36, defining a structure legality constraint set , the structure legality constraint set being composed of legal behavior continuous pairs extracted from structured data;

[0033] S37, performing double-point cross operation on any two paths in the path subset, and completing the cross operation only when the cross segment satisfies the structure legality constraint set

[0034] S38, performing structure legality checking on the sub-paths generated by the cross operation, performing repair processing on the sub-paths that violate the legality constraint, and only retaining the paths that satisfy the constraint after repair;

[0035] ​​​​​​​​​​S39, performing a mutation operation on the crossed path, selecting a behavior node in the path, and querying a replaceable behavior set with the same semantic label in the standard semantic template set and replacing it with a behavior node in the set at random;

[0036] S310, performing semantic consistency verification on the mutated path, confirming that all replaced behaviors meet the semantic classification requirements in the standard semantic template set, and retaining only the path that meets the consistency;

[0037] S311, merging the new path generated by the crossing and mutation operation with the path with the highest fitness ranking in the previous generation to construct a new path set for the next round of genetic evolution iteration;

[0038] S312, when the number of iterations reaches the preset maximum number of generations or the path set fitness changes meet the convergence condition, output the final path set as the medical emergency decision path set.

[0039] Optionally, the S4 specifically comprises:

[0040] S41, performing structural analysis on the medical emergency decision path set and the expert path, dividing the path nodes into receiving stage, judgment stage, execution stage and termination stage according to the stage attributes of the emergency response behaviors, and generating a multi-level path structure diagram;

[0041] S42, constructing a local factor graph based on the multi-level path structure diagram, the nodes in the local factor graph representing the path nodes, the edges representing the direct association relationship between the path nodes, and the edge weights being set according to the structural similarity, semantic similarity and causal embedding value;

[0042] S43, selecting a node from the expert path as a belief propagation source point, setting the initial belief value to 1, and initializing the belief value of the remaining path nodes to 0.5 to generate an initial belief propagation graph;

[0043] S44, using a causal embedding propagation mechanism to quantify the degree of causal dependence between path node pairs, constructing a causal weight matrix representing the causal influence strength between each pair of nodes;

[0044] S45, based on the constructed causal weight matrix, calculating the propagation edge weight for each pair of propagation nodes;

[0045] S46, based on the initial belief propagation graph and the local factor graph, performing layer-by-layer belief propagation operation according to the stage order to which the nodes belong, setting a stage progressive belief decay coefficient, setting an adjustable belief window for each propagation path, and updating the node belief value ;

[0046] S47, set the confidence value contribution coefficient, propagation weight coefficient and path similarity coefficient after each round of propagation is completed, construct a comprehensive evaluation function, and calculate the comprehensive decision evaluation value of the node v ;

[0047] S48, repeat the confidence propagation and pruning operation until the confidence value of all nodes changes less than the confidence deletion threshold or reaches the maximum propagation round;

[0048] S49, take the path structure and confidence value corresponding to the retained node as the trusted fusion basis, complete the confidence fusion between the medical emergency decision path set and the expert path, and generate an emergency decision path;

[0049] S410, deploy the emergency decision path to the medical emergency intelligent decision tree, and provide a trusted path basis for training and emergency drills in the medical emergency situation virtual environment.

[0050] Optionally, the S5 specifically comprises:

[0051] S51, organize medical personnel to enter the medical emergency situation virtual environment, and load the generated emergency decision path to guide the medical personnel to follow the path order to perform emergency response operations;

[0052] S52, real-time collection of key behavior data of medical personnel during execution, including decision response time, operation sequence accuracy, path node arrival condition and cooperation behavior record;

[0053] S53, construct a behavior performance evaluation index system, set key node evaluation dimensions, including execution deviation rate , path offset degree , behavior consistency degree ;

[0054] S54, calculate the comprehensive performance evaluation value of each node ;

[0055] S55, set a trusted node threshold , filter the evaluation value, if , mark the corresponding node from the path as a node that needs to be optimized, and as a path optimization candidate point, and construct a set of nodes that need to be optimized;

[0056] S56, based on the set of nodes that need to be optimized, combine the behavior deviation trajectory in the training process and the expert path to perform behavior backtracking and difference analysis, and form a node fine-tuning suggestion table;

[0057] ​S57, the training instructor combines the node fine-tuning suggestion table to perform manual fine-tuning operation, and manually corrects the nodes and connection relationships deviating from the emergency decision path;

[0058] S58, the path after manual fine-tuning is updated to form an adjusted emergency decision path, and the training performance record and adjustment operation log are kept.

[0059] Optionally, the S6 specifically comprises:

[0060] S61, receiving and analyzing the adjusted emergency decision path after manual fine-tuning and experience summary, and identifying the path node and path structure change information therein;

[0061] S62, mapping the adjusted emergency decision path into the original medical emergency intelligent decision tree structure, replacing the corresponding path branch, and completing the update and reconstruction of the emergency decision tree structure;

[0062] S63, consistency checking of the updated medical emergency intelligent decision tree, deploying the updated medical emergency intelligent decision tree into the medical emergency situation virtual environment, and initializing the key decision points and corresponding interactive actions;

[0063] S64, organizing medical personnel to enter the medical emergency situation virtual environment, and carrying out targeted emergency exercise operation according to the updated medical emergency intelligent decision tree guide path, collecting the interactive behavior data and path execution record of the medical personnel in the exercise process.

[0064] The beneficial effects of the present application are:

[0065] The present application introduces an intelligent decision tree and a situational virtual environment combined medical emergency training method, which makes up for the deficiencies of the prior art in emergency scene simulation, intelligent path generation and dynamic optimization. The expert path and structured historical cases are combined to improve the authority and practicality of the emergency exercise path. Through the integrated application of genetic algorithm and belief propagation algorithm, global optimization and credible fusion of multiple decision paths are realized, effectively overcoming the problems of individual differences difficult to reflect and imperfect feedback closed loop in traditional path training. The artificial fine-tuning and training performance feedback mechanism ensures the dynamic iteration of path optimization, which can continuously adapt to the actual ability changes of medical personnel and complex situation requirements. Relying on the whole process training and evaluation in the virtual simulation environment, medical personnel can repeatedly train and test in the highly restored emergency situation, not only improving the operation proficiency and decision reaction speed, but also greatly enhancing the team cooperation and system ability of emergency response. Overall, the present application can provide a scientific, efficient, personalized and sustainable improvement emergency training system for medical personnel, effectively promoting the intelligentization, professionalization and practicalization process of medical industry emergency training. BRIEF DESCRIPTION OF DRAWINGS

[0066] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and are intended to serve as an aid for explaining the application, and are intended to provide further understanding of the application, and are included as a part of this detailed description, and together with the detailed description explain the application, and do not limit the application. In the drawings:

[0067] Figure 1 A flow chart of the situational medical emergency training method based on intelligent decision tree proposed in the application;

[0068] Figure 2 A schematic diagram of the medical emergency decision path optimization process of the genetic algorithm of the situational medical emergency training method based on intelligent decision tree proposed in the application. DETAILED DESCRIPTION

[0069] The application will now be described in further detail with reference to the drawings. These drawings show only the essential features of the application and are therefore to be regarded only as a schematic illustration of the basic structure of the application. These drawings show only the essential features of the application and are therefore to be regarded only as a schematic illustration of the basic structure of the application.

[0070] Reference Figure 1 and Figure 2 The situational medical emergency training method based on intelligent decision tree comprises the following steps:

[0071] S1, collecting medical emergency historical case data, preprocessing the medical emergency historical case data, generating structured data, extracting paths verified by medical emergency historical cases and conforming to emergency operation specifications, and generating expert paths;

[0072] S2, based on the structured data, constructing a medical emergency situation virtual environment, setting multiple key decision points in the medical emergency situation virtual environment, and initializing a medical emergency intelligent decision tree;

[0073] S3, based on the medical emergency intelligent decision tree, using a genetic algorithm to perform evolutionary operations on multiple decision paths therein, and generating a medical emergency decision path set;

[0074] S4, using a belief propagation algorithm to transfer the confidence of the node information between the medical emergency decision path set and the expert path, realizing the credible fusion of the decision nodes through the information interaction between the paths, and generating an emergency decision path;

[0075] S5, organizing medical personnel to perform emergency training operations in the medical emergency situation virtual environment according to the emergency decision path, recording key behavior data and training performance, comprehensively evaluating the execution effect of the key nodes of the emergency decision path, and according to the evaluation results, performing artificial fine-tuning and experience summary, and forming an adjusted emergency decision path;

[0076] S6, redeploy the adjusted emergency decision path to the medical emergency intelligent decision tree, and guide medical personnel to carry out targeted training in the medical emergency situation virtual environment.

[0077] The application provides a situational medical emergency training method based on an intelligent decision tree, which improves the emergency disposal ability and path execution efficiency of medical personnel by constructing a structured emergency knowledge path, optimizing a decision process and fusing training feedback information, and effectively enhances the reality and pertinence of training by extracting expert paths and constructing a situational virtual environment based on real medical emergency cases; the genetic algorithm is introduced to evolve and optimize multiple decision paths, thereby improving the diversity and adaptability of the path scheme; the belief propagation algorithm is used to realize the reliable fusion of the decision path and the expert experience, thereby guaranteeing the accuracy of the path logic and the scientificity of the operation, and the training process data of the medical personnel is combined to fine-tune and optimize the feedback of the emergency path, thereby forming a cyclic iteration path correction mechanism. Compared with the traditional emergency training method, the application not only improves the response speed, operation accuracy and team cooperation level, but also dynamically adapts to different individual and scene requirements, has high flexibility and promotion value, and is suitable for hospital first-aid training, disaster emergency drilling and other medical emergency fields.

[0078] In the embodiment, the medical emergency historical case data specifically includes an emergency event type, spatial arrangement information and a task element.

[0079] In the embodiment, the preprocessing of the medical emergency historical case data specifically includes data cleaning, denoising, missing completion, format standardization and key field extraction.

[0080] In the embodiment, S2 specifically includes:

[0081] S21, collecting an emergency event type, spatial arrangement information and a task element in the structured data, and constructing a scene factor matrix , wherein, represents the semantic matching degree between the i-th event and the j-th environmental factor;

[0082] S22, performing semantic scene mapping according to the scene factor matrix , and calling a three-dimensional modeling engine to construct a medical emergency situation virtual environment with spatial layout, task flow and interactive element configuration, wherein the medical emergency situation virtual environment includes an operation area, a path node, a device distribution and a personnel station;

[0083] S23, introducing an event label-semantic dictionary mapping mechanism to normalize the concept of the label in the structured data, and forming a standard semantic template set ; ​​

[0084] S24, injecting medical staff role entities from a preset role library according to the scene task type, and performing medical staff role binding operations, mapping each type of medical staff role to executable operation nodes, the operation nodes being matched and selected according to the standard semantic template set

[0085] S25, identifying key operation nodes in the medical emergency scenario virtual environment, and constructing a set of key decision points , and specifying a corresponding input state set and a behavior set for each decision point ;

[0086] S26, automatically generating a medical emergency intelligent decision tree structure based on the configured nodes and behaviors in the scene , wherein , represents a node state transition path triggered by the behavior set , wherein represents a set of nodes in the medical emergency intelligent decision tree

[0087] S27, embedding the constructed medical emergency intelligent decision tree into the medical emergency scenario virtual environment.

[0088] This step realizes the high simulation, controllable structure and strong task adaptability of the medical emergency training environment through the fine construction of the medical emergency scenario virtual environment and the deep integration of the intelligent decision structure. By collecting the event type, spatial layout and task element in the structured data, constructing a scene factor matrix and completing semantic scene mapping, the virtual environment ensures the semantic correspondence and spatial restoration degree of the real scene. Combined with the three-dimensional modeling engine, the virtual environment is generated with integrated operation area, path node and device configuration, providing a complete interaction and execution field for medical staff. The introduction of the standard semantic template set and the role binding mechanism enables the medical staff behavior to be accurately mapped to the operation node, enhances the accuracy of path configuration and the directionality of task execution, automatically constructs the key decision points and input behavior set, realizes the dynamic definition of decision conditions and the accurate generation of execution path. Finally, the medical emergency intelligent decision tree is embedded in the virtual scene, enabling the training system to have dynamic guidance and path exercise ability, effectively improving the intelligent level of virtual environment construction and the structural integrity of decision logic, and providing high-quality basic support for emergency response path optimization and training feedback.

[0089] In this embodiment, the S3 specifically includes:

[0090] S31, extracting a set of paths from the medical emergency intelligent decision tree as a set of paths to be optimized;​

[0091] S32, constructing an initial path set wherein, is an expert path, is a path randomly generated based on structured data, and the initial path set is used as the initial population of the genetic algorithm;

[0092] S33, for each path in the initial path set is represented as a sequence of state-action pairs and is converted into a chromosome representation :

[0093] ;

[0094] wherein, represents the lth decision node in the ith path, represents the action decision taken at the decision node ;

[0095] The chromosome representation Each emergency path is represented as a series of corresponding relationships between states and actions, i.e., each path is composed of a number of "decision node-action decision" pairs, so that the path can be converted into a standardized structure for easy execution of the genetic algorithm optimization process. Each "state-action" pair actually corresponds to a specific decision action made by medical personnel in a certain situation, such as choosing "immediate defibrillation" or "detecting pulse" at a certain first aid node. Combining all the node-action pairs in the path together forms a complete path chromosome, i.e., an individual in the genetic algorithm, which can accurately record the execution process of each path, so that the path can be clearly identified when performing genetic operations such as crossover and mutation, which nodes and behavior combinations can be replaced or recombined, and also provides a basis for fitness evaluation, so that the system can combine the behavior accuracy of each node, time performance data, to make path advantage and disadvantage judgment and evolution selection.

[0096] S34, for each path calculate the fitness value :

[0097] ;

[0098] wherein, is the total execution time of the path, is the number of errors of the path action node, represents the structural similarity of the path to the expert path, , and are weighting coefficients;

[0099] The fitness value This method is used to evaluate the quality of each medical emergency response path. Its core purpose is to measure the overall performance of the path by constructing a fitness function, thus providing a quantitative basis for selection, crossover, and mutation in genetic algorithms. The quality of a path is mainly determined by three factors: execution efficiency, behavioral accuracy, and structural rationality. The first factor is the total execution time of the path. The shorter the time, the more efficient the path is in actual emergency response. Therefore, the reciprocal of the time is used to measure efficiency; higher efficiency means higher fitness. The second factor is the number of errors at behavioral nodes in the path. The reciprocal of the number of errors reflects accuracy; fewer errors mean a more reliable path in actual operation. The third factor is the structural similarity between the path and expert paths, reflecting whether the path follows a verified and effective emergency operation procedure. Higher structural similarity indicates that the path is closer to the ideal execution standard. By setting weighting coefficients for these three indicators, the weight ratio of each factor can be dynamically adjusted according to actual application needs. For example, when rapid response is emphasized in emergency handling, the weight of execution efficiency can be increased; in high-risk scenarios, accuracy and path standardization can be emphasized.

[0100] S35. Based on the fitness function value, adjust the initial path set. The paths in the array are sorted, and a roulette wheel selection method is used to generate a subset of paths for the cross operation.

[0101] S36. Define the set of structural legality constraints. The set of structural legality constraints consists of consecutive pairs of legal behaviors extracted from structured data;

[0102] S37. For any two paths in the path subset Perform a two-point crossover operation only if the crossover segment satisfies the set of structural legality constraints. Under the premise of completing the cross operation;

[0103] S38. Perform structural legality verification on the sub-paths generated by the intersection. Perform repair processing on sub-paths that violate legality constraints and retain only the paths that satisfy the constraints after repair.

[0104] S39. Perform a mutation operation on the crossed paths, select a node representing a behavior in the path, and apply it to the standard semantic template set. The query finds a set of alternative behaviors with the same semantic label and randomly replaces it with one of the behavior nodes.

[0105] S310. Perform semantic consistency verification on the mutated path to confirm that all alternative behaviors meet the semantic classification requirements of the standard semantic template set, and retain only the paths that meet the consistency requirements.

[0106] S311. Merge the new paths generated by crossover and mutation operations with the paths with the highest fitness ranking in the previous generation to construct a new generation of path sets. And then proceed to the next round of genetic evolution iteration;

[0107] S312. When the number of iterations reaches the preset maximum algebra or the fitness change of the path set satisfies the convergence condition, output the final path set. This serves as a set of emergency decision-making pathways for medical staff.

[0108] This step introduces a structured, improved genetic algorithm to systematically optimize multiple paths in the medical emergency intelligent decision tree, enhancing the execution efficiency, logical rationality, and individual adaptability of emergency decision paths. A diverse initial path population is constructed by combining expert paths and structured data, ensuring sufficient initial diversity in the path optimization process. Through state-behavior pair chromosome encoding, the path structure is mapped to standardized evolutionary units, laying the foundation for selection, crossover, and mutation operations. A multi-objective fitness design integrating time cost, error rate, and structural similarity effectively guides genetic operations towards aligning execution efficiency with expert experience. The introduction of structural legality constraints and semantic consistency verification mechanisms ensures that crossover and mutation operations maintain path executability and semantic logical rationality while guaranteeing innovation. A set of alternative behaviors is constructed by integrating a standard semantic template set, achieving semantic-driven mutation and expanding behavioral diversity. Finally, through iterative updates and convergence judgment mechanisms, a set of medical emergency decision paths with reasonable structure, optimized strategies, and semantic consistency is output, providing a high-quality path foundation for reliable integration and drill deployment. This effectively enhances the scientific rigor, relevance, and practicality of path generation.

[0109] In this embodiment, S4 specifically includes:

[0110] S41. Perform structural analysis on the medical emergency decision-making path set and expert path, and divide the path nodes into the receiving stage, judgment stage, execution stage and termination stage according to the stage attributes of emergency response behavior, and generate a multi-level path structure diagram.

[0111] S42. Based on the multi-level path structure graph, construct a local factor graph. The nodes in the local factor graph represent path nodes, and the edges represent the direct relationships between path nodes. Set the edge weights according to structural similarity, semantic similarity and causal embedding value.

[0112] S43. Select nodes from the expert path as confidence propagation source points, set the initial confidence value to 1, initialize the confidence value of the remaining path nodes to 0.5, and generate the initial confidence propagation graph;

[0113] S44. Using a causal embedding propagation mechanism, the degree of causal dependence between path node pairs is quantified, and a causal weight matrix is ​​constructed to represent the strength of causal influence between each pair of nodes.

[0114] S45. Based on the constructed causal weight matrix, calculate the propagation edge weight for each pair of propagation nodes. :

[0115] ;

[0116] in, , and These are weighting coefficients. This represents the semantic similarity index between node u and node v. Structural similarity index between node u and node v These are the causal influence values ​​corresponding to nodes u and v in the causal weight matrix;

[0117] propagation of border rights The goal is to assign reasonable weights to the information dissemination process between different nodes in a medical emergency pathway, thereby improving the accuracy and reliability of confidence propagation. The calculation of propagation edge weights comprehensively considers three key factors: semantic similarity, structural similarity, and the degree of causal influence. Semantic similarity assesses the relevance between two nodes in terms of emergency operational significance, such as whether two nodes express similar medical intentions or execute instructions; its level directly affects the semantic coherence of information dissemination. Structural similarity reflects the similarity of the positions of two nodes in the pathway structure; if the layout and connection methods of two nodes in the pathway are similar, they can be considered to have high structural consistency, which helps enhance pathway stability. The causal influence value reflects the strength of the causal relationship between two nodes in the emergency process, such as whether one operation is a direct prerequisite or result of another operation. By assigning weights to these three indicators and summing them, a propagation edge weight value that comprehensively evaluates the strength of node relationships is constructed. The larger the value, the more credible and important the information dissemination between the two nodes. In the confidence propagation process, the propagation edge weight is used to guide the priority of information diffusion between nodes, thereby achieving more accurate pathway confidence fusion.

[0118] S46. Based on the initial confidence propagation graph and the local factor graph, perform a layer-by-layer confidence propagation operation according to the stage order of the nodes, set a stage-by-stage progressive confidence decay coefficient, set an adjustable confidence window for each propagation path, and set the node confidence value. Update:

[0119] ;

[0120] in, Indicates from node arrive The confidence decay coefficient across the stages, As the normalization factor, For nodes In the The confidence value of the wheel, Let v be the confidence value of node v after the (t+1)th confidence propagation iteration. Let v be the set of adjacent nodes;

[0121] Confidence value of nodes To ensure the rationality and credibility of information dissemination in phased tasks, an updated confidence propagation mechanism is constructed. This mechanism introduces phase decay factors, propagation weights, and adjacency relationships to iteratively update the confidence of each node in different propagation rounds. The node's confidence value is jointly influenced by the confidence of all neighboring nodes, the weights of connecting edges, and the degree of confidence decay between phases. This ensures that information is transmitted along the path while considering contextual relevance and avoiding over-diffusion or trust redundancy. The confidence decay coefficient reflects the natural decay of transmission reliability between different phases; that is, as a node traverses more phases, the reliability of the received information should gradually decrease. The propagation edge weights combine semantic, structural, and causal factors to reflect the propagation intensity between nodes in the knowledge network. The set of neighboring nodes represents the range of sources from which the current node receives information, ensuring comprehensive coverage and logical coherence in the propagation process.

[0122] S47. After each round of propagation, set the confidence contribution coefficient, propagation weight coefficient, and path similarity coefficient, construct a comprehensive evaluation function, and calculate the comprehensive decision evaluation value of node v. :

[0123] ;

[0124] in, , and These are the normalized weights of the three evaluation factors. Indicates the current path With expert path semantic similarity, if Then delete the node. Indicates the confidence level deletion threshold;

[0125] The comprehensive decision evaluation value of computing node v The system comprehensively evaluates the decision credibility of each node in the path to determine whether it should be retained or deleted. The evaluation process integrates the node's own confidence value, the trust propagation strength of neighboring nodes, and the semantic similarity between the overall path and the expert path, constructing a multi-factor weighted comprehensive decision evaluation function. This effectively reflects the value and credibility level of a node in the current path. The node's own confidence value reflects the credibility of the information obtained during the propagation iteration process. The weighted influence of confidence propagation from neighboring nodes considers the contextual relationships in the path structure, ensuring structural consistency in node evaluation. By introducing semantic similarity between the path and the expert path, the system ensures consistency between the evaluation results and professional standards. By setting thresholds, the system can automatically remove nodes with low credibility or no reference value, thereby optimizing the path structure.

[0126] S48. Repeat the confidence propagation and pruning operations until the change in the confidence value of all nodes is less than the confidence deletion threshold or the maximum propagation round is reached.

[0127] S49. Using the path structure and confidence value corresponding to the retained nodes as the basis for reliable fusion, complete the confidence fusion between the medical emergency decision-making path set and the expert path to generate an emergency decision-making path.

[0128] S410. Deploy emergency decision-making paths into the medical emergency intelligent decision tree to provide a reliable path basis for training and emergency drills in the virtual environment of medical emergency scenarios.

[0129] This step achieves a reliable fusion of the medical emergency decision-making path set and expert paths through an improved confidence propagation algorithm, enhancing the scientific nature of the path structure and the credibility of the fusion results. By dividing path nodes into four stages—receiving, judging, executing, and terminating—a multi-level path structure graph is constructed, enabling a systematic analysis of decision-making behavior at the structural level. Edge weight indicators composed of structural similarity, semantic similarity, and causal embedding are introduced to characterize the multi-dimensional correlations between path nodes in the local factor graph, providing an information foundation for propagation. By setting expert path nodes as initial confidence sources, a phased, layer-by-layer confidence propagation operation is executed. While considering causal dependencies, a confidence decay mechanism and a confidence window strategy are introduced to achieve orderly propagation of confidence information among nodes at different levels. An evaluation function constructed by comprehensively considering propagation weights, path similarity, and confidence contribution dynamically assesses the importance of each node. Pruning operations are introduced during propagation to remove nodes with low confidence values, thereby simplifying and strengthening the path structure. Finally, the confidence-fused path results are deployed into the medical emergency intelligent decision tree to form a compact, reasonable, and reliable emergency decision path, providing high-quality decision-making basis for medical training and significantly improving the practicality and accuracy of path execution.

[0130] In this embodiment, S5 specifically includes:

[0131] S51. Organize medical staff to enter the virtual environment of medical emergency situation, load the generated emergency decision path, and guide medical staff to perform emergency response operations in the order of the path;

[0132] S52. Collect key behavioral data of medical staff in real time during the execution process. Key behavioral data includes decision response time, accuracy of operation sequence, arrival status of path nodes, and records of collaborative behavior.

[0133] S53. Construct a behavioral performance evaluation index system and set key evaluation dimensions, including execution deviation rate. Path offset Behavioral consistency ;

[0134] S54, For each node Overall performance evaluation value Perform the calculation:

[0135] ;

[0136] in, , , These are the indicator weight coefficients;

[0137] Overall performance evaluation value The approach involves a multi-indicator fusion method to comprehensively evaluate the performance of each node in the path, thereby quantifying the quality and contribution of each node in the emergency execution process. It comprehensively considers the influencing factors of multiple key performance indicators and uses a weighted combination to form a quantitative score for the final performance of each node. The deviation rate, response time deviation, and behavioral criticality of each node are used as core evaluation indicators. Specifically, the deviation rate reflects whether the node's behavior deviates from the expected path; a high value indicates an anomaly in the node's execution. Response time deviation measures whether medical personnel respond quickly and promptly at that node; a slower response indicates poor training effectiveness. Behavioral criticality assesses the importance of the task undertaken by the node in the overall path; a high weight indicates that the node is crucial in the emergency process. By reasonably weighting and fusing these three dimensions, the node performance is comprehensively and accurately evaluated, ensuring that the evaluation results cover both behavioral accuracy and timeliness and task importance.

[0138] S55, Set trusted node threshold The evaluation values ​​are then filtered, if... Then the corresponding node Mark the nodes in the path that need optimization and use them as candidate points for path optimization to construct a set of nodes that need optimization;

[0139] S56. Based on the set of nodes that need to be tuned, and combined with the behavioral offset trajectory and expert path during the training process, perform behavioral backtracking and difference analysis to form a node fine-tuning suggestion table.

[0140] S57. The training instructors, in conjunction with the node fine-tuning suggestion table, shall perform manual fine-tuning operations to manually correct the nodes and connections that deviate in the emergency decision-making path.

[0141] S58. Update the structure of the path after manual fine-tuning to form an adjusted emergency decision-making path, and retain training performance records and adjustment operation logs.

[0142] This invention enhances the individual adaptability and training relevance of emergency medical training pathways through a pathway fine-tuning mechanism driven by key behavior collection and evaluation. By organizing medical personnel to execute generated emergency decision-making pathways in a virtual emergency environment, and collecting key behavioral data in real time, such as decision response time, accuracy of operation sequence, arrival status of pathway nodes, and collaborative behavior, it provides rich first-hand information support for pathway quality assessment. An evaluation index system covering execution deviation rate, pathway deviation, and behavioral consistency is constructed, enabling quantitative analysis of the performance of each pathway node. By setting a credible node threshold to filter evaluation results, nodes with weak performance in actual execution are identified, and a set of nodes requiring optimization is constructed. Furthermore, by combining behavioral deviation trajectories with expert pathways, behavioral retrospectives and difference analysis are conducted on deviation points, providing precise node fine-tuning suggestions. This process is completed manually by training instructors to ensure the practicality and professionalism of the fine-tuning operations. Finally, the revised pathway structure is updated, and all process data is recorded, forming a continuously optimizing training feedback loop.

[0143] In this embodiment, S6 specifically includes:

[0144] S61. Receive and parse the adjusted emergency decision-making path after manual fine-tuning and experience summary, and identify the path nodes and path structure change information.

[0145] S62. Map the adjusted emergency decision-making path to the original medical emergency intelligent decision-making tree structure, replace the corresponding path branches, and complete the update and reconstruction of the emergency decision-making tree structure.

[0146] S63. Perform consistency verification on the updated medical emergency intelligent decision tree, deploy the updated medical emergency intelligent decision tree to the medical emergency scenario virtual environment, and initialize key decision points and corresponding interactive actions.

[0147] S64. Organize medical staff to enter the virtual environment of medical emergency scenario, and conduct targeted emergency drills according to the updated medical emergency intelligent decision tree guidance path. During the drills, collect the interactive behavior data and path execution records of medical staff.

[0148] This invention focuses on the redeployment and training closed-loop construction of adjusted emergency decision-making paths, enhancing the iterative update capability and environmental adaptability of medical emergency training paths. By analyzing emergency decision-making paths refined through manual adjustments and experience summaries, the system accurately identifies path nodes and structural changes, ensuring structural mappability and update value. The adjusted path content is then replaced with the corresponding path branches in the original medical emergency intelligent decision tree, completing the structural update and reconstruction of the emergency decision tree. Furthermore, consistency checks are performed to ensure the updated decision tree is logically complete and conflict-free, thereby guaranteeing the stability and continuity of the training process. After the update, the medical emergency intelligent decision tree is embedded into a virtual environment of medical emergency scenarios, and key decision points and interactive actions are reinitialized, providing accurate paths and operational guidance for subsequent drills by medical personnel. By organizing medical personnel to re-enter the virtual environment to execute new emergency path drills, the system can collect interactive behavior data and path execution records generated during the drills in real time, continuously supplementing the training data pool and further enhancing the verification effect of the optimized paths.

[0149] Example 1:

[0150] To verify the feasibility of this invention in practice, it was applied to a tertiary hospital. The emergency department and ICU teams jointly applied the contextualized medical emergency training system based on intelligent decision trees proposed in this invention to specifically improve the team's emergency response capabilities in cardiac arrest scenarios. The hospital collected and organized data on key aspects of each emergency response from 112 real-life cardiac arrest cases both inside and outside the hospital over the past three years, including call response, ECG monitoring, cardiopulmonary resuscitation, medication administration, defibrillation procedures, and team communication, constructing a structured medical emergency case database. Based on this data, the system automatically analyzes and extracts expert path nodes, and initializes the intelligent decision tree by combining standardized procedures with case experience, providing decision support for virtual simulation exercises.

[0151] In the application scenario, the hospital selected 20 medical staff with more than 3 years of work experience and divided them into an experimental group (using the system of this invention, 10 people) and a control group (using traditional paper-based processes and a common virtual simulation platform, 10 people). All personnel participated in the same four cardiac arrest scenario simulation assessments. The experimental group members repeatedly practiced in the intelligent decision tree system, with the system automatically guiding them to operate according to the optimal path and recording operation responses, node execution, and team collaboration data in real time. Based on the performance of each training session, the system continuously optimized the decision path using belief propagation and behavioral fine-tuning mechanisms, deploying the new path back into the decision tree to ensure that each round of exercises reflects the latest capabilities and path adjustments of the medical staff.

[0152] After four rounds of systematic training, the experimental group showed significant improvements in key indicators such as average defibrillation response time, accuracy of key procedures, path omission rate, and teamwork score. Data showed that the experimental group's defibrillation response time decreased from 31.2 seconds initially to 19.6 seconds, key procedure accuracy increased from 87.1% to 97.8%, path execution omission rate decreased from 4.2% to 0.7%, and teamwork score improved from 82 to 94. While the control group also showed improvement, the magnitude and absolute value of the improvement were significantly lower than those of the experimental group. Particularly in the simulated emergency rescue assessment, the experimental group achieved a 100% pass rate, compared to only 80% in the control group. A questionnaire survey revealed that 90% of the medical staff in the experimental group indicated that the systematic training method was more targeted, directly reflecting their weaknesses and quickly forming muscle memory for emergency responses.

[0153] This invention can efficiently integrate historical cases and expert experience, and through the fusion of intelligent decision trees and virtual environments, it can achieve personalized optimization and dynamic drills of decision-making paths. This improves the reaction speed, operational accuracy and teamwork level of medical teams in real emergency scenarios, and overcomes the shortcomings of traditional emergency training paths, such as single path, delayed feedback and slow iteration. It provides a new scientific, intelligent and sustainable optimization model for hospital medical training.

[0154] Table 1. Comparison of the main effects of the intelligent decision tree-based medical emergency training system and traditional methods.

[0155] Indicator Initial experiment group 4th round of experiment group Initial control group 4th round of control group Defibrillation operation response time (seconds) 31.2 19.6 29.5 24.8 Node operation accuracy rate 87.1% 97.8% 85.9% 91.2% Path execution omission rate 4.2% 0.7% 3.8% 2.6% Team cooperation score (100 full marks) 82 94 80 87 Pass rate of simulated first aid examination 70% 100% 65% 80%

[0156] According to the data in Table 1, it can be seen that the medical emergency training system based on intelligent decision tree proposed in this invention has significant advantages over traditional training methods in all key training indicators.

[0157] Regarding defibrillation response time, the experimental group had an average response time of 31.2 seconds during the first training session. After four rounds of continuous optimization training using the intelligent decision tree system, this time was significantly reduced to 19.6 seconds, a reduction of 37.2% overall. In contrast, the control group had an average initial response time of 29.5 seconds, which was only reduced to 24.8 seconds in the fourth round, a relatively limited reduction. This demonstrates the significant dynamic optimization effect of the intelligent training path.

[0158] In terms of node operation accuracy, after systematic training, the experimental group's accuracy significantly improved from the initial 87.1% to 97.8%, an increase of 10.7 percentage points. In contrast, the control group using traditional methods only saw an improvement in accuracy from the initial 85.9% to 91.2%, indicating less room for further improvement. This demonstrates that the intelligent decision tree system can more effectively help medical staff accurately master key operational steps.

[0159] Regarding the path execution omission rate, the experimental group also performed exceptionally well, significantly reducing the omission rate from 4.2% initially to 0.7% in the fourth round, effectively resolving the issue of missed items. While the control group also showed some improvement, it still had an omission rate of 2.6%, far less ideal than the experimental group. This demonstrates that intelligent decision trees have a greater advantage in path optimization and training feedback, better ensuring the integrity and standardization of emergency procedures.

[0160] Team collaboration scoring also validated this trend, with the experimental group's score significantly improving from an initial 82 to 94, significantly higher than the control group's 87. This indicates that contextualized intelligent decision tree training led to better teamwork and significantly enhanced emergency collaboration capabilities.

[0161] In the key comprehensive indicator of the simulated emergency rescue assessment pass rate, the experimental group achieved a 100% pass rate after training, while the pass rate of the control group trained in the traditional way only increased to 80%. This significant difference highlights the effectiveness and practical value of the intelligent decision tree training method in real-world scenarios.

[0162] In summary, the intelligent medical emergency training method proposed in this invention significantly improves the emergency response speed, operational accuracy, and teamwork level of medical personnel through dynamic path optimization, real-time feedback, and intelligent decision integration. The overall effect is significantly better than traditional training methods, and it has broad application prospects and practical value.

[0163] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A contextualized medical emergency training method based on intelligent decision trees, characterized in that, Includes the following steps: S1. Collect historical medical emergency case data, preprocess the historical medical emergency case data to generate structured data, extract paths that have been verified by historical medical emergency cases and comply with emergency operation specifications, and generate expert paths. S2. Based on structured data, construct a virtual environment for medical emergency scenarios, set multiple key decision points in the virtual environment for medical emergency scenarios, and initialize the medical emergency intelligent decision tree; S3. Based on the medical emergency intelligent decision tree, a genetic algorithm is used to perform evolutionary operations on multiple decision paths to generate a set of medical emergency decision paths; S4. The confidence propagation algorithm is used to transmit the confidence of node information between the medical emergency decision-making path set and the expert path. The reliable fusion of decision nodes is achieved through information interaction between paths to generate emergency decision-making paths. S5. Organize medical staff to conduct emergency training operations in a virtual environment of medical emergency scenarios, according to the emergency decision-making path, record key behavioral data and training performance, conduct a comprehensive evaluation of the execution effect of key nodes of the emergency decision-making path, and make manual fine-tuning and experience summary based on the evaluation results to form an adjusted emergency decision-making path. S6. Redeploy the adjusted emergency decision-making path to the medical emergency intelligent decision tree to guide medical staff to conduct targeted drills in a virtual environment of medical emergency scenarios.

2. The contextualized medical emergency training method based on intelligent decision tree according to claim 1, characterized in that, The data on historical medical emergency cases specifically includes emergency event types, spatial layout information, and task elements.

3. The contextualized medical emergency training method based on intelligent decision trees according to claim 1, characterized in that, The preprocessing of historical medical emergency case data specifically includes data cleaning, noise reduction, missing data completion, format standardization, and key field extraction.

4. The contextualized medical emergency training method based on intelligent decision tree according to claim 1, characterized in that, S2 specifically includes: S21. Collect emergency event types, spatial layout information, and task elements from structured data to construct a scene factor matrix. ,in, Indicates the first Class events and the first Semantic matching degree between environmental factors; S22, Based on the scene factor matrix Semantic scene mapping is performed, and a 3D modeling engine is called to construct a virtual environment for medical emergency scenarios with spatial layout, task flow and interactive element configuration. The virtual environment for medical emergency scenarios includes operation area, path nodes, equipment distribution and personnel positions. S23. Introduce an event tag-semantic dictionary mapping mechanism to unify the concepts of tags in structured data and form a standard semantic template set. ; S24. Inject medical staff role entities from the preset role library according to the scenario task type, and perform medical staff role binding operation to map each type of medical staff role to an executable operation node. The operation node is based on the standard semantic template set. Match and select the corresponding template items; S25. Identify key operational nodes in a virtual environment of medical emergency scenarios and construct a set of key decision points. And at each decision point Specify the corresponding set of input states With behavior set ; S26. Based on the nodes and behaviors configured in the scenario, automatically generate a medical emergency intelligent decision tree structure. ,in , Represents a set of behaviors The resulting node state transition path, where, This represents the set of nodes in the intelligent decision tree for medical emergency response; S27. Embed the completed intelligent decision tree for medical emergency into the virtual environment of medical emergency scenarios.

5. The contextualized medical emergency training method based on intelligent decision tree according to claim 1, characterized in that, S3 specifically includes: S31, From the Intelligent Decision Tree for Medical Emergency Response Extract path set As a set of paths to be optimized; S32. Construct the initial path set ,in, For expert path, The initial set of paths, randomly generated based on structured data, serves as the initial population for the genetic algorithm. S33. For each path in the initial path set... Represented as state-behavior pairs, and then transformed into chromosome representation. : ； in, This represents the l-th decision node in the i-th path. Indicates at the decision node The behavioral decisions made; S34, for each path Calculate fitness value ; S35. Based on the fitness function value, adjust the initial path set. The paths in the array are sorted, and a roulette wheel selection method is used to generate a subset of paths for the cross operation. S36. Define the set of structural legality constraints. The set of structural legality constraints consists of consecutive pairs of legal behaviors extracted from structured data; S37. For any two paths in the path subset Perform a two-point crossover operation only if the crossover segment satisfies the set of structural legality constraints. Under the premise of completing the cross operation; S38. Perform structural legality verification on the sub-paths generated by the intersection. Perform repair processing on sub-paths that violate legality constraints and retain only the paths that satisfy the constraints after repair. S39. Perform a mutation operation on the crossed paths, select a node representing a behavior in the path, and apply it to the standard semantic template set. The query finds a set of alternative behaviors with the same semantic label and randomly replaces it with one of the behavior nodes. S310. Perform semantic consistency verification on the mutated path to confirm that all alternative behaviors meet the semantic classification requirements of the standard semantic template set, and retain only the paths that meet the consistency requirements. S311. Merge the new paths generated by crossover and mutation operations with the paths with the highest fitness ranking in the previous generation to construct a new generation of path sets. And then proceed to the next round of genetic evolution iteration; S312. When the number of iterations reaches the preset maximum algebra or the fitness change of the path set satisfies the convergence condition, output the final path set. This serves as a set of emergency decision-making pathways for medical staff.

6. The contextualized medical emergency training method based on intelligent decision trees according to claim 1, characterized in that, S4 specifically includes: S41. Perform structural analysis on the medical emergency decision-making path set and expert path, and divide the path nodes into the receiving stage, judgment stage, execution stage and termination stage according to the stage attributes of emergency response behavior, and generate a multi-level path structure diagram. S42. Based on the multi-level path structure graph, construct a local factor graph. The nodes in the local factor graph represent path nodes, and the edges represent the direct relationships between path nodes. Set the edge weights according to structural similarity, semantic similarity and causal embedding value. S43. Select nodes from the expert path as confidence propagation source points, set the initial confidence value to 1, initialize the confidence value of the remaining path nodes to 0.5, and generate the initial confidence propagation graph; S44. Using a causal embedding propagation mechanism, the degree of causal dependence between path node pairs is quantified, and a causal weight matrix is ​​constructed to represent the strength of causal influence between each pair of nodes. S45. Based on the constructed causal weight matrix, calculate the propagation edge weight for each pair of propagation nodes. ; S46. Based on the initial confidence propagation graph and the local factor graph, perform a layer-by-layer confidence propagation operation according to the stage order of the nodes, set a stage-by-stage progressive confidence decay coefficient, set an adjustable confidence window for each propagation path, and set the node confidence value. Update; S47. After each round of propagation, set the confidence contribution coefficient, propagation weight coefficient, and path similarity coefficient, construct a comprehensive evaluation function, and calculate the comprehensive decision evaluation value of node v. ; S48. Repeat the confidence propagation and pruning operations until the change in the confidence value of all nodes is less than the confidence deletion threshold or the maximum propagation round is reached. S49. Using the path structure and confidence value corresponding to the retained nodes as the basis for reliable fusion, complete the confidence fusion between the medical emergency decision-making path set and the expert path to generate an emergency decision-making path. S410. Deploy emergency decision-making paths into the medical emergency intelligent decision tree to provide a reliable path basis for training and emergency drills in the virtual environment of medical emergency scenarios.

7. The contextualized medical emergency training method based on intelligent decision tree according to claim 1, characterized in that, S5 specifically includes: S51. Organize medical staff to enter the virtual environment of medical emergency situation, load the generated emergency decision path, and guide medical staff to perform emergency response operations in the order of the path; S52. Collect key behavioral data of medical staff in real time during the execution process. Key behavioral data includes decision response time, accuracy of operation sequence, arrival status of path nodes, and records of collaborative behavior. S53. Construct a behavioral performance evaluation index system and set key evaluation dimensions, including execution deviation rate. Path offset Behavioral consistency ; S54, For each node Overall performance evaluation value Perform calculations; S55, Set trusted node threshold The evaluation values ​​are then filtered, if... Then the corresponding node Mark the nodes in the path that need optimization and use them as candidate points for path optimization to construct a set of nodes that need optimization; S56. Based on the set of nodes that need to be tuned, and combined with the behavioral offset trajectory and expert path during the training process, perform behavioral backtracking and difference analysis to form a node fine-tuning suggestion table. S57. The training instructors, in conjunction with the node fine-tuning suggestion table, shall perform manual fine-tuning operations to manually correct the nodes and connections that deviate in the emergency decision-making path. S58. Update the structure of the path after manual fine-tuning to form an adjusted emergency decision-making path, and retain training performance records and adjustment operation logs.

8. The contextualized medical emergency training method based on intelligent decision tree according to claim 1, characterized in that, S6 specifically includes: S61. Receive and parse the adjusted emergency decision-making path after manual fine-tuning and experience summary, and identify the path nodes and path structure change information. S62. Map the adjusted emergency decision-making path to the original medical emergency intelligent decision-making tree structure, replace the corresponding path branches, and complete the update and reconstruction of the emergency decision-making tree structure. S63. Perform consistency verification on the updated medical emergency intelligent decision tree, deploy the updated medical emergency intelligent decision tree to the medical emergency scenario virtual environment, and initialize key decision points and corresponding interactive actions. S64. Organize medical staff to enter the virtual environment of medical emergency scenario, and conduct targeted emergency drills according to the updated medical emergency intelligent decision tree guidance path. During the drills, collect the interactive behavior data and path execution records of medical staff.

Images