Emergency team cooperation dynamic resource scheduling method and system based on multi-modal data driving

Abstract

The invention relates to the field of multi-modal data analysis, in particular to an emergency team cooperation dynamic resource scheduling method and system based on multi-modal data driving, and the method comprises the steps: collecting patient physiological time sequence data, team behavior interaction data and resource state data in an emergency scene, and forming a heterogeneous data stream; according to the heterogeneous data flow, a topological dependency relationship of a team cooperation link is deconstructed; calculating a collaborative robustness index based on the topological dependency relationship and the heterogeneous data flow; when the collaborative robustness index is lower than a preset index threshold value, executing a conflict resolution decision of multi-modal constraint driving; and monitoring cooperation efficiency change data after resource allocation, and correcting the weight coefficient of the cooperation robustness index and a preset index threshold in combination with a preset optimization target. Key node overload risk probability calculation and cooperative path resource assurance degree evaluation are fused, quantitative system anti-risk capability evaluation is provided, and compared with traditional single threshold alarm, composite collapse risks are recognized earlier.

Description

Technical Field

[0001] This application relates to the field of multimodal data analysis technology, and in particular to a method and system for dynamic resource scheduling of emergency medical teams based on multimodal data-driven collaboration. Background Technology

[0002] In emergency situations, dynamic resource scheduling refers to a technology that adjusts the allocation of medical resources (personnel / equipment / medications) based on real-time data to optimize team collaboration efficiency. Its core objective is to ensure that critical resources are accurately matched with patient needs and team capabilities under time-sensitive, resource-limited, and highly uncertain conditions.

[0003] Current emergency medical services (EMS) resource dispatching technologies typically rely on pre-defined static procedures or simple response mechanisms based on single-point alarms (such as exceeding a single physiological indicator or a single piece of equipment shortage). For example, dispatchers or team leaders may assign tasks according to standardized operating procedures or passively seek alternatives when receiving an equipment shortage alarm. Due to the static nature of these mechanisms, existing technologies struggle to synchronously correlate and analyze the real-time physiological evolution trends of patients (such as the rate of deterioration), the specific behavioral interactions of team members (such as task parallelism and interaction frequency), and the precise status and consumption of various critical resources (equipment, medications). Decision-making is often based on fragmented or delayed information. Secondly, there are immediate overload risks at critical task execution nodes (such as the probability of calculating overload limits) and the level of resource chain security supporting the entire collaborative path. Managers often struggle to effectively ascertain the real-time resilience and vulnerability of the entire collaborative system, relying instead on experience or simple thresholds for post-event intervention.

[0004] Therefore, existing technologies have shortcomings and need to be improved. Summary of the Invention

[0005] To address one or more problems in the existing technology, the main objective of this application is to provide a method and system for dynamic resource scheduling in emergency rescue team collaboration based on multimodal data-driven approaches.

[0006] To achieve the aforementioned objectives, this application proposes a dynamic resource scheduling method for emergency medical teams based on multimodal data-driven collaboration, the method comprising:

[0007] Collect patient physiological time-series data, team behavioral interaction data, and resource status data in emergency scenarios to form a heterogeneous data stream;

[0008] Based on the heterogeneous data stream, the topological dependencies of the team collaboration link are deconstructed;

[0009] Based on the aforementioned topological dependencies and heterogeneous data streams, a collaborative robustness index is calculated.

[0010] When the collaborative robustness index is lower than the preset index threshold, a multimodal constraint-driven conflict resolution decision is executed.

[0011] Monitor the changes in collaboration efficiency after resource allocation, and adjust the weight coefficients of the collaboration robustness index and the preset index thresholds based on the preset optimization objectives.

[0012] This application also provides a multimodal data-driven dynamic resource scheduling system for emergency medical teams, including:

[0013] The data acquisition module is used to collect patient physiological time-series data, team behavioral interaction data, and resource status data in emergency scenarios, forming a heterogeneous data stream;

[0014] The deconstruction module is used to deconstruct the topological dependencies of the team collaboration link based on the heterogeneous data stream;

[0015] The calculation module is used to calculate the collaborative robustness index based on the topological dependencies and heterogeneous data streams;

[0016] The execution module is used to execute a multimodal constraint-driven conflict resolution decision when the collaborative robustness index is lower than a preset index threshold.

[0017] The monitoring module is used to monitor the changes in collaboration efficiency after resource allocation, and, in conjunction with preset optimization targets, correct the weight coefficients of the collaboration robustness index and the preset index thresholds.

[0018] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0019] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0020] This application presents a multimodal data-driven emergency team collaborative dynamic resource scheduling method and system. It simultaneously collects patient physiological time-series data, team behavioral data, and resource status data to construct a complete data foundation, eliminating the information blind spots of traditional single-dimensional decision-making. Based on real-time events (task changes, resource occupancy), it deconstructs and reconstructs team topological dependencies (task timing + resource competition), accurately capturing bottleneck shifts in the collaborative chain (such as resource conflicts caused by new patients), overcoming the rigidity and sluggishness of static process models. The "Collaborative Robustness Index (CRI)" integrates the calculation of overload risk probability at key nodes with the assessment of resource availability along the collaborative path, providing a quantitative assessment of the system's risk resistance capability, identifying complex collapse risks earlier than traditional single-threshold alarms. When the CRI falls below the threshold, it drives resource reallocation and alternative resource scheduling according to preset priorities, generating primary-backup execution instructions to ensure priority protection of core medical objectives during resource conflicts. Based on the collaborative efficiency data after resource allocation, it dynamically adjusts the CRI weight and threshold, achieving adaptive iteration of the algorithm, allowing the scheduling strategy to continuously approach the preset optimization target. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating an embodiment of the dynamic resource scheduling method for emergency medical teams based on multimodal data-driven collaboration according to this application.

[0022] Figure 2 This is a flowchart illustrating an embodiment of the dynamic resource scheduling method for emergency medical teams based on multimodal data-driven collaboration according to this application.

[0023] Figure 3 This is a schematic block diagram of a multimodal data-driven emergency team collaborative dynamic resource scheduling system according to an embodiment of this application;

[0024] Figure 4 This is a schematic block diagram of the structure of a computer device according to an embodiment of this application.

[0025] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] Reference Figure 1 This application provides a method for dynamic resource scheduling in emergency medical teams based on multimodal data-driven collaboration, the method comprising:

[0028] S1. Collect patient physiological time-series data, team behavioral interaction data, and resource status data in emergency scenarios to form a heterogeneous data stream;

[0029] S2. Based on the heterogeneous data stream, deconstruct the topological dependencies of the team collaboration link;

[0030] S3. Calculate the collaborative robustness index based on the topological dependency relationship and heterogeneous data flow;

[0031] S4. When the collaborative robustness index is lower than the preset index threshold, execute the conflict resolution decision driven by multimodal constraints.

[0032] S5. Monitor the changes in collaboration efficiency after resource allocation, and adjust the weight coefficients of the collaboration robustness index and the preset index thresholds in conjunction with the preset optimization objectives.

[0033] As described in steps S1-S3 above, emergency rescue scenarios involve dynamic information flows across three core dimensions: Patient dimension: Physiological time-series data (such as changes in heart rate, blood pressure, blood oxygen, respiratory rate, and body temperature over time) directly reflects the real-time dynamics and deterioration trends of the patient's vital signs, serving as the most fundamental basis for decision-making. Team dimension: Team behavioral interaction data (such as medical staff action recognition, task start / end timestamps, voice commands, personnel locations, and equipment operation logs) characterizes the execution status, workload, and process bottlenecks of team collaboration. Resource dimension: Resource status data (such as the usage status / location / remaining power or consumables of key equipment, drug inventory, bed occupancy, and oxygen supply) reflects the availability and consumption of the material foundation supporting collaboration. These data, from different sources, in different formats, and potentially with different update frequencies, are synchronously collected, time-aligned, and initially integrated to form a unified "heterogeneous data flow." This constructs a comprehensive real-time data foundation covering all elements of "patient-team-resources," as data from a single dimension cannot fully reflect the complexity of emergency rescue collaboration. Ignoring any aspect (e.g., focusing solely on physiological data without knowing resource availability, or focusing solely on resources without knowing which patient is most critical or which team is busiest) will lead to serious biases in dispatch decisions. By fusing multimodal data, the system can obtain a panoramic view. Complex emergency team collaboration is abstracted into a dynamic network model. Nodes: Identifying key roles in the collaboration (e.g., Doctor A, Nurse B, Ventilator C, Defibrillator D). Task Sequence Constraints: Establishing dependencies on the execution order between nodes based on event logs (e.g., intubation before ventilator use) and patient status (e.g., blood transfusion required when blood pressure drops to a certain threshold). Resource Competition Conflicts: Identifying competition relationships that arise when multiple nodes / tasks simultaneously require the same limited resource (e.g., all needing the same ultrasound equipment or the same bag of type O blood). This process is dynamic (responding to changes in the real-time data stream): changes in task status (e.g., task completion, interruption) and resource events (e.g., equipment failure, removal) trigger real-time reconstruction of the topology. This transforms ambiguous team collaboration into a clear, structured computational model that includes task flow logic (sequence) and resource interaction logic (competition). This study reveals the critical paths, bottlenecks, and potential conflict points in the collaborative process. Dynamic elements ensure the model always reflects the latest situation on-site, avoiding decisions based on outdated information. Emergency response procedures are not static linear processes, but rather complex networks interwoven with personnel, equipment, and tasks. Understanding how nodes are interconnected through tasks and time (sequence) and resources (competition) is a prerequisite for effective resource coordination and conflict management. A key health indicator (CRI) is defined to measure the stability and resilience of the entire collaborative system. It is not a single measurement but integrates two core risk dimensions. For nodes performing critical tasks (especially high-risk tasks), the study analyzes the likelihood of their real-time workload exceeding limits (too high personnel operation frequency? Too many patients being managed simultaneously?).This relies on key nodes identified by topology, node role capabilities, real-time physiological data (determining the severity of urgency), and interaction data (measuring workload). It assesses the sufficiency and reliability of resources supporting the core collaboration path. This requires analyzing the types of necessary resources along the path, their real-time availability, the existence of equivalent alternatives, and whether the maximum delay in acquiring alternatives is acceptable. The assessment results from these two dimensions (potentially through predefined algorithms and models) are dynamically weighted and fused to ultimately output a quantitative Collaboration Robustness Index (CRI) value. This provides a core indicator for quantitatively assessing the overall health of the system. In dynamic and complex environments, a system-level, forward-looking risk metric is needed to guide decision-making. CRI integrates personnel workload and resource availability—two of the most likely sources of collaboration failure—allowing for earlier and more comprehensive perception of systemic risks than a single dimension (such as looking only at equipment status or an individual's physiological parameters).

[0034] As described in steps S4-S5 above, a CRI below the threshold signals insufficient system stability, triggering an automated decision-making process. The core of this decision-making is to resolve conflicts and optimize resource allocation based on pre-defined multi-dimensional constraints: Constraint sources: The rate of deterioration of patient vital signs (first priority): ensuring life-saving resources for the most critical patients. The carrying capacity of overloaded nodes (second priority): alleviating pressure on heavily overloaded personnel and preventing their collapse from causing greater losses. The delay tolerance of alternative resources (third priority): in the event of resource conflicts, choosing operations that can tolerate a certain delay or prioritizing patient care. Decision actions: Under the premise of satisfying the highest priority constraints, reallocate resources (e.g., transferring someone from a secondary task to support a high-risk task or overloaded node), allocate alternative resources (triggered when delay tolerance is exceeded), and generate backup execution plans (increasing redundancy). Ensure that decisions comply with emergency medical ethics and efficiency principles (life priority > team load control > resource scheduling efficiency). Enhance the system's resilience and continuous operation under pressure (avoiding the collapse of the entire collaboration chain due to the failure of a single point). Conflicts (resource conflicts, personnel conflicts, task conflicts) are common in emergency medical care; manual decision-making is slow and prone to errors. Predefine clear, quantifiable, and prioritized constraints to ensure consistency and ethical compliance in machine decision-making. After making scheduling decisions and allocating resources, track the resulting effect data in real time, such as: Has the completion time of critical tasks been shortened? Has the average workload of team members decreased? Has the resource idle rate improved? Have the patient's physiological parameters improved? Compare the actual effects with preset goals (such as "minimizing delays in high-risk tasks," "minimizing team overload risk," and "maximizing resource utilization efficiency"). If a discrepancy is found between the CRI-predicted risk and the actual risk (e.g., the resource availability assessment weight is too low, resulting in no early warning), adjust the relative weights of load risk and resource availability in the CRI calculation. If the current CRI threshold setting is found to be too sensitive (frequent triggering but ineffective) or insensitive (failure to provide timely warnings), adjust the threshold level of the triggering intervention indicator. Continuously improve the accuracy, sensitivity, and effectiveness of CRI decisions. This allows the scheduling strategy to adapt to the fine-tuning needs of different emergency scenarios, different teams, or strategy preferences. Preset weights and thresholds may not be optimal initially, and the optimal strategy may differ in different emergency situations. Closed-loop feedback is a key indicator of system intelligence. It continuously improves internal parameters (weights) and decision points (thresholds) based on actual results, enabling system performance to iteratively improve in practice and reach or exceed preset optimization goals (such as faster, more stable, and more efficient). This overcomes the limitations of static models or rules.

[0035] As described above, synchronous collection of patient physiological time-series data, team behavioral data, and resource status data constructs a complete data foundation, eliminating the information blind spots of traditional single-dimensional decision-making. Based on real-time events (task changes, resource usage), the team's topological dependencies (task timing + resource competition) are deconstructed and reconstructed, accurately capturing bottleneck shifts in the collaboration chain (such as resource conflicts caused by new patients), overcoming the rigidity and sluggishness of static process models. The "Collaboration Robustness Index (CRI)" integrates the calculation of overload risk probability at key nodes with the assessment of resource availability in the collaboration path, providing a quantitative assessment of the system's risk resistance, identifying complex collapse risks earlier than traditional single-threshold alarms. When the CRI falls below the threshold, resource reallocation and alternative resource scheduling are driven according to preset priorities, and master-slave execution instructions are generated to ensure that core medical objectives are prioritized during resource conflicts. Based on the collaboration efficiency data after resource allocation, the CRI weight and threshold are dynamically adjusted to achieve adaptive iteration of the algorithm, continuously bringing the scheduling strategy closer to the preset optimization target.

[0036] Reference Figure 2 In one embodiment, the step of deconstructing the topological dependencies of the team collaboration link based on the heterogeneous data stream includes:

[0037] S21. Extract the emergency rescue task execution entities from the heterogeneous data stream as a set of nodes, wherein the entities include medical personnel roles and key equipment;

[0038] S22. Based on the sequence of task event logs and changes in the patient's physiological state, generate task timing constraints between nodes.

[0039] S23. When multiple nodes request the same resource at the same time, a resource competition conflict relationship is established between the corresponding nodes.

[0040] S24. Reconstruct the node dependency graph in response to task status changes or resource occupancy events in the real-time data stream.

[0041] As described above, the entities that specifically undertake tasks in emergency medical collaboration are abstracted as nodes in the network model. The scope of nodes includes two core entities: medical personnel roles: such as attending physicians, intubation nurses, pharmacists, stretcher bearers, etc. These can be role instances (e.g., "Dr. Zhang is handling patient A's intubation") or general capability units of roles (e.g., "all nurses capable of performing intravenous punctures"). Specific implementations can be chosen based on data granularity and system design. Key equipment: resources with independent "service" capabilities, such as ventilators, defibrillators, ultrasound machines, ECMO equipment, and specific medication kits (e.g., atropine injections). Only by clearly defining the "points" can the "lines" (relationships) be analyzed. Including personnel and equipment ensures the model covers all functional participants in the collaboration, avoiding omissions of critical dependencies (e.g., equipment failure will interrupt node functionality). Task event logs: record the actual sequence of actions, such as "intubation started," "intubation completed," "medication injection started," etc. These logs include timestamps and information about the executing entity (node). Patient physiological state changes: This refers to the dynamic changes in a patient's vital signs (such as a sudden drop in blood pressure or arrhythmia). These changes may trigger new emergency tasks or alter the priority / execution order of existing tasks. By analyzing the logs of consecutive or causally related tasks (e.g., "bleed control" must precede "debridement and suturing"), the system infers the sequential dependencies between nodes performing these tasks (Node A must complete task X before Node B can begin task Y). When a patient's physiological state reaches a preset critical threshold or conforms to a certain deterioration pattern, the system automatically generates the execution requirements for tasks related to that state and their necessary node dependencies. For example, if a patient's blood oxygen saturation remains below 85%, the system may generate an "emergency oxygen therapy" task, which depends on the "respiratory therapist" and "oxygen mask / ventilator" nodes, and this task may have temporal priority over some currently ongoing non-emergency tasks. The basic backbone structure and critical path of the collaborative process are determined. This is crucial for simulating real-world workflows. Emergency tasks are not isolated; they have strict sequential logic (e.g., the airway must be opened before oxygen can be administered). This ensures that the topology reflects the urgent needs arising from the patient's real-time health status. At a given point in time (or within a very short time window), if multiple nodes (whether personnel or equipment nodes) require the same specific, limited, and non-shareable resource (such as a single bag of type O blood, the only available portable ultrasound machine, or a diagnosis confirmation from a specific specialist), the system identifies this resource request conflict. In the topology model, a new relationship is created between these conflicting nodes, called a "resource competition conflict relationship." This identifies a mutually exclusive relationship where these nodes compete for the resource. It explicitly models the conflict and competitive pressure between nodes when critical resources are scarce. Resource conflict is the most common obstacle to collaboration in emergency care.Traditional resource scheduling might only mark the "resource occupancy" status, but this step goes further, directly modeling the conflict as a relationship (edge) between nodes, explicitly embedding the conflict into the collaborative topology graph. This allows for intuitive and efficient location of the source of the conflict (which resource?) and the participants (which nodes are competing?) when calculating CRI (especially the resource availability part), providing direct information input for intelligent decision-making. Without establishing such explicit conflict relationships, the topology model is incomplete. Task status changes: such as task cancellation, unexpected task interruption (e.g., due to equipment failure), early task completion, and the addition of an emergency task (due to physiological changes or other reasons). Resource occupancy events: such as a critical resource being taken, used up, returned, malfunctioning, or newly replenished. Restructuring: When these events occur, the system will adjust the data based on the latest data (heterogeneous data stream). Adding / deleting / modifying nodes: such as removing nodes that have completed all tasks (or marking them as idle), adding nodes for new equipment, and modifying the status of nodes representing a certain role (e.g., a doctor leaving due to an emergency). Adding / deleting / modifying relationships: For example, a task interruption leads to the removal or modification of all temporal dependencies of its subsequent tasks. Resource depletion or occupation leads to the removal (no use competing for resources anymore) or the creation (a new task competing for them) of competition for that resource. Resource release or addition satisfies a task's requirements, thereby removing the waiting relationships for that task and potentially triggering new temporal relationships. Result: Generates a completely new dependency graph (topology) that most accurately reflects the current (or near-future) collaborative chain. It exhibits high responsiveness and adaptability to dynamic environmental changes (interruptions, delays, new requirements).

[0042] In one embodiment, the step of calculating the collaborative robustness index based on the topological dependencies and heterogeneous data streams includes:

[0043] Based on the node roles in the topological dependency relationship and the patient's physiological time-series data, the probability of real-time overload of nodes performing critical tasks is analyzed.

[0044] Assessing the resource availability of collaborative paths includes: identifying essential resource types in critical collaborative paths based on resource status data and task timing constraints; verifying the real-time availability of the essential resource types and the existence of functionally equivalent alternative resources; and calculating the maximum expected delay time for deploying alternative resources.

[0045] The load over-limit probability and the resource availability assessment results are dynamically weighted and integrated.

[0046] Based on the weighted fusion results, a collaborative robustness index value is output.

[0047] As mentioned above, topological dependencies provide crucial information, such as which nodes are performing tasks (especially high-risk tasks, which typically rely on topological relationships to identify their critical paths) and the role types of these nodes (e.g., attending physician, scrub nurse, respiratory therapist; different roles have different workloads). Patient physiological time-series data provides real-time dynamics of patient vital signs (e.g., ECG, blood pressure, blood oxygen trends) to determine the severity of the task and the target patient. The analysis focuses on the node performing the most critical task (e.g., an attending physician performing CPR on a patient experiencing cardiac arrest). It analyzes whether this node is about to become overwhelmed or has already become overwhelmed under the current multiple pressures (workload, complexity of the physiological crisis). The calculation method identifies high-risk tasks (usually based on the degree of deterioration of physiological data). It locates the node performing the task (provided by topological relationships). Using a pre-trained load risk prediction model, the above quantified values ​​are used as input to output a probability value (between 0 and 1) representing the probability that the node will experience overload (e.g., operational error, response delay, task failure) in the near future (e.g., a few minutes). The vulnerability risk of key personnel / equipment nodes due to excessive stress was assessed. This is an important component of CRI (Critical Responsibility Analysis). It provides early warning signals for system intervention (e.g., identifying that "Dr. Zhang has a 70% probability of being unable to effectively handle the current task within 2 minutes"). Personnel overload and insufficient equipment capacity are the primary causes of collaboration failure. This step transforms the abstract concept of "high stress" into a calculable probabilistic risk. The node roles reflect the responsibilities and capacity limits of different personnel (the attending physician bears far greater responsibility and stress than the assistant). Incorporating patient physiological time-series data ensures that the load assessment closely reflects the current level of medical urgency (the load requirements for a doctor are vastly different between a stable patient and a terminally ill patient). The introduction of a pre-trained model addresses the complexity of load risk prediction (nonlinearity, multi-factor coupling). Focusing on key collaboration paths: based on topological relationships, those collaboration paths with the greatest impact on patient outcomes and the highest time constraints are identified (e.g., the core path in cardiac arrest treatment consisting of "defibrillation → medication injection → continuous chest compressions"). Essential Resource Identification: Traverse all tasks on the critical path, combining task definitions and temporal constraints to extract the specific resource types essential for completing these tasks (e.g., defibrillators, epinephrine injection solutions, intravenous access kits). Resource Status Data: Provide the existence, location, and status information of these essential resources. Quantity Verification: Check the real-time available quantity of each essential resource (e.g., quantity in stock, quantity of unused equipment). Alternative Resource Verification: Assess the existence of functionally equivalent alternatives (e.g., other models of defibrillators, alternative drugs with similar chemical structures) to evaluate the basic feasibility and potential vulnerabilities of completing the critical path at the resource level. Insufficient resource availability is a cause of collaboration failure. Traditional systems may only report "XX inventory is 0," but cannot address: "How much impact does this have on core processes? Are there any backup plans?""How long will the backup plan take?" This step focuses on key paths: resource assessment is prioritized, avoiding waste on non-critical paths. Essential resource identification eliminates redundant interference. Alternative resource verification provides potential solutions, increasing system resilience. Dynamic weighting integrates the load overload probability and resource availability assessment results: Load overload probability: a probability value. Resource availability assessment result: a comprehensive evaluation value. This result typically needs to be converted into a "risk" or "reliability" score (e.g., sufficient resources + low-latency alternatives > high availability / low risk score; resource shortage + no alternatives or high-latency alternatives > low availability / high risk score). Specific conversion algorithms can consider factors such as availability quantity satisfaction, alternative resource availability, and latency time conversion factors. Dynamic weighting: dynamically adjusts the relative weights (weighting coefficients) in the final CRI based on the current global situation or preset strategy preferences. This integrates two different characteristics (personnel load vs....). (For material support) However, equally important risk dimensions are uniformly quantified and integrated into a comprehensive indicator (CRI_intermediate). A dynamic weighting mechanism allows CRI to flexibly adapt to the sensitive points of different scenarios, enhancing the indicator's adaptability and relevance. The essence of CRI lies in the comprehensive assessment of the two core collaborative risk sources: "human factors" and "material factors." However, their contributions to system robustness differ in different scenarios: simple addition or fixed weights are not intelligent enough. For example, in a battlefield ambulance with extremely scarce resources, the risk of an overloaded medical staff member (even with a high probability) may not be as fatal as the lack of a tourniquet. Dynamic weighting allows the system to adjust its focus based on real-time "pain points." Based on the weighted fusion result, the collaborative robustness index value is output: the fusion intermediate value calculated in the above steps. The intermediate value is standardized, normalized, or mapped to output the final collaborative robustness index value (CRI).

[0048] In one embodiment, the step of analyzing the real-time overload probability of nodes performing critical tasks based on the node roles in the topological dependency relationship and the patient's physiological time-series data includes:

[0049] Based on the node analysis task type in the topological dependency relationship, the severity level is analyzed according to the patient's physiological time series data.

[0050] Based on the task type and urgency level, determine the execution nodes for high-risk tasks that require immediate execution;

[0051] Based on the identified high-risk and urgent task execution nodes, the real-time load status of the high-risk and urgent task execution nodes is quantified to obtain node quantification values. The quantification of the real-time load status of the nodes includes: quantifying the number of parallel tasks corresponding to the node role; quantifying the deviation of the patient's physiological parameters; and quantifying the real-time interactive action frequency of the node execution subject.

[0052] The node quantization value is input into a pre-trained load risk prediction model, and the load risk prediction model outputs the real-time load over-limit probability.

[0053] As mentioned above, the collaboration relationship graph is used to identify which tasks are currently being performed by which individuals (nodes). Simultaneously, continuous physiological data of the patient (such as heart rate, blood pressure, and blood oxygen) is analyzed in real time. This data is compared with preset thresholds or healthy ranges to determine the patient's level of criticality (e.g., mild risk, moderate risk, high risk, critical). The level of criticality is usually determined by the magnitude of deviation from the baseline value and the rate of deterioration. The urgency of tasks and associated patient states are precisely located. The patient's real-time health risk is assigned to specific tasks, providing crucial information for subsequently identifying key nodes and assessing their workload. For example, the task of "treating patient A in cardiogenic shock" is determined to be significantly more critical than the task of "treating patient B's superficial abrasion." Based on the results of the previous step: Task type screening: Identify task combinations with high timeliness requirements (such as cardiopulmonary resuscitation, hemostasis for massive bleeding) and associated patients in high-risk states (such as cardiac arrest). Node location: Based on the collaboration relationship graph, identify the specific personnel (medical staff nodes) or equipment (critical equipment nodes) responsible for performing these most critical and urgent tasks. Identify the targets for stress assessment. Clearly define the core execution units in the current collaboration chain that pose the greatest stress risk. For example, identifying "Dr. Zhang San performing CPR on a patient experiencing cardiac arrest" or "the only defibrillator operating in high-frequency mode" are high-risk, urgent task execution nodes whose workload must be closely monitored. For these identified core nodes, perform real-time data measurement and numerical calculations across three dimensions: Number of parallel tasks: Calculate the number of specific tasks that the node role is currently handling or closely monitoring simultaneously (e.g., Dr. Zhang San must handle patient A's cardiac arrest while simultaneously instructing nurses on medication errors in patient B). Physiological parameter deviation: For specific patients associated with the node's tasks (e.g., cardiac arrest patients under Dr. Zhang San's care), calculate the degree of difference between the actual values ​​and ideal or safe range benchmarks of their key physiological parameters (e.g., calculate the extent to which the patient's mean arterial pressure is below 60 mmHg). Node execution subject real-time interaction frequency: This measures the frequency of key operations or interactions performed by the node execution subject (person / device) per unit time (e.g., the number of times Dr. Zhang performs chest compressions per minute, or the number of times a defibrillator automatically analyzes an electrocardiogram every 10 minutes). This transforms the abstract concept of "load pressure" into multi-dimensional, concrete numerical values. This set of quantified values ​​(node ​​quantification values) objectively describes the comprehensive state of the target node's workload (parallel tasks), required attention and pressure (physiological deviation), and physical workload (interaction frequency) in the current environment. For example, Dr. Zhang's node quantification values ​​might be: 2 parallel tasks (CPR + guidance), 85% patient physiological deviation (extremely severe), and an interaction frequency of 50 times / minute (high-intensity compressions). The multi-dimensional node quantification values ​​obtained in the previous step (such as the number of parallel tasks, physiological deviation, and interaction frequency) are then combined and input into a pre-trained risk prediction model.It can learn and identify the potential risk probabilities corresponding to different combinations of quantitative values. After processing the input, the model directly outputs a probability value (e.g., 0.75), indicating how likely it is that the node will fail, be delayed, or experience a decline in quality due to overload in the next few minutes. It transforms multidimensional real-time data into scientifically predictive risk values. Based on data and empirical evidence, it accurately judges whether the core executor is "about to collapse" and the "probability of collapse" (e.g., predicting a 75% probability that Dr. Zhang San will make an operational error or experience a sudden drop in efficiency within the next two minutes). This is a direct, quantitative signal that triggers subsequent resource adjustments and conflict resolution.

[0054] In one embodiment, the step of performing multimodal constraint-driven conflict resolution decision-making includes:

[0055] Based on the aforementioned collaborative robustness index value, the constraint priority order is determined according to the first priority, the second priority, and the third priority, wherein the first priority is the rate of physiological deterioration of the patient; the second priority is the task capacity of the overloaded node; and the third priority is the tolerance for delay of alternative resources.

[0056] According to the constraint priority order, after the first priority is met, the remaining resources are allocated to the overloaded nodes with the second priority constraint to reduce their load; when the third priority constraint is broken, alternative resource scheduling is triggered.

[0057] When the resource guarantee level of the collaborative path is lower than a preset redundancy threshold,

[0058] Generate at least one backup instruction that has the same target as the main scheduling instruction but is executed by a different entity.

[0059] As mentioned above, when the collaborative robustness index indicates insufficient system stability, the system activates conflict resolution decision-making. The core of this decision-making follows a pre-defined three-tiered constraint priority order: First priority (highest): Patient physiological deterioration rate. The core objective is to prevent rapid deterioration of the patient's vital signs (e.g., a sudden drop in blood pressure per minute in a patient with massive hemorrhage), which is the fundamental starting point for medical intervention. Second priority: The workload capacity of overloaded nodes. This focuses on how much additional work the medical staff or equipment nodes performing critical tasks can handle (e.g., assessing how many tasks Dr. Zhang can safely take on at present). Third priority: Tolerance for alternative resource delays. This considers the time constraints required for waiting for alternative resources to arrive during resource shortages (e.g., a certain piece of equipment must arrive in 3 minutes, otherwise treatment will be affected). Resources are allocated according to the constraint priority order, prioritizing the first priority: all decisions must prioritize how to reduce the patient's physiological deterioration rate most quickly and effectively (e.g., hemostasis, blood pressure increase, oxygen supply). Resource requests that conflict with this objective are automatically downgraded. Secondly, prioritize the second priority: While ensuring patient safety, allocate remaining available resources (personnel, equipment) to key nodes with the lowest workload and highest risk (overload), reducing their workload (e.g., assigning a new task to a relatively idle nurse, Dr. Wang, instead of an already overloaded doctor). Address the third priority: If resource allocation results in waiting time for alternative resources exceeding the maximum delay tolerated by the operation or patient, immediately initiate alternative resource scheduling (e.g., automatically dispatching to an external resource pool, initiating alternative transport, or calling for higher-level support). Achieve intelligent, step-by-step resource allocation. When the system assessment finds that the resource security level supporting a key emergency chain (considering redundancy, alternative resources, and delays) is below the safe redundancy threshold (i.e., resources are extremely scarce or vulnerable), it means that the path has a high risk of interruption. The system will proactively generate at least one backup instruction. The backup instruction must pursue the same emergency goal as the original main instruction, but the specific executor (subject) is different.

[0060] In one embodiment, after the step of executing a multimodal constraint-driven conflict resolution decision when the collaborative robustness index is lower than a preset index threshold, the method further includes:

[0061] Real-time monitoring of the aforementioned collaborative robustness indicators;

[0062] When the collaborative robustness index is continuously lower than the preset threshold and the resource availability is close to zero, a resource depletion alarm and cross-domain collaboration request are broadcast to the preset external emergency network. The request includes the type of necessary resources, the patient's rate of physiological deterioration, and the geographical location.

[0063] As mentioned above, the system does not cease evaluation after implementing conflict resolution decisions and reallocating resources. It continuously and uninterruptedly tracks changes in the core health indicator—the Collaboration Robustness Index (CRI). This ensures continuous monitoring of the health of the entire collaborative system, assesses the effectiveness of interventions, and alerts to any new or ongoing deterioration. The system detected two extremely serious and simultaneous signals:

[0064] A persistently low CRI (Critical Resource Index): Multiple monitoring of the CRI value, or its continued presence below the safety threshold (preset threshold) for triggering internal intervention over a prolonged period, indicates that internal resource allocation efforts have failed to effectively improve the overall stability of the system. Furthermore, assessments show that the resource availability level supporting critical collaboration links has plummeted to an extremely low point (approaching zero). This means that essential resources are almost completely exhausted; there are no functionally equivalent alternatives available. Even if alternative resources are invoked from outside, the expected delay may far exceed the window of opportunity for intervention. The simultaneous occurrence of these two signals signifies the depletion of internal resources within the local team, making it impossible to maintain the operation of critical collaboration links through internal scheduling. The system automatically issues the highest-level distress signal (resource depletion alert) and a clear cross-domain collaboration request.

[0065] In one embodiment, the method further includes:

[0066] When new patients cause resource overload, the topology dependency is reset in response to the addition of high-risk physiological signals in the patient's physiological time-series data stream.

[0067] Based on the priority order of the constraints, existing resources are reallocated.

[0068] As described above, the system detects a new patient joining the emergency scenario (a "new signal source" appears in the real-time patient physiological data stream), and the patient's physiological signals are identified as high-risk (e.g., weak pulse and blood pressure <60 mmHg, or ventricular fibrillation on ECG). The critical consequence is that overall resource demands exceed the team's current processing capacity (resource overload). The existing collaboration graph (who is doing what, what resources they depend on, and task order) instantly becomes partially or completely invalid. The system immediately uses this moment as a starting point to restart the entire process of deconstructing the collaboration chain. This involves: re-identifying all nodes, including the medical care and equipment required for both existing and newly added patients; re-establishing the task sequence, creating new task logic based on the latest status of all patients; and remapping resource dependencies and conflicts, assessing the resources required for all tasks and identifying new resource contention points (e.g., all critically ill patients needing the same ventilator). Faced with the sudden surge in patients and resource overload, the old model is abandoned, and a new collaboration graph reflecting the current global real-world needs is forcibly established. Utilizing a pre-defined three-tiered priority order (patient deterioration rate first, surplus capacity of overloaded nodes second, and tolerance for delayed alternative resources third), and based on the reset global collaborative topology and the current status of all patients / tasks, a thorough reallocation and planning of all existing limited resources (including human and material resources) is undertaken. This achieves fairness, efficiency, and priority assurance. In extreme cases where resources are insufficient to serve all needs (resource overload), resource allocation is forcibly reorganized according to pre-defined medical ethics priorities (most life-threatening situations prioritized), ensuring that the most scarce resources are used for the most critical issues.

[0069] Reference Figure 3 This application also provides a multimodal data-driven dynamic resource scheduling system for emergency medical teams, comprising:

[0070] Acquisition module 1 is used to collect patient physiological time-series data, team behavioral interaction data, and resource status data in emergency scenarios, forming a heterogeneous data stream;

[0071] Deconstruction module 2 is used to deconstruct the topological dependencies of the team collaboration link based on the heterogeneous data stream;

[0072] Calculation module 3 is used to calculate the collaborative robustness index based on the topological dependency relationship and heterogeneous data flow;

[0073] Execution module 4 is used to execute multimodal constraint-driven conflict resolution decision when the collaborative robustness index is lower than a preset index threshold;

[0074] Monitoring module 5 is used to monitor the changes in collaboration efficiency after resource allocation, and, in conjunction with preset optimization targets, correct the weight coefficients of the collaboration robustness index and the preset index thresholds.

[0075] As described above, it is understood that each component of the multimodal data-driven emergency team collaborative dynamic resource scheduling system proposed in this application can realize the function of any of the multimodal data-driven emergency team collaborative dynamic resource scheduling methods described above, and the specific structure will not be repeated.

[0076] Reference Figure 4 This application also provides a computer device, which may be a server, and its internal structure may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores monitoring data and other data. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a multimodal data-driven dynamic resource scheduling method for emergency medical team collaboration.

[0077] The processor described above executes the multimodal data-driven dynamic resource scheduling method for emergency medical team collaboration, including: collecting patient physiological time-series data, team behavioral interaction data, and resource status data in the emergency medical scenario to form a heterogeneous data stream; deconstructing the topological dependencies of the team collaboration links based on the heterogeneous data stream; calculating a collaboration robustness index based on the topological dependencies and the heterogeneous data stream; executing a multimodal constraint-driven conflict resolution decision when the collaboration robustness index is lower than a preset index threshold; monitoring the collaboration efficiency change data after resource allocation, and correcting the weight coefficient of the collaboration robustness index and the preset index threshold in conjunction with a preset optimization objective.

[0078] One embodiment of this application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements a dynamic resource scheduling method for emergency medical team collaboration based on multimodal data-driven approaches. The method includes the following steps: collecting patient physiological time-series data, team behavioral interaction data, and resource status data in an emergency medical scenario to form a heterogeneous data stream; deconstructing the topological dependencies of the team collaboration links based on the heterogeneous data stream; calculating a collaboration robustness index based on the topological dependencies and the heterogeneous data stream; executing a multimodal constraint-driven conflict resolution decision when the collaboration robustness index is lower than a preset index threshold; monitoring the changes in collaboration efficiency after resource allocation, and correcting the weight coefficient of the collaboration robustness index and the preset index threshold based on a preset optimization objective.

[0079] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0080] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0081] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A dynamic resource scheduling method for emergency medical team collaboration based on multimodal data-driven approaches, characterized in that, The method includes: Collect patient physiological time-series data, team behavioral interaction data, and resource status data in emergency scenarios to form a heterogeneous data stream; Based on the heterogeneous data stream, the topological dependencies of the team collaboration link are deconstructed; Based on the aforementioned topological dependencies and heterogeneous data streams, a collaborative robustness index is calculated. When the collaborative robustness index is lower than the preset index threshold, a multimodal constraint-driven conflict resolution decision is executed. Monitor the changes in collaboration efficiency after resource allocation, and adjust the weight coefficients of the collaboration robustness index and the preset index thresholds based on the preset optimization objectives.

2. The method for dynamic resource scheduling of emergency medical teams based on multimodal data-driven collaboration according to claim 1, characterized in that, The step of deconstructing the topological dependencies of the team collaboration link based on the heterogeneous data stream includes: From the heterogeneous data stream, the entities that perform emergency rescue tasks are extracted as a set of nodes, and the entities include medical personnel roles and key equipment; Based on the sequence of task event logs and changes in the patient's physiological state, task timing constraints are generated between nodes. When multiple nodes request the same resource at the same time, a resource competition conflict relationship is established between the corresponding nodes. Reconstruct the node dependency graph in response to task status changes or resource usage events in the real-time data stream.

3. The method for dynamic resource scheduling of emergency medical teams based on multimodal data-driven collaboration according to claim 1, characterized in that, The step of calculating the collaborative robustness index based on the topological dependencies and heterogeneous data streams includes: Based on the node roles in the topological dependency relationship and the patient's physiological time-series data, the probability of real-time overload of nodes performing critical tasks is analyzed. Assessing the resource availability of collaborative paths includes: identifying essential resource types in critical collaborative paths based on resource status data and task timing constraints; verifying the real-time availability of the essential resource types and the existence of functionally equivalent alternative resources; and calculating the maximum expected delay time for deploying alternative resources. The load over-limit probability and the resource availability assessment results are dynamically weighted and integrated. Based on the weighted fusion results, a collaborative robustness index value is output.

4. The method for dynamic resource scheduling of emergency medical teams based on multimodal data-driven collaboration according to claim 3, characterized in that, The step of analyzing the real-time overload probability of nodes performing critical tasks based on the node roles in the topological dependency relationship and the patient's physiological time-series data includes: Based on the node analysis task type in the topological dependency relationship, the severity level is analyzed according to the patient's physiological time series data. Based on the task type and urgency level, determine the execution nodes for high-risk tasks that require immediate execution; Based on the identified high-risk and urgent task execution nodes, the real-time load status of the high-risk and urgent task execution nodes is quantified to obtain node quantification values. The quantification of the real-time load status of the nodes includes: quantifying the number of parallel tasks corresponding to the node role; quantifying the deviation of the patient's physiological parameters; and quantifying the real-time interactive action frequency of the node execution subject. The node quantization value is input into a pre-trained load risk prediction model, and the load risk prediction model outputs the real-time load over-limit probability.

5. The method for dynamic resource scheduling of emergency medical teams based on multimodal data-driven collaboration according to claim 4, characterized in that, The steps of executing multimodal constraint-driven conflict resolution decision-making include: Based on the aforementioned collaborative robustness index value, the constraint priority order is determined according to the first priority, the second priority, and the third priority, wherein the first priority is the rate of physiological deterioration of the patient; the second priority is the task capacity of the overloaded node; and the third priority is the tolerance for delay of alternative resources. According to the constraint priority order, after the first priority is met, the remaining resources are allocated to the overloaded nodes with the second priority constraint to reduce their load; when the third priority constraint is broken, alternative resource scheduling is triggered. When the resource guarantee level of the collaborative path is lower than a preset redundancy threshold, Generate at least one backup instruction that has the same target as the main scheduling instruction but is executed by a different entity.

6. The method for dynamic resource scheduling of emergency medical teams based on multimodal data-driven collaboration according to claim 3, characterized in that, After the step of executing multimodal constraint-driven conflict resolution decision-making when the collaborative robustness index is lower than a preset index threshold, the method further includes: Real-time monitoring of the aforementioned collaborative robustness indicators; When the collaborative robustness index is continuously lower than the preset threshold and the resource availability is close to zero, a resource depletion alarm and cross-domain collaboration request are broadcast to the preset external emergency network. The request includes the type of necessary resources, the patient's rate of physiological deterioration, and the geographical location.

7. The method for dynamic resource scheduling of emergency medical teams based on multimodal data-driven collaboration according to claim 5, characterized in that, The method further includes: When new patients cause resource overload, the topology dependency is reset in response to the addition of high-risk physiological signals in the patient's physiological time-series data stream. Based on the priority order of the constraints, existing resources are reallocated.

8. A dynamic resource scheduling system for emergency medical teams based on multimodal data-driven collaboration, characterized in that, include: The data acquisition module is used to collect patient physiological time-series data, team behavioral interaction data, and resource status data in emergency scenarios, forming a heterogeneous data stream; The deconstruction module is used to deconstruct the topological dependencies of the team collaboration link based on the heterogeneous data stream; The calculation module is used to calculate the collaborative robustness index based on the topological dependencies and heterogeneous data streams; The execution module is used to execute a multimodal constraint-driven conflict resolution decision when the collaborative robustness index is lower than a preset index threshold. The monitoring module is used to monitor the changes in collaboration efficiency after resource allocation, and, in conjunction with preset optimization targets, correct the weight coefficients of the collaboration robustness index and the preset index thresholds.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images