A medical resource scheduling and risk early warning method, system and electronic device

Abstract

The application discloses a medical resource scheduling and risk early warning method and system and electronic equipment. The method comprises the following steps: acquiring a clinical event record and mapping to a multi-dimensional organ state space; constructing a double-path representation atlas of a patient in the multi-dimensional organ state space in parallel; updating a state transition edge and generating a continuous dynamic risk load sequence by using a clinical event trigger mechanism; monitoring a dynamic evolution direction coupling degree based on time delay mutual information; and generating an active medical resource scheduling suggestion and outputting expert recommendation information when a non-synchronous state offset is captured. The application can realize rolling early warning of risks and active identification of medical intervention conflicts and resource scheduling, and improve the diagnosis and treatment coordination efficiency and resource allocation accuracy in a multi-disease coexistence environment.

Description

Technical Field

[0001] This application relates to the fields of medical information data processing and clinical decision support technology, and in particular to a method, system and electronic device for medical resource scheduling and risk warning. Background Technology

[0002] With the increasing sophistication of modern medical management, accurately assessing and proactively allocating resources for the complex disease progression caused by coexisting multiple illnesses in hospitalized patients has become a core challenge in improving the quality of clinical management. Existing critical care risk assessment systems are primarily based on high-frequency sampling mechanisms within fixed time windows in intensive care environments. These methods rely on static weighted logic and typically only assess organ function status within a single, short timeframe.

[0003] In long-term clinical settings within general wards, clinical data exhibits characteristics such as uneven sampling frequencies and long disease durations, making it difficult to effectively characterize the dynamic evolutionary processes across systems and organs. Traditional static scoring systems cannot quantitatively represent the cumulative impact of acute functional fluctuations and long-term chronic burdens, and struggle to capture clinical decision conflicts arising during the dynamic evolution of multiple organ systems. This often results in the coordination of multidisciplinary collaborative treatment and the allocation of critical rescue resources lagging behind the actual high-risk mutation events occurring in patients, leading to significant early warning blind spots and the risk of resource misallocation. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a method, system, and electronic device for medical resource scheduling and risk early warning. Through dual-path organ state topology construction, event-driven dynamic state transition measurement, and cross-organ coupling analysis, it achieves rolling risk early warning and proactive identification and resource scheduling of medical intervention conflicts.

[0005] In a first aspect, embodiments of this application provide a method for medical resource scheduling and risk warning, including:

[0006] Acquire serialized clinical event records within the clinical diagnosis and treatment cycle, and map the clinical event records to a preset multi-dimensional organ state space based on preset mapping rules;

[0007] In the multi-dimensional organ state space, a dual-path representation map of the patient is constructed in parallel. The dual-path representation map includes a chronic background path constructed using historical and underlying disease characteristics and an acute decompensation evolution path updated using the current acute outbreak.

[0008] The state transition edges are updated using a clinical event triggering mechanism. The instantaneous weights and time change rates of the state transition edges are extracted, and historical shocks are dynamically weighted and aggregated using a preset time decay function to generate a continuous dynamic risk load sequence.

[0009] Based on the time-delay mutual information of different organ dimensions on the dynamic risk load sequence, monitor the dynamic evolution direction coupling degree of cross-organ system;

[0010] When an asynchronous state deviation phenomenon that meets the threshold condition is detected, it is determined that the preset decision conflict condition is met, and proactive medical resource scheduling suggestions are generated and multidisciplinary consultation expert node recommendation information is output.

[0011] Optionally, the multi-dimensional organ state space consists of multiple preset organ-dimensional state vectors; acquiring serialized clinical event records within the clinical treatment cycle, and mapping the clinical event records to the preset multi-dimensional organ state space based on preset mapping rules, including:

[0012] Extract the structured indicator sequences and unstructured text features from the clinical event records;

[0013] The structured index sequence and the unstructured text features are transformed into initial state vectors for each preset organ dimension.

[0014] Optionally, the dual-path representation map consists of multiple nodes representing different physiological subsystems and the state transition edges; the dual-path representation map of the patient is constructed in parallel in the multi-dimensional organ state space, including:

[0015] Extract pre-hospitalization or underlying disease data and map it to the basic potential energy value of nodes in the chronic background path;

[0016] By capturing abnormal physiological index fluctuations or drug intervention events during the clinical diagnosis and treatment cycle, state transition edges between connecting nodes in the acute decompensation evolution path are constructed.

[0017] Optionally, the node's basic potential energy value is configured to reflect the organ's long-term chronic disease burden compensation margin; the state transition edge is configured to represent the direction of influence transmission between physiological systems under acute decompensation.

[0018] Optionally, the state transition edge is updated using a clinical event-triggered mechanism, including:

[0019] Monitor data generation events in the clinical information system and asynchronously wake up the update operation for the dual-path representation map.

[0020] Optionally, the instantaneous weights and time rates of change of the state transition edges are extracted, including:

[0021] Extract the rate of change of the first and second derivatives of the state vector with respect to time;

[0022] The rate of change of the first derivative and the rate of change of the second derivative are combined to form the instantaneous kinetic energy characterizing the evolving impact, which serves as the instantaneous weight and time-varying rate feature. The specific formula for synthesizing the instantaneous kinetic energy is as follows:

[0023] E_instant=W×(α·|v|+β·|a|)

[0024] Where W is the basic risk weight of the state transition edge, calculated as W(A→B)=P(death|A→B)-P(death|no A→B); v is the rate of change of the first derivative (deterioration rate); a is the rate of change of the second derivative (deterioration acceleration); α and β are preset weight coefficients, both ranging from [0,1], and α+β=1, with the default values ​​of α=0.6 and β=0.4, to reflect that the contribution of velocity change to instantaneous kinetic energy is greater than that of acceleration change.

[0025] Optionally, the rate of change of the first derivative is configured as a parameter characterizing the rate of deterioration of organ function indicators; the rate of change of the second derivative is configured as an acceleration parameter characterizing the deterioration trend; and a continuous dynamic risk load sequence is generated by dynamically weighting and aggregating historical shocks through a preset time decay function, including:

[0026] The instantaneous kinetic energy extracted at the current moment is weighted and aggregated with the historical instantaneous kinetic energy accumulated at previous historical moments to construct the cumulative kinetic energy function K(t);

[0027] A half-life decay parameter is introduced into the cumulative kinetic energy function K(t), and the historical instantaneous kinetic energy with a longer historical time interval from the current time is configured to have a smaller weighting value.

[0028] The risk assessment numerical sequence at each moment is continuously output based on the cumulative kinetic energy function K(t) to form the dynamic risk load sequence.

[0029] Optionally, based on the temporal delay mutual information of different organ dimensions on the dynamic risk load sequence, the dynamic evolutionary direction coupling degree of the cross-organ system is monitored, including:

[0030] Calculate the mutual information or covariance values ​​of the dynamic risk load sequence corresponding to the first organ dimension and the dynamic risk load sequence corresponding to the second organ dimension under different time delay windows.

[0031] Optionally, monitoring the dynamic evolutionary directional coupling degree across organ systems also includes:

[0032] By comparing the mutual information values ​​under all the different time delay windows, the target time delay window corresponding to the maximum mutual information value is extracted, and the hysteresis response relationship parameters between the first organ dimension and the second organ dimension system are determined.

[0033] Optionally, monitoring the dynamic evolutionary directional coupling degree across organ systems also includes:

[0034] Construct an evolutionary tension matrix consisting of nodes of multiple different organ dimensions and inter-system connection edges representing evolutionary tension;

[0035] Obtain the first slope parameter of the dynamic risk load sequence of the first organ dimension;

[0036] Obtain the second slope parameter of the node base potential value of the second organ dimension in the chronic background path;

[0037] When the first slope parameter is greater than the first preset slope threshold and the second slope parameter is less than the second preset slope threshold, and the mutual information value between the first organ dimension and the second organ dimension in the current time window is lower than the preset coupling threshold, it is determined that there is a cross-organ physiological compensation failure state, a tension feature vector representing the deviation of organ evolution trajectory is generated, and the tension feature vector is passed to the subsequent conflict determination stage.

[0038] Optionally, when an asynchronous state shift that meets a threshold condition is detected, a preset decision conflict condition is determined, and proactive medical resource scheduling suggestions are generated and multidisciplinary consultation expert node recommendation information is output, including:

[0039] Calculate the deviation of the dynamic evolution directional coupling degree from the initial baseline state parameters;

[0040] If the deviation value is greater than the preset conflict determination threshold, an activation signal is generated to trigger resource allocation operation.

[0041] Optionally, generating proactive medical resource scheduling suggestions and outputting multidisciplinary consultation expert node recommendation information also includes:

[0042] In response to the activation signal, based on the specific organ dimension combination in which the asynchronous state shift phenomenon occurs, a preset medical resource database is matched, and a multidisciplinary consultation notification message containing a target rescue equipment allocation suggestion and a list of experts from related departments is output.

[0043] Optionally, the multi-dimensional organ state space includes a first designated organ dimension and a second designated organ dimension; the output includes a multidisciplinary consultation notification message containing a target rescue equipment allocation suggestion and a list of experts from related departments, including:

[0044] When the index offset of the first specified organ dimension reaches a preset improvement value, and the index offset of the second specified organ dimension reaches a preset deterioration value, thus triggering the asynchronous state offset phenomenon, the life support for the first specified organ dimension is written into the multidisciplinary consultation notification message.

[0045] The system provides assessment recommendations for adjusting equipment operating parameters, generates a reference work order for scheduling personnel to the relevant specialists in the second designated organ dimension, and sends the multidisciplinary consultation notification message containing the target resuscitation equipment allocation recommendations and the list of specialists to the hospital's collaborative management terminal system.

[0046] Secondly, embodiments of this application provide a medical resource scheduling and risk warning system, including:

[0047] The spatial mapping module is configured to acquire serialized clinical event records within the clinical diagnosis and treatment cycle, and map the clinical event records to a preset multi-dimensional organ state space based on preset mapping rules.

[0048] A dual-path construction module is configured to construct a dual-path representation map of a patient in parallel in the multi-dimensional organ state space, wherein the dual-path representation map includes a chronic background path constructed using historical and underlying disease characteristics and an acute decompensation evolution path updated using current acute emergencies.

[0049] The dynamic aggregation module is configured to update the state transition edges using a clinical event triggering mechanism, extract the instantaneous weights and time change rates of the state transition edges, and perform dynamic weighted aggregation of historical shocks using a preset time decay function to generate a continuous dynamic risk load sequence.

[0050] The mutual information monitoring module is configured to monitor the dynamic evolution direction coupling degree across organ systems based on the time-delay mutual information of different organ dimensions on the dynamic risk load sequence.

[0051] The scheduling trigger module is configured to, when it detects asynchronous state deviations that meet threshold conditions, determine that preset decision conflict conditions are met, generate proactive medical resource scheduling suggestions, and output multidisciplinary consultation expert node recommendation information.

[0052] Thirdly, embodiments of this application provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the aforementioned medical resource scheduling and risk warning method.

[0053] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned medical resource scheduling and risk warning method.

[0054] The present invention can bring the following beneficial effects:

[0055] By introducing chronic paths representing historical background potential energy and acute decompensated paths representing event-driven kinetic energy in parallel within a unified topological framework, this approach abandons the traditional static time window sampling model. Through time-delay mutual information and physical-dynamic weighted calculations of feature evolution, this scheme can identify potential decision-making conflicts arising from cross-departmental treatment. This allows for the early generation of resource scheduling recommendations, including expert teams and equipment support, before a patient's condition slides into an irreversible, critical high-risk state. This improves the efficiency of diagnosis and treatment coordination and the accuracy of resource allocation in modern hospitals in environments with multiple coexisting diseases. Attached Figure Description

[0056] Figure 1 This is a flowchart of the medical resource scheduling and risk warning method provided in the embodiments of the present invention.

[0057] Figure 2 This is a logical principle diagram of the multi-dimensional organ state space and dual-path representation map provided in the embodiments of the present invention.

[0058] Figure 3 This is a logical structure block diagram of the medical resource scheduling and risk warning system provided in the embodiments of the present invention.

[0059] Figure 4 This is a hardware structure block diagram of the electronic device provided in the embodiments of the present invention.

[0060] Explanation of reference numerals in the attached figures:

[0061] 300 Medical Resource Scheduling and Risk Early Warning System, 301 Spatial Mapping Module, 302 Dual Path Construction Module, 303 Dynamic Aggregation Module, 304 Mutual Information Monitoring Module, 305 Scheduling Trigger Module, 400 Electronic Equipment, 401 Processor, 402 Memory, 403 Communication Interface, 404 Internal System Communication Bus. Detailed Implementation

[0062] To make the technical solution, purpose, and beneficial effects of this application clearer, the application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0063] Example 1

[0064] This embodiment provides a method for medical resource scheduling and risk early warning. This method primarily relies on a computer system or backend control server with data processing and communication capabilities.

[0065] like Figure 1 As shown, the medical resource scheduling and risk warning method provided in this application mainly includes the following steps S101 to S105:

[0066] Step S101: Obtain serialized clinical event records within the clinical diagnosis and treatment cycle, and map the clinical event records to a preset multi-dimensional organ state space based on preset mapping rules.

[0067] In its implementation, the system connects to the heterogeneous clinical information systems of medical institutions through its built-in data acquisition and conversion gateway module. The data sources collected include, but are not limited to, serialized records such as body fluid laboratory tests, vasoactive drug prescription records, continuous vital sign monitoring data, and nursing records. The gateway module asynchronously acquires these discrete clinical events and normalizes the heterogeneous data. To build a high-quality trajectory model foundation, the processor performs cleaning and filtering operations on the raw clinical event records. When acquiring historical data, the system has a built-in basic queue exclusion logic configured to remove records that do not conform to the characteristics of long-term evolution analysis in general wards. Specifically, the system's exclusion criteria include at least: same-day surgical admission records, emergency admission records, ICU transfer records within the previous 72 hours, direct ICU admission records, and postoperative ICU transfer records. Furthermore, the system sets a pre-outcome cutoff window at the end of the sequence to silence and exclude terminal intervention data occurring within this cutoff window, preventing drastic final resuscitation information from contaminating the smooth evolution trajectory. It should be noted that setting the pre-outcome cutoff window is a preferred data cleansing step in this scheme, but it is not a necessary technical feature for realizing the core technical concept of this application. It should also be noted that although this embodiment uses the Medical Information Mart for Intensive Care IV (MIMIC-IV) database as an example of historical model training or baseline generation, the historical data sources of this application are not limited to this, and those skilled in the art can use electronic health record databases of other regional medical centers as alternatives. By employing the above-mentioned multi-source data fusion and rule-based filtering logic, highly cleaned sequential clinical events can be obtained, providing data support for subsequent dynamic models and avoiding interference from short-term extreme noise on long-term evolutionary characteristics.

[0068] After acquiring the cleaned clinical event records, the processor extracts structured indicator sequences and unstructured text features from the records, and transforms these into initial state vectors for various preset organ dimensions. These state vectors collectively constitute a multi-dimensional organ state space. Specifically, the preset organ dimensions include multiple core system dimensions such as respiratory function, circulatory function, renal function, liver function, coagulation function, neural function, and metabolic function.

[0069] Specifically, the preset organ dimension includes seven subsystems. Among them, renal function, coagulation function, liver function, respiratory function, and metabolic function are scored based on thresholds of laboratory test indicators (similar to the Sequential Organ Failure Assessment (SOFA) score); cardiovascular function and neurological function are classified into two categories based on whether a specific drug prescription event has occurred.

[0070] To construct a directed state transition path, the system dynamically evaluates and determines the dominant organ state at the current moment based on the perturbation values ​​of each preset organ dimension extracted in real time. Specifically, the system calculates the rate of change of the state vector of each preset organ dimension relative to the chronic background path baseline as the perturbation value parameter of the corresponding dimension, and selects the organ dimension with the largest perturbation value parameter as the current dominant node. Further, when multiple preset organ dimensions are found to have the same perturbation value parameter or a difference below a preset tolerance range, the system is configured to respond to this state by determining a unique dominant node according to a preset deterministic disambiguation rule. The deterministic disambiguation rule can be implemented based on preset organ dimension weights or a fixed priority order, and the deterministic disambiguation rule is configured to maintain the uniqueness of trajectory allocation without representing the actual biological advantage order among the preset organ dimensions. If the perturbation value parameters of each dimension do not reach the trigger threshold, then this moment is recorded as the baseline state.

[0071] Step S102: Construct a dual-path representation map of the patient in parallel in the multi-dimensional organ state space. The dual-path representation map includes a chronic background path constructed using historical and underlying disease characteristics and an acute decompensation evolution path updated using current acute emergencies.

[0072] Combination Figure 2The diagram illustrates the logical principle of the multi-dimensional organ state space and dual-path representation map. The multi-dimensional organ state space contains multiple organ nodes reflecting different physiological subsystems. This step topologically defines the evolutionary trajectory at the structural level, quantifying both the static cumulative burden and the dynamic direction of acute deterioration. The system extracts pre-hospitalization or underlying disease data and maps it to the basic potential energy values ​​of nodes in the chronic background path. The quantification method for the basic potential energy values ​​of nodes is as follows: For each organ dimension, long-term chronic disease-related indicators are extracted from the pre-hospitalization or underlying disease data, including but not limited to the duration of chronic disease history, baseline laboratory values ​​(such as baseline creatinine and basal bilirubin), and past medication history. A preset scoring rule or regression model is used to map these indicators to values ​​between 0 and 1, with higher values ​​indicating heavier long-term burden and lower compensatory margin. For example, for the kidney function node, a baseline creatinine value 1.5 times higher than the upper limit of normal is scored as 0.3, 2 times higher as 0.6, and 3 times higher as 0.9. These values ​​are then weighted by the duration of chronic kidney disease history to obtain the final potential energy value. Simultaneously, based on the current acute emergency, the system constructs state transition edges between connecting nodes in the acute decompensation evolution path. These state transition edges are configured to represent the direction of influence transmission between physiological systems under acute decompensation conditions.

[0073] To quantitatively describe the topological tension between chronic base networks and acute transition networks, the processor calculates the state transition changes representing the tendency to deteriorate based on state transition probabilities when constructing the aforementioned two-layer path transition map. Specifically, to implement the probability deviation evaluation logic, the system performs the calculation of the state transition changes representing the tendency to deteriorate. In this embodiment, the following formula is preferably used for calculation:

[0074]

[0075] Among them, variables This represents the change in state transition, used to measure the enhancement degree of a specific cross-organ influence pathway. Specifically, P(acute) and P(chronic) are calculated as follows: based on a large number of patient samples in a historical medical record database (such as MIMIC-IV), the frequency of state transitions from organ node A to organ node B within a preset time window (such as 24 hours) is statistically analyzed, and this frequency is divided by the total number of observed samples as an estimate of the probability. P(chronic) is statistically analyzed using chronic background pathways constructed from pre-hospitalization or underlying disease data, reflecting the metastatic tendency under long-term steady state; P(acute) is statistically analyzed using acute decompensated evolution pathways constructed from data within the current clinical treatment cycle, reflecting the metastatic tendency in the acute phase. If there are insufficient samples in the historical database, a Bayesian smoothing method is used to correct the probability estimates. If the calculated... In the logical graph, this path is identified as the direction of acute enhancement evolution, reflecting a local decompensation risk that deviates from the previous chronic development pattern at the current stage of clinical diagnosis and treatment. By employing a dual-path parallel spatial mapping and difference calculation structure, it is possible to physically isolate and analyze "basic structural defects" and "sudden event impacts" from complex comorbid information, thereby improving the resolution of clinical complexity features. It should be noted that the above-mentioned difference calculation formula for quantifying state evolution deviation is only one implementation form. Those skilled in the art can also use other mathematical models that can characterize distribution divergence to achieve the same purpose of capturing abnormal path deviations.

[0076] Step S103: The state transition edge is updated using a clinical event triggering mechanism. The instantaneous weight and time change rate of the state transition edge are extracted. The historical impacts are dynamically weighted and aggregated using a preset time decay function to generate a continuous dynamic risk load sequence.

[0077] This step breaks through the traditional fixed-time sampling mechanism. The system monitors data generation events in the clinical information system, and once a new test report or medication instruction is received, it asynchronously wakes up the update operation for the dual-path characterization map. This event-driven logic can accommodate the non-uniform distribution of data generated by medical operations in general wards.

[0078] When updating state transition edges, the system first calculates the base risk weight (W) of the transition edge based on the probability of adverse outcomes. For any state transition edge (organ A pointing to organ B), the formula for calculating its base risk weight W is:

[0079] W(A→B) = P(Death|A→B) - P(Death|No A→B)

[0080] That is: the probability of death of patients who have experienced organ metastasis minus the probability of death of patients who have not experienced metastasis.

[0081] After establishing the basic risk weight W for anchoring adverse outcomes, the system further extracts the first derivative (deterioration rate) and second derivative (deterioration acceleration) of this state vector with respect to time, and combines them with the basic risk weight W to synthesize the 'instantaneous kinetic energy' characterizing the evolving shock. Subsequently, the kinetic energy of historical shocks is weighted and aggregated using a time decay function to generate a continuous dynamic risk load sequence.

[0082] When a state transition edge is triggered for update, the processor extracts the first-order derivative rate of change (velocity v) and the second-order derivative rate of change (acceleration a) of the organ state vector represented by that transition edge with respect to time. The first-order derivative rate of change is configured as a parameter representing the rate of deterioration of organ function indicators; the second-order derivative rate of change is configured as an acceleration parameter representing the deterioration trend. Using the instantaneous intensity change rate dW / dt as a reference, the system synthesizes the first and second-order rates of change into an instantaneous kinetic energy representing the evolutionary impact, which serves as the instantaneous weight feature of the edge on the time profile. Furthermore, to filter out meaningless minor perturbations, the system configures a high-risk transition edge determination condition. Specifically, the system marks an edge as a high-risk structural evolution path and includes it in subsequent intensive computation stages only if the instantaneous weight of the currently generated transition edge satisfies W > 0.05. In order to connect the scattered events in time, the processor adopts the event index calculation method, which defines the total number of valid observation events corresponding to the current analysis time as the event index + 1. It only uses the current and previous historical slice data for dynamic inference, and strictly prohibits the use of future event numbers to participate in the current weighting, so as to ensure the forward applicability of the model.

[0083] Based on this, in order to transform the discrete state transition process into a continuous and computable dynamic physical evolution process, the system fuses the instantaneous kinetic energy extracted at the current moment with the historical instantaneous kinetic energy accumulated at previous historical moments, and introduces a physical dynamics operator to construct a cumulative kinetic energy function. To achieve energy aggregation logic with historical memory decay characteristics, specifically, the system performs calculations on a continuous sequence of dynamic risk loads. In this embodiment, the following function model with a time decay term is preferably adopted:

[0084]

[0085] Wherein, the function K(t) represents the system's cumulative kinetic energy or continuous kinetic risk load sequence at the current observation time t; the variable W i The variable t represents the instantaneous weight extracted and assigned when the i-th state transition event in history is triggered; i Represents the absolute timestamp of the i-th historical event recorded by the system; variable This represents the time decay parameter. In the function above, an exponential half-life mechanism is introduced, configured as the historical time interval (tt) from the current time. i The longer the historical instantaneous kinetic energy, the smaller its weighting value, in order to realistically simulate the slow recovery and dissipation process of the human body in response to past stress damage. In this embodiment, for a typical intensive care monitoring environment, the time decay parameter is configured to make the risk influence exhibit a 24-hour half-life characteristic, i.e., the parameter λ is approximately 0.029 (in hours). －1In other alternative implementations, for more rapidly evolving acute cardiovascular wards, the half-life can be shortened to a decay constant corresponding to 12 hours. By employing a function model based on physical decay aggregation to calculate the continuous dynamic risk load sequence, the traditional static slice-based scoring can be abandoned, achieving a nonlinear measure of the continuous evolutionary kinetic energy accumulation. This allows the system to accurately reconstruct the continuous dynamic background of organ compression in data-sparse environments.

[0086] In addition, to characterize the degree of recent risk clustering, the system also extracts K. accel The feature is calculated by summing the instantaneous weights of all triggered transition edges within a preset time window and taking the time average.

[0087] Step S104: Based on the time delay mutual information of different organ dimensions on the dynamic risk load sequence, monitor the dynamic evolution direction coupling degree of the cross-organ system.

[0088] The system goes beyond isolated organ values, constructing an evolutionary tension matrix composed of nodes across multiple organ dimensions and inter-system connections representing evolutionary tension. The processor calculates the mutual information or covariance values ​​of the kinetic risk load sequence corresponding to the first organ dimension and the kinetic risk load sequence corresponding to the second organ dimension under different time delay windows. By comparing the mutual information values ​​under all different time delay windows, the target time delay window corresponding to the maximum mutual information value is extracted, thereby determining the hysteresis response parameter between the first and second organ dimensions. Specifically, the system extracts the duration of the target time delay window as the hysteresis response parameter, which characterizes the physical response delay of the second organ dimension to functional fluctuations in the first organ dimension. By quantifying the "abnormal resonance" in evolutionary dynamics of different systems (such as the circulatory and respiratory systems), this hysteresis response parameter can be directly used to guide frontline physicians in predicting the endpoint of disease deterioration and reserving corresponding treatment windows for clinical intervention.

[0089] Based on the calculation results of mutual information, the system further obtains the first slope parameter of the dynamic risk load sequence in the first organ dimension and the second slope parameter of the node basic potential energy value in the chronic background path in the second organ dimension. When the system detects that the first slope parameter is greater than the first preset slope threshold and the second slope parameter is less than the second preset slope threshold, and the mutual information value of the two in the current time window is lower than the preset coupling threshold, it determines that there is a cross-organ physiological compensation failure state. The system then generates a tension feature vector representing the deviation of the organ evolution trajectory and passes the tension feature vector to the subsequent conflict determination stage. Through the above method based on the composite comparison of mutual information and dual path parameters, the system can detect the hidden "abnormal evolutionary resonance" phenomenon in a purely mathematical way, thereby abstracting the dynamic disconnection characteristics of organ function into an alarm signal that can be recognized by the machine.

[0090] Step S105: When an asynchronous state shift phenomenon that meets the threshold condition is captured, it is determined that the preset decision conflict condition is met, and proactive medical resource scheduling suggestions are generated and multidisciplinary consultation expert node recommendation information is output.

[0091] Based on the tension feature vector derived from the above steps, the processor calculates the deviation of the dynamic evolution directional coupling degree from the initial baseline state parameters. If the deviation value is greater than a preset conflict determination threshold, the system determines that the current evolution is at a "medical decision conflict point" and generates an activation signal to trigger resource allocation operations.

[0092] To achieve proactive and collaborative resource management, the system responds to the activation signal and automatically accesses and matches a preset medical resource database based on the specific organ dimension combination where asynchronous state deviation occurs. Based on the detected deviation and organ combination, the system outputs a multidisciplinary consultation notification message containing a target resuscitation equipment allocation suggestion and a list of relevant departmental experts. Specifically, when the system detects that the index deviation of the first specified organ dimension reaches a preset improvement value, and the index deviation of the second specified organ dimension reaches a preset deterioration value, thus triggering the asynchronous state deviation, the system writes an equipment allocation suggestion into the multidisciplinary consultation notification message. This allocation suggestion may include: an assessment suggestion for reducing the operating parameters of life support equipment for the first specified organ dimension, and the generation of a manpower scheduling reference work order instructing intervention of the departmental experts corresponding to the second specified organ dimension. Through the configured front-end display and alarm push module, the aforementioned multidisciplinary consultation notification message containing the allocation suggestion and list is sent to the hospital collaborative management terminal system or related mobile smart devices.

[0093] Furthermore, in an optional implementation, based on the extracted temporal dynamics features and evolutionary state space parameters, this scheme can further construct a multi-timescale unplanned outcome early warning model using machine learning algorithms. Specifically, the processor inputs the generated risk load sequence and cross-system mutual information features into a pre-trained Light Gradient Boosting Machine (LightGBM) classification tree prediction model, performing rolling prediction evaluations within different prediction time windows such as 24 hours, 48 ​​hours, and 72 hours. The prediction results output by this model represent the probability that the patient's evolution trajectory converges to high-resource-consuming events such as unplanned transfer to the intensive care unit. By employing a gradient boosting tree classification model with feature fusion, it is possible to quickly perform nonlinear classification boundary delineation for high-dimensional kinetic energy combinations generated by high-frequency and low-frequency clinical events, further providing pre-set resource scheduling reference windows with different urgency levels. It should be noted that the LightGBM classification model in this embodiment is only a preferred classification decision-making method. Those skilled in the art can also use other classification algorithm systems such as LSTM or MLP, as long as they can achieve the inference and prediction of the probability of occurrence based on the input time-decaying kinetic energy features.

[0094] This embodiment solves the problem of continuous evaluation under sparse clinical data by decomposing the chronic path representing the basic potential energy and the acute kinetic energy path driving the evolution in a unified multidimensional graph, and replacing the traditional timing detection method with event-driven and physical kinetic energy decay functions. It also establishes an active scheduling bridge from micro data fluctuations to macro medical equipment and expert resources automatic scheduling by quantitatively determining the evolutionary tension and decision-making conflicts of different systems through mutual information operation.

[0095] Example 2

[0096] Based on the method concept provided in Embodiment 1 above, this application also provides a medical resource scheduling and risk early warning system. This system is typically deployed as a back-end service in a hospital information hub or as a cloud-based decision processing service.

[0097] like Figure 3 As shown, the medical resource scheduling and risk early warning system 300 mainly includes a spatial mapping module 301, a dual-path construction module 302, a dynamic aggregation module 303, a mutual information monitoring module 304, and a scheduling triggering module 305. Among them:

[0098] The spatial mapping module 301 is configured to access various peripheral data systems, acquire serialized clinical event records within the clinical diagnosis and treatment cycle, perform preprocessing such as cleaning and truncation, and then convert structured and unstructured features into multi-dimensional vectors based on preset mapping rules, mapping them to a preset multi-dimensional organ state space.

[0099] The dual-path construction module 302 is configured to construct a dual-path representation map of the patient in parallel within the multi-dimensional organ state space. It establishes a chronic background path network reflecting long-term reserves using historical and underlying disease characteristics, and updates the directional acute decompensation evolution path using current acute emergencies.

[0100] The dynamics aggregation module 303 is configured to abandon the static timed window mechanism and adopt an asynchronous triggering mechanism that is woken up by listening for events to update the state transition edges. This module performs mathematical operations to extract the differential rate of change of the transition vector and synthesize instantaneous weights. It has built-in time decay weighted algorithm calculation logic, based on, for example,

[0101]

[0102] Formulas with half-life forgetting properties are used to continuously aggregate and integrate high-risk shock edges that have occurred in history, and output a dynamic risk load sequence.

[0103] The mutual information monitoring module 304 is configured to acquire dynamic sequences of multiple organ modules and calculate mutual information or covariance under different time delay windows. By comparing slope changes, a tension feature matrix is ​​constructed to monitor in real time whether abnormal evolutionary coupling degrees have occurred between cross-organ systems, indicating a disruption of compensatory balance.

[0104] The scheduling trigger module 305, acting as the execution output, is configured to trigger a preset departmental decision-making conflict red line when it receives a non-synchronous offset phenomenon characteristic exceeding the conflict threshold from the mutual information monitoring module 304. It further performs association mapping from a pre-set medical expert and equipment database to generate adjustment suggestions for specific life support equipment and a list of recommended multidisciplinary experts with specific professional backgrounds, and packages these into a proactive medical resource scheduling message to be pushed to the management terminal.

[0105] To ensure the complete data closed-loop operation of the aforementioned virtual processing modules, the system may also include modules not present in the data processing module. Figure 3 The auxiliary modules detailed in the diagram include, for example, a data acquisition and conversion gateway module to ensure seamless handshakes with external systems, and a front-end display and alarm push module for presenting complex path topology networks on healthcare workstations. These modules work together to automate the entire process from data abstraction to proactive resource intervention.

[0106] Example 3

[0107] This application also provides an electronic device. For example... Figure 4 As shown, the hardware structure block diagram of the electronic device 400 mainly includes: processor 401, memory 402 and communication interface 403.

[0108] The processor 401, memory 402, and communication interface 403 can be electrically connected and exchange data via an internal system communication bus 404. The memory 402, as a non-volatile or volatile computer-readable storage medium, is used to store system files of various formats, program code, and intermediate state data generated during the execution of the aforementioned mapping and dynamic integration. The memory 402 stores a computer program, which may specifically be the code instructions for the various virtual functional modules described in Embodiment 2 above.

[0109] The processor 401 can be a general-purpose central processing unit, a microprocessor, an application-specific integrated circuit, a graphics processing unit, a digital signal processor, or a dedicated AI computing chip configured to accelerate the execution of machine learning tree models. When the electronic device 400 is running, the processor 401 calls and executes the computer program from the memory 402, thereby implementing the specific steps in the medical resource scheduling and risk warning method detailed in Embodiment 1, such as calculating exponential decay characteristics, comparing slope parameters, and generating expert recommendations. The communication interface 403 is responsible for sending the finally generated instruction data packet to the external intensive care equipment terminal or the attending physician's portable electronic device via Ethernet or a wireless local area network.

[0110] Furthermore, embodiments of this application also provide a computer-readable storage medium. This computer-readable storage medium stores a computer program or a set of instructions. When the computer program is executed by the processor of the aforementioned computing device, it can implement the steps described in any of the above embodiments. The computer-readable storage medium can take any suitable physical form, such as, but not limited to, various media capable of persistently or temporarily storing program code, such as: USB flash drive, portable hard drive, read-only memory, random access memory, magnetic disk, or optical disk.

[0111] It should be noted that, in this document, relational terms such as “first” and “second” are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus 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, method, article, or apparatus.

[0112] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the principles and technical concepts of this application should be included within the protection scope of this application.

Claims

1. A method for medical resource allocation and risk early warning, characterized in that, include: Acquire serialized clinical event records within the clinical diagnosis and treatment cycle, and map the clinical event records to a preset multi-dimensional organ state space based on preset mapping rules; In the multi-dimensional organ state space, a dual-path representation map of the patient is constructed in parallel. The dual-path representation map includes a chronic background path constructed using historical and underlying disease characteristics and an acute decompensation evolution path updated using the current acute outbreak. The state transition edges are updated using a clinical event triggering mechanism. The instantaneous weights and time change rates of the state transition edges are extracted, and historical shocks are dynamically weighted and aggregated using a preset time decay function to generate a continuous dynamic risk load sequence. Based on the time-delay mutual information of different organ dimensions on the dynamic risk load sequence, monitor the dynamic evolution direction coupling degree of cross-organ system; When an asynchronous state deviation phenomenon that meets the threshold condition is detected, it is determined that the preset decision conflict condition is met, and proactive medical resource scheduling suggestions are generated and multidisciplinary consultation expert node recommendation information is output.

2. The method as described in claim 1, characterized in that, The multi-dimensional organ state space consists of multiple preset organ-dimensional state vectors; acquiring serialized clinical event records within the clinical treatment cycle, and mapping the clinical event records to the preset multi-dimensional organ state space based on preset mapping rules, including: Extract the structured indicator sequences and unstructured text features from the clinical event records; The structured index sequence and the unstructured text features are transformed into initial state vectors for each preset organ dimension.

3. The method as described in claim 1, characterized in that, The dual-path representation map consists of multiple nodes representing different physiological subsystems and the state transition edges; the patient's dual-path representation map is constructed in parallel in the multi-dimensional organ state space, including: Extract pre-hospitalization or underlying disease data and map it to the basic potential energy value of nodes in the chronic background path; Based on the current acute emergency, construct the state transition edges between the connecting nodes in the acute decompensation evolution path.

4. The method as described in claim 1, characterized in that, The state transition edges are updated using a clinical event-triggered mechanism, including: Monitor data generation events in the clinical information system and asynchronously wake up the update operation for the dual-path representation map.

5. The method as described in claim 4, characterized in that, Extracting the instantaneous weights and time-varying rates of change of the state transition edges includes: Extract the rate of change of the first and second derivatives of the organ state vector represented by the state transition edge with respect to time; The rate of change of the first derivative and the rate of change of the second derivative are combined to form the instantaneous kinetic energy characterizing the evolutionary shock, which is used as the instantaneous weight and time variation characteristic.

6. The method as described in claim 5, characterized in that, The rate of change of the first derivative is configured as a parameter characterizing the rate of deterioration of organ function indicators; the rate of change of the second derivative is configured as an acceleration parameter characterizing the deterioration trend. Historical shocks are dynamically weighted and aggregated using a preset time decay function to generate a continuous dynamic risk load sequence, including: The instantaneous kinetic energy extracted at the current moment is weighted and aggregated with the historical instantaneous kinetic energy accumulated at previous historical moments to construct the cumulative kinetic energy function K(t); A half-life decay parameter is introduced into the cumulative kinetic energy function K(t), and the historical instantaneous kinetic energy with a longer historical time interval from the current time is configured to have a smaller weighting value. The risk assessment numerical sequence at each moment is continuously output based on the cumulative kinetic energy function K(t) to form the dynamic risk load sequence.

7. The method as described in claim 1, characterized in that, Based on the temporal delay mutual information of different organ dimensions on the dynamic risk load sequence, the dynamic evolutionary direction coupling degree of cross-organ systems is monitored, including: Calculate the mutual information or covariance values ​​of the dynamic risk load sequence corresponding to the first organ dimension and the dynamic risk load sequence corresponding to the second organ dimension under different time delay windows.

8. The method as described in claim 7, characterized in that, Monitoring the dynamic evolutionary directional coupling degree across organ systems also includes: By comparing the mutual information values ​​under all the different time delay windows, the target time delay window corresponding to the maximum mutual information value is extracted, and the hysteresis response relationship parameters between the first organ dimension and the second organ dimension system are determined.

9. The method as described in claim 7, characterized in that, Monitoring the dynamic evolutionary directional coupling degree across organ systems also includes: Construct an evolutionary tension matrix consisting of nodes of multiple different organ dimensions and inter-system connection edges representing evolutionary tension; Obtain the first slope parameter of the dynamic risk load sequence of the first organ dimension; Obtain the second slope parameter of the node base potential value of the second organ dimension in the chronic background path; When the first slope parameter is greater than the first preset slope threshold and the second slope parameter is less than the second preset slope threshold, and the mutual information value between the first organ dimension and the second organ dimension in the current time window is lower than the preset coupling threshold, it is determined that there is a cross-organ physiological compensation failure state, a tension feature vector representing the deviation of organ evolution trajectory is generated, and the tension feature vector is passed to the subsequent conflict determination stage.

10. The method as described in claim 1, characterized in that, When an asynchronous state deviation that meets a threshold condition is detected, it is determined that a preset decision conflict condition is met, and proactive medical resource scheduling suggestions are generated and multidisciplinary consultation expert node recommendation information is output, including: Calculate the deviation of the dynamic evolution directional coupling degree from the initial baseline state parameters; If the deviation value is greater than the preset conflict determination threshold, an activation signal is generated to trigger resource allocation operation.

11. The method as described in claim 10, characterized in that, It generates proactive medical resource scheduling suggestions and outputs multidisciplinary consultation expert node recommendation information, and also includes: In response to the activation signal, based on the specific organ dimension combination in which the asynchronous state shift phenomenon occurs, a preset medical resource database is matched, and a multidisciplinary consultation notification message containing a target rescue equipment allocation suggestion and a list of experts from related departments is output.

12. The method as described in claim 11, characterized in that, The multi-dimensional organ state space includes a first designated organ dimension and a second designated organ dimension; the output includes a multidisciplinary consultation notification message containing recommendations for the allocation of target rescue equipment and a list of experts from related departments, including: When the index offset of the first designated organ dimension reaches the improvement preset value, and the index offset of the second designated organ dimension reaches the deterioration preset value, thus causing the asynchronous state offset phenomenon, the multidisciplinary consultation notification message is written with the assessment suggestion information for the reduction of the life support equipment operating parameters for the first designated organ dimension, and a reference work order for manpower scheduling of the department experts corresponding to the second designated organ dimension is generated. The multidisciplinary consultation notification message containing the target rescue equipment allocation suggestion and the list of department experts is sent to the hospital collaborative management terminal system.

13. A medical resource scheduling and risk early warning system, characterized in that, include: The spatial mapping module is used to acquire serialized clinical event records within the clinical diagnosis and treatment cycle, and to map the clinical event records to a preset multi-dimensional organ state space based on preset mapping rules. A dual-path construction module is used to construct a dual-path representation map of a patient in parallel in the multi-dimensional organ state space. The dual-path representation map includes a chronic background path constructed using historical and underlying disease characteristics and an acute decompensation evolution path updated using current acute emergencies. The dynamic aggregation module is used to update the state transition edges using a clinical event triggering mechanism, extract the instantaneous weights and time change rates of the state transition edges, and perform dynamic weighted aggregation of historical shocks using a preset time decay function to generate a continuous dynamic risk load sequence. The mutual information monitoring module is used to monitor the dynamic evolution direction coupling degree of cross-organ system based on the time-delay mutual information of different organ dimensions on the dynamic risk load sequence. The scheduling trigger module is used to determine whether the preset decision conflict conditions are met when it detects asynchronous state deviation that meets the threshold conditions, generate proactive medical resource scheduling suggestions and output multidisciplinary consultation expert node recommendation information.

14. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in claim 1.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in claim 1.

Images