Critical patient multi-parameter real-time early warning and first aid cooperation system

Abstract

The invention relates to the technical field of critical patient emergency treatment, and discloses a critical patient multi-parameter real-time early warning and emergency treatment cooperative system, which comprises an anomaly detection module, a causal analysis module, a contribution degree tracing module, a root cause judgment module, a scheme matching module and an instruction arrangement module, performing real-time anomaly detection on the multi-parameter monitoring data flow of the medical information terminal to obtain alarm event data; analyzing a predefined relation topology based on the data to obtain a data flow causal structure; tracing the influence of a causal structure to obtain a root cause contribution degree sequence of each abnormal parameter; judging the significance of a preset clinical standard according to the sequence, and obtaining a root cause parameter judgment result of the early warning and first aid cooperative terminal; matching a first-aid clinical protocol library based on the result to obtain a standardized intervention scheme; analyzing a first-aid execution process of the scheme operation logic, and arranging to obtain a team cooperation instruction set; according to the invention, the efficiency of early warning response and co-processing of emergency treatment of critical patients can be improved.

Description

Technical Field

[0001] This invention relates to the field of critical care technology, and in particular to a multi-parameter real-time early warning and emergency care coordination system for critically ill patients. Background Technology

[0002] In the clinical monitoring of critically ill patients, real-time processing of multi-parameter monitoring data streams and emergency coordination are crucial for ensuring treatment effectiveness. With the development of medical monitoring technology, medical information terminals can simultaneously collect multiple physiological parameters such as heart rate, blood glucose, and blood pressure, forming a continuous and dynamic multi-parameter monitoring data stream. This provides data support for clinical anomaly early warning, disease analysis, and emergency intervention. Currently, anomaly detection, root cause analysis, and intervention plan matching based on multi-parameter monitoring data have become key research areas in critical care. Related technologies, through the analysis and processing of monitoring data, assist medical staff in identifying changes in patient condition, developing emergency strategies, and facilitating team collaboration, aiming to improve the timeliness and accuracy of critical care treatment.

[0003] In existing multi-parameter monitoring anomaly detection technologies, threshold settings often adopt fixed standards, lacking adaptability to individual patient physiological differences and dynamic parameter changes. Furthermore, they fail to incorporate clinical time-series patterns for pseudo-anomaly filtering, resulting in insufficient accuracy in anomaly judgment and an inability to provide a reliable data foundation for root cause analysis. Simultaneously, existing technologies struggle to effectively analyze the causal relationships between multi-parameter anomalies and lack standardized intervention program matching and team collaboration instruction orchestration mechanisms based on root cause contribution quantification. This leads to a lack of targeted emergency interventions, low team collaboration efficiency, and an inability to meet the precision and coordination requirements of critical care. Therefore, improving the efficiency of early warning response and collaborative handling generation has become an urgent problem to be solved. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides a multi-parameter real-time early warning and emergency care coordination system for critically ill patients, characterized in that the system includes an anomaly detection module, a causal analysis module, a contribution tracing module, a root cause determination module, a plan matching module, and an instruction arrangement module, wherein:

[0005] The anomaly detection module is used to perform real-time anomaly detection on the multi-parameter monitoring data stream in the medical information terminal and obtain alarm event data of the multi-parameter monitoring data stream.

[0006] The causal analysis module is used to perform structural analysis on the predefined relationship topology based on the alarm event data to obtain the causal structure of the multi-parameter monitoring data stream;

[0007] The contribution tracing module is used to trace the causal influence of the causal structure and obtain the root cause contribution ranking of each abnormal parameter in the alarm event data.

[0008] The root cause determination module is used to determine the significance of preset clinical standards based on the root cause contribution ranking, and obtain the root cause parameter determination result of the early warning and emergency rescue collaboration terminal.

[0009] The scheme matching module is used to perform strategy matching on a preset emergency clinical protocol library based on the root cause parameter determination results, so as to obtain a standardized intervention scheme for the early warning and emergency collaborative terminal.

[0010] The instruction orchestration module is used to parse the clinical operation logic in the standardized intervention plan, obtain the emergency execution process of the standardized intervention plan, and orchestrate the emergency execution process to obtain the team collaboration instruction set of the early warning and emergency collaboration terminal.

[0011] In a preferred embodiment, when the anomaly detection module performs real-time anomaly detection on the multi-parameter monitoring data stream in the medical information terminal and obtains alarm event data from the multi-parameter monitoring data stream, it is specifically used for:

[0012] The multi-parameter monitoring data stream in the medical information terminal is divided into sliding window segments to obtain the parameter time sequence segments of the multi-parameter monitoring data stream;

[0013] Adaptive threshold learning is performed on the time series segments of the parameters to obtain the dynamic anomaly boundaries of the time series segments of the parameters;

[0014] The parameter time series segment and the corresponding dynamic anomaly boundary are dynamically delimited to obtain the preliminary anomaly marker point of the parameter time series segment;

[0015] Clinical time-series pattern matching is performed on the preliminary abnormality markers, and non-pathological pseudo-abnormality markers are filtered out to obtain the abnormal events of the preliminary abnormality markers.

[0016] The abnormal events are aggregated for multi-parameter clinical correlation to obtain alarm event data from the multi-parameter monitoring data stream.

[0017] In a preferred embodiment, when the causal analysis module performs structural analysis on a predefined relational topology based on the alarm event data to obtain the causal structure of the relationship between the multi-parameter monitoring data streams, it is specifically used for:

[0018] Alarm event mapping is performed on the predefined relationship topology to obtain the activation parameter nodes of the predefined relationship topology;

[0019] Based on the activated parameter nodes, directed edge retrieval is performed on the predefined relation topology to obtain the inter-node association edges of the activated parameter node set;

[0020] Temporal dependency analysis is performed on the edges connecting the nodes within a time window to obtain the causal direction confidence of the edges connecting the nodes.

[0021] Based on the causal direction confidence, directional constraints are applied to the inter-node association edges to obtain directed causal edges of the inter-node association edges;

[0022] By assigning causal weights to the directed causal edges, a weighted directed causal graph of the directed causal edges is obtained.

[0023] By mining the maximum causal flow path in the weighted directed causal graph, the causal structure of the relationship between the multi-parameter monitoring data streams is obtained.

[0024] In a preferred embodiment, when the causal analysis module performs temporal dependency analysis on the inter-node association edges within a time window to obtain the causal direction confidence of the inter-node association edges, it is specifically used for:

[0025] Extract a specified time window from the alarm event data to obtain a multi-parameter time sequence segment of the alarm event data;

[0026] Delay mutual information is measured on the multi-parameter time segment to obtain the delay mutual information matrix between each pair of parameters in the multi-parameter time segment.

[0027] An asymmetric statistical test is performed on the row and column elements of the delayed mutual information matrix to obtain the time-series anticipation measure between the parameters of the delayed mutual information matrix.

[0028] The temporal precedence measure is consistently fused with a pre-built clinical causal knowledge base to obtain the causal direction confidence of the associated edges between the nodes.

[0029] In a preferred embodiment, when the contribution tracing module performs causal influence tracing on the causal structure to obtain the root cause contribution ranking of each abnormal parameter in the alarm event data, it is specifically used for:

[0030] The causal structure is deconstructed into directed paths to obtain the directed path sequence of the causal structure;

[0031] Based on the directed path sequence, abnormal node mapping is performed on the alarm event data to obtain the associated nodes of the alarm event data;

[0032] Influence topology construction is performed on the associated nodes to obtain the influence transmission graph of the associated nodes;

[0033] By tracing back the source nodes of the influence transmission graph, the potential root cause node clusters of the influence transmission graph are obtained.

[0034] The potential root cause node clusters are aggregated by contribution metric to obtain a preliminary contribution list of the abnormal parameters;

[0035] The root cause contribution ranking of the alarm event data is obtained by comparing and calibrating the preliminary contribution list with the preset clinical significance threshold.

[0036] In a preferred embodiment, when the contribution tracing module performs contribution quantification aggregation on the potential root cause node cluster to obtain a preliminary contribution list of the abnormal parameters, it is specifically used for:

[0037] The potential root cause node cluster is weighted to obtain the initial influence weight value of the potential root cause node.

[0038] Based on the directed path sequence, the initial influence weight value is propagated using path dependency to obtain the propagation influence score of the abnormal parameter.

[0039] The propagation impact score is scaled to form a standardized contribution score for the anomaly parameter;

[0040] The standardized contribution scores are sorted and integrated to obtain a preliminary contribution list of the abnormal parameters.

[0041] In a preferred embodiment, when the root cause determination module performs significance determination on preset clinical criteria based on the root cause contribution ranking to obtain the root cause parameter determination result of the early warning and emergency response collaboration terminal, it is specifically used for:

[0042] The contribution values ​​in the root cause contribution ranking are subjected to time-series normalization to obtain the time-series dynamic contribution index of the multi-parameter monitoring data stream.

[0043] Extract the fusion clinical prior knowledge from the preset clinical standards;

[0044] Based on the aforementioned fusion of prior clinical knowledge, a dynamic threshold is calculated for the time-series dynamic contribution index to obtain the consensus integration and judgment threshold for the alarm event data.

[0045] The time-series dynamic contribution index and the consensus integration judgment threshold are compared hierarchically to obtain the significant anomaly level of the alarm event data;

[0046] The significant anomaly hierarchy is mapped to the topological space of the causal structure to obtain the anomaly contribution hierarchy map of the alarm event data;

[0047] By performing physiological path backtracking on the abnormal contribution hierarchy map, a subset of priority association parameters of the alarm event data is obtained;

[0048] The clinical intervention urgency is assigned to the subset of priority-related parameters to obtain the root cause parameter determination results of the early warning and emergency rescue collaboration terminal.

[0049] In a preferred embodiment, when the root cause determination module performs dynamic threshold calculation on the time-series dynamic contribution index based on the fused clinical prior knowledge to obtain the consensus integration determination threshold for the alarm event data, the calculation formula for the consensus integration determination threshold is as follows:

[0050] ;

[0051] In the formula, For a moment The consensus integration determination threshold, The first adjustment coefficient, This is the second adjustment coefficient. This is the third adjustment coefficient. This constitutes the total number of independent clinical guidelines that integrate prior clinical knowledge. To integrate the first clinical prior knowledge in the above-mentioned Independent clinical guidelines A function for quantitative rule extraction. This represents the total number of monitoring parameters in the current alarm event data that are in an abnormal state. For the index of the abnormal parameter, To index based on pathophysiological importance The corresponding abnormal parameters are pre-assigned dimensionless static weights. For index The corresponding abnormal parameters at time The aforementioned time-series dynamic contribution index, This is a function that performs a nonlinear normalized mapping on the input time-series dynamic contribution index. The time from the occurrence of the first alarm event to the current time. The time interval experienced The preset time decay coefficient, This is an S-shaped growth curve function. This is a dimensionless metric value calculated based on the multi-parameter monitoring data stream.

[0052] In a preferred embodiment, when the scheme matching module performs strategy matching on a preset emergency clinical protocol library based on the root cause parameter determination result to obtain a standardized intervention scheme for the early warning and emergency coordination terminal, it is specifically used for:

[0053] The clinical dimension analysis of the root cause parameter determination results is performed to obtain the clinical intervention focus of the root cause parameter determination results;

[0054] Based on the clinical intervention focus, a protocol path traversal is performed on the preset emergency clinical protocol library to obtain candidate clinical intervention paths for the clinical intervention focus.

[0055] The candidate clinical intervention pathways are quantified using multidimensional factors to obtain their urgency priority and confidence assessment values.

[0056] Based on the urgency priority score and the confidence assessment value, the candidate clinical intervention pathways are weighted and ranked to obtain the intervention pathway sequence of the candidate clinical intervention pathways;

[0057] Based on the clinical operation constraint rules in the preset emergency clinical protocol library, the intervention path sequence is fused to obtain the optimized intervention strategy of the emergency clinical protocol library.

[0058] The optimized intervention strategy is deployed in a clinical operation instruction manner to obtain a standardized intervention plan for the early warning and emergency rescue collaborative terminal.

[0059] In a preferred embodiment, when the instruction orchestration module executes the parsing of the clinical operation logic in the standardized intervention plan to obtain the emergency execution flow of the standardized intervention plan, and orchestrates the emergency execution flow to obtain the team collaboration instruction set of the early warning and emergency collaboration terminal, it is specifically used for:

[0060] The standardized intervention protocol is deconstructed into clinical operational logic to obtain a sequence of logical elements of the standardized intervention protocol;

[0061] Analyze the temporal and logical dependencies of the logical element sequence to obtain the dependencies of the standardized intervention scheme;

[0062] Based on the aforementioned dependencies, a directed graph structure is constructed from the sequence of logical elements to obtain the topology graph of the emergency steps of the standardized intervention scheme.

[0063] Role capability mapping is performed on the topology of the emergency rescue steps to obtain the preliminary task assignment table of the early warning and emergency rescue collaboration terminal;

[0064] Spatiotemporal resource conflicts are resolved in the preliminary task assignment table to obtain the collaborative task scheduling scheme of the early warning and emergency rescue collaborative terminal;

[0065] Multimodal instruction synthesis is performed on the task timing and role binding information in the collaborative task scheduling scheme to obtain the team collaboration instruction set of the early warning and emergency rescue collaborative terminal.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] 1. This invention divides the regular data stream by sliding window, combines individual patient physiological differences, dynamic parameter characteristics and clinical normal range for adaptive threshold learning to form dynamic abnormal boundaries, then marks preliminary abnormal points by point-by-point comparison, filters non-pathological pseudo-abnormalities by clinical time sequence pattern matching, and finally aggregates multi-parameter related abnormal events. This solves the problems of poor adaptability of fixed thresholds and high false alarm rate in the prior art, and ensures that abnormal data is consistent with the actual condition of patients.

[0068] 2. This invention analyzes the multi-parameter causal relationship structure, quantifies the root cause contribution of each abnormal parameter, and determines the core root cause in combination with clinical standards. Based on the standardized intervention plan optimized by root cause matching, it further deconstructs the operational logic, analyzes the dependency relationship, matches role capabilities, resolves resource conflicts, and compiles a multimodal team collaboration instruction set. This solves the problems of unclear causal relationships, lack of targeted intervention, and low collaboration in existing technologies, and meets the precision and collaboration requirements of critical care emergency treatment. Attached Figure Description

[0069] Figure 1 This is a system architecture diagram of a multi-parameter real-time early warning and emergency care coordination system for critically ill patients provided in an embodiment of the present invention;

[0070] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0073] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0074] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0075] In practice, the server-side equipment deployed in the multi-parameter real-time early warning and emergency care coordination system for critically ill patients may consist of one or more devices. This system can be implemented as a business instance, a virtual machine, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node to provide real-time early warning and emergency care coordination for critically ill patients to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage various user terminals. Or, it can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more devices configured to provide the system to various user terminals.

[0076] In terms of implementation, the real-time multi-parameter early warning and emergency care coordination system for critically ill patients and the user terminal are mutually compatible. That is, if the real-time multi-parameter early warning and emergency care coordination system for critically ill patients is implemented as an application installed on a cloud service platform, then the user terminal is a client that establishes a communication connection with the application; or if the real-time multi-parameter early warning and emergency care coordination system for critically ill patients is implemented as a website, then the user terminal is implemented as a webpage; or if the real-time multi-parameter early warning and emergency care coordination system for critically ill patients is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0077] like Figure 1 The figure shown is a system architecture diagram of a multi-parameter real-time early warning and emergency care coordination system for critically ill patients provided in an embodiment of the present invention.

[0078] The real-time early warning and emergency care coordination system 100 for critically ill patients described in this invention can be set up in a cloud server. In terms of implementation, it can be implemented as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed as a website. Depending on the functions implemented, the real-time early warning and emergency care coordination system 100 for critically ill patients may include an anomaly detection module 101, a causal analysis module 102, a contribution tracing module 103, a root cause determination module 104, a solution matching module 105, and an instruction orchestration module 106. The modules described in this invention can also be called units, referring to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0079] In this embodiment of the invention, in the multi-parameter real-time early warning and emergency care coordination system for critically ill patients, each of the above modules can be implemented independently and can call other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the multi-parameter real-time early warning and emergency care coordination system for critically ill patients provided by this embodiment of the invention, the applicable scope of the system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion to quickly and flexibly expand the system. In practical applications, the above modules can be set in the same device or different devices, or they can be set in a virtual device, such as a service instance in a cloud server.

[0080] The following describes the components and specific workflow of the multi-parameter real-time early warning and emergency care coordination system for critically ill patients, using specific embodiments as examples:

[0081] The anomaly detection module 101 is used to perform real-time anomaly detection on the multi-parameter monitoring data stream in the medical information terminal and obtain alarm event data of the multi-parameter monitoring data stream.

[0082] In this embodiment of the invention, when the anomaly detection module performs real-time anomaly detection on the multi-parameter monitoring data stream in the medical information terminal and obtains alarm event data of the multi-parameter monitoring data stream, it is specifically used for:

[0083] The multi-parameter monitoring data stream in the medical information terminal is divided into sliding window segments to obtain the parameter time sequence segments of the multi-parameter monitoring data stream;

[0084] Adaptive threshold learning is performed on the time series segments of the parameters to obtain the dynamic anomaly boundaries of the time series segments of the parameters;

[0085] The parameter time series segment and the corresponding dynamic anomaly boundary are dynamically delimited to obtain the preliminary anomaly marker point of the parameter time series segment;

[0086] Clinical time-series pattern matching is performed on the preliminary abnormality markers, and non-pathological pseudo-abnormality markers are filtered out to obtain the abnormal events of the preliminary abnormality markers.

[0087] The abnormal events are aggregated for multi-parameter clinical correlation to obtain alarm event data from the multi-parameter monitoring data stream.

[0088] Based on the actual update frequency of monitoring parameters for critically ill patients, determine an appropriate window duration and movement step size. The window duration should balance data integrity and timely analysis, while the movement step size should ensure data continuity between adjacent windows.

[0089] The distribution of data within each time series segment is statistically analyzed, including the data concentration range and fluctuation degree. This is combined with the clinically common normal fluctuation range for this parameter in critically ill patients, fully considering the impact of individual physiological differences on the parameter. During the learning process, historical data segments of this parameter during the monitoring period are continuously collected for each patient, and their unique fluctuation patterns are analyzed. Based on this information, the threshold value for judging abnormalities is continuously adjusted to ensure that each parameter time series segment corresponds to a set of threshold standards that fit its own data characteristics and individual patient conditions, ultimately forming a dynamic abnormal boundary specific to that parameter time series segment.

[0090] Each data point in the parameter time series segment is extracted one by one. The specific value of each data point is compared with the upper and lower limits of the dynamic anomaly boundary. Each data point is carefully checked to see if it exceeds the set boundary range. If the value of a data point is higher than the upper limit of the boundary or lower than the lower limit of the boundary, the data point is identified as a suspected anomaly. The specific monitoring time corresponding to the data point is accurately recorded to ensure that each abnormal data point that exceeds the boundary can be accurately identified and marked. Finally, all the preliminary anomaly marking points in the parameter time series segment are obtained.

[0091] First, we compile a list of common non-pathological abnormality patterns encountered in clinical practice. These patterns include parameter fluctuations caused by changes in patient positioning, transient malfunctions in monitoring equipment, etc., and clarify the temporal characteristics of these non-pathological fluctuations, such as duration, trend, and whether they are accompanied by synchronous changes in other parameters. We then comprehensively compare the duration and numerical change curves of each preliminary abnormality marker with the characteristics of these non-pathological abnormality patterns. If the characteristics of a preliminary abnormality marker perfectly match a non-pathological abnormality pattern, the marker is considered a false abnormality and excluded. Only markers whose duration matches the characteristics of pathological abnormalities or are correlated with changes in other parameters are retained; these retained markers collectively constitute the abnormal event.

[0092] Based on clinical pathology knowledge, we conduct in-depth analysis of the physiological correlations between the monitoring parameters corresponding to various abnormal events to determine whether different abnormal events are caused by the same disease progression. For example, abnormal events related to blood glucose, heart rate, and sweating have a clear pathological correlation in clinical practice, belonging to multiple abnormal parameter manifestations caused by the same disease progression. By integrating these physiologically correlated abnormal events and comprehensively summarizing the abnormal values, occurrence time, and trends of all relevant parameters, we form a complete and systematic alarm event record. This record fully presents the status of all relevant parameters that triggered the alarm and their interrelationships, ultimately yielding the alarm event data corresponding to the multi-parameter monitoring data stream.

[0093] The beneficial effects are as follows: Sliding window partitioning can transform continuous data streams into regular parameter time-series segments, ensuring data continuity, avoiding the omission of key changes, and providing a clearly structured data foundation for subsequent anomaly detection, thus improving the orderliness and efficiency of anomaly detection. Adaptive threshold learning dynamically adjusts thresholds by combining parameter data features, individual patient patterns, and clinical normal ranges, avoiding the problem that fixed thresholds cannot adapt to individual differences and parameter changes, making the abnormal boundary closely match the patient's actual condition, and improving the pertinence and accuracy of anomaly judgment. Dynamic boundary clarification accurately identifies and marks suspected abnormal points and clarifies their location and time by comparing data with abnormal boundaries point by point, reducing interference from invalid data. This provides precise targets for subsequent screening of real abnormal events, further improving detection accuracy and efficiency. Clinical time-series pattern matching and false anomaly filtering can accurately identify and exclude non-pathological false anomalies, significantly reducing false alarms and ensuring that the retained abnormal events are related to the patient's actual condition. This provides reliable data for clinical judgment and emergency decision-making, avoiding interference from false alarms for medical staff. Multi-parameter clinical correlation aggregation can integrate related abnormal events caused by the same condition, comprehensively presenting parameter abnormalities and their interrelationships. This avoids misjudging the condition caused by viewing a single abnormal event in isolation, helping medical staff quickly grasp the full picture of the condition and providing complete and systematic data support for timely and accurate emergency measures.

[0094] The causal analysis module 102 is used to perform structural analysis on the predefined relationship topology based on the alarm event data to obtain the causal structure of the multi-parameter monitoring data stream.

[0095] In this embodiment of the invention, when the causal analysis module performs structural analysis on the predefined relationship topology based on the alarm event data to obtain the causal structure of the relationship between the multi-parameter monitoring data streams, it is specifically used for:

[0096] Alarm event mapping is performed on the predefined relationship topology to obtain the activation parameter nodes of the predefined relationship topology;

[0097] Based on the activated parameter nodes, directed edge retrieval is performed on the predefined relation topology to obtain the inter-node association edges of the activated parameter node set;

[0098] Temporal dependency analysis is performed on the edges connecting the nodes within a time window to obtain the causal direction confidence of the edges connecting the nodes.

[0099] Based on the causal direction confidence, directional constraints are applied to the inter-node association edges to obtain directed causal edges of the inter-node association edges;

[0100] By assigning causal weights to the directed causal edges, a weighted directed causal graph of the directed causal edges is obtained.

[0101] By mining the maximum causal flow path in the weighted directed causal graph, the causal structure of the relationship between the multi-parameter monitoring data streams is obtained.

[0102] When the causal analysis module performs temporal dependency analysis on the inter-node association edges within a time window to obtain the causal direction confidence of the inter-node association edges, it is specifically used for:

[0103] Extract a specified time window from the alarm event data to obtain a multi-parameter time sequence segment of the alarm event data;

[0104] Delay mutual information is measured on the multi-parameter time segment to obtain the delay mutual information matrix between each pair of parameters in the multi-parameter time segment.

[0105] An asymmetric statistical test is performed on the row and column elements of the delayed mutual information matrix to obtain the time-series anticipation measure between the parameters of the delayed mutual information matrix.

[0106] The temporal precedence measure is consistently fused with a pre-built clinical causal knowledge base to obtain the causal direction confidence of the associated edges between the nodes.

[0107] The predefined relational topology is a pre-built framework containing potential correlations among various monitoring parameters of critically ill patients, covering the correlation logic of physiological and pathological parameters. When mapping alarm events, key information of abnormal parameters in the alarm event data is first extracted comprehensively. Then, this information is compared one by one with the parameter nodes in the predefined relational topology to filter out nodes directly related to the abnormal parameters. These nodes are the activated parameter nodes.

[0108] Based on the identified active parameter nodes, the position and association range of each node in the predefined relational topology are first determined. Then, with each active node as the core, all lines connecting these nodes are comprehensively searched in the topology. During the retrieval, potential connection paths between nodes are examined one by one, and the node information connected by each associated edge is recorded, ultimately obtaining the associated edges between nodes.

[0109] First, based on the patterns of critically ill patients' conditions and clinical monitoring needs, a time window is set to cover key time periods before and after alarm events to ensure the complete capture of parameter change trajectories. Within the time window, real-time change data of the two parameters connected by each associated edge are continuously collected. The order of parameter changes is analyzed to determine the degree of influence of the earlier changing parameter on the later changing parameter, thereby forming a preliminary judgment on the causal direction of each associated edge and determining the confidence level of the causal direction.

[0110] By combining clinical data and medical research, a confidence threshold for causal direction is set to determine the clarity of the causal direction. The confidence of each associated edge is compared with the threshold: if the confidence is higher than the threshold, the direction from the cause node to the result node is marked according to the analyzed causal direction; if the confidence is lower than the threshold, the original data is retrieved again, auxiliary parameters are added, or the time window is adjusted and analyzed again until the confidence reaches the threshold and the direction is marked, thus forming a directed causal edge.

[0111] By combining clinical cases, medical research, and treatment guidelines, this study analyzes the importance and intensity of causal relationships represented by different directed causal edges in disease progression, and establishes a multi-dimensional weighting standard covering factors such as urgency, scope, and criticality of intervention. Each directed causal edge is evaluated and scored according to the standard, assigned a corresponding weight, and then the weighted directed causal edges are integrated and arranged according to node position to construct a weighted directed causal graph.

[0112] By sorting out all possible causal transmission paths in the weighted directed causal graph, calculating the sum of the weights of all directed causal edges on each path, i.e., the causal flow strength, comparing and sorting the causal flow strengths of all paths, and selecting the path with the highest strength, i.e. the maximum causal flow path, organizing and summarizing the nodes, directed causal edges and weight relationships contained in the path, and finally forming the causal structure of the multi-parameter monitoring data flow.

[0113] Based on the characteristics of disease progression and monitoring accuracy requirements, a time window covering the period before and after the alarm event is set. This means the start time includes the pre-alarm warning period, and the end time includes the post-alarm change period, ensuring complete capture of parameter changes. Data for all monitoring parameters within this time window, including numerical values, change times, and fluctuation amplitudes, is extracted from the complete alarm data and sorted chronologically to form a multi-parameter time series segment.

[0114] Each parameter in the time segment is paired with other parameters, and a time delay value is set for each group of parameters to cover the possible response delay. For each delay value, the mutual information value of the group of parameters is calculated, which reflects the degree of correlation between the parameters under that delay. Then, according to the parameter combination and the order of the delay value, all mutual information values ​​are arranged into a two-dimensional table, namely the delay mutual information matrix.

[0115] Observe the mutual information values ​​of any two sets of parameters in the matrix, such as the mutual information values ​​of parameter A to B and B to A, and compare the differences between the two to analyze whether there is significant asymmetry: if the mutual information value of A to B is larger, it means that A is more likely to change first and affect B. Perform statistical analysis on the asymmetry of all row and column elements to quantify the degree of precedence of each parameter relative to other parameters and form a time-series precedence measurement result.

[0116] A pre-built clinical causal knowledge base contains proven parametric causal relationships. The time-series prior measurement results are compared with the knowledge base entries one by one: if they match, the credibility of the causal direction is enhanced; if there are differences, the measurement results are corrected or the knowledge base is improved by combining clinical reality and patient conditions. Finally, a comprehensive evaluation is conducted to determine the confidence level of the causal direction of each associated edge.

[0117] The beneficial effects are as follows: By accurately extracting abnormal parameter information and matching it with the topology, the core nodes related to the alarm can be quickly identified, and interference from irrelevant nodes can be eliminated. This provides clear targets for subsequent parameter correlation analysis, improving the targeting and initial efficiency of causal structure analysis. It focuses on activating the correlation between nodes, efficiently obtaining accurate correlation edges, eliminating interference from irrelevant correlation edges in the topology, providing a clear foundation for subsequent causal direction analysis, reducing the amount of invalid data processing, focusing on the temporal relationship of parameters within the time window, accurately capturing the dynamic influence of parameters, accurately judging the potential causal direction of correlation edges and quantifying their credibility, providing data support for subsequent determination of directed causal edges, improving analysis accuracy, and ensuring that directed causal edges have clear and reliable causal directions through scientific threshold constraints, avoiding deviations caused by ambiguity in causal direction, laying the foundation for constructing accurate causal graphs. Through quantitative weights and graph construction, the important levels of different causal relationships are intuitively reflected, laying the foundation for subsequent mining of the maximum causal flow. The pathway provides clear evidence, making the causal structure more comprehensive and accurate. It precisely locates the core causal transmission relationship, forming a logically clear and focused causal structure. This provides core evidence for subsequent tracing of root cause parameters and formulating intervention plans, improving the accuracy and effectiveness of critical care early warning. It focuses on data in key time periods related to alarms, eliminating interference from irrelevant time periods, providing a precise data foundation for subsequent measurement and analysis, and improving the targeting of analysis. The mutual information calculation under excessive delays comprehensively captures the dynamic correlation of parameters, avoiding information omissions in fixed-time point analysis, and providing rich quantitative support for subsequent time-series precedence analysis. It judges time-series precedence relationships based on quantitative mutual information differences, avoiding subjective bias, and provides key time-series evidence for determining the causal direction, enhancing the scientific nature of the judgment. It uses mature clinical knowledge to verify data results, avoiding the bias of simply relying on data, significantly improving the accuracy of causal direction confidence, and providing clinically relevant support for causal structure analysis.

[0118] The contribution tracing module 103 is used to trace the causal influence of the causal structure and obtain the root cause contribution ranking of each abnormal parameter in the alarm event data.

[0119] In this embodiment of the invention, when the contribution tracing module performs causal influence tracing on the causal structure to obtain the root cause contribution ranking of each abnormal parameter in the alarm event data, it is specifically used for:

[0120] The causal structure is deconstructed into directed paths to obtain the directed path sequence of the causal structure;

[0121] Based on the directed path sequence, abnormal node mapping is performed on the alarm event data to obtain the associated nodes of the alarm event data;

[0122] Influence topology construction is performed on the associated nodes to obtain the influence transmission graph of the associated nodes;

[0123] By tracing back the source nodes of the influence transmission graph, the potential root cause node clusters of the influence transmission graph are obtained.

[0124] The potential root cause node clusters are aggregated by contribution metric to obtain a preliminary contribution list of the abnormal parameters;

[0125] The root cause contribution ranking of the alarm event data is obtained by comparing and calibrating the preliminary contribution list with the preset clinical significance threshold.

[0126] When the contribution tracing module performs contribution quantification aggregation on the potential root cause node cluster to obtain a preliminary contribution list of the abnormal parameters, it is specifically used for:

[0127] The potential root cause node cluster is weighted to obtain the initial influence weight value of the potential root cause node.

[0128] Based on the directed path sequence, the initial influence weight value is propagated using path dependency to obtain the propagation influence score of the abnormal parameter.

[0129] The propagation impact score is scaled to form a standardized contribution score for the anomaly parameter;

[0130] The standardized contribution scores are sorted and integrated to obtain a preliminary contribution list of the abnormal parameters.

[0131] The causal structure is defined as a whole composed of multiple interconnected parameter nodes and directed causal edges. Directed path deconstruction involves breaking down all complete related paths in the causal structure one by one according to the direction of causal relationship transmission, ensuring that no ordered link from the starting node to the ending node is omitted during the breakdown. After breakdown, these independent paths are arranged according to their logical order in the causal structure, ultimately forming a directed path sequence.

[0132] First, extract all parameter information marked as abnormal from the alarm event data. This information includes the temporal characteristics of the abnormality and the attributes of the parameters themselves. Then, using the previously obtained directed path sequence as a reference framework, compare the extracted abnormal parameter information with each node in the path sequence one by one, accurately mapping each abnormal parameter to a matching parameter node in the sequence. These successfully matched nodes are the associated nodes of the alarm event data. The entire process ensures that the associated nodes can completely cover all parameter nodes related to the abnormality.

[0133] First, the attributes and characteristics of each associated node and its specific position in the directed path sequence are analyzed to determine the scope and intensity of each node's potential influence on other nodes. Then, with each associated node as the core and the actual influence relationships between nodes as the connecting links, all associated nodes are systematically connected according to the true logic of influence transmission, forming a structural map that visually presents the paths and levels of mutual influence between nodes—an influence transmission map—ensuring that this map accurately reflects the influence transmission between associated nodes.

[0134] Starting with all terminal nodes in an abnormal state in the influence transmission diagram, we work backwards along the opposite direction of influence transmission. During this backwards process, we examine each upstream influencing node to determine if it is an initial node capable of autonomously triggering anomalies, unaffected by other nodes. All confirmed initial nodes are categorized and integrated to form a cluster of potential root cause nodes, ensuring that no possible initial anomalous node is overlooked.

[0135] First, identify the core dimensions for measuring contribution, including the propagation range of the anomaly triggered by the node, the intensity of its impact, and its criticality in the causal structure. For each node in the potential root cause node cluster, conduct a specific numerical evaluation based on the aforementioned core dimensions to obtain a quantified contribution value for each node. Then, according to the correlation between nodes and anomaly parameters, summarize and integrate the quantified values ​​of all potential root cause nodes corresponding to the same anomaly parameter to form preliminary contribution data related to the anomaly parameter, and then compile it into a preliminary contribution list.

[0136] First, a comprehensive analysis is conducted to identify the physiological functional importance of each node in the potential root cause node cluster, its criticality in clinical diagnosis and treatment, and the frequency of its influence in similar past abnormal cases. Based on this information, a unified and reasonable weight allocation standard is established. Each node is assigned a corresponding initial value according to this standard; this initial value is the initial influence weight value, ensuring that the weight allocation accurately reflects the importance and potential influence of the node itself.

[0137] Using directed path sequences as the channel for influence propagation, starting from each potential root cause node, the initial influence weight value of that node is progressively passed down to downstream related nodes according to the association order and propagation direction of the nodes in the path. During the propagation process, the weight values ​​are adjusted accordingly based on factors such as the length of the path and the tightness of the association between nodes. Finally, all the propagated weight values ​​received by each anomalous parameter are summarized to obtain the propagation influence score of that anomalous parameter.

[0138] First, a unified standard scale range is determined, which must meet the routine quantitative requirements of clinical assessment. For each abnormal parameter's propagation impact score, it is converted into a corresponding value within the standard scale range according to a pre-set, reasonable mapping rule. During the conversion process, the consistency and rationality of the mapping are strictly ensured, so that the propagation impact scores of different abnormal parameters can be compared under the same standard. The converted value is the standardized contribution score.

[0139] First, define the sorting rules, using the standardized contribution score from highest to lowest as the primary sorting criterion. For outliers with the same score, a secondary sort is performed based on the importance of their corresponding potential root cause nodes. Arrange all outliers in an ordered manner according to this sorting rule, while simultaneously integrating the relevant basic information for each outlier to form a clear and complete preliminary contribution list, ensuring that the list intuitively presents the order of contribution of each outlier.

[0140] The beneficial effects include: clearly outlining the causal transmission of parameters, simplifying complex causal relationships, laying the foundation for accurately locating the root cause of anomalies, avoiding biases in root cause tracing, quickly identifying relevant parameter nodes of alarm events, eliminating interference from irrelevant nodes, improving the targeting and efficiency of root cause tracing, transforming complex influence relationships of related nodes into structured maps, facilitating the understanding of the scope and path of node influence, providing clear evidence for tracing back to source nodes, focusing on core nodes that may cause anomalies, narrowing the scope of root cause investigation, improving the accuracy and efficiency of root cause identification, converting the impact of nodes on anomalies into quantitative information, providing data support for contribution ranking, improving the objectivity and scientific rigor of root cause analysis, considering the inherent importance of nodes, laying the foundation for accurate quantification of contribution, avoiding biases in contribution assessment due to differences in node importance, considering the transmission process and path characteristics of abnormal influences, accurately reflecting the actual impact of nodes on abnormal parameters, improving the accuracy of contribution assessment, eliminating the incomparability of different propagation influence scores, establishing a unified standard for measuring contribution, facilitating ranking and integration, presenting the priority of abnormal parameter contribution, providing an orderly data foundation for comparison and calibration with clinical significance thresholds, and facilitating the rapid location of core abnormal parameters.

[0141] The root cause determination module 104 is used to determine the significance of preset clinical standards based on the root cause contribution ranking, and obtain the root cause parameter determination result of the early warning and emergency rescue collaboration terminal.

[0142] In this embodiment of the invention, when the root cause determination module performs a significance determination on preset clinical standards based on the root cause contribution ranking to obtain the root cause parameter determination result of the early warning and emergency response collaboration terminal, it is specifically used for:

[0143] The contribution values ​​in the root cause contribution ranking are subjected to time-series normalization to obtain the time-series dynamic contribution index of the multi-parameter monitoring data stream.

[0144] Extract the fusion clinical prior knowledge from the preset clinical standards;

[0145] Based on the aforementioned fusion of prior clinical knowledge, a dynamic threshold is calculated for the time-series dynamic contribution index to obtain the consensus integration and judgment threshold for the alarm event data.

[0146] The time-series dynamic contribution index and the consensus integration judgment threshold are compared hierarchically to obtain the significant anomaly level of the alarm event data;

[0147] The significant anomaly hierarchy is mapped to the topological space of the causal structure to obtain the anomaly contribution hierarchy map of the alarm event data;

[0148] By performing physiological path backtracking on the abnormal contribution hierarchy map, a subset of priority association parameters of the alarm event data is obtained;

[0149] The clinical intervention urgency is assigned to the subset of priority-related parameters to obtain the root cause parameter determination results of the early warning and emergency rescue collaboration terminal.

[0150] When the root cause determination module performs dynamic threshold calculation on the time-series dynamic contribution index based on the fused clinical prior knowledge to obtain the consensus integration determination threshold for the alarm event data, the specific formula for calculating the consensus integration determination threshold is as follows:

[0151] ;

[0152] In the formula, For a moment The consensus integration determination threshold, The first adjustment coefficient, This is the second adjustment coefficient. This is the third adjustment coefficient. This constitutes the total number of independent clinical guidelines that integrate prior clinical knowledge. To integrate the first clinical prior knowledge in the above-mentioned Independent clinical guidelines A function for quantitative rule extraction. This represents the total number of monitoring parameters in the current alarm event data that are in an abnormal state. For the index of the abnormal parameter, To index based on pathophysiological importance The corresponding abnormal parameters are pre-assigned dimensionless static weights. For index The corresponding abnormal parameters at time The aforementioned time-series dynamic contribution index, This is a function that performs a nonlinear normalized mapping on the input time-series dynamic contribution index. The time from the occurrence of the first alarm event to the current time. The time interval experienced The preset time decay coefficient, This is an S-shaped growth curve function. This is a dimensionless metric value calculated based on the multi-parameter monitoring data stream.

[0153] When the root cause parameter determination unit performs root cause parameter determination, it first extracts all contribution values ​​from the acquired root cause contribution ranking results and organizes them into an ordered numerical sequence according to the chronological order of data collection. Then, using a unified numerical transformation rule, each contribution value in this sequence is adjusted to eliminate interference caused by differences in data collection conditions and parameter monitoring ranges at different time points. This ensures that each adjusted value accurately reflects the root cause contribution of the parameter at the corresponding time point. Through this process, an index data that dynamically reflects the changes in root cause contribution of multi-parameter monitoring data streams at different time stages is ultimately formed.

[0154] The pre-defined clinical standards are diagnostic and treatment references established by professional medical institutions or industry associations based on extensive clinical practice. The root cause parameter determination unit comprehensively reviews the content related to the determination of abnormal monitoring parameters in critically ill patients and root cause analysis within this clinical standard system. Each piece of content is screened, eliminating information irrelevant to the current alarm event data and lacking practical reference value. Key knowledge points that guide root cause parameter determination, reflect pathophysiological relationships, and clarify the principles of abnormal parameter assessment are extracted. These scattered knowledge points are then systematically integrated to form a complete and targeted set of integrated clinical prior knowledge.

[0155] The root cause parameter determination unit uses integrated clinical prior knowledge as the basis for calculation, comprehensively analyzing the numerical trend, fluctuation range, and degree of conformity with relevant standards in the time-series dynamic contribution index. Based on the anomaly assessment principles and parameter importance classifications clearly defined in the clinical prior knowledge, corresponding calculation rules are formulated, and the values ​​of each item in the time-series dynamic contribution index are substituted into these rules for calculation. During the calculation process, the changes in the index at different time points are fully considered, allowing the calculated determination threshold to be adjusted in real time according to the dynamic changes of the index. This ensures that the threshold accurately adapts to the characteristics of alarm event data at different stages, ultimately resulting in a consensus-integrated determination threshold that comprehensively reflects clinical standard requirements and actual data conditions.

[0156] The parameter determination unit pre-divides the data into multiple anomaly severity levels, each corresponding to a clearly defined numerical range. Then, the time-series dynamic contribution index at each time point is compared with the corresponding consensus integration threshold to determine whether the index is higher or lower than the threshold, and by what extent. Based on the comparison results and referring to the preset level division standards, the anomaly level corresponding to the alarm event data at each time point is determined. The level results for all time points are then summarized and organized to form a significant anomaly level result that comprehensively reflects the changes in the anomaly severity of the alarm event data.

[0157] The root cause parameter determination unit first clarifies the topological space composition of the causal structure, including each parameter node and the associated paths between nodes. Then, for each parameter node, the salience anomaly level is accurately mapped to the corresponding node position in the topological space. Simultaneously, based on the relationships between nodes, the anomaly level information of different nodes is correlated and presented. Through this correlation operation, the scattered salience anomaly level data is transformed into a visual map combined with the causal structure. This map clearly shows the degree of anomaly contribution of each parameter node and the correlation of anomaly contributions between different nodes, ultimately forming an anomaly contribution hierarchy map.

[0158] The root cause parameter determination unit is based on an abnormal contribution hierarchy map, combined with clinically known physiological metabolic pathways and pathological change transmission pathways. Starting from nodes with high abnormal contribution levels in the map, it traces back to their associated upstream parameter nodes. During the tracing process, it focuses on parameter nodes that significantly impact the abnormal contribution of downstream nodes and are directly related to physiological and pathological changes. The traced parameter nodes are then selected based on their abnormal contribution level and the strength of their association with core abnormal nodes. The selected high-priority parameter nodes directly related to the abnormal events are then aggregated to form a priority-associated parameter subset.

[0159] The root cause parameter determination unit references the intervention time limits and urgency levels of various monitoring parameter abnormalities in clinical treatment guidelines. It combines the abnormal contribution level of each parameter in the priority-related parameter subset with its impact on the patient's condition, assigning a corresponding urgency weight to each parameter. The weighting fully considers factors such as the risk of disease deterioration if intervention is not timely after parameter abnormality and the parameter's core role in the physiological and pathological process. Parameters with higher urgency are assigned higher weights, and parameters with lower urgency are assigned lower weights. Subsequently, a comprehensive evaluation of all parameters is conducted, considering their weights, abnormal contributions, and correlations with other parameters. Finally, the parameters that accurately reflect the root cause of the abnormal event are determined, forming the root cause parameter determination result, which is then sent to the early warning and emergency response collaboration terminal.

[0160] The formula for calculating the consensus integration threshold is used to calculate the time step. consensus integration judgment threshold The core principle is to integrate prior clinical knowledge, the dynamic contribution of current abnormal monitoring parameters, the time decay effect, and the overall state of multiple parameters to form a quantitative judgment benchmark. This benchmark is then balanced by adjusting coefficients to improve the scientific rigor of the judgment. , , The adjustment coefficients control the influence intensity of the three types of information respectively. This represents the pre-defined total number of independent clinical guidelines. Reflects the overall quantitative contribution of prior clinical knowledge. This represents the total number of currently monitored anomaly parameters. For static weights of parameters, The time-series dynamic contribution index of the parameters is determined by... After normalization and weighted summation, the overall dynamic contribution of the outlier parameters is obtained. The time interval from the first alarm to time t. Controlling the time decay rate, This reflects the continuous cumulative effect of alarms, and , The formula combines the combined effects of time and the overall state of multiple parameters. By weightedly fusing the three types of information, it ultimately arrives at a judgment threshold that takes into account clinical consensus, real-time dynamics, and time effects.

[0161] The beneficial effects include: eliminating inconsistencies in time-dimensional data, enabling horizontal comparison of contribution values, accurately capturing the characteristics of root cause contribution changes, providing precise and unified data support for subsequent significance determination, avoiding judgment bias, utilizing mature clinical experience to provide a scientific and authoritative basis for dynamic threshold calculation and root cause determination, ensuring compliance with clinical logic, improving the clinical applicability and reliability of root cause determination results, calculating dynamic thresholds based on prior clinical knowledge to align with clinical needs and data patterns, avoiding judgment bias of fixed thresholds, improving the accuracy of alarm event data determination, clearly defining the degree of abnormality through hierarchical comparison, avoiding ambiguity in judgment, providing a clear hierarchical basis for subsequent atlas construction and root cause screening, facilitating medical staff to quickly grasp the severity of abnormalities, and intuitively presenting abnormalities by combining abnormal hierarchy and causal topology space. The system analyzes the distribution and correlation of common contributions, helping healthcare professionals quickly identify key abnormal nodes and transmission paths, providing a clear basis for subsequent analysis, and improving the efficiency of root cause analysis. By tracing back through physiological pathways, it accurately filters key correlation parameters, eliminates irrelevant interference, focuses on core factors, and provides a precise parameter range for subsequent urgency weighting and root cause determination, improving the targeting and efficiency of the determination. It assigns urgency weights to clinical interventions to highlight key parameters, ensuring that the root cause determination results align with intervention needs. This facilitates healthcare professionals in quickly determining priority intervention parameters, guiding emergency response implementation, and improving the efficiency and targeting of emergency care. It comprehensively calculates thresholds based on multiple factors, ensuring they conform to clinical patterns, adapt to real-time data and duration characteristics, and consider the overall state of the data flow, providing a scientific basis for significance determination, avoiding threshold bias, and improving the accuracy and reliability of the determination.

[0162] The scheme matching module 105 is used to perform strategy matching on a preset emergency clinical protocol library based on the root cause parameter determination results, so as to obtain a standardized intervention scheme for the early warning and emergency collaborative terminal.

[0163] In this embodiment of the invention, when the scheme matching module performs strategy matching on a preset emergency clinical protocol library based on the root cause parameter determination result to obtain a standardized intervention scheme for the early warning and emergency coordination terminal, it is specifically used for:

[0164] The clinical dimension analysis of the root cause parameter determination results is performed to obtain the clinical intervention focus of the root cause parameter determination results;

[0165] Based on the clinical intervention focus, a protocol path traversal is performed on the preset emergency clinical protocol library to obtain candidate clinical intervention paths for the clinical intervention focus.

[0166] The candidate clinical intervention pathways are quantified using multidimensional factors to obtain their urgency priority and confidence assessment values.

[0167] Based on the urgency priority score and the confidence assessment value, the candidate clinical intervention pathways are weighted and ranked to obtain the intervention pathway sequence of the candidate clinical intervention pathways;

[0168] Based on the clinical operation constraint rules in the preset emergency clinical protocol library, the intervention path sequence is fused to obtain the optimized intervention strategy of the emergency clinical protocol library.

[0169] The optimized intervention strategy is deployed in a clinical operation instruction manner to obtain a standardized intervention plan for the early warning and emergency rescue collaborative terminal.

[0170] First, a comprehensive collection of all abnormal information related to the root cause parameter determination results is gathered, including the specific manifestations of the abnormal parameters, the associated physiological systems, and the severity of the abnormality. Then, from a practical clinical application perspective, multiple analytical dimensions are defined. These dimensions include the lesion site corresponding to the parameter, the potential health risks that the abnormality may cause, and common intervention directions for such parameter abnormalities in clinical practice. Following these defined dimensions, the root cause parameter determination results are comprehensively and meticulously analyzed. Through this progressive analytical process, the core issues requiring clinical intervention are precisely identified, thereby determining the focus of clinical intervention.

[0171] The core characteristics of the clinical intervention focus are clearly defined. These characteristics include the specific type of abnormality to be addressed, the range of physiological systems involved, and the urgency of the required intervention. Based on these core characteristics, specific search criteria are then established. A comprehensive search is conducted from a pre-designed emergency clinical protocol database, which stores complete intervention protocols for various clinical conditions. Each protocol includes specific intervention steps and procedures, i.e., protocol paths. During the search, each protocol in the database is meticulously compared, and all protocols that match the core characteristics of the clinical intervention focus are selected. The corresponding intervention procedures for these protocols are then extracted, ultimately forming candidate clinical intervention paths.

[0172] Several key elements for quantification were identified, including the time required for intervention implementation, the complexity of the intervention itself, the expected improvement effect of the intervention on abnormal parameters, the potential risks during the intervention, and the actual frequency of clinical application of the pathway. For each candidate clinical intervention pathway, each of these elements was evaluated individually. A reasonable scoring range was established for each element, combining standard clinical practice data and long-term accumulated data. Then, each candidate pathway was scored on its performance across each element according to the established scoring range, transforming the qualitative descriptions into intuitively comparable quantitative values. Based on the scores for each element, a comprehensive calculation was performed to determine the urgency priority of each candidate pathway, reflecting the urgent need for its implementation in the current clinical situation; simultaneously, a confidence assessment value was derived, reflecting the reliability and expected outcome assurance level of the pathway in clinical application.

[0173] Based on the actual clinical needs of critical care, a reasonable weighting of emergency priority and confidence assessment values ​​was determined. Considering the time-sensitive nature of emergency care for critically ill patients, the weight of emergency priority was higher than that of confidence assessment values, ensuring that urgent and effective intervention pathways were given priority consideration. For each candidate clinical intervention pathway, its emergency priority score was multiplied by its corresponding weight, and then its confidence assessment value was multiplied by its corresponding weight. The two products were then added together to obtain a comprehensive score for each candidate pathway. All candidate clinical intervention pathways were ranked in descending order of comprehensive score to form an intervention pathway sequence, with pathways having higher comprehensive scores appearing earlier in the sequence.

[0174] Clinical operation constraints were extracted from a pre-defined emergency clinical protocol library. These rules cover the required sequence of clinical operations, contraindications between different interventions, and operational limitations specific to a patient's condition. Each intervention pathway in the sequence was then analyzed in detail to identify its core intervention steps, applicable conditions, and advantages. Based on strict adherence to these constraints, the highest-priority pathways in the sequence were used as the core framework. Subsequent intervention steps that did not conflict with or could supplement this core framework were then incorporated. For conflicting interventions across different pathways, the better intervention was retained based on the constraints and actual clinical outcomes. Redundant or repetitive steps were merged and simplified to form an optimized intervention strategy.

[0175] Each intervention step in the optimized intervention strategy is meticulously broken down, transforming each step into specific clinical operational instructions. Each instruction clearly defines the role of the medical personnel executing it, such as doctors or nurses; specifies the details of the operation, including the site, method, dosage, and frequency; clarifies the execution timeframes to ensure compliance with the intervention sequence; and outlines precautions and quality control standards to guarantee accuracy and safety. All these specific clinical operational instructions are systematically organized and integrated according to the intervention sequence, forming a clearly structured, concise, and directly clinically applicable standardized intervention plan. This plan is directly adaptable to the needs of early warning and emergency response collaboration terminals, facilitating rapid understanding and execution by medical personnel.

[0176] The beneficial effects include: accurately identifying the focus of clinical intervention, avoiding deviations in intervention direction caused by single-dimensional analysis, laying an accurate foundation for matching subsequent intervention plans, quickly screening candidate intervention paths that fit the current clinical situation, ensuring the relevance of the paths, avoiding the omission of optimal paths, providing ample selection space for plan optimization, transforming abstract candidate paths into comparable numerical indicators, clearly presenting path differences, providing objective basis for weighted ranking, reducing subjective judgment bias, highlighting urgent and reliable intervention paths, ensuring that the path sequence meets the core needs of critical care, providing clear priorities for strategy integration, improving emergency response efficiency and effectiveness, absorbing the advantages of each candidate path, avoiding intervention operation conflicts and redundancies, forming a scientific and efficient optimized intervention strategy that is better adapted to the current clinical situation, transforming the optimized strategy into standardized operating instructions, reducing the difficulty of understanding and execution for medical staff, avoiding operational deviations, ensuring accurate implementation of the plan, and adapting to emergency collaborative terminals to improve emergency collaborative efficiency.

[0177] The instruction orchestration module 106 is used to parse the clinical operation logic in the standardized intervention plan, obtain the emergency execution process of the standardized intervention plan, and orchestrate the emergency execution process to obtain the team collaboration instruction set of the early warning and emergency collaboration terminal.

[0178] In this embodiment of the invention, when the instruction orchestration module executes and parses the clinical operation logic in the standardized intervention plan to obtain the emergency execution flow of the standardized intervention plan, and orchestrates the emergency execution flow to obtain the team collaboration instruction set of the early warning and emergency collaboration terminal, it is specifically used for:

[0179] The standardized intervention protocol is deconstructed into clinical operational logic to obtain a sequence of logical elements of the standardized intervention protocol;

[0180] Analyze the temporal and logical dependencies of the logical element sequence to obtain the dependencies of the standardized intervention scheme;

[0181] Based on the aforementioned dependencies, a directed graph structure is constructed from the sequence of logical elements to obtain the topology graph of the emergency steps of the standardized intervention scheme.

[0182] Role capability mapping is performed on the topology of the emergency rescue steps to obtain the preliminary task assignment table of the early warning and emergency rescue collaboration terminal;

[0183] Spatiotemporal resource conflicts are resolved in the preliminary task assignment table to obtain the collaborative task scheduling scheme of the early warning and emergency rescue collaborative terminal;

[0184] Multimodal instruction synthesis is performed on the task timing and role binding information in the collaborative task scheduling scheme to obtain the team collaboration instruction set of the early warning and emergency rescue collaborative terminal.

[0185] Standardized intervention protocols encompass a series of coherent operational requirements surrounding critical care. To accurately break down the core execution elements, the protocol needs to be deconstructed according to its clinical operational logic. During this deconstruction, based on the differences in the nature of emergency procedures and the interconnectedness of execution stages, each core action, prerequisite, operational requirement, and outcome assessment standard involved in the protocol is broken down one by one. This ensures that each component has a clear and independent function and defined scope. Then, according to the original clinical execution logic sequence in the protocol, these specific parts are arranged in an orderly manner, ultimately forming a logical element sequence.

[0186] For the established sequence of logical elements, a temporal and logical dependency analysis is conducted. First, the clinical procedures corresponding to each logical element are systematically reviewed to clarify which procedure must be performed before and after the procedure corresponding to another element during emergency treatment, precisely defining the temporal order of each element. Then, the relationships and constraints between the elements are further examined to determine whether there are logical connections such as the completion of the procedure corresponding to one element before the procedure corresponding to another can be initiated, or the result of the procedure corresponding to one element directly affects the procedure corresponding to another element. These temporal sequences and logical constraints are systematically organized to ultimately clarify the dependencies of the standardized intervention plan.

[0187] Based on the previously defined dependencies, a directed graph structure is constructed. Each element in the logical element sequence is treated as an independent node. According to the execution order of each element, a dedicated connection line is drawn from the first element node to the subsequent element nodes. The direction of the line directly corresponds to the order of operations, ensuring that the connection line accurately reflects the temporal dependencies between elements. At the same time, all element nodes are comprehensively reviewed to ensure that each node can establish a complete connection with other related nodes based on the actual dependencies, without omitting any associations. The final result is a topology diagram that can intuitively present the execution order and relationships of each element.

[0188] First, comprehensively review all roles involved in the early warning and emergency rescue collaboration terminal, clarifying the professional skill areas, clinical operation authority, actual execution capability range, and past experience in performing related tasks for each role. Then, compare each emergency rescue step in the emergency rescue step topology diagram, deeply analyze the key information such as the operational difficulty, professional technical requirements, and required operational qualifications of each step, and select suitable roles with the ability to perform the step based on the capabilities of each role. Subsequently, clearly associate and bind each emergency rescue step with the corresponding suitable role. Finally, systematically organize the association results of all steps and roles according to the order of emergency rescue steps to form a preliminary task assignment table.

[0189] A comprehensive and detailed review of the preliminary task assignment table is conducted, focusing on identifying conflicts such as the same role being assigned multiple tasks requiring execution within the same time period, multiple tasks simultaneously requesting and occupying necessary medical facilities or specialized equipment, and insufficient supply of critical medicines and consumables for emergency care to meet the concurrent needs of multiple tasks. For any conflicts identified, the execution time of some tasks is adjusted based on their urgency and importance, or conflicting tasks are reassigned to other roles with the appropriate capabilities. Existing resources can also be optimized, prioritizing the resource needs of high-priority tasks to ensure all tasks can be executed smoothly within a reasonable timeframe and with available resources, ultimately forming a collaborative task scheduling solution.

[0190] From the collaborative task scheduling scheme, the system extracts the specific execution time of each task, the sequence of tasks, and other timing information, as well as the role binding information such as the execution role, specific operation requirements, and precautions for each task. Combining the actual needs of critical care scenarios and the work habits of different roles, this extracted information is transformed into various forms of instructions, such as text prompts, voice broadcasts, and terminal pop-up reminders, to ensure that medical staff can quickly obtain relevant task information in a way that suits them, regardless of their work status. Finally, according to the task relationships and the classification of execution roles, all the various forms of instructions are systematically organized and categorized to form a complete team collaboration instruction set.

[0191] The beneficial effects include transforming abstract, standardized intervention plans into a clear sequence of logical elements, avoiding logical confusion in subsequent steps, laying the foundation for streamlining emergency procedures, clearly understanding the timing and logical constraints of each logical element, avoiding chaotic timing or logical conflicts in the emergency procedure, ensuring the scientific and rational nature of the procedure, transforming abstract dependencies into intuitive visual graphics, facilitating medical staff's understanding of the emergency procedure, reducing the difficulty of subsequent task assignment, achieving a scientific match between emergency steps and execution roles, avoiding operational errors or inefficiencies due to mismatched capabilities, laying the foundation for efficient emergency response in personnel allocation, eliminating time, space, and resource conflicts in task execution in advance, ensuring the orderly and smooth progress of emergency tasks, buying time for critically ill patients, providing diversified methods for obtaining instructions, meeting the information needs of different roles in emergency scenarios, improving teamwork and tacit understanding, and ensuring the efficient and orderly progress of the emergency procedure.

[0192] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0193] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multi-parameter real-time early warning and emergency care coordination system for critically ill patients, characterized in that, The system includes an anomaly detection module, a causal analysis module, a contribution tracing module, a root cause determination module, a solution matching module, and an instruction orchestration module, wherein: The anomaly detection module is used to perform real-time anomaly detection on the multi-parameter monitoring data stream in the medical information terminal and obtain alarm event data of the multi-parameter monitoring data stream. The causal analysis module is used to perform structural analysis on the predefined relationship topology based on the alarm event data to obtain the causal structure of the multi-parameter monitoring data stream; The contribution tracing module is used to trace the causal influence of the causal structure and obtain the root cause contribution ranking of each abnormal parameter in the alarm event data. The root cause determination module is used to determine the significance of preset clinical standards based on the root cause contribution ranking, and obtain the root cause parameter determination result of the early warning and emergency rescue collaboration terminal. The scheme matching module is used to perform strategy matching on a preset emergency clinical protocol library based on the root cause parameter determination results, so as to obtain a standardized intervention scheme for the early warning and emergency collaborative terminal. The instruction orchestration module is used to parse the clinical operation logic in the standardized intervention plan, obtain the emergency execution process of the standardized intervention plan, and orchestrate the emergency execution process to obtain the team collaboration instruction set of the early warning and emergency collaboration terminal.

2. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 1, characterized in that, When the anomaly detection module performs real-time anomaly detection on the multi-parameter monitoring data stream in the medical information terminal and obtains alarm event data from the multi-parameter monitoring data stream, it is specifically used for: The multi-parameter monitoring data stream in the medical information terminal is divided into sliding window segments to obtain the parameter time sequence segments of the multi-parameter monitoring data stream; Adaptive threshold learning is performed on the time series segments of the parameters to obtain the dynamic anomaly boundaries of the time series segments of the parameters; The parameter time series segment and the corresponding dynamic anomaly boundary are dynamically delimited to obtain the preliminary anomaly marker point of the parameter time series segment; Clinical time-series pattern matching is performed on the preliminary abnormality markers, and non-pathological pseudo-abnormality markers are filtered out to obtain the abnormal events of the preliminary abnormality markers. The abnormal events are aggregated for multi-parameter clinical correlation to obtain alarm event data from the multi-parameter monitoring data stream.

3. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 1, characterized in that, When the causal analysis module performs structural analysis on the predefined relationship topology based on the alarm event data to obtain the causal structure of the relationship between the multi-parameter monitoring data streams, it is specifically used for: Alarm event mapping is performed on the predefined relationship topology to obtain the activation parameter nodes of the predefined relationship topology; Based on the activated parameter nodes, directed edge retrieval is performed on the predefined relation topology to obtain the inter-node association edges of the activated parameter node set; Temporal dependency analysis is performed on the edges connecting the nodes within a time window to obtain the causal direction confidence of the edges connecting the nodes. Based on the causal direction confidence, directional constraints are applied to the inter-node association edges to obtain directed causal edges of the inter-node association edges; By assigning causal weights to the directed causal edges, a weighted directed causal graph of the directed causal edges is obtained. By mining the maximum causal flow path in the weighted directed causal graph, the causal structure of the relationship between the multi-parameter monitoring data streams is obtained.

4. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 3, characterized in that, When the causal analysis module performs temporal dependency analysis on the inter-node association edges within a time window to obtain the causal direction confidence of the inter-node association edges, it is specifically used for: Extract a specified time window from the alarm event data to obtain a multi-parameter time sequence segment of the alarm event data; Delay mutual information is measured on the multi-parameter time segment to obtain the delay mutual information matrix between each pair of parameters in the multi-parameter time segment. An asymmetric statistical test is performed on the row and column elements of the delayed mutual information matrix to obtain the time-series anticipation measure between the parameters of the delayed mutual information matrix. The temporal precedence measure is consistently fused with a pre-built clinical causal knowledge base to obtain the causal direction confidence of the associated edges between the nodes.

5. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 1, characterized in that, When the contribution tracing module performs causal impact tracing on the causal structure to obtain the root cause contribution ranking of each abnormal parameter in the alarm event data, it is specifically used for: The causal structure is deconstructed into directed paths to obtain the directed path sequence of the causal structure; Based on the directed path sequence, abnormal node mapping is performed on the alarm event data to obtain the associated nodes of the alarm event data; Influence topology construction is performed on the associated nodes to obtain the influence transmission graph of the associated nodes; By tracing back the source nodes of the influence transmission graph, the potential root cause node clusters of the influence transmission graph are obtained. The potential root cause node clusters are aggregated by contribution metric to obtain a preliminary contribution list of the abnormal parameters; The root cause contribution ranking of the alarm event data is obtained by comparing and calibrating the preliminary contribution list with the preset clinical significance threshold.

6. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 5, characterized in that, When the contribution tracing module performs contribution quantification aggregation on the potential root cause node cluster to obtain a preliminary contribution list of the abnormal parameters, it is specifically used for: The potential root cause node cluster is weighted to obtain the initial influence weight value of the potential root cause node. Based on the directed path sequence, the initial influence weight value is propagated using path dependency to obtain the propagation influence score of the abnormal parameter. The propagation impact score is scaled to form a standardized contribution score for the anomaly parameter; The standardized contribution scores are sorted and integrated to obtain a preliminary contribution list of the abnormal parameters.

7. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 3, characterized in that, When the root cause determination module performs a significance determination based on preset clinical standards according to the root cause contribution ranking to obtain the root cause parameter determination result of the early warning and emergency rescue collaboration terminal, it is specifically used for: The contribution values ​​in the root cause contribution ranking are subjected to time-series normalization to obtain the time-series dynamic contribution index of the multi-parameter monitoring data stream. Extract the fusion clinical prior knowledge from the preset clinical standards; Based on the aforementioned fusion of prior clinical knowledge, a dynamic threshold is calculated for the time-series dynamic contribution index to obtain the consensus integration and judgment threshold for the alarm event data. The time-series dynamic contribution index and the consensus integration judgment threshold are compared hierarchically to obtain the significant anomaly level of the alarm event data; The significant anomaly hierarchy is mapped to the topological space of the causal structure to obtain the anomaly contribution hierarchy map of the alarm event data; By performing physiological path backtracking on the abnormal contribution hierarchy map, a subset of priority association parameters of the alarm event data is obtained; The clinical intervention urgency is assigned to the subset of priority-related parameters to obtain the root cause parameter determination results of the early warning and emergency rescue collaboration terminal.

8. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 7, characterized in that, When the root cause determination module performs dynamic threshold calculation on the time-series dynamic contribution index based on the fused clinical prior knowledge to obtain the consensus integration determination threshold for the alarm event data, the specific formula for calculating the consensus integration determination threshold is as follows: ； In the formula, For a moment The consensus integration determination threshold, The first adjustment coefficient, This is the second adjustment coefficient. This is the third adjustment coefficient. This constitutes the total number of independent clinical guidelines that integrate prior clinical knowledge. To integrate the first clinical prior knowledge in the above-mentioned Independent clinical guidelines A function for quantitative rule extraction. This represents the total number of monitoring parameters in the current alarm event data that are in an abnormal state. For the index of the abnormal parameter, To index based on pathophysiological importance The corresponding abnormal parameters are pre-assigned dimensionless static weights. For index The corresponding abnormal parameters at time The aforementioned time-series dynamic contribution index, This is a function that performs a nonlinear normalized mapping on the input time-series dynamic contribution index. The time from the occurrence of the first alarm event to the current time. The time interval experienced The preset time decay coefficient, This is an S-shaped growth curve function. This is a dimensionless metric value calculated based on the multi-parameter monitoring data stream.

9. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 1, characterized in that, When the scheme matching module performs strategy matching on a preset emergency clinical protocol library based on the root cause parameter determination result to obtain a standardized intervention scheme for the early warning and emergency coordination terminal, it is specifically used for: The clinical dimension analysis of the root cause parameter determination results is performed to obtain the clinical intervention focus of the root cause parameter determination results; Based on the clinical intervention focus, a protocol path traversal is performed on the preset emergency clinical protocol library to obtain candidate clinical intervention paths for the clinical intervention focus. The candidate clinical intervention pathways are quantified using multidimensional factors to obtain their urgency priority and confidence assessment values. Based on the urgency priority score and the confidence assessment value, the candidate clinical intervention pathways are weighted and ranked to obtain the intervention pathway sequence of the candidate clinical intervention pathways; Based on the clinical operation constraint rules in the preset emergency clinical protocol library, the intervention path sequence is fused to obtain the optimized intervention strategy of the emergency clinical protocol library. The optimized intervention strategy is deployed in a clinical operation instruction manner to obtain a standardized intervention plan for the early warning and emergency rescue collaborative terminal.

10. The multi-parameter real-time early warning and emergency care coordination system for critically ill patients as described in claim 1, characterized in that, When the instruction orchestration module executes and parses the clinical operation logic in the standardized intervention plan to obtain the emergency execution flow of the standardized intervention plan, and orchestrates the emergency execution flow to obtain the team collaboration instruction set of the early warning and emergency collaboration terminal, it is specifically used for: The standardized intervention protocol is deconstructed into clinical operational logic to obtain a sequence of logical elements of the standardized intervention protocol; Analyze the temporal and logical dependencies of the logical element sequence to obtain the dependencies of the standardized intervention scheme; Based on the aforementioned dependencies, a directed graph structure is constructed from the sequence of logical elements to obtain the topology graph of the emergency steps of the standardized intervention scheme. Role capability mapping is performed on the topology of the emergency rescue steps to obtain the preliminary task assignment table of the early warning and emergency rescue collaboration terminal; Spatiotemporal resource conflicts are resolved in the preliminary task assignment table to obtain the collaborative task scheduling scheme of the early warning and emergency rescue collaborative terminal; Multimodal instruction synthesis is performed on the task timing and role binding information in the collaborative task scheduling scheme to obtain the team collaboration instruction set of the early warning and emergency rescue collaborative terminal.

Images