Dynamic evaluation device for enteral nutrition tolerance of critical patient

Abstract

The invention relates to the technical field of intelligent medical treatment, and discloses a critical patient enteral nutrition tolerance dynamic evaluation device, which comprises a multi-source data acquisition module, a causal topology construction module, a causal path tracking module, a risk quantitative evaluation module, an early warning traceability module and a decision fusion module, synchronously collecting enteral nutrition and key physiological parameters of the critical patient, and obtaining multi-dimensional time sequence data; a causal topology is constructed for the multi-dimensional time sequence data, and a time-varying causal inference map is obtained; taking a preset gastrointestinal intolerance core index as a root node, tracking the map topological structure, and obtaining a key causal path; quantitatively evaluating the key causal path risk to obtain a path risk measure; interpreting a path risk measure threshold, generating a causal early warning signal, and reversely tracing to determine a priority intervention target; fusing the early warning signal, the key path, the risk measure and the intervention target, and outputting a clinical decision basis; according to the method, the efficiency of dynamic evaluation of the enteral nutrition tolerance of the critical patient can be improved.

Description

Technical Field

[0001] This invention relates to the field of smart medical technology, and in particular to a device for dynamically assessing the enteral nutrition tolerance of critically ill patients. Background Technology

[0002] Enteral nutrition is a key supportive treatment for critically ill patients, but its tolerability is dynamically affected by a variety of physiological factors. Current clinical practice mainly relies on intermittent observation and experience-based judgment of individual indicators by medical staff, lacking systematic and continuous objective data support. The collection of enteral nutrition infusion data and other multi-source physiological parameters such as vital signs is often independent and asynchronous in time, and existing analytical methods are unable to reveal the complex time-varying causal relationships between these heterogeneous time-series data, resulting in a delayed and superficial assessment of tolerability risks.

[0003] Existing technologies face two main limitations in achieving precise and prospective risk management of enteral nutrition tolerance in critically ill patients: First, current data acquisition methods struggle to achieve high-precision time synchronization and lossless fusion of dynamic enteral nutrition infusion with multi-dimensional key physiological parameters, leading to inherent time biases and insufficient information completeness in the data foundation upon which subsequent analyses are based. Second, existing analytical models generally lack the ability to effectively mine and structurally represent the complex, time-varying causal relationships inherent in multi-source heterogeneous time-series data, failing to construct causal reasoning maps that truly reflect the internal logic and dynamic evolution of physiological processes. This limits the depth and accuracy of risk warning and intervention decisions. Therefore, improving the efficiency of dynamic assessment of enteral nutrition tolerance in critically ill patients has become an urgent problem to be solved. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides a dynamic assessment device for enteral nutrition tolerance in critically ill patients, characterized in that the device comprises a multi-source data acquisition module, a causal topology construction module, a causal path tracing module, a risk quantification assessment module, an early warning and source tracing module, and a decision fusion module, wherein:

[0005] The multi-source data acquisition module is used to simultaneously acquire enteral nutrition and key physiological parameters of critically ill patients from multiple sources, thereby obtaining multi-dimensional time-series data of the critically ill patients.

[0006] The causal topology construction module is used to construct causal topology on the multi-dimensional time series data to obtain the time-varying causal inference graph of the multi-dimensional time series data.

[0007] The causal path tracing module is used to trace the causal path of the topology in the time-varying causal reasoning graph with a preset gastrointestinal intolerance core index as the root node, and to obtain the key causal path of the time-varying causal reasoning graph.

[0008] The risk quantification assessment module is used to quantify the path risk of the key causal path and obtain the path risk measure of the key causal path.

[0009] The early warning and tracing module is used to perform threshold judgment on the path risk measurement, obtain the causal early warning signal of the path risk measurement, and perform reverse causal tracing on the key causal path corresponding to the causal early warning signal to obtain the priority intervention target of the causal early warning signal.

[0010] The decision fusion module is used to perform decision fusion on the causal early warning signal, the key causal path, the path risk measurement, and the priority intervention target to obtain the clinical decision basis for the critically ill patient.

[0011] In a preferred embodiment, when the multi-source data acquisition module performs multi-source synchronous acquisition of enteral nutrition and key physiological parameters of critically ill patients to obtain multi-dimensional time-series data of the critically ill patients, it is specifically used for:

[0012] The dynamics of enteral nutrition infusion in critically ill patients were recorded in a time-series non-destructive manner to obtain a refined time series of the enteral nutrition infusion dynamics.

[0013] The key physiological parameters of the critically ill patients were monitored holographically to obtain the time series of the key physiological parameters;

[0014] The refined time series is timestamped with the time series to obtain the synchronized time stream of the critically ill patient;

[0015] The original monitoring values ​​in the synchronized time series are standardized in terms of dimensions to obtain the standardized value set of the synchronized time series.

[0016] The standardized numerical set is discretized to obtain the standardized numerical distribution of the synchronized time-series stream;

[0017] By integrating the indicator data from the standardized numerical distribution, multi-dimensional time-series data of the critically ill patients are obtained.

[0018] In a preferred embodiment, when the causal topology construction module performs causal topology construction on the multi-dimensional time series data to obtain the time-varying causal inference graph of the multi-dimensional time series data, it is specifically used for:

[0019] Explore the causal structure of the multi-dimensional time series data and extract the candidate causal edge set of the multi-dimensional time series data;

[0020] Dynamic causal efficacy quantification is performed on the candidate causal edge set to obtain the time-varying causal strength matrix of the candidate causal edge set;

[0021] By performing topological derivation on the time-varying causal intensity matrix, a time-varying causal reasoning graph of the time-varying causal intensity matrix is ​​obtained.

[0022] In a preferred embodiment, when the causal topology construction module performs topological derivation on the time-varying causal intensity matrix to obtain the time-varying causal inference graph of the time-varying causal intensity matrix, it is specifically used for:

[0023] Based on the preset judgment criteria, a set of valid causal edges is selected from the time-varying causal intensity matrix;

[0024] Reconstruct the directed associations of the effective causal edge set to obtain the directed causal skeleton of the effective causal edge set;

[0025] Cross-window topological aggregation is performed on the directed causal skeleton to obtain the condensed topological pattern of the directed causal skeleton;

[0026] The time-varying causal structure in the condensed topological mode is fused with a graph structure to obtain the fused graph skeleton of the time-varying causal structure.

[0027] Time-varying attribute annotation is performed on the fusion graph skeleton to obtain the time-varying causal inference graph of the time-varying causal intensity matrix.

[0028] In a preferred embodiment, when the causal path tracing module performs causal path tracing on the topology of the time-varying causal inference graph with a preset gastrointestinal intolerance core index as the root node to obtain the key causal paths of the time-varying causal inference graph, it is specifically used for:

[0029] The preset core index of gastrointestinal intolerance is set as the root node, and the root node is used as the starting point to perform directed causal tracing of the topology of the time-varying causal reasoning graph to obtain the penetration causal path of the time-varying causal reasoning graph.

[0030] Determine the consistency between the causal direction and temporal logic between nodes in the penetration causal path to obtain the effective causal path subset of the penetration causal path;

[0031] The contribution of the effective causal path subset is quantified to obtain the contribution degree of the effective causal path subset;

[0032] The contribution degree and the temporal span information of the effective causal path subset are weighted and fused to obtain the path contribution priority sequence of the contribution degree and the temporal span information;

[0033] The key causal paths of the time-varying causal inference graph are obtained by performing saliency screening on the path contribution priority sequence.

[0034] In a preferred embodiment, when the risk quantification assessment module performs a quantification assessment of the path risk of the key causal path to obtain a path risk measure for the key causal path, it is specifically used for:

[0035] Analyze the node state time sequence of the key causal path to obtain the abnormal state of the node state time sequence;

[0036] Based on the abnormal state, the time-varying effect of the causal path edges of the nodes in the key causal path is quantified to obtain the dynamic influence intensity of the causal path edges.

[0037] By performing time-weighted aggregation on the abnormal state and the intensity of the dynamic impact, the risk evolution sequence of the key causal path is obtained;

[0038] By performing contour fitting on the risk evolution sequence, the comprehensive risk evolution trajectory of the key causal path is obtained;

[0039] Based on preset risk assessment rules, peak period detection is performed on the comprehensive risk evolution trajectory to obtain the risk peak period interval of the comprehensive risk evolution trajectory;

[0040] Feature extraction is performed on the risk peak period interval to obtain the temporal features of the risk peak period interval;

[0041] The peak intensity and duration of the time-domain features are quantitatively characterized to obtain the path risk measure of the key causal path.

[0042] In a preferred embodiment, the formula for calculating the intensity of the dynamic influence is as follows:

[0043] ;

[0044] In the formula, In time From the nodes in the key causal path To the node The intensity of the dynamic influence, Based on the nodes in the key causal path The function that calculates the abnormal state. For the nodes in the key causal path In time The node state timing values, For the nodes in the key causal path The average state value over the reference period. For the nodes in the key causal path The standard deviation of the state values ​​over the reference period. Based on the nodes in the key causal path The function that calculates the abnormal state. For the nodes in the key causal path In time The node state timing values, For the nodes in the key causal path The average state value over the reference period. For the nodes in the key causal path The standard deviation of the state values ​​over the reference period. For a function that takes a time delay parameter as input, In time From the nodes in the key causal path To the node The time delay parameter or time-dependent parameter for the transmission of causal effects.

[0045] In a preferred embodiment, when the early warning and tracing module performs threshold judgment on the path risk measure to obtain a causal early warning signal of the path risk measure, and performs reverse causal tracing on the key causal path corresponding to the causal early warning signal to obtain the priority intervention target of the causal early warning signal, it is specifically used for:

[0046] Threshold identification is performed on the path risk measure to obtain the risk index of the path risk measure;

[0047] The statistical distribution of core physiological parameters in the multi-dimensional time-series data is analyzed to determine the parameter-specific dynamic range of the critically ill patients;

[0048] Within the parameter-specific dynamic range, a multi-level risk interval corresponding to the parameter within the parameter-specific dynamic range is calibrated.

[0049] Extract the boundary values ​​of the multi-level risk intervals to obtain the dynamic risk threshold set of the critically ill patient;

[0050] Establish the correspondence between the range of parameter values ​​in the dynamic risk threshold set and the risk level;

[0051] The correspondence is formalized into an executable set of judgment rules to construct an individualized risk spectrum for the critically ill patients;

[0052] The risk indicators are assigned to the individualized risk spectrum to obtain the risk level of the risk indicators.

[0053] The risk attribution level is signal-encoded to obtain the causal early warning signal of the path risk measure;

[0054] Based on the risk spatiotemporal information indicated by the causal early warning signal, risk path segments are located from the key causal paths;

[0055] Reconstruct the causal transmission relationship between each node in the risk path segment to obtain the abnormal transmission network of the key causal path;

[0056] Based on the risk attribution level and the topology of the abnormal transmission network, the intervention effectiveness of the nodes in the abnormal transmission network is evaluated to determine the priority score of the abnormal transmission network.

[0057] Based on the priority score, starting from the root node, the abnormal transmission network is traversed in reverse to obtain the candidate key nodes of the abnormal transmission network.

[0058] The node with the highest priority score among the candidate key nodes is selected as the priority intervention target for the causal early warning signal.

[0059] In a preferred embodiment, when the early warning and tracing module performs an evaluation of the intervention effectiveness of each node in the abnormal transmission network to obtain a priority score for the abnormal transmission network, it is specifically used for:

[0060] Analyze the topological connectivity of the anomalous conduction network to obtain the causal edge set of the anomalous conduction network;

[0061] Based on the causal edge set, the causal control capability of nodes in the abnormal transmission network on downstream nodes is quantified to obtain the quantification result of the node influence of the abnormal transmission network.

[0062] The node impact quantification results are fused with the node historical response archives of the abnormal transmission network to obtain the comprehensive node characteristics of the abnormal transmission network.

[0063] Based on the comprehensive characteristics of the nodes, the potential blocking strength of the nodes in the abnormal transmission network to downstream risk transmission is analyzed to assess the node strength of the abnormal transmission network.

[0064] Based on the node strength, the node blocking effectiveness of the abnormal conduction network is determined, and the node blocking effectiveness is normalized to obtain the priority score of the abnormal conduction network.

[0065] In a preferred embodiment, when the decision fusion module performs decision fusion on the causal early warning signal, the key causal path, the path risk measure, and the priority intervention target to obtain the clinical decision basis for the critically ill patient, it is specifically used for:

[0066] The causal early warning signal, the key causal path, the path risk measure, and the priority intervention target are integrated into a decision factor set for the critically ill patient;

[0067] By analyzing the warning level of the causal warning signal and the topological stability of the key causal path, the timeliness attribute of the critical patient can be obtained;

[0068] Based on the timeliness attribute, the mapping relationship between the path risk measure and the priority intervention target is calibrated to obtain the dynamic intervention rules for the priority intervention target;

[0069] Based on the dynamic intervention rules, the specific operation sequence and objectives of the priority intervention targets are arranged to obtain the targeted intervention plan for the priority intervention targets;

[0070] By integrating the decision factor set, the timeliness attribute, and the targeted intervention plan, the clinical decision-making basis for the critically ill patient is obtained.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] 1. This invention constructs a more complete and reliable data foundation through high-precision synchronous and lossless fusion of multi-source data acquisition. Specifically, it utilizes sensor linkage to achieve time-series lossless recording of the dynamics of enteral nutrition infusion and simultaneously and non-invasively collects key physiological parameters. Then, through millisecond-level timestamp alignment and standardization, it eliminates time deviations and dimensional differences, thereby providing high-quality, comparable multi-dimensional time-series data for subsequent analysis and fundamentally improving the accuracy of the analysis.

[0073] 2. This invention achieves in-depth analysis and precise early warning of risk mechanisms by constructing a time-varying causal reasoning graph and performing dynamic path tracing. Specifically, it first mines the causal structure in the data and quantifies the time-varying causal strength to construct a causal graph that reflects dynamic associations; then, it uses core indicators as roots to trace paths and quantify contributions, screening out key causal paths and assessing their risk measures; finally, it generates early warning signals through threshold interpretation and reverse-traces to determine priority intervention targets. This process transforms complex physiological interactions into a clearly structured, quantifiable, and time-sequential causal network, making early warning and intervention more forward-looking and targeted. Attached Figure Description

[0074] Figure 1 This is a schematic diagram of the architecture of a dynamic assessment device for enteral nutrition tolerance in critically ill patients according to an embodiment of the present invention;

[0075] 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

[0076] 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.

[0077] 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.

[0078] 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)."

[0079] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0080] In practice, the server-side equipment deployed by the dynamic assessment device for enteral nutrition tolerance in critically ill patients may consist of one or more devices. This dynamic assessment device can be implemented as a business instance, a virtual machine, or a hardware device. 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, providing dynamic assessment of enteral nutrition tolerance to various users. 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 users. 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 dynamic assessment of enteral nutrition tolerance to various users.

[0081] In terms of implementation, the dynamic assessment device for enteral nutrition tolerance in critically ill patients and the user terminal are mutually compatible. That is, if the dynamic assessment device for enteral nutrition tolerance in critically ill patients is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the dynamic assessment device for enteral nutrition tolerance in critically ill patients is implemented as a website, then the user terminal is implemented as a webpage; or if the dynamic assessment device for enteral nutrition tolerance in 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.

[0082] like Figure 1 The diagram shown is a device architecture diagram of a dynamic assessment device for enteral nutrition tolerance in critically ill patients provided in an embodiment of the present invention.

[0083] The dynamic assessment device 100 for enteral nutrition tolerance in critically ill patients described in this invention can be installed on a cloud server. In terms of implementation, it can be one or more service devices, or an application installed on the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the dynamic assessment device 100 for enteral nutrition tolerance in critically ill patients may include a multi-source data acquisition module 101, a causal topology construction module 102, a causal path tracing module 103, a risk quantification assessment module 104, an early warning and source tracing module 105, and a decision fusion 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 an electronic device's processor and perform a fixed function, stored in the electronic device's memory.

[0084] In this embodiment of the invention, each of the above-mentioned modules in the dynamic assessment device for enteral nutrition tolerance in critically ill patients can be implemented independently and called upon other modules. Here, "called upon" can be understood as a module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the dynamic assessment device for enteral nutrition tolerance in critically ill patients provided by this embodiment of the invention, the applicable scope of the device 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 dynamic assessment device for enteral nutrition tolerance in critically ill patients. In practical applications, the above-mentioned 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.

[0085] The following describes the components and workflow of the dynamic assessment device for enteral nutrition tolerance in critically ill patients, using specific embodiments as examples:

[0086] The multi-source data acquisition module 101 is used to simultaneously acquire enteral nutrition and key physiological parameters of critically ill patients from multiple sources, thereby obtaining multi-dimensional time-series data of the critically ill patients.

[0087] In this embodiment of the invention, when the multi-source data acquisition module performs multi-source synchronous acquisition of enteral nutrition and key physiological parameters of critically ill patients to obtain multi-dimensional time-series data of the critically ill patients, it is specifically used for:

[0088] The dynamics of enteral nutrition infusion in critically ill patients were recorded in a time-series non-destructive manner to obtain a refined time series of the enteral nutrition infusion dynamics.

[0089] The key physiological parameters of the critically ill patients were monitored holographically to obtain the time series of the key physiological parameters;

[0090] The refined time series is timestamped with the time series to obtain the synchronized time stream of the critically ill patient;

[0091] The original monitoring values ​​in the synchronized time series are standardized in terms of dimensions to obtain the standardized value set of the synchronized time series.

[0092] The standardized numerical set is discretized to obtain the standardized numerical distribution of the synchronized time-series stream;

[0093] By integrating the indicator data from the standardized numerical distribution, multi-dimensional time-series data of the critically ill patients are obtained.

[0094] For the enteral nutrition infusion process of critically ill patients, data acquisition is carried out using a sensor recording method linked to the infusion equipment, ensuring that normal infusion operation is not affected throughout the process and achieving non-destructive data recording. The recorded content covers key dynamic information such as the start and end times of infusion, real-time changes in infusion rate, cumulative infusion volume, and the number of interruptions and duration of each interruption during the infusion process. Fixed short-time acquisition intervals are set to fully capture the infusion dynamics at each time point. Through continuous and stable acquisition, a refined time series containing multiple aspects of infusion details is ultimately formed.

[0095] The monitoring scope is clearly focused on core physiological parameters related to enteral nutrition tolerance, specifically including gastrointestinal indicators, basic vital signs, and metabolic indicators. Non-invasive monitoring devices are used, with sensors deployed at corresponding monitoring sites on the patient's body to avoid additional physical trauma and suffering for critically ill patients. Real-time values ​​of each parameter are continuously collected at fixed time intervals, maintaining stability and continuity during the collection process. All collected physiological parameter values ​​are then organized chronologically to form a time series of key physiological parameters that clearly reflects the changes in physiological parameters over time.

[0096] First, a uniformly formatted time stamp, accurate to the millisecond level, is added to each data point in the refined time series and key physiological parameter time series to ensure the accuracy of time recording and lay the foundation for data pairing. Using the time stamp of one set of time series as a benchmark, data with the same or closest time stamp in another set of time series are retrieved one by one. The infusion dynamic data and physiological parameter data under the same time dimension are paired one by one. For data points with slight time differences, the data correspondence is improved by linear supplementation to ensure that each time node has the corresponding infusion dynamic data and physiological parameter data at the same time, and finally a synchronized time series with consistent time dimension and interconnected data is formed.

[0097] First, the original units of each data item in the synchronized time-series stream are systematically sorted out to clarify the physical meaning and measurement standards of different data. Referring to the clinical standards and data reference ranges recognized in the industry, a unified normalization processing method is adopted for each type of data to convert the original values ​​of each type of data into a unified numerical range of 0-1. During the conversion process, precise calculations are performed strictly according to the actual change range of each data item to ensure that the converted values ​​can not only truly reflect the relative size and change trend of the original data, but also do not change the clinical meaning of the data itself. Finally, a standardized numerical set of all data under the same dimension is obtained.

[0098] First, a comprehensive analysis of the overall distribution of the standardized numerical set is conducted. Combining the data distribution density, dispersion, and the needs of actual clinical applications, a reasonable number of discrete intervals is determined to ensure that the interval divisions not only conform to the data characteristics but also have clinical reference value. The interval range corresponding to each value in the standardized numerical set is determined one by one, and a clear interval identifier is marked for each value. At the same time, detailed statistics are compiled on the distribution information such as the number of values, the proportion of values, the maximum and minimum values ​​within each discrete interval, etc., to form a complete standardized numerical distribution result, clearly presenting the distribution characteristics of each data in different intervals.

[0099] First, the categories of indicators in the standardized numerical distribution are clearly defined, and the inherent relationships between different indicators are identified. A hierarchical integration approach is then adopted for data fusion: data is first aggregated within the same category of indicators to clarify their distribution characteristics and temporal relationships; then, data from different categories of indicators are integrated along the time dimension to ensure that data from various indicators at the same time point correspond to each other. During the integration process, the original distribution characteristics and temporal relationship information of each indicator are fully preserved, without losing key data details, ultimately forming multi-dimensional time-series data containing multiple indicators and exhibiting both time series and distribution characteristics.

[0100] The beneficial effects include: non-invasive recording that does not interfere with treatment; complete preservation of infusion dynamic details; provision of accurate basic data for analyzing the correlation between the infusion process and enteral nutrition tolerance; coverage of key physiological indicators; non-invasive monitoring suitable for critically ill patients; continuous collection to ensure the continuity of physiological data; provision of interference-free physiological state basis for multi-dimensional analysis; elimination of data time bias; synchronization of infusion and physiological data time; laying the foundation for exploring the causal relationship between the two; ensuring the accuracy of subsequent analysis; elimination of differences in data units and scales; making different types of data comparable; providing a unified scale basis for cross-data type analysis; improving the feasibility and accuracy of analysis; simplifying data complexity; intuitively displaying data distribution patterns; facilitating the extraction of causal relationship features; reducing the difficulty of causal topology construction; improving the efficiency of causal identification; achieving organic integration of multi-source data; preserving original data characteristics; forming a comprehensive multi-dimensional dataset; and providing complete data support for subsequent evaluation processes.

[0101] The causal topology construction module 102 is used to construct causal topology on the multi-dimensional time series data to obtain a time-varying causal reasoning graph of the multi-dimensional time series data.

[0102] In this embodiment of the invention, when the causal topology construction module performs causal topology construction on the multi-dimensional time series data to obtain the time-varying causal inference graph of the multi-dimensional time series data, it is specifically used for:

[0103] Explore the causal structure of the multi-dimensional time series data and extract the candidate causal edge set of the multi-dimensional time series data;

[0104] Dynamic causal efficacy quantification is performed on the candidate causal edge set to obtain the time-varying causal strength matrix of the candidate causal edge set;

[0105] By performing topological derivation on the time-varying causal intensity matrix, a time-varying causal reasoning graph of the time-varying causal intensity matrix is ​​obtained.

[0106] When the causal topology construction module performs topological derivation on the time-varying causal intensity matrix to obtain the time-varying causal inference graph of the time-varying causal intensity matrix, it is specifically used for:

[0107] Based on the preset judgment criteria, a set of valid causal edges is selected from the time-varying causal intensity matrix;

[0108] Reconstruct the directed associations of the effective causal edge set to obtain the directed causal skeleton of the effective causal edge set;

[0109] Cross-window topological aggregation is performed on the directed causal skeleton to obtain the condensed topological pattern of the directed causal skeleton;

[0110] The time-varying causal structure in the condensed topological mode is fused with a graph structure to obtain the fused graph skeleton of the time-varying causal structure.

[0111] Time-varying attribute annotation is performed on the fusion graph skeleton to obtain the time-varying causal inference graph of the time-varying causal intensity matrix.

[0112] We comprehensively analyze the potential correlations between various indicators in multi-dimensional time series data, meticulously examine the interaction relationships between different time series data sequences during the time evolution process, and identify which indicators may have causal relationships. From these potential correlations, we further screen out the correlation relationships between indicators that meet the basic correlation characteristics, and integrate these screened correlation relationships to form a candidate causal edge set.

[0113] For each association in the candidate causal edge set, the influence of the association on relevant indicators is continuously monitored at different time stages, taking into account the temporal characteristics of multi-dimensional time series data. According to a unified quantitative standard, this influence is converted into specific values ​​that can be directly used for analysis. The quantitative values ​​corresponding to all candidate causal edges are systematically integrated to form a matrix that can clearly reflect the intensity of causal influence at different time stages.

[0114] Using the quantitative data in the time-varying causal intensity matrix as the core basis, and combining the basic logical rules of causal association, a structural framework that can reflect the causal relationship between indicators is gradually derived and constructed. In the derivation process, the intensity data of different time stages in the matrix are fully integrated to ensure that the constructed structural framework not only conforms to the inherent logic of causal association, but also accurately reflects the characteristics of causal association changing over time and the overall association logic.

[0115] Clear judgment criteria are pre-defined, which cover core content such as the minimum threshold for causal association strength and association stability requirements. According to the set criteria, the quantitative strength data corresponding to each causal edge in the time-varying causal strength matrix is ​​checked one by one to determine whether it meets the conditions for being recognized as a valid causal association. All causal edges that meet the conditions are selected and integrated to form a set of valid causal edges.

[0116] By deeply analyzing the interaction between the two indicators connected by each causal edge in the effective causal edge set, we can clearly determine which indicator is the cause of the change and which indicator is the result of the change, and determine the causal direction of each causal edge. Based on the determined causal direction, we can systematically organize all effective causal edges to build a basic structural framework that can clearly show the causal flow between indicators.

[0117] The entire time series is divided into multiple continuous and non-overlapping time windows according to a preset time length. For the directed causal skeleton in each time window, its structural characteristics are analyzed in detail, including the connection mode of causal edges and the correlation density of indicator nodes. The structural characteristics under all time windows are integrated to extract stable and common topological structures. These common structures are merged and converged to form a cohesive topological pattern.

[0118] For various causal structures in condensed topological models that reflect changes at different time stages, we comprehensively analyze the internal connections and differences between the structures; according to the logical order of time evolution and the internal logic of causal relationships, we systematically integrate these structures; in the integration process, we eliminate conflicting content and redundant information between different structures, and finally form a unified graph structure framework that can fully reflect the overall characteristics of time-varying causal structures.

[0119] For each causal edge and each indicator node in the fusion graph skeleton, the corresponding quantitative data in the time-varying causal strength matrix is ​​retrieved; combined with this quantitative data, the strength characteristics of each causal edge and the state characteristics of each node at different time stages are labeled with attribute information that changes over time; through system labeling, the fusion graph skeleton is endowed with attribute characteristics that can reflect changes in the time dimension, and finally a time-varying causal reasoning graph is formed.

[0120] The beneficial effects are as follows: It accurately identifies potential causal relationship combinations, providing comprehensive and targeted foundational materials for subsequent causal topology construction, avoiding omissions of key relationships, ensuring the comprehensiveness of causal analysis, accurately capturing dynamic changes in causal relationships, and forming a matrix that clearly presents the influence intensity of each candidate causal edge at different times. This provides precise quantitative evidence for topology derivation, ensuring the graph reflects causal changes over time. The structure is derived based on quantitative data, ensuring a precise correspondence between the topology and causal strength data. Incorporating time-dimensional characteristics makes the inference graph time-varying and adaptable, accurately presenting the causal relationship structure at different time stages. Causal edges are selected according to preset criteria, accurately eliminating invalid or weakly correlated edges, ensuring the effective set of causal edges is complete. It possesses reliable strength, provides high-quality materials for topology construction, improves graph accuracy, clarifies the causal direction of effective causal edges and organizes the framework, clearly presents the causal flow between indicators, and constructs a directed causal skeleton that provides a clear structural foundation for subsequent topology aggregation and graph generation, making the graph logic clearer. It divides time windows and integrates structural features, extracts stable and common topological structures in the directed causal skeleton, and forms a cohesive topological pattern that reflects core structural features, avoids local bias in a single window, improves graph stability and reliability, integrates different time-varying causal structures, eliminates conflicts and redundancy, and forms a unified and complete fusion graph skeleton that covers core features and time evolution logic, laying a unified structural foundation for subsequent attribute labeling and graph generation.

[0121] By annotating the time-varying attributes of the fusion graph skeleton, the time-varying causal reasoning graph has both a clear topological structure and intuitive time-varying attribute features, providing rich and accurate information for subsequent path tracing and risk assessment, and enhancing the practical value of the graph.

[0122] The causal path tracing module 103 is used to trace the causal path of the topology in the time-varying causal reasoning graph by taking the preset gastrointestinal intolerance core index as the root node, and to obtain the key causal path of the time-varying causal reasoning graph.

[0123] In this embodiment of the invention, when the causal path tracing module performs causal path tracing on the topological structure of the time-varying causal inference graph with a preset gastrointestinal intolerance core index as the root node to obtain the key causal path of the time-varying causal inference graph, it is specifically used for:

[0124] The preset core index of gastrointestinal intolerance is set as the root node, and the root node is used as the starting point to perform directed causal tracing of the topology of the time-varying causal reasoning graph to obtain the penetration causal path of the time-varying causal reasoning graph.

[0125] Determine the consistency between the causal direction and temporal logic between nodes in the penetration causal path to obtain the effective causal path subset of the penetration causal path;

[0126] The contribution of the effective causal path subset is quantified to obtain the contribution degree of the effective causal path subset;

[0127] The contribution degree and the temporal span information of the effective causal path subset are weighted and fused to obtain the path contribution priority sequence of the contribution degree and the temporal span information;

[0128] The key causal paths of the time-varying causal inference graph are obtained by performing saliency screening on the path contribution priority sequence.

[0129] First, identify the clinically recognized core indicators of gastrointestinal intolerance. Set these indicators as root nodes. Starting from the root node, explore the upstream and downstream nodes related to it layer by layer along the causal relationship between nodes in the time-varying causal reasoning graph. Record the connection relationship and influence direction between nodes in detail, and finally form a permeation causal path that includes all related nodes and connections.

[0130] For each causal path, the occurrence time of each node's state is extracted, and the chronological sequence of the node states is outlined. The causal direction between nodes in each path is checked against this chronological sequence: if the causal direction matches the chronological order, the path segment is retained; if the causal direction is reversed and there is no reasonable basis, the segment is discarded. All valid path segments are integrated to obtain a subset of valid causal paths.

[0131] We collect clinical data from previously critically ill patients, analyze each effective causal pathway based on this data, and assign specific values ​​to each pathway by statistically analyzing key clinical indicators. These values ​​represent the pathway contribution rate, and the higher the value, the stronger the influence of the pathway on the core indicators of gastrointestinal intolerance.

[0132] Weights are assigned based on clinical needs: contribution value is given a higher weight, while time span weight is allocated to balance immediate risks and long-term impacts. The contribution value and time span quantification value of each effective causal pathway are weighted and combined to calculate a comprehensive score. All pathways are then ranked from highest to lowest score to form a pathway contribution priority sequence.

[0133] Based on extensive previous diagnostic and treatment data and clinical intervention effect thresholds, a significance screening threshold was established. From the path contribution priority sequence, paths with a comprehensive score higher than the threshold and a significant impact on gastrointestinal intolerance were selected, while paths with a score lower than the threshold and a weak impact were removed, ultimately yielding the key causal paths in the time-varying causal inference graph.

[0134] The beneficial effects include ensuring the targeted starting point of tracing, avoiding confusion in direction, fully capturing related paths, providing comprehensive basic data for subsequent screening, reducing the risk of missing key paths, eliminating illogical and unreasonable paths, ensuring the reliability of the effective path subset, avoiding interference from erroneous paths in subsequent evaluation, laying a solid foundation for accurate screening of key paths, quantifying contribution based on real clinical data, transforming path impact into intuitive numerical values, replacing vague qualitative judgments, making path evaluation more objective and accurate, providing scientific quantitative support for priority ranking, comprehensively considering contribution and time span and reasonably allocating weights, avoiding the one-sidedness of single-dimensional evaluation, making the priority sequence more aligned with dual clinical needs, facilitating priority focus on core paths, setting thresholds for path screening based on clinical practice, accurately focusing on major key paths, reducing interference from irrelevant paths, making subsequent evaluation and intervention more targeted, and improving clinical efficiency.

[0135] The risk quantification assessment module 104 is used to quantify the path risk of the key causal path and obtain the path risk measure of the key causal path.

[0136] In this embodiment of the invention, when the risk quantification assessment module performs a quantification assessment of the path risk of the key causal path to obtain a path risk measure for the key causal path, it is specifically used for:

[0137] Analyze the node state time sequence of the key causal path to obtain the abnormal state of the node state time sequence;

[0138] Based on the abnormal state, the time-varying effect of the causal path edges of the nodes in the key causal path is quantified to obtain the dynamic influence intensity of the causal path edges.

[0139] By performing time-weighted aggregation on the abnormal state and the intensity of the dynamic impact, the risk evolution sequence of the key causal path is obtained;

[0140] By performing contour fitting on the risk evolution sequence, the comprehensive risk evolution trajectory of the key causal path is obtained;

[0141] Based on preset risk assessment rules, peak period detection is performed on the comprehensive risk evolution trajectory to obtain the risk peak period interval of the comprehensive risk evolution trajectory;

[0142] Feature extraction is performed on the risk peak period interval to obtain the temporal features of the risk peak period interval;

[0143] The peak intensity and duration of the time-domain features are quantitatively characterized to obtain the path risk measure of the key causal path.

[0144] The formula for calculating the intensity of the dynamic influence is as follows:

[0145] ;

[0146] In the formula, In time From the nodes in the key causal path To the node The intensity of the dynamic influence, Based on the nodes in the key causal path The function that calculates the abnormal state. For the nodes in the key causal path In time The node state timing values, For the nodes in the key causal path The average state value over the reference period. For the nodes in the key causal path The standard deviation of the state values ​​over the reference period. Based on the nodes in the key causal path The function that calculates the abnormal state. For the nodes in the key causal path In time The node state timing values, For the nodes in the key causal path The average state value over the reference period. For the nodes in the key causal path The standard deviation of the state values ​​over the reference period. For a function that takes a time delay parameter as input, In time From the nodes in the key causal path To the node The time delay parameter or time-dependent parameter for the transmission of causal effects.

[0147] A comprehensive and detailed analysis of the state sequence of each node in the critical causal path was conducted. The state sequence of the nodes in the critical causal path covers the specific state records of each node at different time points in the path. During the analysis, the state data of each node was checked one by one in chronological order. At the same time, the standard range of normal working state of each node was defined. The node state data at each time point was compared with the corresponding standard range. Any node state data that exceeded the standard range was identified as an abnormal state, thereby completely identifying all abnormal situations in the node state sequence.

[0148] Based on the identified abnormal states, the transmission process of each abnormal state between nodes in the key causal path is tracked. The role of causal relationships between nodes at different time points is observed in detail. The severity of the abnormal state and the transmission speed between nodes are analyzed. Combining this key information, the impact of the causal path edges between nodes changing over time is transformed into specific and measurable values. These values ​​represent the dynamic impact strength of the causal path edges, which can clearly reflect the magnitude of the role of the causal path edges in risk transmission at different time periods.

[0149] First, clarify the importance of abnormal states and dynamic impact intensity at different time points. Assign higher weights to abnormal states and dynamic impact intensity that are closer to the current moment and have a more significant impact on the subsequent risk development trend. Then, according to the determined weights, integrate and calculate the abnormal state data and dynamic impact intensity data corresponding to each time point, and summarize the originally scattered single time point data into a continuous sequence. This sequence can fully present the changes in risk over time, which is the risk evolution sequence of the key causal path.

[0150] By comprehensively reviewing all data points in the risk evolution sequence, deeply analyzing the intrinsic relationships and overall trends between each data point, and constructing a continuous curve based on the fluctuation patterns of the data, this curve will fully cover all key data points in the sequence and accurately present the core characteristics such as the rise and fall of risk in different time periods and the magnitude of fluctuations. The final curve is the comprehensive risk evolution trajectory of the key causal path.

[0151] Pre-set risk assessment rules, clearly define the numerical standards corresponding to high risk levels and the duration standards that must be met for high risk levels. Then, conduct segment-by-segment investigation and analysis along the comprehensive risk evolution trajectory to accurately identify the time periods in which the risk value reaches the preset high-level numerical standards and the duration meets the set requirements. At the same time, record the start and end times of each such time period in detail. These time periods are the risk peak intervals of the comprehensive risk evolution trajectory.

[0152] For each identified risk peak period interval, we conduct in-depth analysis of the changing patterns of risk values ​​within the interval, and extract key information that reflects the core characteristics of the risk in that interval. Specifically, this includes the highest level of risk value reached within the interval, the average level of risk value within the interval, the rate of change of risk value from the rising phase to the falling phase, and the frequency of risk fluctuations within the peak period interval. These key pieces of information together constitute the temporal characteristics of the risk peak period.

[0153] By combining the peak intensity and duration of the risk peak period in the time domain, a scientific and reasonable comprehensive quantitative calculation method is set. The specific values ​​corresponding to the peak intensity and duration are substituted into the calculation method for comprehensive calculation, and finally a comprehensive quantitative result that can fully reflect the risk level of the key causal path is obtained. This result is the path risk measurement of the key causal path.

[0154] In the formula for calculating the intensity of dynamic influence, It is a core indicator to be determined, used to quantify time. Nodes in the key causal path To the node The path edge time-varying effect, its value is related to the node , The degree of anomaly is positively correlated with the causal transmission delay, which is the causal characteristic of node status and time series topology analysis from multi-source data collection. and Convert nodes separately , The standardized deviation contributes to the impact of path edges; the larger the input deviation, the larger the function value. This is based on the abnormal state parsing logic, where... , It is a node , Real-time status, It is a node , Reference period average, , It corresponds to the standard deviation; Adjusting for the time lag in causal transmission: the smaller the time lag, the larger the function value. It is a node To the node The transmission delay is considered. The overall formula integrates the dual-node anomaly contribution and delay adjustment, transforming the data collection and causal analysis results into path risk quantification values. This provides support for subsequent path risk measurement, early warning signal generation, and intervention target determination, and aligns with the processing logic of each module and actual physiological laws.

[0155] The beneficial effects include: accurately locating all abnormal states and ensuring no risk-related data is overlooked, laying an accurate and comprehensive foundation for subsequent path risk assessment; objectively quantifying the dynamic impact of causal path edges, improving the objectivity and accuracy of subsequent risk assessment; providing reliable data for understanding the patterns of risk transmission; highlighting the impact of data at key time points; constructing a risk evolution sequence that truly reflects the risk development process; providing realistic basic data for analyzing risk change trends; transforming discrete sequences into continuous trajectories, intuitively displaying the patterns of risk change, facilitating medical staff's observation of the risk situation; providing clear analytical objects for identifying high-risk periods; accurately pinpointing key time periods of concentrated risk outbreaks, enabling medical staff to grasp high-risk periods and providing clear time guidance for targeted interventions; extracting the core characteristics of risk peaks; comprehensively mining key information; providing comprehensive evidence for risk quantification and characterization; enabling risk assessment to more deeply reflect the essence of risk; comprehensively quantifying peak intensity and duration to obtain intuitive and accurate path risk measurement; providing clear reference for risk warning and clinical intervention; and improving the scientific nature of clinical decision-making.

[0156] The early warning and tracing module 105 is used to perform threshold interpretation on the path risk measurement to obtain the causal early warning signal of the path risk measurement, and to perform reverse causal tracing on the key causal path corresponding to the causal early warning signal to obtain the priority intervention target of the causal early warning signal.

[0157] In this embodiment of the invention, when the early warning and tracing module performs threshold judgment on the path risk measure to obtain the causal early warning signal of the path risk measure, and performs reverse causal tracing on the key causal path corresponding to the causal early warning signal to obtain the priority intervention target of the causal early warning signal, it is specifically used for:

[0158] Threshold identification is performed on the path risk measure to obtain the risk index of the path risk measure;

[0159] The statistical distribution of core physiological parameters in the multi-dimensional time-series data is analyzed to determine the parameter-specific dynamic range of the critically ill patients;

[0160] Within the parameter-specific dynamic range, a multi-level risk interval corresponding to the parameter within the parameter-specific dynamic range is calibrated.

[0161] Extract the boundary values ​​of the multi-level risk intervals to obtain the dynamic risk threshold set of the critically ill patient;

[0162] Establish the correspondence between the range of parameter values ​​in the dynamic risk threshold set and the risk level;

[0163] The correspondence is formalized into an executable set of judgment rules to construct an individualized risk spectrum for the critically ill patients;

[0164] The risk indicators are assigned to the individualized risk spectrum to obtain the risk level of the risk indicators.

[0165] The risk attribution level is signal-encoded to obtain the causal early warning signal of the path risk measure;

[0166] Based on the risk spatiotemporal information indicated by the causal early warning signal, risk path segments are located from the key causal paths;

[0167] Reconstruct the causal transmission relationship between each node in the risk path segment to obtain the abnormal transmission network of the key causal path;

[0168] Based on the risk attribution level and the topology of the abnormal transmission network, the intervention effectiveness of the nodes in the abnormal transmission network is evaluated to determine the priority score of the abnormal transmission network.

[0169] Based on the priority score, starting from the root node, the abnormal transmission network is traversed in reverse to obtain the candidate key nodes of the abnormal transmission network.

[0170] The node with the highest priority score among the candidate key nodes is selected as the priority intervention target for the causal early warning signal.

[0171] When the early warning and source tracing module evaluates the intervention effectiveness of each node in the abnormal transmission network to obtain a priority score for the abnormal transmission network, it is specifically used for:

[0172] Analyze the topological connectivity of the anomalous conduction network to obtain the causal edge set of the anomalous conduction network;

[0173] Based on the causal edge set, the causal control capability of nodes in the abnormal transmission network on downstream nodes is quantified to obtain the quantification result of the node influence of the abnormal transmission network.

[0174] The node impact quantification results are fused with the node historical response archives of the abnormal transmission network to obtain the comprehensive node characteristics of the abnormal transmission network.

[0175] Based on the comprehensive characteristics of the nodes, the potential blocking strength of the nodes in the abnormal transmission network to downstream risk transmission is analyzed to assess the node strength of the abnormal transmission network.

[0176] Based on the node strength, the node blocking effectiveness of the abnormal conduction network is determined, and the node blocking effectiveness is normalized to obtain the priority score of the abnormal conduction network.

[0177] First, a complete numerical sequence of pathway risk measurement is collected. Then, combined with the individual circumstances of critically ill patients, such as age, underlying diseases, and current treatment stage, as well as the risk assessment benchmarks related to enteral nutrition tolerance in clinical practice, these values ​​are characterized and classified one by one. The degree of risk association corresponding to each value is clarified, and finally, a risk indicator that can directly reflect the risk status is formed.

[0178] Core physiological parameters such as heart rate, blood pressure, and gastrointestinal motility are extracted from multi-dimensional time-series data. All recorded values ​​of these parameters are classified and statistically analyzed to identify the distribution characteristics of each parameter, such as frequency of occurrence, concentration range, and fluctuation amplitude. Then, referring to the normal range of physiological parameters in healthy individuals and considering the individual circumstances of the critically ill patient, such as underlying diseases and treatment progress, abnormal interference data generated during the testing process are removed, and the specific dynamic range of each core physiological parameter suitable for the patient is defined, namely the parameter-specific dynamic range.

[0179] Within the parameter-specific dynamic range of each core physiological parameter, based on the clinical classification criteria for the severity of enteral nutrition intolerance, and combined with the patient's disease severity and treatment tolerance, the dynamic range is divided into multiple risk levels, such as low risk, medium risk, and high risk. The starting and ending values ​​of each level range are clearly defined to ensure that the division of each range accurately corresponds to different risk levels, thus completing the calibration of multi-level risk ranges.

[0180] Each core physiological parameter corresponding to a multi-level risk interval is investigated one by one. The starting and ending boundary values ​​of each interval are collected. These values ​​are deduplicated, sorted and organized to ensure that each boundary value is accurate and without omission. All the organized boundary values ​​are integrated together to form a dynamic risk threshold set specific to the critically ill patient.

[0181] Based on the previously defined matching relationship between multi-level risk intervals and risk levels, the dynamic risk thresholds are concentrated in the parameter value intervals formed by each adjacent boundary value. These intervals are then associated with the corresponding risk levels such as low risk, medium risk, and high risk, clarifying the specific risk level corresponding to each parameter value interval. This ensures that the correspondence between each interval and level is clear and unique, thus completing the establishment of the correspondence.

[0182] The correspondence between the parameter value range of each core physiological parameter and the risk level is transformed into clear rules that the module can identify and execute. Each rule clearly defines the risk level when the value falls into a certain range. Then, these judgment rules corresponding to all core physiological parameters are systematically integrated to form a risk assessment system that comprehensively covers all core physiological parameters of the patient and is tailored to their individual circumstances, namely, an individualized risk spectrum.

[0183] The risk indicators corresponding to the previously obtained path risk measurements are compared one by one with the various judgment rules in the individualized risk spectrum to see which core physiological parameter value range the value of the risk indicator falls into. Based on the corresponding judgment rules, the risk level is determined, and the classification of the risk indicator is completed, resulting in a clear risk classification level.

[0184] Based on different risk attribution levels, corresponding coding rules are set. Different risk levels, such as low risk, medium risk, and high risk, correspond to different specific signal forms. The obtained risk attribution levels are converted according to the rules to form signals with clear identification that can be received by subsequent modules and recognized by clinical personnel, namely causal warning signals.

[0185] The system analyzes the spatiotemporal information contained in the causal warning signal, such as the time of risk occurrence and the physiological parameter nodes involved. Based on this information, it then checks each key causal path to find the path segment that matches the spatiotemporal information. It clarifies the starting and ending positions of this path segment in the entire key causal path, accurately identifies the path segment with risk, and completes the location of the risk path segment.

[0186] A detailed analysis of the identified risk path segments is conducted, outlining the sequence of each node in the segment, the interaction relationships between nodes, reconstructing the transmission process of causal signals between nodes, identifying which nodes exhibit abnormal transmission, and organizing and integrating these abnormal transmission relationships according to the actual transmission logic to form a clear abnormal transmission network reflecting the risk transmission path.

[0187] First, clarify the severity of the risk reflected by the risk attribution level, and then analyze the topological structure of the abnormal transmission path, including the number of nodes, the tightness of the connection between nodes, and the position of nodes in the path. Combining these two aspects of information, analyze the potential role and expected effect of each node in blocking the transmission of risk and mitigating the degree of risk. Assign corresponding scores to each node according to the magnitude of its role and the quality of its effect to form a priority score for each node.

[0188] Starting from the root node corresponding to the preset core indicators of gastrointestinal intolerance, each node in the abnormal transmission network is examined one by one in the order opposite to the risk transmission direction in the abnormal transmission network. The priority score of each node is recorded, and the nodes that meet the preset standard are selected and integrated to form a set of candidate key nodes of the abnormal transmission network.

[0189] By comparing and analyzing the priority scores of each node in the candidate key node set, the node with the highest score is identified as the most effective node in intervening in risk transmission and improving patients' enteral nutrition tolerance, and it is determined as the priority intervention target corresponding to the causal early warning signal.

[0190] The abnormal transmission path is meticulously dissected to examine the connections between each node and other nodes, identify which nodes have direct causal relationships, sort out the direction and form of these relationships, combine each pair of nodes with direct causal relationships as a causal edge, and collect all such causal edges to form a complete set of causal edges.

[0191] Based on causal edge sets, we analyze the strength and scope of each node's influence on downstream related nodes through the corresponding causal edges. Combining the statistical effects of similar node regulation in clinical data, we use a reasonable numerical conversion method to transform the causal regulation ability of each node on downstream nodes into specific values. We then integrate these values ​​of all nodes to form a quantified result of node influence.

[0192] Collect historical response files for each node in the abnormal transmission network, including data on the node's past response in similar risk scenarios and the effects of intervention. Then, integrate these historical data with the corresponding node impact quantification results, complement and improve the data, remove redundant information, and retain key data that can comprehensively reflect the characteristics of the node to form the comprehensive characteristics of each node, i.e., the node comprehensive characteristics.

[0193] Based on the comprehensive characteristics of each node, we analyze its position in the abnormal transmission network, its closeness to other nodes, and the effects of past interventions to determine the strength of its potential role in blocking the transmission of risk to downstream nodes. Combining these judgments, we assess the blocking capability of each node by level or numerical value, thus completing the evaluation of node strength.

[0194] Based on the node strength obtained from the assessment, and combined with the actual role of the nodes in the abnormal transmission network, the actual effectiveness of each node in blocking risk transmission is determined. Then, the blocking effectiveness values ​​of all nodes are adjusted according to a unified standard to make them fall within a fixed numerical range, eliminating the difference in the dimensions of effectiveness values ​​between different nodes. The adjusted values ​​are the priority scores of each node.

[0195] The beneficial effects include: combining individual patient circumstances with clinical benchmarks to identify risks, avoiding bias from uniform standards, improving the accuracy of risk indicators, laying the foundation for early warning; analyzing the distribution of core physiological parameters and considering individual differences, breaking through the limitations of general scope, making risk assessment more relevant to patients' actual situations; finely calibrating multi-level risk intervals within a specific dynamic range, avoiding ambiguity in risk judgment, providing a clear basis for accurate early warning; extracting interval boundary values ​​to form a dynamic risk threshold set, eliminating subjectivity in risk judgment, making assessment systematic; establishing a correspondence between parameter value intervals and risk levels, standardizing the judgment process, improving the efficiency of risk assessment, reducing errors; constructing an individualized risk spectrum, making risk assessment more systematic and operable, improving standardization and accuracy; determining risk levels based on individualized spectrums, avoiding the problem of inapplicability of general standards, improving the accuracy and reliability of judgment; encoding risk levels as early warning signals, facilitating subsequent module processing and clinical identification, improving signal transmission efficiency and convenience; locating risk path segments based on spatiotemporal information, quickly locking the core of the risk, improving risk localization efficiency. This approach focuses on improving efficiency and accuracy, reconstructing the causal transmission relationship of nodes, clearly presenting the risk transmission path and abnormal links, providing a basis for determining intervention targets, evaluating node effectiveness by combining risk level and topological structure to avoid the one-sidedness of a single dimension, providing quantitative basis for target screening, and using priority scoring to reversely screen candidate nodes, narrowing the range of intervention targets and improving the efficiency of determining priority targets. The node with the highest score is selected as the priority target, clarifying the core of clinical intervention, improving intervention effect, alleviating patient risk, sorting out the causal relationships of nodes to form causal edge sets, clearly presenting core relationships, providing basic data for quantifying node impact, quantifying the causal regulation capacity of nodes, transforming abstract influence into intuitive values, providing precise basis for integrating node characteristics, integrating quantitative results with historical archives, improving node characteristics, avoiding the one-sidedness of single data, supporting node strength assessment, evaluating blocking strength based on comprehensive node characteristics, avoiding subjective assumptions, improving the credibility of node strength assessment, normalizing node blocking efficacy, unifying scoring comparison standards, eliminating dimensional differences, and improving the scientific nature of intervention plans.

[0196] The decision fusion module 106 is used to perform decision fusion on the causal early warning signal, the key causal path, the path risk measurement and the priority intervention target to obtain the clinical decision basis for the critically ill patient.

[0197] In this embodiment of the invention, when the decision fusion module performs decision fusion on the causal early warning signal, the key causal path, the path risk measure, and the priority intervention target to obtain the clinical decision basis for the critically ill patient, it is specifically used for:

[0198] The causal early warning signal, the key causal path, the path risk measure, and the priority intervention target are integrated into a decision factor set for the critically ill patient;

[0199] By analyzing the warning level of the causal warning signal and the topological stability of the key causal path, the timeliness attribute of the critical patient can be obtained;

[0200] Based on the timeliness attribute, the mapping relationship between the path risk measure and the priority intervention target is calibrated to obtain the dynamic intervention rules for the priority intervention target;

[0201] Based on the dynamic intervention rules, the specific operation sequence and objectives of the priority intervention targets are arranged to obtain the targeted intervention plan for the priority intervention targets;

[0202] By integrating the decision factor set, the timeliness attribute, and the targeted intervention plan, the clinical decision-making basis for the critically ill patient is obtained.

[0203] First, extract the warning type and risk level of the causal warning signal, the node association sequence and influence transmission direction of the key causal path, the risk intensity and risk persistence of the path risk measurement, and the core nodes and intervention priority of the priority intervention target. Then, classify and sort this information to remove duplicate and invalid content. Finally, according to the clinical assessment logic of critically ill patients, the classified information system is collected to form a set of decision factors with a clear structure to ensure that key information can be quickly retrieved for subsequent clinical decisions.

[0204] First, clarify the criteria for classifying warning levels in causal warning signals, and combine clinical data on enteral nutrition tolerance in critically ill patients to determine the urgency of the current warning; then, analyze the topological stability of key causal pathways, and judge whether the pathway is prone to change and whether the influence transmission is stable by observing the tightness of node associations and changes in the path structure over time; based on the judgment results of the warning urgency and topological stability, comprehensively determine the time attributes such as risk response time requirements and effective intervention time windows, and clarify the optimal intervention time range.

[0205] Using the response time requirements and effective time window in the timeliness attribute as a reference, first analyze whether the correspondence between path risk measurement and priority intervention targets is reasonable, and determine whether high-risk measurement matches high-priority targets and whether different risk durations correspond to different targets; adjust the mismatch relationship, such as strengthening the correspondence between high-risk measurement with urgent timeliness and core priority targets, and optimizing the correspondence between risks with long duration but low urgency and secondary priority targets; through adjustment, establish dynamic intervention rules that are adapted to timeliness attributes and conform to the actual risk, and clarify the priority targets and intervention requirements under different risks.

[0206] Based on the target priority in the dynamic intervention rules, the operation sequence of the priority intervention targets is first determined; then, intervention goals are set for each target, clarifying the standards for improvement of physiological parameters and the extent of improvement in nutritional tolerance; then, combined with clinical norms and individual patient conditions, the operation methods for each target are formulated, integrating the operation sequence, goals, and methods to form a targeted intervention plan that can guide clinical practice.

[0207] First, the basic information of the decision factor set is combined with the time requirements of the timeliness attribute to clarify the time constraints and information basis for clinical decision-making. Then, the results are integrated with the operation sequence, intervention goals, and operation methods of the targeted intervention plan. Finally, the integrated information is sorted out and optimized, and organized into a hierarchical content according to the conventional logic of clinical decision-making, covering the core basis for decision-making, intervention plan, and implementation time requirements, forming a basis for clinical decision-making for critically ill patients for medical staff to refer to.

[0208] The beneficial effects include avoiding information fragmentation and omission of key content, providing comprehensive and centralized basic information support for clinical decision-making, improving the availability and retrieval efficiency of decision-making information, accurately grasping the key time information for risk response, providing a time dimension basis for intervention rule formulation, avoiding poor results due to intervention being too early or too late, adapting intervention rules to the time characteristics of risk, avoiding inappropriate selection of intervention targets due to fixed mapping, improving the pertinence and adaptability of intervention measures, avoiding chaotic intervention operations and vague goals, providing clear and executable guidance for clinical operations, improving the standardization and effectiveness of intervention operations, avoiding decision-making bias caused by a single information dimension, providing complete and reliable decision support for medical staff, and improving the scientific nature and accuracy of clinical decision-making.

[0209] 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.

[0210] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application device 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.

[0211] 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 device for dynamically assessing enteral nutrition tolerance in critically ill patients, characterized in that, The device includes a multi-source data acquisition module, a causal topology construction module, a causal path tracing module, a risk quantification and assessment module, an early warning and source tracing module, and a decision fusion module, wherein: The multi-source data acquisition module is used to simultaneously acquire enteral nutrition and key physiological parameters of critically ill patients from multiple sources, thereby obtaining multi-dimensional time-series data of the critically ill patients. The causal topology construction module is used to construct causal topology on the multi-dimensional time series data to obtain the time-varying causal inference graph of the multi-dimensional time series data. The causal path tracing module is used to trace the causal path of the topology in the time-varying causal reasoning graph with a preset gastrointestinal intolerance core index as the root node, and to obtain the key causal path of the time-varying causal reasoning graph. The risk quantification assessment module is used to quantify the path risk of the key causal path and obtain the path risk measure of the key causal path. The early warning and tracing module is used to perform threshold judgment on the path risk measurement, obtain the causal early warning signal of the path risk measurement, and perform reverse causal tracing on the key causal path corresponding to the causal early warning signal to obtain the priority intervention target of the causal early warning signal. The decision fusion module is used to perform decision fusion on the causal early warning signal, the key causal path, the path risk measurement, and the priority intervention target to obtain the clinical decision basis for the critically ill patient.

2. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 1, characterized in that, When the multi-source data acquisition module performs multi-source synchronous acquisition of enteral nutrition and key physiological parameters of critically ill patients to obtain multi-dimensional time-series data of the critically ill patients, it is specifically used for: The dynamics of enteral nutrition infusion in critically ill patients were recorded in a time-series non-destructive manner to obtain a refined time series of the enteral nutrition infusion dynamics. The key physiological parameters of the critically ill patients were monitored holographically to obtain the time series of the key physiological parameters; The refined time series is timestamped with the time series to obtain the synchronized time stream of the critically ill patient; The original monitoring values ​​in the synchronized time series are standardized in terms of dimensions to obtain the standardized value set of the synchronized time series. The standardized numerical set is discretized to obtain the standardized numerical distribution of the synchronized time-series stream; By integrating the indicator data from the standardized numerical distribution, multi-dimensional time-series data of the critically ill patients are obtained.

3. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 1, characterized in that, When the causal topology construction module performs causal topology construction on the multi-dimensional time series data to obtain the time-varying causal inference graph of the multi-dimensional time series data, it is specifically used for: Explore the causal structure of the multi-dimensional time series data and extract the candidate causal edge set of the multi-dimensional time series data; Dynamic causal efficacy quantification is performed on the candidate causal edge set to obtain the time-varying causal strength matrix of the candidate causal edge set; By performing topological derivation on the time-varying causal intensity matrix, a time-varying causal reasoning graph of the time-varying causal intensity matrix is ​​obtained.

4. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 3, characterized in that, When the causal topology construction module performs topological derivation on the time-varying causal intensity matrix to obtain the time-varying causal inference graph of the time-varying causal intensity matrix, it is specifically used for: Based on the preset judgment criteria, a set of valid causal edges is selected from the time-varying causal intensity matrix; Reconstruct the directed associations of the effective causal edge set to obtain the directed causal skeleton of the effective causal edge set; Cross-window topological aggregation is performed on the directed causal skeleton to obtain the condensed topological pattern of the directed causal skeleton; The time-varying causal structure in the condensed topological mode is fused with a graph structure to obtain the fused graph skeleton of the time-varying causal structure. Time-varying attribute annotation is performed on the fusion graph skeleton to obtain the time-varying causal inference graph of the time-varying causal intensity matrix.

5. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 2, characterized in that, When the causal path tracing module performs causal path tracing on the topological structure of the time-varying causal inference graph, using a preset gastrointestinal intolerance core index as the root node, to obtain the key causal paths of the time-varying causal inference graph, it is specifically used for: The preset core index of gastrointestinal intolerance is set as the root node, and the root node is used as the starting point to perform directed causal tracing of the topology of the time-varying causal reasoning graph to obtain the penetration causal path of the time-varying causal reasoning graph. Determine the consistency between the causal direction and temporal logic between nodes in the penetration causal path to obtain the effective causal path subset of the penetration causal path; The contribution of the effective causal path subset is quantified to obtain the contribution degree of the effective causal path subset; The contribution degree and the temporal span information of the effective causal path subset are weighted and fused to obtain the path contribution priority sequence of the contribution degree and the temporal span information; The key causal paths of the time-varying causal inference graph are obtained by performing saliency screening on the path contribution priority sequence.

6. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 1, characterized in that, When the risk quantification assessment module performs a quantitative assessment of the path risk of the key causal path to obtain a path risk measure for the key causal path, it is specifically used for: Analyze the node state time sequence of the key causal path to obtain the abnormal state of the node state time sequence; Based on the abnormal state, the time-varying effect of the causal path edges of the nodes in the key causal path is quantified to obtain the dynamic influence intensity of the causal path edges. By performing time-weighted aggregation on the abnormal state and the intensity of the dynamic impact, the risk evolution sequence of the key causal path is obtained; By performing contour fitting on the risk evolution sequence, the comprehensive risk evolution trajectory of the key causal path is obtained; Based on preset risk assessment rules, peak period detection is performed on the comprehensive risk evolution trajectory to obtain the risk peak period interval of the comprehensive risk evolution trajectory; Feature extraction is performed on the risk peak period interval to obtain the temporal features of the risk peak period interval; The peak intensity and duration of the time-domain features are quantitatively characterized to obtain the path risk measure of the key causal path.

7. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 6, characterized in that, The formula for calculating the intensity of the dynamic influence is as follows: ； In the formula, In time From the nodes in the key causal path To the node The intensity of the dynamic influence, Based on the nodes in the key causal path The function that calculates the abnormal state. For the nodes in the key causal path In time The node state timing values, For the nodes in the key causal path The average state value over the reference period. For the nodes in the key causal path The standard deviation of the state values ​​over the reference period. Based on the nodes in the key causal path The function that calculates the abnormal state. Nodes in the critical causal path In time The node state timing values, For the nodes in the key causal path The average state value over the reference period. For the nodes in the key causal path The standard deviation of the state values ​​over the reference period. This is a function that takes a time delay parameter as input. In time From the nodes in the key causal path To the node The time delay parameter or time-dependent parameter for the transmission of causal effects.

8. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 5, characterized in that, When the early warning and tracing module performs threshold judgment on the path risk measurement to obtain the causal early warning signal of the path risk measurement, and performs reverse causal tracing on the key causal path corresponding to the causal early warning signal to obtain the priority intervention target of the causal early warning signal, it is specifically used for: Threshold identification is performed on the path risk measure to obtain the risk index of the path risk measure; The statistical distribution of core physiological parameters in the multi-dimensional time-series data is analyzed to determine the parameter-specific dynamic range of the critically ill patients; Within the parameter-specific dynamic range, a multi-level risk interval corresponding to the parameter within the parameter-specific dynamic range is calibrated. Extract the boundary values ​​of the multi-level risk intervals to obtain the dynamic risk threshold set of the critically ill patient; Establish the correspondence between the range of parameter values ​​in the dynamic risk threshold set and the risk level; The correspondence is formalized into an executable set of judgment rules to construct an individualized risk spectrum for the critically ill patients; The risk indicators are assigned to the individualized risk spectrum to obtain the risk level of the risk indicators. The risk attribution level is signal-encoded to obtain the causal early warning signal of the path risk measure; Based on the risk spatiotemporal information indicated by the causal early warning signal, risk path segments are located from the key causal paths; Reconstruct the causal transmission relationship between each node in the risk path segment to obtain the abnormal transmission network of the key causal path; Based on the risk attribution level and the topology of the abnormal transmission network, the intervention effectiveness of the nodes in the abnormal transmission network is evaluated to determine the priority score of the abnormal transmission network. Based on the priority score, starting from the root node, the abnormal transmission network is traversed in reverse to obtain the candidate key nodes of the abnormal transmission network. The node with the highest priority score among the candidate key nodes is selected as the priority intervention target for the causal early warning signal.

9. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 8, characterized in that, When the early warning and source tracing module evaluates the intervention effectiveness of each node in the abnormal transmission network to obtain a priority score for the abnormal transmission network, it is specifically used for: Analyze the topological connectivity of the anomalous conduction network to obtain the causal edge set of the anomalous conduction network; Based on the causal edge set, the causal control capability of nodes in the abnormal transmission network on downstream nodes is quantified to obtain the quantification result of the node influence of the abnormal transmission network. The node impact quantification results are fused with the node historical response archives of the abnormal transmission network to obtain the comprehensive node characteristics of the abnormal transmission network. Based on the comprehensive characteristics of the nodes, the potential blocking strength of the nodes in the abnormal transmission network to downstream risk transmission is analyzed to assess the node strength of the abnormal transmission network. Based on the node strength, the node blocking effectiveness of the abnormal conduction network is determined, and the node blocking effectiveness is normalized to obtain the priority score of the abnormal conduction network.

10. The dynamic assessment device for enteral nutrition tolerance in critically ill patients as described in claim 1, characterized in that, When the decision fusion module performs decision fusion on the causal early warning signal, the key causal path, the path risk measure, and the priority intervention target to obtain the clinical decision basis for the critically ill patient, it is specifically used for: The causal early warning signal, the key causal path, the path risk measure, and the priority intervention target are integrated into a decision factor set for the critically ill patient; By analyzing the warning level of the causal warning signal and the topological stability of the key causal path, the timeliness attribute of the critical patient can be obtained; Based on the timeliness attribute, the mapping relationship between the path risk measure and the priority intervention target is calibrated to obtain the dynamic intervention rules for the priority intervention target; Based on the dynamic intervention rules, the specific operation sequence and objectives of the priority intervention targets are arranged to obtain the targeted intervention plan for the priority intervention targets; By integrating the decision factor set, the timeliness attribute, and the targeted intervention plan, the clinical decision-making basis for the critically ill patient is obtained.

Images