Contextual awareness information recommendation method and system in patient monitoring

Abstract

The application provides a context-aware information recommendation method and system in patient monitoring, and relates to the technical field of computer of medical patient monitoring, first, dynamic evolution characteristics are perceived according to real-time state data of a monitoring scene, and a dynamic demand model containing demand association rules and change trends is constructed based on this; then, scene adaptation knowledge modules in a monitoring knowledge repository are restructured and structured; then, initial context-aware recommendation information is generated; then, a dynamic evolution characteristic change track is tracked, and dynamic demand model parameters are adjusted to optimize the initial recommendation information; finally, a recommendation information set is generated and pushed to a monitoring execution terminal. The application can adapt to dynamic changes of a monitoring scene, provide accurate information for medical staff, and improve monitoring quality.

Description

Technical Field

[0001] This invention relates to the technical field of computers for medical patient monitoring, and more specifically, to a context-aware information recommendation method and system for patient monitoring. Background Technology

[0002] In patient monitoring within the medical field, providing timely and accurate information relevant to the patient's current condition to medical staff is directly related to the patient's treatment outcome and safety. Traditional patient monitoring information recommendation methods are usually based on fixed rules and pre-set knowledge bases. These methods can play a certain role when dealing with relatively stable patient conditions, but they are clearly insufficient when facing complex and ever-changing real-world monitoring scenarios.

[0003] Real-world patient monitoring scenarios are dynamic, with factors such as a patient's physiological indicators, disease progression, and treatment environment constantly changing over time. For example, a patient's vital signs may fluctuate drastically in a short period, or the treatment plan may be adjusted according to the progression of the disease. However, traditional methods struggle to perceive these dynamic changes in real time and cannot update information recommendations promptly based on the evolving scenario. This results in potentially outdated or inaccurate recommendations that fail to meet the monitoring needs of healthcare professionals at different stages.

[0004] Furthermore, the knowledge structure in traditional knowledge bases is often static, lacking a mechanism for flexible adjustment and optimization based on actual needs. This makes it difficult to quickly and accurately extract the most relevant information from the knowledge base when facing diverse monitoring scenarios, affecting the quality and efficiency of information recommendation. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a context-aware information recommendation method in patient monitoring, the method comprising: Based on real-time status data in the monitoring scenario, perceive the dynamic evolution characteristics of the patient monitoring scenario; A dynamic demand model for patient monitoring is constructed based on the aforementioned dynamic evolution characteristics. The dynamic demand model includes the association rules and changing trends of demand as the scenario evolves. The scenario adaptation knowledge module in the guardianship knowledge reserve is reconstructed based on the dynamic demand model constructed according to the dynamic evolution characteristics, and the scenario adaptation knowledge module is restructured according to the demand change trend. Initial context-aware recommendation information is generated based on the reconstructed scene adaptation knowledge module; The system tracks the real-time changes of the dynamic evolution features, adjusts the parameters of the dynamic demand model, optimizes the initial context-aware recommendation information based on the adjusted dynamic demand model, generates the final recommendation information set, and pushes it to the monitoring execution terminal.

[0006] Furthermore, embodiments of the present invention also provide a context-aware information recommendation system for patient monitoring, comprising: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to perform the aforementioned context-aware information recommendation method in patient monitoring by executing the machine-executable instructions.

[0007] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions stored in a computer-readable storage medium, a processor of a context-aware information recommendation system in patient monitoring reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the context-aware information recommendation system in patient monitoring to perform the aforementioned context-aware information recommendation method in patient monitoring.

[0008] Based on the above, by perceiving the dynamic evolution characteristics of the patient monitoring scenario based on real-time status data, it is possible to understand the changes in the patient's environment. A dynamic demand model for patient monitoring can be constructed based on these dynamic evolution characteristics. This model includes the association rules and trends of demand evolution with the scenario, enabling in-depth analysis of the intrinsic connection between patient needs and the monitoring scenario, and accurate prediction of the development direction of patient needs. The scenario-adaptive knowledge modules in the monitoring knowledge base are reconstructed according to the dynamic demand model and structurally reorganized according to the changing trends of demand. This allows the knowledge in the knowledge base to dynamically adjust with changes in scenario and demand, improving the efficiency of knowledge retrieval and utilization, and ensuring that recommended information is highly matched to the current scenario and demand. By tracking the real-time changes in dynamic evolution characteristics and adjusting the parameters of the dynamic demand model, the initial context-aware recommendation information is optimized, generating the final recommendation information set and pushing it to the monitoring execution terminal. This ensures that the recommended information always keeps pace with scenario changes, providing medical staff with effective patient monitoring information and effectively improving the quality and efficiency of patient monitoring. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the execution flow of the context-aware information recommendation method in patient monitoring provided in an embodiment of the present invention.

[0010] Figure 2 This is a schematic diagram of exemplary hardware and software components of a context-aware information recommendation system for patient monitoring provided in an embodiment of the present invention. Detailed Implementation

[0011] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1This is a flowchart illustrating a context-aware information recommendation method in patient monitoring according to an embodiment of the present invention. The following is a detailed description of this context-aware information recommendation method in patient monitoring.

[0012] Step S110: Perceive the dynamic evolution characteristics of the patient monitoring scenario based on real-time status data in the monitoring scenario.

[0013] In this embodiment, the monitoring process of an adult patient admitted to the intensive care unit is used as the application scenario throughout the text. This patient suffers from respiratory failure due to severe lung infection, requiring continuous vital sign monitoring and respiratory support. Real-time status data in the monitoring scenario includes physiological data recorded by multi-parameter monitors, operating parameters of equipment such as ventilators, operation records performed by nursing staff, and ward environment monitoring data. Through continuous collection and analysis of the above real-time status data, the dynamic changes of the patient monitoring scenario over time are perceived, thereby extracting key features that reflect the evolution of the scenario.

[0014] Step S111: Deploy scene-aware terminals to collect real-time status data in the monitoring scene. The real-time status data includes execution data of monitoring operations, physiological monitoring data of patients, and environmental parameter data.

[0015] In this intensive care unit setting, multiple scene-sensing terminals were deployed. Physiological monitoring data was collected via multi-parameter monitors, including parameters such as heart rate, blood pressure (systolic, diastolic, and mean arterial pressure), blood oxygen saturation, respiratory rate, and body temperature, sampled every second. This data was transmitted in real-time to the central data processing system in digital signal form. Data on monitoring procedures was recorded through a nursing information system. When performing procedures such as suctioning, turning over, and administering medication, nursing staff selected the corresponding operation type in the system and recorded the start and end times, as well as key parameters during the operation, such as suctioning duration and medication dosage. Environmental parameter data was collected through environmental sensors within the ward, including data recorded every five minutes on indoor temperature, humidity, light intensity, and air oxygen and carbon dioxide concentrations.

[0016] Step S112: Separate different types of data in the real-time status data, distinguish between monitoring operation data, physiological monitoring data and environmental parameter data, and arrange each type of data in the order of collection time to generate an independent time-series data sequence.

[0017] After receiving the aforementioned real-time status data, the central data processing system first classifies the data. In this embodiment, based on the data source and format, the data is divided into three main categories: monitoring operation data, physiological monitoring data, and environmental parameter data. For monitoring operation data, it is arranged in ascending order of the start time of the operation. Each record includes an operation type code, an operation start timestamp, an operation end timestamp, the operator ID, and operation-related parameters, forming a monitoring operation time-series data sequence. Physiological monitoring data is arranged in sampling time order. Each sampling time point corresponds to a set of physiological parameter values, forming a multi-dimensional array. Each element of the array corresponds to a physiological parameter, and the entire physiological monitoring data constitutes a two-dimensional time-series data sequence with time as the horizontal axis and parameter type as the vertical axis. Environmental parameter data is also arranged in collection time order, with each time point recording a set of environmental parameter values, forming an environmental parameter time-series data sequence.

[0018] Step S113: Extract the operation execution node information from the monitoring operation data, record the start time, key actions and completion time of each monitoring operation, and generate a monitoring operation time sequence node record. The monitoring operation time sequence node record marks the specific content and time coordinate of each operation node.

[0019] From the time sequence data of monitoring operations, further information on operation execution nodes is extracted. Taking suctioning as an example, after identifying the start timestamp, this time point is marked as the initiation node, and the operation type is recorded as "suspension via artificial airway." Key actions during the operation are extracted, such as "connecting the suction catheter," "inserting the suction catheter to the appropriate depth," "starting negative pressure suction," and "rotating and withdrawing the suction catheter." Each key action corresponds to a time point, which is obtained by analyzing sub-operation records during the operation or by key action markers manually entered by nursing staff. The operation completion timestamp is marked as the completion node. The initiation nodes, key action nodes, and completion nodes are arranged in chronological order. Each node includes a node type (initiation, key action, completion), a node time coordinate (timestamp accurate to the second), and a detailed description of the node content (e.g., "starting negative pressure suction"), thereby generating a monitoring operation time sequence node record. For example, the timing record of a suctioning operation may include: start node (time T1, content "sputum suctioning begins through artificial airway"), key action node 1 (time T1+15 seconds, content "connect suction tube"), key action node 2 (time T1+30 seconds, content "insert suction tube to appropriate depth"), key action node 3 (time T1+45 seconds, content "start negative pressure suction"), key action node 4 (time T1+60 seconds, content "rotate and withdraw suction tube"), and completion node (time T1+75 seconds, content "sputum suctioning through artificial airway completed").

[0020] Step S114: Analyze the change characteristics in the physiological monitoring data, compare the physiological monitoring data at different time points, identify the change pattern of data values, record the process data of the physiological state transitioning from one stable state to another, and generate a time-series record of physiological state changes. The time-series record of physiological state changes fully presents the process of physiological state change.

[0021] For physiological monitoring data, a sliding window analysis method is used to identify change characteristics. A sliding window with a length of 5 minutes and a sliding step of 1 minute is set, and the mean and standard deviation of each physiological parameter within each window are calculated. When the difference between the means of two consecutive windows exceeds a preset threshold (for example, the threshold for heart rate is set at 15% of the baseline heart rate, and the threshold for blood pressure is set at 10% of the baseline blood pressure), the physiological state is considered to have begun to change. Subsequently, the steady-state data before the change began (usually physiological data within 30 minutes before the change began, and the standard deviation of the data during this period is less than the preset stability threshold) are traced, and the mean values ​​of each parameter in the steady-state are recorded. Next, all physiological data points from the beginning of the change to the formation of a new steady state are tracked. The new steady state is defined as the difference between the means of three consecutive sliding windows being less than the stability threshold. The initial steady-state data, all data points during the transition process, and the new steady-state data are arranged in chronological order to generate a time-series record of physiological state changes. For example, a patient's heart rate is stable at around 85 beats / minute for 30 minutes before time T2, with a standard deviation of 3 beats / minute. Starting from time T2, the heart rate gradually increases, reaching 110 beats / minute at T2+10 minutes, and then stabilizes at around 105 beats / minute between T2+15 minutes and T2+45 minutes, with a standard deviation of 4 beats / minute. The time-series record of physiological state changes would then include stable heart rate data from T2-30 minutes to time T2, transitional heart rate data from T2 to T2+15 minutes, and new stable heart rate data from T2+15 minutes to T2+45 minutes, with the start and end times of each phase marked.

[0022] Step S115: Capture the change information in the environmental parameter data, record the specific values ​​of the environmental parameters at different time points, and generate a time sequence record of environmental parameter changes. The time sequence record contains the change data of the environmental parameters and the corresponding time points.

[0023] Environmental parameter data is collected every five minutes. First, the difference between each collected environmental parameter value and the previous collected value is checked. When the difference exceeds a preset change threshold (e.g., temperature change exceeding 1 degree Celsius, humidity change exceeding 5%), that time point is marked as a change node. For each change node, the current timestamp, parameter type, value before the change, value after the change, and change magnitude (value after the change minus the value before the change) are recorded. Simultaneously, even if the parameter value does not exceed the change threshold, the current value of the environmental parameter is recorded hourly as a routine monitoring record. All change node records and routine monitoring records are arranged chronologically to generate a time-series record of environmental parameter changes. For example, if the temperature in the ward is 23 degrees Celsius at time T3, and changes to 24.5 degrees Celsius at T3+5 minutes, exceeding the change threshold of 1 degree Celsius, then T3+5 minutes is marked as a temperature change node, and the timestamp, parameter type "temperature", value before the change (23 degrees Celsius), value after the change (24.5 degrees Celsius), and change magnitude +1.5 degrees Celsius are recorded. If the temperature remains around 24.5 degrees Celsius at subsequent time points such as T3+10 minutes and T3+15 minutes, and does not exceed the fluctuation threshold, then a normal temperature value will only be recorded once at T3+60 minutes.

[0024] Step S116: Using time as a unified dimension, align and integrate the monitoring operation time sequence node records, physiological state change time sequence records, and environmental parameter change time sequence records to generate a scene comprehensive time sequence evolution record.

[0025] In this embodiment, the timelines of the three types of time-series records are unified into millisecond-level timestamps. Then, using the timestamps as indexes, events and data points from different records are integrated onto the same timeline. For events at the same timestamp or very close timestamps (time difference less than 1 second), they are sorted and recorded according to the priority of monitoring operation nodes, physiological state change nodes, and environmental parameter change nodes. During the integration process, each time point may contain information from one or more records. For example, at time T4, the monitoring operation time-series record contains the key action node "starting negative pressure suction," the physiological state change time-series record shows that the heart rate has increased by 5 beats / minute compared to the previous second, and the environmental parameter change time-series record has no change information. Therefore, at time T4 in the scene comprehensive time-series evolution record, the key monitoring operation action and heart rate change information will be recorded simultaneously. Thus, through the above method, the three originally independent time-series records are integrated into a comprehensive record that can reflect various events and changes occurring in the monitoring scene at different time points.

[0026] Step S117: Extract change events from the scene comprehensive temporal evolution record and generate a scene change event set. The change event is a specific data change that causes a change in the scene state. Each change event includes the event occurrence time, event type and specific event representation.

[0027] From the comprehensive temporal evolution record of the scene, key data changes that significantly alter the scene's state are identified and defined as change events. Event types include monitoring operation events (such as suctioning and medication administration), physiological state events (such as increased heart rate and decreased blood oxygen saturation), and environmental change events (such as increased temperature and decreased humidity). For monitoring operation events, the specific characteristics include operation type and start / end time; for physiological state events, they include parameter type, value before change, value after change, and magnitude of change; for environmental change events, they include parameter type, value before change, value after change, and magnitude of change. In this embodiment, the entire comprehensive temporal evolution record of the scene is traversed, and each data change that meets the preset event definition is extracted and arranged in chronological order to generate a set of scene change events. For example, change events such as "At time T5, suctioning begins" (monitoring operation event), "At T5+1 minute, blood oxygen saturation drops from 95% to 90%" (physiological state event), and "At T5+5 minutes, ward temperature rises from 24 degrees Celsius to 25 degrees Celsius" (environmental change event) are extracted from the comprehensive record.

[0028] Step S118: Based on the preset causal rule base, perform correlation analysis on each event in the scene change event set, generate causal chains between events, and generate scene event causal correlation records. The scene event causal correlation records contain causal logic identifiers and correlation strength values ​​between events.

[0029] The pre-defined causal rule base is built based on medical knowledge and clinical experience, containing rules governing potential causal relationships between various events. For example, the rule base may include rules such as "suctioning may cause a temporary decrease in blood oxygen saturation" and "an increase in body temperature may cause an increase in heart rate." In this embodiment, events in the set of scene change events are paired and checked to see if they conform to a rule in the causal rule base. For event pairs that conform to the rule, the time interval between them is further calculated to see if it is within the time range allowed by the rule (e.g., a decrease in blood oxygen saturation occurring within 1-3 minutes after suctioning may have a causal relationship). If it does, a causal relationship is marked between the two events, and the association strength value is calculated based on the rule's confidence level and the event's time. The association strength value ranges from 0 to 1, with a higher value indicating a higher probability of a causal relationship. For example, the events "suctioning begins (T5)" and "blood oxygen saturation decreases (T5+1 minutes)" conform to the relevant rule in the rule base, and the time interval of 1 minute is within the allowed range, therefore a causal relationship is marked between them, and the association strength value is calculated to be 0.85 based on the rule's confidence level and time. Record all identified causal event pairs, their corresponding causal logic identifiers (such as "sucking leads to a decrease in blood oxygen"), and their correlation strength values ​​to generate a causal relationship record for scene events.

[0030] Step S119: Integrate the set of scene change events and the causal relationship records of scene events, and combine them in a structured manner according to time sequence and causal logic to present the evolution path of the scene from the initial state to the current state.

[0031] The events in the scene change event set are arranged in chronological order to form an event timeline. Then, based on the causal relationship record of scene events, directed arrows are used to connect events with causal relationships on the event timeline, with the arrows pointing from the cause event to the result event, and the correlation strength value is marked on the arrows. For a result event that may have multiple cause events, all related cause events are pointed to the result event via arrows. At the same time, events without a clear causal relationship are kept in their independent position on the timeline. In this way, a scene evolution path that reflects both the chronological order of events and the causal relationships between events is constructed. For example, the event timeline sequentially shows "Suctioning operation begins (T5)", "Blood oxygen saturation decreases (T5+1 minute)", and "Heart rate increases (T5+2 minutes)", where "Suctioning operation begins" points to "Blood oxygen saturation decreases" (correlation strength 0.85), and "Blood oxygen saturation decreases" points to "Heart rate increases" (correlation strength 0.7), thus showing the evolution path from the start of suctioning operation to the increase in heart rate.

[0032] Step S1110: Extract the change features in the evolution path, summarize the direction of change and the representation form in the scene evolution process, and generate dynamic evolution features that represent the dynamic changes of the scene.

[0033] Feature extraction is performed on the constructed scenario evolution path. First, the frequency and temporal distribution of various events are statistically analyzed, such as the number of suctioning operations in the past 2 hours and the proportion of changes in various parameters among physiological state events. Second, the length and complexity of the causal chain are analyzed, such as the average number of result events triggered by each causal event and the number of events contained in the longest causal chain. Then, key turning points in the evolution path are identified, i.e., those events that trigger a series of subsequent events or cause significant changes in the scenario state, such as endotracheal intubation or severe arrhythmia events. Finally, the overall trend of scenario evolution is summarized, such as whether the physiological state tends to stabilize or deteriorate, and whether the frequency of monitoring operations increases or decreases. The above statistical data, causal chain features, key turning point information, and overall trend information are integrated to form dynamic evolution features that comprehensively characterize the dynamic changes of the scenario. For example, dynamic evolution features might include: 2 suctioning operations in the past 2 hours, 3 events of decreased blood oxygen saturation, an average causal chain length of 2.5, a key turning point being the suctioning operation at time T5, and an overall trend of fluctuating deterioration in the physiological state.

[0034] Step S120: Construct a dynamic demand model for patient monitoring based on the dynamic evolution characteristics. The dynamic demand model includes the association rules and changing trends of demand as the scenario evolves.

[0035] In the aforementioned intensive care unit (ICU) scenario, based on the extracted dynamic evolution features, a dynamic demand model is constructed to reflect how patient monitoring needs change with the evolution of the scenario. This dynamic demand model will comprehensively consider the dynamic changes in monitoring operations, physiological states, and environmental factors, as well as the causal relationships between them, thereby predicting and describing the types, priorities, and content depth of monitoring information needs at different stages of scenario evolution.

[0036] Step S121: Extract the scene evolution driving elements from the dynamic evolution features. The scene evolution driving elements include the advancement nodes of the monitoring operation, the representation of changes in the patient's physiological state, and the change type of environmental factors.

[0037] From the dynamic evolutionary characteristics, we identify and extract the key elements that directly drive the scene evolution. The progression nodes of monitoring operations are the key action nodes and completion nodes in the previously generated monitoring operation sequence node records, such as "insertion of suction catheter" and "completion of medication administration." These nodes mark the progress of the monitoring operation and directly affect the subsequent scene state. The representation of changes in the patient's physiological state includes changes in various physiological parameters recorded in the physiological state change sequence records, such as "heart rate increases from 85 beats / minute to 110 beats / minute" and "blood oxygen saturation decreases from 95% to 90%." These representations reflect changes in the patient's condition and are the core factors driving changes in monitoring needs. The type of change in environmental factors is determined based on the parameter type and direction of change in the environmental parameter change sequence records, such as "increased temperature" and "decreased humidity." Specific types of environmental changes may affect the patient's comfort and condition, thereby indirectly driving adjustments in monitoring needs. These elements are extracted as the main driving forces of scene evolution.

[0038] Step S122: Divide the types and dimensions of the driving elements of scene evolution, classify them according to the source attributes of the driving elements, and distinguish between monitoring operation type driving elements, physiological state type driving elements and environmental change type driving elements.

[0039] The extracted scene evolution drivers were categorized according to their source attributes. Monitoring operation drivers originate from key moments in various monitoring procedures performed by nursing staff, such as suctioning, turning over, administering medication, and changing dressings. These drivers typically have clear human operational characteristics and timeframes. Physiological state drivers originate from changes in the patient's own physiological parameters, such as increases, decreases, or fluctuations in heart rate, blood pressure, blood oxygen saturation, respiratory rate, and body temperature. These drivers directly reflect changes in the patient's physiological state. Environmental change drivers originate from changes in ward environmental parameters, such as variations in temperature, humidity, light, and air composition. While these drivers do not directly reflect the patient's condition, they may affect patient comfort and treatment outcomes, thus indirectly influencing monitoring needs.

[0040] Step S123: Analyze the evolution trajectory characteristics of the scene evolution driving elements under each type dimension, extract the change path of the elements over time, record the specific state data of the elements at different time nodes, and generate independent evolution path records for each type of element. The independent evolution path records for each type of element present all the change details of the elements from the initial state to the current state.

[0041] For each type of driving factor, its evolution trajectory over time is analyzed. For monitoring operation-related driving factors, all operation progression nodes are arranged chronologically, and the operation type, occurrence time, duration, and state changes before and after the operation (e.g., patient's blood pressure before and after drug administration) are recorded for each node. For example, the evolution path record for monitoring operation-related factors may include: intravenous injection of drug A begins at time T6, lasting 5 minutes, with blood pressure of 100 / 60 mmHg before injection and rising to 110 / 65 mmHg after injection; a turning operation is performed at time T7, lasting 10 minutes, with the patient in a supine position before turning and a lateral position after turning, etc. For physiological state-related driving factors, for each physiological parameter, all numerical changes from the start of monitoring to the current time are recorded, including the numerical range during the stable period, the numerical fluctuations during the changing period, and the time points and durations of the changes. For example, an independent evolutionary path record for heart rate might include: from T0 to T10 minutes, the heart rate stabilizes at 80-85 beats per minute; from T10 to T15 minutes, the heart rate gradually increases to 95 beats per minute; from T15 to T25 minutes, the heart rate fluctuates between 90-100 beats per minute, and so on. For environmental change-related driving factors, the numerical changes of each environmental parameter are also recorded chronologically, including the time point of the change, the values ​​before and after the change, and the duration of the change. Through this method, an independent evolutionary path record is generated for each type of driving factor, presenting a detailed picture of its entire change process from its initial state to its current state.

[0042] Step S124: Integrate the evolution paths of scene evolution driving elements under different types and dimensions, locate the interaction nodes between elements, capture the associated time points when changes in elements of one type and dimension trigger changes in elements of other types and dimensions, and construct a multi-dimensional element association evolution record. The multi-dimensional element association evolution record marks the association logic and triggering order between each element.

[0043] Step S1241: Identify key time nodes from the scene evolution-driven element evolution path under each type dimension. The key time nodes are specific time points of element change, and each key time node corresponds to a determined element state.

[0044] In the evolutionary path of monitoring operations, key time nodes include the start time, execution time, and completion time of each operation. For example, the start time T5 of suctioning, the insertion time of the suction catheter T5+30 seconds, and the completion time T5+75 seconds, etc., with each node corresponding to a specific state of the operation. In the evolutionary path of physiological states, key time nodes are the time when physiological parameters begin to change, the time when they reach their peak or trough, and the time when they stabilize in the new state. For example, the time when heart rate begins to rise T2, the time when heart rate reaches its peak of 110 beats / minute T2+10 minutes, and the time when it stabilizes at 105 beats / minute T2+15 minutes, etc. In the evolutionary path of environmental changes, key time nodes are the time when environmental parameters begin to change, the time when they reach their peak, and the time when they stabilize. For example, the time when temperature begins to rise T3, and the time when temperature reaches 24.5 degrees Celsius T3+5 minutes, etc. Therefore, the above key time nodes are extracted from their respective evolutionary paths.

[0045] Step S1242: Using time as a unified benchmark, align the key time nodes to a unified time axis to generate a multi-dimensional key node set, obtain the state combination data of different types of dimensional elements at the same key time node, and determine the interaction relationship of each element at the key time node according to the preset relationship rules.

[0046] All extracted key time nodes are converted into uniform millisecond-level timestamps and then arranged on the same timeline to form a multi-dimensional key node set. For each key time node on the timeline, it is checked whether there is element status data from different types of dimensions at that time point. For example, at the key node of time T5+1 minutes, there may be a monitoring operation status of "suctioning in progress", a physiological status status of "blood oxygen saturation 90%" and an environmental change status of "temperature 24 degrees Celsius" at the same time. The preset relationship rule base contains the possible interaction relationships between different types of elements, such as "suctioning may cause a decrease in blood oxygen saturation" and "an increase in temperature may cause an increase in heart rate". In this embodiment, based on these rules, it is determined whether there is an interaction relationship between different types of elements at the same key time node.

[0047] Step S1243: Mark the combination of elements that have an interaction relationship at the same key time node, and generate node element association pairs. Each node element association pair contains elements of different types and dimensions and their corresponding interaction representations.

[0048] For elements identified as having an interaction relationship at the same critical time point, they are grouped into node element association pairs. For example, at time T5+1 minutes, "suctioning in progress" (a monitoring operation element) and "blood oxygen saturation drops from 95% to 90%" (a physiological state element) have an interaction relationship, and they are marked as a node element association pair, with the interaction characterized as "suctioning leads to a decrease in blood oxygen saturation". Each association pair includes the types of the two elements, a specific state description, and the direction and nature of the interaction.

[0049] Step S1244: Track the changes in the node element association pairs between adjacent key time nodes, record the changes in the state of the elements in the association pairs and the adjustments in the interaction relationships, and generate the association pair evolution sequence.

[0050] Arrange key time nodes in chronological order and observe the changes in the correlation pairs between adjacent nodes. For example, at time T5+1, the correlation pair is "suctioning in progress" and "decreasing blood oxygen saturation." By the next key time node T5+2, the correlation pair may change to "suctioning in progress" and "increased heart rate." At the same time, a new correlation pair may also form between "decreasing blood oxygen saturation" and "increased heart rate." Record the addition, disappearance, and changes of these correlation pairs, including specific changes in the state of the elements (such as blood oxygen saturation continuing to decrease to 88%) and adjustments in the interaction relationship (such as changing from "causing" to "accompanying"), thereby generating a correlation pair evolution sequence.

[0051] Step S1245: Identify common change patterns from the evolutionary sequences of the associated pairs, and generate general rules for cross-type element association evolution based on the common change patterns.

[0052] Pattern recognition is performed on the evolutionary sequences of association pairs to identify recurring patterns of change in association pairs across different time periods or events. For example, the association pair of "decreased blood oxygen saturation" followed by "increased heart rate" was repeatedly observed after "suspension procedure began." This pattern can be summarized as a common change pattern of "suspension procedure → decreased blood oxygen saturation → increased heart rate." Based on this common change pattern, general rules for the evolution of cross-type element associations are generated, such as "suspension procedure usually first leads to a decrease in blood oxygen saturation, which then causes an increase in heart rate."

[0053] Step S1246: Identify variation cases that do not conform to the general rules from the evolutionary sequence of the association pairs, record the triggering conditions and representation forms of the variation cases, and generate special association evolution supplementary rules.

[0054] In the evolutionary sequence of association pairs, there may be some special cases that do not conform to the general rules. For example, after a suctioning procedure, blood oxygen saturation may increase instead of decrease, or heart rate may decrease instead of increase. In this embodiment, these special change cases are identified, and the triggering conditions at the time are recorded in detail, such as the patient's underlying condition, the specific method of operation, environmental factors, etc., as well as the manifestation of the special change, such as the magnitude of the increase in blood oxygen saturation and the time delay of heart rate changes. Based on the above special cases, supplementary special association evolution rules are generated to improve the description of the element association evolution.

[0055] Step S1247: Construct a structured presentation of the multi-dimensional element association evolution record. The structured presentation includes a time axis, a type dimension axis, and an association strength axis. The three axes are perpendicular to each other to form a three-dimensional record.

[0056] Design a three-dimensional structured representation to record the evolution of multi-dimensional element relationships. The time axis extends horizontally, marking key time nodes; the type dimension axis extends vertically, distinguishing elements related to monitoring operations, physiological states, and environmental changes; the relationship strength axis extends along the depth direction, representing the magnitude of the relationship between elements. The three axes are perpendicular to each other, forming a three-dimensional coordinate system.

[0057] Step S1248: Incorporate node element association pairs, association pair evolution sequences, and special association evolution rule sets into a structured presentation format, arrange them in chronological order and by type dimension, label the association strength of each association pair, and generate a multi-dimensional element association evolution record.

[0058] In the aforementioned three-dimensional coordinate system, node element association pairs are placed in their corresponding positions according to their key time nodes and element type dimensions. Interrelated elements are connected by lines, with the line thickness or color depth adjusted based on the association strength value (the higher the association strength value, the thicker the line or the darker the color). The evolution sequence of association pairs is represented by dynamic changes in the lines, such as their appearance, disappearance, or color changes. A set of special association evolution rules is appended as annotation information to the corresponding special change cases. Through this method, all element association evolution information is integrated into a three-dimensional structure, generating a multi-dimensional element association evolution record.

[0059] Step S125: Extract key evolution nodes from the multi-dimensional element association evolution record. The key evolution nodes are the core time points that trigger changes in demand direction. The location of key evolution nodes is determined by analyzing the element association strength and the degree of demand influence. Each key evolution node corresponds to a specific scenario state and element combination.

[0060] First, a comprehensive impact index is calculated for each node element association pair in the multi-dimensional element association evolution record. The calculation of the comprehensive impact index considers the element association strength, the importance weight of the associated elements, and the position of the association pair in the entire evolution path. The element association strength is the previously calculated association strength value; the element importance weight is pre-set based on the importance of the element to patient monitoring, such as physiological state elements typically having a higher weight than environmental change elements; the position weight is determined according to the node's order in the evolution path, with nodes closer to the current moment having a higher weight. The formula for calculating the comprehensive impact index is: Comprehensive Impact Index = Association Strength × Element Importance Weight × Position Weight. Then, a comprehensive impact index threshold is set. When the comprehensive impact index of a node element association pair exceeds this threshold, its corresponding key time node is marked as a key evolution node. Each key evolution node corresponds to a specific scenario state (e.g., the patient is in a hypoxic state after suctioning) and element combination (e.g., the combination of "suctioning procedure" and "decreased blood oxygen saturation"). For example, in the above-mentioned intensive care scenario, the combined impact index of the node element association at T5+1 minutes on "sputum suction in progress" and "blood oxygen saturation drops from 95% to 90%" exceeds the threshold. Therefore, T5+1 minutes is identified as a key evolution node, and its corresponding scenario state is "hypoxia after suctioning". The element combination is "sputum suction" in the monitoring operation category and "decrease in blood oxygen saturation" in the physiological state category.

[0061] Step S126: Generate the demand representation content corresponding to each key evolution node. The generation of the demand representation content is based on the specific state of each type of driving element at the key evolution node. The demand representation content includes the information types and application directions required for patient monitoring under the key evolution node, and generates a node demand representation set.

[0062] For each critical evolutionary node, the needs representation content for patient monitoring is determined based on the specific status of each type of driving factor at that node. Information types include physiological parameter monitoring information (such as real-time trends in blood oxygen saturation and heart rate variability analysis), operational guidance information (such as procedures for managing post-suction hypoxia and medication adjustment suggestions), and equipment status information (such as whether ventilator parameter settings are appropriate and whether monitor alarm thresholds need adjustment). The application direction clarifies the purpose of this information, such as assessing changes in the patient's condition, guiding immediate actions, and optimizing treatment plans. For example, at the critical evolutionary node of T5+1 minutes (post-suction hypoxia), the needs representation content might include: real-time blood oxygen saturation monitoring curve (information type: physiological parameter monitoring information; application direction: assessing the degree and duration of hypoxia), post-suction hypoxia management guidelines (information type: operational guidance information; application direction: guiding nursing staff to administer oxygen or adjust ventilator parameters), and ventilator oxygen concentration setting suggestions (information type: equipment status information; application direction: optimizing ventilator parameters to improve oxygenation). The needs representation content for each critical evolutionary node is collected to form a node needs representation set.

[0063] Step S127: Analyze the demand representation change trajectory between adjacent key evolution nodes, extract the transition logic of demand from one key evolution node to another, summarize the triggering conditions and representation forms of demand changes, and generate a set of demand transition rules.

[0064] By arranging key evolutionary nodes chronologically, the demand representation content of adjacent nodes is compared to analyze how the demand transitions from one node to another. For example, the demand representation of the first key evolutionary node (T5+1 minutes) is "post-suction hypoxia management," while the demand representation of the second key evolutionary node (T5+5 minutes, assuming blood oxygen saturation recovers to 93%) is "respiratory function assessment." The trajectory of demand representation change is from "managing hypoxia emergencies" to "assessing respiratory function recovery." The transition logic might be that "when blood oxygen saturation recovers to above 92% and stabilizes for 5 minutes, the demand shifts from emergency management to functional assessment." Triggering conditions include the numerical range of blood oxygen saturation and the stabilization time. The change in representation form is from emphasizing operational guidance information to emphasizing physiological parameter trend analysis and functional assessment indicators. This analysis is performed on each pair of adjacent key evolutionary nodes to summarize the triggering conditions for demand changes (such as specific physiological parameters reaching a certain threshold, a certain time after operation completion, etc.) and representation forms (such as the increase or decrease of information types, changes in application direction, etc.). These summaries are then compiled into a set of demand transition rules.

[0065] Step S128: Construct a requirement evolution correlation matrix, which integrates the requirement representation content and requirement transition rule set of key evolution nodes into a matrix. The row dimension of the matrix represents the key evolution nodes, and the column dimension represents the requirement representation and transition rules of the corresponding key evolution nodes.

[0066] A two-dimensional matrix is ​​constructed, where the rows represent key evolutionary nodes arranged chronologically, with each row representing a key evolutionary node. The columns are divided into two parts: a demand representation column, containing various information types within the demand representation content of that node, and a transition rule column, containing the transition rules from the current node to the next. The values ​​of the matrix elements are determined based on specific circumstances. For the demand representation column, if the demand representation of a node contains a certain information type, the element value is "1"; otherwise, it is "0". For the transition rule column, the element value is a description of the transition rule from the current node to the next. For example, in the demand representation column of a certain row (corresponding to the key evolutionary node T5+1 minutes), the elements for "physiological parameter monitoring information" and "operational guidance information" are both "1", and the element for "equipment status information" is also "1". The element value of the transition rule column is "When blood oxygen saturation rises above 92% and remains stable for 5 minutes, transition to respiratory function assessment requirements". Through this matrix integration, the demand representations of each key evolutionary node and the transition rules between them are displayed.

[0067] Step S129: Generate a predictor of demand change trends. The predictor is obtained based on the trajectory characteristics of demand changes in historical scenario evolution data. The predictor is associated with the demand representation of key evolution nodes and the demand transition direction of subsequent key evolution nodes.

[0068] We collected a large amount of evolutionary data from historical monitoring scenarios, including key evolutionary nodes, demand representations, and demand transitions of past patients. For the demand representations at each key evolutionary node, we analyzed the possible subsequent demand transition directions and their probabilities. For example, the demand representation of "post-suction hypoxia treatment" may subsequently transition to "respiratory function assessment," "continued hypoxia treatment," or "transition to other emergency needs," each with a certain probability of occurrence. Based on the above historical data, we trained a predictive model using statistical analysis methods (such as logistic regression and decision trees). The model's input is the demand representation characteristics of the current key evolutionary node (such as information type combinations and application directions), and the output is each possible subsequent demand transition direction and its probability. The trained model parameters are used as predictive factors, which can correlate the demand representation of the current key evolutionary node with possible subsequent demand transition directions. For example, predictors might include: "When the demand profile includes 'decreased blood oxygen saturation' and 'operational instructions' is the primary information type, the probability of transitioning to a 'respiratory function assessment' demand is 0.7, the probability of transitioning to a 'continued hypoxia treatment' demand is 0.2, and the probability of transitioning to other emergency demands is 0.1."

[0069] Step S1210: Integrate the demand representation set, demand transition rule set, demand evolution correlation matrix and prediction factors of key evolution nodes, and combine them in a structured manner according to the time sequence and logical relationship of scenario evolution to generate a dynamic demand model.

[0070] Key evolutionary nodes are arranged chronologically, with each node associated with specific content in its demand representation set. Based on a demand transition rule set, logical transition connections are established between adjacent nodes, clarifying how the demand from one node transitions to the demand of the next. A demand evolution correlation matrix serves as an auxiliary structure, visually displaying the correspondence between demand representations and transition rules between nodes in matrix form. Predictive factors are appended to each key evolutionary node to predict the possible subsequent development direction of demand at that node. Through this structured combination, the various components are organically integrated to form a dynamic demand model capable of dynamically describing the changes in patient monitoring needs as the scenario evolves. This dynamic demand model can predict future demand directions based on the current scenario state and evolutionary trends.

[0071] Step S130: Reconstruct the scenario adaptation knowledge module in the guardianship knowledge reserve based on the dynamic demand model constructed according to the dynamic evolution characteristics, and restructure the scenario adaptation knowledge module according to the demand change trend.

[0072] In this intensive care unit (ICU) scenario, the monitoring knowledge base contains a wealth of information related to patient monitoring, such as disease treatment guidelines, nursing operation standards, equipment usage instructions, and drug information. Based on a dynamic requirements model, this knowledge is filtered, organized, and reorganized to construct a scenario-adaptive knowledge module that matches the evolving needs of the current scenario. This scenario-adaptive knowledge module will structure the knowledge content according to the changing needs described by the dynamic requirements model.

[0073] Step S131: Analyze the demand representation content of the key evolution nodes in the dynamic demand model, and extract the information demand type and application scenario requirements corresponding to each key evolution node.

[0074] For each key evolutionary node in the dynamic demand model, its demand representation content is analyzed in detail. The required information types for this node are identified, such as physiological parameter monitoring information, operational guidance information, equipment status information, and medication information. The specific content requirements for each information type are clarified; for example, physiological parameter monitoring information may require a real-time trend graph of blood oxygen saturation and statistical data from the most recent hour. Simultaneously, the application scenario requirements corresponding to this key evolutionary node are analyzed, including the urgency level of the scenario (e.g., emergency treatment, routine monitoring, assessment and analysis), the personnel involved (e.g., nurses, doctors), and the usage environment (e.g., bedside, central monitoring station). For example, at the key evolutionary node of "post-suction hypoxia management," the information demand types include real-time blood oxygen saturation monitoring information (requiring the inclusion of the trend curve of the past 5 minutes and the current value), post-suction hypoxia management operational guidance information (requiring specific treatment steps and precautions), and ventilator parameter adjustment information (requiring the inclusion of oxygen concentration and PEEP adjustment recommendations); the application scenario requirement is an emergency treatment scenario, primarily for bedside nurses.

[0075] Step S132: Extract basic knowledge units from the monitoring knowledge reserve that match the requirements of each key evolution node. The basic knowledge unit is the smallest knowledge fragment that can directly meet the requirements of the corresponding key evolution node. The basic knowledge units are extracted one by one in the order of the key evolution nodes. Each basic knowledge unit fully covers the key requirements of the corresponding key evolution node.

[0076] Based on the information needs and application scenarios of each key evolutionary node, searches and matching are performed within the monitoring knowledge reserve. Basic knowledge units are the smallest indivisible knowledge fragments in the knowledge reserve, such as a specific operational procedure description, a definition of a normal range for a physiological parameter, or instructions for the use and dosage of a medication. For example, for the key evolutionary node of "post-suction hypoxia management," the basic knowledge unit "post-suction hypoxia management procedure" is extracted from the knowledge reserve. This basic knowledge unit includes immediate measures to be taken when blood oxygen saturation is below 90%, such as increasing oxygen concentration, checking airway patency, and notifying the doctor, fully covering the key operational guidance requirements for this node. Following the chronological order of the key evolutionary nodes, corresponding basic knowledge units are extracted and matched for each node, ensuring that each unit directly meets the needs of that node.

[0077] Step S133: Verify the matching relationship between the basic knowledge unit and the corresponding key evolution node requirement representation. When the content of the basic knowledge unit covers the information requirement type of the key evolution node, record it as a node-knowledge unit matching pair.

[0078] The extracted basic knowledge units are matched and verified with the corresponding key evolutionary nodes' requirement representations. The verification process includes checking whether the information type of the basic knowledge unit matches the information type in the requirement representation, whether the content fully covers the key requirements, and whether the application scenario matches. For example, verifying whether the basic knowledge unit "Post-suction hypoxia treatment procedure" covers the operational guidance information requirement of the key evolutionary node "Post-suction hypoxia treatment," and checking whether the unit includes all necessary treatment steps, precautions, and judgment criteria. If the verification passes, meaning the content of the basic knowledge unit fully covers the information requirement type of the key evolutionary node, then the key evolutionary node and the basic knowledge unit are recorded as a node-knowledge unit matching pair. For example, the "Post-suction hypoxia treatment" node and the "Post-suction hypoxia treatment procedure" knowledge unit form a matching pair.

[0079] Step S134: Connect the basic knowledge units corresponding to adjacent key evolution nodes. Based on the demand transition rule set in the dynamic demand model, determine the connection logic for the transition of the basic knowledge unit of the previous key evolution node to the basic knowledge unit of the next key evolution node, and construct the knowledge unit transition link. The knowledge unit transition link includes the connection order and association method of the knowledge units.

[0080] Based on the demand transition rule set in the dynamic demand model, the demand transition logic between two adjacent key evolution nodes is analyzed to determine the connection logic between their corresponding basic knowledge units. For example, the first key evolution node is "post-suction hypoxia treatment," and the corresponding basic knowledge unit is "post-suction hypoxia treatment procedure"; the second key evolution node is "respiratory function assessment," and the corresponding basic knowledge unit is "respiratory function assessment indicators and methods." The demand transition rule set stipulates that when blood oxygen saturation rises above 92% and remains stable for 5 minutes, the demand transitions from "post-suction hypoxia treatment" to "respiratory function assessment." Therefore, the connection logic between the two basic knowledge units is: when the measures in the "post-suction hypoxia treatment procedure" knowledge unit are implemented and the patient's blood oxygen saturation reaches the above condition, the connection to the "respiratory function assessment indicators and methods" knowledge unit is established. The above connection logic is recorded in the connection order of knowledge units (first "post-suction hypoxia treatment procedure," then "respiratory function assessment indicators and methods") and the association method (triggered by blood oxygen saturation conditions) to construct the knowledge unit transition link.

[0081] Step S135: For the information gap between adjacent basic knowledge units, extract transitional knowledge fragments that can fill the gap from the monitoring knowledge reserve to supplement the connecting knowledge content in the transition link between adjacent basic knowledge units.

[0082] Step S1351: Based on the content association logic of adjacent basic knowledge units, determine the classification of information gaps, including conceptual connection gaps, logical deduction gaps, and application guidance gaps.

[0083] Analyze the content relationships between adjacent basic knowledge units to determine if there are information gaps and their types. Conceptual connection gaps refer to differences in concepts or terminology between two knowledge units, requiring an intermediate concept for connection. For example, the "Post-Suctioning Hypoxia Treatment Procedure" mentions "PEEP adjustment," while "Respiratory Function Assessment Indicators and Methods" directly uses "the effect of PEEP settings on oxygenation." If the former unit doesn't explain the concept of PEEP in detail, and the latter doesn't review it, then there is a conceptual connection gap. Logical deduction gaps refer to a lack of necessary logical reasoning steps between the conclusion of one knowledge unit and the premise of another. For example, after the "Post-Suctioning Hypoxia Treatment Procedure," it directly transitions to "Respiratory Function Assessment," but doesn't explain why respiratory function assessment is necessary after hypoxia treatment; this lack of logical deduction constitutes a logical deduction gap. Application guidance gaps refer to a lack of specific operational guidance when transitioning from one knowledge unit to another in practical application. For example, after completing the hypoxia treatment procedure following suctioning, how to switch to the respiratory function assessment procedure, and how to connect the steps, create a gap in application guidelines.

[0084] Step S1352: To address the gaps in concept connection, extract transitional concept explanations from the guardianship knowledge reserve that can connect adjacent basic knowledge unit concepts.

[0085] For gaps in conceptual connections, explanatory knowledge fragments related to the missing concept are retrieved from the monitoring knowledge base. For example, for the aforementioned gap in the concept of PEEP, the explanatory content of "definition, mechanism of action and preliminary introduction to clinical application of positive end-expiratory pressure (PEEP)" is extracted from the knowledge base. This explanatory content can explain the basic concept of PEEP and connect the use of PEEP in the two knowledge units before and after it.

[0086] Step S1353: For the gap in logical deduction, extract the reasoning basis content that conforms to the professional logic of the guardianship field from the guardianship knowledge reserve. The reasoning basis content is used to support the logical transition from one knowledge unit to another.

[0087] For gaps in logical deduction, extract knowledge fragments that provide logical support. For example, for the logical gap "Why is a respiratory function assessment necessary after hypoxia treatment?", extract the knowledge fragment "The necessity of assessing the impact of hypoxia events on respiratory function" from the knowledge reserve. This knowledge fragment explains that hypoxia may indicate potential respiratory function deterioration, therefore, respiratory function needs to be assessed after treatment to determine if there are persistent problems or if further intervention is needed, thus supporting the logical transition from hypoxia treatment to respiratory function assessment.

[0088] Step S1354: For the application guidance gap, extract the operation connection content that can guide the smooth transition between adjacent basic knowledge units in actual application. The operation connection content includes the coordination method and sequence between the application operations corresponding to the two knowledge units.

[0089] For gaps in application guidance, specific operational guidance segments were extracted. For example, regarding the gap in the connection between the post-suction hypoxia treatment procedure and the respiratory function assessment procedure, the knowledge segment "Connection Procedure between Post-Suction Hypoxia Treatment and Respiratory Function Assessment" was extracted. This knowledge segment details that after completing the hypoxia treatment procedure, the current blood oxygen saturation, heart rate, and other parameters should be recorded first. Then, temporary high-concentration oxygen inhalation (if applicable) should be stopped, and the respiratory function assessment indicators should be measured after 5 minutes. This clarifies the sequence of operations and the coordination method.

[0090] Step S1355: Perform consistency verification on the extracted supplementary content to ensure that the expression attributes of the supplementary content are consistent with the expression attributes of the basic knowledge unit. The supplementary content includes explanations of transitional concepts, reasoning basis content, and operational connection content.

[0091] Check whether the style of expression, use of technical terminology, and data units in the supplementary content are consistent with those in the basic knowledge unit. For example, if the basic knowledge unit uses "blood oxygen saturation" instead of "oxygen saturation," the supplementary content should also use "blood oxygen saturation"; if the basic knowledge unit uses mmHg for blood pressure, the supplementary content should also consistently use mmHg. If any inconsistencies are found, modify the supplementary content to ensure consistency with the expression attributes of the basic knowledge unit.

[0092] Step S1356: Trim or expand the supplementary content to match the coverage of the information gaps.

[0093] Adjust the supplementary content according to the size and specific content of the information gap. If the supplementary content is too lengthy and contains information beyond the scope of the gap, it should be trimmed, retaining only the parts directly related to the gap. If the supplementary content is too brief and fails to fully cover the gap, it should be appropriately expanded to add necessary details. For example, if the explanation of the transitional concept is too detailed in its historical development of PEEP, this part is irrelevant to the concept connection and should be trimmed; if the reasoning only briefly mentions the importance of assessment without explaining the specific reasons, it needs to be expanded to include information such as the specific changes in respiratory function that hypoxia may cause.

[0094] Step S1357: Generate a position index for the supplementary content in the knowledge unit transition link. The position index is generated based on the information gap type corresponding to the supplementary content and the positional relationship between adjacent basic knowledge units. It is used to insert the supplementary content between two basic knowledge units without changing the internal structure of the original knowledge unit.

[0095] The location index is generated using the format "Previous Knowledge Unit ID - Gap Type - Subsequent Knowledge Unit ID". For example, for a conceptual connection gap between "Post-Suctioning Hypoxia Treatment Procedure" (ID: KU001) and "Respiratory Function Assessment Indicators and Methods" (ID: KU002), the location index for the supplementary content is "KU001-Conceptual Connection-KU002". This location index allows for accurate insertion of supplementary content into the corresponding position between two basic knowledge units without affecting the internal structure of the original knowledge units.

[0096] Step S1358: Associate the corresponding information gap type identifier with the supplementary content, and set the type identifier as the metadata field of the supplementary content.

[0097] Add a metadata field to each piece of supplementary content to identify its corresponding information gap type, such as "conceptual connection," "logical deduction," or "application guidance." This facilitates the subsequent management and retrieval of supplementary content and also helps clarify the role of supplementary content during the use of the knowledge module.

[0098] Step S1359: Arrange all basic knowledge units and their corresponding supplementary content according to the position index and order to generate a complete knowledge unit transition link.

[0099] Based on the location index, the basic knowledge units and supplementary content are arranged sequentially. For example, first place "Hypoxia Management Procedure After Suctioning" (KU001), then insert supplementary content with the location index "KU001-Conceptual Connection-KU002" (PEEP Concept Explanation), followed by supplementary content with the location index "KU001-Logical Derivation-KU002" (Explanation of Assessment Necessity), then supplementary content with the location index "KU001-Application Guidelines-KU002" (Operational Connection Procedure), and finally place "Respiratory Function Assessment Indicators and Methods" (KU002). This method forms a complete knowledge unit transition link containing basic knowledge units and supplementary content.

[0100] Step S13510: Check the knowledge unit transition link to confirm that all information gaps between adjacent basic knowledge units have been filled with supplementary content, and that all basic knowledge units are connected sequentially through the supplementary content.

[0101] Traverse the entire knowledge unit transition chain, checking for any unfilled information gaps between each pair of adjacent basic knowledge units. For any gaps found, repeat the step of extracting supplementary content. Simultaneously, check whether the basic knowledge units are correctly connected through the supplementary content to form a continuous knowledge chain. For example, ensure that from the first basic knowledge unit, the supplementary content allows for a smooth transition to the next basic knowledge unit, up to the last basic knowledge unit in the chain, without any breaks or jumps.

[0102] Step S136: Based on the time sequence of key evolution nodes and the connection logic of knowledge unit transition links, reorganize basic knowledge units and transition knowledge fragments to generate a preliminary scene-adaptive knowledge module structure.

[0103] All the basic knowledge units corresponding to the key evolution nodes are arranged in chronological order in the dynamic requirements model. Then, according to the connection logic of the knowledge unit transition links, corresponding transitional knowledge fragments (i.e., the supplementary content mentioned above) are inserted between adjacent basic knowledge units. For example, the basic knowledge units corresponding to the key evolution nodes arranged in chronological order are KU001 (post-suction hypoxia treatment procedure), KU002 (respiratory function assessment indicators and methods), KU003 (ventilator parameter optimization suggestions), etc. Transitional knowledge fragments are inserted between KU001 and KU002, and corresponding transitional knowledge fragments are inserted between KU002 and KU003, thus forming a preliminary scenario adaptation knowledge module structure connected by chronological order and logical relationship.

[0104] Step S137: Divide the hierarchy of the scene-adaptive knowledge module structure, assign hierarchy identifiers to key evolution nodes according to preset rules, place the basic knowledge units corresponding to key evolution nodes with basic hierarchy identifiers in the basic hierarchy, and place the knowledge content of the connecting hierarchy in the connecting hierarchy.

[0105] The pre-defined hierarchical classification rules are determined based on the importance, urgency, and information type of key evolutionary nodes. For example, key evolutionary nodes related to emergency treatment that are directly related to patient life safety are assigned to the first-level basic level; key evolutionary nodes related to routine monitoring and assessment are assigned to the second-level basic level; and transitional knowledge fragments connecting nodes at different basic levels are assigned to the connecting level. For example, the key evolutionary node "post-suction hypoxia management" belongs to the emergency treatment category and is assigned to the first-level basic level, with its corresponding basic knowledge unit KU001 placed at the first-level basic level; the key evolutionary node "respiratory function assessment" belongs to the routine assessment category and is assigned to the second-level basic level, with its corresponding basic knowledge unit KU002 placed at the second-level basic level; and the transitional knowledge fragments connecting KU001 and KU002 are placed at the connecting level.

[0106] Step S138: Mark the application sequence of knowledge content in each level to determine the activation time and application duration of knowledge content in the process of scenario evolution.

[0107] For basic-level knowledge content, the activation time is the occurrence time of the corresponding key evolutionary node, and the application duration is determined based on the duration of the key evolutionary node or the duration of the requirement. For example, the key evolutionary node for "post-suction hypoxia management" occurs at T5+1 minutes, and the requirement continues until blood oxygen saturation stabilizes and recovers. Therefore, the activation time for KU001 is T5+1 minutes, and the application duration is from T5+1 minutes to T5+5 minutes (assuming blood oxygen is stable at this time). For transitional knowledge fragments at the connecting level, the activation time is a point in time before the end of the application duration of the previous basic knowledge unit (e.g., 30 seconds in advance), and the application duration is the transition process covering two basic knowledge units. For example, the transitional knowledge fragment connecting KU001 and KU002 has an activation time of T5+4.5 minutes (30 seconds before the end of the application duration of KU001), and an application duration of T5+4.5 minutes to T5+5.5 minutes, ensuring that monitoring personnel can obtain the connecting knowledge in a timely manner during the transition process. Mark the above activation time and application duration on the knowledge content.

[0108] Step S139: Integrate the knowledge content and application time sequence annotations in the hierarchy to generate a structured scenario adaptation knowledge module. The scenario adaptation knowledge module includes basic level knowledge, connecting level knowledge and corresponding application time planning.

[0109] Knowledge content at different levels (basic knowledge units at the foundational level and transitional knowledge fragments at connecting levels) is organized hierarchically, and application timing is associated with the corresponding knowledge content. Simultaneously, an application time plan is developed, specifying when each knowledge content will be activated and for how long. For example, the first-level foundational level includes KU001, activated at T5+1 minutes, with an application duration of 4 minutes; the connecting level includes transitional knowledge fragments linking KU001 and KU002, activated at T5+4.5 minutes, with an application duration of 1 minute; the second-level foundational level includes KU002, activated at T5+5 minutes, with an application duration of 5 minutes, and so on.

[0110] Step S140: Generate initial context-aware recommendation information based on the reconstructed scene adaptation knowledge module.

[0111] In this intensive care unit scenario, the reconstructed scenario-adaptive knowledge module includes basic knowledge units and transitional knowledge fragments organized hierarchically and chronologically. Based on this module, according to the current scenario's evolutionary stage and dynamic demand model, knowledge content matching the needs of the current key evolutionary nodes is selected and organized according to a preset presentation format and order to generate initial context-aware recommendation information.

[0112] Step S141: Analyze the hierarchical structure of the scenario-adaptive knowledge module, extract the specific content of basic-level knowledge and connecting-level knowledge, as well as their related logic and application order, and generate a knowledge module structure analysis record.

[0113] A detailed analysis of the hierarchical structure of the scenario-adaptive knowledge module is performed. This involves identifying the specific basic knowledge units contained in different basic levels, such as the first-level and second-level basic levels, as well as the transitional knowledge fragments contained in the connecting levels. The logical relationships between the knowledge content at each level are analyzed, such as how first-level basic level knowledge transitions to second-level basic level knowledge through connecting level knowledge. Simultaneously, the application order of the knowledge content is determined, i.e., which basic knowledge unit is applied first, which transitional knowledge fragment is activated subsequently, and so on. This information, including the list of knowledge content at each level, the description of the logical relationships, and the order of application, is recorded to generate a knowledge module structure analysis record. For example, the knowledge module structure parsing record may include: Level 1 basic knowledge: KU001 (post-suction hypoxia treatment procedure); Transitional knowledge: TC001 (PEEP concept explanation), TC002 (explanation of assessment necessity), TC003 (operational transition procedure); Level 2 basic knowledge: KU002 (respiratory function assessment indicators and methods); Association logic: KU001 transitions to KU002 through TC001, TC002, and TC003; Application order: KU001→TC001→TC002→TC003→KU002.

[0114] Step S142: Extract information units from the basic level knowledge one by one in the order of key evolutionary nodes. The information units are specific contents that directly meet the requirements of the corresponding key evolutionary nodes.

[0115] Based on the chronological order of key evolutionary nodes, corresponding information units are extracted from the basic-level knowledge. Each information unit is a specific content part of the basic knowledge unit that directly meets the needs of the corresponding key evolutionary node. For example, for the key evolutionary node of "post-suction hypoxia treatment," the information units extracted from KU001 (post-suction hypoxia treatment procedure) include: "When blood oxygen saturation <90%, immediately increase the oxygen concentration to 100%", "Check if the airway is open, suction if necessary", and "If blood oxygen saturation remains <90% for more than 2 minutes, notify the doctor," etc. These information units directly correspond to the operational guidance requirements of this key evolutionary node. Similar information units are extracted sequentially from the basic knowledge units of each basic-level knowledge according to the order of the key evolutionary nodes.

[0116] Step S143: Extract transition information units from the knowledge at the connection level according to the order of the knowledge unit transition link. The transition information units and information units are combined according to a preset connection logic.

[0117] Based on the order of transitional knowledge segments in the knowledge unit transition link, transitional information units are extracted from the knowledge at the connection level. A transitional information unit is the specific content within a transitional knowledge segment used to connect preceding and following information units. For example, the transitional information unit extracted from TC001 (PEEP concept explanation) is "PEEP refers to positive airway pressure applied at the end of expiration, which can prevent alveolar collapse and improve oxygenation"; the transitional information unit extracted from TC002 (explanation of assessment necessity) is "Hypoxia events may indicate impaired lung ventilation or gas exchange function; post-treatment assessment of respiratory function helps determine changes in the patient's condition"; and the transitional information unit extracted from TC003 (operational connection procedure) is "After completing hypoxia treatment, record the current parameters, stop high-concentration oxygen inhalation for 5 minutes, and then begin respiratory function assessment." These transitional information units are combined with preceding and following information units according to a pre-defined connection logic (e.g., first explaining the concept, then stating the necessity, and finally guiding the operational connection).

[0118] Step S144: Obtain the predefined presentation format of the information unit and the transition information unit. The predefined presentation format is set according to the application scenario requirements of the corresponding key evolution node and conforms to the information transmission standard in the field of monitoring, including the information display structure, content organization method and transmission format.

[0119] The predefined presentation formats of information units and transitional information units are pre-set in the system. For information units in emergency treatment scenarios, the presentation format is usually a striking list format, using bold headings to highlight key steps, with concise and clear content, such as "[Emergency Treatment Steps] 1. Immediately increase oxygen concentration to 100%; 2. Check airway patency...". For information units in routine assessment scenarios, assessment indicators and normal ranges may be displayed in tabular form, such as "Respiratory Function Assessment Indicators: |Indicator Name|Normal Range|Current Value|". Transitional information units are usually presented as short paragraphs of text, used to naturally connect preceding and following information units, such as "[Conceptual Connection] PEEP refers to positive airway pressure applied at the end of expiration, which can prevent alveolar collapse and improve oxygenation. This parameter is of great significance in subsequent respiratory function assessments." These presentation formats conform to the information transmission standards in the monitoring field.

[0120] Step S145: Mark the information unit with the corresponding scene evolution driving element, and mark the state of the driving element and the application triggering condition that it is adapted to in the information unit.

[0121] Each information unit is labeled with its associated scenario-driven evolution element. For example, the information unit "When blood oxygen saturation <90%, immediately increase oxygen concentration to 100%" is matched with the driving element state "physiological state driving element - blood oxygen saturation drops to <90%", and the application trigger condition is "blood oxygen saturation <90% after suctioning". These labels are added at the beginning or end of the information unit, such as "[Matching Element] Blood Oxygen Saturation <90% (after suctioning) [Trigger Condition] Blood Oxygen Saturation <90% after suctioning". Through the above labeling, the monitoring personnel can clearly understand which scenario conditions the information unit is applicable to.

[0122] Step S146: Arrange the presentation order of information units and transitional information units according to the time sequence of scene evolution and the application sequence of knowledge modules, and supplement the application instructions of information units. The application instructions include the usage method of information units in actual monitoring operations, the corresponding monitoring actions, and the expected effects after application.

[0123] Based on the chronological order of scenario evolution and the application sequence of knowledge modules, information units and transitional information units are arranged sequentially. For example, information units related to the key evolution node of "post-suctioning hypoxemia management" are arranged first, followed by transitional information units connecting to the key evolution node of "respiratory function assessment," and finally, information units related to the key evolution node of "respiratory function assessment." Each information unit is followed by application instructions. These instructions detail how the information unit is used in actual monitoring operations, such as "When increasing oxygen concentration, it should be done using the oxygen concentration adjustment knob on the ventilator panel; rotating it clockwise increases the concentration"; the corresponding monitoring actions, such as "After adjustment, observe the change in the pulse oximetry monitor reading"; and the expected effects after application, such as "Pulse oximetry is expected to begin to rise within 1-2 minutes after adjustment; if it does not rise, proceed to the next step." This arrangement and supplementation make the recommended information more complete and practical.

[0124] Step S147: Integrate information units, transitional information units, application instructions, and presentation formats to generate an information recommendation structure. In the order of the information recommendation structure, convert all information units and application instructions into standardized textual expressions to generate initial context-aware recommendation information.

[0125] The pre-arranged information units, transitional information units, and application instructions are integrated according to their predefined presentation formats to form a complete information recommendation structure. For example, the information recommendation structure might include: a title (e.g., "Recommended Information for Hypoxia Management After Suctioning"), emergency treatment steps (a list of information units), application instructions (how to use each step), conceptual connections (transitional information unit paragraphs), and respiratory function assessment indicators (an information unit table). Then, following the order of this information recommendation structure, all content is transformed into standardized textual expressions, ensuring accurate, concise language that conforms to medical terminology standards, ultimately generating initial context-aware recommendation information.

[0126] Step S150: Track the real-time change trajectory of the dynamic evolution features, adjust the parameters of the dynamic demand model, optimize the initial context-aware recommendation information based on the adjusted dynamic demand model, generate the final recommendation information set and push it to the monitoring execution terminal.

[0127] In this intensive care unit scenario, the patient's condition and the monitoring environment are dynamically changing. In this embodiment, the real-time trajectory of dynamic evolution characteristics is continuously tracked, and the parameters of the dynamic demand model are adjusted according to the new changes, so that the model can accurately reflect the current demand status. Then, based on the adjusted dynamic demand model, the initial context-aware recommendation information is optimized, the knowledge content is updated, the recommendation order and presentation format are adjusted, the final set of recommendation information is generated, and it is pushed to the monitoring execution terminal or mobile nursing device at the nurse station.

[0128] Step S151: Acquire dynamic evolution characteristics in the monitoring scenario through the real-time acquisition channel at preset time intervals.

[0129] In this embodiment, real-time status data of the monitoring scenario is continuously collected at preset time intervals (e.g., once per second) using previously deployed scene-aware terminals. This includes physiological monitoring data, monitoring operation data, and environmental parameter data. Then, following the method described in step S110, dynamic evolution features are extracted from this real-time status data, including scene change events, causal relationships between events, and evolution paths. For example, at T5+6 minutes, the real-time acquisition channel shows that the patient's blood oxygen saturation has recovered to 94%, the heart rate is stable at 95 beats / minute, the ward temperature is maintained at 24.5 degrees Celsius, and no new monitoring operations have been performed. Based on this, the dynamic evolution features are updated.

[0130] Step S152: Extract the newly added driving elements from the dynamic evolution features collected in real time. The newly added driving elements are newly emerging elements that were not previously included in the dynamic evolution features. The original representation and emergence time of the newly added driving elements are retained.

[0131] By comparing the currently extracted dynamic evolution features with the previously extracted dynamic evolution features, newly emerging scene evolution driving elements are identified. These newly added driving elements may be new monitoring operations (such as new medical orders issued by doctors), new physiological state changes (such as the occurrence of arrhythmias), or new environmental changes (such as sudden power outages). For example, at T5+7 minutes, the patient experienced brief premature ventricular contractions (PVCs). This is a physiological state driving element that did not appear in the previous dynamic evolution features, so it is extracted as a new driving element, and its original characteristics ("PVCs, frequency 5 times / minute") and occurrence time (T5+7 minutes) are recorded.

[0132] Step S153: Compare the newly added driving elements with the existing scenario evolution driving element types in the dynamic demand model, determine the type dimension to which the newly added driving element belongs, and classify the newly added driving element into the corresponding monitoring operation category, physiological state category, or environmental change category dimension.

[0133] The characteristics of the newly added driving factors are compared with the types of driving factors for scenario evolution defined in the dynamic demand model. For example, the aforementioned "ventricular premature beats" factor originates from the patient's physiological monitoring data and reflects changes in the patient's physiological state; therefore, it is determined to belong to the physiological state category of driving factors and is classified under the physiological state dimension. If the newly added driving factor is "the doctor's order for intravenous injection of drug B," it belongs to the monitoring operation category of driving factors; if the newly added driving factor is "the humidity in the ward suddenly drops to 30%," it belongs to the environmental change category of driving factors.

[0134] Step S154: Compare the newly added driving element with the existing multi-dimensional element association evolution record, and determine the impact category of the newly added driving element on the existing evolution path based on the comparison result. The impact category includes path fine-tuning and direction change, and record the specific representation and associated node corresponding to the impact category.

[0135] The emergence time, type dimension, and specific characteristics of newly added driving elements are compared with the existing node elements in the multi-dimensional element correlation evolution record. The analysis examines whether the newly added driving elements are correlated with existing elements and whether such correlations alter the original evolutionary path. Path fine-tuning refers to situations where the newly added driving elements have a minor impact on the existing evolutionary path, requiring only adjustments to the demand characteristics or transition rules of some nodes without changing the overall evolutionary direction. For example, if the aforementioned "ventricular premature beats" element has a low frequency (5 times / minute) and short duration, it may only lead to increased attention to heart rate monitoring, which falls under path fine-tuning, specifically characterized by "increased heart rate variability," with the associated node being the current key evolutionary node for "respiratory function assessment." A change in direction refers to a fundamental alteration in the evolutionary path caused by the addition of new driving factors. This necessitates the addition of key evolutionary nodes or a change in the order of existing nodes. For example, if a patient experiences a severe arrhythmia (ventricular tachycardia), the original evolutionary path, which was primarily focused on respiratory function assessment, will shift to one primarily focused on arrhythmia management. This is a change in direction, specifically characterized by a "severe arrhythmia event," with the associated node being a new key evolutionary node for "emergency management of arrhythmias."

[0136] Step S155: Adjust the position of key evolution nodes in the dynamic demand model based on the impact results of the newly added driving factors. If the newly added driving factors cause path fine-tuning, then modify the demand representation content of the corresponding key evolution nodes; if they cause a change in direction, then add key evolution nodes and define the demand representation of the newly added key evolution nodes.

[0137] For minor adjustments to the pathway, identify the key evolutionary nodes affected by the newly added driving factors and revise their requirement representation content. For example, in the key evolutionary node of "respiratory function assessment," due to the appearance of the "ventricular premature beats" element, add the "heart rate variability monitoring information" type to the requirement representation content, and adjust the application direction to "simultaneously assess respiratory function and heart rate stability." For changes in direction, add key evolutionary nodes to the dynamic requirement model, determine their occurrence time, and define their requirement representation content. For example, add the key evolutionary node of "emergency treatment of arrhythmias," with a time point of T5+7 minutes, and the requirement representation content includes "arrhythmia type identification information," "guidance information on the use of antiarrhythmic drugs," and "information on the electrical cardioversion procedure," etc.

[0138] Step S156: Update the demand transition rule set in the dynamic demand model, incorporate the demand change logic caused by the newly added driving factors into the demand transition rule set, and supplement the new demand transition conditions and representation forms.

[0139] For path fine-tuning, update the demand transition rules between existing key evolution nodes. For example, after the "Respiratory Function Assessment" node, if a new driving factor of "ventricular premature beat frequency > 10 beats / minute" appears, the demand transition rule is adjusted from "Respiratory Function Assessment → Ventilator Parameter Optimization" to "Respiratory Function Assessment → Heart Rate Abnormality Management". For directional changes, add new demand transition rules, such as "Emergency Management of Arrhythmia → Respiratory Function Assessment after Heart Rate Stabilization", with the transition condition being "Arrhythmia is controlled, heart rate recovers to 60-100 beats / minute and remains stable for 5 minutes". This is represented by the demand transitioning from "Emergency Management of Arrhythmia" to "Assessing Respiratory Function". Add the above new or updated transition rules to the demand transition rule set.

[0140] Step S157: Adjust the predictor parameters in the dynamic demand model, and update the numerical representation of the predictor by combining the evolution trajectory characteristics and influence of the newly added driving factors. The numerical representation of the predictor corresponds to the demand change trend after including the newly added driving factors.

[0141] Based on the evolutionary trajectory characteristics (such as duration and magnitude of change) and impact (such as the degree of influence on the patient's condition) of the newly added driving factors, the parameters of the predictor are adjusted. For example, for the newly added driving factor "ventricular premature beats," if its duration is short and its impact is low, the probability of the subsequent transition from the "respiratory function assessment" node to the "heart rate abnormality management" requirement in the predictor is increased from 0.1 to 0.3; if the newly added driving factor "ventricular tachycardia," which has a high impact, appears, the predictor is significantly adjusted so that the predicted probability of the "emergency management of arrhythmia" requirement reaches 0.95. The updated predictor numerical representation can more accurately reflect the trend of demand changes after including the newly added driving factors.

[0142] Step S158: Based on the adjusted dynamic demand model, re-retrieve the reconstructed scenario adaptation knowledge module, extract the knowledge content that matches the adjusted key evolution node demand representation, and replace the corresponding outdated knowledge units in the initial context-aware recommendation information.

[0143] Based on the key evolution nodes in the adjusted dynamic demand model (including existing nodes and newly added nodes that revise the demand representation), matching knowledge content is re-retrieved in the scenario adaptation knowledge module. For example, for the "Respiratory Function Assessment" node that revises the demand representation, basic knowledge units containing "Heart Rate Variability Monitoring" are retrieved to replace the original knowledge units in the initial recommendation information that only focused on respiratory function indicators. For the newly added "Emergency Treatment of Arrhythmias" node, basic knowledge units such as "Management Procedures for Ventricular Tachycardia" are retrieved and added to the recommendation information, replacing outdated knowledge units in the initial recommendation information that are inconsistent with the current demand (such as the originally planned "Ventilator Parameter Optimization Suggestions").

[0144] Step S159: Supplement the knowledge content corresponding to the newly added requirements in the adjusted dynamic requirement model, extract knowledge units that can meet the requirements of the newly added key evolution nodes from the guardian knowledge reserve, and integrate them into the initial context-aware recommendation information.

[0145] For newly added key evolutionary nodes in the dynamic demand model, corresponding basic knowledge units and transitional knowledge fragments are extracted from the monitoring knowledge reserve and added to the initial context-aware recommendation information. For example, for the "emergency treatment of arrhythmia" node, basic knowledge units such as "identification criteria for ventricular tachycardia", "dosage and administration of amiodarone", and "electrical cardioversion procedure" are extracted, as well as transitional knowledge fragments connecting the "emergency treatment of arrhythmia" node and subsequent nodes, such as "necessity of hemodynamic assessment after arrhythmia treatment". The above knowledge content is integrated into the recommendation information to ensure that the new demand is met.

[0146] Step S1510: Reorder the knowledge content in the initial context-aware recommendation information according to the application sequence of the adjusted dynamic demand model, and generate the optimized final recommendation information set.

[0147] Based on the temporal order and application sequence of key evolutionary nodes in the adjusted dynamic demand model, the knowledge content in the initial context-aware recommendation information is reordered. Newly added knowledge units are placed at their corresponding new key evolutionary nodes, and the order of existing knowledge units is adjusted to adapt to the new transition rules. For example, knowledge units related to "emergency treatment of arrhythmia" are placed before knowledge units related to "respiratory function assessment" because the change in direction alters the evolutionary path order. Simultaneously, based on the new application sequence annotations, the activation time and presentation order of each knowledge content are adjusted to ensure that the temporal order of the recommendation information aligns with the actual progress of the scenario evolution, ultimately generating an optimized final recommendation information set.

[0148] For example, step S1511: parse the adjusted dynamic demand model and extract the update information of key evolution nodes, including newly added key evolution nodes, deleted key evolution nodes and key evolution nodes whose demand representations have been adjusted.

[0149] Analyze the adjusted dynamic requirements model to identify changes in key evolutionary nodes. For example, it may be found that a new key evolutionary node, "Emergency Management of Cardiac Arrhythmia," has been added; the original key evolutionary node, "Ventilator Parameter Optimization," has been deleted due to changes in the scenario; and the requirements representation content of the key evolutionary node, "Respiratory Function Assessment," has been adjusted. Record all the above updates in detail, including the time, type, and requirements representation of the newly added nodes; the identifier of the deleted nodes; and the original and new requirements representation content of the nodes whose requirements representation has been adjusted.

[0150] Step S1512: For the newly added key evolution nodes, extract matching new knowledge units from the reconstructed scene adaptation knowledge module. The new knowledge units meet the requirements of the new key evolution nodes.

[0151] For newly identified key evolutionary nodes, such as "emergency treatment of arrhythmias", matching new knowledge units are retrieved and extracted from the reconstructed scenario adaptation knowledge module based on their demand representation content (such as arrhythmia type identification, drug use guidance, operation procedures, etc.). The above knowledge units should be able to fully cover the information demand type and application scenario requirements of the new nodes, such as knowledge units like "Guidelines for Emergency Treatment of Ventricular Tachycardia" and "Specifications for Intravenous Injection of Amiodarone".

[0152] Step S1513: For the corresponding deleted key evolution nodes, remove the basic knowledge units and related transitional knowledge fragments corresponding to the deleted key evolution nodes from the initial context-aware recommendation information.

[0153] For key evolutionary nodes that have been deleted, such as "ventilator parameter optimization", find the corresponding basic knowledge units (such as "ventilator parameter optimization recommendations") and transitional knowledge fragments connecting this node to other nodes (such as "transition instructions from respiratory function assessment to ventilator parameter optimization") in the initial context-aware recommendation information, and completely remove the above knowledge content from the recommendation information to avoid irrelevant information interfering with the monitoring personnel.

[0154] Step S1514: For the key evolution nodes where the corresponding demand representation is adjusted, replace the original knowledge units in the initial context-aware recommendation information with updated knowledge units that match the adjusted demand representation.

[0155] For key evolutionary nodes where the demand representation has been adjusted, such as "respiratory function assessment," the demand representation may now include "heart rate variability monitoring." Therefore, it is necessary to find the original "respiratory function assessment indicators and methods" knowledge unit in the initial recommendation information and replace it with an updated knowledge unit that includes "heart rate variability monitoring indicators," such as "joint assessment indicators and methods for respiratory function and heart rate variability."

[0156] Step S1515: Adjust the order of knowledge units in the initial context-aware recommendation information, and re-plan the presentation order of all knowledge units based on the application sequence of the adjusted dynamic demand model.

[0157] Based on the chronological order and application sequence of key evolution nodes in the adjusted dynamic demand model, all knowledge units in the recommended information are reordered. For example, if a new "emergency treatment of arrhythmia" node occurs before the "respiratory function assessment" node, its corresponding knowledge unit is arranged before the "respiratory function assessment" knowledge unit. If the order of some nodes changes due to changes in demand transition rules, the order of knowledge units is also adjusted accordingly to ensure that the arrangement of knowledge units is consistent with the chronological and logical order of scenario evolution.

[0158] Step S1516: Supplement the application instructions corresponding to the newly added knowledge unit. The application instructions include specific details on usage methods, operation coordination, and effect correlation.

[0159] Supplement the newly added knowledge units with application instructions. For example, for the knowledge unit "Amiodarone Intravenous Injection Procedure Guidelines", the application instructions include: usage method (e.g., "Add 150 mg of amiodarone to 20 ml of 5% glucose injection and slowly inject intravenously over a period of not less than 10 minutes"), operation coordination (e.g., "Closely monitor the electrocardiogram during the injection process and observe changes in heart rate and rhythm"), and effect correlation (e.g., "It is expected that arrhythmias will be controlled within 10-15 minutes after the injection; if ineffective, electrical cardioversion should be considered"). The above application instructions ensure that monitoring personnel can correctly use the newly added knowledge units.

[0160] Step S1517: Optimize the transition descriptions between knowledge units. Adjust the transition descriptions for the connection points after replacing or adding knowledge units.

[0161] When knowledge units are replaced or added, the transitional statements between them may need to be adjusted. For example, between the knowledge unit "Emergency Management of Cardiac Arrhythmias" and the knowledge unit "Respiratory Function Assessment," a new transitional statement needs to be added: "[Transition] After the arrhythmia is controlled, respiratory function needs to be further assessed to determine whether the underlying cause of hypoxia is related to the arrhythmia." Ensure that the transitional statements are natural and logically coherent, helping caregivers smoothly transition from one knowledge unit to the next.

[0162] Step S1518: Integrate the updated knowledge units, application instructions, and transitional statements to generate optimized initial recommendation information. The structure of the initial recommendation information is consistent with the initial context-aware recommendation information, and the content is fully adapted to the adjusted dynamic demand model.

[0163] The updated knowledge units (including new, replaced, and retained ones), supplementary application instructions, and optimized transitional expressions are integrated together. The overall structure of the initial context-aware recommendation information (such as title, hierarchy, and presentation format) remains unchanged, but the content is completely replaced to adapt to the adjusted dynamic requirements model. For example, the title of the recommendation information may still be "Patient Monitoring Context-Aware Recommendation Information," but the internal knowledge units, application instructions, and transitional expressions have all been updated to reflect the current scenario requirements.

[0164] Step S1519: Adjust the level of detail in the initial recommendation information according to the real-time rhythm of scenario evolution. The knowledge units corresponding to key evolution nodes contain complete descriptions, while the knowledge units corresponding to transition nodes retain specific descriptions. Finally, integrate them to form an optimized set of context-aware recommendation information.

[0165] Based on the real-time pace of scenario evolution, the level of detail in each knowledge unit within the recommended information is adjusted. For currently occurring critical evolutionary nodes, the corresponding knowledge units present complete descriptions, including all operational steps, application instructions, and precautions. For transitional nodes or past critical evolutionary nodes, the knowledge units retain detailed descriptions but can be appropriately simplified to highlight core content. For example, if the current scenario is at the critical evolutionary node of "emergency treatment of arrhythmia," the knowledge unit for this node will display detailed treatment steps and medication guidance; while the previous knowledge unit of "treatment of hypoxia after suctioning" will be simplified to a list of core steps. Through these adjustments, the recommended information becomes more focused on current needs while retaining necessary historical information, ultimately integrating to form an optimized set of context-aware recommended information, which is then pushed to the monitoring execution terminal.

[0166] In one exemplary embodiment, a context-aware information recommendation system for patient monitoring is provided. This system can be a terminal, a server, etc., and its internal structure diagram can be as follows: Figure 2As shown, the context-aware information recommendation system for patient monitoring includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements a context-aware information recommendation method for patient monitoring. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or a button, trackball, or touchpad set on the shell of the context-aware information recommendation system in patient monitoring, or an external keyboard, touchpad, or mouse, etc.

[0167] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A context-aware information recommendation method in patient monitoring, characterized in that, The method includes: Based on real-time status data in the monitoring scenario, perceive the dynamic evolution characteristics of the patient monitoring scenario; A dynamic demand model for patient monitoring is constructed based on the aforementioned dynamic evolution characteristics. The dynamic demand model includes the association rules and changing trends of demand as the scenario evolves. The scenario adaptation knowledge module in the guardianship knowledge reserve is reconstructed based on the dynamic demand model constructed according to the dynamic evolution characteristics, and the scenario adaptation knowledge module is restructured according to the demand change trend. Initial context-aware recommendation information is generated based on the reconstructed scene adaptation knowledge module; The system tracks the real-time changes of the dynamic evolution features, adjusts the parameters of the dynamic demand model, optimizes the initial context-aware recommendation information based on the adjusted dynamic demand model, generates the final recommendation information set, and pushes it to the monitoring execution terminal.

2. The context-aware information recommendation method for patient monitoring according to claim 1, characterized in that, The construction of a dynamic demand model for patient monitoring based on the aforementioned dynamic evolutionary characteristics includes: The scene evolution driving elements are extracted from the dynamic evolution features. The scene evolution driving elements include the advancement nodes of the monitoring operation, the representation of changes in the patient's physiological state, and the types of changes in environmental factors. The driving factors of scene evolution are classified into different types and dimensions, and classified according to the source attributes of the driving factors, distinguishing between monitoring operation driving factors, physiological state driving factors, and environmental change driving factors. The evolution trajectory characteristics of scene evolution driving elements under each type dimension are analyzed, the change path of elements over time is extracted, the specific state data of elements at different time nodes are recorded, and independent evolution path records of each type of element are generated. The independent evolution path records of each type of element present all the change details of the element from the initial state to the current state. By integrating the evolution paths of scene evolution-driving elements under different types and dimensions, locating the interaction nodes between elements, capturing the associated time points where changes in elements of one type and dimension trigger changes in elements of other types and dimensions, and constructing a multi-dimensional element association evolution record, the association logic and triggering order between each element are marked in the multi-dimensional element association evolution record. Key evolution nodes are extracted from the multi-dimensional element association evolution record. These key evolution nodes are the core time points that trigger changes in demand direction. The location of key evolution nodes is determined by analyzing the element association strength and the degree of demand impact. Each key evolution node corresponds to a specific scenario state and element combination. Generate demand representation content corresponding to each key evolution node. The generation of demand representation content is based on the specific state of each type of driving element at the key evolution node. The demand representation content includes the information types and application directions required for patient monitoring under the key evolution node, and generate a node demand representation set. Analyze the demand representation change trajectory between adjacent key evolution nodes, extract the transition logic of demand from one key evolution node to another, summarize the triggering conditions and representation forms of demand changes, and generate a set of demand transition rules. Construct a demand evolution correlation matrix, which integrates the demand representation content and demand transition rule set of key evolution nodes into a matrix. The row dimension of the matrix represents the key evolution nodes, and the column dimension represents the demand representation and transition rules of the corresponding key evolution nodes. A predictor of demand change trends is generated. The predictor is obtained based on the trajectory characteristics of demand changes in historical scenario evolution data. The predictor is associated with the demand representation of key evolution nodes and the demand transition direction of subsequent key evolution nodes. By integrating the demand representation set, demand transition rule set, demand evolution correlation matrix and predictive factors of key evolution nodes, and combining them in a structured manner according to the time sequence and logical relationship of scenario evolution, a dynamic demand model is generated.

3. The context-aware information recommendation method for patient monitoring according to claim 1, characterized in that, The scenario adaptation knowledge module in the guardianship knowledge reserve base, which is reconstructed based on the dynamic demand model constructed according to the dynamic evolution characteristics, includes: The key evolution node requirement representation content in the dynamic requirement model is analyzed, and the information requirement type and application scenario requirements corresponding to each key evolution node are extracted. Extract basic knowledge units from the monitoring knowledge reserve that match the requirements of each key evolution node. The basic knowledge unit is the smallest knowledge fragment that can directly meet the requirements of the corresponding key evolution node. The basic knowledge units are extracted one by one in the order of the key evolution nodes. Each basic knowledge unit fully covers the key requirements of the corresponding key evolution node. Verify the matching relationship between basic knowledge units and the corresponding key evolution nodes' requirement representations. When the content of the basic knowledge unit covers the information requirement type of the key evolution node, record it as a node-knowledge unit matching pair. Connect the basic knowledge units corresponding to adjacent key evolution nodes, and determine the connection logic of the transition from the basic knowledge unit of the previous key evolution node to the basic knowledge unit of the next key evolution node based on the demand transition rule set in the dynamic demand model. Construct a knowledge unit transition link, which includes the connection order and association method of knowledge units. To address the information gaps between adjacent basic knowledge units, transitional knowledge fragments that can fill the gaps are extracted from the monitoring knowledge reserve to supplement the connecting knowledge content in the transition links between adjacent basic knowledge units. Based on the chronological order of key evolution nodes and the connection logic of knowledge unit transition links, basic knowledge units and transitional knowledge fragments are reorganized to generate a preliminary scenario-adaptive knowledge module structure. Divide the knowledge module structure into levels for scene adaptation, assign level identifiers to key evolution nodes according to preset rules, place the basic knowledge units corresponding to key evolution nodes with basic level identifiers in the basic level, and place the knowledge content of the connecting level in the connecting level. Mark the application sequence of knowledge content in each level to determine the activation time and application duration of knowledge content in the process of scenario evolution; The knowledge content and application time sequence annotations in the integration hierarchy are used to generate a structured scenario-adaptive knowledge module. The scenario-adaptive knowledge module includes basic level knowledge, connecting level knowledge and corresponding application time planning.

4. The context-aware information recommendation method in patient monitoring according to claim 1, characterized in that, The process of tracking the real-time changes of the dynamic evolution features, adjusting the parameters of the dynamic demand model, and optimizing the initial context-aware recommendation information based on the adjusted dynamic demand model includes: The system acquires dynamic evolution characteristics of the monitored scenario at preset time intervals through a real-time acquisition channel. Newly added driving elements are extracted from the dynamic evolution features collected in real time. These newly added driving elements are newly emerging elements that were not previously included in the dynamic evolution features. The original representation and emergence time of the newly added driving elements are retained. By comparing the newly added driving elements with the existing scenario evolution driving elements in the dynamic demand model, the type dimension to which the newly added driving elements belong is determined, and the newly added driving elements are classified into the corresponding monitoring operation category, physiological state category, or environmental change category dimension. The newly added driving elements are compared with the existing multi-dimensional element evolution records. Based on the comparison results, the impact category of the newly added driving elements on the existing evolution path is determined. The impact category includes path fine-tuning and direction change. The specific representation and associated nodes corresponding to the impact category are recorded. Adjust the positions of key evolution nodes in the dynamic demand model based on the impact of the newly added driving factors. If the newly added driving factors cause path fine-tuning, then revise the demand representation content of the corresponding key evolution nodes; if they cause a change in direction, then add key evolution nodes and define the demand representation of the new key evolution nodes. Update the demand transition rule set in the dynamic demand model, incorporate the demand change logic caused by the new driving factors into the demand transition rule set, and supplement the new demand transition conditions and representation forms. Adjust the predictor parameters in the dynamic demand model, combine the evolution trajectory characteristics and influence of the newly added driving factors, update the numerical representation of the predictor, and the numerical representation of the predictor corresponds to the demand change trend after including the newly added driving factors. Based on the adjusted dynamic demand model, the reconstructed scenario-adaptive knowledge module is retrieved again, and knowledge content matching the adjusted key evolution node demand representation is extracted to replace the corresponding outdated knowledge units in the initial context-aware recommendation information. The knowledge content corresponding to the newly added requirements in the dynamic requirements model after the supplementation and adjustment is extracted from the guardianship knowledge reserve to meet the requirements of the newly added key evolution nodes and integrated into the initial context-aware recommendation information; Based on the application sequence of the adjusted dynamic demand model, the knowledge content in the initial context-aware recommendation information is reordered to generate the optimized final recommendation information set.

5. The context-aware information recommendation method for patient monitoring according to claim 1, characterized in that, The process of perceiving the dynamic evolution characteristics of the patient monitoring scenario based on real-time status data in the monitoring scenario includes: Deploy scene-aware terminals to collect real-time status data in the monitoring scene. The real-time status data includes execution data of monitoring operations, physiological monitoring data of patients, and environmental parameter data. Different types of data in real-time status data are separated, and monitoring operation data, physiological monitoring data and environmental parameter data are distinguished. Each type of data is arranged in the order of collection time to generate an independent time-series data sequence. Extract the operation execution node information from the monitoring operation data, record the start time, key actions and completion time of each monitoring operation, and generate a monitoring operation time sequence node record. The monitoring operation time sequence node record marks the specific content and time coordinate of each operation node. Analyze the changing characteristics in physiological monitoring data, compare physiological monitoring data at different time points, identify the changing patterns of data values, record the process data of physiological state transitioning from one stable state to another, and generate a time-series record of physiological state changes. The time-series record of physiological state changes fully presents the process of physiological state change. Capture changes in environmental parameter data, record the specific values ​​of environmental parameters at different time points, and generate a time-series record of environmental parameter changes, wherein the time-series record contains the change data of environmental parameters and the corresponding time points; Using time as a unified dimension, the monitoring operation time sequence records, physiological state change time sequence records, and environmental parameter change time sequence records are aligned and integrated to generate a comprehensive scene time sequence evolution record. Extract change events from the scene comprehensive temporal evolution record to generate a scene change event set. The change event is a specific data change that causes a change in the scene state. Each change event includes the event occurrence time, event type and specific event representation. Based on a pre-defined causal rule base, correlation analysis is performed on each event in the set of scene change events to generate causal chains between events and generate scene event causal correlation records. The scene event causal correlation records contain causal logic identifiers and correlation strength values ​​between events. Integrate the set of scene change events and the causal relationship records of scene events, and combine them in a structured way according to time sequence and causal logic to present the evolution path of the scene from the initial state to the current state; Extract the change features in the evolution path, summarize the direction of change and the representation form in the scene evolution process, and generate dynamic evolution features that represent the dynamic changes of the scene.

6. The context-aware information recommendation method in patient monitoring according to claim 1, characterized in that, The initial context-aware recommendation information generated based on the reconstructed scene adaptation knowledge module includes: The hierarchical structure of the scene-adaptive knowledge module is analyzed, and the specific content of the basic level knowledge and the connecting level knowledge, as well as their related logic and application order, are extracted to generate a knowledge module structure analysis record. Information units are extracted from the basic-level knowledge in the order of key evolutionary nodes. The information units are specific contents that directly meet the needs of the corresponding key evolutionary nodes. Transitional information units are extracted from the knowledge at the connection level according to the sequence of the knowledge unit transition link, and the transitional information units are combined with information units according to a preset connection logic; The predefined presentation format of the information acquisition unit and the transition information unit is set according to the application scenario requirements of the corresponding key evolution nodes and conforms to the information transmission standards in the field of monitoring, including the information display structure, content organization method and transmission format. Mark the information unit with the corresponding scene evolution driving element, and mark the state of the driving element and the application triggering conditions that it is adapted to in the information unit. The presentation order of information units and transitional information units is arranged according to the time sequence of scenario evolution and the application sequence of knowledge modules, as well as the application instructions of supplementary information units. The application instructions include the usage method of information units in actual monitoring operations, the corresponding monitoring actions, and the expected effects after application. Integrate information units, transitional information units, application instructions, and presentation formats to generate an information recommendation structure. Following the order of the information recommendation structure, convert all information units and application instructions into standardized textual expressions to generate initial context-aware recommendation information.

7. The context-aware information recommendation method in patient monitoring according to claim 1, characterized in that, The step of tracking the real-time changes in the dynamic evolutionary characteristics and adjusting the parameters of the dynamic demand model includes: Extract the parameter types from the dynamic demand model, including the weight parameters of the demand evolution correlation matrix, the triggering parameters of the demand transition rule set, and the characterization parameters of the predictor. Analyze the impact of the real-time change trajectory of the dynamic evolution characteristics on each parameter type, identify the changes in the real-time change trajectory that cause the parameter values ​​to change, and sort out the correspondence between the changes and the parameter changes. Based on the correspondence between the changed content and the parameter changes, determine the parameter adjustment direction and adjustment range logic triggered by different types of changes, and generate the association rules for parameter adjustment; Record the specific parameter values ​​in the parameter type, and record the initial value of each parameter in the current dynamic requirement model; By comparing the changes in the real-time trajectory with the correlation rules of parameter adjustments, the type and specific parameters that need to be adjusted are determined. Based on the correlation rules of parameter adjustment, combined with the quantitative intensity of the changed content and the predefined sensitivity coefficient of the parameter, the adjustment range of the parameter to be adjusted is calculated; The initial parameter values ​​are adjusted according to the calculated adjustment range, and the adjusted parameter values ​​replace the original parameter values. At the same time, the time, triggering conditions and adjustment range of the parameter adjustment are recorded to generate a parameter adjustment log. Integrate all the adjusted parameters and reconstruct the parameter set of the dynamic demand model.

8. The context-aware information recommendation method for patient monitoring according to claim 2, characterized in that, The evolutionary path that integrates scene evolution driving elements under different types and dimensions includes: Key time nodes are identified from the scene evolution-driven element evolution path under each type dimension. The key time nodes are the specific time points of element change, and each key time node corresponds to a determined element state. Using time as a unified benchmark, the key time nodes are aligned to a unified time axis to generate a multi-dimensional key node set. The state combination data of different types of dimensional elements at the same key time node are obtained, and the interaction relationship of each element at the key time node is determined according to the preset relationship rules. Mark the combination of elements that have an interaction relationship at the same key time node, and generate node element association pairs. Each node element association pair contains elements of different types and dimensions and their corresponding interaction representations. Track the changes in the node element association pairs between adjacent key time nodes, record the changes in the state of the elements in the association pairs and the adjustment of the interaction relationship, and generate the association pair evolution sequence. Identify common change patterns from the evolutionary sequences of associated pairs, and generate general rules for the cross-type element association evolution based on the common change patterns; Identify variation cases that do not conform to the general rules from the evolutionary sequence of association pairs, record the triggering conditions and representation forms of the variation cases, and generate special association evolution supplementary rules; A structured presentation format for recording the evolution of multi-dimensional element associations is constructed. The structured presentation format includes a time axis, a type dimension axis, and an association strength axis. The three axes are perpendicular to each other to form a three-dimensional recording format. The node element association pairs, association pair evolution sequences, and special association evolution rule sets are incorporated into a structured presentation format, arranged in chronological order and by type dimension, and the association strength of each association pair is marked to generate a multi-dimensional element association evolution record.

9. The context-aware information recommendation method in patient monitoring according to claim 3, characterized in that, To address information gaps between adjacent basic knowledge units, transitional knowledge fragments capable of filling these gaps are extracted from the monitoring knowledge reserve to supplement the connecting knowledge content in the transition links between adjacent basic knowledge units. This includes: Based on the content association logic of adjacent basic knowledge units, the classification of information gaps is determined. The classification includes conceptual connection gaps, logical deduction gaps, and application guidance gaps. To address the gaps in conceptual connections, we extract transitional conceptual explanations from the guardianship knowledge reserve that can connect adjacent basic knowledge units. To address the gap in logical deduction, reasoning content that conforms to the professional logic of the guardianship field is extracted from the guardianship knowledge reserve. This reasoning content is used to support the logical transition from one knowledge unit to another. To address the gaps in application guidance, operational connection content that can guide the smooth transition between adjacent basic knowledge units in practical applications is extracted. The operational connection content includes the coordination method and sequence of application operations corresponding to two knowledge units. The extracted supplementary content is subjected to consistency verification to ensure that the expression attributes of the supplementary content are consistent with the expression attributes of the basic knowledge unit. The supplementary content includes explanations of transitional concepts, reasoning basis, and operational connection content. The supplementary content can be trimmed or expanded to match the coverage of the information gaps. To supplement the content, a position index is generated in the knowledge unit transition link. The position index is generated based on the information gap type corresponding to the supplementary content and the positional relationship between adjacent basic knowledge units. It is used to insert the supplementary content between two basic knowledge units without changing the internal structure of the original knowledge unit. To supplement the content, associate the corresponding information gap type identifier and set the type identifier as the metadata field of the supplemented content; Arrange all basic knowledge units and their corresponding supplementary content according to the location index and order to generate a complete knowledge unit transition link; The knowledge unit transition links are checked to confirm that all information gaps between adjacent basic knowledge units have been filled with supplementary content, and that all basic knowledge units are sequentially connected through the supplementary content.

10. A context-aware information recommendation system for patient monitoring, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the context-aware information recommendation method in patient monitoring according to any one of claims 1 to 9 by executing the machine-executable instructions.

Images