An artificial intelligence-based medical data information analysis method and system

Abstract

This application provides a medical data information analysis method and system based on artificial intelligence, including: acquiring verbal communication information from medical staff and recognizing the verbal communication information to obtain instructions or statements in text form; performing semantic analysis on the text instructions or statements to extract operation type, patient identification, time information, and intent to adjust parameters; generating temporary calibration rules for correcting data interpretation logic based on the semantic analysis results, and injecting the temporary calibration rules into the medical data analysis system; when the medical data analysis system processes physiological monitoring data, the temporary calibration rules are preferentially applied to correct the judgment of data anomalies. This application can distinguish between data anomalies caused by normal clinical operations and actual changes in the patient's condition, effectively reducing false alarms and false negatives, and improving the accuracy and reliability of data analysis in intensive care environments.

Description

Technical Field

[0001] This application relates to the field of medical data information analysis, and more specifically, to a method and system for medical data information analysis based on artificial intelligence. Background Technology

[0002] In the intensive care unit, a medical data analysis system has been introduced to enhance real-time monitoring and early warning capabilities for critically ill patients. This system is responsible for continuously collecting and analyzing various physiological parameters of patients, including real-time monitoring data such as electrocardiograms, blood pressure, blood oxygen saturation, and respiratory rate, as well as past information recorded in electronic medical records, such as laboratory test results, medication information, and imaging reports. Through a sophisticated predictive mechanism, the system aims to identify early signs of deterioration in a patient's condition and promptly alert medical staff, thereby providing support for gaining valuable intervention time.

[0003] However, after the system was put into actual operation, medical staff quickly discovered that although the system performed excellently when processing continuous and stable data streams, the accuracy and reliability of its warnings were significantly reduced in certain specific situations. Specifically, when patients' physiological monitoring data experienced intermittent interruptions or drastic fluctuations, the system often generated false alarms or missed alarms. For example, during routine nursing procedures such as turning over, changing dressings, performing airway suction, or bedside ultrasound examinations, it is necessary to temporarily remove some sensors or adjust the position of monitoring equipment. These operations inevitably lead to gaps, abnormal values, or drastic fluctuations in real-time data streams such as electrocardiograms, blood oxygen saturation, or blood pressure for several minutes to tens of minutes. When the system receives these intermittently interrupted data, because its internal judgment mechanism is mainly trained based on continuous and relatively stable data patterns, it cannot effectively distinguish whether these data gaps or abnormalities are caused by normal clinical operations or are a true reflection of a sudden deterioration in the patient's condition. The system usually treats these data interruptions or abnormal fluctuations as potential physiological abnormalities according to its preset logic and immediately triggers high-priority warnings. Summary of the Invention

[0004] This application discloses a medical data information analysis method and system based on artificial intelligence, which aims to solve the problem that existing medical data analysis systems cannot accurately distinguish between data abnormalities caused by normal operations and actual deterioration of the condition when faced with frequent clinical operations and data fluctuations in the intensive care environment, resulting in false alarms or omissions.

[0005] To achieve the above objectives, this application adopts the following technical solution:

[0006] In a first aspect, this application discloses an artificial intelligence-based medical data information analysis method for calibrating the judgment of abnormal physiological monitoring data in an intensive care environment, including:

[0007] Acquire verbal communication information from healthcare workers and identify such information to obtain instructions or statements in text form;

[0008] Semantic analysis is performed on text-based instructions or statements to extract operation type, patient identification, time information, and intent to adjust parameters;

[0009] Based on the semantic analysis results, temporary calibration rules are generated to correct the data interpretation logic, and these temporary calibration rules are injected into the medical data analysis system.

[0010] When processing physiological monitoring data in a medical data analysis system, temporary calibration rules should be applied first to correct any judgments of data anomalies.

[0011] Secondly, this application also discloses an artificial intelligence-based medical data information analysis system for calibrating the judgment of abnormal physiological monitoring data in an intensive care environment, including:

[0012] The information acquisition module is used to acquire verbal communication information from medical staff and to identify the verbal communication information in order to obtain instructions or statements in text form.

[0013] The semantic analysis module is used to perform semantic analysis on text-based instructions or statements to extract operation type, patient identification, time information, and intent to adjust parameters;

[0014] The rule generation and injection module is used to generate temporary calibration rules for correcting data interpretation logic based on semantic analysis results, and inject the temporary calibration rules into the medical data analysis system.

[0015] The data processing module is used to prioritize the application of temporary calibration rules when processing physiological monitoring data in the medical data analysis system, in order to correct the judgment of data anomalies.

[0016] Compared with the prior art, this application has at least the following beneficial effects:

[0017] This application captures verbal instructions from healthcare professionals in real time, transforming unstructured operational context information into structured calibration rules. This allows the system to understand the potential impact of ongoing clinical procedures on physiological data. For example, when a healthcare professional verbally instructs the patient to turn over, potentially interrupting the electrocardiogram (ECG), the system can generate corresponding temporary calibration rules to suppress or adjust abnormal ECG data during the turn, thus preventing normal physiological data fluctuations from being misinterpreted as deterioration of the patient's condition. This mechanism enables the system to interpret data more intelligently, significantly reducing false alarms caused by data interruptions or abnormal fluctuations due to clinical procedures, and improving the accuracy and reliability of early warnings. Simultaneously, by prioritizing the application of temporary calibration rules, the system can more accurately identify genuine deterioration of the patient's condition, avoiding the risk of missed reports and providing clinicians with more accurate and data-driven diagnostic and treatment references, thereby improving patient safety and the quality of care in intensive care settings. Attached Figure Description

[0018] Figure 1 A flowchart illustrating an artificial intelligence-based medical data information analysis method provided in this application;

[0019] Figure 2 This application provides a schematic diagram of the structure of an artificial intelligence-based medical data information analysis system. Detailed Implementation

[0020] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0021] like Figure 1 As shown, this application proposes an artificial intelligence-based medical data information analysis method for calibrating the judgment of abnormal physiological monitoring data in an intensive care environment, including:

[0022] Acquire verbal communication information from healthcare workers and identify such information to obtain instructions or statements in text form;

[0023] Semantic analysis is performed on text-based instructions or statements to extract operation type, patient identification, time information, and intent to adjust parameters;

[0024] Based on the semantic analysis results, temporary calibration rules are generated to correct the data interpretation logic, and these temporary calibration rules are injected into the medical data analysis system.

[0025] When processing physiological monitoring data in a medical data analysis system, temporary calibration rules should be applied first to correct any judgments of data anomalies.

[0026] Verbal communication among medical staff refers to verbal communication between medical staff or between medical staff and patients in the intensive care environment. This communication may include descriptions of the patient's condition, instructions for upcoming procedures, and statements of parameter adjustments.

[0027] The judgment of abnormal physiological monitoring data refers to the medical data analysis system analyzing the patient's physiological parameters (such as heart rate, blood pressure, blood oxygen saturation, etc.) according to preset thresholds and patterns to identify whether there are values ​​or trends that exceed the normal range or indicate changes in the condition.

[0028] Temporary calibration rules refer to a set of corrective logic dynamically generated based on real-time verbal communication information from healthcare professionals. These rules are used to adjust the medical data analysis system's criteria for judging abnormalities in specific physiological parameters within a specific time period. These rules are temporary and designed to address specific clinical procedures or situations, rather than permanently altering the system's fundamental judgment logic.

[0029] A medical data analysis system is a software or hardware system that integrates artificial intelligence algorithms. It is responsible for collecting, processing, and analyzing various medical data of patients and providing functions such as diagnostic assistance and early warning.

[0030] The following is a further detailed description of the artificial intelligence-based medical data information analysis method provided in the embodiments of this application:

[0031] Various technologies can be employed to acquire and recognize verbal communication information from healthcare workers to extract textual instructions or statements. For example, a high-sensitivity microphone array can be deployed to capture real-time speech information from the intensive care unit. This speech information is then transmitted to a speech recognition module, which can convert continuous speech signals into text based on deep learning models such as Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, or Transformer models. To improve recognition accuracy, the speech recognition model can be pre-trained on a large amount of medical scenario dialogue data and optimized for medical terminology and specific accents. Alternatively, healthcare workers can directly record their verbal instructions using smart voice assistant devices and convert these instructions into text in real time.

[0032] Natural Language Processing (NLP) techniques can be used to perform semantic analysis on textual instructions or statements to extract operation types, patient identifiers, time information, and intent to adjust parameters. For example, a semantic analysis engine combining rule-based and machine learning technologies can be built. This engine first uses Named Entity Recognition (NER) to identify patient identifiers such as patient name and bed number, operation types such as turning over, airway suction, and adjusting infusion pumps, and time information such as now, five minutes later, and for half an hour. Then, an intent recognition model (such as one based on Support Vector Machines (SVM) or a deep learning classifier) ​​is used to determine whether the verbal communication from healthcare workers is an instruction (e.g., adjusting the lower limit of the blood oxygen alarm to 88%) or a statement (e.g., the patient is undergoing a bedside ultrasound examination), and the specific intent to adjust parameters (e.g., adjusting the lower limit of the blood oxygen alarm to 88%) is extracted from it.

[0033] To generate temporary calibration rules based on semantic analysis results to correct data interpretation logic, and to inject these temporary calibration rules into the medical data analysis system, a rule generator can be designed. For example, when semantic analysis indicates that a patient identified as XXX is undergoing a turning operation expected to last 10 minutes, the rule generator can generate a temporary calibration rule stipulating that, in the next 10 minutes, the abnormal judgment thresholds for physiological parameters such as blood pressure, heart rate, and blood oxygen saturation of patient XXX can be appropriately relaxed, or that high-priority warnings will not be immediately triggered in the event of a brief data interruption. This rule is then encapsulated into a specific data structure and injected into the rule engine of the medical data analysis system in real time through an API interface or message queue mechanism.

[0034] When processing physiological monitoring data, medical data analysis systems prioritize the application of temporary calibration rules to correct for anomalies. The system's internal rule engine is designed with priority management capabilities. For example, when the system receives real-time physiological monitoring data from a patient, it first checks if any temporary calibration rules are currently active for that patient. If so, the system will prioritize applying these temporary calibration rules to interpret the current data. For instance, if a temporary calibration rule instructs the suppression of low blood oxygen saturation alarms within a specific time period, then even if the real-time blood oxygen value is below the regular alarm threshold, the system will not trigger an alarm as long as it is within the range allowed by the temporary rule. Only when data anomalies exceed the correction range set by the temporary calibration rule will the system process the data according to its basic judgment logic.

[0035] This application acquires, identifies, and semantically analyzes verbal communication information from healthcare professionals, transforming this unstructured, real-time clinical context into temporary calibration rules that the system can understand and apply. These rules dynamically correct the medical data analysis system's judgment logic regarding abnormal physiological monitoring data, enabling it to adaptively adjust according to the actual clinical situation. For example, when healthcare professionals verbally declare that a patient is turning over and the electrocardiogram (ECG) may be temporarily interrupted, a temporary rule can be generated to suppress abnormal ECG alarms during the turning process. This dynamic calibration mechanism allows the system to more intelligently determine the nature of data interruptions or fluctuations, thus avoiding false alarms caused by the lack of contextual information in traditional systems and improving the accuracy of identifying genuine deterioration of the patient's condition. Therefore, this application significantly improves the practicality and reliability of medical data analysis systems in complex clinical environments, providing clinicians with more accurate and effective decision support.

[0036] In some implementations, the steps of generating temporary calibration rules for correcting data interpretation logic based on semantic analysis results and injecting these temporary calibration rules into the medical data analysis system include:

[0037] Obtain the patient's basic physiological parameters;

[0038] Impact patterns of query operations;

[0039] Calculate the dynamic calibration threshold range based on the patient's baseline physiological parameters and the impact pattern of the operation;

[0040] Determine the effective duration of the temporary calibration rule based on the impact pattern of the operation;

[0041] Based on the semantic analysis results, temporary calibration rules containing dynamic calibration threshold ranges and effective durations are generated and injected into the medical data analysis system.

[0042] Specifically, obtaining the patient's baseline physiological parameters refers to the system acquiring the patient's current physiological indicators, such as heart rate, blood pressure, blood oxygen saturation, body temperature, and respiratory rate, in real time or periodically from electronic medical records, physiological monitoring devices, or other clinical information systems. These parameters provide baseline data for subsequent calibration threshold calculations.

[0043] The impact pattern of a query operation refers to the system's search in a pre-defined knowledge base, based on the operation type extracted through semantic analysis (e.g., drug infusion, body position adjustment, equipment calibration), to determine the type, degree, and duration of the impact of that operation on different physiological parameters. For example, if a certain vasopressor may cause an increase in blood pressure, its impact pattern would include the expected range of blood pressure increase and the duration of its effect.

[0044] The dynamic calibration threshold range is calculated based on the patient's baseline physiological parameters and the impact pattern of the procedure. This means dynamically adjusting the abnormal judgment threshold of physiological monitoring data by combining the patient's current physiological baseline and the expected impact of a specific procedure. For example, if a patient has low baseline blood pressure and is receiving vasopressor therapy, the system will appropriately adjust the lower or upper limit of blood pressure abnormality based on the impact pattern of the vasopressor drug to avoid misjudging normal physiological fluctuations caused by treatment as abnormalities.

[0045] Determining the duration of effectiveness for temporary calibration rules based on the impact pattern of the operation refers to setting the length of time the calibration rule is effective in the medical data analysis system based on the pharmacodynamics, mechanism of action, or clinical practice experience of a specific operation. For example, a calibration rule for a single-use drug injection may only need to be effective for a few hours, while a continuously infused drug may require a longer effective time.

[0046] Therefore, based on the semantic analysis results, a temporary calibration rule containing the dynamic calibration threshold range and the effective duration is generated. This temporary calibration rule is then injected into the medical data analysis system. This means that the dynamic threshold range and effective duration calculated above are encapsulated into a complete temporary calibration rule and integrated into the data interpretation logic of the medical data analysis system, so that it can be preferentially applied when processing subsequent physiological monitoring data.

[0047] This application refines and contextualizes the generation process of temporary calibration rules by introducing the acquisition of patients' basic physiological parameters, the querying of operation influence patterns, the calculation of dynamic calibration threshold ranges, and the determination of the effective duration. Specifically, after the verbal communication information of medical staff is identified and semantically analyzed, the system not only identifies the operation type but also combines the patient's current specific physiological state with the known influence patterns of the operation on physiological parameters. Therefore, the calibration rules are no longer static or preset but can be dynamically adjusted according to individual patient differences and the expected effects of the operation, thus ensuring the accuracy and timeliness of calibration. This dynamic adjustment mechanism effectively avoids misjudging normal physiological fluctuations caused by treatment or nursing operations as abnormalities, while also more accurately identifying true physiological abnormalities and improving the reliability of data analysis.

[0048] In some implementations, the step of generating a temporary calibration rule containing a dynamic calibration threshold range and an effective duration based on the semantic analysis results, and injecting the temporary calibration rule into the medical data analysis system, further includes:

[0049] Query currently active temporary calibration rules for the same patient identifier;

[0050] Conflict detection and overlay analysis are performed on the currently generated temporary calibration rules and the already activated temporary calibration rules.

[0051] When a conflict is detected between a currently generated temporary calibration rule and an already activated temporary calibration rule, the conflict is adjudicated according to a preset priority strategy to generate a correction rule.

[0052] When an additive effect is detected between the currently generated temporary calibration rule and the already activated temporary calibration rule, a composite calibration rule is generated according to the preset additive calculation logic.

[0053] Inject correction rules or composite calibration rules into the medical data analysis system.

[0054] Specifically, querying currently active temporary calibration rules for the same patient identifier means that before generating a new temporary calibration rule, the system accesses a rule management module or database to retrieve all currently effective temporary calibration rules for a specific patient that have not expired or been revoked. These rules may originate from previous verbal instructions or statements from healthcare professionals and have already been incorporated into the medical data analysis system. The purpose is to obtain the set of all existing rules that might interact with the new rule.

[0055] The conflict detection and overlay analysis of the currently generated temporary calibration rules and the activated temporary calibration rules can be understood as evaluating the potential interactions between the new and old rules. Conflict detection aims to identify whether there are contradictions between the new and existing rules in terms of calibration objectives, threshold ranges, or effective times. For example, one rule requires raising the abnormal threshold of a certain physiological parameter, while another rule requires lowering that threshold. Overlay analysis aims to identify whether there are synergistic effects between the new and existing rules, i.e., their combined effect may lead to more significant or complex changes in the interpretation of physiological data than a single rule. For example, both rules may cause the threshold of a certain physiological parameter to adjust in the same direction, but to different degrees.

[0056] In practical applications, when a conflict is detected between a currently generated temporary calibration rule and an already activated temporary calibration rule, the conflict is adjudicated according to a preset priority strategy to generate a corrective rule. The priority strategy can be based on various factors, such as the rule's source (physician instructions take precedence over nurse instructions), the rule's urgency, the rule's generation time (latest rule overrides older rule), or predefined clinical safety guidelines. The adjudication process aims to ensure that the final set of rules is logically consistent and clinically safe. For example, if one rule instructs that the upper limit of abnormal heart rate be increased to 120 beats / minute, while another rule instructs that it be decreased to 100 beats / minute, the system will select one of them or generate a compromise corrective rule based on the priority strategy.

[0057] When an additive effect is detected between the currently generated temporary calibration rule and an already activated temporary calibration rule, a composite calibration rule is generated according to a preset additive calculation logic. The additive calculation logic can be customized based on specific physiological parameters and operation types. For example, for threshold adjustment, it could be a simple weighted average, maximum or minimum value selection, or a more complex combined algorithm based on physiological models or machine learning. Its purpose is to accurately reflect the combined impact of multiple operations on the interpretation of physiological data. For example, if both operations are expected to cause a slight increase in blood pressure, the composite calibration rule will reflect this cumulative increase effect.

[0058] Injecting correction rules or composite calibration rules into the medical data analysis system ensures that the calibration rules applied by the system when processing physiological monitoring data have undergone conflict resolution and superposition integration, thereby avoiding misjudgments caused by contradictions between rules or failure to consider superposition effects.

[0059] The proposed solution obtains a comprehensive rule context by first querying currently activated temporary calibration rules for the same patient identifier when generating new temporary calibration rules. Based on this, conflict detection and overlay analysis are performed between the newly generated rule and the activated rules, enabling the system to identify and quantify potential interactions between rules. This proactive assessment allows for adjudication based on a pre-defined priority strategy when conflicts are detected, ensuring the uniqueness and security of data interpretation logic in complex clinical scenarios. Furthermore, when an overlay effect exists, composite calibration rules are generated through overlay calculation logic, enabling the system to more accurately reflect the combined impact of multiple clinical operations on physiological parameters, avoiding the bias that a single rule might introduce. This mechanism effectively solves the problems of logical contradictions and inaccurate interpretation that may arise when multiple temporary calibration rules coexist, thereby improving the accuracy and reliability of anomaly detection in physiological monitoring data.

[0060] In some implementations, the steps of conflict detection and overlay analysis between the currently generated provisional calibration rule and the already activated provisional calibration rule include:

[0061] Identify calibration rules caused by different operation types in the currently generated temporary calibration rules and the already activated temporary calibration rules;

[0062] Query the complex context knowledge base to obtain the interaction patterns of different operation types on multiple physiological parameters;

[0063] Based on the interaction mode, the currently generated temporary calibration rules and the activated temporary calibration rules are superimposed for calculation or conflict resolution to generate calibration rules that reflect complex situations.

[0064] Inject calibration rules that reflect complex scenarios into the medical data analysis system.

[0065] Specifically, identifying calibration rules triggered by different operation types means that the system needs to distinguish which specific medical operations (e.g., intravenous infusion, medication administration, mechanical ventilation parameter adjustments, etc.) triggered the currently generated calibration rules and the activated calibration rules. The purpose is to provide a foundation for subsequent interaction pattern queries and complex scenario analysis.

[0066] The querying of the complex context knowledge base can be understood as the system accessing a pre-built database containing the interaction patterns between various medical and nursing procedures. This knowledge base stores the interaction patterns of different procedure types on multiple physiological parameters of patients (such as blood pressure, heart rate, and blood oxygen saturation), showing their synergistic, antagonistic, or independent effects. Its purpose is to provide a scientific basis for handling calibration rules in multi-procedural contexts.

[0067] Based on the interaction pattern, superimposed calculations or conflict resolution can be performed. For example, based on specific interaction patterns defined in the knowledge base, complex calculations such as weighted averaging, nonlinear combination, or conditional judgment can be applied to calibration rules caused by different operations to generate a composite calibration rule that more accurately reflects the current clinical context. The aim is to ensure that the judgment of abnormal physiological monitoring data can be more accurate and reliable in multi-operation scenarios.

[0068] Injecting calibration rules that reflect complex scenarios into the medical data analysis system ensures that the calibration rules, which have been analyzed and processed through complex interaction patterns, can be applied in subsequent physiological monitoring data processing, thereby improving the accuracy of data anomaly detection.

[0069] This application's solution, by introducing the identification of calibration rules arising from different operation types and combining them with interaction patterns stored in a composite context knowledge base, enables a deeper understanding of the comprehensive impact of various medical and nursing operations on patients' physiological parameters. It is precisely because of the ability to acquire and apply these complex interaction patterns that, when faced with multiple temporary calibration rules, the system no longer simply performs conflict resolution or superposition calculations, but instead generates a more clinically meaningful composite calibration rule based on the simulation of actual physiological response mechanisms. This mechanism effectively solves the potential bias problem when processing calibration rules under multiple operation scenarios, ensuring that the generated calibration rules are closer to real clinical situations.

[0070] The following is a specific example to illustrate this.

[0071] Suppose a critically ill patient is simultaneously receiving both a sedative infusion and a vasopressor infusion. Sedatives typically cause a drop in blood pressure, while vasopressors are designed to raise it. If a simple conflict resolution is used, the system might assume that the calibration rules for these two actions cancel each other out, or select one based on a preset priority, which may not accurately reflect the patient's true blood pressure trend.

[0072] The proposed solution first identifies calibration rules arising from two different administration types: sedative infusion and vasopressor infusion. The system then queries a composite context knowledge base to obtain specific interaction patterns between sedatives and vasopressors on blood pressure. For example, the knowledge base might indicate that at a specific dose, the blood pressure-raising effect of a vasopressor partially offsets the blood pressure-lowering effect of a sedative, but may simultaneously cause compensatory changes in heart rate. Based on this interaction pattern, the system generates a calibration rule reflecting this composite context. This rule may set a narrower normal blood pressure range and dynamically adjust the threshold for judging abnormal heart rate. Therefore, when a patient's blood pressure or heart rate fluctuates, the medical data analysis system can more accurately determine whether these fluctuations are due to the expected interaction of drugs or genuine physiological abnormalities, thus avoiding false alarms or missed alarms caused by simplistic processing.

[0073] In some implementations, the step of identifying calibration rules caused by different operation types in the currently generated temporary calibration rules and the already activated temporary calibration rules includes:

[0074] The operation type descriptions in the currently generated temporary calibration rules and the activated temporary calibration rules are standardized, and the operation type descriptions are mapped to one or more standard operation categories in the preset operation type classification system.

[0075] Adjust the granularity of operation type identification based on the patient's current clinical status;

[0076] Based on the standard operation category and the adjusted identification granularity, identify the calibration rules caused by different operation types.

[0077] Specifically, standardizing operation type descriptions involves transforming informal and diverse operational descriptions that may arise in verbal communication among medical staff into unified and standardized expressions using natural language processing technologies (such as lexical analysis, syntactic analysis, and entity recognition). For example, expressions like "increase infusion rate" or "speed up infusion" can be standardized to "adjust infusion rate." The pre-defined operation type classification system can be a hierarchical structure containing multiple standard operation categories, such as medication management, vital sign monitoring, and equipment operation. Through mapping, it is ensured that operations with different expressions but essentially the same function are categorized under the same standard operation category, thereby improving consistency in recognition.

[0078] Adjusting the granularity of procedure type identification based on the patient's current clinical status can be understood as dynamically adjusting the level of precision in procedure type identification based on information such as the severity of the patient's condition, the trend of fluctuations in physiological parameters, and diagnostic results. For example, for patients with stable conditions, a coarser-grained identification might be used, distinguishing only broad categories of procedures; while for critically ill patients or those with frequently fluctuating physiological parameters, a finer-grained identification might be needed, such as distinguishing different routes of administration or dosage adjustments for the same drug. This adjustment aims to make the identification results more consistent with actual clinical needs, avoiding over- or under-identification.

[0079] Based on the standard operation category and adjusted identification granularity, the system identifies calibration rules arising from different operation types. Specifically, after standardizing the operation type descriptions and dynamically adjusting the identification granularity, the system uses this information to accurately filter and match temporary calibration rules triggered or affected by different operation types. For example, if the identification granularity is adjusted to fine mode, the system will distinguish between intravenous insulin injection and subcutaneous insulin injection, even though they both belong to the standard operation category of medication administration, thus more accurately identifying calibration rules arising from these specific operations.

[0080] This application addresses the issues of diversity and non-standardization in verbal communication by standardizing operation type descriptions, ensuring that operations with different expressions but essentially the same nature can be consistently identified and categorized. Furthermore, by dynamically adjusting the granularity of operation type identification based on the patient's current clinical state, the system can flexibly select the level of precision based on the complexity of the actual clinical context. It is precisely this standardization and dynamic granularity adjustment that enables more accurate identification of calibration rules arising from different operation types, thus providing more precise and reliable input for subsequent complex context knowledge base queries and interaction pattern analysis. In this way, the proposed solution effectively improves the accuracy and adaptability of calibration rule identification in complex clinical operation scenarios, laying a solid foundation for subsequent conflict detection and overlay analysis.

[0081] Suppose a critically ill patient has persistently low blood pressure, and the medical staff verbally instructs the patient to receive an infusion of vasopressors at a faster rate.

[0082] First, the information acquisition module acquires and identifies the verbal communication information. The semantic analysis module performs semantic analysis on it, extracting the operation type as infusion of vasopressors and adjustment of infusion rate, the patient identifier as the patient, and the intention to adjust the parameters.

[0083] When generating temporary calibration rules and performing conflict detection and overlay analysis, it is necessary to identify calibration rules caused by different operation types.

[0084] Specifically, the system standardizes descriptions of operation types such as infusing vasopressors and increasing the infusion rate. For example, infusing vasopressors may be standardized as medication management - vasopressor infusion, and increasing the infusion rate may be standardized as equipment operation - infusion pump rate adjustment, and mapped to the corresponding standard operation category in the preset operation type classification system.

[0085] Simultaneously, the system considers the patient's current clinical state. For example, if the patient is in the pre-shock stage with drastic blood pressure fluctuations, the system will determine if a finer granularity of identification is needed. At this finer granularity, the system will not only identify the two broad categories of medication administration and equipment operation, but will also further distinguish between the two specific operations of vasopressor infusion and infusion pump rate adjustment, even if they may be classified as the same category under some coarse-grained classifications.

[0086] Based on these standardized standard operation categories and adjusted fine-grained identification, the system can accurately identify temporary calibration rules caused by two different operation types: vasopressor infusion and infusion pump rate adjustment. For example, vasopressor infusion may lead to increased blood pressure, while infusion pump rate adjustment may affect infusion volume and drug concentration. Through this precise identification, the system can more accurately query the composite context knowledge base to obtain the interaction patterns of these two operations on physiological parameters such as blood pressure and heart rate, thereby performing more accurate conflict detection and overlay analysis to generate calibration rules that reflect composite contexts. For example, during vasopressor infusion, the threshold for judging abnormal blood pressure can be dynamically adjusted to avoid misjudging physiological changes under normal drug action as abnormal.

[0087] In some implementations, the above-mentioned querying of a complex contextual knowledge base to obtain interaction patterns of different operation types on multiple physiological parameters further includes:

[0088] The knowledge base for complex contexts will be updated as follows:

[0089] In the composite contextual knowledge base, a confidence parameter and a verification period are set for each known operation type interaction pattern;

[0090] When physiological data fluctuation patterns that contradict the currently activated temporary calibration rules are observed in actual clinical data, or when the activated temporary calibration rules are frequently manually intervened or modified within a specific time period, these events are recorded as potential interactive pattern update trigger events.

[0091] In response to a potential interaction pattern update trigger event, an interaction pattern verification process is initiated, which includes: collecting patient physiological data, healthcare staff operation logs, and medication usage records related to the potential interaction pattern update trigger event, and performing correlation analysis on this data;

[0092] If the association analysis results indicate the existence of a new operation type interaction pattern, or the confidence parameter of an existing operation type interaction pattern decreases, generate new interaction pattern suggestions or generate suggestions to correct the existing interaction pattern;

[0093] Submit new interaction mode suggestions or revision suggestions to clinical experts for review and confirmation;

[0094] After confirmation by clinical experts, the new or revised interaction mode is injected into the composite context knowledge base, and its confidence parameters and validation cycle are updated.

[0095] At the same time, the verification cycle of interaction patterns in the composite context knowledge base is checked regularly, and the verification process is automatically triggered for patterns that exceed the verification cycle.

[0096] Specifically, the confidence parameter can be understood as a quantitative assessment of the reliability or accuracy of a specific operation type interaction pattern. Its value can be dynamically adjusted based on factors such as the number of times the pattern has been validated in historical data, its consistency with clinical expert consensus, and the frequency of corrections in practical applications. The validation cycle refers to a time interval set for each interaction pattern. After this time interval ends, the system will automatically trigger a re-evaluation and validation of the pattern to ensure its continued effectiveness and accuracy. Potential interaction pattern update trigger events refer to two situations identified by the system that may require updating the composite context knowledge base: first, when actual clinical data exhibits physiological data fluctuation patterns inconsistent with currently activated temporary calibration rules; for example, the system predicts that a certain operation will lead to an increase in physiological parameters based on existing rules, but a decrease is actually observed; second, when an activated temporary calibration rule is frequently manually intervened or corrected by medical staff within a specific time period, which usually indicates that the rule may no longer be fully applicable to the current clinical context. The interaction pattern validation process is a systematic data collection and analysis process designed to delve into the causes of triggering events. It collects multi-source information, including patient physiological data, healthcare worker operation logs, and medication usage records, and performs correlation analysis on this data to identify potential new interaction patterns or changes in existing patterns. The correlation analysis results are conclusions drawn from the comprehensive analysis of the collected multi-source data. These conclusions indicate whether previously unknown interaction patterns exist, or whether the confidence parameters of existing interaction patterns need to be lowered due to inconsistencies with reality. New interaction pattern suggestions or correction suggestions are technical solutions automatically generated by the system based on the correlation analysis results, used to update or adjust interaction patterns in the composite context knowledge base. Clinical expert review and confirmation are crucial steps to ensure the accuracy and security of knowledge base updates. Suggestions generated by the system are evaluated and approved by clinicians with specialized knowledge. The validation cycle of interaction patterns in the composite context knowledge base is checked regularly. For patterns exceeding the validation cycle, the validation process is automatically triggered. This proactive maintenance mechanism ensures that the knowledge base is regularly reviewed and updated, even in the absence of obvious triggering events.

[0097] This application effectively addresses the problem of inaccurate calibration rules caused by the lag in the knowledge base due to changes in clinical practice by introducing a dynamic update mechanism for the composite context knowledge base. Specifically, by setting confidence parameters and verification cycles for each interaction pattern, the system can quantitatively assess the reliability of the knowledge and proactively plan its review. When actual clinical data contradicts existing rules, or when rules are frequently manually intervened, these events are recorded as potential update triggers, prompting the system to initiate a comprehensive interaction pattern verification process. This process, through the collection and correlation analysis of patient physiological data, medical staff operation logs, and medication usage records, can delve into the underlying causes of data anomalies or rule failures, thereby identifying new interaction patterns or changes in existing patterns. As a result, the system can generate new interaction pattern suggestions or correction suggestions, and after review and confirmation by clinical experts, these updates are injected into the composite context knowledge base. This continuous, data-driven update mechanism ensures that the composite context knowledge base remains synchronized with the latest clinical practices, thus providing the most accurate and reliable interaction pattern information for the aforementioned conflict detection and overlay analysis.

[0098] Suppose a high-confidence interaction pattern exists in a composite context knowledge base regarding the impact of infusion pump adjustments and antihypertensive medication use on a patient's blood pressure, with a validation period of six months. In a clinical practice, the system generates a temporary calibration rule based on this interaction pattern. However, subsequent observations show a persistent and significant difference between the patient's blood pressure fluctuation pattern and the rule's predictions, and healthcare professionals manually correct the system's assessment of abnormal blood pressure multiple times within a short period. These events are identified and recorded by the system as potential interaction pattern update triggers. In response, the system initiates an interaction pattern validation process, automatically collecting detailed physiological monitoring data of the patient within the relevant time period, healthcare professionals' operation logs (including specific infusion pump adjustment records, antihypertensive medication dosage and timing), and other medication usage records. Through correlation analysis of these multi-source data, the system discovers that the patient was simultaneously using a novel diuretic, and the combined effect of this diuretic and the antihypertensive medication altered the original blood pressure interaction pattern. Based on this correlation analysis, the system generated a new interaction pattern suggestion, pointing out that the effect of infusion pump adjustment + antihypertensive drugs + novel diuretics on blood pressure differed significantly from the original pattern, and recommended updating the knowledge base. This suggestion was subsequently submitted to clinical experts for review. After the clinical experts confirmed that the novel diuretic did indeed change the original interaction pattern, the new interaction pattern was injected into the composite context knowledge base, and corresponding confidence parameters and validation periods were set. Simultaneously, the system also periodically checks the validation periods of all interaction patterns. For example, for the original interaction pattern of infusion pump adjustment and antihypertensive drug use, even without triggered events, the system automatically initiates the validation process after the six-month validation period to reassess its effectiveness. In this way, the composite context knowledge base is continuously updated and optimized, ensuring that the system can consistently provide accurate and reliable anomaly detection for physiological monitoring data in complex and ever-changing clinical environments.

[0099] In some implementations, the step of recording these events as potential interactive pattern update trigger events when physiological data fluctuation patterns that contradict the currently activated temporary calibration rule are observed in actual clinical data, or when the activated temporary calibration rule is frequently manually intervened or modified within a specific time period, includes:

[0100] When physiological data fluctuations that contradict the currently activated temporary calibration rules are observed in actual clinical data, the device operating status and data transmission links corresponding to the physiological data fluctuations should be checked.

[0101] Based on the inspection results of equipment operating status and data transmission links, trend analysis is performed on physiological data fluctuations to determine whether physiological data fluctuations exhibit patterns related to known equipment failures or data transmission anomalies.

[0102] At the same time, recent operation logs and medication records for patients are obtained and compared with the time points of physiological data fluctuations to identify whether there are clinical operations or drug effects corresponding to physiological data fluctuations.

[0103] Based on the inspection results of the equipment operating status and data transmission link, the trend analysis results, and the comparison results, it is determined whether the fluctuation of physiological data is caused by non-interactive mode factors or indicates potential changes in interactive modes.

[0104] When fluctuations in physiological data indicate potential changes in interaction patterns, these physiological data fluctuation events are recorded as potential interaction pattern update trigger events.

[0105] This includes checking the operational status of equipment and the data transmission link corresponding to fluctuations in physiological data, aiming to initially rule out data anomalies caused by hardware failures or network anomalies. For example, the internal logs of relevant monitoring equipment can be queried to check whether its sensors are working properly, whether the power supply is stable, and whether there are error reports in the data transmission module. Simultaneously, checking the data transmission link can include evaluating indicators such as network connectivity status and packet loss rate to ensure data integrity and real-time performance.

[0106] Based on the inspection results of equipment operation status and data transmission links, trend analysis is performed on physiological data fluctuations to determine whether they exhibit patterns related to known equipment malfunctions or data transmission anomalies. For example, certain equipment malfunctions may cause periodic data loss, fixed offsets, or sudden jumps, while data transmission anomalies may manifest as data delays or intermittent interruptions. By comparing the currently observed physiological data fluctuation trends with these pre-defined known patterns, abnormalities caused by non-clinical factors can be effectively identified.

[0107] Acquiring recent patient procedure logs and medication usage records and comparing them with time points of physiological data fluctuations aims to identify any clinical procedures or medication effects corresponding to these fluctuations. For example, certain procedures performed on patients by healthcare professionals (such as changes in position, suctioning, or intravenous infusion) or newly administered medications (such as vasopressors or sedatives) may cause normal fluctuations in physiological parameters within a short period. While these fluctuations may contradict current calibration rules, they do not indicate a fundamental change in interaction patterns. Precise time comparisons can distinguish these physiological responses induced by clinical interventions.

[0108] Based on the inspection results of equipment operation status and data transmission links, trend analysis results, and comparison results, the system will comprehensively determine whether the fluctuations in physiological data are caused by non-interaction mode factors or indicate potential changes in interaction modes. Only when non-interaction mode factors such as equipment failure, abnormal data transmission, and known clinical operations or drug effects are ruled out, and the fluctuations in physiological data still significantly contradict the currently activated temporary calibration rules, will they be identified as indicating potential changes in interaction modes and recorded as potential interaction mode update trigger events.

[0109] This application effectively addresses potential misjudgment issues in the basic solution by introducing a multi-dimensional and systematic verification mechanism. Specifically, firstly, by examining the equipment operating status and data transmission links corresponding to fluctuations in physiological data, it can initially rule out data anomalies caused by hardware failures or network anomalies, avoiding misjudging fluctuations caused by non-clinical factors as changes in interaction patterns. Secondly, combined with trend analysis, it can identify whether physiological data fluctuations conform to known equipment failure or data transmission anomaly patterns, further enhancing the ability to identify non-interaction pattern factors. Simultaneously, by acquiring and comparing patients' recent operation logs and medication records, it can identify physiological changes caused by clinical operations or drug effects by medical staff. While these changes may contradict calibration rules, they do not indicate fundamental changes in interaction patterns. Finally, by integrating these inspection, analysis, and comparison results, the system can more accurately determine whether physiological data fluctuations originate from non-interaction pattern factors or truly indicate potential changes in the interaction pattern of the operation type. This hierarchical and progressive judgment logic ensures that only rigorously verified events are recorded as potential interaction pattern update trigger events, thereby avoiding inaccurate knowledge base updates due to misjudgments.

[0110] Suppose that a critically ill patient's blood pressure monitoring data suddenly shows a significant drop, contradicting the currently activated temporary calibration rule for that patient (e.g., a rule that sets the normal blood pressure range at a higher level based on the patient's current medication). In this case, the scheme of this application will initiate the following process:

[0111] First, the system automatically checks the operational status of the blood pressure monitoring device (e.g., a non-invasive blood pressure monitor or an invasive arterial pressure monitor) and its data transmission link. For example, by querying the device log, it confirms whether the device is in normal operating mode and whether there are any power outages, sensor malfunctions, or unstable network connections. If the check results show that the device and the link are normal, it proceeds to the next step.

[0112] Secondly, trend analysis is performed on the blood pressure fluctuation data. The system compares the current fluctuation pattern with preset known equipment failure patterns (e.g., sensor drift, intermittent data loss due to low battery) or data transmission anomaly patterns (e.g., step-like drops due to packet loss). If the trend analysis indicates that the fluctuation does not conform to any known equipment failure or data transmission anomaly patterns, clinical factors are further considered.

[0113] At the same time, the system will access the patient's recent (e.g., within the past 30 minutes) medical staff operation logs and medication usage records. For example, by comparing the time points when blood pressure drops, it can identify whether new antihypertensive drugs have been injected, whether the patient has changed position, or whether other clinical procedures that may affect blood pressure have been performed.

[0114] Finally, a comprehensive judgment is made based on the above examination, trend analysis, and comparison results. If the blood pressure drop is found to closely coincide with the timing of a recently injected new drug (e.g., a sedative or vasodilator), and this drug is known to cause a drop in blood pressure, then the system will determine that the physiological data fluctuation is caused by a non-interaction mode factor (i.e., drug effect), rather than a potential change in the interaction mode indicating the operation type. In this case, the event will not be recorded as a potential interaction mode update trigger event, but may instead trigger a fine-tuning of the drug effect parameters of the current temporary calibration rule. Conversely, if factors such as equipment, data transmission, and known clinical operations or drug effects are excluded, and the fluctuation pattern persists and is inconsistent with the existing interaction mode, then the event will be recorded as a potential interaction mode update trigger event to initiate the verification process of the composite context knowledge base.

[0115] In some implementations, the step of performing trend analysis on physiological data fluctuations to determine whether the physiological data fluctuations exhibit a pattern related to known equipment failures or data transmission anomalies includes:

[0116] Obtain the operation logs of medical staff that overlap with the time periods of physiological data fluctuations;

[0117] Identify whether there are operation records related to equipment calibration or maintenance in the operation log;

[0118] Based on the patterns of impact on physiological data identified in the operation logs, a corrective assessment of physiological data fluctuations is performed.

[0119] Determine whether fluctuations in physiological data conform to the range and duration of fluctuations caused by normal operation as determined by the corrective assessment;

[0120] When physiological data fluctuations fall within the range and duration of fluctuations caused by normal operation, the physiological data fluctuations are marked as fluctuations caused by normal operation and are not classified as equipment failures or abnormal data transmission patterns.

[0121] Specifically, acquiring healthcare worker operation logs that overlap with the time period of physiological data fluctuations means that when the system detects abnormal fluctuations in physiological data, it automatically retrieves healthcare worker operation records corresponding to the time period in which the fluctuation occurred. These operation logs can come from electronic medical record systems, nursing record systems, or dedicated equipment operation record systems, and their purpose is to provide contextual information for subsequent analysis of the causes of the fluctuations.

[0122] Identifying whether there are operation records related to equipment calibration or maintenance in the operation log can be understood as performing keyword matching, semantic analysis, or comparison according to preset rules on the acquired operation log to find information indicating equipment calibration or maintenance activities, such as equipment calibration, sensor cleaning, parameter reset, and maintenance checks. This step aims to initially screen out normal operational events that may cause fluctuations in physiological data.

[0123] Based on the patterns of impact on physiological data identified in the operation logs from equipment calibration or maintenance operations, a corrective assessment of physiological data fluctuations is performed. Specifically, the system uses a pre-stored knowledge base or model to understand which physiological parameters are typically affected by different equipment calibration or maintenance operations, and what type (e.g., transient increase, decrease, fluctuation amplitude) and duration of these effects. For example, a certain ventilator calibration may cause a transient increase in respiratory rate data over a short period. Through this assessment, a model of expected fluctuations can be established.

[0124] Determining whether fluctuations in physiological data conform to the range and duration of fluctuations caused by normal operation as determined by the corrective assessment involves comparing the actual observed fluctuations in physiological data with the expected fluctuation model established by the corrective assessment. The comparison includes the amplitude, duration, and morphological characteristics of the fluctuations. For example, if the actual fluctuations are within the expected range and the duration matches the typical effect time of the calibration operation, then it is considered to conform to the fluctuation pattern caused by normal operation.

[0125] When fluctuations in physiological data fall within the range and duration of fluctuations caused by normal operation, these fluctuations are marked as normal operational fluctuations and are not categorized as equipment malfunctions or abnormal data transmission patterns. This means the system will treat such fluctuations as predictable and harmless events, thus avoiding misjudging them as anomalies requiring further investigation, such as potential changes in interaction patterns or equipment malfunctions.

[0126] This application, by introducing the analysis of medical staff's operation logs and combining them with preset patterns of the impact of equipment calibration or maintenance operations on physiological data, enables a refined corrective assessment of physiological data fluctuations. Because the system can proactively identify and exclude physiological data fluctuations caused by normal clinical operations during trend analysis, subsequent judgments are more focused on fluctuations caused by abnormal factors. In this way, the system can effectively distinguish between normal noise and true signals, thereby avoiding misjudging transient data changes caused by routine operations as abnormal events requiring the triggering of interactive mode update verification processes.

[0127] This application significantly improves the accuracy of identifying abnormal physiological data. Compared to trend analysis relying solely on the physiological data itself, this solution, by combining medical staff operation logs, effectively filters out physiological data fluctuations caused by normal operations such as equipment calibration and maintenance, thereby reducing the false alarm rate. This precise identification capability allows the system to more accurately determine whether physiological data fluctuations truly indicate potential changes in interaction patterns or equipment malfunctions, avoiding unnecessary verification processes and saving computational resources and medical staff review time. Furthermore, by marking fluctuations caused by normal operations, it also provides medical staff with a clearer data context, helping them better understand the patient's physiological state and equipment operation.

[0128] The following is a specific example to illustrate this.

[0129] Suppose that in the intensive care unit, a patient's blood oxygen saturation data suddenly and briefly drops, then quickly returns to normal. According to the above-described approach, the system might identify this fluctuation as a potential interactive mode update trigger event and initiate a verification process. However, according to the optimized approach of this application, the system first retrieves the medical staff operation logs that overlap with the time period of the blood oxygen saturation fluctuation. If the operation logs show that a few minutes before the fluctuation occurred, the nursing staff performed probe position adjustments or equipment calibration operations on the blood oxygen saturation monitoring device, the system will identify these calibration or maintenance-related operation records.

[0130] Based on a pre-set knowledge base, the system understands that probe position adjustments or equipment calibration typically cause slight fluctuations or brief drops in blood oxygen saturation data over a short period (e.g., 10-30 seconds), with the fluctuation range usually within a specific range. The system compares the actual observed blood oxygen saturation fluctuations with this expected pattern of influence from probe position adjustments. If the actual decrease in blood oxygen saturation, its duration, and recovery pattern all match the fluctuation range and duration caused by probe position adjustments, the system determines that the blood oxygen saturation fluctuation is consistent with fluctuations caused by normal operation.

[0131] The system will classify the fluctuation in blood oxygen saturation as a normal operational fluctuation, rather than categorizing it as a device malfunction or abnormal data transmission pattern, and will not trigger potential interactive mode update verification processes. In this way, the system avoids misjudging and overreacting to harmless fluctuations caused by normal operation, thereby improving the accuracy of data analysis and the system's operational efficiency.

[0132] In some implementations, the step of performing trend analysis on physiological data fluctuations to determine whether the physiological data fluctuations exhibit a pattern related to known equipment failures or data transmission anomalies includes:

[0133] Acquire equipment operating parameters and environmental parameters during periods of physiological data fluctuation;

[0134] Based on equipment operating parameters and environmental parameters, multi-channel synchronous analysis of physiological data fluctuations is performed.

[0135] Based on the results of multi-channel synchronous analysis, the fluctuations in physiological data are subjected to spectral decomposition.

[0136] The spectral decomposition results are compared with the preset non-interactive mode factor characteristic spectra;

[0137] Based on the comparison results, the fluctuations in physiological data were classified as being caused by specific non-interactive pattern factors.

[0138] Specifically, when acquiring equipment operating parameters and environmental parameters during periods of physiological data fluctuation, parameters directly related to the physiological monitoring equipment can be collected, such as sensor impedance, power supply voltage, and signal-to-noise ratio, as well as environmental factors such as room temperature, humidity, and electromagnetic interference intensity. The acquisition of these parameters aims to provide comprehensive background information for subsequent analysis of potential non-physiological causes of physiological data fluctuations.

[0139] Multi-channel synchronous analysis of physiological data fluctuations based on equipment operating parameters and environmental parameters can be understood as performing time synchronization and correlation analysis on the fluctuations of multiple physiological parameters (such as heart rate, blood pressure, blood oxygen saturation, etc.) with their corresponding equipment operating parameters and environmental parameters. The purpose is to identify whether physiological data fluctuations occur synchronously with specific changes in equipment or environmental parameters, thereby making a preliminary judgment on whether the fluctuations originate from non-interactive factors. For example, if an abnormal fluctuation in a certain physiological parameter occurs simultaneously with a momentary drop in power supply voltage, it may indicate a power supply problem.

[0140] Based on the results of multi-channel synchronous analysis, spectral decomposition of physiological data fluctuations refers to using Fourier transform or other time-frequency analysis techniques to decompose the physiological data fluctuation signal into different frequency components. The purpose is to reveal whether specific periodicity or frequency characteristics exist in the data fluctuations, which are often associated with equipment failure, environmental noise, or specific artifact patterns. For example, power line interference typically exhibits significant spectral peaks at specific frequencies (such as 50Hz or 60Hz).

[0141] Comparing the spectral decomposition results with the pre-defined non-interactive pattern factor characteristic spectrum refers to matching the spectral characteristics of actually observed physiological data fluctuations with a pre-established characteristic spectrum library containing known equipment failures, environmental interference, or artifact patterns. This characteristic spectrum library can be constructed through historical data analysis, experimental simulation, or expert knowledge. Its purpose is to accurately identify the specific non-interactive pattern factors causing physiological data fluctuations through pattern recognition.

[0142] Based on the comparison results, classifying fluctuations in physiological data as caused by specific non-interactional modal factors means that, based on the similarity or matching degree of the spectral comparison, the currently observed fluctuations in physiological data are explicitly attributed to a known non-interactional modal factor, such as sensor detachment, poor electrode contact, power supply interference, motion artifacts, etc. This classification helps to distinguish fluctuations caused by non-interactional modal factors from true physiological changes or interaction patterns.

[0143] This application acquires equipment operating parameters and environmental parameters during periods of physiological data fluctuation and performs multi-channel synchronous analysis of these parameters to more comprehensively capture potential non-interactive pattern factors causing data anomalies. Traditional trend analysis may only focus on the anomalies in the physiological data itself, ignoring the underlying equipment or environmental causes. By introducing multi-channel synchronous analysis, the synchronous relationship between physiological data fluctuations and changes in equipment or environmental parameters can be identified, thus initially screening out non-physiological fluctuations. Furthermore, by performing spectral decomposition on physiological data fluctuations, hidden periodic or frequency characteristics in the data can be revealed. These characteristics are often unique fingerprints of equipment failure or environmental interference. Finally, the spectral decomposition results are compared with the preset non-interactive pattern factor characteristic spectrum, enabling the system to accurately identify and classify the specific non-interactive pattern factors causing the fluctuations. This multi-dimensional and refined analysis method allows the system to more accurately distinguish physiological data fluctuations caused by non-interactive pattern factors such as equipment failure and environmental interference, avoiding misjudging them as potential events requiring an update of the interaction pattern.

[0144] like Figure 2 As shown in the embodiments of this application, an artificial intelligence-based medical data information analysis system is also disclosed for calibrating the judgment of abnormal physiological monitoring data in an intensive care environment. The system includes:

[0145] Information acquisition module 1 is used to acquire verbal communication information from medical staff and to identify the verbal communication information in order to obtain instructions or statements in text form;

[0146] Semantic analysis module 2 is used to perform semantic analysis on text-based instructions or statements to extract operation type, patient identification, time information, and intent to adjust parameters;

[0147] Rule generation and injection module 3 is used to generate temporary calibration rules for correcting data interpretation logic based on semantic analysis results, and inject the temporary calibration rules into the medical data analysis system;

[0148] Data processing module 4 is used to prioritize the application of temporary calibration rules when processing physiological monitoring data in the medical data analysis system, in order to correct the judgment of data anomalies.

[0149] The system proposed in this application can significantly improve the accuracy and reliability of data anomaly detection in complex intensive care environments.

[0150] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A medical data information analysis method based on artificial intelligence, used to calibrate the judgment of abnormal physiological monitoring data in an intensive care environment, characterized in that, include: Acquire verbal communication information from healthcare workers and identify such information to obtain instructions or statements in text form; Semantic analysis is performed on text-based instructions or statements to extract operation type, patient identification, time information, and intent to adjust parameters; Based on the semantic analysis results, temporary calibration rules are generated to correct the data interpretation logic, and these temporary calibration rules are injected into the medical data analysis system. When processing physiological monitoring data in a medical data analysis system, temporary calibration rules should be applied first to correct any judgments of data anomalies.

2. The method for analyzing medical data based on artificial intelligence according to claim 1, characterized in that, The step of generating temporary calibration rules for correcting data interpretation logic based on semantic analysis results, and injecting the temporary calibration rules into the medical data analysis system includes: Obtain the patient's basic physiological parameters; Impact patterns of query operations; Calculate the dynamic calibration threshold range based on the patient's baseline physiological parameters and the impact pattern of the operation; Determine the effective duration of the temporary calibration rule based on the impact pattern of the operation; Based on the semantic analysis results, temporary calibration rules containing dynamic calibration threshold ranges and effective durations are generated and injected into the medical data analysis system.

3. The method for analyzing medical data based on artificial intelligence according to claim 2, characterized in that, The step of generating temporary calibration rules containing dynamic calibration threshold ranges and effective durations based on semantic analysis results, and injecting the temporary calibration rules into the medical data analysis system, further includes: Query currently active temporary calibration rules for the same patient identifier; Conflict detection and overlay analysis are performed on the currently generated temporary calibration rules and the already activated temporary calibration rules. When a conflict is detected between a currently generated temporary calibration rule and an already activated temporary calibration rule, the conflict is adjudicated according to a preset priority strategy to generate a correction rule. When an additive effect is detected between the currently generated temporary calibration rule and the already activated temporary calibration rule, a composite calibration rule is generated according to the preset additive calculation logic. Inject correction rules or composite calibration rules into the medical data analysis system.

4. The method for analyzing medical data based on artificial intelligence according to claim 3, characterized in that, The steps for conflict detection and overlay analysis between the currently generated temporary calibration rules and the already activated temporary calibration rules include: Identify calibration rules caused by different operation types in the currently generated temporary calibration rules and the already activated temporary calibration rules; Query the complex context knowledge base to obtain the interaction patterns of different operation types on multiple physiological parameters; Based on the interaction mode, the currently generated temporary calibration rules and the activated temporary calibration rules are superimposed for calculation or conflict resolution to generate calibration rules that reflect complex situations. Inject calibration rules that reflect complex scenarios into the medical data analysis system.

5. The method for analyzing medical data based on artificial intelligence according to claim 4, characterized in that, The step of identifying calibration rules caused by different operation types in the currently generated temporary calibration rules and the already activated temporary calibration rules includes: The operation type descriptions in the currently generated temporary calibration rules and the activated temporary calibration rules are standardized, and the operation type descriptions are mapped to one or more standard operation categories in the preset operation type classification system. Adjust the granularity of operation type identification based on the patient's current clinical status; Based on the standard operation category and the adjusted identification granularity, identify the calibration rules caused by different operation types.

6. The method for analyzing medical data based on artificial intelligence according to claim 4, characterized in that, The step of querying the composite context knowledge base to obtain the interaction patterns of different operation types on multiple physiological parameters includes: The knowledge base for complex contexts will be updated as follows: In the composite contextual knowledge base, a confidence parameter and a verification period are set for each known operation type interaction pattern; When physiological data fluctuation patterns that contradict the currently activated temporary calibration rules are observed in actual clinical data, or when the activated temporary calibration rules are frequently manually intervened or modified within a specific time period, these events are recorded as potential interactive pattern update trigger events. In response to a potential interaction pattern update trigger event, an interaction pattern verification process is initiated, which includes: collecting patient physiological data, healthcare staff operation logs, and medication usage records related to the potential interaction pattern update trigger event, and performing correlation analysis on this data; If the association analysis results indicate the existence of a new operation type interaction pattern, or the confidence parameter of an existing operation type interaction pattern decreases, generate new interaction pattern suggestions or generate suggestions to correct the existing interaction pattern; Submit new interaction mode suggestions or revision suggestions to clinical experts for review and confirmation; After confirmation by clinical experts, the new or revised interaction mode is injected into the composite context knowledge base, and its confidence parameters and validation cycle are updated. At the same time, the verification cycle of interaction patterns in the composite context knowledge base is checked regularly, and the verification process is automatically triggered for patterns that exceed the verification cycle.

7. The method for analyzing medical data based on artificial intelligence according to claim 6, characterized in that, The steps for recording these events as potential interactive pattern update trigger events when physiological data fluctuation patterns that contradict the currently activated temporary calibration rules are observed in actual clinical data, or when the activated temporary calibration rules are frequently manually intervened or modified within a specific time period, include: When physiological data fluctuations that contradict the currently activated temporary calibration rules are observed in actual clinical data, the device operating status and data transmission links corresponding to the physiological data fluctuations should be checked. Based on the inspection results of equipment operating status and data transmission links, trend analysis is performed on physiological data fluctuations to determine whether physiological data fluctuations exhibit patterns related to known equipment failures or data transmission anomalies. At the same time, recent operation logs and medication records for patients are obtained and compared with the time points of physiological data fluctuations to identify whether there are clinical operations or drug effects corresponding to physiological data fluctuations. Based on the inspection results of the equipment operating status and data transmission link, the trend analysis results, and the comparison results, it is determined whether the fluctuation of physiological data is caused by non-interactive mode factors or indicates potential changes in interactive modes. When fluctuations in physiological data indicate potential changes in interaction patterns, these physiological data fluctuation events are recorded as potential interaction pattern update trigger events.

8. The method for analyzing medical data based on artificial intelligence according to claim 7, characterized in that, The step of performing trend analysis on physiological data fluctuations to determine whether the physiological data fluctuations exhibit a pattern related to known equipment failures or abnormal data transmission includes: Obtain the operation logs of medical staff that overlap with the time periods of physiological data fluctuations; Identify whether there are operation records related to equipment calibration or maintenance in the operation log; Based on the patterns of impact on physiological data identified in the operation logs, a corrective assessment of physiological data fluctuations is performed. Determine whether fluctuations in physiological data conform to the range and duration of fluctuations caused by normal operation as determined by the corrective assessment; When physiological data fluctuations fall within the range and duration of fluctuations caused by normal operation, the physiological data fluctuations are marked as fluctuations caused by normal operation and are not classified as equipment failures or abnormal data transmission patterns.

9. The method for analyzing medical data based on artificial intelligence according to claim 7, characterized in that, The step of performing trend analysis on physiological data fluctuations to determine whether the physiological data fluctuations exhibit a pattern related to known equipment failures or abnormal data transmission includes: Acquire equipment operating parameters and environmental parameters during periods of physiological data fluctuation; Based on equipment operating parameters and environmental parameters, multi-channel synchronous analysis of physiological data fluctuations is performed. Based on the results of multi-channel synchronous analysis, the fluctuations in physiological data are subjected to spectral decomposition. The spectral decomposition results are compared with the preset non-interactive mode factor characteristic spectra; Based on the comparison results, the fluctuations in physiological data were classified as being caused by specific non-interactive pattern factors.

10. A medical data information analysis system based on artificial intelligence, used to calibrate the judgment of abnormal physiological monitoring data in an intensive care environment, characterized in that, The system includes: The information acquisition module is used to acquire verbal communication information from medical staff and to identify the verbal communication information in order to obtain instructions or statements in text form. The semantic analysis module is used to perform semantic analysis on text-based instructions or statements to extract operation type, patient identification, time information, and intent to adjust parameters; The rule generation and injection module is used to generate temporary calibration rules for correcting data interpretation logic based on semantic analysis results, and inject the temporary calibration rules into the medical data analysis system. The data processing module is used to prioritize the application of temporary calibration rules when processing physiological monitoring data in the medical data analysis system, in order to correct the judgment of data anomalies.

Images