Self-adaptive infusion control method and system based on physiological parameters

Abstract

The invention provides a self-adaptive infusion control method and system based on physiological parameters. Vital signs and infusion states of a patient are monitored in real time, change trends of the parameters are analyzed, whether the infusion rate needs to be adjusted or infusion needs to be stopped or not is judged, a corresponding infusion adjusting instruction is generated, and infusion equipment is dynamically adjusted according to the infusion adjusting instruction. Under the abnormal condition, the risk level is evaluated, an alarm is given to inform medical staff, and meanwhile all related data are recorded to form an infusion process log. After each infusion is finished, the relationship between the multi-dimensional physiological parameters and abnormal events is analyzed, potential influence factors and interaction of the potential influence factors are identified, and a decision algorithm and a control strategy in the future infusion process are subjected to predictive optimization through a machine learning model, so that the infusion control strategy of a patient is adjusted in a personalized mode. According to the invention, the safety and flexibility of infusion are improved, and personalized medical care is realized.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data processing technology, and in particular to an adaptive infusion control and system based on physiological parameters. Background Art

[0002] In the modern medical environment, infusion therapy is a common treatment method widely used in various diseases and postoperative care. In order to ensure the safety and effectiveness of the infusion process, medical institutions have strict requirements for real-time monitoring of patients' vital signs and infusion status.

[0003] Currently, the infusion control systems used in most hospitals mainly rely on manual settings and fixed rules to control the infusion rate. Some advanced systems have begun to introduce automated monitoring equipment that can collect physiological parameters in real time and trigger alarms through preset thresholds.

[0004] However, the decision logic of these systems is usually static, based on fixed clinical guidelines or empirical values, and lacks dynamic adaptability. Some high-end systems may integrate simple machine learning models for preliminary risk assessment, but their optimization is limited and they fail to make full use of historical case data for continuous improvement, resulting in a lack of flexibility and personalization in the infusion control system. Summary of the invention

[0005] The embodiments of the present application provide an adaptive infusion control method and system based on physiological parameters, so as to solve the problem that the infusion control system in the prior art lacks flexibility and personalization.

[0006] In a first aspect, an embodiment of the present application provides an adaptive infusion control method based on physiological parameters, comprising:

[0007] Monitor the patient's vital signs and infusion status in real time to obtain monitoring data, including multi-dimensional physiological parameters, current infusion rate, and cumulative infusion volume;

[0008] Based on the monitoring data, a multi-level decision tree algorithm is used to analyze the change trend of the multi-dimensional physiological parameters, determine whether it is necessary to adjust the infusion rate or stop the infusion, and generate an infusion adjustment instruction; the multi-level decision tree is constructed according to clinical guidelines, individualized treatment principles and historical case data analysis, and the decision rule set of the multi-level decision tree is continuously optimized in combination with a reinforcement learning algorithm;

[0009] According to the infusion adjustment instruction, the infusion equipment is dynamically adjusted to ensure the safety and effectiveness of the infusion process, and when an abnormal situation is detected, the risk level is evaluated using the Bayesian network, and an alarm of a corresponding level is issued based on the risk level to notify medical staff; at the same time, all multi-dimensional physiological parameters during each infusion process and the abnormal information corresponding to the abnormal situation are recorded to form an infusion process log;

[0010] After each infusion process is completed, based on the infusion process log, an association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during this infusion, so as to identify potential influencing factors and their interactions. Based on the influencing factors and their interactions, a machine learning model is used to predictively optimize the patient's decision algorithm and infusion control strategy in future infusion processes, so as to adjust the patient's infusion control strategy.

[0011] Optionally, it also includes:

[0012] Aiming at the unique physiological response pattern of the patient, based on the adjusted infusion control strategy, the individual differences between different patients are analyzed by a learning algorithm, key influencing factors are identified and the dependency of the key influencing factors in time series is captured to obtain an identification result; the key influencing factors refer to factors that affect the infusion effect and are used to adjust the specific content of the infusion control strategy;

[0013] Based on the recognition result, a gradient boosting decision tree and a long short-term memory network are used to perform personalized calibration processing on the infusion control parameters in the infusion control strategy to obtain a personalized infusion control strategy that meets the immediate needs of the patient.

[0014] Optionally, based on the infusion process log, an association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during this infusion to identify potential influencing factors and their interactions, including:

[0015] Based on the infusion process log, an association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events, wherein the association rule mining algorithm is intended to be used to discover whether a specific physiological parameter change pattern or combination is associated with the occurrence of abnormal events, and also to discover the interaction between physiological parameters in different dimensions, and generate a preliminary list of influencing factors and their interactions;

[0016] Further screening and refining the preliminary list to identify and list key influencing factors and their interactions, and determine a final list of influencing factors and their interactions;

[0017] Among them, the key influencing factors are the key causes of abnormal events, or important indicators related to the infusion effect; each entry in the final list describes in detail the specific properties of the influencing factors and the specific process of the influencing factors acting together on the infusion process.

[0018] Optionally, the predictive optimization of the decision algorithm and infusion control strategy of the patient in the future infusion process based on the influencing factors and their interactions by a machine learning model to adjust the infusion control strategy of the patient includes:

[0019] According to the final list, an appropriate machine learning model is selected, and the multi-dimensional physiological parameters and abnormal events contained in the historical infusion process log of the patient are used as training samples to obtain a trained machine learning model;

[0020] Based on the trained machine learning model, the predicted abnormal events of the patient during the future infusion process are obtained, and based on the predicted abnormal events, the predicted risk of the patient during the infusion process in the future period of time is determined according to the latest monitoring data of the patient obtained, and a prediction result is generated;

[0021] The prediction result is used to dynamically adjust the infusion control strategy of the patient.

[0022] Optionally, based on the monitoring data, using a multi-level decision tree algorithm to analyze the change trend of the multi-dimensional physiological parameters to determine whether it is necessary to adjust the infusion rate or stop the infusion, includes:

[0023] According to clinical guidelines, individualized treatment principles and historical case data analysis, a multi-level decision tree model is constructed, and the multi-level decision tree model is used to analyze the change trend of the multi-dimensional physiological parameters to determine whether it is necessary to adjust the infusion rate or stop the infusion;

[0024] Inputting the monitoring data into the multi-level decision tree model, evaluating the change trend and combination pattern of the physiological parameters of each dimension in the monitoring data layer by layer, and generating analysis results of the change trend of the multi-dimensional physiological parameters;

[0025] Based on the analysis results, determine whether the patient is currently at risk, and if so, adjust the infusion rate or stop the infusion based on the risk situation.

[0026] Optionally, when an abnormal situation is detected, the risk level is evaluated using a Bayesian network, and an alarm of a corresponding level is issued based on the risk level to notify medical staff, including:

[0027] When an abnormal situation is detected, the abnormal situation and its corresponding multi-dimensional physiological parameters are input into the activated Bayesian network model, so that the Bayesian network model can dynamically update the posterior probability distribution of the abnormal event corresponding to the abnormal situation according to the interaction between physiological parameters in different dimensions and the prior knowledge learned from historical data; the prior knowledge refers to the risk level of a specific combination of physiological parameters in historical cases;

[0028] Evaluate the probability of risk occurring in the current state according to the posterior probability distribution, and classify the risk level according to the evaluation result;

[0029] Based on the risk level, an alert of the corresponding level is issued to medical staff.

[0030] Optionally, selecting an appropriate machine learning model according to the final list, and using the multi-dimensional physiological parameters and abnormal events contained in the patient's historical infusion process log as training samples to obtain a trained machine learning model includes:

[0031] The trained machine learning model is obtained through the following calculation formula:

[0032] M=f(X train ,Y train ;θ)

[0033] Where M represents the trained machine learning model; X train Y is a training set composed of multi-dimensional physiological parameters extracted from historical nighttime logs; train is the corresponding historical abnormal event label; θ is the parameter set of the model, including weights and bias terms;

[0034] The method comprises: obtaining predicted abnormal events of the patient during the future infusion process based on the trained machine learning model, and determining the predicted risk of the patient during the infusion process within a future period of time based on the predicted abnormal events and the latest monitoring data of the patient obtained, and generating a prediction result, including:

[0035] The trained machine learning model is applied to simulated monitoring data in the future to predict abnormal events that may occur in the future. The definition of predicting abnormal events is:

[0036]

[0037] in, represents the predicted abnormal event at the future time point t+Δt; X sim (t+Δt) is the multi-dimensional physiological parameter input in the simulated future period; θ is the trained model parameter;

[0038] Based on the predicted abnormal event, according to the latest monitoring data of the patient obtained, the predicted risk in the infusion process in the future period is determined, and an assessment result of the predicted risk is generated as a prediction result; wherein the assessment result of the predicted risk is defined as:

[0039]

[0040] Among them, R(t+Δt) represents the evaluation result of the predicted risk at the future time point t+Δt; X hist is a multidimensional physiological parameter in the historical time period; φ is the parameter set of the risk assessment function g(·);

[0041] The method of dynamically adjusting the infusion control strategy of the patient by using the prediction result includes:

[0042] Based on the evaluation result of the predicted risk, the infusion control strategy of the patient is dynamically adjusted to ensure the safety and effectiveness of the infusion process; wherein the adjusted infusion control strategy is defined as:

[0043] C'=h(C,R(t+Δt);ω)

[0044] Among them, C' represents the adjusted infusion control strategy; C is the current infusion control parameter vector; ω is the parameter set of the adjustment function h(·).

[0045] In a second aspect, an embodiment of the present application provides an adaptive infusion control system based on physiological parameters, comprising:

[0046] A monitoring module is used to monitor the patient's vital signs and infusion status in real time and obtain monitoring data, which includes multi-dimensional physiological parameters, current infusion rate, and cumulative infusion volume;

[0047] An analysis module, for analyzing the change trend of the multi-dimensional physiological parameters based on the monitoring data using a multi-level decision tree algorithm, determining whether it is necessary to adjust the infusion rate or stop the infusion, and generating an infusion adjustment instruction; the multi-level decision tree is constructed based on clinical guidelines, individualized treatment principles, and historical case data analysis, and is combined with a reinforcement learning algorithm to continuously optimize the decision rule set of the multi-level decision tree;

[0048] The notification module is used to dynamically adjust the infusion device according to the infusion adjustment instruction to ensure the safety and effectiveness of the infusion process, and when an abnormal situation is detected, the Bayesian network is used to evaluate the risk level, and an alarm of a corresponding level is issued based on the risk level to notify medical staff; at the same time, all multi-dimensional physiological parameters and abnormal information corresponding to the abnormal situation during each infusion process are recorded to form an infusion process log;

[0049] The adjustment module is used to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during the infusion process after each infusion process, based on the infusion process log, using an association rule mining algorithm to identify potential influencing factors and their interactions, and to predictively optimize the decision-making algorithm and infusion control strategy of the patient in the future infusion process based on the influencing factors and their interactions through a machine learning model, so as to adjust the infusion control strategy of the patient.

[0050] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an adaptive infusion control method based on physiological parameters as described in the first aspect above.

[0051] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements an adaptive infusion control method based on physiological parameters as described in the first aspect.

[0052] In the embodiment of the present application, the patient's vital signs and infusion status are monitored in real time to obtain monitoring data, which includes multi-dimensional physiological parameters, current infusion rate, and cumulative infusion volume; based on the monitoring data, a multi-level decision tree algorithm is used to analyze the change trend of the multi-dimensional physiological parameters to determine whether it is necessary to adjust the infusion rate or stop the infusion, and generate an infusion adjustment instruction; the multi-level decision tree is constructed based on clinical guidelines, individualized treatment principles, and historical case data analysis, and the decision rule set of the multi-level decision tree is continuously optimized in combination with a reinforcement learning algorithm; according to the infusion adjustment instruction, the infusion device is dynamically adjusted to ensure the safety and effectiveness of the infusion process, and when abnormalities are detected, the infusion device is automatically adjusted. In normal situations, the Bayesian network is used to evaluate the risk level, and an alarm of the corresponding level is issued based on the risk level to notify medical staff; at the same time, all multidimensional physiological parameters during each infusion process and the abnormal information corresponding to the abnormal situation are recorded to form an infusion process log; after each infusion process, based on the infusion process log, the association rule mining algorithm is used to analyze the relationship between the multidimensional physiological parameters and abnormal events during this infusion, so as to identify potential influencing factors and their interactions, and through a machine learning model, based on the influencing factors and their interactions, the decision-making algorithm and infusion control strategy of the patient in the future infusion process are predictively optimized to adjust the infusion control strategy of the patient.

[0053] The technical solution of this application has the following beneficial effects:

[0054] This application monitors the patient's vital signs and infusion status in real time to ensure timely response to abnormal situations and reduce possible risks during the infusion process. Multi-level decision tree algorithms and reinforcement learning are used to continuously optimize the infusion rate and control strategy to make the infusion process more accurate and efficient. Customize the infusion plan according to the specific situation and historical data of each patient to achieve individualized treatment and improve the treatment effect. Through association rule mining and machine learning models, the relationship between multi-dimensional physiological parameters and abnormal events during the infusion process is analyzed to provide predictive optimization suggestions for future infusion processes and further improve the infusion control strategy. When an abnormality is detected, the Bayesian network is used to assess the risk level to help medical staff respond quickly and accurately, enhancing the emergency response capabilities of the medical team. The entire system continuously improves its performance through continuous learning and optimization to adapt to the needs of different patients and the changing medical environment.

[0055] Furthermore, according to the unique physiological response patterns of patients, the individual differences between different patients are analyzed by learning algorithms, key influencing factors are identified and the dependencies of these factors in time series are captured to obtain identification results; the key influencing factors refer to factors that affect the infusion effect and are used to adjust the specific content of the infusion control strategy. Based on the above identification results, the gradient boosting decision tree and long short-term memory network are used to perform personalized calibration processing on the infusion control parameters in the infusion control strategy, so as to obtain a personalized infusion control strategy that adapts to the immediate needs of patients. This method can not only accurately identify the key factors affecting the infusion effect and their dynamic changes, but also realize the fine adjustment of the infusion control parameters through advanced machine learning technology, ensuring that each patient can obtain the treatment plan that best suits their current physiological state. This move significantly improves the safety and effectiveness of infusion, reduces the treatment deviation caused by individual differences, and enhances the adaptive ability of the system, providing strong technical support for personalized medicine. In addition, by capturing the time series dependencies of key influencing factors, the system can better predict and respond to potential risks, improving the overall treatment effect and patient satisfaction.

[0056] In summary, the present invention not only improves the safety and effectiveness of the infusion process, but also provides strong technical support for achieving personalized medical care.

[0057] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0059] Figure 1 A flowchart of an adaptive infusion control method based on physiological parameters provided in an embodiment of the present application;

[0060] Figure 2 A schematic diagram of the structure of an adaptive infusion control system based on physiological parameters provided in an embodiment of the present application;

[0061] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0063] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0065] Figure 1 A flowchart of an adaptive infusion control method based on physiological parameters is provided for an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0066] 101. Monitor the patient's vital signs and infusion status in real time to obtain monitoring data, wherein the monitoring data includes multi-dimensional physiological parameters, current infusion rate, and cumulative infusion volume;

[0067] In this step, real-time monitoring of the patient's vital signs and infusion status is achieved through sensors and other medical monitoring devices integrated in the infusion equipment. These devices can continuously collect multi-dimensional physiological parameters (such as heart rate, blood pressure, blood oxygen saturation, body temperature, etc.), current infusion rate, and cumulative infusion volume. Multi-dimensional physiological parameters include not only basic vital signs, but also deeper data reflecting metabolic status, organ function, etc., which are used to comprehensively evaluate the patient's health status. Real-time monitoring ensures the immediacy and accuracy of the data, providing a reliable basis for subsequent analysis.

[0068] In the embodiment of the present application, during the implementation process, a series of high-precision sensors are first deployed around the patient, such as an electrocardiogram (ECG) monitor, a pulse oximeter, a blood pressure cuff, etc., and these sensors are connected to a central monitoring system. The system is responsible for collecting and integrating data streams from different sensors, while recording the status information of the infusion device, such as the infusion rate and cumulative volume. All of this data is transmitted to a local or cloud server for processing and storage for further analysis and decision-making.

[0069] For example, in an intensive care unit (ICU) environment, nurses will install a complete set of monitoring equipment at each patient's bedside, including an intelligent infusion pump that can accurately control the infusion rate and automatically record the cumulative amount of infusion. In addition, it will be equipped with a variety of vital signs monitoring instruments, such as non-invasive blood pressure monitors, pulse oximeters, and ECG monitors. All of these devices are connected to the hospital's information management system (HIS) and synchronize data to the cloud server via a wireless network. Medical staff can view patients' real-time data at any time through mobile terminals to ensure that any abnormal situation is responded to in the first time.

[0070] 102. Based on the monitoring data, a multi-level decision tree algorithm is used to analyze the change trend of the multi-dimensional physiological parameters, determine whether it is necessary to adjust the infusion rate or stop the infusion, and generate an infusion adjustment instruction;

[0071] In this step, based on the monitoring data, a multi-level decision tree algorithm is used to analyze the changing trends of multi-dimensional physiological parameters, aiming to identify potential patterns that may lead to adverse consequences, and to determine whether it is necessary to adjust the infusion rate or stop the infusion. A multi-level decision tree is a hierarchical machine learning model that gradually splits data according to a predefined set of rules to eventually form a series of decision paths. The multi-level decision tree used in the present invention is not only constructed based on clinical guidelines and individualized treatment principles, but also combines historical case data analysis and reinforcement learning algorithms to continuously optimize its decision rule set to meet the specific needs of different patients.

[0072] In the embodiment of the present application, after the monitoring data is transmitted to the server, the system will start a multi-level decision tree algorithm to conduct an in-depth analysis of the data. First, the algorithm preliminarily screens out key parameters based on preset clinical guidelines and individualized treatment principles; then, by comparing historical case data, similar cases and their corresponding optimal treatment methods are identified; finally, the decision rules are dynamically adjusted in combination with the reinforcement learning algorithm to ensure that each decision can maximize the treatment effect and safety. If it is found that the infusion rate needs to be adjusted or the infusion needs to be stopped, the system will automatically generate the corresponding infusion adjustment instructions.

[0073] Continuing with the I CU example, suppose a heart patient shows an abnormally accelerated heart rate during infusion. After the system receives this change, it immediately starts a multi-level decision tree algorithm for analysis. The algorithm first confirms whether the increased heart rate conforms to the known symptom pattern of heart disease, and then refers to the best practices of similar situations in the historical case library to decide to appropriately reduce the infusion rate. At the same time, the system also considers other relevant factors, such as the patient's blood oxygen level and blood pressure fluctuations, to ensure that the adjusted infusion strategy does not cause new problems. Finally, the system generates an instruction to reduce the infusion rate and sends it to the smart infusion pump for execution.

[0074] 103. According to the infusion adjustment instruction, dynamically adjust the infusion device to ensure the safety and effectiveness of the infusion process, and when an abnormal situation is detected, use the Bayesian network to evaluate the risk level, and issue an alarm of a corresponding level based on the risk level to notify medical staff; at the same time, record all multi-dimensional physiological parameters during each infusion process and the abnormal information corresponding to the abnormal situation to form an infusion process log;

[0075] In this step, the infusion equipment is dynamically adjusted according to the infusion adjustment instructions to ensure the safety and effectiveness of the infusion process. On this basis, when the system detects an abnormal situation, it will use the Bayesian network to evaluate the risk level, and based on this, issue an alarm of the corresponding level to notify medical staff. The Bayesian network is a probabilistic graphical model that simulates complex causal relationships by constructing conditional dependencies between nodes, thereby achieving efficient risk assessment of complex systems. All multi-dimensional physiological parameters and abnormal information during each infusion process will be recorded in detail to form an infusion process log for subsequent analysis.

[0076] In an embodiment of the present application, once an infusion adjustment instruction is generated, the system will immediately send it to the infusion device, which will then adjust the infusion rate or take other necessary measures. At the same time, the system continuously monitors the patient's vital signs, and once an abnormality (such as arrhythmia, hypotension, etc.) is detected, the Bayesian network is immediately started for risk assessment. The evaluation results will determine the risk level and trigger an alarm mechanism of the corresponding level, such as sending text messages or email reminders to medical staff. All monitoring data and abnormal information will be fully recorded to form a detailed infusion process log for future review and analysis.

[0077] In the same I CU scenario, assume that the smart infusion pump quickly adjusted the infusion rate after receiving the instruction to reduce the infusion rate. However, in the next few minutes, the system detected that the patient's heart rate still did not return to normal, but instead there was a slight drop in blood pressure. At this time, the system immediately started the Bayesian network for risk assessment, and the results showed that there was a moderate risk, so the second-level alarm was triggered and an emergency notification was sent to the doctor on duty. After receiving the notification, the doctor rushed to the scene and adjusted the treatment plan in time according to the detailed data and suggestions provided by the system, successfully stabilizing the patient's condition. All data of the entire process are recorded in detail in the infusion process log for subsequent analysis and improvement.

[0078] 104. After each infusion process is completed, based on the infusion process log, an association rule mining algorithm is used to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during this infusion period to identify potential influencing factors and their interactions, and through a machine learning model, based on the influencing factors and their interactions, the patient's decision algorithm and infusion control strategy in the future infusion process are predictively optimized to adjust the patient's infusion control strategy.

[0079] In this step, after each infusion process, based on the infusion process log, an association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during this infusion to identify potential influencing factors and their interactions. Association rule mining is a data mining technology that aims to find frequently occurring patterns and associations from a large amount of data to help understand the intrinsic connection between variables. Through machine learning models, especially those that can capture time series characteristics (such as long short-term memory networks), decision algorithms and infusion control strategies in future infusion processes can be predictively optimized to better meet the immediate needs of patients.

[0080] In an embodiment of the present application, after the infusion is completed, the system will automatically call the association rule mining algorithm to conduct a comprehensive analysis of the infusion process log. The algorithm first identifies which changes in multi-dimensional physiological parameters are closely related to the occurrence of abnormal events, and then reveals potential influencing factors and their interactions. Subsequently, the system uses machine learning models (such as gradient boosting decision trees and long short-term memory networks) to perform predictive modeling of future infusion processes based on these influencing factors. The model not only takes into account the impact of static parameters, but also pays special attention to dynamic changes in time series to provide a more accurate personalized infusion control strategy. Finally, the system will generate a detailed report containing optimized infusion control strategy recommendations for reference and implementation by medical staff.

[0081] Continuing with the previous I CU case, let's assume that a patient has completed a week-long infusion treatment. After the treatment, the system automatically starts the association rule mining algorithm to analyze all the data during this period, and finds the complex associations between heart rate, blood pressure, and blood oxygen saturation, especially during the night time, when changes in certain parameters are more likely to cause abnormal events. Based on these findings, the system uses gradient boosting decision trees and long short-term memory networks to build a prediction model to predict the risks that may occur in the next few days and make personalized recommendations for infusion control strategies. For example, it is recommended to appropriately reduce the infusion rate during the night time and closely monitor changes in specific parameters. Based on these suggestions, the medical staff adjusted the subsequent treatment plan, significantly reducing the frequency of abnormal events and improving the overall treatment effect. All these analysis results and optimization suggestions are recorded in detail, providing valuable experience and guidance for future treatment.

[0082] Optionally, the method also includes: targeting the patient's unique physiological response pattern, based on the adjusted infusion control strategy, analyzing the individual differences between different patients through a learning algorithm, identifying key influencing factors and capturing the dependencies of the key influencing factors in time series to obtain an identification result; based on the identification result, using a gradient boosting decision tree and a long short-term memory network to perform personalized calibration processing on the infusion control parameters in the infusion control strategy to obtain a personalized infusion control strategy that meets the patient's immediate needs.

[0083] This method further expands the personalized treatment of patients' unique physiological response patterns, and deeply analyzes the individual differences between different patients through learning algorithms. Key influencing factors refer to those factors that significantly affect the effect of infusion, such as the patient's age, weight, underlying disease conditions, drug sensitivity, etc. These factors include not only static characteristics (such as demographic information), but also dynamically changing data (such as real-time vital signs and metabolic status). By capturing the time series dependencies of these key influencing factors, the system can more accurately understand the special needs of each patient during the infusion process.

[0084] Dependencies in time series refer to the trends of key influencing factors over time and their correlations. For example, changes in heart rate may be related to fluctuations in blood pressure, and these changes may be affected by the infusion rate. Identifying these dependencies helps predict future physiological responses and optimize infusion control strategies accordingly.

[0085] In the embodiment of the present application, first, the system extracts key influencing factors from the multidimensional physiological parameters accumulated during each infusion process. These data are cleaned and standardized for subsequent analysis. Machine learning algorithms (such as cluster analysis or principal component analysis) are used to compare the key influencing factors between different patients, identify patient groups with similar characteristics, and highlight the uniqueness of each patient. Time series analysis techniques (such as autoregressive integral moving average model ARIMA or long short-term memory network LSTM) are used to explore the changing trends of key influencing factors over time and their interactions. Based on the above analysis results, gradient boosting decision tree (GBDT) and long short-term memory network (LSTM) are used to personalize the infusion control parameters in the infusion control strategy. GBDT is used to process static features and nonlinear relationships, while LSTM focuses on capturing dynamic changes in time series to ensure that the calibrated strategy can adapt to the immediate needs of patients.

[0086] In the intensive care unit (ICU) environment of a general hospital, the application of this optional solution can significantly improve the accuracy and safety of infusion therapy. The specific implementation steps are as follows:

[0087] First, for every patient admitted to the ICU, the system automatically collects and records multi-dimensional physiological parameters such as vital signs (such as heart rate, blood pressure, blood oxygen saturation), current infusion rate and cumulative volume. All data are transmitted to the central server for cleaning and standardization to ensure data quality and consistency.

[0088] Second, the system uses cluster analysis to divide patients into several groups, each with similar underlying health conditions and physiological characteristics. Then, for each patient in each group, the system further analyzes their unique characteristics, especially those key influencing factors that may lead to differences in infusion effects, such as tolerance to specific drugs or the presence of certain complications.

[0089] Next, taking an elderly patient with chronic kidney disease as an example, the system found that his heart rate and blood pressure fluctuated greatly during the night, and these fluctuations were significantly correlated with the infusion rate. By analyzing these time series data with the LSTM model, the system identified the potential patterns of heart rate and blood pressure changes and determined the complex dependencies between them and the infusion rate.

[0090] Furthermore, based on the above analysis results, the system used GBDT and LSTM to calibrate the infusion control strategy for this patient. GBDT adjusted the initial infusion rate according to the patient's static characteristics (such as age, weight, and underlying diseases), while LSTM appropriately reduced the infusion rate and increased the monitoring frequency during the night period based on the dynamic changes in the time series. In addition, the system also recommends that medical staff pay special attention to changes in heart rate and blood pressure during the night period and take necessary intervention measures in a timely manner.

[0091] Finally, after a period of observation, the medical staff found that the patient's nighttime heart rate and blood pressure fluctuations were significantly reduced, and the overall treatment effect was significantly improved. All data and analysis results of the entire process were recorded in detail, providing valuable reference and guidance for similar cases in the future.

[0092] Through this personalized infusion control strategy, the system not only improves the safety and effectiveness of treatment, but also enhances the work efficiency of medical staff and reduces unnecessary waste of medical resources. More importantly, it provides each patient with more intimate and scientific medical services, reflecting the humanistic care of modern medical technology.

[0093] Optionally, in step 104, based on the infusion process log, an association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during the infusion period to identify potential influencing factors and their interactions, including:

[0094] Based on the infusion process log, an association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events. The association rule mining algorithm is intended to be used to discover that a specific physiological parameter change pattern or combination is associated with the occurrence of abnormal events, and is also used to discover the interactions between physiological parameters of different dimensions, and generate a preliminary list of influencing factors and their interactions; the preliminary list is further screened and refined to identify and list key influencing factors and their interactions, and to determine a final list of influencing factors and their interactions.

[0095] In this scheme, the association rule mining algorithm is a data mining technique that aims to discover frequently occurring patterns and associations from a large amount of data to help understand the intrinsic connection between variables. In this step, the algorithm is applied to the infusion process log to analyze the relationship between multi-dimensional physiological parameters and abnormal events.

[0096] Specifically:

[0097] Multi-dimensional physiological parameters include real-time monitoring of vital signs such as heart rate, blood pressure, blood oxygen saturation, body temperature, as well as metabolic status and other data reflecting the patient's health status.

[0098] Abnormal events refer to any deviation from the normal range or changes in the combination pattern of multi-dimensional physiological parameters detected during the infusion process, such as arrhythmia, hypotension, etc.

[0099] Influencing factors and their interactions refer to those factors that significantly affect the effect of infusion (such as the patient's age, weight, underlying disease condition, drug sensitivity, etc.), as well as the complex interactions between these factors.

[0100] The preliminary list is a series of possible influencing factors and their interactions generated by the association rule mining algorithm. This list contains all potential influencing factors, but has not yet been screened and refined.

[0101] The final list is the result of further screening and refinement of the preliminary list. Each entry describes in detail the specific nature of the key influencing factors and how they work together on the infusion process to determine the key causes of abnormal events or important indicators related to the infusion effect.

[0102] In the present application, first, based on the multi-dimensional physiological parameters and abnormal event data in the infusion process log, an association rule mining algorithm is used to identify the association between a specific physiological parameter change pattern or combination and the occurrence of abnormal events. This step not only finds direct associations, but also reveals the interaction between physiological parameters in different dimensions.

[0103] Second, based on the above analysis results, a preliminary list of all potential influencing factors and their interactions was generated. This list covers a wide range of possibilities but has not yet been rigorously verified.

[0104] Next, the preliminary list is analyzed in depth to remove irrelevant or redundant influencing factors and retain the truly critical influencing factors and their interactions. This process may involve statistical testing, expert review or other verification methods to ensure the accuracy and reliability of the final list.

[0105] Finally, after screening and refinement, a final list was determined, in which each item described in detail the specific nature of the influencing factors and how they work together on the infusion process. This information provides a solid foundation for the subsequent optimization of infusion control strategies.

[0106] Through the above steps, the system can not only accurately identify the key factors affecting the infusion effect and their interactions, but also provide personalized infusion control strategies for each patient, thereby achieving more intelligent, safe and efficient medical care.

[0107] Optionally, in step 104, predictive optimization of the decision algorithm and infusion control strategy of the patient in the future infusion process is performed based on the influencing factors and their interactions through a machine learning model to adjust the infusion control strategy of the patient, including:

[0108] According to the final list, an appropriate machine learning model is selected, and the multi-dimensional physiological parameters and abnormal events contained in the patient's historical infusion process log are used as training samples to obtain a trained machine learning model; based on the trained machine learning model, the predicted abnormal events of the patient in the future infusion process are obtained, and on the basis of the predicted abnormal events, according to the latest monitoring data of the patient obtained, the predicted risk of the infusion process in the future period of time is determined, and a prediction result is generated; using the prediction result, the infusion control strategy of the patient is dynamically adjusted.

[0109] In this scenario, a machine learning model is a type of algorithm that is trained with data to make predictions or decisions, and is able to learn patterns from historical data and apply them to new data. In this step, the appropriate machine learning model is selected to predictively optimize the patient's future infusion process based on the influencing factors and their interactions in the final list.

[0110] The latest monitoring data refers to the multi-dimensional physiological parameters collected at the current moment, which are used to assess risks in the future.

[0111] The prediction results are based on the trained machine learning model to evaluate the predicted risks that may occur during future infusions. These prediction results will guide the dynamic adjustment of infusion control strategies to ensure the safety and effectiveness of treatment.

[0112] In the embodiment of this application, in an intensive care unit (ICU) environment of a general hospital, it is assumed that a heart patient has completed a week of infusion treatment. After the treatment, the system starts the machine learning model to predictively optimize the future infusion process. The specific implementation steps are as follows:

[0113] First, the system selected the long short-term memory network (LSTM) as the machine learning model based on the key influencing factors in the final list (such as the combination of increased heart rate and decreased blood oxygen saturation), because LSTM is good at processing time series data and can capture complex dynamic changes. Then, the system used the multi-dimensional physiological parameters (such as heart rate, blood pressure, blood oxygen saturation) and abnormal events (such as arrhythmia and hypotension) in the patient's infusion process log over the past week as training samples to train a high-precision LSTM model.

[0114] After training, the system uses the LSTM model to predict specific abnormal events that may occur to the patient in the next few days. For example, the model predicts that if the current infusion rate continues, the patient's heart rate may further increase, accompanied by a significant decrease in blood oxygen saturation, leading to the occurrence of arrhythmia.

[0115] Based on the above predicted abnormal events, the system combines the patient's latest monitoring data (such as current heart rate and blood oxygen saturation) to assess the potential risks during the infusion process in the future. The system found that if the heart rate continues to rise and the blood oxygen saturation drops below 95%, the patient's risk of arrhythmia will increase significantly. Therefore, the system generates a detailed prediction result report, indicating that there is a moderate risk in the next three days and recommends taking preventive measures.

[0116] Based on the prediction results, the system dynamically adjusted the patient's infusion control strategy. Specifically, it was recommended to appropriately reduce the infusion rate and increase the monitoring frequency during the night, especially paying close attention to changes in heart rate and blood oxygen saturation. In addition, the system also proposed other auxiliary measures, such as preparing emergency drugs and equipment, so as to respond quickly when necessary.

[0117] Finally, after a period of observation, the medical staff found that the patient's nighttime heart rate and blood oxygen saturation fluctuations were significantly reduced, and the overall treatment effect was significantly improved. All these analysis results and optimization suggestions were recorded in detail, providing valuable experience and guidance for future treatment. In this way, the system not only improves the safety and effectiveness of treatment, but also enhances the quality of personalized medical services.

[0118] Through the above steps, the system can not only accurately identify the key factors affecting the infusion effect and their interactions, but also provide personalized infusion control strategies for each patient, thereby achieving more intelligent, safe and efficient medical care. This method not only reduces the frequency of abnormal events, but also improves the overall treatment effect, reflecting the humanistic care of modern medical technology.

[0119] Optionally, in step 102, based on the monitoring data, using a multi-level decision tree algorithm to analyze the change trend of the multi-dimensional physiological parameters to determine whether it is necessary to adjust the infusion rate or stop the infusion includes:

[0120] According to clinical guidelines, individualized treatment principles and historical case data analysis, a multi-level decision tree model is constructed, and the multi-level decision tree model is used to analyze the changing trends of the multi-dimensional physiological parameters and determine whether it is necessary to adjust the infusion rate or stop the infusion; the monitoring data is input into the multi-level decision tree model, and the changing trends and combination patterns of the physiological parameters of each dimension in the monitoring data are evaluated layer by layer to generate analysis results of the changing trends of the multi-dimensional physiological parameters; based on the analysis results, it is determined whether the patient is currently at risk, and if so, the infusion rate is adjusted or the infusion is stopped according to the risk situation.

[0121] In this scheme, the multi-level decision tree model is a hierarchical machine learning algorithm that gradually splits the data according to preset rules and eventually forms a series of decision paths. In this step, the multi-level decision tree model is used to analyze the changing trends of multi-dimensional physiological parameters and determine whether it is necessary to adjust the infusion rate or stop the infusion.

[0122] Clinical guidelines: Refers to recognized treatment standards and recommendations in the medical field, such as guidelines issued by international cardiology societies.

[0123] Principle of individualized treatment: Consider the specific conditions of each patient (such as age, weight, underlying diseases, etc.) and develop a personalized treatment plan.

[0124] Historical case data analysis: Utilize monitoring data and treatment results from past cases to optimize the rule set of the decision tree model and improve its accuracy and applicability.

[0125] Change trends and their combination patterns: refers to the change trends of various physiological parameters over time and the interaction patterns between them, such as an increase in heart rate accompanied by a decrease in blood oxygen saturation.

[0126] Risk assessment is based on the analysis results of a multi-level decision tree model to determine whether the current patient is at risk and decide whether to adjust the infusion rate or stop the infusion based on the risk situation.

[0127] In the present application, first, a multi-level decision tree model is constructed based on clinical guidelines, individualized treatment principles, and historical case data analysis. The model can analyze the changing trends and combination patterns of multi-dimensional physiological parameters and identify potential patterns that may lead to adverse consequences.

[0128] Secondly, the multi-dimensional physiological parameters monitored in real time are input into the multi-level decision tree model to evaluate the change trend and combination pattern of each dimension of physiological parameters layer by layer to generate detailed analysis results. This process not only considers the changes of a single parameter, but also pays special attention to the interaction between different parameters.

[0129] Then, through the evaluation of the multi-level decision tree model, the system generates a detailed analysis report describing the changing trends of various physiological parameters and their combination patterns. This report provides an important basis for subsequent risk assessment.

[0130] Finally, based on the above analysis results, the system evaluates whether the current patient is at risk. If there is a risk, the system will decide whether to adjust the infusion rate or stop the infusion based on the specific situation (such as the risk level) and generate corresponding instructions. In addition, the system will record all evaluation and decision-making processes for future review and improvement.

[0131] Through the above steps, the system can not only accurately identify the key factors affecting the infusion effect and their interactions, but also provide personalized infusion control strategies for each patient, thereby achieving more intelligent, safe and efficient medical care. This method not only reduces the frequency of abnormal events, but also improves the overall treatment effect, reflecting the humanistic care of modern medical technology.

[0132] Optionally, when an abnormal situation is detected in step 103, the risk level is evaluated using a Bayesian network, and an alarm of a corresponding level is issued based on the risk level to notify medical staff, including:

[0133] When an abnormal situation is monitored, the abnormal situation and its corresponding multi-dimensional physiological parameters are input into the activated Bayesian network model, so that the Bayesian network model can dynamically update the posterior probability distribution of the abnormal event corresponding to the abnormal situation according to the interaction between physiological parameters in different dimensions and the prior knowledge learned from historical data; the prior knowledge refers to the risk level of a specific combination of physiological parameters in historical cases; the probability of risk occurring in the current state is evaluated according to the posterior probability distribution, and the risk level is divided according to the evaluation result; based on the risk level, an alarm of the corresponding level is issued to medical staff.

[0134] In this scheme, the Bayesian network model is a probabilistic graphical model that simulates complex causal relationships by building conditional dependencies between nodes. In this step, the activated Bayesian network model is used to assess the risk level of abnormal situations and issue corresponding levels of alarms based on the risk level.

[0135] Prior knowledge: refers to the risk level of a specific combination of physiological parameters in historical cases. This information comes from learning and statistical analysis of past case data.

[0136] Posterior probability distribution: Based on the currently monitored abnormal conditions and their corresponding multi-dimensional physiological parameters, combined with the prior knowledge in historical data, the probability distribution of abnormal events is dynamically updated.

[0137] Risk level classification: Evaluate the probability of risk occurring in the current state based on the posterior probability distribution, and classify the risk level (such as low, medium, and high) accordingly to guide subsequent treatment measures.

[0138] In the embodiment of the present application, first, when the monitoring system detects an abnormal situation, the pre-trained Bayesian network model is immediately activated. The model has learned the prior knowledge extracted from historical data and can understand the interaction between physiological parameters in different dimensions.

[0139] Secondly, the detected abnormalities and their corresponding multi-dimensional physiological parameters are input into the activated Bayesian network model. The model dynamically updates the posterior probability distribution of abnormal events based on these data, taking into account the complex interactions between different parameters.

[0140] Next, the probability of risk occurring in the current state is evaluated based on the updated posterior probability distribution. Then, based on the evaluation results, the risks are divided into different levels (such as low, medium, and high). This risk assessment not only takes into account the changes in a single parameter, but also pays special attention to the interaction between different parameters to ensure the comprehensiveness and accuracy of the assessment.

[0141] Finally, based on the determined risk level, the system issues a corresponding level of alert to medical staff. For example, a low risk may trigger a yellow alert to alert medical staff to pay attention; a medium risk may trigger an orange alert to recommend preventive measures; and a high risk may trigger a red alert to instruct immediate emergency action.

[0142] Through the above steps, the system can not only accurately assess the risk level in abnormal situations, but also provide personalized early warning and intervention measures for each patient, thereby achieving more intelligent, safe and efficient medical care. This method not only reduces the frequency of abnormal events, but also improves the overall treatment effect, reflecting the humanistic care of modern medical technology.

[0143] Optionally, selecting an appropriate machine learning model according to the final list, and using the multi-dimensional physiological parameters and abnormal events contained in the patient's historical infusion process log as training samples to obtain a trained machine learning model includes:

[0144] The trained machine learning model is obtained through the following calculation formula:

[0145] M=f(X train ,Y train ;θ)

[0146] Where M represents the trained machine learning model; X train Y is a training set composed of multi-dimensional physiological parameters extracted from historical nighttime logs; train is the corresponding historical abnormal event label; θ is the parameter set of the model, including weights and bias terms.

[0147] In the embodiment of the present application, assuming that a deep neural network (DNN) model is used, the trained machine learning model can be expressed as:

[0148] M=σ(W L ·σ(W L-1 ·…·σ(W1·X train +b1)+b L-1 )+b L )

[0149] Among them, M represents the trained machine learning model; X train Y is a training set composed of multi-dimensional physiological parameters extracted from historical nighttime logs; train is the corresponding historical abnormal event label, W i and b i are the weight matrix and bias vector of the i-th layer respectively; σ(·) is the activation function, such as ReLU or Sigmoid; L is the number of layers of the neural network.

[0150] To better understand the various parameters in the above formula, the following are detailed annotations:

[0151] M: represents the final trained machine learning model. This model can receive new multi-dimensional physiological parameter inputs and output corresponding prediction results (such as whether an abnormal event will occur). In this example, it is assumed that a deep neural network (DNN) model is used.

[0152] Training set X train :It is a training set composed of multi-dimensional physiological parameters extracted from historical infusion process logs. Each sample is usually a vector containing multiple features (such as heart rate, blood pressure, blood oxygen saturation, etc.). These features are used to describe the patient's physiological state at a certain point in time.

[0153] Feature Examples:

[0154] Heart rate (HR)

[0155] Blood pressure (BP)

[0156] Blood oxygen saturation (SpO2)

[0157] Body temperature(Temp)

[0158] Infusion rate

[0159] Cumulative infusion volume

[0160] Other metabolic or organ function indicators

[0161] Historical abnormal event label Y train : is the corresponding historical abnormal event label. Each label is usually a binary variable (0 or 1), indicating whether an abnormal event (such as arrhythmia, hypotension, etc.) has occurred under the corresponding multi-dimensional physiological parameters. Labels are used to supervise the learning of the model to help it understand which feature combinations may lead to abnormal events.

[0162] Tag example:

[0163] 0: No abnormal event occurred

[0164] 1: Abnormal events (such as arrhythmia) occur

[0165] Model parameter set θ: represents the set of all parameters of the model, including weights and bias terms. These parameters are continuously adjusted during the training process through optimization algorithms (such as gradient descent) to minimize the prediction error of the model.

[0166] Weight matrix W i : The weight matrix of the i-th layer, which represents the connection strength between the neurons in this layer and the neurons in the previous layer. Each element w ij Represents the weight from the jth previous layer neuron to the i-th current layer neuron.

[0167] Bias vector b i : The bias vector of the i-th layer, which represents the bias term of each layer of neurons. The bias term is used to adjust the threshold of the activation function so that the model can better fit the data.

[0168] Deep Neural Network (DNN) Model Expression

[0169] For the deep neural network model, the trained machine learning model M can be expressed as:

[0170] M=σ(W L ·σ(W L-1 ·…·σ(W1·Xtrain +b1)+b L-1 )+b L )

[0171] σ(·): Activation function, used to introduce nonlinearity so that the model can capture complex patterns. Common activation functions include ReLU (Rectified Linear Unit), Sigmoid, Tanh, etc.

[0172] ReLU:σ(x)=max(0,x)

[0173] Sigmoid:

[0174] Tanh:σ(x)=tanh(x)

[0175] W i : The weight matrix of the i-th layer. The weight matrix of each layer determines the connection strength between the neurons in this layer and the neurons in the previous layer. i : The bias vector of the i-th layer. Each layer bias vector is used to adjust the threshold of the activation function. L: The number of layers in the neural network. The more layers there are, the more complex the model is, but it also requires more training data and computing resources to avoid overfitting.

[0176] The method comprises: obtaining predicted abnormal events of the patient during the future infusion process based on the trained machine learning model, and determining the predicted risk of the patient during the infusion process within a future period of time based on the predicted abnormal events and the latest monitoring data of the patient obtained, and generating a prediction result, including:

[0177] The trained machine learning model is applied to simulated monitoring data in the future to predict abnormal events that may occur in the future. The definition of predicting abnormal events is:

[0178]

[0179] in, represents the predicted abnormal event at the future time point t+Δt; X sim (t+Δt) is the multi-dimensional physiological parameter input in the simulated future period; θ is the trained model parameter;

[0180] In the embodiment of the present application, assuming that Bayesian optimization is used to tune the model hyperparameters, the predicted abnormal event can be expressed as:

[0181]

[0182] in, represents the predicted abnormal event at the future time point t+Δt; X sim(t+Δt) is the multi-dimensional physiological parameter input in the simulated future period of time, M opt is the best model after Bayesian optimization; θ opt are the optimized model parameters.

[0183] To better understand the various parameters in the above formula, the following are detailed annotations:

[0184] Predicting unusual events Represents the predicted abnormal event at the future time point t+Δt. This value is a binary variable (0 or 1) that indicates whether an abnormal event (such as arrhythmia, hypotension, etc.) will occur at that time point.

[0185] 0: No abnormal event occurred

[0186] 1: An abnormal event occurred

[0187] The multi-dimensional physiological parameter input X in the simulated future period sim (t+Δt): is the multi-dimensional physiological parameter input for a simulated period of time in the future. Each sample is usually a vector containing multiple features (such as heart rate, blood pressure, blood oxygen saturation, etc.). These features are used to describe the patient's expected physiological state at a certain point in the future.

[0188] Feature Examples:

[0189] Heart rate (HR)

[0190] Blood pressure (BP)

[0191] Blood oxygen saturation (SpO2)

[0192] Body temperature(Temp)

[0193] Transmission rate

[0194] Cumulative infusion volume

[0195] Other metabolic or organ function indicators

[0196] Trained model parameters θ: represents the set of all parameters of the trained machine learning model, including weights and bias terms. These parameters are continuously adjusted during the training process through optimization algorithms (such as gradient descent) to minimize the prediction error of the model.

[0197] Weight matrix W i : The weight matrix of the i-th layer, which represents the connection strength between the neurons in this layer and the neurons in the previous layer.

[0198] Bias vector b i : The bias vector of the i-th layer, representing the bias term of each layer of neurons.

[0199] The best model M after tuning using Bayesian optimization opt :The best machine learning model after Bayesian optimization. Bayesian optimization is an efficient hyperparameter optimization method that can find the optimal hyperparameter combination in a small number of iterations, thereby improving the performance of the model.

[0200] The optimized model parameters θ opt : represents the best set of model parameters obtained after Bayesian optimization. These parameters further optimize the model's predictive ability, enabling it to more accurately predict future abnormal events.

[0201] For the unoptimized model, the predicted abnormal events can be expressed as:

[0202]

[0203] For the model after Bayesian optimization, the prediction of abnormal events can be expressed as:

[0204]

[0205] Suppose a heart patient is receiving infusion therapy, and the system has trained a deep neural network (DNN) model to predict abnormal events that may occur in the future. The specific implementation steps are as follows:

[0206] The system applies the trained DNN model to simulated monitoring data in the future to predict abnormal events that may occur in the future. For example, at the future time point t+Δt, the patient's physiological parameters are expected to be:

[0207] Heart rate: 80 bpm

[0208] Blood pressure: 125 / 80 mmHg

[0209] Blood oxygen saturation: 97%

[0210] Body temperature: 36.6℃

[0211] Infusion rate: 105mL / h

[0212] Cumulative infusion volume: 600mL

[0213] The system uses these parameters as input X sim (t+Δt), and use the trained model M to make predictions:

[0214]

[0215] If the prediction result is 1, it means that the system predicts that an abnormal event may occur at the future time point t+Δt.

[0216] In order to further improve the prediction accuracy of the model, the system uses the Bayesian optimization method to tune the model's hyperparameters. opt and its parameter θ opt It is more adaptable to the specific situation of the current patient and improves the reliability of the prediction.

[0217]

[0218] In this way, the system not only improves the accuracy of predictions of future abnormal events, but also enhances the quality of personalized medical services, ensuring that each patient receives the treatment plan that best suits his or her current physiological state.

[0219] Based on the predicted abnormal event, according to the latest monitoring data of the patient obtained, the predicted risk in the infusion process in the future period is determined, and an assessment result of the predicted risk is generated as a prediction result; wherein the assessment result of the predicted risk is defined as:

[0220]

[0221] Among them, R(t+Δt) represents the evaluation result of the predicted risk at the future time point t+Δt; X hist is a multidimensional physiological parameter in the historical time period; φ is the parameter set of the risk assessment function g(·);

[0222] In the embodiment of the present application, assuming that the risk assessment function is a risk scoring model based on logistic regression, the assessment result of the predicted risk can be expressed as:

[0223]

[0224] Among them, R(t+Δt) represents the evaluation result of the predicted risk at the future time point t+Δt; X hist is the multidimensional physiological parameter in the historical time period; w and v are the historical multidimensional physiological parameters x hist and predicting abnormal events The weight vector of ; b is a constant term (bias).

[0225] To better understand the various parameters in the above formula, the following are detailed annotations:

[0226] The evaluation result of the predicted risk R(t+Δt): represents the evaluation result of the predicted risk at the future time point t+Δt. This value is a probability value between 0 and 1, which is used to indicate the risk level of an abnormal event occurring at that time point.

[0227] Close to 0: Low risk

[0228] Close to 1: High risk

[0229] Multidimensional physiological parameters X in historical time periods hist : It is a multi-dimensional physiological parameter in a historical period. Each sample is usually a vector containing multiple features (such as heart rate, blood pressure, blood oxygen saturation, etc.). These features are used to describe the changes in the patient's physiological state over a period of time.

[0230] Feature Examples:

[0231] Heart rate (HR)

[0232] Blood pressure (BP)

[0233] Blood oxygen saturation (SpO2)

[0234] Body temperature(Temp)

[0235] Infusion rate

[0236] Cumulative infusion volume

[0237] Other metabolic or organ function indicators

[0238] Predicting unusual events Represents the predicted abnormal event at the future time point t+Δt. This is a binary variable (0 or 1) that indicates whether an abnormal event (such as arrhythmia, hypotension, etc.) will occur at that time point.

[0239] 0: No abnormal event occurred

[0240] 1: An abnormal event occurred

[0241] Risk assessment function g(·): is a function used to assess the predicted risk. In this example, it is assumed that the risk assessment function is a risk scoring model based on logistic regression.

[0242] Parameter set φ: represents the parameter set of the risk assessment function g(·), including the weight vectors w and v, and the bias term b.

[0243] w: is the historical multidimensional physiological parameter X hist The weight vector of . Each element w i Indicates the influence of the i-th historical physiological parameter on risk assessment.

[0244] v: is to predict abnormal events The weight vector of . Each element v i Indicates the impact of the i-th predicted abnormal event on risk assessment.

[0245] Bias term b: is a constant term (bias) used to adjust the threshold of the risk assessment function so that the model can better fit the data.

[0246] For the risk scoring model based on logistic regression, the evaluation result of the predicted risk can be expressed as:

[0247]

[0248] Where: T x hist :Historical multi-dimensional physiological parameters X hist The linear combination of represents the impact of historical data on risk. Predicting unusual events The linear combination of represents the impact of predicted abnormal events on risk. b: Bias term, used to adjust the threshold of risk assessment.

[0249] Suppose there is a heart patient who is undergoing infusion therapy, and the system has predicted abnormal events that may occur in the future. The specific implementation steps are as follows:

[0250] Multidimensional physiological parameters X in historical time periods hist :The system collects multi-dimensional physiological parameters of the patient over a period of time, such as:

[0251] Heart rate: 78 bpm

[0252] Blood pressure: 120 / 80 mmHg

[0253] Blood oxygen saturation: 98%

[0254] Body temperature: 36.5℃

[0255] Infusion rate: 100mL / h

[0256] Cumulative infusion volume: 500mL

[0257] Predicting future abnormal events

[0258] Predicting unusual events The system uses a trained machine learning model to predict abnormal events at a future time point t+Δt, for example, predicting that an increased heart rate accompanied by a decrease in blood oxygen saturation may lead to arrhythmia.

[0259] The system is based on historical multi-dimensional physiological parameters X hist and predicting abnormal events Apply the logistic regression model to assess the risk at the future time point t+Δt:

[0260]

[0261] If R(t+Δt) is close to 1, it means that there is a high risk of an abnormal event occurring at time t+Δt in the future.

[0262] If R(t+Δt) is close to 0, it means the risk is low.

[0263] In this way, the system not only improves the accuracy of predictions of future abnormal events, but also enhances the quality of personalized medical services, ensuring that each patient receives the treatment plan that best suits their current physiological state.

[0264] The method of dynamically adjusting the infusion control strategy of the patient by using the prediction result includes:

[0265] Based on the evaluation result of the predicted risk, the infusion control strategy of the patient is dynamically adjusted to ensure the safety and effectiveness of the infusion process; wherein the adjusted infusion control strategy is defined as:

[0266] C'=h(C,R(t+Δt);ω)

[0267] Among them, C' represents the adjusted infusion control strategy; C is the current infusion control parameter vector; ω is the parameter set of the adjustment function h(·).

[0268] In the embodiment of the present application, assuming that the adjustment function is a proportional adjustment based on the risk level, the adjustment formula is:

[0269] C' k =C k ×(1+α×R k (t+Δt))

[0270] Among them, C' k is the adjusted kth infusion control parameter; C k is the current kth infusion control parameter; α is a proportional factor used to adjust the impact of risk on the control parameter; R k (t+Δt) is an estimate of the predicted risk specific to that control parameter.

[0271] To better understand the various parameters in the above formula, the following are detailed annotations:

[0272] Adjusted infusion control strategy C': represents the adjusted infusion control strategy. This is a vector containing multiple control parameters (such as infusion rate, infusion volume, etc.) to guide how the infusion device should be adjusted to ensure the safety and effectiveness of the treatment.

[0273] Current infusion control parameter vector C: represents the current infusion control parameter vector. Each element Ck represents the kth specific infusion control parameter, for example:

[0274] Infusion rate

[0275] Infusion volume

[0276] Monitoring frequency

[0277] Other related control parameters

[0278] The evaluation result of the predicted risk R(t+Δt): represents the evaluation result of the predicted risk at the future time point t+Δt. This value is a probability value between 0 and 1, which is used to indicate the risk level of an abnormal event occurring at that time point.

[0279] Close to 0: Low risk

[0280] Close to 1: High risk

[0281] Adjustment function h(·): is a function used to adjust the infusion control strategy. In this example, it is assumed that the adjustment function is a proportional adjustment based on the risk level.

[0282] Parameter set ω: represents the parameter set of the adjustment function h(·), including the scale factor α, etc.

[0283] Adjusted kth infusion control parameter C' k : represents the adjusted kth infusion control parameter. This is the specific parameter value that is dynamically adjusted according to the evaluation results of the predicted risk.

[0284] The current kth input control parameter C k : represents the current kth infusion control parameter. Each element C k Is a specific control parameter, such as infusion rate or infusion volume.

[0285] Proportional factor α: It is a proportional factor used to adjust the degree of influence of risk on control parameters. A larger α value will make the influence of risk on control parameters more significant; a smaller α value will weaken this influence.

[0286] The estimated risk R specific to the control parameter k (t+Δt): represents the evaluation result of the predicted risk specific to the kth control parameter. This value is a probability value between 0 and 1, which is specifically for the risk evaluation of this control parameter.

[0287] Assume that a heart patient is undergoing transfusion therapy, and the system has predicted abnormal events and their risks that may occur in the future. The specific implementation steps are as follows:

[0288] The system obtains the current infusion control parameters, such as:

[0289] Infusion rate: 100mL / h

[0290] Cumulative infusion volume: 500mL

[0291] Monitoring frequency: once every hour

[0292] Assessing forecast risk

[0293] Predicted risk assessment result R(t+Δt): The system assesses the risk at the future time point t+Δt based on historical multi-dimensional physiological parameters and predicted abnormal events. For example, it predicts that an increased heart rate accompanied by a decrease in blood oxygen saturation may lead to arrhythmia, and the risk assessment result is R(t+Δt)=0.7.

[0294] The system dynamically adjusts the infusion control strategy based on the evaluation results of the predicted risk using the adjustment formula:

[0295] C'k=C k ×(1+α×R k (t+Δt))

[0296] Infusion rate adjustment: Assumption C k is the infusion rate (100 mL / h), the proportionality factor α = 0.5, and the risk assessment result R specific to this control parameter k (t+Δt)=0.7, then the adjusted infusion rate is defined as:

[0297] C' rate =100×(1+0.5×0.7)=100×1.35=135mL / h

[0298] Cumulative infusion volume adjustment: Assumption C k is the cumulative infusion volume (500 mL), the proportionality factor α = 0.5, and the risk assessment result R specific to this control parameter k (t+Δt)=0.7, then the adjusted cumulative infusion volume is defined as:

[0299] C' volume =500×(1+0.5×0.7)=500×1.35=675mL

[0300] Monitoring frequency adjustment: Assume C k is the monitoring frequency (once per hour), the scaling factor α = 0.5, and the risk assessment result R specific to this control parameter k (t+Δt)=0.7, then the adjusted monitoring frequency is defined as:

[0301] C' frequency = Increase to once every half hour

[0302] In this way, the system not only improves the accuracy of predicting future abnormal events, but also enhances the quality of personalized medical services, ensuring that each patient can receive the treatment plan that best suits their current physiological state. In addition, dynamically adjusting the infusion control strategy helps improve the safety and effectiveness of treatment and reduce potential medical risks.

[0303] Figure 2 A schematic diagram of a structure of an adaptive infusion control system based on physiological parameters is provided for an embodiment of the present application, such as Figure 2 As shown, the system includes:

[0304] Monitoring module 21, used to monitor the patient's vital signs and infusion status in real time and obtain monitoring data, the monitoring data including multi-dimensional physiological parameters, current infusion rate and cumulative infusion volume;

[0305] The analysis module 22 is used to analyze the change trend of the multi-dimensional physiological parameters based on the monitoring data using a multi-level decision tree algorithm, determine whether it is necessary to adjust the infusion rate or stop the infusion, and generate an infusion adjustment instruction; the multi-level decision tree is constructed according to clinical guidelines, individualized treatment principles and historical case data analysis, and the decision rule set of the multi-level decision tree is continuously optimized in combination with a reinforcement learning algorithm;

[0306] The notification module 23 is used to dynamically adjust the infusion device according to the infusion adjustment instruction to ensure the safety and effectiveness of the infusion process, and when an abnormal situation is detected, use the Bayesian network to evaluate the risk level, and issue an alarm of a corresponding level based on the risk level to notify medical staff; at the same time, record all multi-dimensional physiological parameters during each infusion process and the abnormal information corresponding to the abnormal situation to form an infusion process log;

[0307] The adjustment module 24 is used to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during the infusion process based on the infusion process log after each infusion process, so as to identify potential influencing factors and their interactions, and to predictively optimize the decision-making algorithm and infusion control strategy of the patient in the future infusion process based on the influencing factors and their interactions through a machine learning model, so as to adjust the infusion control strategy of the patient.

[0308] Figure 2 The adaptive infusion control system based on physiological parameters can be implemented Figure 1The implementation principle and technical effect of the adaptive infusion control method based on physiological parameters described in the illustrated embodiment will not be described in detail. The specific manner in which each module and unit performs operations in the adaptive infusion control system based on physiological parameters in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

[0309] In one possible design, Figure 2 The adaptive infusion control system based on physiological parameters of the illustrated embodiment can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0310] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0311] The processing component 32 is used to perform the above Figure 1 The illustrated embodiment is an adaptive infusion control method based on physiological parameters.

[0312] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0313] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0314] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0315] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0316] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0317] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0318] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is an adaptive infusion control method based on physiological parameters.

[0319] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0320] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0321] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0322] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An adaptive infusion control method based on physiological parameters, characterized in that: include: Monitor the patient's vital signs and infusion status in real time to obtain monitoring data, including multi-dimensional physiological parameters, current infusion rate, and cumulative infusion volume; Based on the monitoring data, a multi-level decision tree algorithm is used to analyze the change trend of the multi-dimensional physiological parameters, determine whether it is necessary to adjust the infusion rate or stop the infusion, and generate an infusion adjustment instruction; the multi-level decision tree is constructed according to clinical guidelines, individualized treatment principles and historical case data analysis, and the decision rule set of the multi-level decision tree is continuously optimized in combination with a reinforcement learning algorithm; According to the infusion adjustment instruction, the infusion equipment is dynamically adjusted to ensure the safety and effectiveness of the infusion process, and when an abnormal situation is detected, the risk level is evaluated using the Bayesian network, and an alarm of a corresponding level is issued based on the risk level to notify medical staff; at the same time, all multi-dimensional physiological parameters during each infusion process and the abnormal information corresponding to the abnormal situation are recorded to form an infusion process log; After each infusion process is completed, based on the infusion process log, an association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during this infusion, so as to identify potential influencing factors and their interactions. Based on the influencing factors and their interactions, a machine learning model is used to predictively optimize the patient's decision algorithm and infusion control strategy in future infusion processes, so as to adjust the patient's infusion control strategy.

2. The method according to claim 1, characterized in that Also includes: Aiming at the unique physiological response pattern of the patient, based on the adjusted infusion control strategy, the individual differences between different patients are analyzed by a learning algorithm, key influencing factors are identified and the dependency of the key influencing factors in time series is captured to obtain an identification result; the key influencing factors refer to factors that affect the infusion effect and are used to adjust the specific content of the infusion control strategy; Based on the recognition result, a gradient boosting decision tree and a long short-term memory network are used to perform personalized calibration processing on the infusion control parameters in the infusion control strategy to obtain a personalized infusion control strategy that meets the immediate needs of the patient.

3. The method according to claim 1, characterized in that Based on the infusion process log, the association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during the infusion period to identify potential influencing factors and their interactions, including: Based on the infusion process log, an association rule mining algorithm is applied to analyze the relationship between the multi-dimensional physiological parameters and abnormal events, wherein the association rule mining algorithm is intended to be used to discover whether a specific physiological parameter change pattern or combination is associated with the occurrence of abnormal events, and also to discover the interaction between physiological parameters in different dimensions, and generate a preliminary list of influencing factors and their interactions; Further screening and refining the preliminary list to identify and list key influencing factors and their interactions, and determine a final list of influencing factors and their interactions; Among them, the key influencing factors are the key causes of abnormal events, or important indicators related to the infusion effect; each entry in the final list describes in detail the specific properties of the influencing factors and the specific process of the influencing factors acting together on the infusion process.

4. The method according to claim 3, characterized in that The predictive optimization of the decision-making algorithm and infusion control strategy of the patient in the future infusion process based on the influencing factors and their interactions by the machine learning model to adjust the infusion control strategy of the patient includes: According to the final list, an appropriate machine learning model is selected, and the multi-dimensional physiological parameters and abnormal events contained in the historical infusion process log of the patient are used as training samples to obtain a trained machine learning model; Based on the trained machine learning model, the predicted abnormal events of the patient during the future infusion process are obtained, and based on the predicted abnormal events, the predicted risk of the patient during the infusion process within a future period of time is determined according to the latest monitoring data of the patient obtained, and a prediction result is generated; The prediction result is used to dynamically adjust the patient's infusion control strategy.

5. The method according to claim 1, characterized in that The method of analyzing the change trend of the multi-dimensional physiological parameters based on the monitoring data using a multi-level decision tree algorithm to determine whether it is necessary to adjust the infusion rate or stop the infusion includes: According to clinical guidelines, individualized treatment principles and historical case data analysis, a multi-level decision tree model is constructed, and the multi-level decision tree model is used to analyze the changing trend of the multi-dimensional physiological parameters to determine whether it is necessary to adjust the infusion rate or stop the infusion; Inputting the monitoring data into the multi-level decision tree model, evaluating the change trend and combination pattern of the physiological parameters of each dimension in the monitoring data layer by layer, and generating analysis results of the change trend of the multi-dimensional physiological parameters; Based on the analysis results, determine whether the patient is currently at risk, and if so, adjust the infusion rate or stop the infusion based on the risk situation.

6. The method according to claim 1, characterized in that When an abnormal situation is detected, the risk level is evaluated using the Bayesian network, and an alarm of a corresponding level is issued based on the risk level to notify medical staff, including: When an abnormal situation is detected, the abnormal situation and its corresponding multi-dimensional physiological parameters are input into the activated Bayesian network model, so that the Bayesian network model can dynamically update the posterior probability distribution of the abnormal event corresponding to the abnormal situation according to the interaction between physiological parameters in different dimensions and the prior knowledge learned from historical data; the prior knowledge refers to the risk level of a specific combination of physiological parameters in historical cases; Evaluate the probability of risk occurring in the current state according to the posterior probability distribution, and classify the risk level according to the evaluation result; Based on the risk level, an alert of the corresponding level is issued to medical staff.

7. The method according to claim 4, characterized in that The method of selecting an appropriate machine learning model according to the final list and using the multi-dimensional physiological parameters and abnormal events contained in the patient's historical infusion process log as training samples to obtain a trained machine learning model includes: The trained machine learning model is obtained through the following calculation formula: M=f(X train ,Y train ;θ) Where M represents the trained machine learning model; X train Y is a training set composed of multi-dimensional physiological parameters extracted from historical nighttime logs; train is the corresponding historical abnormal event label; θ is the parameter set of the model, including weights and bias terms; The method comprises: obtaining predicted abnormal events of the patient during the future infusion process based on the trained machine learning model, and determining the predicted risk of the patient during the infusion process within a future period of time based on the predicted abnormal events and the latest monitoring data of the patient obtained, and generating a prediction result, including: The trained machine learning model is applied to simulated monitoring data in the future to predict abnormal events that may occur in the future. The definition of predicting abnormal events is: in, represents the predicted abnormal event at the future time point t+Δt; X sim (t+Δt) is the multi-dimensional physiological parameter input in the simulated future period; θ is the trained model parameter; Based on the predicted abnormal event, according to the latest monitoring data of the patient obtained, the predicted risk in the infusion process in the future period is determined, and an assessment result of the predicted risk is generated as a prediction result; wherein the assessment result of the predicted risk is defined as: Among them, R(t+Δt) represents the evaluation result of the predicted risk at the future time point t+Δt; X hist is a multidimensional physiological parameter in the historical time period; φ is the parameter set of the risk assessment function g(·); The method of dynamically adjusting the infusion control strategy of the patient by using the prediction result includes: Based on the evaluation result of the predicted risk, the infusion control strategy of the patient is dynamically adjusted to ensure the safety and effectiveness of the infusion process; wherein the adjusted infusion control strategy is defined as: C'=h(C,R(t+Δt);ω) Among them, C' represents the adjusted infusion control strategy; C is the current infusion control parameter vector; ω is the parameter set of the adjustment function h(·).

8. An adaptive infusion control system based on physiological parameters, characterized in that: include: A monitoring module is used to monitor the patient's vital signs and infusion status in real time and obtain monitoring data, which includes multi-dimensional physiological parameters, current infusion rate, and cumulative infusion volume; An analysis module, for analyzing the change trend of the multi-dimensional physiological parameters based on the monitoring data using a multi-level decision tree algorithm, determining whether it is necessary to adjust the infusion rate or stop the infusion, and generating an infusion adjustment instruction; the multi-level decision tree is constructed based on clinical guidelines, individualized treatment principles, and historical case data analysis, and is combined with a reinforcement learning algorithm to continuously optimize the decision rule set of the multi-level decision tree; The notification module is used to dynamically adjust the infusion device according to the infusion adjustment instruction to ensure the safety and effectiveness of the infusion process, and when an abnormal situation is detected, the Bayesian network is used to evaluate the risk level, and an alarm of a corresponding level is issued based on the risk level to notify medical staff; at the same time, all multi-dimensional physiological parameters and abnormal information corresponding to the abnormal situation during each infusion process are recorded to form an infusion process log; The adjustment module is used to analyze the relationship between the multi-dimensional physiological parameters and abnormal events during the infusion process after each infusion process, based on the infusion process log, using an association rule mining algorithm to identify potential influencing factors and their interactions, and to predictively optimize the decision-making algorithm and infusion control strategy of the patient in the future infusion process based on the influencing factors and their interactions through a machine learning model, so as to adjust the infusion control strategy of the patient.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an adaptive infusion control method based on physiological parameters as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, an adaptive infusion control method based on physiological parameters as claimed in any one of claims 1 to 7 is implemented.

Images