Intelligent vascular active drug closed-loop infusion system for cardiac surgery ICU (Intensive Care Unit)

Abstract

The invention discloses an intelligent vasoactive drug closed-loop infusion system for a cardiac surgery ICU, and relates to the technical field of medical equipment and control, in particular to the intelligent vasoactive drug closed-loop infusion system for the cardiac surgery ICU, which comprises a multi-channel physiological signal acquisition unit for continuously acquiring blood pressure and heart rate signals of a patient, and the central processing and control unit is internally provided with a medicine regulation and control logic module. The regulation and control module automatically generates a control instruction without delay based on comparison between the real-time physiological signal and a preset target range, and directly drives the medicine infusion execution unit to adjust the infusion rate. The system further comprises a safety boundary monitoring unit and a nursing record synchronization interface, so that treatment safety and data synchronization are ensured. Besides, the system can integrate third-party monitoring equipment data, supports multi-drug collaborative management and AI-based individualized target dynamic setting, and realizes intelligent and automatic closed-loop accurate regulation and control of vasoactive drug infusion of critical patients.

Description

Technical Field

[0001] This invention relates to the field of medical equipment and control technology, specifically to an intelligent closed-loop infusion system for vasoactive drugs in cardiac surgery ICUs. Background Technology

[0002] Maintaining hemodynamic stability is crucial in the postoperative management of cardiac surgery. Vasoactive drugs (such as positive inotropic agents, vasoconstrictors, and vasodilators) are key treatments for regulating vital signs such as blood pressure, cardiac output, and organ perfusion. Currently, in cardiac surgery ICUs, the infusion of vasoactive drugs generally employs an open-loop, manually adjusted mode. This means that medical staff manually set and adjust the infusion rate of the infusion pump based on intermittently collected physiological parameters such as arterial blood pressure, heart rate, and central venous pressure, combined with clinical experience. This traditional mode has a series of inherent defects and potential risks, becoming a bottleneck for refined and efficient postoperative management.

[0003] First, significant delays and lags are the main problems. A patient's physiological state, especially in the early postoperative period, is in a state of rapid and dynamic change. From the identification of parameter changes to the assessment and decision-making by healthcare professionals, and then to the manual adjustment of the infusion pump, there is a human delay of several minutes or even longer. During this period, the patient may already be in a dangerous state such as excessively low or high blood pressure, increasing the risk of insufficient organ perfusion or excessive cardiac load. This regulatory lag not only affects the optimal timing of treatment but may also cause physiological parameters to fluctuate around target values, which is detrimental to maintaining a stable internal environment.

[0004] Secondly, the high reliance on human resources and individual differences in experience introduce uncertainty. Closed-loop management requires medical staff to maintain high levels of attention for extended periods, frequently checking monitors and calculating dosage adjustments. In the busy ICU environment, this significantly increases the workload of medical staff and easily leads to oversights due to fatigue. Simultaneously, drug dosage adjustments are highly dependent on the personal experience and immediate judgment of medical staff; differences in strategy and scale may exist between different operators, making it difficult to achieve standardized, homogeneous, and precise treatment, thus affecting the controllability of overall medical quality.

[0005] Furthermore, the ability to handle complex dynamic processes is limited. Hemodynamic regulation is a complex process involving multiple variables, nonlinearity, and interdependent interactions. For example, adjusting a drug may simultaneously affect blood pressure and heart rate, with effect peaks exhibiting different time constants. When multiple vasoactive drugs are used in combination, their interactions (synergistic or antagonistic) further complicate manual regulation. Traditional manual methods struggle to analyze and coordinate these complex relationships in real time and quantitatively, typically employing relatively conservative or simplified approaches, making it difficult to achieve optimal control based on multi-parameter fusion.

[0006] Furthermore, there are blind spots in safety monitoring. In the traditional model, monitoring the status of the infusion system itself (such as infusion line blockage, infusion pump communication failure, and impending completion of drug infusion) is relatively separate from monitoring the patient's physiological state. When abnormalities occur, early warning and intervention still rely on manual detection, which may lead to delays in treatment. In extreme cases, such as sudden signal loss or equipment failure, the lack of automatic safety redundancy mechanisms may directly endanger patient safety.

[0007] To address these challenges, the field of medical automation control has begun exploring closed-loop infusion technology. However, most existing research and product solutions focus on single-parameter (e.g., blood pressure) monotherapy control, with relatively simple control strategies (e.g., classic PID controllers). These solutions are insufficiently adaptable to the complex, multi-stage pathophysiological characteristics of post-cardiac surgery patients and the practical clinical needs of multi-drug combination therapy. These systems often lack advanced intelligent decision-making logic to handle rapid fluctuations in physiological signals, multi-parameter coupling, multi-drug synergy, and differentiated goals at different clinical stages. They also lack in-depth, forward-looking safety boundaries and abnormal state handling mechanisms.

[0008] Therefore, there is a need to develop an intelligent closed-loop infusion system for vasoactive drugs specifically for cardiac surgery ICUs. This system needs to be able to continuously and in real-time acquire multi-channel physiological signals, automatically and without delay make control decisions that are close to or even superior to those made by experienced medical staff, and intelligently handle complex situations such as signal noise, physiological trends, and drug interactions. More importantly, the system must have a robust safety architecture, capable of automatically executing preset protective operations under any abnormal circumstances, thereby freeing medical staff from the heavy and stressful task of continuous manual adjustments, and ultimately achieving more stable, precise, and safer hemodynamic closed-loop management than manual management, thus improving patient outcomes. Summary of the Invention

[0009] The purpose of this invention is to provide an intelligent closed-loop infusion system for vasoactive drugs in cardiac surgery ICUs. By monitoring the patient's physiological signals in real time and comparing them with preset target values, the system automatically and without delay generates precise drug infusion control commands, realizing intelligent, closed-loop, and precise regulation of vasoactive drug administration to maintain the patient's hemodynamic stability, improve treatment safety and efficiency, and automatically synchronize key information to nursing records.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a closed-loop infusion system for intelligent vasoactive drugs in a cardiac surgery ICU, comprising: A multi-channel physiological signal acquisition unit is used to continuously and in real time acquire the patient's arterial blood pressure, central venous pressure, heart rate and blood oxygen saturation signals, and convert the signals into standard digital signals for output; The central processing and control unit has its input end connected to the output end of the physiological signal acquisition unit via a wired or wireless data interface, and is used to receive and process the digital signals. The drug infusion execution unit has its control terminal connected to the digital output terminal of the central processing and control unit via a control bus. It is used to receive control commands and drive at least one high-precision injection pump to adjust the infusion rate of vasoactive drugs. The central processing and control unit is also bidirectionally connected to a nursing record synchronization interface, which is used to automatically write or update key information, including abnormal physiological signal events, drug infusion rate adjustment instructions, system alarm events and operation logs, into the electronic nursing record sheet according to a preset format during system operation. The central processing and control unit has a built-in core drug regulation logic module. This module is configured to: compare the real-time blood pressure and heart rate signal values ​​with the patient's personalized target blood pressure range and target heart rate range stored in the module at a preset sampling period; automatically generate digital control instructions without manual review delay based on the comparison results and the built-in regulation rules; and send them directly and in real time to the drug infusion execution unit to drive the infusion pump to adjust the infusion rate of vasoactive drugs.

[0011] Furthermore, the drug regulation logic module specifically includes a first-level regulation submodule and a second-level regulation submodule connected in series; The first-level regulation submodule is configured to: when the real-time monitored physiological signal value continuously exceeds the internally preset target value range and reaches the first time threshold, calculate and output a preliminary adjustment amount of vasoactive drug infusion rate based on a pre-stored first type of rule set established based on clinical guidelines and pharmacological principles. The first type of rule set includes at least a function relating the signal deviation direction, deviation magnitude and adjustment amount. The second-level control submodule has its input end connected to the output end of the first-level control submodule and is configured to: receive the preliminary adjustment amount, simultaneously access the real-time physiological signal change trend data stream, dynamically correct and smooth the preliminary adjustment amount according to the pre-stored second-class rule set that focuses on system stability and safety, and finally generate control instructions for driving the drug infusion execution unit.

[0012] Furthermore, the correction logic included in the second type of rule set specifically includes: Oscillation suppression logic: When the absolute value of the rate of change of a specific physiological signal value per unit time exceeds the preset rapid change threshold, this logic is triggered to apply a suppression coefficient of less than 1 to the initial adjustment amount calculated by the first-level control submodule, so as to slow down the adjustment amplitude and avoid signal oscillation caused by the system response being too fast. Boundary fine-tuning logic: When a specific physiological signal value is detected to fluctuate slightly near the boundary of a preset target value range, enter a state of stagnation, and reach a second time threshold, this logic is triggered to apply a fine-tuning coefficient to the initial adjustment amount. This fine-tuning coefficient is dynamically calculated based on the amplitude and frequency of the stagnation, enabling the system to make more precise and gradual rate adjustments to promote the stabilization of physiological signals within the target range.

[0013] Furthermore, the system also includes an independently operating security boundary monitoring unit; The safety boundary monitoring unit has an independent data input channel for continuously and in parallel monitoring the raw or preprocessed physiological signal data stream from the physiological signal acquisition unit, while continuously monitoring the working status, communication link status and infusion rate of each injection pump in the drug infusion execution unit. The safety boundary monitoring unit is preset with several safety limit range conditions and equipment abnormal state conditions. When any physiological signal value exceeds its corresponding safety limit range, or when the physiological signal data stream loss continues for more than a third time threshold, the infusion pump communication is interrupted, or the infusion pump reports a hardware error, the safety boundary monitoring unit can immediately bypass the main control logic of the central processing and control unit and directly send instructions to the drug infusion execution unit and the system alarm unit to trigger the system to enter the preset advanced safety mode and drive the audible and visual alarm device to issue an alarm. When the safety boundary monitoring unit triggers an alarm or executes a mode switching command, it will simultaneously synchronize the event type code, the precise event timestamp, the relevant physiological data snapshot at the time of the event trigger, and the safety mode to be executed to the specific safety event section of the electronic nursing record through the nursing record synchronization interface.

[0014] Furthermore, the advanced security mode includes at least two preset modes: Safety Hold Mode: When the trigger condition is a brief signal abnormality or uncertain interference, the system will execute this mode, which will immediately lock the infusion rate of all current vasoactive drugs and maintain the rate unchanged, while simultaneously sounding an alarm and waiting for confirmation from medical staff. Safety Degradation Mode: When the trigger condition is a clear and continuous physiological signal exceeding the limit, the system executes this mode, that is, according to the deceleration program with multiple time steps pre-stored in the safety boundary monitoring unit, the infusion rate of the relevant vasoactive drugs is gradually reduced until a preset minimum baseline maintenance infusion rate or zero point is reached.

[0015] Furthermore, the central processing and control unit is also connected to the existing monitoring equipment network in the ICU through a standardized monitoring data integration interface; The monitoring data integration interface is used to obtain extended physiological parameters, including at least central venous pressure and blood oxygen saturation values, from third-party patient monitors at fixed intervals. In the process of generating control commands based on blood pressure and heart rate, the drug regulation logic module incorporates the acquired central venous pressure and blood oxygen saturation data as auxiliary judgment parameters into the calculation to assess the patient's volume status and peripheral perfusion, thereby enhancing the contextual awareness of regulation decisions. In addition, all data from third-party monitoring devices received through the monitoring data integration interface, along with the infusion rate adjustment instructions generated by the drug regulation logic module each time and the actual infusion rate after the instructions are executed, are encapsulated into data packets and automatically filled into the corresponding vital signs record and medication record fields in the electronic nursing record through the nursing record synchronization interface.

[0016] Furthermore, the system also includes a simulation and deduction unit running in the background; The simulation and deduction unit can take the current system state as the initial condition, including real-time physiological signal values ​​and current drug infusion rates, and combine it with the patient physiological response model built based on the population pharmacokinetic-pharmacodynamic model and physiological system model stored in its internal storage. After receiving the regulation request and before the actual execution of the instruction, it can perform rapid calculation and simulation to deduce the predicted change trajectory of the patient's key physiological signals when different candidate drug adjustment strategies are used within a preset time window in the future. The simulation and deduction unit outputs the simulation and deduction results of each strategy, including the predicted physiological signal curves, possible extreme values ​​and stability indicators, to the drug regulation logic module, so that it can perform internal multi-strategy pre-evaluation and optimal selection reference before finally generating and issuing control commands.

[0017] Furthermore, the system includes a multi-drug synergistic management module, which is suitable for the closed-loop infusion management of two or more vasoactive drugs with different or complementary pharmacological effects simultaneously. The collaborative management module contains pre-stored drug priority rules set by clinical experts to determine the order of use of various drugs when it is necessary to raise blood pressure or adjust hemodynamics; it also contains pre-stored interaction compensation rules based on known pharmacodynamic interaction studies between drugs. When a patient's physiological state requires adjustments to the infusion rates of multiple drugs, the collaborative management module first determines the adjustment order and master-slave relationship of various drugs according to the priority rules. Then, when calculating the adjustment amount of the master drug, it simultaneously calculates the superimposed or antagonistic impact that this adjustment may have on the target effects of other related drugs according to the interaction compensation rules, and generates compensatory adjustment suggestions for related drugs accordingly. Finally, it outputs a comprehensive adjustment scheme that includes the adjustment rates and timing of multiple drugs. When performing any collaborative adjustment operation involving multiple drugs, the multi-drug collaborative management module will synchronize the complete adjustment plan, the reason for the adjustment, the original rate of each drug before adjustment, and the target rate after adjustment to the medication change record section of the electronic nursing record form in a structured list form through the nursing record synchronization interface.

[0018] Furthermore, the central processing and control unit also includes an individualized dynamic goal setting module based on machine learning algorithms, which is configured to: While the system is running for the current patient, it continuously collects and stores time-series physiological signal data and drug infusion rate data. Through machine learning algorithms, it analyzes the historical response patterns of the current patient to the adjustment of the vasoactive drug infusion rate, including sensitivity, delay time, and steady-state gain characteristics. Based on the learned individualized response pattern characteristics and combined with the patient's basic pathophysiological state, the module can make small-scale dynamic fine-tuning and personalized offsets to the default target blood pressure and heart rate value ranges that are pre-stored in the drug regulation logic module and are applicable to the general population, thereby generating a set of individualized target value ranges that are more suitable for the current patient's unique physiological characteristics and response patterns, and setting it as a new comparison benchmark for the drug regulation logic module. After each automatic adjustment of the individualized target value range, the individualized dynamic target setting module will generate a change description record through the nursing record synchronization interface, which will include the specific adjustment range, the main data feature patterns on which the adjustment is based, and the effective start time of the new target value range. This record will then be synchronized to the individualized treatment parameter setting section of the electronic nursing record.

[0019] Furthermore, the central processing and control unit also includes a regulation policy self-optimization module based on a reinforcement learning framework, which is configured as follows: During the closed-loop control process of the system, system status data, data of the executed control actions, and changes in the stability of the patient's physiological signals over a period of time after the actions are continuously collected at fixed intervals as control effect feedback data, which constitute empirical data. Based on reinforcement learning algorithms, with the dual optimization objectives of "maintaining the stability of the patient's physiological signals within the target range in the long term" and "minimizing the number of unnecessary frequent adjustments to the drug infusion rate", the module uses accumulated empirical data to perform offline, iterative optimization learning and updating of key parameters in the pre-stored control rule set in the drug regulation logic module. After each round of strategy optimization iteration, the self-optimization module of the control strategy will synchronize the key node time of optimization, the numerical comparison of the control rule parameters involved in the optimization before and after optimization, and the expected control effect improvement evaluation report based on historical data backtesting to the system self-learning record department of the electronic nursing record through the nursing record synchronization interface in the form of version update log.

[0020] This invention provides an intelligent closed-loop infusion system for vasoactive drugs in cardiac surgery ICUs, which has the following beneficial effects: 1. Improve the accuracy and timeliness of hemodynamic management. This system continuously acquires physiological signals through multiple channels, combines them with built-in drug regulation logic, compares them in real-time with preset targets, and automatically generates control commands without human delay, directly driving the infusion device. This process eliminates delays and individual experience differences inherent in traditional manual observation, judgment, and adjustment. Especially on the basis of the first-level regulation, the second-level regulation module can suppress or fine-tune the initial adjustment based on signal change trends, achieving rapid and stable regulation of key indicators such as blood pressure and heart rate. This helps to maintain the patient's physiological state more accurately and promptly within the target range, providing strong support for circulatory stability in critically ill patients, such as those after cardiac surgery.

[0021] A multi-layered safety protection system is constructed to enhance system reliability. The system has an independently configured safety boundary monitoring unit that continuously monitors signal data flow and equipment status. Once an anomaly is detected, such as physiological signals exceeding safety limits, signal loss, or equipment communication interruption, a preset safety hold or safety degradation mode is immediately triggered, along with a simultaneous alarm. This design, which separates core control logic from proactive safety monitoring, is equivalent to adding a real-time monitoring "insurance" layer to automated control. In the event of extreme situations or technical failures, it can adopt conservative strategies (such as rate locking or stepped speed reduction) to mitigate risks, buying time for medical personnel to intervene and significantly reducing the possibility of a sudden increase in patient risk due to a single system failure.

[0022] The system achieves multi-source information fusion and decision support, optimizing control quality. It integrates key third-party parameters such as central venous pressure and blood oxygen saturation through a monitoring data integration interface, providing the control module with auxiliary judgment criteria for more comprehensive decision-making. The added simulation and deduction unit can pre-simulate the future effects of different adjustment strategies based on the current state and physiological models, providing internal pre-evaluation references for control command generation. Furthermore, the collaborative management module for multi-drug combination therapy can comprehensively calculate compensatory adjustment schemes based on pre-stored priorities and interaction rules. These functions collectively expand the system's perception and "foresight" capabilities, enabling automatic control not only based on current deviations but also considering multi-parameter correlations and future trends, thereby improving the overall robustness and adaptability of control decisions.

[0023] The system incorporates individualization and self-optimization capabilities to enhance clinical adaptability. It features an AI-based individualized dynamic target setting module that learns patients' unique response patterns to medication during operation and fine-tunes the default target range accordingly, making control standards more aligned with individual physiological characteristics. Simultaneously, a reinforcement learning-based strategy self-optimization module continuously iterates and optimizes control rule parameters with the goal of stabilizing physiological indicators and reducing ineffective operations. This allows the system to transcend fixed, universal control logic, gradually adapting to specific patients and optimizing its own strategies. This promises to provide more personalized and time-dependent control performance in long-term monitoring, reducing healthcare professionals' reliance on one-size-fits-all parameter settings.

[0024] Strengthening human-machine collaboration and the closed loop of medical data ensures clinical compliance and traceability. Through a nursing record synchronization interface, the system automatically synchronizes key monitoring data, alarm events, adjustment instructions, execution results, and even individualized adjustment basis and strategy optimization records to nursing records, ensuring the integrity, accuracy, and timely archiving of data throughout the entire process, greatly facilitating medical documentation and post-event auditing. An integrated multimodal interaction intent confirmation unit performs secondary confirmation and records the process during major operations or manual interventions, enhancing the reliability and accountability of human-machine interaction. These designs ensure that the intelligent system is deeply embedded in clinical workflows while strictly adhering to medical safety regulations, achieving traceable operations and auditable decisions, and effectively supporting the final decision-making and accountability management of medical staff. Attached Figure Description

[0025] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0026] Figure 1This is a flowchart illustrating the overall system architecture of the present invention; Figure 2 This is a flowchart illustrating the two-level drug regulation logic of the present invention. Figure 3 This is a flowchart of the security boundary monitoring and anomaly handling process of the present invention; Figure 4 This is a flowchart illustrating the multi-drug synergistic adjustment process of the present invention. Figure 5 This is a flowchart of the AI ​​individualized target + reinforcement learning self-optimization process of the present invention. Detailed Implementation

[0027] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses consistent with some aspects of this disclosure as detailed in the appended claims.

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0029] How to use: 1. System startup and target setting First, connect the multi-channel physiological signal acquisition unit to ensure that at least the patient's blood pressure and heart rate signals can be continuously acquired.

[0030] Through the human-machine interface of the central processing and control unit, medical staff can preset target ranges for physiological signals such as blood pressure and heart rate in the drug regulation logic module according to the patient's condition. This serves as the benchmark for the system's automatic regulation.

[0031] Automatic closed-loop control execution After the system starts up, it enters automatic operation mode. The central processing and control unit will continuously receive physiological signals.

[0032] The drug regulation logic module compares real-time physiological signal values ​​with preset target value ranges and automatically and directly generates control commands to drive the drug infusion execution unit (such as an injection pump) to adjust the drug infusion rate in order to maintain stable physiological parameters.

[0033] The regulation process employs a tiered strategy: the first-level regulation submodule calculates an initial drug infusion rate adjustment based on a first-class rule set when the signal deviates from the target range. The second-level regulation submodule then combines physiological signal trend data and makes corrections based on a second-class rule set. For example, when the signal changes too rapidly (rate exceeds the threshold), an inhibition coefficient is applied to the initial adjustment; when the signal hovers around the target boundary, a fine-tuning coefficient is applied to generate the final control command, avoiding over-regulation.

[0034] Multi-source data integration and collaborative management Third-party devices can be connected via a monitoring data integration interface to obtain data such as central venous pressure and blood oxygen saturation. This data will be used as auxiliary judgment parameters in regulatory decisions.

[0035] If two or more vasoactive drugs need to be used simultaneously, the multi-drug synergistic management module can be activated. This module will determine the adjustment order based on pre-stored priority rules and, according to interaction compensation rules, calculate the impact of adjusting one drug on the effects of other drugs and make compensatory adjustments accordingly. All synergistic adjustment schemes, changes in the rates of each drug, and their reasons will be recorded synchronously.

[0036] Safety monitoring and abnormal handling The safety boundary monitoring unit continuously monitors the physiological data stream and equipment status in the background. Once the physiological signal exceeds the safety limit, or if there is an anomaly such as signal loss or pump communication interruption, the unit will immediately trigger the preset contingency plan.

[0037] The contingency plan includes: a safety hold mode (locking the current infusion rate) or a safety degrade mode (reducing to the basic maintenance dose according to a preset step sequence), while issuing an alarm. Relevant event types, times, and data will be synchronized to the nursing records in real time.

[0038] Application of advanced features Simulation and extrapolation: In complex situations, the simulation and extrapolation unit can be activated. Based on current patient data, infusion rate, and built-in physiological models, this unit simulates the physiological trajectory under different adjustment strategies in the future, providing internal pre-assessment references for regulatory decisions.

[0039] Individualization and self-optimization: The AI-based personalized dynamic target setting module learns patients' historical response patterns to medication during operation and dynamically fine-tunes the default target value range accordingly, generating a more personalized target range.

[0040] The self-optimization module of the regulation strategy based on reinforcement learning continuously collects control process data to maintain stability and reduce frequent adjustments, and uses algorithms to iteratively optimize the regulation rule parameters.

[0041] Key information regarding the aforementioned individualized adjustments and strategy optimizations (such as adjustment range, basis, parameter comparison, etc.) will be recorded in detail.

[0042] Human-computer interaction and recording synchronization All critical system operations, medication rate adjustment commands, and execution results are automatically synchronized to the electronic nursing record sheet through the nursing record synchronization interface, ensuring data consistency and traceability.

[0043] When medical staff need to intervene actively or the system detects important command input, the multimodal human-computer interaction and intent confirmation unit (which may integrate voice, gesture and other recognition methods) will be activated to confirm the operation intent for a second time. All interaction events and confirmation processes will also be recorded and synchronized.

[0044] Example: Example 1: Application of graded regulation in postoperative hypertension management The patient was transferred to the ICU after coronary artery bypass surgery, experiencing blood pressure fluctuations due to stress. Medical staff set target ranges for blood pressure and heart rate for the system. Upon system startup, the multi-channel physiological signal acquisition unit continuously acquired the patient's arterial blood pressure and electrocardiogram (ECG) heart rate signals. The central processing and control unit received this data in real time. Initially, the patient's blood pressure remained above the upper limit of the preset target range. At this point, the drug regulation logic module was activated. Its built-in first-level regulation submodule quickly calculated an initial adjustment amount to reduce the infusion rate of vasopressors based on the degree of blood pressure deviation and a first-class rule set. Immediately afterwards, the second-level regulation submodule intervened. It analyzed that although the blood pressure was high, the rate of change did not exceed the preset rapid threshold, therefore no inhibition coefficient was applied. However, the module detected a persistent trend of blood pressure hovering near the target range boundary, so it applied a fine-tuning coefficient to the initial adjustment amount, generating a smoother final control command. This command directly drove the drug infusion execution unit, precisely reducing the infusion rate of the infusion pump. The entire regulation process was automatic and continuous, without any delays caused by human observation, judgment, or operation. A few minutes later, the patient's blood pressure returned to the target range. Furthermore, key time points during the adjustment process, blood pressure values ​​before and after adjustment, the calculated adjustment amount, and the final infusion rate were all automatically and accurately synchronized to the patient's electronic nursing record via the nursing record synchronization interface, forming a complete treatment record.

[0045] Example 2: Implementation of multi-drug synergy and safety monitoring in the management of complex shock A patient in cardiogenic shock requires simultaneous administration of two vasoactive drugs, norepinephrine and dopamine, to maintain circulation. Healthcare professionals activated the multi-drug synergistic management module in the system and preset priority rules for the relevant drugs. During system operation, the multi-channel physiological signal acquisition unit detected a downward trend in the patient's blood pressure. The drug regulation logic module calculated an initial instruction requiring increased vasoactive drug support based on the blood pressure signal. This instruction was first sent to the multi-drug synergistic management module. Based on pre-stored rules, this module determined that norepinephrine was the first-line drug for maintaining blood pressure and had a higher priority, thus deciding to adjust norepinephrine first. Next, based on pre-stored interaction compensation rules, the module calculated that increasing norepinephrine might have a slight impact on some of the cardiac effects of dopamine, and therefore simultaneously calculated a small compensatory increase in the dopamine infusion rate. Finally, the system generated a composite control instruction containing the sequential adjustment and synergistic compensation of the two drugs, driving the drug infusion execution unit to precisely adjust the two infusion pumps. The entire coordinated adjustment plan, the original and target infusion rates of each medication, and the reasons for adjustments based on interactions were all recorded and synchronized through the nursing record synchronization interface. During treatment, the safety boundary monitoring unit continuously monitored the patient in the background. Suddenly, due to a change in patient position, the blood pressure monitoring signal was transiently lost. The safety boundary monitoring unit immediately detected the abnormal signal loss, instantly triggering the system to enter the preset safety hold mode, immediately locking all medication infusion rates to maintain a constant level, and issuing an audible and visual alarm to alert medical staff. The time and type of this abnormal event, along with the physiological data from the preceding moment, were also simultaneously recorded in the nursing record.

[0046] Example 3: Data Integration and Simulation to Support Decision Making A patient who had undergone valve replacement surgery had blood pressure within the target range but elevated central venous pressure (CVP). This system, through its monitoring data integration interface, received real-time CVP and oxygen saturation data from third-party monitoring devices. In routine adjustments, the drug regulation logic module used not only the blood pressure and heart rate data collected by this system but also these integrated CVP and oxygen saturation data as auxiliary parameters. At a certain point, the system detected a slow downward trend in the patient's blood pressure, while the CVP data indicated a potential change in volume status. To develop a better adjustment strategy, the medical staff invoked the simulation simulation unit. Based on the current physiological signal values, drug infusion rate, and a built-in physiological response model, this unit simulated and predicted the possible trajectories of the patient's blood pressure, heart rate, and CVP over the next thirty minutes under three different strategies: "small increase in drug A," "maintaining the current rate and observing," and "combined adjustment of drugs A and B." The graphical results of the simulation provided valuable internal pre-assessment references for the drug regulation logic module before generating the final control command. The medical staff, referring to these simulation results and combining them with clinical judgment, approved the robust adjustment plan subsequently generated by the system. The instructions, execution results, and reasons for triggering the simulation were all integrated and recorded through the nursing record synchronization interface.

[0047] Example 4: AI Personalized Goal Setting and Self-Optimizing Learning Process When managing a heart failure patient who had been long-term dependent on vasoactive drugs, the system's AI-based individualized dynamic target setting module continuously operated. Over several days of system operation, this module silently learned and deeply analyzed the patient's unique physiological signals and their historical response patterns to each adjustment of the drug infusion rate. For example, it discovered that the patient's response to minute dose adjustments was more sensitive and milder than that of typical patients. Based on this learned individualized response pattern, the module dynamically fine-tuned the default blood pressure target range pre-stored in the drug regulation logic module, narrowing it by 5%, generating an individualized target range more suited to the patient's current physiological characteristics. The magnitude of this adjustment, the basis for historical response sensitivity analysis, and the specific time of its effect were all synchronized to the nursing record sheet via the nursing record synchronization interface, ensuring that changes in treatment goals were traceable. Simultaneously, the system's reinforcement learning-based self-optimization module for regulation strategies was also working. It continuously collected state data, adjustment action data, and feedback data on the stabilization of physiological signals after adjustments during the closed-loop control process. With the optimization goal of maintaining long-term stability of physiological signals and reducing unnecessary frequent adjustments, this module uses reinforcement learning algorithms to iteratively optimize the parameters of the pre-stored control rule set in the drug regulation logic module, such as fine-tuning the response gain coefficient under certain conditions. The key nodes of this strategy optimization, the parameter comparison table before and after optimization, and the expected control effect evaluation report are also recorded and synchronized, realizing the transparent evolution of the control strategy.

[0048] Example 5: Multimodal Interaction and Emergency Intervention Recording During routine system operation, the multimodal human-computer interaction and intent confirmation unit remains in standby mode. One day, while monitoring the patient at the bedside, medical staff, based on newly emerging vital signs, determined it was necessary to immediately pause the current automated closed-loop infusion, switch to manual control, and significantly reduce the dosage of a certain medication. When the medical staff pressed the virtual "Emergency Intervention - Pause Closed-Loop" button on the main interface of the central processing and control unit, the system recognized this as a pre-set critical operation command input. Immediately, the multimodal human-computer interaction and intent confirmation unit was activated. A prominent confirmation dialog box popped up on the system interface, and the integrated microphone activated, clearly prompting via voice synthesis: "Pause closed-loop infusion command received, please confirm again." While uttering the "confirm" command, the medical staff performed a specific confirmation gesture as prompted on the screen. The unit completed a dual-modal secondary confirmation of the operation intent through voice recognition and gesture recognition. After confirmation, the system immediately exited the automated closed-loop mode, transferring control to the medical staff and recording the complete process of this critical interaction event: including the trigger command, voice confirmation content, gesture recognition result, and the final executed mode switching command. All these detailed logs of interactions and confirmations are synchronized to the "Special Operation Records" section of the nursing record sheet through the nursing record synchronization interface, ensuring the traceability of any major human intervention and meeting the core requirements of medical safety management.

[0049] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A closed-loop infusion system for intelligent vasoactive drugs in a cardiac surgery ICU, characterized in that, include: A multi-channel physiological signal acquisition unit is used to continuously acquire at least the patient's blood pressure and heart rate signals; The central processing and control unit has its input terminal connected to the physiological signal acquisition unit; A drug infusion execution unit, whose control terminal is connected to the output terminal of the central processing and control unit, is used to drive at least one infusion pump to infuse vasoactive drugs; The central processing and control unit is also connected to a nursing record synchronization interface, which is used to synchronize key detection content during system operation with nursing record sheets; The central processing and control unit has a built-in drug regulation logic module. This module compares the real-time acquired physiological signal values ​​with its internally preset target value range and automatically generates control commands without human delay, directly driving the drug infusion execution unit to adjust the infusion rate.

2. The intelligent vasoactive drug closed-loop infusion system for cardiac surgery ICU according to claim 1, characterized in that: The drug regulation logic module includes a first-level regulation submodule and a second-level regulation submodule; The first-level regulation submodule is used to calculate and output a preliminary drug infusion rate adjustment amount according to a pre-stored first-class rule set when the physiological signal value deviates from the target value range; The second-level control submodule receives the preliminary adjustment amount and physiological signal change trend data, corrects the preliminary adjustment amount according to the pre-stored second type of rule set, and generates the final control command.

3. The intelligent vasoactive drug closed-loop infusion system for cardiac surgery ICU according to claim 2, characterized in that: The correction logic of the second type of rule set includes: when the rate of change of the detected physiological signal value exceeds a preset threshold, applying an inhibition coefficient to the initial adjustment amount; when the detected physiological signal value continues to hover at the target value boundary, applying a fine adjustment coefficient to the initial adjustment amount.

4. The intelligent vasoactive drug closed-loop infusion system for cardiac surgery ICU according to claim 1, characterized in that: The system also includes a security boundary monitoring unit; The safety boundary monitoring unit continuously monitors the data stream of the physiological signal acquisition unit and the status of the drug infusion execution unit; When the physiological signal value exceeds the preset safety limit range, or when at least one abnormal state such as signal loss or interruption of infusion pump communication is detected, the safety boundary monitoring unit can immediately trigger the system to enter the preset safety maintenance mode or safety degradation mode and issue an alarm. When the safety boundary monitoring unit triggers an alarm or switches modes, it will synchronize the event type, occurrence time, and related physiological data to the nursing record sheet through the nursing record synchronization interface.

5. The intelligent vasoactive drug closed-loop infusion system for cardiac surgery ICU according to claim 4, characterized in that: The safety maintenance mode is to immediately lock the current drug infusion rate and maintain it unchanged; the safety degradation mode is to gradually reduce the drug infusion rate to a baseline maintenance level according to a preset step procedure.

6. The intelligent vasoactive drug closed-loop infusion system for cardiac surgery ICU according to claim 1, characterized in that: The central processing and control unit is also connected to a monitoring data integration interface for receiving at least central venous pressure and blood oxygen saturation data from third-party monitoring devices; when generating control commands, the drug regulation logic module uses the central venous pressure and blood oxygen saturation data as auxiliary judgment parameters. The data received by the monitoring data integration interface from third-party monitoring devices, as well as the infusion rate adjustment instructions and execution results generated by the drug regulation logic module, are all synchronized to the nursing record sheet through the nursing record synchronization interface.

7. The intelligent vasoactive drug closed-loop infusion system for cardiac surgery ICU according to claim 1, characterized in that: The system also includes a simulation and deduction unit; The simulation unit can simulate the trajectory of physiological signal changes under different adjustment strategies over a period of time based on the current physiological signal values, drug infusion rate, and built-in patient physiological response model. The output of the simulation unit can be used by the drug regulation logic module for internal pre-evaluation before generating control commands.

8. The intelligent vasoactive drug closed-loop infusion system for cardiac surgery ICU according to claim 1, characterized in that: The system includes a multi-drug synergistic management module, which is suitable for the simultaneous infusion of two or more vasoactive drugs; The collaborative management module pre-stores priority rules and interaction compensation rules among different drugs; When it is necessary to adjust the infusion rate of multiple drugs, the collaborative management module determines the adjustment order according to the priority rule, and calculates the impact of adjusting one drug on the target effect of another drug according to the interaction compensation rule and makes compensatory adjustments. When performing any collaborative adjustment operation involving multiple drugs, the multi-drug collaborative management module will synchronize the adjustment plan, the original rate, target rate, and reason for adjustment of each drug to the nursing record sheet through the nursing record synchronization interface.

9. The intelligent vasoactive drug closed-loop infusion system for cardiac surgery ICU according to claim 1, characterized in that: The central processing and control unit also includes an AI-based individualized dynamic goal setting module, which is configured to: During system operation, it continuously learns and analyzes the historical response patterns of current patient physiological signals to drug infusion rate adjustments; Based on the learned individual response patterns, the default target value range stored in the drug regulation logic module is dynamically fine-tuned to generate an individualized target value range that is more suitable for the current physiological characteristics of the patient. The adjustment range, basis, and effective time of the individualized target value range are synchronized to the nursing record sheet through the nursing record synchronization interface.

10. A closed-loop intelligent vasoactive drug infusion system for cardiac surgery ICU according to claim 1 or 9, characterized in that: The central processing and control unit also includes a reinforcement learning-based regulation policy self-optimization module, which is configured as follows: Continuously collect state data, action data, and control effect feedback data of the closed-loop control process; Based on reinforcement learning algorithms, with the optimization goal of maintaining stable physiological signals and reducing unnecessary frequent adjustments, the parameters of the pre-stored control rule set in the drug regulation logic module are iteratively optimized. Key nodes of strategy optimization, parameter comparisons before and after optimization, and expected effect assessments are synchronized to the nursing record sheet through the aforementioned nursing record synchronization interface. The system also includes a multimodal human-computer interaction and intent confirmation unit, which integrates at least one interaction modality among speech recognition, gesture recognition, and eye tracking; When the system detects a preset critical operation command input or when medical staff actively intervene, it can confirm the operation intention a second time through multimodal interaction. The human-computer interaction and intent confirmation unit records all major interaction events, confirmation processes, and final instructions, and synchronizes them to the nursing record sheet through the nursing record synchronization interface.

Images