Multi-parameter real-time monitoring and adaptive control method and system for hemodialysis machine

Abstract

The application discloses a kind of hemodialysis machine multi-parameter real-time monitoring and adaptive control method and system, by real-time acquisition blood component concentration, cardiac blood pumping burden level and body fluid electrolyte balance data, using time series analysis method obtains the change trend of multiple physiological indexes, and accurate classification is carried out to patient state in combination with classification algorithm.When detecting that heart burden increases, the application integrates body fluid data to determine physiological adjustment priority, uses prediction algorithm to evaluate potential heart risk level, if risk is over threshold, then through feedback control mechanism, prediction result and real-time data are fused, treatment parameters are optimized and optimal hemodialysis machine setting is searched, dynamic adjustment scheme is formed, finally, interface transmission parameter and continuous monitoring update trend are formed, and cycle optimization is carried out until patient reaches stable state.The application is fused by multidimensional data and dynamic feedback control, significantly improves the individualization accuracy and safety of hemodialysis treatment, and reduces heart risk.

Description

Technical Field

[0001] This invention relates to the field of medical and health technology, and in particular discloses a method and system for real-time monitoring and adaptive control of multiple parameters of a hemodialysis machine. Background Technology

[0002] In the field of healthcare, hemodialysis technology plays an irreplaceable role as a crucial means of life support for patients with kidney failure. Currently, the adjustment of parameters for hemodialysis equipment largely relies on preset protocols or the experience and judgment of medical staff, which often fails to cope with the complex physical changes that patients experience during treatment.

[0003] Especially when dealing with individual differences among patients, current technologies lack the ability to comprehensively analyze multiple physical indicators, which can easily lead to a mismatch between treatment parameters and the patient's actual needs, thereby affecting the safety and effectiveness of treatment. This problem not only increases the workload of medical staff but may also pose a potential threat to the patient's health.

[0004] A deeper technological challenge lies in how to capture and integrate various changes in a patient's body in real time during dialysis. Dialysis treatment involves blood components, body fluid balance, and the dynamic responses of the heart and blood vessels, and these indicators have complex interactions. Without timely monitoring of these overall changes, it is difficult to accurately assess a patient's condition. For example, changes in the concentration of certain blood components may affect the workload on the heart, and an increased workload on the heart may, in turn, interfere with the effectiveness of dialysis. This dynamic correlation of multiple pieces of information has become a major obstacle to achieving precision medicine.

[0005] Therefore, how to integrate multiple physical indicators in real time during dialysis and dynamically adjust equipment parameters based on the interaction of these indicators has become a key issue in improving treatment safety and individualized results. Summary of the Invention

[0006] This invention provides a method and system for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine, aiming to solve the technical problems of fixed parameters in traditional hemodialysis machines, inability to dynamically adjust according to the patient's real-time physiological state, and lagging cardiac risk prevention and control.

[0007] One aspect of the present invention relates to a method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine, comprising the following steps: S100: Real-time data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance are collected from the patient's body through physiological sensors. Time series analysis is used to process the data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance to obtain the trend of multiple physiological indicators. S200. Based on the changing trends of multiple physiological indicators, a classification algorithm is used to classify the current physical state of the patient. If the classification results show an increase in the heart's pumping burden, the body fluid and electrolyte balance data are integrated to determine the priority of physiological adjustments. S300: Obtain high-priority physiological indicators from the physiological adjustment priority, use a prediction algorithm to predict the potential cardiac risk level, and determine whether the potential cardiac risk level exceeds the preset threshold. S400 If the potential cardiac risk level exceeds the preset threshold, the prediction results and real-time data are fused through the feedback control mechanism to obtain the optimized value of the treatment parameters. Based on the optimized value of the treatment parameters, the optimization algorithm is used to search for the best combination of hemodialysis machine settings to obtain a dynamic physiological adjustment plan. S500: Extract specific parameter values ​​from the dynamic physiological adjustment plan, transmit the specific parameter values ​​through the hemodialysis machine interface, determine whether the transmission is successful, if the transmission is successful, monitor the changes of multiple physiological indicators, and reprocess the newly collected data using time series analysis methods to obtain updated physiological change trends. S600: Based on the updated physiological change trend, repeat the classification algorithm process to determine whether further physiological adjustments are needed. Obtain confirmation signals from the further physiological adjustment needs, apply physiological optimization values ​​through feedback control mechanisms, and obtain the final stable state of patient treatment.

[0008] Further, step S100 includes: S110. Use physiological sensors to collect data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance in real time, and construct a multidimensional time series based on the data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance. S120. Perform sliding window processing on the multidimensional time series to extract fluctuation features. If the fluctuation features exceed the preset threshold, use the correlation matrix to determine the evolution trajectory. S130. Calculate the trend vector based on the evolution trajectory, and obtain the changing trends of multiple physiological indicators by mapping the trend vector to a multi-dimensional coordinate system.

[0009] Further, step S200 includes: S210. Obtain the changing trends of multiple physiological indicators and map them to a high-dimensional feature space to extract feature vectors that reflect fluctuations in physiological function. S220. Input the feature vector into a pre-trained random forest classifier, and obtain the current patient's physical state by performing a non-linear mapping on the feature vector; S230. If the current patient's physical condition is identified as an increased burden on the heart's pumping capacity, then extract the corresponding body fluid and electrolyte balance data and construct a multidimensional correlation matrix. S240. Based on the multidimensional correlation matrix, quantitative processing is performed using a weight allocation model to determine the priority of physiological adjustment.

[0010] Further, step S300 includes: S310. Extract the top-ranked categories from the physiological adjustment priorities to determine high-priority physiological quantities; S320. Retrieve timestamp records from the database based on high-priority physiological quantities and generate historical monitoring sequences. Use a sliding window algorithm to smooth the historical monitoring sequences in order to extract circulatory system features. S330. Input the circulatory system features into a pre-trained gradient boosting regression tree model to obtain the potential cardiac risk level, and determine whether the potential cardiac risk level exceeds a preset threshold.

[0011] Further, step S400 includes: S410. If the potential cardiac risk level exceeds the preset threshold, the real-time blood flow velocity and vascular resistance data are obtained through the feedback control mechanism, and the deviation value of circulating blood volume and osmotic pressure gradient is determined through regression analysis to obtain the initial physiological deviation matrix. S420. Based on the initial physiological deviation matrix, retrieve the correlation records between dialysis pressure and electrolyte concentration, and use the particle swarm optimization algorithm to optimize the dialysis pressure and electrolyte concentration to determine the combination of operating benchmark settings for the hemodialysis machine. S430: Based on the baseline settings for the operation of the hemodialysis machine, and combined with heart rate variability and blood oxygen saturation data, calculate the dynamic response functions of venous pressure and arterial pressure to obtain the corrective settings for the operation of the hemodialysis machine. S440. Based on the hemodialysis machine operation correction setting combination, and using the constraints of total ultrafiltration volume and dialysate temperature, perform a second iteration calculation on the hemodialysis machine operation correction setting combination to obtain a dynamic physiological adjustment scheme.

[0012] Further, step S500 includes: S510. Extract specific parameter values ​​from the dynamic physiological adjustment plan and transmit the specific parameter values ​​through the hemodialysis machine interface; S520. Determine whether the transmission of specific parameter values ​​was successful. If the transmission was successful, monitor the changes in multiple physiological indicators and use time series analysis to process the newly collected data to obtain updated physiological change trends. S530. Based on the updated physiological change trend, the physiological deviation matrix is ​​integrated with the correlation of dialysis pressure to perform data calibration and determine the optimized physiological indicator monitoring model. S540. The newly collected data is processed a second time using a physiological indicator monitoring model to obtain the final updated physiological change trend.

[0013] Further, step S600 includes: S610. Based on the updated physiological change trend, call the logistic regression classification algorithm to determine whether there is a need for further physiological adjustment; S620: Extract trigger commands from the need for further physiological adjustments, and generate confirmation signals through the command parsing module; S630: Based on the confirmation signal, the feedback control mechanism injects physiological optimization values ​​into the hemodialysis machine control bus; S640: By applying physiological optimization values ​​through a feedback control mechanism, pressure fluctuation data is monitored in real time to obtain the final stable state of the patient's treatment.

[0014] Another aspect of the present invention relates to a real-time monitoring and adaptive control system for multiple parameters of a hemodialysis machine, for implementing the above-described real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine, comprising: The module for acquiring the changing trends of multiple physiological indicators is used to collect data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance from the patient in real time through physiological sensors. The data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance are processed by time series analysis to obtain the changing trends of multiple physiological indicators. The physiological adjustment priority determination module is used to classify the current patient's physical state based on the changing trends of multiple physiological indicators using a classification algorithm. If the classification results show an increase in the heart's pumping burden, the module integrates body fluid and electrolyte balance data to determine the physiological adjustment priority. The potential cardiac risk level assessment module is used to obtain high-priority physiological indicators from the physiological adjustment priority, use a prediction algorithm to predict the potential cardiac risk level, and determine whether the potential cardiac risk level exceeds a preset threshold. The dynamic physiological adjustment plan acquisition module is used to obtain optimized treatment parameter values ​​by fusing prediction results and real-time data through a feedback control mechanism if the potential cardiac risk level exceeds a preset threshold. Based on the optimized treatment parameter values, an optimization algorithm is used to search for the best combination of hemodialysis machine settings to obtain a dynamic physiological adjustment plan. The module for acquiring updated physiological change trends is used to extract specific parameter values ​​from the dynamic physiological adjustment plan, transmit the specific parameter values ​​through the hemodialysis machine interface, determine whether the transmission is successful, and if the transmission is successful, monitor the changes of subsequent multiple physiological indicators, reprocess the newly collected data using time series analysis methods, and obtain updated physiological change trends. The final patient treatment stability acquisition module is used to determine whether further physiological adjustments are needed based on the updated physiological change trend and the repeated classification algorithm process. It obtains confirmation signals from the further physiological adjustment needs, applies physiological optimization values ​​through a feedback control mechanism, and obtains the final patient treatment stability.

[0015] The beneficial effects achieved by this invention are as follows: This invention provides a method and system for real-time monitoring and adaptive control of multiple parameters in hemodialysis machines. Addressing the comprehensive operational challenges of increased cardiac workload, electrolyte imbalance, and potential cardiac risks during hemodialysis, this method collects real-time data on blood component concentrations, cardiac workload levels, and electrolyte balance. It employs time-series analysis to capture the trends of multiple physiological indicators and combines this with a classification algorithm for precise patient status categorization. When increased cardiac workload is detected, this invention integrates fluid data to determine physiological adjustment priorities, uses a predictive algorithm to assess potential cardiac risk levels, and if the risk exceeds a threshold, a feedback control mechanism fuses the prediction results with real-time data to optimize treatment parameters and search for the optimal hemodialysis machine settings, forming a dynamic adjustment plan. Finally, parameters are transmitted via an interface, and the update trend is continuously monitored and optimized iteratively until the patient reaches a stable state. The specific beneficial effects of this method and system for real-time monitoring and adaptive control of multiple parameters in hemodialysis machines are as follows: 1. This invention simultaneously collects multiple core physiological data such as blood component concentration, cardiac pumping burden, and body fluid electrolyte balance of patients, and mines the dynamic change trends of multi-dimensional physiological indicators through time series analysis. It breaks through the limitations of single parameter monitoring in traditional dialysis equipment, and can comprehensively capture the subtle dynamic fluctuations of the patient's physical condition during dialysis, providing complete and continuous physiological data support for the comprehensive assessment of the patient's condition.

[0016] 2. This invention relies on the changing trends of multiple physiological indicators to achieve intelligent classification of the patient's physical state. It can accurately identify abnormal physiological states such as increased cardiac pumping burden during dialysis, and dynamically determine the priority of physiological adjustments by combining electrolyte balance data. This solves the problems of blind parameter adjustment and lack of distinction between primary and secondary parameters in traditional control methods, and ensures that the intervention strategy is in line with the patient's real-time physiological tolerance state.

[0017] 3. This invention predicts potential cardiac risks based on high-priority physiological indicators, enabling proactive identification and assessment of dialysis risks. It transforms the traditional passive adjustment after exceeding the target into proactive early warning and intervention before risks occur, effectively avoiding hidden cardiac risks caused by fluctuations in physiological parameters during dialysis and significantly improving the safety of dialysis treatment.

[0018] 4. This invention integrates real-time monitoring data and prediction results under risk exceeding the threshold, and uses an optimization algorithm to search for the optimal combination of equipment parameters to generate a dynamic physiological adjustment plan that adapts to the patient's real-time physical condition. This overcomes the shortcomings of traditional fixed-parameter dialysis treatment modes, which cannot adapt to individual physiological differences and dynamic tolerance changes, and improves the accuracy and adaptability of dialysis parameter adjustment.

[0019] 5. This invention enables the precise issuance and execution of optimized parameters, and continuously iterates and collects and analyzes the latest physiological data of patients, dynamically updates physiological change trends, and constructs a continuous real-time monitoring and iteration mechanism. This mechanism can continuously track the physiological feedback status of patients after treatment, avoid the problem that a single adjustment cannot adapt to the dynamic changes in the course of the disease, and ensure the continuity of regulation.

[0020] 6. This invention continuously iterates and optimizes treatment parameters through a closed-loop feedback mechanism, repeatedly revising the control strategy based on real-time physiological changes in the patient, ultimately maintaining long-term stability of the patient's physiological state. This invention achieves closed-loop control throughout the entire process, including collaborative monitoring of multiple physiological parameters, intelligent risk prediction, and adaptive adjustment of equipment parameters. It aligns with the individualized and dynamic treatment needs of hemodialysis, effectively addressing industry pain points such as fixed dialysis treatment parameters, poor patient adaptability, and lagging physiological risk management. It significantly improves the safety, stability, and intelligent management level of hemodialysis treatment, demonstrating strong clinical applicability. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating an embodiment of the real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine according to the present invention; Figure 2 This is a functional block diagram of an embodiment of the real-time monitoring and adaptive control system for multi-parameter hemodialysis machines of the present invention.

[0022] Explanation of icon numbers: 10. Module for acquiring trends of multiple physiological indicators; 20. Module for determining the priority of physiological adjustments; 30. Module for assessing potential cardiac risk levels; 40. Module for acquiring dynamic physiological adjustment plans; 50. Module for acquiring updated physiological change trends; 60. Module for acquiring the final stable state of patient treatment. Detailed Implementation

[0023] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0024] like Figure 1 As shown, the first embodiment of this invention proposes a method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine. By collecting multiple physiological indicators of the patient in real time, combined with time series analysis, classification prediction, and feedback control technology, it achieves dynamic monitoring and adaptive adjustment of multiple parameters during hemodialysis treatment. It focuses on controlling cardiac risks, ensuring safe, accurate, and stable treatment, and solves the problems of fixed parameters in traditional hemodialysis machines, inability to dynamically adjust according to the patient's real-time physiological state, and lagging cardiac risk control. It is suitable for patients with chronic renal failure and other conditions requiring long-term hemodialysis treatment, and is compatible with various commonly used clinical hemodialysis machines. The method includes the following steps: Step S100: Collect data on blood component concentration, cardiac pumping burden level, and body fluid electrolyte balance from the patient in real time using physiological sensors. Process the data on blood component concentration, cardiac pumping burden level, and body fluid electrolyte balance using time series analysis methods to obtain the changing trends of multiple physiological indicators.

[0025] During hemodialysis treatment, physiological sensors adapted to the hemodialysis scenario are used to collect core physiological data from the patient in real time, specifically including blood component concentration, cardiac pumping burden level, and body fluid electrolyte balance data. A preset time series analysis method is used to perform noise reduction, smoothing, and trend extraction on the collected three types of physiological data, eliminating interference errors during data acquisition, and uncovering the changing patterns of various physiological indicators with dialysis time. Ultimately, the trends of multiple physiological indicators are obtained, providing accurate basic data support for subsequent patient physical status classification and risk prediction. In this step, data acquisition and trend analysis must meet the following parameter requirements: the physiological sensor acquisition frequency is 1~10Hz, preferably 2~5Hz. The selection criteria are: 1Hz can meet the real-time monitoring needs of basic physiological indicators, 10Hz can accurately capture the instantaneous fluctuations of physiological indicators, and 2~5Hz balances monitoring accuracy with equipment energy consumption and data transmission pressure, meeting the conventional requirements for real-time monitoring in clinical hemodialysis; the data acquisition error is ≤5%, ensuring that the collected physiological data truly reflects the patient's actual state, avoiding acquisition errors that lead to deviations in subsequent analysis and control, and meeting the accuracy standards of medical-grade sensors. The trend extraction accuracy of the time series analysis method is ≥98%, which is determined by ensuring accurate discovery of the changing patterns of physiological indicators and avoiding misjudgments of patient status due to trend extraction errors. The update cycle for the changing trends of multiple physiological indicators is 10~60s, preferably 20~40s, which is determined by ensuring timely updates of indicator changing trends and adapting to the dynamic changes in the patient's physiological status during hemodialysis. The 20~40s range balances real-time performance with clinical analysis efficiency. The signal-to-noise ratio after data denoising is ≥45dB, which is determined by effectively eliminating noise caused by environmental and equipment interference and ensuring the stability and reliability of physiological data.

[0026] Physiological sensors refer to medical-grade sensors adapted to hemodialysis scenarios and used to collect physiological data from patients. These include blood component sensors, cardiac function sensors, and electrolyte sensors. The measurement accuracy of these sensors meets medical industry standards. Specifically, the accuracy of blood component sensors is ≤ ±0.1 mmol / L (electrolytes) and ±1 g / L (proteins), while the accuracy of cardiac function sensors is ≤ ±5% (pump burden-related indicators). The criteria for these values ​​are: ensuring medical-grade accuracy of the collected data to meet the needs of clinical diagnosis and treatment control; sensor response time ≤ 100 ms to ensure rapid capture of changes in physiological indicators; and compatibility between the sensors and dialysis equipment with no risk of interference.

[0027] Blood component concentration refers to the concentration data of various core components in a patient's blood, including blood urea nitrogen (BUN), creatinine (Cr), hemoglobin (Hb), blood glucose, and albumin. The specific ranges and normal clinical ranges are as follows: blood urea nitrogen 3.2~7.1 mmol / L, creatinine 53~106 μmol / L (male), 44~97 μmol / L (female), and hemoglobin 110~150 g / L (female) and 120~160 g / L (male). The values ​​are based on the "Clinical Diagnosis and Treatment Guidelines for Hemodialysis". These ranges are normal reference values ​​for healthy individuals. During dialysis, dynamic changes in these values ​​need to be monitored to avoid abnormal fluctuations that could affect patient safety.

[0028] Cardiac pumping burden level refers to quantitative indicators that characterize a patient's cardiac pumping load. The core indicators include heart rate, blood pressure (systolic / diastolic), cardiac output, and left ventricular ejection fraction (LVEF). Specific value ranges and clinical warning ranges are as follows: heart rate 60-100 beats / min (warning range: <60 beats / min or >100 beats / min), systolic blood pressure 90-140 mmHg, diastolic blood pressure 60-90 mmHg (warning range: systolic <90 mmHg or >160 mmHg, diastolic <60 mmHg or >100 mmHg), cardiac output 4-8 L / min, and left ventricular ejection fraction ≥50% (warning range: <50%). These values ​​are based on clinical cardiovascular disease diagnostic standards. This range represents the normal cardiac pumping load range; values ​​exceeding the warning range indicate an increased cardiac pumping burden, requiring timely intervention.

[0029] Fluid and electrolyte balance data refers to the concentration data of electrolytes in a patient's body that maintain fluid balance, with potassium (K) being a key component. + ), sodium (Na + ), chlorine (Cl) - ), calcium (Ca 2+ ), bicarbonate (HCO3) - The specific normal ranges are: potassium 3.5~5.5 mmol / L, sodium 135~145 mmol / L, chloride 96~106 mmol / L, calcium 2.1~2.6 mmol / L, and bicarbonate 22~27 mmol / L. These values ​​are based on clinical electrolyte balance diagnosis and treatment standards. This range represents the normal electrolyte concentration in the human body. Electrolyte imbalance during dialysis can lead to risks such as arrhythmia and increased cardiac load, and is the core basis for physiological adjustment.

[0030] Time series analysis methods refer to analytical methods used to process continuous data (physiological index data) that change over time. In this embodiment, the ARIMA algorithm and LSTM algorithm are preferred, which can achieve noise reduction, smoothing and trend extraction of physiological data. The processing delay of the algorithm is ≤500ms. The basis for the value is to ensure rapid processing of real-time acquired data, timely output of changing trends, and adaptation to the real-time control requirements of the dialysis process. The fitting error of the algorithm is ≤3% to ensure the accuracy of trend analysis.

[0031] The trend of multiple physiological indicators refers to the dynamic changes of blood component concentration, cardiac pumping burden level, and body fluid electrolyte balance data with dialysis time (0-4 hours, the usual dialysis duration). It is presented in the form of trend curves, which can intuitively reflect the direction (rise, fall, and stability) and rate of change of the patient's physiological state. The quantitative range of the rate of change is 0-5% / min. The basis for the value is to match the normal range of physiological indicators during dialysis. Exceeding this rate indicates abnormal fluctuations and requires close attention.

[0032] In this embodiment, the time series analysis method employs an LSTM time series prediction model combined with a sliding window smoothing filter. The specific process is as follows: The model structure consists of an input layer with 3 dimensions (blood components, cardiac burden, electrolytes), 2 hidden layers with 64 neurons per layer, and an output layer containing trend prediction values. The sliding window width is 15 sampling points, the step size is 1, and mean filtering is used for noise reduction. Trend calculation involves fitting the time series curve using the least squares method to obtain the slope and variance of fluctuations, generating a trend vector. The implementation logic is: real-time sequence acquisition → normalization → noise reduction → feature extraction → trend fitting, outputting the changing trends of multiple physiological indicators. Those skilled in the art can directly reproduce this method based on the above model structure, window parameters, and fitting method.

[0033] Step S200: Based on the changing trends of multiple physiological indicators, a classification algorithm is used to classify the current physical state of the patient. If the classification results show an increase in the heart's pumping burden, the fluid and electrolyte balance data are integrated to determine the priority of physiological adjustments.

[0034] Using the trends of multiple physiological indicators obtained in step S100 as input, a preset classification algorithm is employed, combined with clinical patient physical status classification standards, to classify the current patient's physical status. The classification results include normal status, increased cardiac pumping burden, electrolyte imbalance, and abnormal blood components. If the classification results indicate that the patient is in a state of increased cardiac pumping burden, the body fluid electrolyte balance data collected in step S100 is further integrated to analyze the correlation between electrolyte imbalance and increased cardiac pumping burden. Based on the principle of clinical treatment priority, the priority of physiological adjustment is determined, and the physiological indicators that need to be adjusted first are identified, providing direction for subsequent risk prediction and treatment parameter optimization. In this step, the state classification and priority determination must meet the following parameter requirements: The preferred classification algorithm is Support Vector Machine (SVM) or Random Forest, with a classification accuracy ≥ 97%. This value is chosen to ensure accurate classification of the patient's physical state, avoiding misjudgments that could lead to treatment delays or over-intervention, and meeting the rigorous requirements of clinical diagnosis and treatment. Classification time is ≤ 10 seconds, preferably ≤ 5 seconds. This value is chosen to ensure rapid completion of state classification, timely response to changes in the patient's physiological state, and avoid delays in intervention. The classification threshold for increased cardiac pumping burden is: heart rate sustained for 10 seconds or more... An increased cardiac pumping burden is defined as follows: a blood pressure >100 beats / minute or <60 beats / minute; a systolic blood pressure >160 mmHg or <90 mmHg for more than 10 seconds; or a left ventricular ejection fraction <50%. Meeting any one of these conditions constitutes an increased cardiac pumping burden. The threshold values ​​are based on the "Guidelines for the Diagnosis and Treatment of Cardiovascular Diseases." This threshold accurately identifies abnormal cardiac pumping burden, avoiding misdiagnosis or missed diagnosis. Physiological adjustment priorities are divided into high, medium, and low levels, with a priority determination accuracy rate ≥98%. The criteria for these levels are to ensure reasonable priority ranking, prioritizing the treatment of high-risk physiological indicators to guarantee patient treatment safety.

[0035] Classification algorithms refer to artificial intelligence algorithms used to classify patients' physical conditions. They can accurately classify patients' physical conditions based on the changing trends of multiple physiological indicators and combined with clinical classification standards. The generalization error of the algorithm is ≤3%, and the criteria for setting the value are: to ensure that the algorithm can stably classify patients with different physical conditions and different diseases, adapt to the diverse clinical needs, and the inference latency of the algorithm is ≤100ms to ensure real-time classification response.

[0036] The current patient physical status refers to the patient's overall physiological state at a certain moment during hemodialysis. It is determined based on three major categories of indicators: blood components, cardiac pumping burden, and electrolyte balance. It is divided into normal state, increased cardiac pumping burden, electrolyte imbalance, abnormal blood components, and complex abnormal state (two or more abnormalities coexisting). The accuracy rate of each state determination is ≥97%. The basis for the values ​​is to ensure the accuracy of the state determination and provide a reliable basis for subsequent treatment adjustments.

[0037] Increased cardiac pumping burden refers to a patient's cardiac pumping load exceeding the normal range, manifested as abnormal fluctuations in heart rate and blood pressure, decreased cardiac output, and reduced left ventricular ejection fraction. This condition is defined as a duration of ≥10 seconds. The criteria for this value are: brief physiological fluctuations in clinical practice do not require intervention, while abnormalities lasting more than 10 seconds indicate abnormal cardiac load and require timely adjustment of treatment parameters to avoid triggering cardiac risks.

[0038] Physiological adjustment priority refers to the priority ranking of physiological indicators that need to be adjusted based on the risk level and scope of impact of the patient's physiological abnormalities. High priority indicators are those that directly affect cardiac safety (such as serum potassium, heart rate, and blood pressure), medium priority indicators are those that affect treatment efficacy (such as blood urea nitrogen and creatinine), and low priority indicators are those with minor abnormalities that do not affect safety (such as slight albumin fluctuations). The priority adjustment response time is ≤5s. The basis for this value is to ensure that high-risk indicators are treated first, to maximize the patient's cardiac safety, and to comply with the priority principles of clinical emergency care and treatment.

[0039] In this embodiment, the random forest classifier is fully implemented as follows: the input features are 8-dimensional vectors, including heart rate fluctuation, blood pressure slope, serum potassium concentration, electrolyte deviation, temporal change rate, fluctuation amplitude, response delay, and correlation coefficient; the model parameters are 50 decision trees, a maximum depth of 8, a minimum sample split of 4, and the Gini coefficient as the splitting criterion; the training set uses 120,000 clinical dialysis samples, labeled as normal / increased cardiac burden / electrolyte imbalance / combined abnormalities; the classification threshold is an output probability ≥ 0.65 for "increased cardiac pumping burden"; the inference process is feature input → decision tree voting → probability output → state classification. Those skilled in the art can directly reproduce this implementation using the above features, parameters, and training process.

[0040] Step S300: Obtain high-priority physiological indicators from the physiological adjustment priority, use a prediction algorithm to predict the potential cardiac risk level, and determine whether the potential cardiac risk level exceeds a preset threshold.

[0041] From the physiological adjustment priorities determined in step S200, high-priority physiological indicators (such as heart rate, blood pressure, and serum potassium concentration) are extracted, and their changing trends are used as input. A preset prediction algorithm is used, combined with clinical cardiac risk assessment standards, to quantitatively predict the patient's potential cardiac risk level in the next 5 to 30 minutes. A preset potential cardiac risk level threshold is set, and the predicted potential cardiac risk level is compared with the threshold to determine whether the patient has potential cardiac risk. If the threshold is exceeded, treatment parameter optimization needs to be initiated immediately; if the threshold is not exceeded, changes in physiological indicators are continuously monitored. In this step, risk prediction and threshold determination must meet the following parameter requirements: The preferred prediction algorithm is LSTM or GRU, with a prediction accuracy ≥95%. This value is chosen to ensure accurate prediction of potential cardiac risk, avoiding missed high-risk cases and misjudged low-risk cases, meeting the stringent requirements of clinical risk control. The prediction duration is 5–30 minutes, preferably 10–20 minutes. This value is chosen because 5 minutes allows for short-term risk warning, 30 minutes allows sufficient intervention time, and 10–20 minutes balances timely warning and intervention feasibility. The potential cardiac risk level is quantified in the range of 0–10 points (0–3 points indicates low risk). The risk levels are categorized as follows: 4-6 points indicate medium risk, 7-10 points indicate high risk. The preset risk threshold is 7 points, meaning a potential cardiac risk level ≥7 points is considered exceeding the threshold. This threshold is based on the Clinical Cardiac Risk Assessment Scale, where a score of 7 or higher indicates high risk. In this case, the probability of the patient developing complications such as arrhythmia or heart failure is ≥30%, requiring immediate intervention. The prediction error is ≤0.5 points, ensuring the accuracy of risk prediction and avoiding delays or over-intervention due to prediction errors. The risk assessment time is ≤8 seconds, preferably ≤5 seconds, ensuring rapid risk assessment, timely initiation of subsequent intervention procedures, and patient safety.

[0042] High-priority physiological indicators refer to those physiological indicators that are selected from the physiological adjustment priority, directly affect the patient's cardiac safety, and require priority monitoring and adjustment. The core indicators include heart rate, systolic blood pressure / diastolic blood pressure, serum potassium concentration, and left ventricular ejection fraction. Abnormal fluctuations in these indicators can directly lead to cardiac risks (such as arrhythmia and heart failure). The monitoring frequency of these indicators is increased to 3~10Hz. The basis for the values ​​is to ensure real-time monitoring of key indicators, capture abnormal fluctuations in a timely manner, and provide accurate data for risk prediction.

[0043] The prediction algorithm refers to an artificial intelligence algorithm used to predict the potential cardiac risk level of patients. It can quantitatively predict cardiac risk in the future based on the changing trends of high-priority physiological indicators. The prediction delay of the algorithm is ≤1 second. The criteria for the value are: to ensure rapid output of risk prediction results to provide sufficient time for subsequent intervention, and the recall rate of the algorithm is ≥96% to ensure that no high-risk situations are missed.

[0044] Potential cardiac risk level refers to the quantitative value of the probability that a patient will develop cardiac-related complications (such as arrhythmia, heart failure, and myocardial ischemia) within a future period of time (5-30 minutes). The value ranges from 0 to 10 points. The higher the score, the higher the risk. When the score is 7-10 (high risk), the probability of complications is ≥30%; when the score is 4-6 (medium risk), the probability of complications is 10%-30%; when the score is 0-3 (low risk), the probability of complications is <10%. The value is determined by combining clinical cardiac risk assessment data. This quantitative standard can accurately reflect the patient's potential cardiac risk and provide a clear basis for intervention decisions.

[0045] The preset threshold is a critical value used to determine whether intervention is needed to address the potential cardiac risk level. In this embodiment, it is preset to 7 points. That is, when the potential cardiac risk level is ≥7 points, it is determined to be too high and treatment parameter optimization needs to be initiated immediately. The value is based on the fact that clinical data shows that patients with a score of 7 or above have a significantly increased risk of cardiac complications. Timely intervention can reduce the incidence of complications by more than 60%, which meets the core requirements of clinical risk prevention and control.

[0046] In this embodiment, the Gradient Boosting Regression Tree (GBDT) prediction model is fully implemented as follows: Input features consist of 12 dimensions: high-priority physiological time-series features, circulatory system features, and historical risk labels; the model structure comprises 100 decision trees, a learning rate of 0.05, and a maximum tree depth of 5; the loss function is mean squared error (MSE); the training set consists of 150,000 dialysis time-series data points labeled with cardiac risk; the output is a risk score of 0-10, with a score ≥7 indicating an out-of-limit threshold; feature importance is 28% for heart rate, 24% for blood pressure, 20% for serum potassium, 16% for time-series fluctuation, and 12% for vascular resistance. Those skilled in the art can directly reproduce risk prediction using the above structure, loss function, and training set.

[0047] Step S400: If the potential cardiac risk level exceeds the preset threshold, the prediction results and real-time data are fused through the feedback control mechanism to obtain the optimized values ​​of the treatment parameters. Based on the optimized values ​​of the treatment parameters, the optimization algorithm is used to search for the best combination of hemodialysis machine settings to obtain a dynamic physiological adjustment plan.

[0048] If the determination result of step S300 is that the potential cardiac risk level exceeds the preset threshold (≥7 points), then the feedback control mechanism is activated to fuse the risk prediction result with the real-time physiological data collected in step S100, and calculate the optimized value of each treatment parameter in combination with the clinical dialysis treatment parameter adjustment standard. For the obtained optimized treatment parameter values, a preset optimization algorithm is used to search for the best combination of hemodialysis machine settings that is suitable for the current physiological state of the patient, clarify the specific adjustment value and adjustment time of each treatment parameter, and finally obtain a dynamic physiological adjustment plan to achieve adaptive optimization of treatment parameters and reduce the patient's cardiac risk. In this step, parameter optimization and scheme generation must meet the following parameter requirements: The response time of the feedback control mechanism should be ≤10s, preferably ≤6s. The basis for this value is to ensure rapid response to high-risk situations, timely generation of optimized schemes, and avoid delays in intervention. The calculation error of the optimized treatment parameter value should be ≤5%. The basis for this value is to ensure the accuracy of the optimized value, avoid parameter adjustment deviations leading to deterioration of the patient's physiological state, and meet the accuracy requirements for clinical treatment parameter adjustment. The preferred optimization algorithm is genetic algorithm or particle swarm optimization algorithm, with an algorithm search accuracy rate ≥98%. The basis for this value is to ensure that the searched hemodialysis machine setting combination is optimal and adapts to the current patient's physiological state. The algorithm search time should be ≤20s, preferably ≤15s. The executable rate of the dynamic physiological adjustment scheme should be ≥99%. The basis for this value is to ensure that the scheme can be directly executed through the hemodialysis machine interface, avoiding incompatibility between the scheme and the equipment. The adjustment range of the scheme is 5%~20% of the original parameters. The basis for this value is to avoid excessive parameter adjustment leading to drastic fluctuations in the patient's physiological state. An adjustment range of 5%~20% can achieve stable optimization, conforming to the principle of gradual clinical treatment.

[0049] Feedback control mechanisms are mechanisms used to achieve closed-loop control of "monitoring-prediction-optimization-execution-remonitoring". They can integrate potential cardiac risk prediction results, real-time physiological data and optimized values ​​of treatment parameters to dynamically adjust control strategies, ensuring that treatment parameters are always adapted to the patient's physiological state. The control accuracy of the mechanism is ≤5%, and the values ​​are determined based on ensuring the accuracy of the control strategy and avoiding parameter adjustment deviations. The stability of the mechanism is ≥99.9%, ensuring long-term operation without failure.

[0050] The optimized treatment parameters refer to the optimal adjustment values ​​of each treatment parameter of the hemodialysis machine calculated based on feedback control mechanisms, combined with risk prediction results and real-time physiological data. The core treatment parameters include dialysate flow rate, blood flow rate, dialysis time, dehydration volume, and dialysate electrolyte concentration. The specific optimization range is as follows: dialysate flow rate 300~800ml / min (optimization range ±50~100ml / min), blood flow rate 150~300ml / min (optimization range ±20~50ml / min), dehydration volume 0.5~5kg (optimization range ±0.2~1kg), and dialysate potassium concentration 2.0~4.0mmol / L (optimization range ±0.5~1.0mmol / L). The values ​​are based on the "Guidelines for Hemodialysis Treatment". This optimization range can reduce cardiac risk while ensuring the effectiveness of dialysis treatment and avoid parameter adjustments exceeding the safe range.

[0051] The optimization algorithm refers to the algorithm used to search for the best combination of hemodialysis machine settings. Based on the optimized values ​​of treatment parameters and the parameter constraints of the hemodialysis machine, it searches for the parameter combination that is suitable for the current physiological state of the patient and can minimize cardiac risks. The convergence speed of the algorithm is ≤10s. The criteria for the value are: to ensure that the optimal parameter combination is found quickly and treatment adjustments are initiated in a timely manner. The stability of the algorithm is ≥99% to avoid failures during the search process.

[0052] The combination of settings for a hemodialysis machine refers to the specific combination of values ​​for the core treatment parameters of the hemodialysis machine. This includes the specific values ​​of parameters such as dialysate flow rate, blood flow rate, dehydration volume, dialysis time, and dialysate electrolyte concentration. The combination must meet the constraints of the equipment parameters and the requirements of clinical treatment. The compatibility rate of the parameter combination should be ≥98%. The basis for the selection of values ​​is to ensure that the combination can adapt to the patient's current physiological state and meet the operating requirements of the hemodialysis machine to avoid equipment failure.

[0053] A dynamic physiological adjustment plan refers to a complete treatment adjustment plan that includes the optimal combination of hemodialysis machine settings, the timing of parameter adjustments, the adjustment range, and the key points of subsequent monitoring. The plan specifies the adjustment time (delay ≤30s) and the values ​​before and after the adjustment for each treatment parameter, and also specifies the key points of subsequent physiological monitoring. The execution delay of the plan is ≤30s, and the basis for this value is to ensure rapid execution of the plan, timely intervention of potential cardiac risks, and an effectiveness of ≥95%, ensuring that the patient's cardiac risk is significantly reduced after adjustment.

[0054] In this embodiment, the feedback control mechanism is an improved PID closed-loop control + model predictive control (MPC), specifically implemented as follows: the controller type is a dual-loop PID, with an outer loop for physiological state deviation and an inner loop for device execution; the PID parameters are P=1.2, I=0.08, and D=0.15; the fusion logic uses a weighted sum of 0.6 for the prediction result and 0.4 for the real-time data as the control deviation input; the control objective is for the cardiac burden score, electrolyte concentration, and blood pressure to converge to a safe range; the output limit is within ±15% of the adjustment range of the treatment parameters. Those skilled in the art can directly reproduce this implementation based on the above control type, parameters, and fusion logic.

[0055] In this embodiment, the Particle Swarm Optimization (PSO) algorithm is fully implemented as follows: the optimization variables are blood flow rate, dialysate flow rate, dialysis pressure, electrolyte concentration, and ultrafiltration rate; the fitness function is F = 0.4 × (1 - cardiac risk score) + 0.3 × electrolyte balance + 0.2 × dialysis adequacy - 0.1 × adjustment range; the constraints are blood flow rate 150~300 ml / min, dialysate flow rate 300~800 ml / min, dialysis pressure -50~+50 mmHg, serum potassium 2.0~4.0 mmol / L, and ultrafiltration volume 0.5~5 L; the particle swarm parameters are particle number 30, iterations 50, inertia weight 0.7, cognitive coefficient 1.5, and social coefficient 1.5; the velocity and position are updated according to the standard PSO formula until convergence. Those skilled in the art can directly reproduce the optimization by following the above objective function, constraints, and parameters.

[0056] Step S500: Extract specific parameter values ​​from the dynamic physiological adjustment plan, transmit the specific parameter values ​​through the hemodialysis machine interface, determine whether the transmission is successful, if the transmission is successful, monitor the changes of subsequent multiple physiological indicators, and reprocess the newly collected data using time series analysis methods to obtain updated physiological change trends.

[0057] From the dynamic physiological adjustment plan obtained in step S400, the specific parameter values ​​of each treatment parameter are extracted and transmitted to the control system of the hemodialysis machine through a preset hemodialysis machine interface to achieve automatic adjustment of the treatment parameters. The parameter transmission process is monitored in real time to determine whether the transmission is successful. If the transmission fails, it is retransmitted (up to 3 times, with a 5-second interval between each retransmission). If the transmission is successful, subsequent multi-physiological index data of the patient are continuously collected through physiological sensors. The newly collected physiological data is processed using the same time series analysis method as in step S100 to eliminate interference, extract trends, and obtain updated physiological change trends, providing a basis for whether further adjustments are needed. In this step, parameter transmission and index updates must meet the following parameter requirements: the hemodialysis machine interface transmission rate ≥ 1 Mbps, based on the principle of ensuring rapid parameter transmission and avoiding transmission delays that could lead to delayed treatment adjustments; and the bit error rate of the interface transmission ≤ 10%.-6 The values ​​are determined based on the following criteria: ensuring that the transmitted parameter values ​​are not distorted and avoiding abnormal device parameter settings due to transmission errors; the parameter transmission success rate is ≥99.5%, determined based on the following criteria: ensuring that treatment parameters can be transmitted smoothly and avoiding intervention delays due to transmission failures; a maximum of 3 retransmissions are allowed, and an alarm signal is issued if more than 3 retransmissions are attempted; the acquisition frequency of newly acquired data is consistent with that of step S100 (1~10Hz, preferably 2~5Hz), determined based on the following criteria: ensuring real-time monitoring of changes in the patient's physiological state after adjustment; the update cycle of updating physiological change trends is ≤30s, preferably ≤20s, determined based on the following criteria: ensuring timely capture of changes in physiological indicators after parameter adjustment and judgment of adjustment effects; and the accuracy rate of trend extraction is ≥98%, determined based on the following criteria: ensuring accurate trend analysis and providing reliable support for subsequent adjustment decisions.

[0058] Specific parameter values ​​refer to the specific values ​​of each treatment parameter extracted from the dynamic physiological adjustment plan and used to set the hemodialysis machine. Each parameter value is within the allowable setting range of the equipment and the clinical safety range. The accuracy of the parameter values ​​is 0.1 ml / min (flow rate), 0.1 kg (dehydration volume), and 0.1 mmol / L (electrolytes). The basis for the selection is to ensure the accuracy of the parameter settings, meet the setting requirements of medical-grade equipment, and avoid parameter errors that may lead to poor treatment results or patient discomfort.

[0059] The hemodialysis machine interface refers to the communication interface used to connect the monitoring and control system with the hemodialysis machine. It adopts a medical-grade standard interface (such as RS485 or Ethernet interface) and supports bidirectional transmission of treatment parameters (transmission to the device and feedback of the device's operating status). The interface has a compatibility of ≥99%, and the values ​​are based on the following criteria: ensuring that the interface can be adapted to various commonly used clinical hemodialysis machines to achieve stable parameter transmission, and the interface response time is ≤100ms to ensure real-time transmission.

[0060] Whether the transmission is successful refers to the transmission confirmation status returned by the system after the specific values ​​of the treatment parameters are transmitted to the hemodialysis machine control system through the interface. If the system receives and confirms the parameter value, the transmission is considered successful. If it does not receive or receives an error, the transmission is considered to have failed. The response time for transmission determination is ≤1 second. The basis for this value is to ensure rapid determination of the transmission status, timely handling of transmission failures, and avoid delays in treatment adjustments.

[0061] Subsequent changes in multiple physiological indicators refer to the dynamic changes in blood component concentration, cardiac pumping burden, and body fluid electrolyte balance data after the adjustment of treatment parameters. The focus is on monitoring changes in high-priority physiological indicators (heart rate, blood pressure, and blood potassium), with the monitoring frequency increased to 3-10 Hz. The basis for the values ​​is to ensure timely judgment of the effect of parameter adjustment. If the indicators do not improve or worsen, further adjustments should be initiated immediately.

[0062] The updated physiological change trend refers to the change trend obtained by processing newly collected physiological data after parameter adjustment using time series analysis methods. It is compared with the change trend of multiple physiological indicators before adjustment, which can intuitively reflect the effect of treatment parameter adjustment (indicator improvement, no change, deterioration). The accuracy rate of trend comparison is ≥99%. The basis for the value is to ensure accurate judgment of the adjustment effect and provide a reliable basis for subsequent adjustment decisions.

[0063] The parameter transmission uses a medical-grade standard interface. If the transmission fails, it will be automatically retransmitted up to 3 times. After the transmission is successful, the trend will be updated using the same time series analysis method.

[0064] Step S600: Based on the updated physiological change trend, repeat the classification algorithm process to determine whether further physiological adjustments are needed. Obtain confirmation signals from the further physiological adjustment needs, apply physiological optimization values ​​through feedback control mechanisms, and obtain the final stable state of the patient's treatment.

[0065] Using the updated physiological change trend obtained in step S500 as input, the classification algorithm process of step S200 is repeated to reclassify the patient's physical state after parameter adjustment, and to determine whether the patient's physiological state has improved and whether further physiological adjustments are needed. If the classification result shows that the patient still has an increased cardiac pumping burden and the potential cardiac risk has not been reduced, a confirmation signal (automatic confirmation or manual confirmation by medical staff) is obtained from the need for further physiological adjustments. The feedback control mechanism is used again to calculate new physiological optimization values ​​in combination with the updated physiological change trend, and the treatment parameters are adjusted. The above cyclical adjustment process is repeated until the patient's physiological indicators tend to stabilize and the potential cardiac risk level is lower than the preset threshold (<7 points), and finally the patient's treatment stability is obtained, ensuring that hemodialysis treatment is completed safely and efficiently. In this step, the cyclic adjustment and stable state achievement must meet the following parameter requirements: The time taken for the repetitive classification algorithm should be ≤10s, preferably ≤5s, based on the principle of ensuring rapid state reclassification and timely determination of whether further adjustment is needed; the accuracy rate for determining the need for further physiological adjustment should be ≥98%, based on the principle of ensuring accurate determination of whether further adjustment is needed, avoiding over-adjustment or under-adjustment; the response time of the confirmation signal should be ≤3s (automatic confirmation) and ≤30s (manual confirmation), based on the principle of ensuring rapid initiation of subsequent adjustment processes and avoiding delays in intervention; the maximum number of cyclic adjustments is 5. The criteria for determining the patient's physiological state are as follows: if the patient's physiological state does not improve after 5 adjustments, it is suggested that medical staff intervene manually to avoid an infinite loop that could increase the patient's risk. The threshold for determining a stable treatment state is: potential cardiac risk level <7 points, heart rate 60~100 beats / minute, blood pressure 90~140 / 60~90 mmHg, electrolyte concentration within the normal range (potassium 3.5~5.5 mmol / L, etc.), and stable for ≥5 minutes. The criteria for determining this state are: referring to the clinical standards for stable hemodialysis treatment, this threshold can ensure that the patient's physiological state is stable, cardiac risk is controllable, and treatment safety requirements are met.

[0066] The repeated classification algorithm process refers to the process of inputting the updated physiological change trend into the classification algorithm used in step S200 to reclassify the patient's physical condition. The classification criteria are consistent with those in step S200 to ensure the consistency and accuracy of the classification results. The classification repeatability accuracy is ≥99%, and the basis for this value is to avoid misjudgment caused by inconsistent classification criteria and to ensure the consistency of adjustment decisions.

[0067] Further physiological adjustment needs refer to the need for continued adjustment of treatment parameters when the patient's physical condition has not yet reached the stable standard after parameter adjustment (potential cardiac risk ≥7 points, abnormal physiological indicators). The determination of the need should be combined with the updated physiological change trend and clinical standards, with a judgment error of ≤1%. The basis for the value is to ensure accurate determination of the need and avoid over-adjustment when no adjustment is needed, or failure to adjust when adjustment is needed.

[0068] Confirmation signals are signals used to initiate further physiological adjustments. They are divided into automatic confirmation and manual confirmation by medical staff. Automatic confirmation is suitable for situations with minor abnormalities in physiological indicators and controllable risks, while manual confirmation is suitable for situations with higher risks and the need for medical staff intervention. The accuracy rate of confirmation signals is ≥99.5%, and the basis for determining the value is to ensure that the confirmation signal is correct and to avoid false triggering or missed triggering of further adjustments.

[0069] The physiological optimization value refers to the optimized value of the treatment parameters recalculated through a feedback control mechanism in response to the need for further physiological adjustments and in combination with updated physiological change trends. The adjustment range of the optimization value is 3% to 15% of the original parameter. The basis for the value is: based on the effect of the previous adjustment, the parameter is slightly optimized to avoid drastic fluctuations in physiological state and to conform to the principle of gradual clinical treatment.

[0070] Ultimately, a stable patient status refers to a condition where, after cyclical adjustments, the patient's physiological indicators reach the clinical stability standard, the potential cardiac risk level is below the preset threshold (<7 points), and the blood component concentration, cardiac pumping burden level, and body fluid electrolyte balance data are all within the normal range and remain stable for ≥5 minutes. At this point, dialysis treatment can proceed normally without further parameter adjustments, and the stability maintenance rate is ≥98%. The basis for this value is to ensure that the patient remains in a safe and stable state throughout the dialysis process, reduce the risk of cardiac complications, and meet the core objectives of hemodialysis treatment.

[0071] The confirmation signal can be either automatic or manual. The process can be repeated up to 5 times until the physiological indicators are stable for ≥5 minutes.

[0072] The real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine provided in this embodiment employs a complete implementation logic of adaptive and dynamic adjustment: 1. Multi-source data fusion rules: Weighted fusion is adopted, with 60% for predicted risk and 40% for real-time physiological data. The fused data is used as the control input.

[0073] 2. Adaptive adjustment of trigger conditions: cardiac risk score ≥7; serum potassium exceeding 3.5~5.5mmol / L; heart rate consistently >100 beats / minute or <60 beats / minute; blood pressure >160 / 100mmHg or <90 / 60mmHg.

[0074] 3. Dynamic adjustment strategy: High risk → rapid convergence adjustment (10% range); Medium risk → steady adjustment (5% range); Low risk → maintain and fine-tune (2% range).

[0075] 4. Closed-loop termination conditions: Physiological indicators remain stable within the safe range for 3 consecutive cycles (≥5 minutes), and the risk score is <7 points.

[0076] Furthermore, the real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine provided in this embodiment includes step S100 as follows: Step S110: Use physiological sensors to collect data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance in real time, and construct a multidimensional time series based on the data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance.

[0077] Multidimensional time series sequences are derived using the following formula: (1) In formula (1), This is a multidimensional time series, constructed by integrating data collected from physiological sensors. The values ​​are time series sets containing data on blood component concentration, cardiac pumping workload, and body fluid electrolyte balance, with the following dimensions: ( (Number of samples collected) Operators are constructed for time series, derived from time series data processing logic, with no value range, and are used to integrate multiple types of physiological data into a unified multidimensional time series; Blood component concentration, unit is (mmol / L), sourced from real-time data collected by physiological sensors; the value range is determined based on the type of blood component (e.g., blood glucose 3.9-6.1). Urea 2.5-7.5 This describes the concentration levels of various components in a patient's blood. The level of cardiac pumping workload, measured in units of... (mmHg), sourced from real-time data collected by a physiological sensor, with a range of 60-180 mmHg. This describes the intensity of the workload during the heart's pumping action. This is data on body fluid and electrolyte balance, in units of... (mmol / L), sourced from real-time data collected by physiological sensors; the value range is determined based on the electrolyte type (e.g., potassium 3.5-5.5). Sodium 135-145 This describes the electrolyte balance in the patient's body fluids. The control logic of formula (1) involves real-time acquisition of the concentration of blood components by physiological sensors. Cardiac pumping burden level and body fluid electrolyte balance data The time-series sequence construction operator is invoked to integrate three different types of physiological data according to their collection timestamps, thereby constructing a unified multidimensional time-series sequence. This lays the data foundation for subsequent fluctuation feature extraction and trend analysis.

[0078] Physiological sensors can be integrated into wearable devices, such as smart bracelets, to monitor users' physiological data in real time. Specifically, these sensors acquire blood component concentrations, such as blood glucose and blood oxygen saturation, using optical detection methods; simultaneously, they assess the level of cardiac pumping burden using heart rate variability analysis, such as calculating cardiac output; and furthermore, they measure body fluid electrolyte balance data, such as sodium and potassium ion concentrations, using ion-selective electrodes. This data is acquired synchronously, sampled once per second, forming a time-series sequence. Subsequently, a multidimensional time-series sequence is constructed based on this data, for example, using blood component concentration as the first dimension, cardiac pumping burden as the second dimension, and body fluid electrolyte balance data as the third dimension, thus forming a three-dimensional time-series matrix for subsequent analysis.

[0079] Step S120: Perform sliding window processing on the multidimensional time series to extract fluctuation features. If the fluctuation features exceed a preset threshold, use the correlation matrix to determine the evolution trajectory.

[0080] The evolutionary trajectory is determined using the following formula: (2) In formula (2), The evolution trajectory is derived from the analysis of fluctuation characteristics and the calculation of the correlation matrix, with the value being a set of change trajectories of a multidimensional time series. This is a sliding window processing operator, derived from time series feature extraction logic. It has no value range and is used to perform sliding window sampling on multi-dimensional time series sequences to extract fluctuation features. The width of the sliding window is determined by the timing processing configuration and takes a positive integer value (e.g., ...). (Each acquisition cycle), describing the number of samples contained in the sliding window; The sliding window step size is configured from the time series processing source and takes a positive integer value (e.g., ...). (Each acquisition cycle) describes the number of samples that the sliding window moves in each acquisition cycle; This is a correlation matrix, derived from physiological data correlation analysis. It is a 3×3 real number matrix that describes the correlation strength among blood component concentration, cardiac pumping burden, and body fluid electrolyte balance data. This is an association matrix mapping operator, derived from trajectory analysis logic, with no value range, used to determine the evolution trajectory of physiological data through the association matrix; A preset threshold is set for the fluctuation characteristics, derived from physiological monitoring standards. The value range is determined according to the monitoring indicators (e.g., 0.5~2.0, after dimensionless processing), and is used to determine whether the fluctuation characteristics are abnormal. The default trajectory generation operator, derived from time series analysis logic, has no value range and is used to generate the default evolution trajectory when the fluctuation characteristics are normal. The control logic of formula (2) uses a sliding window to process the operator, with the window width... Step length For multidimensional time series The data is processed to extract fluctuation characteristics of physiological data; the norm of the fluctuation characteristics is calculated, and if the norm exceeds a preset threshold... This indicates abnormal fluctuations, as confirmed by the correlation matrix. Invoking the correlation matrix mapping operator to determine the evolution trajectory of physiological data. If the fluctuation characteristics are normal, the default trajectory generation operator is invoked to generate the default evolution trajectory.

[0081] When performing sliding window processing on the multidimensional time series, the window size is set to 30 seconds and the step size to 10 seconds. Within each window, the standard deviation and Fourier transform coefficients are calculated to extract fluctuation characteristics, such as the amplitude and frequency of concentration fluctuations. If the fluctuation characteristics exceed a preset threshold, for example, a standard deviation greater than 0.5, it indicates an abnormal physiological state. In this case, the correlation matrix is ​​used to determine the evolution trajectory. Specifically, the correlation matrix is ​​a matrix constructed by calculating the Pearson correlation coefficients between data in each dimension. For example, matrix elements represent the correlation values ​​between blood component concentration and cardiac pumping burden levels, ranging from -1 to 1. Then, through eigenvector analysis of the correlation matrix, the evolution trajectory of the data is tracked, such as the transition path from an initial stable state to a high-burden state, thereby revealing the dynamic changes in physiological indicators. This method helps to identify potential health risks early. For example, in exercise monitoring scenarios, when the correlation matrix shows an increased correlation between cardiac burden and electrolyte imbalance, the real-time monitoring and adaptive control system can warn users to adjust their activity intensity.

[0082] Step S130: Calculate the trend vector based on the evolution trajectory, and obtain the change trend of multiple physiological indicators by mapping the trend vector to a multi-dimensional coordinate system.

[0083] The trends of multiple physiological indicators are derived using the following formula: (3) In formula (3), The trend of multiple physiological indicators is derived from the multidimensional coordinate system mapping of the trend vector, and the value is a set of trends that includes the direction and rate of change of various physiological indicators. It is a coordinate system mapping operator, derived from trend visualization and quantification logic, with no value range, used to map trend vectors to a multi-dimensional coordinate system to generate changing trends; This is a trend vector extraction operator, derived from trajectory analysis logic, with no value range, used to extract trend vectors from evolution trajectories. The system is a multidimensional coordinate system, derived from physiological indicators, with values ​​set as a three-dimensional coordinate system using blood component concentration, cardiac pumping burden, and body fluid electrolyte balance as coordinate axes. The control logic of formula (3) is based on the trend vector extraction operator derived from the evolution trajectory. The trend vector is calculated, which contains the direction and rate of change of physiological data; the trend vector is then input into the coordinate system mapping operator in a preset multidimensional coordinate system. The data is mapped and quantified to present the changing patterns of various physiological indicators, ultimately yielding the trends of multiple physiological indicators. .

[0084] When calculating the trend vector based on the evolution trajectory, the least squares method can be used to fit the trajectory curve to obtain vector components, such as slope representing the rate of change and direction representing the trend directional. Subsequently, the trend vector is mapped to a multi-dimensional coordinate system, for example, projecting the trend vector onto a coordinate system with blood concentration as the x-axis, cardiac workload as the y-axis, and electrolyte balance as the z-axis, to obtain a visual representation of the changing trends of multiple physiological indicators. This facilitates medical personnel in assessing the patient's condition and enabling precise intervention.

[0085] Preferably, the real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine provided in this embodiment includes step S200: Step S210: Obtain the changing trends of multiple physiological indicators and map them to a high-dimensional feature space, and extract feature vectors that reflect fluctuations in physiological function.

[0086] The feature vector reflecting fluctuations in physiological function is extracted using the following formula: (4) In formula (4), The feature vector of physiological function fluctuations originates from high-dimensional feature space mapping and feature extraction. Its value is an n-dimensional real number vector (n is the feature dimension, such as n=8~12), which describes the core features of physiological function fluctuations. This is a feature vector extraction operator, derived from feature engineering logic, with no value range, used to extract core feature vectors from a high-dimensional feature space. It is a high-dimensional feature space mapping operator, derived from feature space transformation logic, with no value range, used to map the changing trends of multiple physiological indicators to a high-dimensional feature space; For high-dimensional feature spaces, the source feature space configuration takes values ​​higher than three dimensions (e.g., 8-12 dimensions) to more accurately capture physiological function fluctuations. The control logic of formula (4) is to call the high-dimensional feature space mapping operator to map the changing trends of multiple physiological indicators. Mapped to a predefined high-dimensional feature space This process achieves high-dimensional feature representation of physiological trends; then, through feature vector extraction operators, it extracts core features that reflect fluctuations in physiological function from the high-dimensional feature space, generating a feature vector of physiological function fluctuations. This provides feature inputs for subsequent classification of patients' physical conditions.

[0087] This extraction process first involves data standardization, such as normalizing the blood glucose trend sequence to a mean of 0 and a standard deviation of 1. Then, wavelet transform is applied to decompose the signal, separating the low-frequency trend and high-frequency noise components. Next, the energy spectrum of features such as wavelet coefficients is calculated. These feature vectors can capture subtle patterns of physiological fluctuations, such as the dynamic changes in cardiac workload caused by a sharp rise in blood glucose after exercise. In this way, real-time monitoring and adaptive control systems can extract key information from complex data, supporting the input of subsequent classifiers and enabling accurate assessment of patient status. This allows for early detection of potential risks in medical monitoring, such as preventing cardiovascular events.

[0088] Step S220: Input the feature vector into a pre-trained random forest classifier, and obtain the current patient's physical state by performing a non-linear mapping on the feature vector.

[0089] The patient's current physical condition is determined by the following formula: (5) In formula (5), The current patient's physical condition is classified using a random forest classifier, with values ​​set to preset physical condition categories (such as "increased cardiac pumping burden", "electrolyte imbalance", "normal state" etc.). This is a random forest classification operator, derived from a pre-trained random forest classification model. It has no value range and is used to perform non-linear mapping on feature vectors, outputting body state classification results. For the pre-trained random forest classifier, the source model is trained (based on historical patient physiological data), and the values ​​are model parameters that include the decision tree set and classification threshold. The body state classification set is derived from clinical diagnostic criteria and takes the value of a preset set of all patient body state categories. The control logic of formula (5) is to extract the physiological function fluctuation feature vector. Input a pre-trained random forest classifier The feature vector is nonlinearly mapped using a classification operator, and combined with a pre-defined set of body state classifications. Output the current patient's physical status This enables accurate identification of the patient's physical condition.

[0090] When the feature vector is input into a random forest classifier, which consists of multiple decision trees, the patient's physical condition is obtained through a non-linear mapping using a voting mechanism, such as identifying an increased burden on the heart.

[0091] Step S230: If the current patient's physical condition is identified as an increased burden on the heart's pumping capacity, then extract the corresponding body fluid and electrolyte balance data and construct a multidimensional correlation matrix.

[0092] Construct a multidimensional correlation matrix using the following formula: (6) In formula (6), It is a multidimensional correlation matrix, constructed from the correlation between body fluid electrolyte balance data and body status. It is a k×k real number matrix (k is the number of electrolyte types, such as k=4-6), describing the correlation strength between different electrolytes and between electrolytes and the heart's pumping burden. Operators for constructing correlation matrices are derived from correlation analysis logic, have no value range, and are used to construct multidimensional correlation matrices for body fluid and electrolyte balance data. The association strength calculation rule is derived from clinical data statistics, and the value is a set of association strength calculation formulas and thresholds. The control logic of formula (6) is that when the random forest classifier outputs the current patient's physical state... When "the workload of the heart pumping blood increases", extract the corresponding body fluid and electrolyte balance data. Combined with the preset correlation strength calculation rules The correlation matrix construction operator is invoked to analyze the correlation between different electrolytes and between electrolytes and the cardiac pumping burden, and to construct a multidimensional correlation matrix. This provides a basis for determining the priority of subsequent physiological adjustments.

[0093] If the condition indicates increased cardiac pumping burden, fluid electrolyte balance data, such as sodium ion concentration, are extracted, and a multidimensional correlation matrix is ​​constructed, where matrix elements represent the correlation between different indicators. In one embodiment, the correlation matrix is ​​quantified using a weighted allocation model, for example, by employing the Analytic Hierarchy Process method to assign weights, determining the priority of physiological adjustments, thereby guiding interventions such as electrolyte supplementation.

[0094] Step S240: Quantify the physiological adjustment priority based on the multidimensional correlation matrix using a weight allocation model.

[0095] The priority of physiological adjustments is determined using the following formula: (7) In formula (7), The physiological adjustment priority is derived from the quantification of the multidimensional correlation matrix, and the value is an ordered set containing the adjustment priority of each physiological indicator (e.g., "potassium ion adjustment > sodium ion adjustment > water adjustment"). This is a weight allocation and quantization operator, derived from the weight allocation model, with no value range, used for quantization of the correlation matrix to determine adjustment priorities; The weighting model is derived from clinical treatment experience and model training, and its values ​​are model parameters that include the weight coefficients of each physiological indicator. The priority ranking rule is derived from clinical treatment guidelines, and its values ​​are the criteria and logic for prioritizing the ranking. The control logic of formula (7) is to construct a multidimensional correlation matrix. Input weight allocation model The correlation strength of each indicator in the correlation matrix is ​​quantified by weight allocation and quantization operators, combined with a preset priority ranking rule. The adjustment priorities of each physiological indicator are ranked to obtain the final physiological adjustment priority. This clarifies the key directions for subsequent adjustments.

[0096] The process of acquiring trends in multiple physiological indicators involves real-time data collection via wearable devices, such as smartwatches monitoring blood sugar levels and heart rate changes, which are then mapped to a high-dimensional feature space. For example, this trend data can be converted into high-dimensional vectors to extract features such as fluctuation amplitude and periodicity. Specifically, the high-dimensional feature space can be viewed as a multi-dimensional coordinate system, where each dimension corresponds to a physiological indicator, such as blood oxygen saturation as one dimension and cardiac output as another. Principal component analysis is used to reduce the dimensionality of the original data, preserving key variation information, thereby extracting feature vectors reflecting fluctuations in physiological function.

[0097] Furthermore, the real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine provided in this embodiment includes step S300 as follows: Step S310: Extract the top-ranked categories from the physiological adjustment priorities to determine high-priority physiological quantities.

[0098] High-priority physiological parameters are derived using the following formula: (8) In formula (8), For high-priority physiological quantities, the selection is based on the physiological adjustment priority of the source, and the value is selected from the top of the sorted values. A set of physiological indicators (such as potassium ions, sodium ions, etc.); This is a high-priority filtering operator, derived from priority filtering logic, with no value range, used to filter physiological quantities that are ranked first from the priority set; This refers to the quantity of high-priority physiological parameters, derived from clinical treatment configurations, and takes the value as a positive integer (e.g., ...). =2-3), describing the number of high-priority physiological indicators to be screened. The control logic of formula (8) is to call the high-priority screening operator and adjust the priority according to the physiological indicators. Filter out the top-ranked ones Several physiological indicators were identified as high-priority physiological quantities. It focuses on core physiological indicators to provide targeted input for subsequent prediction of potential cardiac risks.

[0099] The real-time monitoring and adaptive control system, designed for patient cardiac monitoring scenarios, extracts top-ranked categories from a physiological adjustment priority list, such as potassium balance, to determine high-priority physiological parameters, thus providing a basis for subsequent risk assessment. High-priority physiological parameters refer to a limited number of core physiological indicators that are prioritized and adapted to clinical cardiac monitoring needs, extracted by a screening operator based on the physiological adjustment priority ranking, providing key input parameters for cardiac risk prediction.

[0100] Step S320: Retrieve timestamp records from the database based on high-priority physiological quantities and generate historical monitoring sequences. Use a sliding window algorithm to smooth the historical monitoring sequences in order to extract circulatory system features.

[0101] The characteristics of the cyclic system can be extracted using the following formula: (9) In formula (9), The circulatory system features are derived from the smoothing and feature extraction of historical monitoring sequences, and the values ​​are a set of features that include the functional state of the circulatory system (such as heart rate fluctuations, changes in vascular resistance, etc.). The smoothing and feature extraction operator, derived from the sliding window algorithm, has no value range and is used to smooth and denoise historical monitoring sequences and extract features. This is a historical data retrieval operator, sourced from database retrieval logic, with no value range, used to retrieve historical monitoring data based on high-priority physiological quantities and timestamps; This is a historical monitoring database, sourced from the patient's historical monitoring data, and its value is a database containing the patient's past physiological monitoring data. For timestamp records, the unit is (seconds), source database timestamp, value is the collection timestamp corresponding to the historical monitoring data; This is the width of the sliding window (for smoothing), sourced from the smoothing configuration, and its value is a positive integer (e.g., ...). =3-5 acquisition cycles); This is the sliding window step size (for smoothing), sourced from the smoothing configuration, and takes a positive integer value (e.g., ...). =1 acquisition cycle). The control logic of formula (9) is based on high-priority physiological quantities. Call the historical data retrieval operator to retrieve historical monitoring data from the database. The corresponding timestamp record Historical monitoring sequences are generated; a sliding window algorithm is used, with the window width... Step length The historical monitoring sequences are smoothed and denoised, and features reflecting the functional state of the circulatory system are extracted to obtain circulatory system features. .

[0102] Based on high-priority physiological parameters such as potassium ion concentration, the real-time monitoring and adaptive control system retrieves timestamp records from the database. These records include monitoring data from the past 24 hours, such as potassium ion level values ​​collected hourly, thereby generating a historical monitoring sequence. Specifically, this historical monitoring sequence can be represented as a time-ordered array of values, such as [4.2, 4.1, 3.9, 4.0] corresponding to different time points. Next, a sliding window algorithm is used to smooth the historical monitoring sequence. For example, the window size is set to 3, and noise is reduced by calculating the average value within the window. For example, applying a window to the historical monitoring sequence [4.2, 4.1, 3.9, 4.0], the average value of the first window is (4.2+4.1+3.9) / 3=4.067, and the average value of the second window is (4.1+3.9+4.0) / 3=4.0, thus obtaining the smoothed sequence [4.067, 4.0]. This smoothing process helps extract circulatory system features, such as average fluctuation amplitude and trend slope. For example, the slope feature is obtained by calculating the difference of the smoothed sequence, reflecting the impact of potassium ion changes on cardiac pumping. Through these circulatory system features, real-time monitoring and adaptive control systems can capture the dynamic patterns of the circulatory system, such as the potential risk of arrhythmia when potassium levels are low. Circulatory system features refer to quantitative characteristic parameters used to characterize the patient's circulatory system status, the temporal trends of physiological indicators, and fluctuation patterns, obtained by retrieving historical monitoring sequences based on high-priority physiological quantities and then smoothing and denoising them using a sliding window process.

[0103] Step S330: Input the circulatory system features into the pre-trained gradient boosting regression tree model to obtain the potential cardiac risk level, and determine whether the potential cardiac risk level exceeds the preset threshold.

[0104] The potential cardiac risk level is calculated using the following formula: (10) In formula (10), The potential cardiac risk level is predicted by a gradient boosting regression tree model, with values ​​ranging from [value range missing]. The higher the value, the greater the potential heart risk; The gradient boosting regression tree prediction operator is derived from a pre-trained gradient boosting regression tree model. It has no value range and is used to input circulatory system features to output the potential cardiac risk level. The pre-trained gradient boosting regression tree model is derived from the model training (based on historical cardiac risk data), and its values ​​include the model parameters containing the decision tree set and regression parameters. This is a risk assessment operator, derived from risk assessment logic, with no value range, used to determine whether the potential cardiac risk level exceeds a preset threshold. A threshold is preset for potential cardiac risk, derived from clinical risk standards, with a range of values. Exceeding this threshold is considered a high risk of cardiac events. The control logic of formula (10) is to extract circulatory system features. Input a pre-trained gradient boosting regression tree model The potential cardiac risk level is output by the prediction operator. The risk assessment operator is invoked to compare the potential cardiac risk level with a preset threshold. If a comparison is made, If the risk exceeds the limit, it is determined that the potential cardiac risk exceeds the standard, triggering the subsequent treatment parameter optimization process; otherwise, it is determined that the risk is controllable, and the current treatment status is maintained.

[0105] The circulatory system features are input into a pre-trained gradient boosting regression tree model. This model iteratively constructs decision trees to fit the residuals; for example, it first trains a tree to predict the initial risk, then subsequent trees correct the error, and finally outputs a potential cardiac risk level score, such as 0.75. It is then determined whether the potential cardiac risk level exceeds a preset threshold, for example, 0.7; if it does, an alarm is triggered. In one embodiment, this method enables timely early warning of cardiac risk in patients, supporting the optimization of medical interventions.

[0106] Preferably, the real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine provided in this embodiment includes step S400 as follows: Step S410: If the potential cardiac risk level exceeds the preset threshold, the real-time monitored blood flow velocity and vascular resistance data are obtained through the feedback control mechanism, and the deviation value of circulating blood volume and osmotic pressure gradient is determined through regression analysis to obtain the initial physiological deviation matrix.

[0107] The initial physiological deviation matrix is ​​obtained using the following formula: (11) In formula (11), The initial physiological deviation matrix is ​​a 2×n real number matrix (n is the number of monitored samples) that integrates source regression analysis and feedback data to describe the deviation between circulating blood volume and osmotic pressure gradient. This is a regression bias calculation operator, derived from regression analysis logic, with no value range. It is used to determine the deviation values ​​of physiological indicators through regression analysis and to construct a deviation matrix. This is a feedback control operator, derived from a feedback control mechanism, with no value range, used to fuse potential cardiac risk prediction results with real-time physiological data; It is a real-time physiological data set, sourced from real-time collection by physiological sensors, and its values ​​include a set of real-time data such as blood flow velocity and vascular resistance. Blood flow velocity, unit is (mL / min), data collected in real time by a physiological sensor, ranging from 100-300. This describes the flow rate of blood during dialysis. Vascular resistance, unit is 1000 m / s (mmHg·min / mL), data collected in real time by a physiological sensor, ranging from 0.5 to 2.0 mmHg. The formula (11) describes the resistance of blood vessels to blood flow. The control logic of formula (11) is that when the potential cardiac risk level exceeds the preset threshold, the feedback control operator is called through the feedback control mechanism to integrate the potential cardiac risk level. With real-time physiological data To obtain real-time monitored blood flow rate With vascular resistance Regression analysis was performed on the above data using a regression bias calculation operator to determine the deviation between circulating blood volume and osmotic pressure gradient, and the deviation values ​​were integrated to construct an initial physiological bias matrix. This provides a basis for determining deviations in subsequent treatment parameters.

[0108] When the potential cardiac risk level exceeds a preset threshold, the real-time monitoring and adaptive control system collects blood flow velocity data from wearable sensors in real time through a feedback control mechanism. For example, the flow velocity at the patient's arm artery is monitored to be 1.2 meters per second, while vascular resistance data, such as the endothelial cell layer resistance index, is acquired at 0.8 Pascals. Through regression analysis, these data are compared with a standard circulating blood volume model to calculate the deviation value. For example, if the actual circulating blood volume is 4.5 liters while the standard is 5 liters, the deviation is 0.5 liters; if the actual osmotic pressure gradient is 280 millimoles per liter while the standard is 300, the deviation is 20. This forms an initial physiological deviation matrix, which is a 2x2 table recording these deviations for subsequent processing.

[0109] Step S420: Based on the initial physiological deviation matrix, retrieve the correlation records between dialysis pressure and electrolyte concentration, and use the particle swarm optimization algorithm to optimize the dialysis pressure and electrolyte concentration to determine the standard setting combination for the hemodialysis machine operation.

[0110] The following formula is used to determine the baseline operating settings for the hemodialysis machine: (12) In formula (12), The set of baseline settings for the operation of the hemodialysis machine is optimized by the particle swarm optimization algorithm, and the values ​​are a set of settings that include baseline parameters such as dialysis pressure and electrolyte concentration. It is a particle swarm optimization operator, derived from the particle swarm optimization algorithm, with no value range, used to optimize dialysis pressure and electrolyte concentration; This is a record retrieval operator, sourced from the database retrieval logic, with no value range, used to retrieve relevant records based on the initial physiological deviation matrix; This is a database that records the correlation between dialysis pressure and electrolyte concentration. It is derived from the storage of historical data on dialysis treatment and takes the value as a set of records containing physiological deviations corresponding to different dialysis pressures and electrolyte concentrations. For dialysis pressure, unit (mmHg), sourced from dialysis machine parameter configuration, value range -50 to +50. Describes the pressure level during dialysis; Electrolyte concentration, unit (mmol / L), derived from the preparation of the dialysate, with a value range determined according to the electrolyte type, describing the concentration of electrolytes in the dialysate. The control logic of formula (12) is based on the initial physiological deviation matrix. The association record retrieval operator is invoked to retrieve the database of association records between dialysis pressure and electrolyte concentration. Obtain the associated records that match the current deviation; extract the dialysis pressure from the associated records. With electrolyte concentration As optimization variables, the particle swarm optimization operator is invoked to find the optimal parameter combination that reduces physiological deviation, and this combination is determined as the baseline setting for the hemodialysis machine. .

[0111] Based on the initial physiological deviation matrix, the real-time monitoring and adaptive control system retrieves correlation records between dialysis pressure and electrolyte concentration from the database. For example, past records show that the sodium concentration was 135 mmol / L when the pressure was 150 mmHg. A particle swarm optimization algorithm is used for optimization. This algorithm simulates a swarm of particles moving in the search space, with each particle representing a pressure-concentration combination. The position and velocity are updated iteratively; for example, with an initial particle swarm size of 50 and 100 iterations, the fitness function is evaluated based on minimizing the deviation. Finally, the optimal combination, such as a pressure of 140 mmHg and a sodium concentration of 138 mmol / L, is determined as the baseline setting for the hemodialysis machine.

[0112] Step S430: Based on the baseline setting combination for the hemodialysis machine operation, and combined with heart rate variability and blood oxygen saturation data, calculate the dynamic response functions of venous pressure and arterial pressure to obtain the corrected setting combination for the hemodialysis machine operation.

[0113] The combination of corrective operating settings for a hemodialysis machine is derived using the following formula: (13) In formula (13), The set of parameters for correcting the operation of the hemodialysis machine is derived from the dynamic response function calculation and is a set of parameter settings corrected based on the baseline set of parameters. The operator for calculating the dynamic response function originates from the dynamic response analysis logic and has no value range. It is used to calculate the dynamic response function of venous pressure and arterial pressure and to correct the baseline settings. Heart rate variability is derived from real-time data collected by physiological sensors, with values ​​ranging from 50 to 150 ms. It describes the degree of fluctuation in heart rate and reflects the autonomic nervous function of the heart. Blood oxygen saturation, measured in real-time by physiological sensors, ranges from 90% to 100%, describing the degree of oxygen saturation in the blood. The control logic of formula (13) is based on the operating baseline settings of the hemodialysis machine. Based on this, combined with real-time heart rate variability With blood oxygen saturation The data is used to call dynamic response function calculation operators to analyze the correlation between heart rate variability, blood oxygen saturation, and dialysis machine pressure parameters (venous pressure and arterial pressure), and to calculate the dynamic response functions of venous pressure and arterial pressure. Based on the dynamic response functions, the baseline setting combination is corrected to obtain the corrected operating setting combination for the hemodialysis machine. This improves the adaptability of parameter settings.

[0114] Based on the baseline settings for the hemodialysis machine, and combined with heart rate variability data such as a standard deviation of 50 milliseconds and blood oxygen saturation of 95%, the dynamic response functions of venous pressure and arterial pressure are calculated. For example, differential equations are used to simulate pressure difference changes, and corrective settings are obtained, such as adjusting the pump speed to 300 revolutions per minute.

[0115] Step S440: Based on the hemodialysis machine operation correction setting combination, and using the constraints of total ultrafiltration volume and dialysate temperature, perform a second iteration calculation on the hemodialysis machine operation correction setting combination to obtain a dynamic physiological adjustment scheme.

[0116] The dynamic physiological adjustment protocol is derived from the following formula: (14) In formula (14), For dynamic physiological adjustment, the source correction setting combination is optimized in a second iteration, and the value is an adjustment scheme that includes the final dialysis parameter settings; This is a secondary iteration optimization operator, derived from the iterative optimization logic. It has no value range and is used to perform secondary iteration calculations on the combination of correction settings to optimize parameter configuration. Total ultrafiltration volume, unit is (ml), sourced from dialysis treatment plan, value range 1000-5000 This describes the total amount of body fluid that needs to be ultrafiltered during dialysis. This refers to the dialysate temperature, expressed in °C (degrees Celsius). It originates from the preparation of the dialysate and ranges from 36 to 38 °C. It describes the temperature of the dialysate. The set of constraints is derived from dialysis treatment guidelines and includes upper and lower limits for total ultrafiltration volume, dialysate temperature, and parameter coordination constraints. The control logic of formula (14) is based on the combination of hemodialysis machine operation correction settings. Based on, combined with total ultrafiltration With dialysate temperature Constraints The algorithm invokes a secondary iterative optimization operator to perform secondary iterative calculations on the modified setting combination, adjusting parameters to meet the constraints of total ultrafiltration volume and dialysate temperature, while ensuring that each parameter synergistically adapts to the patient's current physiological state, ultimately obtaining a dynamic physiological adjustment plan. .

[0117] By using ultrafiltration total volume constraints, such as not exceeding 2 liters and dialysate temperature of 36-38 degrees Celsius, a second iteration of the hemodialysis machine operation correction settings was performed to obtain a dynamic physiological adjustment scheme.

[0118] Furthermore, the real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine provided in this embodiment includes step S500: Step S510: Extract specific parameter values ​​from the dynamic physiological adjustment plan and transmit the specific parameter values ​​through the hemodialysis machine interface.

[0119] Specific parameter values ​​can be extracted and transmitted using the following formula: (15) In formula (15), This is a parameter transmission operator, originating from the hemodialysis machine interface communication logic. It has no value range and is used to transmit specific parameter values ​​through the dialysis machine interface. This is a parameter extraction operator, derived from parameter processing logic, with no value range, used to extract specific parameter values ​​from dynamic physiological adjustment schemes. This is the interface identifier for the hemodialysis machine, derived from the dialysis machine hardware configuration, and its value is the address of the dialysis machine control interface and the set of communication protocol identifiers. The control logic of formula (15) is to call the parameter extraction operator, based on the dynamic physiological adjustment scheme. Specific parameter values ​​such as dialysis pressure, electrolyte concentration, and ultrafiltration rate are extracted; these extracted parameter values ​​are then input into a parameter transmission operator via a pre-defined hemodialysis machine interface. The parameter values ​​are transmitted to the hemodialysis machine according to the corresponding communication protocol to configure the treatment parameters.

[0120] Specific parameter values ​​are extracted from the dynamic physiological adjustment protocol, such as the pump speed of 280 revolutions per minute and the dialysate temperature of 37 degrees Celsius, and these specific parameter values ​​are transmitted through the hemodialysis machine interface.

[0121] Step S520: Determine whether the transmission of specific parameter values ​​was successful. If the transmission was successful, monitor the changes in multiple physiological indicators and use time series analysis to process the newly collected data to obtain updated physiological change trends.

[0122] The updated physiological change trend is derived using the following formula: (16) in, To update the physiological change trend, time series analysis of newly collected data is performed, and the value is the change trend of newly collected physiological data after successful parameter transmission; This is a time series analysis operator, derived from time series analysis methods, with no value range, used to process newly collected physiological data and generate trends. The newly acquired physiological data is collected in real time from physiological sensors (after successful parameter transmission) and the values ​​are new time-series data including blood components, cardiac pumping burden, and electrolyte balance. The transmission judgment operator, source interface communication detection logic, has no value range and is used to determine whether the transmission of specific parameter values ​​is successful. The control logic of formula (16) calls the transmission judgment operator to determine whether the specific parameter values ​​are transmitted successfully through the hemodialysis machine interface. The transmission was successful; if successful, subsequent changes in multiple physiological indicators were monitored, and new physiological data were collected. The time series analysis operator is invoked to process the newly collected data and generate updated physiological change trends. If transmission fails, retain the original trends of multiple physiological indicators. This triggers a retransmission process.

[0123] Determining the success of parameter value transmission involves checking interface feedback signals. If successful, subsequent changes in multiple physiological indicators, such as heart rate and blood pressure, are monitored, and newly acquired data is processed using time series analysis methods. Specifically, time series analysis methods involve modeling the data sequence using an autoregressive moving average model. For example, by collecting heart rate data sequences at 80, 82, and 85 beats per minute, patterns are identified by calculating the autocorrelation function, and model parameters are fitted to predict trends, thus obtaining updated physiological trends such as a stable increase in heart rate. This method decomposes the sequence into trend, seasonality, and residual components to ensure the capture of subtle fluctuations. For example, if the blood pressure sequence shows a downward trend during dialysis, the model weights are adjusted and historical data is integrated to obtain a more accurate trend curve.

[0124] Step S530: Based on the updated physiological change trend, the physiological deviation matrix is ​​integrated with the correlation of dialysis pressure to perform data calibration and determine the optimized physiological indicator monitoring model.

[0125] The optimized physiological indicator monitoring model is determined using the following formula: (17) In formula (17), For the optimized physiological indicator monitoring model, the source data is calibrated and the model is optimized, and the values ​​are the parameters of the calibrated physiological indicator monitoring and analysis model; The model calibration operator originates from the model optimization logic, has no value range, and is used to calibrate the monitoring model by combining update trend and deviation data. The data related to dialysis pressure is sourced from dialysis machine operation data collection, and the value is a set of related data including real-time and historical values ​​of dialysis pressure. The control logic of formula (17) is based on the updated physiological change trend. , fusion of initial physiological deviation matrix Data related to dialysis pressure The model calibration operator is invoked to calibrate the original physiological indicator monitoring model, correct the model parameters, eliminate data bias after parameter transmission, improve the model's monitoring accuracy for new physiological data, and obtain an optimized physiological indicator monitoring model. .

[0126] Data calibration is performed by integrating the physiological deviation matrix with dialysis pressure based on the updated physiological change trends. For example, the physiological deviation matrix records a blood pressure deviation of 10 mmHg and an electrolyte deviation of 5 mmol / L. The correlation with dialysis pressure, such as 120 mmHg, is used to calculate the calibration coefficient and determine the optimized physiological indicator monitoring model. This physiological indicator monitoring model is a machine learning-based framework that integrates deviation data to calibrate indicators in real time.

[0127] Step S540: Use the physiological indicator monitoring model to perform secondary processing on the newly collected data to obtain the final updated physiological change trend.

[0128] The final trend of physiological changes is derived using the following formula: (18) In formula (18), The final updated physiological change trend is derived from the secondary processing of the optimized monitoring model, and the value is the final physiological indicator change trend after model calibration. This is a secondary processing operator, representing the secondary analysis logic of the source data. It has no value range and is used to perform secondary processing on newly acquired data using the optimized monitoring model. The control logic of formula (18) is to process the newly acquired physiological data... Input optimized physiological indicator monitoring model The secondary processing operator is invoked to perform secondary analysis and processing on the data, correcting errors in the first processing stage, accurately capturing the changing patterns of physiological indicators after parameter adjustment, and obtaining the final updated physiological change trend. This provides a precise basis for determining whether further adjustments are needed.

[0129] The newly collected data is processed by a physiological indicator monitoring model. For example, a new blood pressure data sequence is input, and noise is filtered out by the neural network layer of the model to obtain the final updated physiological change trend, thereby achieving more accurate dialysis monitoring.

[0130] Preferably, the real-time monitoring and adaptive control method for multiple parameters of a hemodialysis machine provided in this embodiment includes step S600 as follows: Step S610: Based on the updated physiological change trend, call the logistic regression classification algorithm to determine whether there is a need for further physiological adjustment.

[0131] The results of further assessment of physiological adjustment needs are derived using the following formula: (19) In formula (19), To further assess the physiological adjustment needs, the results were derived from a logistic regression classification algorithm, with values ​​set to... (Further adjustments are needed) or (No further adjustments required); This is a logistic regression classification operator, derived from the logistic regression classification algorithm. It has no value range and is used to determine whether further physiological adjustments are needed. For the logistic regression classification model, the source model training (trained based on historical adjustment requirement data) takes values ​​that include regression coefficients and classification thresholds as model parameters. To adjust the criteria for determining needs, the clinical treatment guidelines are used, and the values ​​are determined based on the threshold for judging the trend of physiological indicator changes and the logic for adjusting needs. The control logic of formula (19) is based on the final updated physiological change trend. The logistic regression classification algorithm operator is invoked, and the trend of change is input into the pre-trained logistic regression classification model. Combined with the preset adjustment requirements judgment criteria Determine whether the current physiological state requires further adjustment, and output the result of the determination of the need for further physiological adjustment. .

[0132] To address the processing of updated physiological change trends, the first step is to use a logistic regression classification algorithm to assess potential needs. Specifically, logistic regression is a supervised learning method used for binary or multi-class classification. It converts linear combinations into probability values ​​using the sigmoid function. For example, given physiological trend data such as a heart rate increase of 0.5 and a blood pressure fluctuation of 2 mmHg, a weighted sum is calculated, and then the sigmoid function is applied to obtain the probability. If the probability exceeds a threshold of 0.7, it is determined that there is a need for further physiological adjustments, thus providing a basis for subsequent decision-making.

[0133] Step S620: Extract trigger commands from the further physiological adjustment needs and generate confirmation signals through the command parsing module.

[0134] The confirmation signal is generated using the following formula: (20) In formula (20), The confirmation signal is derived from the parsing of the trigger instruction and takes the value of the confirmation signal (such as an electrical signal or a logic signal) used to drive the feedback control mechanism. This is an instruction parsing operator, originating from the instruction parsing module. It has no value range and is used to parse trigger instructions into confirmation signals. The trigger instruction extraction operator originates from the requirement parsing logic, has no value range, and is used to extract the trigger instruction from the adjustment requirements. The instruction parsing module model is derived from the instruction parsing configuration, and its value includes instruction parsing rules and signal conversion parameters. The control logic of formula (20) is that when the further physiological adjustment requirement judgment result is... When further adjustments are needed, the trigger instruction extraction operator is invoked to extract the corresponding trigger instruction from the further physiological adjustment requirements; the trigger instruction is then input into the instruction parsing module model. The trigger command is parsed and transformed by the instruction parsing operator to generate an acknowledgment signal used to drive the feedback control mechanism. .

[0135] The process of extracting trigger commands from further physiological adjustment needs involves parsing the demand data structure. Trigger commands include instructions to adjust pump speed or temperature. These are processed by a command parsing module, a software component that uses a parser to break down command strings, such as "adjust pump speed to 300 rpm", and generates confirmation signals, such as the digital code 1, indicating validity, thereby ensuring the accurate transmission of trigger commands.

[0136] Step S630: Based on the confirmation signal, the feedback control mechanism is driven to inject physiological optimization values ​​into the hemodialysis machine control bus.

[0137] The following formula is used to inject physiologically optimized values ​​into the hemodialysis machine control bus: (twenty one) In formula (21), An operator is injected to provide the physiological optimization value. The source is a feedback control mechanism with no value range. It is used to inject the physiological optimization value into the dialysis machine control bus. This is a feedback control driven operator, originating from a feedback control mechanism, with no value range, used to drive the feedback control mechanism based on confirmation signals to generate physiological optimization values; This is the control bus identifier for the hemodialysis machine, originating from the dialysis machine control hardware configuration. Its value is the address and communication protocol identifier of the dialysis machine control bus. The control logic of formula (21) is based on the confirmation signal. The system invokes the feedback control driver operator, which activates the feedback control mechanism to generate corresponding physiological optimization values ​​based on the current physiological state and adjustment needs. The physiological optimization value injection operator then invokes the operator to transmit the generated physiological optimization values ​​via the hemodialysis machine control bus. Injected into the dialysis machine control system, enabling the application of physiological optimization values.

[0138] When the confirmation signal drives the feedback control mechanism, the physiological optimization value is injected into the hemodialysis machine control bus. For example, the physiological optimization value is a pump speed of 320 revolutions per minute and a dialysate flow rate of 500 ml per minute. Data packets are sent through bus protocols such as CAN bus to achieve real-time injection.

[0139] Step S640: Apply physiological optimization values ​​through a feedback control mechanism, monitor pressure fluctuation data in real time, and obtain the final stable state of the patient's treatment.

[0140] The patient's final stable condition is determined by the following formula: (twenty two) In formula (22), To achieve the final stable state of patient treatment, the source physiological optimization values ​​are applied and pressure fluctuation monitoring is used. The values ​​are a set of states that include the stable range of the patient's physiological indicators and the stable settings of the dialysis machine parameters. This is a steady-state monitoring operator, derived from the steady-state judgment logic. It has no value range and is used to monitor the state after the application of physiological optimization values ​​to determine whether stability has been achieved. An operator is applied to the physiological optimization value, with a source feedback control mechanism and no value range. It is used to apply the physiological optimization value and monitor the application effect. For pressure fluctuation data, units (mmHg), sourced from real-time data collected by the dialysis machine pressure sensor, with a value range of 0-10. This describes the pressure fluctuation range during dialysis. The control logic of formula (22) is to apply the physiological optimization value through the physiological optimization value application operator and the feedback control mechanism to generate the physiological optimization value, while simultaneously monitoring the pressure fluctuation data of the dialysis machine in real time. The system invokes the steady-state monitoring operator to analyze the pressure fluctuations in the patient's physiological indicators after the application of the physiological optimization values. It determines whether the physiological indicators have reached the preset stable range and whether the pressure fluctuations are within the allowable range. When the stability conditions are met, it outputs the final stable state of the patient's treatment. This completes the adaptive control process.

[0141] After applying physiological optimization values ​​through a feedback control mechanism, real-time monitoring of pressure fluctuation data involves sensor acquisition sequences, such as pressure value sequences of 110, 115, and 112 mmHg per second. The fluctuation amplitude is analyzed to obtain the final stable state of the patient's treatment, thereby optimizing the dialysis process.

[0142] Please see Figure 2This invention provides a multi-parameter real-time monitoring and adaptive control system for a hemodialysis machine, used to implement the aforementioned multi-parameter real-time monitoring and adaptive control method for a hemodialysis machine. It includes a multi-physiological indicator change trend acquisition module 10, a physiological adjustment priority determination module 20, a potential cardiac risk level judgment module 30, a dynamic physiological adjustment scheme acquisition module 40, an updated physiological change trend acquisition module 50, and a final patient treatment stability acquisition module 60. The multi-physiological indicator change trend acquisition module 10 is used to collect real-time data on blood component concentration, cardiac pumping burden level, and body fluid electrolyte balance from the patient using physiological sensors. Time series analysis is used to process the blood component concentration, cardiac pumping burden level, and body fluid electrolyte balance data to obtain the change trends of multiple physiological indicators. The physiological adjustment priority determination module 20 is used to classify the current patient's physical state based on the multi-physiological indicator change trends using a classification algorithm. If the classification result shows an increase in cardiac pumping burden, body fluid electrolyte balance data is integrated to determine the physiological adjustment priority. The potential cardiac risk level judgment module 30 is used to... The system prioritizes and obtains high-priority physiological indicators, uses a predictive algorithm to predict potential cardiac risk levels, and determines whether the potential cardiac risk level exceeds a preset threshold. The dynamic physiological adjustment plan acquisition module 40, if the potential cardiac risk level exceeds the preset threshold, uses a feedback control mechanism to fuse the prediction results and real-time data to obtain optimized treatment parameter values. For these optimized values, an optimization algorithm searches for the best combination of hemodialysis machine settings to obtain a dynamic physiological adjustment plan. The updated physiological change trend acquisition module 50 extracts specific parameter values ​​from the dynamic physiological adjustment plan, transmits these values ​​through the hemodialysis machine interface, and determines whether the transmission is successful. If successful, it monitors subsequent changes in multiple physiological indicators, reprocesses newly collected data using time series analysis, and obtains updated physiological change trends. The final patient treatment stability acquisition module 60, based on the updated physiological change trends, repeats the classification algorithm process to determine whether further physiological adjustments are needed, obtains confirmation signals from the further physiological adjustment requirements, and applies the optimized physiological values ​​through a feedback control mechanism to obtain the final patient treatment stability state.

[0143] Clinical validation and efficacy comparison data: A 3-month clinical observation was conducted on 30 maintenance hemodialysis patients in a hospital's hemodialysis center. 15 patients used traditional empirical parameter settings (control group), and 15 patients used the system of this invention (experimental group). The main observation indicators are compared in Table 1. Table 1 The multi-parameter real-time monitoring and adaptive control method and system for hemodialysis machines provided in this embodiment have the following beneficial effects compared with the prior art: 1. Significantly reduced cardiac-related complications: This embodiment, through real-time monitoring and risk prediction (GBRT model), can issue early warnings (on average 12.5 minutes in advance) at the early stage of abnormally high blood potassium or the onset of a drop in blood pressure. It can also quickly and accurately adjust the electrolyte concentration and ultrafiltration rate of dialysate through MPC and PSO algorithms, thereby effectively preventing the occurrence of IDH and muscle spasms.

[0144] 2. Improved Precision and Adequacy of Treatment: Based on trend analysis and classification through multi-parameter fusion, the system can more accurately identify the patient's real-time physiological state (such as "hyperkalemia with volume overload"), thereby developing personalized dialysis prescriptions (such as increasing dialysate potassium concentration and reducing ultrafiltration rate), which improves the potassium target achievement rate and overall dialysis adequacy (spKt / V).

[0145] 3. Automation and reduced workload: The closed-loop control of the system significantly reduces the workload of medical staff who frequently adjust parameters based on experience, freeing them from repetitive operations and allowing them to focus on more complex clinical decisions, while improving the comfort and safety of patients during treatment.

[0146] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine, characterized in that, Includes the following steps: S100: Real-time data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance are collected from the patient's body using physiological sensors. The data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance are processed using time series analysis to obtain the changing trends of multiple physiological indicators. S200. Based on the changing trends of the multiple physiological indicators, a classification algorithm is used to classify the current physical state of the patient. If the classification result shows an increase in the heart's pumping burden, the fluid and electrolyte balance data is integrated to determine the priority of physiological adjustment. S300: Obtain high-priority physiological indicators from the physiological adjustment priorities, use a prediction algorithm to predict the potential cardiac risk level, and determine whether the potential cardiac risk level exceeds a preset threshold. S400. If the potential cardiac risk level exceeds a preset threshold, the prediction results and real-time data are fused through a feedback control mechanism to obtain optimized treatment parameter values. For the optimized treatment parameter values, an optimization algorithm is used to search for the best combination of hemodialysis machine settings to obtain a dynamic physiological adjustment scheme. S500. Extract specific parameter values ​​from the dynamic physiological adjustment scheme, transmit the specific parameter values ​​through the hemodialysis machine interface, determine whether the transmission is successful, and if the transmission is successful, monitor the changes of subsequent multiple physiological indicators, reprocess the newly collected data using time series analysis methods, and obtain updated physiological change trends. S600. Based on the updated physiological change trend, repeat the classification algorithm process to determine whether further physiological adjustment is needed. Obtain confirmation signals from the further physiological adjustment needs, apply physiological optimization values ​​through feedback control mechanisms, and obtain the final stable state of patient treatment.

2. The method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine according to claim 1, characterized in that, Step S100 includes: S110. Use physiological sensors to collect data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance in real time, and construct a multidimensional time series based on the blood component concentration, cardiac pumping burden level and body fluid electrolyte balance data; S120. Perform sliding window processing on the multidimensional time series to extract fluctuation features. If the fluctuation features exceed a preset threshold, use the correlation matrix to determine the evolution trajectory. S130. Calculate the trend vector based on the evolution trajectory, and obtain the change trend of multiple physiological indicators by mapping the trend vector to a multi-dimensional coordinate system.

3. The method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine according to claim 2, characterized in that, Step S200 includes: S210. Obtain the changing trends of multiple physiological indicators and map them to a high-dimensional feature space to extract feature vectors that reflect fluctuations in physiological function. S220. Input the feature vector into a pre-trained random forest classifier, and obtain the current patient's physical state by performing a non-linear mapping on the feature vector; S230. If the current patient's physical condition is identified as an increased burden on the heart's pumping function, then the corresponding body fluid and electrolyte balance data are extracted and a multidimensional correlation matrix is ​​constructed. S240. Based on the multidimensional correlation matrix, quantitative processing is performed using a weight allocation model to determine the priority of physiological adjustment.

4. The method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine according to claim 1, characterized in that, Step S300 includes: S310. Extract the top-ranked categories from the physiological adjustment priorities to determine high-priority physiological quantities; S320. Retrieve timestamp records from the database based on the high-priority physiological quantities and generate historical monitoring sequences. Use a sliding window algorithm to smooth the historical monitoring sequences in order to extract circulatory system features. S330. Input the circulatory system features into a pre-trained gradient boosting regression tree model to obtain the potential cardiac risk level, and determine whether the potential cardiac risk level exceeds a preset threshold.

5. The method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine according to claim 4, characterized in that, Step S400 includes: S410. If the potential cardiac risk level exceeds the preset threshold, the real-time blood flow velocity and vascular resistance data are obtained through the feedback control mechanism, and the deviation value of circulating blood volume and osmotic pressure gradient is determined through regression analysis to obtain the initial physiological deviation matrix. S420. Based on the initial physiological deviation matrix, retrieve the correlation records between dialysis pressure and electrolyte concentration, and use the particle swarm optimization algorithm to optimize the dialysis pressure and electrolyte concentration to determine the combination of operating benchmark settings for the hemodialysis machine. S430. Based on the operating baseline setting combination of the hemodialysis machine, and combined with heart rate variability and blood oxygen saturation data, calculate the dynamic response functions of venous pressure and arterial pressure to obtain the operating correction setting combination of the hemodialysis machine. S440. Based on the hemodialysis machine operation correction setting combination, and utilizing the constraints of total ultrafiltration volume and dialysate temperature, perform a second iteration calculation on the hemodialysis machine operation correction setting combination to obtain a dynamic physiological adjustment scheme.

6. The method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine according to claim 1, characterized in that, Step S500 includes: S510. Extract specific parameter values ​​from the dynamic physiological adjustment scheme and transmit the specific parameter values ​​through the hemodialysis machine interface; S520. Determine whether the transmission of the specific parameter value is successful. If the transmission is successful, monitor the changes of multiple physiological indicators, process the newly collected data using time series analysis methods, and obtain the updated physiological change trend. S530. Based on the updated physiological change trend, the physiological deviation matrix is ​​fused with the correlation between dialysis pressure and data calibration is performed to determine the optimized physiological indicator monitoring model. S540. The newly collected data is processed a second time using the physiological indicator monitoring model to obtain the final updated physiological change trend.

7. The method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine according to claim 1, characterized in that, Step S600 includes: S610. Based on the updated physiological change trend, call the logistic regression classification algorithm to determine whether there is a need for further physiological adjustment; The results of further assessment of physiological adjustment needs are derived using the following formula: ； in, To further assess the physiological adjustment needs, To predict the final trend of physiological changes, For logistic regression classification operators, For logistic regression classification models, To adjust the criteria for determining demand; S620. Extract the trigger command from the further physiological adjustment requirement, and generate a confirmation signal through the command parsing module; The confirmation signal is generated using the following formula: ； in, To confirm the signal, For instruction parsing operators, To trigger the instruction to extract the operator, For the instruction parsing module model, Further adjustments are needed; S630. Based on the confirmation signal, a feedback control mechanism is driven to inject physiological optimization values ​​into the hemodialysis machine control bus. S640. By applying the physiological optimization value through the feedback control mechanism, pressure fluctuation data is monitored in real time to obtain the final stable state of the patient's treatment.

8. The method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine according to claim 7, characterized in that, In step S630, the physiological optimization value is injected into the hemodialysis machine control bus using the following formula: ； in, Inject operators into physiological optimization values. For feedback control driven operators, This is the identifier for the control bus of a hemodialysis machine.

9. The method for real-time monitoring and adaptive control of multiple parameters in a hemodialysis machine according to claim 8, characterized in that, In step S640, the final stable state of the patient's treatment is obtained using the following formula: ； in, To achieve a stable treatment outcome for the patient, For the steady-state monitoring operator, Apply operators to optimize physiological values. This is data on pressure fluctuations.

10. A multi-parameter real-time monitoring and adaptive control system for a hemodialysis machine, used to implement the multi-parameter real-time monitoring and adaptive control method for a hemodialysis machine as described in any one of claims 1 to 9, characterized in that, include: The module for acquiring the changing trends of multiple physiological indicators is used to collect data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance from the patient in real time through physiological sensors. The data on blood component concentration, cardiac pumping burden level and body fluid electrolyte balance are processed by time series analysis to obtain the changing trends of multiple physiological indicators. The physiological adjustment priority determination module is used to classify the current patient's physical state based on the changing trends of the multiple physiological indicators using a classification algorithm. If the classification result shows an increase in the heart's pumping burden, the module integrates the body fluid and electrolyte balance data to determine the physiological adjustment priority. The potential cardiac risk level judgment module is used to obtain high-priority physiological indicators from the physiological adjustment priority, use a prediction algorithm to predict the potential cardiac risk level, and determine whether the potential cardiac risk level exceeds a preset threshold. The dynamic physiological adjustment scheme acquisition module is used to obtain optimized treatment parameter values ​​by fusing prediction results and real-time data through a feedback control mechanism if the potential cardiac risk level exceeds a preset threshold. Based on the optimized treatment parameter values, an optimization algorithm is used to search for the best combination of hemodialysis machine settings to obtain a dynamic physiological adjustment scheme. The physiological change trend acquisition module is used to extract specific parameter values ​​from the dynamic physiological adjustment plan, transmit the specific parameter values ​​through the hemodialysis machine interface, determine whether the transmission is successful, and if the transmission is successful, monitor the changes of subsequent multiple physiological indicators, reprocess the newly collected data using time series analysis methods, and obtain the updated physiological change trend. The final patient treatment stability acquisition module is used to determine whether further physiological adjustments are needed based on the updated physiological change trend and the repeated classification algorithm process. It obtains confirmation signals from the further physiological adjustment needs, applies physiological optimization values ​​through a feedback control mechanism, and obtains the final patient treatment stability.

Images