Infusion pump control method and system

Abstract

The invention relates to the technical field of control of infusion pumps, in particular to a control method and system of an infusion pump. The method comprises the following steps: firstly, acquiring physiological parameters of an infusion patient, and extracting physiological parameters in an abnormal state; then, on the basis of the abnormal data, fitting of blood glucose change and blood viscosity increment is carried out, and blood vessel infusion resistance data are obtained through regression analysis; the infusion pressure range is dynamically regulated and controlled according to the resistance data, and the flow gradient in unit time is further adjusted; and finally, constructing an infusion self-adaptive parameter control model based on a strategy gradient algorithm, and embedding the infusion self-adaptive parameter control model into an infusion pump control system to realize accurate infusion adjustment and optimization. The control technology of the infusion pump is optimized, so that the control technology of the infusion pump is more perfect.

Description

Technical Field

[0001] The present invention relates to the technical field of infusion pump control, and particularly to a control method and system for an infusion pump. Background Art

[0002] With the progress of medical technology, as an important medical device, infusion pumps are widely used in clinical treatment in hospitals, especially playing a crucial role in situations such as intensive care, surgery, cancer treatment, and long-term infusion. Infusion pumps ensure that patients can accurately receive liquid drugs according to medical orders during treatment by precisely controlling the infusion flow rate. However, due to the patient's physical condition, changes in the condition, and the complexity of treatment, the previous infusion pump control methods usually relied on manual setting of flow rate and pressure, which were prone to deviations in dynamic conditions, resulting in adverse reactions such as too fast, too slow, or uneven infusion for patients, thus affecting the treatment effect and even posing safety hazards. However, there is a problem that the traditional control method of an infusion pump has poor adaptability to the physiological reactions of infusion patients, resulting in the inability to dynamically control infusion according to the physiological needs of infusion patients. Summary of the Invention

[0003] Based on this, it is necessary to provide a control method and system for an infusion pump to solve at least one of the above technical problems.

[0004] To achieve the above object, a control method for an infusion pump, the method includes the following steps:

[0005] Step S1: Obtain the physiological parameters of the infusion patient; extract the abnormal state parameters from the physiological parameters of the infusion patient to obtain the abnormal state physiological parameters;

[0006] Step S2: Perform fitting of the blood viscosity increment with blood glucose change based on the abnormal state physiological parameters to obtain the blood viscosity increment data with blood glucose change; perform regression analysis of the vascular infusion resistance based on the blood viscosity increment data with blood glucose change to obtain the vascular infusion resistance data;

[0007] Step S3: Dynamically regulate the infusion pressure range interval according to the vascular infusion resistance data to obtain the dynamically regulated data of the infusion pressure range interval; perform flow gradient control matching per unit time based on the dynamically regulated data of the infusion pressure range interval to obtain the infusion flow gradient control data;

[0008] Step S4: Build an infusion adaptive parameter control model based on the policy gradient algorithm for the dynamically regulated data of the infusion pressure range interval and the infusion flow gradient control data to obtain the infusion adaptive parameter control model; embed the infusion adaptive parameter control model into the infusion pump control system to execute the control of the infusion pump.

[0009] Preferably, step S1 includes the following steps:

[0010] Step S11: Obtain the physiological parameters of the infusion patient;

[0011] Step S12: Fill in the missing values of the physiological parameters of the infusion patient to obtain the physiological filling parameters of the infusion patient;

[0012] Step S13: Perform time series marking processing on the physiological filling parameters of the infusion patient to obtain the time series marked physiological parameters of the infusion patient;

[0013] Step S14: Extract the abnormal state parameters from the time series marked physiological parameters of the infusion patient to obtain the abnormal state physiological parameters.

[0014] Preferably, step S2 includes the following steps:

[0015] Step S21: Quantify the hypertension fluctuation value of the abnormal state physiological parameters to obtain the hypertension fluctuation quantification value;

[0016] Step S22: Perform a simulation assessment of the blood vessel wall pressure bearing according to the hypertension fluctuation quantification value to obtain the blood vessel wall pressure bearing data;

[0017] Step S23: Fit the increment of blood viscosity due to blood glucose change based on the abnormal state physiological parameters to obtain the increment data of blood viscosity due to blood glucose change;

[0018] Step S24: Perform a regression analysis of the blood vessel infusion resistance based on the blood vessel wall pressure bearing data and the increment data of blood viscosity due to blood glucose change to obtain the blood vessel infusion resistance data.

[0019] Preferably, step S23 includes the following steps:

[0020] Step S231: Extract the blood glucose concentration change from the abnormal state physiological parameters to obtain the blood glucose concentration change data;

[0021] Step S232: Identify the blood vessel endothelial cell damage gradient based on the blood glucose concentration change data to obtain the blood vessel endothelial cell damage gradient;

[0022] Step S233: Predict the increment of plasma fibrinogen content according to the blood vessel endothelial cell damage gradient and the blood glucose concentration change data to obtain the plasma fibrinogen increment data;

[0023] Step S234: Fit the increment of blood viscosity due to blood glucose change based on the plasma fibrinogen increment data to obtain the increment data of blood viscosity due to blood glucose change.

[0024] Preferably, step S24 includes the following steps:

[0025] Step S241: Perform a temporal gradient prediction of vascular wall sclerosis on the vascular wall pressure-bearing data to obtain the temporal gradient of vascular wall sclerosis;

[0026] Step S242: Conduct an isoproportional regression analysis of vascular elasticity weakening based on the temporal gradient of vascular wall sclerosis and the increment data of blood viscosity change due to blood glucose to obtain the regression data of vascular elasticity weakening;

[0027] Step S243: Perform a calculation of increased vascular permeability based on the temporal gradient of vascular wall sclerosis and the regression data of vascular elasticity weakening to obtain the data of increased vascular permeability;

[0028] Step S244: Determine the vascular flow resistance coefficient based on the regression data of vascular elasticity weakening, the temporal gradient of vascular wall sclerosis, and the data of increased vascular permeability to obtain the vascular flow resistance coefficient;

[0029] Step S245: Conduct a regression analysis of vascular infusion resistance based on the vascular flow resistance coefficient to obtain the vascular infusion resistance data.

[0030] Preferably, step S244 includes the following steps:

[0031] Quantify the dynamic contraction radius of the blood vessel based on the regression data of vascular elasticity weakening and the temporal gradient of vascular wall sclerosis to obtain the dynamic contraction radius of the blood vessel;

[0032] Estimate the probability of local coagulation factor formation based on the data of increased vascular permeability to obtain the probability of local coagulation formation;

[0033] Perform a piecewise integration of blood flow dynamic loss based on the dynamic contraction radius of the blood vessel and the probability of local coagulation formation to obtain the piecewise data of blood flow dynamic loss;

[0034] Conduct an analysis of the vascular flow resistance distribution on the piecewise data of blood flow dynamic loss to obtain the vascular flow resistance distribution data;

[0035] Determine the vascular flow resistance coefficient based on the piecewise data of blood flow dynamic loss and the vascular flow resistance distribution data to obtain the vascular flow resistance coefficient.

[0036] Preferably, step S3 includes the following steps:

[0037] Step S31: Perform a normalization process on the vascular infusion resistance data to obtain the normalized infusion resistance data;

[0038] Step S32: Dynamically regulate the infusion pressure range interval based on the normalized infusion resistance data to obtain the dynamically regulated data of the infusion pressure range interval;

[0039] Step S33: Match the infusion rate with the dynamically regulated data of the infusion pressure range interval to obtain the infusion rate gradient matching data;

[0040] Step S34: Based on the dynamic regulation data of the infusion pressure range interval and the infusion rate gradient matching data, perform flow gradient control matching per unit time to obtain infusion flow gradient control data.

[0041] Preferably, step S32 includes the following steps:

[0042] Step S321: Perform segmented trend analysis on the normalized infusion resistance data, calculate the resistance gradient change rate within adjacent time windows, and screen abnormal resistance mutation points based on the gradient change rate to obtain resistance gradient mutation data;

[0043] Step S322: Perform mutation point clustering analysis on the resistance gradient mutation data to obtain resistance mutation mode data;

[0044] Step S323: Perform analysis on the numerical interval of vascular fluid pressure bearing according to the resistance mutation mode data to obtain the numerical interval of vascular fluid pressure bearing;

[0045] Step S324: Perform dynamic regulation of the infusion pressure range interval according to the numerical interval of vascular fluid pressure bearing to obtain dynamic regulation data of the infusion pressure range interval.

[0046] Preferably, step S4 includes the following steps:

[0047] Step S41: Perform iterative learning of the infusion logic of the infusion pump according to the dynamic regulation data of the infusion pressure range interval, the infusion rate gradient matching data, and the infusion flow gradient control data to obtain infusion logic iterative data;

[0048] Step S42: Based on the policy gradient algorithm, construct an infusion adaptive parameter control model for the infusion logic iterative data to obtain an infusion adaptive parameter control model;

[0049] Step S43: Embed the infusion adaptive parameter control model into the infusion pump control system to execute the control of the infusion pump.

[0050] Preferably, the present invention also provides a control system for an infusion pump, which is used to execute the control method of the infusion pump as described above. The control system of the infusion pump includes:

[0051] An abnormal physiological parameter extraction module, which is used to obtain the physiological parameters of the infusion patient; extract abnormal state parameters from the physiological parameters of the infusion patient to obtain abnormal state physiological parameters;

[0052] An infusion resistance regression analysis module, which is used to perform fitting of the blood viscosity increment due to blood glucose change based on the abnormal state physiological parameters to obtain blood viscosity increment data due to blood glucose change; perform vascular infusion resistance regression analysis according to the blood viscosity increment data due to blood glucose change to obtain vascular infusion resistance data;

[0053] A flow gradient control module, which is used to dynamically regulate the infusion pressure range according to the vascular infusion resistance data to obtain the dynamic regulation data of the infusion pressure range; and perform flow gradient control matching per unit time based on the dynamic regulation data of the infusion pressure range to obtain the infusion flow gradient control data;

[0054] An adaptive parameter control model construction module, which is used to construct an infusion adaptive parameter control model based on the policy gradient algorithm for the dynamic regulation data of the infusion pressure range and the infusion flow gradient control data to obtain the infusion adaptive parameter control model; and embed the infusion adaptive parameter control model into the infusion pump control system to execute the control of the infusion pump.

[0055] The beneficial effects of the present invention are as follows. By obtaining the physiological parameters of the infusion patient in real time, such as heart rate, blood pressure, body temperature, blood glucose, etc., the physical condition of the patient can be comprehensively understood. Extracting abnormal state parameters from these physiological data can identify potential problems such as blood glucose fluctuations and blood viscosity in the patient. This process provides accurate basic data for subsequent analysis, can timely detect the health risks of the patient, and lays a foundation for the formulation of personalized treatment plans, thus providing necessary guarantees for the precise regulation of the infusion process. Based on the extracted abnormal state physiological parameters, by fitting the relationship between blood glucose changes and the increment of blood viscosity, the influence of blood glucose changes on blood viscosity can be obtained. The acquisition of this data can help doctors evaluate the change of blood fluidity in the patient and predict the change trend of vascular resistance. By performing regression analysis on the relationship between blood glucose and blood viscosity, the vascular infusion resistance data can be obtained, providing a scientific basis for dynamically adjusting parameters such as infusion flow rate and pressure in the subsequent process, thereby more effectively controlling the risks during the infusion process. According to the vascular infusion resistance data obtained from the fitting of blood glucose changes and the increment of blood viscosity, further dynamic regulation of the infusion pressure range interval is carried out. This dynamic adjustment mechanism can adapt to the change of vascular resistance in the patient's body in real time, accurately control the infusion pressure, and avoid complications caused by improper pressure. At the same time, this process can be automatically adjusted according to the individual differences of the patient, thereby improving the safety and comfort of treatment and avoiding the errors and risks that occur during traditional manual adjustment. By integrating and optimizing the dynamic regulation data of the infusion pressure range interval and the flow gradient control data based on the policy gradient algorithm, an infusion adaptive parameter control model is constructed. This model can learn and adjust the control strategy in real time, and adaptively adjust the infusion parameters according to the physiological state of the patient during the infusion process. This intelligent control system can improve the automation level of the infusion pump, reduce the errors of human intervention, and improve the treatment accuracy and efficiency. By embedding this control model into the infusion pump control system, not only the intelligent level of the system is improved, but also the patient can obtain the most suitable infusion plan during the treatment process, significantly reducing the medical risk and improving the treatment effect. Therefore, the present invention is an optimization process for the control method of a traditional infusion pump, solving the problem that the control method of a traditional infusion pump has poor adaptability to the physiological response of the infusion patient, resulting in the inability to dynamically control the infusion according to the physiological needs of the infusion patient, and improving the adaptability of the infusion pump to the physiological response of the infusion patient, thereby improving the ability to dynamically control the infusion according to the physiological needs of the infusion patient. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic flow chart of the steps of a control method for an infusion pump;

[0057] Figure 2 For Figure 1Schematic diagram of the detailed implementation steps of step S2 in

[0058] Figure 3 is Figure 1 Schematic diagram of the detailed implementation steps of step S3 in Specific implementation manner

[0059] Please refer to Figures 1 to 3 , a control method for an infusion pump, the method comprising the following steps:

[0060] Step S1: Obtain physiological parameters of the infusion patient; extract abnormal state parameters from the physiological parameters of the infusion patient to obtain abnormal state physiological parameters;

[0061] Step S2: Perform fitting of the blood viscosity increment with blood glucose change based on the abnormal state physiological parameters to obtain blood viscosity increment data with blood glucose change; perform regression analysis of the vascular infusion resistance based on the blood viscosity increment data with blood glucose change to obtain vascular infusion resistance data;

[0062] Step S3: Perform dynamic regulation and control of the infusion pressure range interval according to the vascular infusion resistance data to obtain dynamic regulation and control data of the infusion pressure range interval; perform flow gradient control matching per unit time based on the dynamic regulation and control data of the infusion pressure range interval to obtain infusion flow gradient control data;

[0063] Step S4: Based on the policy gradient algorithm, construct an infusion adaptive parameter control model for the dynamic regulation and control data of the infusion pressure range interval and the infusion flow gradient control data to obtain an infusion adaptive parameter control model; embed the infusion adaptive parameter control model into the infusion pump control system to execute the control of the infusion pump.

[0064] In an embodiment of the present invention, referring to Figure 1 as described, it is a schematic diagram of the step flow of a control method for an infusion pump of the present invention. In this example, the control method for the infusion pump comprises the following steps:

[0065] Step S1: Obtain physiological parameters of the infusion patient; extract abnormal state parameters from the physiological parameters of the infusion patient to obtain abnormal state physiological parameters;

[0066] In the embodiments of the present invention, first, physiological parameters of the infusion patient are obtained. High-precision physiological monitoring devices are used, including but not limited to continuous glucose monitors (CGM), ambulatory blood pressure monitors (ABPM), hemorheology analyzers, etc. These devices can provide data with high temporal resolution. The physiological parameters include blood glucose level, blood viscosity, blood pressure, vascular elasticity coefficient, heart rate, blood oxygen saturation, etc. After data collection, the linear interpolation method is used to fill in the missing data to ensure data integrity, and wavelet transform is used for denoising. Subsequently, the time series is windowed, with a 10-second window unit. Statistical features such as blood glucose mean, standard deviation, coefficient of variation, maximum value, and minimum value are calculated within each window. At the same time, blood pressure fluctuation features such as pulse pressure and mean arterial pressure are extracted. Next, an abnormal state analysis of the physiological parameters is performed. Based on the three-sigma principle (3σ rule), a threshold range is set to screen abnormal data points. For the blood glucose level, if it exceeds 10 mmol / L or is lower than 3 mmol / L, it is marked as an abnormal state; for the blood viscosity, if it is higher than 5 mPa·s, it is judged as a high blood viscosity state. After the abnormal state parameters are extracted, the Fourier transform is used to analyze the frequency domain characteristics of the data, and then the blood glucose change rate and blood viscosity change rate are calculated, and the change trend within the next 30 minutes is predicted based on regression analysis.

[0067] Step S2: Fit the blood glucose change and blood viscosity increment based on the physiological parameters in the abnormal state to obtain the blood glucose change and blood viscosity increment data; perform a regression analysis of the vascular infusion resistance based on the blood glucose change and blood viscosity increment data to obtain the vascular infusion resistance data;

[0068] In the embodiments of the present invention, based on the physiological parameters in the abnormal state, the blood glucose change and blood viscosity increment are fitted. The nonlinear least squares method (Levenberg-Marquardt algorithm) is used to perform curve fitting on the relationship between blood glucose and blood viscosity changes, establish a second-order polynomial regression equation, and calculate the goodness of fit R 2To evaluate the fitting effect. Meanwhile, in combination with the blood glucose change gradient, calculate the change in plasma fibrinogen concentration, establish the relationship between blood glucose level and fibrinogen change using Logistic regression, and further obtain the predicted value of blood viscosity. Subsequently, based on the blood glucose change data and blood viscosity data, calculate the characteristics of fluid flow in blood vessels through the Navier-Stokes equation, and perform vascular wall pressure simulation using the finite element method (FEM) to obtain the vascular wall pressure bearing data. Based on the vascular wall pressure data and blood viscosity increment data, conduct a regression analysis of vascular infusion resistance. Using a multiple linear regression model, with vascular wall pressure, blood viscosity, blood glucose change rate, fibrinogen concentration, etc. as independent variables, construct an infusion resistance calculation equation. During the calculation process, use Ridge Regression to prevent the problem of multicollinearity, ensure the stability of the model, and perform feature selection through LASSO regression to remove redundant variables. Finally, based on the obtained regression equation, calculate the vascular infusion resistance and perform time series analysis to smooth the resistance data using moving average filtering.

[0069] Step S3: Dynamically regulate the infusion pressure range interval according to the vascular infusion resistance data to obtain dynamically regulated data for the infusion pressure range interval; perform flow gradient control matching per unit time based on the dynamically regulated data for the infusion pressure range interval to obtain infusion flow gradient control data;

[0070] In the embodiment of the present invention, based on the calculated vascular infusion resistance data, dynamically regulate the infusion pressure range interval. First, use the Z-score normalization method to normalize the vascular infusion resistance and map the resistance value to the interval [0, 1]. Then, based on the Kalman Filter, dynamically estimate the resistance change and predict the change trend of the infusion resistance in the next 30 seconds. On this basis, calculate the infusion pressure range adapted to the vascular resistance, calculate the appropriate infusion pressure value using the pressure-flow characteristic curve (Poiseuille's law), and adjust it in combination with the pressure-flow response curve of the infusion pump to ensure flow stability. Subsequently, according to the infusion pressure range data, perform flow gradient control matching per unit time. Based on the adaptive gradient descent algorithm, dynamically adjust the infusion flow. Calculate the current flow gradient and adjust the step size according to the real-time change rate of the vascular infusion resistance to make the change of the infusion rate smoother. Use the third-order spline interpolation method to smooth the flow curve to eliminate the influence of mutation points on the infusion system. Finally, output the infusion flow gradient control data.

[0071] Step S4: Based on the policy gradient algorithm, construct an infusion adaptive parameter control model for the infusion pressure range interval dynamic regulation data and the infusion flow gradient control data, and obtain the infusion adaptive parameter control model; embed the infusion adaptive parameter control model into the infusion pump control system to execute the control of the infusion pump.

[0072] In the embodiment of the present invention, in order to further optimize the control of the infusion pump, a policy gradient algorithm (Policy Gradient) is used to construct an infusion adaptive parameter control model. First, set up a reinforcement learning environment, with infusion flow, vascular resistance, blood glucose level, etc. as state variables, and the infusion pressure adjustment value as the action variable, and define a reward function to make the infusion flow stable and meet the patient's needs. Use the proximal policy optimization algorithm (PPO) to train the adaptive control strategy, use the replay buffer to store historical state data, and update the value function based on the temporal difference method (TD Learning). After training is completed, extract the optimal policy and construct an adaptive parameter control model. Finally, embed the adaptive parameter control model into the infusion pump control system to achieve intelligent control of the infusion pump. The infusion pump receives the patient's physiological data and calculates the infusion pressure and flow adjustment plan in real time to ensure the stability and safety of the infusion process.

[0073] Step S1 includes the following steps:

[0074] Step S11: Obtain the physiological parameters of the infusion patient;

[0075] Step S12: Fill in the missing values of the physiological parameters of the infusion patient to obtain the physiological filled parameters of the infusion patient;

[0076] Step S13: Perform time series marking processing on the physiological filled parameters of the infusion patient to obtain the time series marked physiological parameters of the infusion patient;

[0077] Step S14: Extract the abnormal state parameter from the time series marked physiological parameters of the infusion patient to obtain the abnormal state physiological parameter.

[0078] In the embodiments of the present invention, in the control method of the infusion pump, first, the physiological parameters of the infusion patient are obtained. A continuous glucose monitor (CGM) is used to collect blood glucose levels, and the sampling frequency is set to once per minute, and the real-time changes in blood glucose concentration are recorded. A dynamic blood pressure monitor (ABPM) is used to obtain the patient's blood pressure data, and the sampling period is set to once every 30 seconds, and the systolic blood pressure, diastolic blood pressure, and pulse pressure difference are recorded respectively. A hemodynamic analyzer is used to measure blood viscosity, and the sampling frequency is once every 5 minutes, and parameters such as whole blood viscosity, plasma viscosity, and red blood cell deformability are extracted. In addition, an electrocardiogram monitor is used to record the patient's heart rate changes, and the sampling frequency is once per second, and the peripheral blood perfusion index is measured in combination with photoplethysmography (PPG) to evaluate the vasoconstriction state. All data are transmitted to the central data processing system through a serial interface and synchronized according to the time stamp to ensure data consistency. There are missing values in the obtained physiological parameter data, so filling processing is required. First, the integrity of the collected time series data is detected, and for the missing data points, linear interpolation is used for filling. When the interval of the missing data points is less than 5 sampling points, weighted averaging of the adjacent front and rear values is used for filling. When the missing values continuously exceed 5 sampling points, the local regression estimation method (LOESS) is used for smoothing and completion. After the data filling is completed, cubic spline interpolation is used to perform curve fitting on the data to ensure the continuity of data changes. At the same time, the filled data is standardized, and all parameters are converted to the same order of magnitude range to avoid affecting subsequent analysis due to different numerical scales of different parameters. The processed physiological parameter data needs to be subjected to time series labeling processing to construct time series features. First, a time stamp is added to each data point, and the data is sorted in chronological order. On this basis, the sliding window features are calculated, the window length is set to 5 minutes, and statistical indicators such as the mean, variance, maximum value, and minimum value of blood glucose level, blood pressure, and blood viscosity are calculated within each window. At the same time, the first-order difference and the second-order difference are calculated to evaluate the rate and acceleration of parameter changes. For heart rate data, the short-time Fourier transform (STFT) is used to calculate the spectral features, and the low-frequency and high-frequency power ratios are extracted to evaluate the regulation state of the autonomic nervous system. In addition, based on the Kalman filter method, the time series data is smoothed to reduce the influence of measurement noise on the data and ensure the continuity of the data. After the time series labeling is completed, abnormal state parameters need to be extracted from the data to screen out abnormal physiological states that affect infusion control. First, the three-sigma principle (3σ rule) is used to set the outlier threshold, and abnormal point detection is performed on parameters such as blood glucose level, blood pressure, and blood viscosity. When the blood glucose level exceeds the normal range (below 3.9 mmol / L or above 11.1 mmol / L), it is marked as an abnormal blood glucose state; when the blood pressure exceeds 140 / 90 mmHg or is below 90 / 60 mmHg, it is marked as a hypertension or hypotension state.For blood viscosity, the K-means clustering method is used to divide the data into three categories: normal, slightly high, and extremely high, and the data points in the high blood viscosity state are extracted. In addition, based on wavelet transform to decompose the hemodynamic data, the vascular compliance index is calculated, and patients with vascular sclerosis or decreased elasticity are screened. After all abnormal state parameters are extracted, they are stored in the dataset and used as a reference for subsequent infusion control.

[0079] Step S2 includes the following steps:

[0080] Step S21: Quantify the hypertension fluctuation value of the abnormal state physiological parameters to obtain the hypertension fluctuation quantification value;

[0081] Step S22: Conduct a simulation assessment of the vascular wall pressure bearing according to the hypertension fluctuation quantification value to obtain the vascular wall pressure bearing data;

[0082] Step S23: Fit the increment of blood viscosity with blood glucose change based on the abnormal state physiological parameters to obtain the increment data of blood viscosity with blood glucose change;

[0083] Step S24: Conduct a regression analysis of the vascular infusion resistance according to the vascular wall pressure bearing data and the increment data of blood viscosity with blood glucose change to obtain the vascular infusion resistance data.

[0084] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0085] Step S21: Quantify the hypertension fluctuation value of the abnormal state physiological parameters to obtain the hypertension fluctuation quantification value;

[0086] In the embodiment of the present invention, when quantifying the hypertension fluctuation value of the abnormal state physiological parameters, first obtain the marked hypertension abnormal state data, including the time series data of systolic blood pressure, diastolic blood pressure, and pulse pressure difference. The sliding window technique is adopted, and each window is set to 10 minutes. Calculate the mean value, variance, and maximum and minimum fluctuation amplitudes of blood pressure during this time period, and extract the blood pressure fluctuation characteristics. Use discrete wavelet transform to decompose the blood pressure time series, extract the low-frequency and high-frequency components, and analyze the short-term fluctuation and long-term trend change of blood pressure. Use autocorrelation analysis to calculate the periodic change characteristics of blood pressure fluctuation to determine whether there is circadian rhythm disorder in blood pressure. Based on the variational mode decomposition method, decompose the blood pressure data and extract the energy proportion of each order component to quantify the severity of hypertension fluctuation. Finally, through data normalization, convert the blood pressure fluctuation value into a quantization index within the standard range for subsequent calculation of the vascular wall pressure bearing capacity.

[0087] Step S22: Conduct a simulation assessment of the vascular wall pressure bearing according to the hypertension fluctuation quantification value to obtain the vascular wall pressure bearing data;

[0088] In the embodiments of the present invention, when performing a simulation evaluation of the pressure bearing of the blood vessel wall according to the quantified value of blood pressure fluctuation, first, a calculation framework for blood vessel wall pressure is established. Combining the principles of fluid mechanics and the elastic mechanics characteristics of the blood vessel wall, the dynamic pressure borne by the blood vessel wall is calculated. Using the finite element method, the blood vessel wall structure is divided into multiple grid units to simulate the stress conditions of the blood vessel wall under different blood pressure fluctuation conditions. Based on the Poiseuille flow hypothesis, the shear stress of blood in the blood vessel is calculated, and combined with the blood vessel compliance parameter, the deformation amount of the blood vessel wall is deduced. Using the strain energy density function of the arterial wall, the elastic energy storage of the blood vessel wall is calculated to evaluate the pressure bearing capacity of the blood vessel. Combining the real patient data, inputting the quantified values of blood pressure fluctuation at different stages, analyzing the deformation degree of the blood vessel wall under different pressure conditions, and extracting key pressure bearing parameters, including the maximum stress point of the blood vessel wall, the average strain distribution, and the change trend of the blood vessel wall thickness.

[0089] Step S23: Based on the physiological parameters in the abnormal state, perform fitting on the increment of blood viscosity due to blood glucose change to obtain the data of the increment of blood viscosity due to blood glucose change;

[0090] In the embodiments of the present invention, when performing fitting on the increment of blood viscosity due to blood glucose change based on the physiological parameters in the abnormal state, first, the data of abnormal fluctuations in blood glucose concentration are screened, and the time series data of blood viscosity are obtained. Pearson correlation analysis is used to calculate the correlation between blood glucose change and blood viscosity, and the time lag parameter with significant correlation is determined. The sample entropy method is used to evaluate the complexity of blood glucose fluctuation, calculate the slope of blood glucose change, and extract key fluctuation characteristics. Linear regression analysis is performed on the blood glucose change and blood viscosity data to calculate the incremental contribution of blood glucose concentration change to blood viscosity. For the non-linear relationship, the locally weighted regression method is used for fitting to analyze the influence of short-term and long-term changes in blood glucose level on blood viscosity. In order to further improve the fitting accuracy, multiple regression analysis is adopted, comprehensively considering factors such as blood glucose level, plasma fibrinogen content, and red blood cell aggregation index, and finally the data of the increment of blood viscosity caused by blood glucose change are obtained.

[0091] Step S24: Perform regression analysis on the blood vessel infusion resistance based on the blood vessel wall pressure bearing data and the data of the increment of blood viscosity due to blood glucose change to obtain the blood vessel infusion resistance data.

[0092] In the embodiments of the present invention, when performing regression analysis on the vascular infusion resistance based on the blood vessel wall pressure bearing data and the blood viscosity increment data due to blood glucose changes, first, the two types of data are integrated to establish a multi-variable data set. The principal component analysis method is used to reduce the data dimension and extract the characteristic variables that have a greater impact on the vascular infusion resistance. Based on the logistic regression model, the contribution weights of the blood vessel wall bearing capacity and the blood viscosity change to the infusion resistance are calculated. Piecewise regression analysis is performed on the data, corresponding resistance regression equations are established under different blood vessel states, and the regression coefficients are calculated. The non-linear least squares method is used to optimize the regression parameters to ensure that the model can accurately fit the infusion resistance under different blood flow states. Finally, the infusion resistance under different blood vessel states is calculated and stored for subsequent adjustment of the infusion control parameters.

[0093] Step S23 includes the following steps:

[0094] Step S231: Extract the blood glucose concentration change from the physiological parameters in the abnormal state to obtain the blood glucose concentration change data;

[0095] Step S232: Identify the vascular endothelial cell injury gradient based on the blood glucose concentration change data to obtain the vascular endothelial cell injury gradient;

[0096] Step S233: Predict the increment of plasma fibrinogen content according to the vascular endothelial cell injury gradient and the blood glucose concentration change data to obtain the plasma fibrinogen increment data;

[0097] Step S234: Fit the blood viscosity increment due to blood glucose change based on the plasma fibrinogen increment data to obtain the blood viscosity increment data due to blood glucose change.

[0098] In the embodiment of the present invention, when extracting the change of blood glucose concentration for abnormal physiological parameters, first obtain the continuously monitored blood glucose data, including fasting blood glucose, postprandial blood glucose and all-day blood glucose change curves. Select the blood glucose measurement data with a time window of 5 minutes, use the sliding average method to smooth the short-term fluctuations, and use the linear interpolation method to fill the missing data points. Calculate the blood glucose concentration change rate based on the time series difference method, and extract the instantaneous gradient of blood glucose change. Combined with the baseline blood glucose value, calculate the relative blood glucose fluctuation amplitude, and identify the drastic rise and fall change points. Use the dynamic time warping method to align the blood glucose curves of different patients to eliminate the influence of individual differences, and finally obtain a standardized blood glucose concentration change data set. When identifying the gradient of vascular endothelial cell damage based on blood glucose concentration change data, first extract the high blood glucose peak point and calculate the peak duration. Combined with the blood glucose fluctuation frequency analysis, screen the short-term high-frequency fluctuation feature points, and calculate the blood glucose amplitude change within the fluctuation period. Based on the blood glucose concentration change rate, set the damage threshold, and mark the high fluctuation area exceeding the threshold. Calculate the long-term vascular endothelial damage risk index using the high blood glucose exposure time parameter. Combined with the data of endothelial function biomarkers, including endothelin-1, angiotensin, and nitric oxide levels, the degree of influence of blood glucose fluctuation on endothelial cell damage was calculated. The degree of damage was divided into three gradients of low, medium, and high by a hierarchical clustering algorithm, and the endothelial cell damage gradient data was output. When predicting the increment of plasma fibrinogen content based on the endothelial cell damage gradient and blood glucose concentration change data, the time interval with a higher damage gradient was first screened, and the corresponding blood glucose fluctuation characteristics were extracted. The partial least squares regression method was used to calculate the weight of the influence of blood glucose concentration changes on plasma fibrinogen content. Combined with historical blood test data, samples with high blood glucose accompanied by high fibrinogen levels were screened, and the association matrix between blood glucose changes and fibrinogen content was established. The regression relationship between the rate of change of blood glucose concentration, the degree of endothelial damage, and the increment of plasma fibrinogen was calculated by using the multivariate linear regression method, and the model parameters were adjusted to improve the prediction accuracy. Finally, the increment data of plasma fibrinogen at different time points were generated and stored for subsequent blood viscosity calculation. When fitting the blood viscosity increment of blood glucose changes based on the plasma fibrinogen increment data, first obtain the plasma viscosity measurement data and extract the time series features related to the fibrinogen content. Combined with the blood glucose change data, the blood viscosity increment data is aligned and the outliers are removed. The local regression method is used to calculate the contribution rate of plasma fibrinogen increment to blood viscosity, and a data fitting curve is established. The kernel density estimation method is further used to calculate the probability distribution of blood viscosity changes caused by changes in blood glucose concentration, and analyze the blood rheological properties under different blood glucose levels. Finally, combined with the changes in blood glucose concentration and fibrinogen increment data at different time points, the blood viscosity increment data of blood glucose changes is output and used for subsequent infusion resistance analysis and infusion flow rate control.

[0099] In another embodiment, the time series data of the patient's blood glucose concentration within 72 hours is obtained, with a sampling frequency of 1 time per 5 minutes, and the data format is (timestamp, blood glucose value in mg / dL). The sliding window standard deviation method is used for trend analysis. The window length is set to 20 sampling points (100 minutes), and the standard deviation of the blood glucose values within each window is calculated. When the standard deviations of three consecutive windows all exceed the preset threshold (15 mg / dL), this period is marked as a significantly fluctuating interval of blood glucose. The absolute difference between the blood glucose values of adjacent sampling points is extracted through differential operation, and the data points with a difference greater than 10 mg / dL are selected as key fluctuation points. The timestamps of the key fluctuation points and the corresponding blood glucose values are combined into a triple (timestamp, baseline blood glucose value, fluctuation amplitude) to form a dataset of blood glucose concentration changes. For example, a certain period of data contains records such as (08:00, 120 mg / dL, +18 mg / dL), (08:05, 138 mg / dL, -22 mg / dL), etc. This dataset will be used as the input feature for subsequent vascular injury assessment. For the identification of the vascular endothelial cell injury gradient, based on the blood glucose concentration change data, an improved Leaky ReLU regression network is used for injury gradient quantification. The input layer of the network consists of three neurons, corresponding to the blood glucose fluctuation amplitude, duration, and frequency characteristics respectively. The hidden layer is set with 15 neurons, using the ReLU activation function. The output layer divides the injury degree into five levels through the Softmax function: mild (0 - 20%), moderate (21 - 40%), significant (41 - 60%), severe (61 - 80%), and extreme (81 - 100%). During training, the labeled data from the MIMIC-III database is used, and the weight parameters are optimized through the backpropagation algorithm. The learning rate is set to 0.001, and the batch size is 32. In the test phase, the samples with a blood glucose fluctuation amplitude ≥ 15 mg / dL and a duration > 30 minutes are input into the network, and the output result is normalized to obtain the injury gradient value. For example, the blood glucose fluctuation characteristics (amplitude 18 mg / dL, duration 45 minutes, frequency 0.8 times per hour) of a certain patient from 08:00 to 09:00 output an injury gradient value of 0.67 after being calculated by the network, corresponding to the significant injury level. A multiple linear regression model is used to predict the change in fibrinogen content. The model input includes four features: blood glucose injury gradient value (0 - 1), platelet count (×10^9 / L), C-reactive protein concentration (mg / L), and erythrocyte sedimentation rate (mm / h). The test data for the past 7 days is extracted from the laboratory information system to establish a training set containing 200 patient data. Through the stepwise regression method, significant variables are screened, and the final model is determined as: ΔF = 0.35 injury gradient + 0.22 platelet count - 0.18 C-reactive protein + 0.15 erythrocyte sedimentation rate + 2.1.During prediction, the damage gradient value (0.67), platelet count (280×10^9 / L), C-reactive protein (8mg / L), and erythrocyte sedimentation rate (32mm / h) of the current period are substituted into the model, and the predicted value of fibrinogen increment is calculated to be 3.82g / L. This predicted value characterizes the degree of enhancement of blood coagulation tendency and serves as a key parameter for subsequent viscosity calculation. An extended model of the Hagen-Poiseuille equation is used for viscosity calculation, and the formula is: η = η0×(1 + αΔF + βHct), where η0 is the base viscosity (3.5mPa·s), α = 0.12 is the fibrinogen coefficient, β = 0.08 is the hematocrit coefficient, and Hct is the hematocrit value (40%). First, the Hct value (42%) is obtained from the blood routine data. Combining with the fibrinogen increment (3.82g / L) in step S233, the blood viscosity increment is calculated: Δη = 0.12×3.82 + 0.08×(42 - 40) = 0.46 + 0.16 = 0.62mPa·s. Adding this increment to the base viscosity, the current blood viscosity is 4.12mPa·s. This result reflects the change in blood fluidity caused by blood glucose fluctuations and serves as an input parameter for vascular infusion resistance analysis. The entire calculation process is executed in real time by an embedded system, and the data update interval is 5 minutes, which is synchronized with the infusion pump control cycle.

[0100] Step S24 includes the following steps:

[0101] Step S241: Perform a temporal gradient prediction of vascular wall sclerosis on the vascular wall pressure-bearing data to obtain the temporal gradient of vascular wall sclerosis;

[0102] Step S242: Perform an isoproportional regression analysis of vascular elasticity weakening based on the temporal gradient of vascular wall sclerosis and the blood viscosity increment data due to blood glucose changes to obtain the regression data of vascular elasticity weakening;

[0103] Step S243: Perform a calculation of increased vascular permeability based on the temporal gradient of vascular wall sclerosis and the regression data of vascular elasticity weakening to obtain the data of increased vascular permeability;

[0104] Step S244: Determine the vascular flow resistance coefficient based on the regression data of vascular elasticity weakening, the temporal gradient of vascular wall sclerosis, and the data of increased vascular permeability to obtain the vascular flow resistance coefficient;

[0105] Step S245: Perform a regression analysis of vascular infusion resistance based on the vascular flow resistance coefficient to obtain the vascular infusion resistance data.

[0106] In the embodiments of the present invention, when predicting the time-series gradient of vascular wall sclerosis for vascular wall pressure-bearing data, first, long-term blood pressure monitoring data is extracted, including the changes in systolic blood pressure, diastolic blood pressure, and pulse pressure, and combined with indicators such as patient age, blood lipid level, and inflammatory factors to screen out samples with a tendency of vascular sclerosis. The compliance change of the vascular wall is calculated using the blood pressure fluctuation range, and combined with historical ultrasound image data, the growth rate of the vascular wall thickness is calculated by fitting. The exponential smoothing method is used to process the time-series data to generate the time-series gradient of the vascular sclerosis rate. Combining with the vascular compliance reduction rate, the degree of vascular wall sclerosis at different time points is calculated and normalized to obtain the time-series gradient dataset of vascular wall sclerosis. When performing proportional regression analysis of vascular elasticity weakening based on the time-series gradient of vascular wall sclerosis and the incremental data of blood viscosity due to blood glucose changes, first, the vascular sclerosis data and the incremental data of blood viscosity are aligned, and the feature points within the same time window are selected. The least squares method is used to calculate the regression coefficient between the degree of vascular wall sclerosis and the change in blood viscosity, and the goodness of fit of the regression model is verified. Based on the historical vascular elasticity measurement data, the vascular dilation ability at different sclerosis degrees is calculated, and the hierarchical regression method is used for elasticity weakening analysis. Through the partial least squares regression method, the influence ratio of the increase in blood viscosity caused by blood glucose changes on the decrease in vascular elasticity is calculated, and the regression data of vascular elasticity weakening is output. When performing vascular permeability exacerbation calculation based on the time-series gradient of vascular wall sclerosis and the regression data of vascular elasticity weakening, first, the high-risk time interval in the vascular sclerosis data is extracted, and the corresponding vascular elasticity attenuation rate is selected. Using hemodynamic measurement data, vascular endothelial permeability parameters are calculated, including the transendothelial transport rate, the perivascular space fluid exchange rate, etc. The stepwise regression method is used to calculate the quantitative relationship between the decrease in vascular elasticity and the change in permeability, and through the feature screening method, the low-correlation data points are removed. Combining with the vascular leakage experiment data, the vascular permeability change rate at different sclerosis degrees is calculated, and the vascular permeability exacerbation data is generated. When determining the vascular flow resistance coefficient based on the regression data of vascular elasticity weakening, the time-series gradient of vascular wall sclerosis, and the vascular permeability exacerbation data, first, the time period with the highest vascular elasticity attenuation rate is extracted, and the change in vascular diameter is calculated. Combining with the vascular internal fluid shear force data, the multivariate linear regression method is used to calculate the flow resistance change trend under different vascular states. Based on the blood flow viscosity calculation method, the change in the internal resistance of the blood vessel is evaluated, and combined with the degree of vascular wall sclerosis, the calculation parameters of the flow resistance coefficient are adjusted. Through the curve fitting method, the formula for calculating the vascular flow resistance coefficient is generated, and cross-validation is performed using the data of different patients to obtain the final vascular flow resistance coefficient. When performing vascular infusion resistance regression analysis based on the vascular flow resistance coefficient, first, the historical infusion data corresponding to different vascular flow resistance coefficients is selected, and the key parameters such as infusion flow rate and infusion pressure are extracted. The multivariate regression method is used to calculate the mapping relationship between the infusion pressure and the vascular flow resistance coefficient.Combined with the hemodynamic model, calculate the decay of the infusion flow rate under different vascular states, and based on the flow control algorithm, evaluate the changing trend of the vascular infusion resistance. By means of the piecewise linear regression method, calculate the optimal infusion parameters under different vascular resistance states and generate infusion resistance data for subsequent infusion pressure regulation and flow matching control.

[0107] Step S244 includes the following steps:

[0108] Quantify the dynamic contraction radius of the blood vessel according to the vascular elasticity weakening regression data and the temporal gradient of vascular wall sclerosis to obtain the dynamic contraction radius of the blood vessel;

[0109] Estimate the formation probability of local coagulation factors based on the data of increased vascular permeability to obtain the local coagulation formation probability;

[0110] Perform piecewise integration of hemodynamic loss according to the dynamic contraction radius of the blood vessel and the local coagulation formation probability to obtain piecewise data of hemodynamic loss;

[0111] Analyze the vascular flow resistance distribution of the piecewise data of hemodynamic loss to obtain vascular flow resistance distribution data;

[0112] Determine the vascular flow resistance coefficient according to the piecewise data of hemodynamic loss and the vascular flow resistance distribution data to obtain the vascular flow resistance coefficient.

[0113] In the embodiments of the present invention, when quantifying the dynamic contraction radius of blood vessels according to the blood vessel elasticity weakening regression data and the blood vessel wall sclerosis time series gradient, first, the changes in blood vessel diameter at different time points in the blood vessel elasticity weakening regression data are extracted and data-matched with the blood vessel wall sclerosis time series gradient. Using high-resolution blood vessel ultrasound image data, the average thickness change value of the blood vessel wall in different contraction states is calculated. Based on the multi-point discrete measurement method, the change trend of the blood vessel contraction radius within a specific time window is recorded, and the spline interpolation method is used to generate a data set of the blood vessel dynamic contraction radius with a continuous time series. When estimating the local coagulation factor formation probability based on the blood vessel permeability exacerbation data, first, a time period with a high endothelial permeability in the blood vessel permeability exacerbation data is selected, and the corresponding changes in plasma protein concentration are extracted. Combining the platelet aggregation rate data measured in the experiment, the Poisson distribution model is used to calculate the probability density of coagulation factor aggregation per unit time. Based on the blood flow velocity distribution data, the shear force magnitude in different regions of the blood vessel is calculated, and the conditional probability calculation method is used to evaluate the formation probability of local coagulation factors in the high permeability region, generating a local coagulation formation probability data set. When performing piecewise integration of the blood flow dynamic loss according to the blood vessel dynamic contraction radius and the local coagulation formation probability, first, the blood vessel contraction radius is classified into layers, and multiple interval ranges are set according to different contraction amplitudes. Combining the blood flow velocity data, the flow energy loss within each interval is calculated, and the numerical integration method is used to perform piecewise summation of the blood flow dynamic loss in different contraction intervals. Using the boundary condition correction method, the additional energy loss caused by blood vessel sclerosis and permeability changes is compensated, and the piecewise data of the blood flow dynamic loss is generated. When performing an analysis of the blood vessel flow resistance distribution on the piecewise data of the blood flow dynamic loss, first, the region with high energy loss in the blood flow dynamic loss data is extracted, and the hydrodynamic characteristics of this region are analyzed. Based on the pipe flow theory, the flow resistance changes at different cross-sections of the blood vessel are calculated, and the gradient recursion method is used to evaluate the non-linear distribution relationship of the flow resistance with respect to the blood vessel diameter change. Combining the blood vessel image data, the blood vessel wall shear stress distribution is calculated, and the equal-proportion normalization method is used to generate the blood vessel flow resistance distribution data. When determining the blood vessel flow resistance coefficient according to the piecewise data of the blood flow dynamic loss and the blood vessel flow resistance distribution data, first, the blood flow dynamic loss data is normalized, and the change trend of the flow resistance under different infusion pressures is extracted. Combining the blood vessel flow resistance distribution data, the weighted average method is used to calculate the comprehensive flow resistance coefficient of different blood vessel regions. Using the numerical regression method, the flow resistance change rate in different elastic states of the blood vessel is calculated and corrected based on the historical infusion data, and finally, the blood vessel flow resistance coefficient data is generated for use in subsequent calculation of infusion control parameters.

[0114] Step S3 includes the following steps:

[0115] Step S31: Normalize the blood vessel infusion resistance data to obtain the normalized infusion resistance data;

[0116] Step S32: Dynamically regulate the infusion pressure range interval according to the normalized infusion resistance data to obtain dynamically regulated data for the infusion pressure range interval;

[0117] Step S33: Match the infusion rate with the dynamically regulated data for the infusion pressure range interval to obtain gradient-matched data for the infusion rate;

[0118] Step S34: Based on the dynamically regulated data for the infusion pressure range interval and the gradient-matched data for the infusion rate, perform flow gradient control matching per unit time to obtain infusion flow gradient control data.

[0119] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0120] Step S31: Normalize the vascular infusion resistance data to obtain normalized infusion resistance data;

[0121] In the embodiment of the present invention, when normalizing the vascular infusion resistance data, first extract the resistance values at different time points from the vascular infusion resistance dataset, and calculate the maximum value, minimum value, and average value according to the statistical feature analysis method. Using the min-max normalization method, convert the resistance data to the numerical interval from 0 to 1, and adopt the Z-score standardization method to eliminate the influence of data deviation. Combining historical infusion data, perform trend analysis on the resistance changes in different infusion stages, and eliminate abnormal data through the outlier detection method to generate normalized infusion resistance data to provide a unified standard data input format.

[0122] Step S32: Dynamically regulate the infusion pressure range interval according to the normalized infusion resistance data to obtain dynamically regulated data for the infusion pressure range interval;

[0123] In the embodiment of the present invention, when dynamically regulating the infusion pressure range interval according to the normalized infusion resistance data, first extract the normalized infusion resistance values, and calculate the optimal infusion pressure interval corresponding to different resistance levels according to the statistical distribution. Adopt the gradient optimization method to dynamically adjust the infusion pressure range, and combine the vascular compliance data to calculate the bearing capacity of the blood vessel at different pressures. Based on the real-time blood pressure monitoring data, adopt the adaptive control strategy to set the safety pressure threshold, and dynamically adjust the infusion pressure range through the differential correction method to ensure the stability and safety of the infusion pressure.

[0124] Step S33: Match the infusion rate with the dynamically regulated data for the infusion pressure range interval to obtain gradient-matched data for the infusion rate;

[0125] In the embodiments of the present invention, when matching the infusion rate with the dynamic regulation data of the infusion pressure range interval, first, the optimal infusion rates within different pressure ranges are extracted, and based on the hydrodynamic calculation method, the changes in the infusion rate under different blood vessel resistance levels are evaluated. The multivariate regression analysis method is used to calculate the response relationship of the infusion rate with respect to the pressure change, and in combination with the blood viscosity data, the influence of the blood flow characteristics on the infusion rate is calculated. The interpolation algorithm is used to generate a continuous infusion rate adjustment curve, and the boundary constraint optimization method is adopted to ensure that the infusion rate fluctuates within a reasonable range, and finally the infusion rate gradient matching data is generated.

[0126] Step S34: Based on the dynamic regulation data of the infusion pressure range interval and the infusion rate gradient matching data, perform flow gradient control matching per unit time to obtain the infusion flow gradient control data.

[0127] In the embodiments of the present invention, when performing flow gradient control matching per unit time based on the dynamic regulation data of the infusion pressure range interval and the infusion rate gradient matching data, first, the rate change conditions in different time periods are extracted from the infusion rate gradient matching data, and in combination with the infusion pressure adjustment data, the flow change trend per unit time is calculated. The time series analysis method is used to predict the flow gradient changes in future time periods, and the dynamic weighted average method is adopted to smooth the flow fluctuation data. Based on the optimal control theory, the optimal flow adjustment strategy is calculated, and in combination with the mechanical response characteristics of the infusion pump, the flow adjustment step size is set, and finally the infusion flow gradient control data is generated to optimize the flow output accuracy and stability of the infusion pump.

[0128] Step S32 includes the following steps:

[0129] Step S321: Perform segmented trend analysis on the normalized infusion resistance data, calculate the resistance gradient change rate within adjacent time windows, and based on the gradient change rate, screen out the abnormal resistance mutation points to obtain the resistance gradient mutation data;

[0130] Step S322: Perform mutation point clustering analysis on the resistance gradient mutation data to obtain the resistance mutation mode data;

[0131] Step S323: Perform numerical interval analysis of the blood vessel fluid bearing pressure according to the resistance mutation mode data to obtain the blood vessel fluid bearing pressure numerical interval;

[0132] Step S324: Perform dynamic regulation of the infusion pressure range interval according to the blood vessel fluid bearing pressure numerical interval to obtain the dynamic regulation data of the infusion pressure range interval.

[0133] In the embodiments of the present invention, when performing piecewise trend analysis on the normalized infusion resistance data, first divide the infusion resistance data according to a preset time window. For example, the length of each time window is 5 seconds. Within each time window, calculate the gradient change rate of the resistance value, that is, the ratio of the resistance difference between two adjacent time points to the time interval. By calculating the resistance gradient change rate within adjacent time windows, signs of mutation can be captured. Based on the gradient change rate, set a threshold. When the gradient change rate exceeds this threshold, it is determined as a resistance mutation point. Through this method, abnormal resistance mutation points caused by blood vessel changes or other factors during the infusion process can be accurately screened out. The data of these abnormal resistance mutation points constitutes the resistance gradient mutation data. When performing mutation point clustering analysis on the resistance gradient mutation data, first, based on the similarity of the time series, select a suitable distance measurement method (such as the Euclidean distance) to cluster different mutation points. Through this clustering method, similar mutation points are classified into the same group to identify different types of resistance mutation patterns. For example, one type is the mutation caused by blood vessel constriction, and the other is the mutation caused by blood vessel dilation. Using the K-means clustering algorithm, group the mutation points according to their change amplitude and occurrence frequency to obtain the resistance mutation pattern data. Each resistance mutation pattern represents a potential blood vessel dynamic characteristic, providing a basis for subsequent analysis. When performing blood vessel fluid pressure-bearing numerical interval analysis based on the resistance mutation pattern data, first, according to different resistance mutation pattern data, combined with the physiological parameters of the blood vessel, calculate the maximum fluid pressure that the blood vessel can bear under each pattern. For example, for the constriction type mutation pattern, it is necessary to calculate the maximum pressure-bearing capacity of the blood vessel wall, while for the dilation type mutation pattern, the dilation limit of the blood vessel needs to be considered. Combining the actual infusion data, using the multiple regression analysis method, divide the blood vessel pressure-bearing capacity corresponding to different resistance mutation patterns into numerical intervals. These numerical intervals represent the pressure ranges required for infusion under different blood vessel states. Finally, when performing dynamic regulation of the infusion pressure range interval based on the blood vessel fluid pressure-bearing numerical interval, first, combine the currently measured blood vessel state and historical resistance mutation pattern data to dynamically adjust the output pressure of the infusion pump. For example, when it is detected that the blood vessel is in a constricted state, automatically increase the output pressure of the infusion pump to ensure smooth blood flow; while when the blood vessel dilates, reduce the infusion pressure. Through the regulation strategy, optimize the infusion pressure range in real time to ensure the stability of the blood vessel and the effectiveness of the infusion during the infusion process. Finally, obtain the dynamic regulation data of the infusion pressure range interval.

[0134] Step S4 includes the following steps:

[0135] Step S41: Perform iterative learning on the infusion logic of the infusion pump according to the dynamic regulation data of the infusion pressure range interval, the infusion rate gradient matching data, and the infusion flow gradient control data to obtain the iterative infusion logic data;

[0136] Step S42: Based on the policy gradient algorithm, construct an infusion adaptive parameter control model for the infusion logic iterative data to obtain the infusion adaptive parameter control model;

[0137] Step S43: Embed the infusion adaptive parameter control model into the infusion pump control system to execute the control of the infusion pump.

[0138] In the embodiments of the present invention, when iteratively learning the infusion logic of an infusion pump, first, dynamic regulation data of the infusion pressure range interval, infusion rate gradient matching data, and infusion flow gradient control data are collected. Each type of data represents the infusion conditions at different stages and the corresponding changes in blood vessel states. By analyzing this data, the rules and correlations during the infusion process can be discovered. By repeatedly training this data, the control parameters of the infusion pump are gradually adjusted to ensure the stability of each operation. Specifically, in the initial stage, a default parameter value range is set, and then through repeated experiments, the infusion pressure, rate, and flow are gradually adjusted, and the parameters are adjusted according to real-time feedback. As the experimental data gradually accumulates, the infusion pump will gradually learn how to optimize the infusion operation according to different blood vessel conditions and blood flow states. Finally, through this iterative learning process, optimized infusion logic iterative data is obtained. When constructing an infusion adaptive parameter control model based on the policy gradient algorithm for the infusion logic iterative data, first, a policy gradient algorithm is selected, such as Proximal Policy Optimization (PPO) or deep Q-learning, etc. These algorithms adjust the control strategy of the infusion pump in real time through the input of the infusion logic iterative data, so as to achieve adaptive parameter control. For example, during the infusion process, if it is detected that the pressure of the blood vessel is too high or too low, the policy gradient algorithm will adjust the infusion rate or pressure to avoid blood vessel damage. The algorithm optimizes the effect of each operation by continuously updating the model and maximizing the reward function. Specifically, a loss function can be set to evaluate the effect of the current model output, and the parameters of the infusion pump are optimized according to the feedback value. In each training, by continuously adjusting the weights and parameters in the model, the infusion pump can make the best response according to the changes in external conditions. Finally, a well-trained infusion adaptive parameter control model is obtained. When embedding the trained infusion adaptive parameter control model into the infusion pump control system, first, the model is imported into the control unit of the pump, and the parameters of the model are interfaced with the control hardware of the infusion pump through a programming language (such as C++ or Python). The control system calculates and adjusts the parameters such as the infusion pressure, rate, and flow of the pump in real time according to the input physiological parameter data and sensor feedback. The embedded control system can dynamically adjust the infusion process according to the real-time situation to ensure that each operation during the infusion process meets the physiological requirements of the human blood vessels. In actual operation, when it is detected that data such as blood glucose concentration and blood vessel elasticity change, the infusion pump control system will adjust the working parameters of the infusion pump in real time according to the optimal control strategy provided by the model. In this way, it can be ensured that the infusion pump operates stably under different physiological states, avoiding the situation where the parameter range set manually cannot adapt to the actual needs, and finally realizing automated and precise infusion control.

[0139] The present invention also provides a control system for an infusion pump, which is used to execute the control method of the infusion pump as described above. The control system of the infusion pump includes:

[0140] An abnormal physiological parameter extraction module, which is used to obtain the physiological parameters of the infusion patient; extract the abnormal state parameters from the physiological parameters of the infusion patient to obtain the abnormal state physiological parameters;

[0141] An infusion resistance regression analysis module, which is used to perform fitting of the blood viscosity increment with blood glucose change based on the abnormal state physiological parameters to obtain the blood viscosity increment data with blood glucose change; perform regression analysis on the vascular infusion resistance based on the blood viscosity increment data with blood glucose change to obtain the vascular infusion resistance data;

[0142] A flow gradient control module, which is used to perform dynamic regulation of the infusion pressure range interval according to the vascular infusion resistance data to obtain the dynamic regulation data of the infusion pressure range interval; perform flow gradient control matching per unit time based on the dynamic regulation data of the infusion pressure range interval to obtain the infusion flow gradient control data;

[0143] An adaptive parameter control model construction module, which is used to construct an infusion adaptive parameter control model based on the policy gradient algorithm for the dynamic regulation data of the infusion pressure range interval and the infusion flow gradient control data to obtain the infusion adaptive parameter control model; embed the infusion adaptive parameter control model into the infusion pump control system to execute the control of the infusion pump.

[0144] The above are only the specific embodiments of the present invention, which enable those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A control method for an infusion pump, characterized in that: The following steps are involved: Step S1: Acquire physiological parameters of the infusion patient; extract abnormal state parameters of the infusion patient's physiological parameters to obtain abnormal state physiological parameters; Step S2: performing blood sugar change blood viscosity increment fitting based on abnormal physiological parameters to obtain blood sugar change blood viscosity increment data; performing vascular infusion resistance regression analysis based on blood sugar change blood viscosity increment data to obtain vascular infusion resistance data; Step S3: dynamically control the infusion pressure range interval according to the vascular infusion resistance data to obtain the infusion pressure range interval dynamic control data; perform flow gradient control matching within a unit time based on the infusion pressure range interval dynamic control data to obtain the infusion flow gradient control data; Step S4: Based on the policy gradient algorithm, an infusion adaptive parameter control model is constructed for the dynamic control data of the infusion pressure range and the infusion flow gradient control data to obtain the infusion adaptive parameter control model; the infusion adaptive parameter control model is embedded into the infusion pump control system to execute the control of the infusion pump.

2. The control method of the infusion pump according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquiring physiological parameters of the infusion patient; Step S12: Filling missing values ​​of the physiological parameters of the infusion patient to obtain the physiological filling parameters of the infusion patient; Step S13: performing time series marking processing on the physiological filling parameters of the infusion patient to obtain the time series marked physiological parameters of the infusion patient; Step S14: extracting abnormal state parameters from the time-series marked physiological parameters of the infusion patient to obtain abnormal state physiological parameters.

3. The control method of the infusion pump according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: quantifying the hypertension fluctuation numerical value of the abnormal physiological parameters to obtain the hypertension fluctuation quantification numerical value; Step S22: performing a simulation evaluation of the vascular wall pressure bearing capacity according to the quantified value of the hypertension fluctuation to obtain vascular wall pressure bearing capacity data; Step S23: performing blood sugar change and blood viscosity increment fitting based on abnormal physiological parameters to obtain blood sugar change and blood viscosity increment data; Step S24: Perform vascular infusion resistance regression analysis based on the vascular wall pressure bearing data and the blood sugar change blood viscosity increment data to obtain vascular infusion resistance data.

4. The control method of the infusion pump according to claim 3, characterized in that: Step S23 includes the following steps: Step S231: extracting blood glucose concentration change of abnormal physiological parameters to obtain blood glucose concentration change data; Step S232: performing vascular endothelial cell damage gradient identification based on blood glucose concentration change data to obtain the vascular endothelial cell damage gradient; Step S233: predicting the plasma fibrinogen content increment according to the vascular endothelial cell injury gradient and the blood glucose concentration change data to obtain the plasma fibrinogen content increment data; Step S234: performing blood glucose change blood viscosity increment fitting based on the plasma fibrinogen increment data to obtain blood glucose change blood viscosity increment data.

5. The control method of the infusion pump according to claim 3, characterized in that: Step S24 includes the following steps: Step S241: predicting the time series gradient of vascular wall hardening based on the vascular wall pressure bearing data to obtain the time series gradient of vascular wall hardening; Step S242: performing proportional regression analysis of vascular elasticity weakening according to the vascular wall hardening time series gradient and blood sugar change blood viscosity increment data to obtain vascular elasticity weakening regression data; Step S243: performing vascular permeability aggravation calculation based on the vascular wall hardening time series gradient and vascular elasticity weakening regression data to obtain vascular permeability aggravation data; Step S244: determining the vascular flow resistance coefficient according to the vascular elasticity weakening regression data, the vascular wall hardening time series gradient and the vascular permeability aggravation data, and obtaining the vascular flow resistance coefficient; Step S245: Perform vascular infusion resistance regression analysis based on the vascular flow resistance coefficient to obtain vascular infusion resistance data.

6. The control method of the infusion pump according to claim 5, characterized in that: Step S244 includes the following steps: The dynamic contraction radius of blood vessels is quantified based on the regression data of vascular elastic weakening and the time series gradient of vascular wall stiffening to obtain the dynamic contraction radius of blood vessels; Based on the data of increased vascular permeability, the probability of local coagulation factor formation is estimated to obtain the probability of local coagulation formation; The hemodynamic loss is segmentedly integrated according to the dynamic contraction radius of the blood vessels and the probability of local coagulation formation to obtain segmented data of the hemodynamic loss; Performing vascular flow resistance distribution analysis on the segmented data of hemodynamic loss to obtain vascular flow resistance distribution data; The vascular flow resistance coefficient is determined according to the segmented data of hemodynamic loss and the vascular flow resistance distribution data to obtain the vascular flow resistance coefficient.

7. The control method of the infusion pump according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: normalizing the blood vessel infusion resistance data to obtain infusion resistance normalized data; Step S32: dynamically control the infusion pressure range according to the normalized data of the infusion resistance to obtain dynamic control data of the infusion pressure range; Step S33: performing infusion rate matching on the dynamic control data of the infusion pressure range interval to obtain infusion rate gradient matching data; Step S34: Based on the infusion pressure range interval dynamic control data and the infusion rate gradient matching data, flow gradient control matching is performed within a unit time to obtain infusion flow gradient control data.

8. The control method of the infusion pump according to claim 7, characterized in that: Step S32 includes the following steps: Step S321: performing segmented trend analysis on the normalized data of infusion resistance, calculating the resistance gradient change rate in adjacent time windows, and screening abnormal resistance mutation points based on the gradient change rate to obtain resistance gradient mutation data; Step S322: performing mutation point clustering analysis on the resistance gradient mutation data to obtain resistance mutation pattern data; Step S323: performing a vascular fluid pressure numerical range analysis based on the resistance mutation pattern data to obtain a vascular fluid pressure numerical range; Step S324: dynamically control the infusion pressure range according to the blood vessel fluid pressure range to obtain dynamic control data of the infusion pressure range.

9. The control method of the infusion pump according to claim 7, characterized in that: Step S4 includes the following steps: Step S41: performing iterative learning of the infusion logic of the infusion pump according to the dynamic control data of the infusion pressure range, the infusion rate gradient matching data and the infusion flow gradient control data to obtain infusion logic iteration data; Step S42: constructing an infusion adaptive parameter control model for the infusion logic iteration data based on a policy gradient algorithm to obtain an infusion adaptive parameter control model; Step S43: embedding the infusion adaptive parameter control model into the infusion pump control system to execute the control of the infusion pump.

10. A control system for an infusion pump, characterized in that: For executing the control method of the infusion pump according to claim 1, the control system of the infusion pump comprises: The abnormal physiological parameter extraction module is used to obtain the physiological parameters of the infusion patient; the abnormal state parameters are extracted from the physiological parameters of the infusion patient to obtain the abnormal state physiological parameters; The infusion resistance regression analysis module is used to perform blood sugar change blood viscosity increment fitting based on abnormal physiological parameters to obtain blood sugar change blood viscosity increment data; perform vascular infusion resistance regression analysis based on blood sugar change blood viscosity increment data to obtain vascular infusion resistance data; A flow gradient control module is used to dynamically control the infusion pressure range according to the vascular infusion resistance data to obtain the dynamic control data of the infusion pressure range; and to match the flow gradient control within a unit time based on the dynamic control data of the infusion pressure range to obtain the infusion flow gradient control data; The adaptive parameter control model construction module is used to construct an infusion adaptive parameter control model for the dynamic control data of the infusion pressure range and the infusion flow gradient control data based on the policy gradient algorithm to obtain the infusion adaptive parameter control model; the infusion adaptive parameter control model is embedded in the infusion pump control system to execute the control of the infusion pump.

Images