Search

CN-121983269-A - Blood purification equipment real-time control method based on edge calculation

CN121983269ACN 121983269 ACN121983269 ACN 121983269ACN-121983269-A

Abstract

The application relates to the crossing field of computer technology and edge calculation, and discloses a real-time control method of blood purification equipment based on edge calculation. The method comprises the steps of fusing the multi-mode real-time physiological parameters, the historical dialysis data and the current treatment target through a local edge computing unit, constructing an individual body fluid dynamic response prediction model, and optimizing and generating an ultrafiltration rate instruction within a millisecond level, wherein the system comprises a multi-mode sensor array, a historical data storage module, a task planning interface, an edge computing unit and an ultrafiltration pump controller. The application realizes the accurate maintenance of the hemodynamic state, obviously reduces the risk of hypotension and improves the treatment safety and individuation level.

Inventors

  • TIAN XIUJUAN
  • WANG DI
  • SUN SHIREN

Assignees

  • 中国人民解放军空军军医大学

Dates

Publication Date
20260505
Application Date
20260114

Claims (10)

  1. 1. A method for controlling a blood purification device in real time based on edge calculation, comprising: real-time physiological parameter sequences of a patient are acquired in real time through a multi-mode physiological sensor array on blood purification equipment, wherein the real-time physiological parameter sequences comprise arterial pressure signals, venous pressure signals, transmembrane pressure signals, blood flow signals, heart rate variability signals, blood oxygen saturation signals and bioelectrical impedance spectrum signals; Acquiring the history dialysis record data of the real-time acquired patient through a history dialysis data storage module arranged in the equipment; Obtaining a target ultrafiltration volume, a target treatment duration and a preset safety constraint boundary of the current dialysis treatment through a task planning interface; inputting the real-time physiological parameter sequence, the historical dialysis record data and the current dialysis treatment parameters to an edge computing unit arranged locally on the equipment; in the edge calculation unit, extracting time sequence characteristics of the real-time physiological parameter sequence to generate a current physiological state dynamic representation; performing personalized baseline modeling on the historical dialysis record data to generate a patient-specific body fluid regulation capacity benchmark characterization; Fusing the current physiological state dynamic representation with the patient-specific body fluid regulation capacity reference representation in a cross-period feature to generate an individualized body fluid dynamic response prediction representation; generating an ultrafiltration rate instruction of the next control period through an ultrafiltration rate online optimization solver based on the personalized body fluid dynamic response prediction characterization and the current dialysis treatment parameters; and sending the ultrafiltration rate instruction to an ultrafiltration pump driving controller so as to adjust the working parameters of the ultrafiltration pump in real time.
  2. 2. The edge computing-based blood purification device real-time control method of claim 1, wherein performing time series feature extraction on the real-time physiological parameter sequence to generate a current physiological state dynamic characterization comprises: organizing the real-time physiological parameter sequence into seven-channel time sequence vectors according to a time dimension, wherein each channel contains data of not less than 500 sampling points in the last 50 seconds; inputting the seven-channel time sequence vector into a one-dimensional convolutional neural network, wherein the one-dimensional convolutional neural network comprises 3 convolutional layers which are sequentially connected, the number of convolution kernels of each convolutional layer is 32, 64 and 128, the convolution kernel sizes are 5, an activation function is a correction linear unit, and each convolutional layer is connected with a maximum pooling layer and the pooling window size is 2; and outputting a 128-dimensional current physiological state dynamic characterization vector as the current physiological state dynamic characterization.
  3. 3. The edge computing-based blood purification device real-time control method of claim 2, wherein individualizing baseline modeling the historical dialysis record data to generate a patient-specific body fluid regulation capacity benchmark characterization comprises: using ultrafiltration rate in historical dialysis record data as an input variable and arterial pressure falling slope as an output variable to construct a Gaussian process regression model; Adopting a radial basis function as a covariance function, and learning a length scale and a signal variance parameter from historical data through maximum likelihood estimation; And respectively encoding the sampling value of the mean function of the Gaussian process regression model on a standard ultrafiltration rate grid and the square root of the corresponding square difference point into front 128 dimensions and rear 128 dimensions, and forming 256-dimension body fluid regulation capacity reference representation vectors serving as the patient-specific body fluid regulation capacity reference representation.
  4. 4. The edge computing-based blood purification device real-time control method of claim 3, wherein fusing the current physiological state dynamic characterization with the patient-specific body fluid regulation capacity benchmark characterization across time period features to generate an individualized body fluid dynamic response prediction characterization comprises: Transforming the current physiological state dynamic representation into a query vector through a first full connection layer; transforming the body fluid regulation capacity reference representation into a key vector and a value vector through a second full connection layer; Calculating attention weight through an attention mechanism based on the query vector, the key vector and the value vector and generating fusion output; And taking the fusion output as 256-dimensional individualized fluid dynamic response prediction characterization vector as the individualized fluid dynamic response prediction characterization.
  5. 5. The method for controlling an edge-based blood purification apparatus in real time according to claim 4, wherein the attention weight calculation formula of the attention mechanism is: wherein Is a matrix of values.
  6. 6. The edge-calculation-based blood purification device real-time control method according to claim 5, wherein generating an ultrafiltration rate instruction of a next control period by an ultrafiltration rate online optimization solver based on the personalized body fluid dynamic response predictive representation and the current dialysis treatment parameters, comprises: a model prediction control framework is adopted, a prediction time domain is set to be 30 seconds in the future, and the control time domain is set to be 5 seconds in the future; constructing a nonlinear state space equation description system dynamic model, wherein state variables comprise arterial pressure, venous pressure and effective circulating blood volume estimated values, control input is ultrafiltration rate, and output is a measurable physiological parameter; Taking the sum of squares of deviation of the minimized predicted arterial pressure and the target arterial pressure as an objective function, and applying hard limiting conditions of all safety constraint boundaries; and solving an ultrafiltration rate set value per second in the future 5 seconds by adopting a sequence quadratic programming algorithm to serve as the ultrafiltration rate instruction.
  7. 7. The method for controlling an edge-based blood purification apparatus in real time according to claim 6, wherein the objective function expression is: Wherein, the Is the first The second predicts the arterial pressure and, In order to achieve the target arterial pressure, Is the first The ultrafiltration rate in seconds, To smooth the penalty coefficients.
  8. 8. The method according to claim 7, wherein the model parameters of the system dynamic model are updated in real time by an online recursive least square method, and the forgetting factor is set to 0.98.
  9. 9. The method of claim 8, wherein sending the ultrafiltration rate command to an ultrafiltration pump drive controller to adjust the ultrafiltration pump operating parameters in real time comprises: Converting the ultrafiltration rate command to a pulse width modulated signal; driving a direct current brushless motor to control the rotating speed of an ultrafiltration pump, so that the adjustment accuracy of the volume flow of the ultrafiltration liquid reaches +/-5 milliliters per minute; The ultrafiltration rate command is read every 10 ms and the pwm signal output is updated.
  10. 10. The edge-calculation-based blood purification device real-time control method according to claim 9, wherein the bioelectrical impedance spectrum signal is obtained by applying an alternating excitation current in a frequency range of 5000 hz to 500 khz by a four electrode method and measuring a voltage response, for calculating a ratio of extracellular fluid to total body fluid, the ratio value being involved as a key input feature in generation of a dynamic characterization of a current physiological state.

Description

Blood purification equipment real-time control method based on edge calculation Technical Field The invention belongs to the field of intersection of computer technology and edge calculation, and particularly relates to a real-time control method of blood purification equipment based on edge calculation. Background Along with the wide application of blood purification technology in end-stage renal disease treatment, higher requirements are put on the accurate control and individuation adaptation of the dialysis process. One of the core goals of hemodialysis is to effectively remove excess water from a patient's body by ultrafiltration while maintaining hemodynamic stability. The body fluid distribution of human body has high nonlinearity and time-varying characteristics, and is influenced by multiple factors such as cardiac function, vascular permeability, autonomic nerve regulation, feeding in dialysis, body position change and the like, so that the capacity state is dynamic and complex. The traditional ultrafiltration control strategy mostly adopts open-loop setting or simple closed-loop feedback based on fixed parameters (such as transmembrane pressure or venous pressure is used as single input), lacks the sensing and modeling capability of the real-time physiological state of a patient, is difficult to adapt to the difference between individuals and in individuals, is easy to cause complications such as hypotension, muscle spasm or volume overload, and seriously threatens the treatment safety and comfort. The real-time control method based on edge calculation becomes a key direction for improving the intelligent level of the blood purification equipment. The direction aims at sinking the computing power to the equipment end, and realizing the localization fusion analysis and the rapid decision of multisource physiological signals (such as blood pressure, heart rate variability, biological impedance and ultrafiltration rate) in the dialysis process while guaranteeing the data privacy and communication efficiency. The basic principle is that a lightweight intelligent algorithm is deployed by utilizing limited but low-delay computing power resources at the edge side, the capacity state of a patient is evaluated on line, and the ultrafiltration rate and the displacement fluid flow are dynamically adjusted so as to maintain circulation stability. The prior art faces multiple contradictions in realizing the aim, on one hand, a high-precision physiological model can better characterize a body fluid transfer mechanism, but has high computational complexity, and is difficult to meet the severe requirement of edge equipment on real-time performance, and on the other hand, a simplified control strategy (such as proportional integral derivative control) has good robustness, but lacks self-adaptive capability on nonlinear dynamics and individual specificity, and cannot realize real intelligent regulation. Furthermore, existing systems generally lack the ability to continually learn and optimize long-term control strategies from historical treatment experience, resulting in a stiff control logic. Especially in the face of sudden hemodynamic fluctuations or special pathological conditions (such as heart failure and dialysis), the system response is often delayed or excessive, and it is difficult to combine safety and dehydration efficiency. Therefore, an edge-side real-time control architecture integrating the advantages of offline learning and online execution and having both intelligence and robustness is needed to solve the problem of dynamic and accurate regulation and control of capacity management in the blood purification process. Disclosure of Invention The invention provides a real-time control method of blood purification equipment based on edge calculation, and aims to solve the technical problems that most of traditional ultrafiltration control is open loop or simple closed loop, and complex body fluid dynamic changes in individual differences of patients and dialysis process are difficult to cope with, and hypotension or capacity overload are easy to cause. According to the method, the edge computing unit is deployed locally on the blood purification equipment, the multisource physiological parameters, the historical dialysis data and the current treatment process information are fused in real time, the dynamic self-adaptive ultrafiltration rate regulation model is built, and the on-line optimization and execution of an ultrafiltration strategy are completed within a millisecond time scale, so that the accurate maintenance of the hemodynamic state of a patient is realized. According to one aspect of the invention, a blood purification device real-time control method based on edge calculation is provided, which comprises the steps of collecting a real-time physiological parameter sequence of a patient in real time through a multi-mode physiological sensor array on the blood purification device, obtaining h