Personalized cardiovascular system modeling method based on multi-objective optimization

Adjusting the parameters of the cardiovascular system model through a multi-objective optimization algorithm, solving the problem of insufficient personalization and accuracy of central vasculature modeling in the existing technology, and achieving high-precision modeling and hemodynamic data of the cardiovascular system Access to scientific guidance for clinical diagnosis and treatment.

CN118942713APending Publication Date: 2024-11-12DALIAN NATIONALITIES UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410936624.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing cardiovascular system modeling methods are insufficient in personalization and accuracy, and it is difficult to effectively simulate the hemodynamic status of individual cardiovascular systems, limiting the accuracy of clinical diagnosis and treatment.

Method used

Using a personalized cardiovascular system modeling method based on multi-objective optimization, the zero-dimensional cardiovascular system simulation model and in vitro simulation circulatory system are constructed, and the model parameters are adjusted in combination with a multi-objective evolution algorithm to optimize blood viscosity resistance and vascular compliance. and blood flow moment of inertia to match simulated blood pressure waveforms and measured blood pressure waveforms.

Benefits of technology

It realizes personalized modeling of the cardiovascular system, improves the accuracy and convenience of modeling, can obtain hemodynamic data without direct anthropometric measurement, reduces hardware costs, improves parameter adjustment accuracy, and provides Scientific guidance on the diagnosis and treatment of cardiovascular diseases.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942713A_ABST
    Figure CN118942713A_ABST
Patent Text Reader

Abstract

The invention discloses a personalized cardiovascular system modeling method based on multi-objective optimization. The method comprises the following steps: constructing a zero-dimensional cardiovascular system simulation model based on a cardiovascular system anatomical structure of a human body; building a cardiovascular in-vitro simulation circulatory system for collecting hemodynamic data on the basis of the simulation model; the circulation system simulates the structure of the cardiovascular system and simulates the cardiovascular system to work through motor movement; determining an objective function and a decision variable of a multi-objective algorithm; constructing an objective function according to a root mean square error, a systolic pressure difference value, a diastolic pressure difference value and a dicrotic wave error between the simulated blood pressure waveform of the model and the actually measured blood pressure waveform of the system; taking blood viscous resistance, blood vessel compliance and blood flow inertia as model decision variables; and optimizing the simulation model by adopting a multi-objective evolutionary algorithm to obtain a personalized cardiovascular zero-dimensional model with optimal parameters. The modeling requirements of the personalized cardiovascular system can be met, and an efficient, universal and robust method is provided for parameter estimation of the cardiovascular system model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of personalized cardiovascular system modeling, and in particular to a personalized cardiovascular system modeling method based on multi-objective optimization. Background Art

[0002] At present, the research on diagnostic methods, pathological mechanisms and treatment methods related to cardiovascular diseases often adopts clinical diagnosis and treatment methods. However, due to the difficulty of in vivo measurement technology, individual measurement differences and potential ethical issues, the improvement of the symptomatic diagnosis and treatment level of cardiovascular diseases by clinicians is limited. With the development of computer technology and computing power, the in vitro modeling method of personalized cardiovascular system has received widespread attention. By mathematically modeling the cardiovascular system and solving the model with a computer, the movement of blood in the cardiovascular circulatory system can be simulated, allowing clinicians to more fully understand the internal mechanism and development process of the disease, and carry out some experiments that are not convenient for clinical practice, such as analyzing the causes of cardiovascular disease, observing the changes in hemodynamic parameters during the onset of cardiovascular disease, and evaluating the treatment effect, etc., to provide scientific guidance for future clinical diagnosis and also help the design and development of new medical instruments and equipment.

[0003] Prior art application number: CN202410362378.X discloses a method, device, computer-readable storage medium and product for determining the prognosis effect before coronary intervention surgery, which involves constructing a 0-dimensional model of the whole body cardiovascular system, and based on the individualized data of patients collected clinically, the parameters of the 0-dimensional model are individually calibrated using a parameter optimization algorithm, and after obtaining the calibrated 0-dimensional model, the 1-3-dimensional coupling model of the coronary tree is embedded in the calibrated 0-dimensional model to obtain a 0-1-3-dimensional geometric multi-scale coupling model. Application number: CN202210588962.8 discloses a human cardiovascular system based on lumped parameters, a modeling method and application thereof, which uses a power pump to simulate the heart, an elastic time-varying model to simulate the ventricle, a unidirectional diode to simulate the valve, an elastic circular tube to simulate the blood vessel, and an incompressible Newtonian fluid to simulate the blood, and combines the standard structure of the human cardiovascular system to construct a human cardiovascular system model, and constructs the pressure-volume curve relationship of the ventricle by establishing a ground model, which can effectively simulate the behavior of the ventricle. The above two existing technologies are both aimed at building corresponding models in clinical practice, which are of reference significance for diagnosis or surgery, and the technical problems solved by this method are different. Summary of the invention

[0004] 1. Technical problems to be solved:

[0005] In view of the above technical problems, the present invention provides a personalized cardiovascular system modeling method based on multi-objective optimization.

[0006] 2. Technical solution:

[0007] A personalized cardiovascular system modeling method based on multi-objective optimization, characterized by comprising the following steps:

[0008] Step 1: Based on the anatomical structure of the human cardiovascular system, a zero-dimensional cardiovascular system simulation model including the main arteries, arterioles, capillaries and veins of the human body is constructed;

[0009] Step 2: Building a cardiovascular in vitro simulated circulatory system capable of collecting hemodynamic data based on the zero-dimensional cardiovascular system simulation model; the cardiovascular simulated circulatory system simulates the ventricles, atria, arterial vascular trees, compliance devices, and heart valve structures that constitute the cardiovascular system, and can simulate the work of the cardiovascular system through the reciprocating motion of the servo motor and its control system;

[0010] Step 3: Determine the objective function and decision variables of the multi-objective algorithm; construct the objective function by using the root mean square error, systolic pressure difference, diastolic pressure difference and dicrotic wave error physiological information between the simulated blood pressure waveform simulated by the zero-dimensional cardiovascular system simulation model and the measured blood pressure waveform collected by the cardiovascular in vitro simulation circulation system; use the blood viscosity resistance, vascular compliance and blood flow inertia of the model as the decision variables of the model;

[0011] Step 4: Use a multi-objective evolutionary algorithm to adjust the parameters of the zero-dimensional cardiovascular system simulation model to obtain the optimal parameters, and then obtain a personalized cardiovascular zero-dimensional model with optimal parameters.

[0012] Furthermore, in step one, the constructed zero-dimensional cardiovascular system simulation model includes 166 blood vessels, among which the blood vessels numbered 1-71 are the main medium and large arteries in various parts of the human body, all of which are bifurcated structures; the blood vessels numbered 72-73 are two connected venous structures, and the venous structure is simplified into a series structure of two non-bifurcated vascular segments; the blood vessels numbered 74-76 represent the pulmonary artery, pulmonary arterioles and capillaries and pulmonary veins, respectively, and do not contain bifurcated structures; the blood vessels numbered 77-166 represent arterioles and capillaries; in the in vitro simulated circulatory system constructed in step two, the arterioles and capillaries are connected in series at the ends of the medium and large arteries in a three-section manner, and are uniformly aggregated and flow into the venous structure numbered 72.

[0013] Furthermore, in step 2, the cardiovascular in vitro simulation circulation system can simulate and collect different hemodynamic states under normal cardiovascular conditions and multiple stenosis states, specifically: first measure the blood pressure waveforms of the four arteries, namely the radial artery, splenic artery, internal iliac artery and anterior tibial artery, under normal conditions through the cardiovascular simulation circulation system; then change the degree of stenosis of the femoral artery of the cardiovascular simulation circulation system multiple times, and measure the blood pressure waveforms of the above four arteries; the multiple stenosis states are different degrees of stenosis of the femoral artery.

[0014] Furthermore, the objective function constructed in step 3 is as follows:

[0015] loss 1 =RMSE (1)

[0016] loss 2 =ΔSBP+ΔDBP (2)

[0017] loss 3 =ΔP 1 +ΔP 2 (3)

[0018] loss 4 =ΔT 1 +ΔT 2 +ΔT 3 +ΔT 4 (4)

[0019] loss all1 =a 1 *loss 1 +a 2 *loss 2 +a 3 *loss 3 +a 4 *loss 4 (5)

[0020] In the above formula, loss1 is the overall error RMSE between the measured blood pressure waveform collected by the cardiovascular in vitro simulation circulation system and the simulated blood pressure waveform simulated by the zero-dimensional cardiovascular system simulation model; loss2 is the systolic pressure ΔSBP and diastolic pressure error ΔDBP between the measured blood pressure waveform and the model simulated blood pressure waveform; loss3 is the error of the replay wave part of the measured blood pressure waveform and the model simulated blood pressure waveform, where ΔP1 is the difference in blood pressure between the two at the moment of replay notch; ΔP2 is the difference in the maximum value of the dicrotic wave blood pressure between the two; loss4 is the weighted sum of the errors at each moment, where ΔT1 is the difference in the corresponding moment of the systolic pressure between the two, ΔT2 is the difference in the corresponding moment of the diastolic pressure between the two, ΔT3 is the difference in the corresponding moment of the maximum value of the dicrotic wave blood pressure between the two; ΔT4 is the difference in the corresponding moment of the replay notch between the two; loss all1 is the weighted sum of the errors between the measured blood pressure waveform and the simulated blood pressure waveform, where a1, a2, a3 and a4 are loss all1 Weighting coefficients for different error terms;

[0021] The specific decision variables are: the blood viscosity resistance, vascular compliance, and blood flow inertia of the 166 blood vessels in the zero-dimensional cardiovascular system simulation model constitute the decision variable X, as shown in the following formula:

[0022] X=[R 1 ,R 2 ,…,R 166 ,L 1 ,L 2 ,…,L 76 ,C 1 ,C 2 ,…,C 166 ] (6)

[0023] In the above formula, Ri represents the vascular blood viscosity resistance of each medium and large artery and the small arteries and capillaries at the end of the model, where 1≤i≤166; Cj represents the vascular compliance of each medium and large artery, small artery and capillary, where 1≤j≤166; Lk represents the blood flow inertia of each medium and large artery, 1≤k≤76.

[0024] Furthermore, in step 4, a multi-objective evolutionary algorithm is used to adjust the parameters of the zero-dimensional cardiovascular system simulation model to obtain the optimal parameters, which specifically includes the following steps:

[0025] S41: randomly generating initial values ​​of each decision variable in the decision space; according to the different parameter ranges of each component in the decision variable, using direct uniform distribution or logarithmic transformation random generation method respectively; wherein the direct uniform distribution random generation method is used for larger arteries such as the aorta, and the logarithmic random generation method is used for the parameters of the components whose parameter ranges span more than two orders of magnitude;

[0026] S42: randomly generate the initial value of each decision variable as an individual in the initial population, form a set, and form an initialized population;

[0027] S43: Analyze the type attributes of each decision variable; the type attributes specifically include convergence-related, diversity-related and sensitivity-related type attributes;

[0028] S44: Optimize the decision variables of convergence-related attributes, diversity-related attributes, and sensitivity-related attributes respectively;

[0029] S45: Based on the matching degree between the simulated waveform and the measured waveform, the best individual in the population is selected as a feasible solution to the parameter estimation problem in the cardiovascular in vitro simulated circulation system; specifically, a preset number of decision variables closest to the origin of the decision space are screened out through the Euclidean distance, and the decision variable with the best waveform matching degree is selected as the optimal solution.

[0030] 3. Beneficial effects:

[0031] (1) In a personalized cardiovascular system modeling method based on multi-objective optimization disclosed in this method, first, a zero-dimensional cardiovascular system simulation model is constructed based on the typical human cardiovascular system anatomical structure, which can represent the complete blood circulation system of the human body. The overall structure is composed of the heart, large arteries, small arteries, capillaries and veins, and can achieve complete systemic and pulmonary circulation functions. The application scope of the model is met by designing the structure of medium and large arteries, pulmonary arteries, pulmonary arterioles, capillaries and pulmonary veins, that is, focusing on arterial hemodynamics. The model constructed at the same time contains a total of 166 main medium and large arteries, small arteries and capillaries, which can simulate parameters such as blood pressure, blood flow, and vascular volume of each vascular segment, providing a verification and comparison basis for personalized modeling of the cardiovascular system, improving the accuracy and convenience of modeling, and facilitating clinical promotion and application.

[0032] (2) In this method, in order to collect the hemodynamic state of the human body, an in vitro simulated circulatory system is designed based on the constructed simulation model. The system simulates the structure and working principle of the cardiovascular system; the simulation of different cardiovascular states is achieved by controlling the relevant parameters of the system. In this method, different cardiovascular states are characterized by controlling the different degrees of stenosis of the femoral artery, thereby achieving the hemodynamic data of the human body without direct measurement of the human body. Compared with the commonly used physical modeling method, this method greatly improves the parameter adjustment accuracy while reducing the hardware cost.

[0033] (3) In this method, when optimizing the parameters of the simulation model, the model parameters can be classified into three types of variables with different properties: convergence-related variables, diversity-related variables, and sensitivity-related variables by analyzing the different effects of the parameters in the model on the target values. The convergence-related variables and sensitivity-related variables are optimized by non-dominated ranking, and the diversity-related variables are optimized based on the angle selection strategy. By using different optimization methods for parameters of different properties, the algorithm can converge better.

[0034] In summary, this method realizes personalized modeling of the cardiovascular system by combining physical models with mathematical models, which provides scientific guidance for analyzing the causes of cardiovascular diseases, obtaining data on changes in hemodynamic parameters during the onset of cardiovascular diseases, and evaluating treatment effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is the overall flow chart of the present invention;

[0036] Figure 2 A flowchart of a process of calculating optimization parameters of simulation model parameters using a multi-objective evolutionary algorithm in the present invention;

[0037] Figure 3A schematic diagram of a zero-dimensional cardiovascular system simulation model constructed in the present invention;

[0038] Figure 4 Schematic diagram of the in vitro simulated circulation system designed in this method;

[0039] Figure 5 This is a comparison diagram of the modeling simulation and actual measurement results of the radial artery, splenic artery, internal iliac artery and anterior tibial artery under normal conditions of four targets in a specific embodiment;

[0040] Figure 6 This is a comparison diagram of the modeling simulation and actual measurement results of the radial artery, splenic artery, internal iliac artery and anterior tibial artery in the case of the sixth stenosis stage of the four target femoral arteries in a specific embodiment. DETAILED DESCRIPTION

[0041] The present invention will be described in detail below with reference to the accompanying drawings.

[0042] Attached Figure 1 To Attachment Figure 4 As shown, a personalized cardiovascular system modeling method based on multi-objective optimization is characterized by comprising the following steps:

[0043] Step 1: Based on the anatomical structure of the human cardiovascular system, a zero-dimensional cardiovascular system simulation model including the main arteries, arterioles, capillaries and veins of the human body is constructed;

[0044] Step 2: Building a cardiovascular in vitro simulated circulatory system capable of collecting hemodynamic data based on the zero-dimensional cardiovascular system simulation model; the cardiovascular simulated circulatory system simulates the ventricles, atria, arterial vascular trees, compliance devices, and heart valve structures that constitute the cardiovascular system, and can simulate the work of the cardiovascular system through the reciprocating motion of the servo motor and its control system;

[0045] Step 3: Determine the objective function and decision variables of the multi-objective algorithm; construct the objective function by using the root mean square error, systolic pressure difference, diastolic pressure difference and dicrotic wave error physiological information between the simulated blood pressure waveform simulated by the zero-dimensional cardiovascular system simulation model and the measured blood pressure waveform collected by the cardiovascular in vitro simulation circulation system; use the blood viscosity resistance, vascular compliance and blood flow inertia of the model as the decision variables of the model;

[0046] Step 4: Use a multi-objective evolutionary algorithm to adjust the parameters of the zero-dimensional cardiovascular system simulation model to obtain the optimal parameters, and then obtain a personalized cardiovascular zero-dimensional model with optimal parameters.

[0047] Furthermore, in step one, the constructed zero-dimensional cardiovascular system simulation model includes 166 blood vessels, among which the blood vessels numbered 1-71 are the main medium and large arteries in various parts of the human body, all of which are bifurcated structures; the blood vessels numbered 72-73 are two connected venous structures, and the venous structure is simplified into a series structure of two non-bifurcated vascular segments; the blood vessels numbered 74-76 represent the pulmonary artery, pulmonary arterioles and capillaries and pulmonary veins, respectively, and do not contain bifurcated structures; the blood vessels numbered 77-166 represent arterioles and capillaries; in the in vitro simulated circulatory system constructed in step two, the arterioles and capillaries are connected in series at the ends of the medium and large arteries in a three-section manner, and are uniformly aggregated and flow into the venous structure numbered 72.

[0048] The zero-dimensional cardiovascular system simulation model constructed in this method is shown in the attached Figure 3 As shown, the numbers 1-166 in the figure represent numbers 1-166; the circles with numbers correspond to numbers 1-73; the blood vessels 1-71 are the main large and medium arteries in various parts of the human body, and 72-73 are two connected venous structures; the boxes with numbers correspond to numbers 77-166, representing arterioles and capillaries.

[0049] Furthermore, in step 2, the cardiovascular in vitro simulation circulation system can simulate and collect different hemodynamic states under normal cardiovascular conditions and multiple stenosis states, specifically: first measure the blood pressure waveforms of the four arteries, namely the radial artery, splenic artery, internal iliac artery and anterior tibial artery, under normal conditions through the cardiovascular simulation circulation system; then change the degree of stenosis of the femoral artery of the cardiovascular simulation circulation system multiple times, and measure the blood pressure waveforms of the above four arteries; the multiple stenosis states are different degrees of stenosis of the femoral artery.

[0050] The cardiovascular in vitro simulated circulation system designed in this method is as shown in the attached Figure 4 As shown in the figure, the system includes structures such as ventricles, atria, arterial vascular trees, compliance devices, and heart valves; its working principle is to drive the servo motor to move periodically through the reciprocating motion of the servo motor and the host computer to burn the pre-written servo motor control program into the single-chip microcomputer, and the single-chip microcomputer controls the servo motor to move periodically, so that the punch at the end of the motor periodically squeezes the ventricular structure, thereby simulating the rhythmic contraction and relaxation of the heart in each cardiac cycle, thereby driving the blood circulation of the entire cardiovascular system. The process of collecting data in this method is: first, measure the blood pressure waveforms of the four arteries of radial artery, splenic artery, internal iliac artery and anterior tibial artery under normal conditions, and then measure the blood pressure waveforms of the above four arteries at different stages of femoral artery stenosis by artificially changing the degree of stenosis of the femoral artery. The purpose of this is to create a variety of different hemodynamic states in the system to test the personalized ability of the cardiovascular system model and optimization algorithm designed in this study.

[0051] Furthermore, the objective function constructed in step 3 is as follows:

[0052] loss 1 =RMSE (1)

[0053] loss 2 =ΔSBP+ΔDBP (2)

[0054] loss 3 =ΔP 1 +ΔP 2 (3)

[0055] loss 4 =ΔT 1 +ΔT 2 +ΔT 3 +ΔT 4 (4)

[0056] loss all1 =a 1 *loss 1 +a 2 *loss 2 +a 3 *loss 3 +a 4 *loss 4 (5)

[0057] In the above formula, loss1 is the overall error RMSE between the measured blood pressure waveform collected by the cardiovascular in vitro simulation circulation system and the simulated blood pressure waveform simulated by the zero-dimensional cardiovascular system simulation model; loss2 is the systolic pressure ΔSBP and diastolic pressure error ΔDBP between the measured blood pressure waveform and the model simulated blood pressure waveform; loss3 is the error of the replay wave part of the measured blood pressure waveform and the model simulated blood pressure waveform, where ΔP1 is the difference in blood pressure between the two at the moment of replay notch; ΔP2 is the difference in the maximum value of the dicrotic wave blood pressure between the two; loss4 is the weighted sum of the errors at each moment, where ΔT1 is the difference in the corresponding moment of the systolic pressure between the two, ΔT2 is the difference in the corresponding moment of the diastolic pressure between the two, ΔT3 is the difference in the corresponding moment of the maximum value of the dicrotic wave blood pressure between the two; ΔT4 is the difference in the corresponding moment of the replay notch between the two; loss all1 is the weighted sum of the errors between the measured blood pressure waveform and the simulated blood pressure waveform, where a1, a2, a3 and a4 are loss all1 Weighting coefficients for different error terms;

[0058] The specific decision variables are: the blood viscosity resistance, vascular compliance, and blood flow inertia of the 166 blood vessels in the zero-dimensional cardiovascular system simulation model constitute the decision variable X, as shown in the following formula:

[0059] X=[R1 ,R 2 ,…,R 1+6 ,L 1 ,L 2 ,…,L 76 ,C 1 ,C 2 ,…,C 166 ] (6)

[0060] In the above formula, Ri represents the vascular blood viscosity resistance of each medium and large artery and the small arteries and capillaries at the end of the model, where 1≤i≤166; Cj represents the vascular compliance of each medium and large artery, small artery and capillary, where 1≤j≤166; Lk represents the blood flow inertia of each medium and large artery, 1≤k≤76.

[0061] In this method, the objective function is constructed by using the physiological information of the root mean square error between the simulated blood pressure waveform and the measured blood pressure waveform, the systolic pressure difference, the diastolic pressure difference, and the dicrotic wave error to quantitatively evaluate the optimization results. The calculation formula of the constructed objective function is as shown in the above formulas (1)-(6).

[0062] Furthermore, in step 4, a multi-objective evolutionary algorithm is used to adjust the parameters of the zero-dimensional cardiovascular system simulation model to obtain the optimal parameters, which specifically includes the following steps:

[0063] S41: randomly generating initial values ​​of each decision variable in the decision space; according to the different parameter ranges of each component in the decision variable, using direct uniform distribution or logarithmic transformation random generation method respectively; wherein the direct uniform distribution random generation method is used for larger arteries such as the aorta, and the logarithmic random generation method is used for the parameters of the components whose parameter ranges span more than two orders of magnitude;

[0064] S42: randomly generate the initial value of each decision variable as an individual in the initial population, form a set, and form an initialized population;

[0065] S43: Analyze the type attributes of each decision variable; the type attributes specifically include convergence-related, diversity-related and sensitivity-related type attributes;

[0066] S44: Optimize the decision variables of convergence-related attributes, diversity-related attributes, and sensitivity-related attributes respectively;

[0067] S45: Based on the matching degree between the simulated waveform and the measured waveform, the best individual in the population is selected as a feasible solution to the parameter estimation problem in the cardiovascular in vitro simulated circulation system; specifically, a preset number of decision variables closest to the origin of the decision space are screened out through the Euclidean distance, and the decision variable with the best waveform matching degree is selected as the optimal solution. Specific embodiment 1:

[0069] This specific embodiment is used to illustrate the specific implementation process of step 4 of this application, and its process is as shown in the attached Figure 2 shown.

[0070] The initialization of step S42 is: randomly generate the initial value of each decision variable as an individual in the initial population, form a set, and form an initialized population. The population is denoted as pop n , its specific expression is as follows:

[0071] pop n ={x 1 ,x 2 ,…,x i ,…x m}

[0072] Wherein, n represents the evolutionary generation, and each evolutionary generation n=n+1, when the population is an initialization population, n=0; m represents the number of individuals contained in the population, and in this embodiment, m=50; x i represents the i-th individual in the population.

[0073] In step S43, the type attributes of each decision variable are analyzed, wherein the specific steps of the convergence-related and diversity-related type attribute analysis method include:

[0074] S431: Input initialization population pop 0 ,pop 0 Contains m individuals, each individual x i is a d-dimensional decision variable, set the number of random samplings t for each component of the decision variable, and the number of random selections k for individuals in the population. In this experiment, m = 50, t = 5, k = 10;

[0075] S432: From pop 0 Randomly select an individual x i , for the decision variable component x j i Randomly sample t times in the decision space to obtain t new individuals, which constitute the sampled population sampop; when this step is performed for the first time, j = 1;

[0076] S433: performing non-dominated sorting on sampop based on a fast non-dominated sorting algorithm, wherein the highest non-dominated rank after sorting is maxRankNum;

[0077] S434: If the number of randomly selected individuals is less than k, the number of randomly selected individuals is increased by 1, and the process returns to S432 to continue executing, and the maxRankNum calculated each time is added, recorded as sum(maxRankNum), otherwise, S435 is executed;

[0078] S435: If sum(maxRankNum)>k*t / 2, then x j is the convergence-related variable, otherwise x j are diversity-related variables;

[0079] S436: If j≤d, j++, return to step S432 to determine the next decision variable component, otherwise end the variable analysis and output all convergence-related variables and diversity-related variables.

[0080] The specific steps of the type attribute analysis method of decision variable sensitivity include:

[0081] S437: Input initialization population pop 0 The number of random sampling times t for each component in the decision variable and the number of random selection times k for individuals in the population are the same as those in the decision variable grouping step, and are also set to t = 5 and k = 10. Define count j , used to count the number of times the current component is judged as the sensitivity SV of the decision variable in k random draws, and count j Initialized to 0;

[0082] S438: From pop 0 Randomly select an individual x i , for the decision variable component x j i Randomly sample t times in the decision space to obtain t new individuals, which constitute the sampled population sampop; when this step is performed for the first time, j = 1;

[0083] S439: Calculate the sum of the target values ​​of each individual in sampop on all the target functions, and calculate the average of the sum of all the target values ​​of the individuals, which is recorded as mean(sum(loss)). i The sum of the objective values ​​of all objective functions is sum(loss i ) has changed by more than 20%, that is, Then count j =count j +1;

[0084] S4310: If the number of randomly selected individuals is less than k, the number of randomly selected individuals is increased by 1, and the process returns to S438 to continue, otherwise, S431 is executed;

[0085] S4311: If countj>k / 2, then x i is a sensitivity-related variable; otherwise x i Not a sensitivity-related variable.

[0086] At the same time, let countj=0;

[0087] S4312: If j≤d, j++, and return to S432 to continue, otherwise end the variable sensitivity analysis and output all sensitivity-related variables.

[0088] In the optimization process of step S44, the optimization steps of the variables related to convergence and diversity include:

[0089] S441: Input the current nth generation population pop n , and convergence-related variables;

[0090] S442: Based on the binary tournament method from pop n Select an individual x from i,n , extract x i,n The decision variable components that belong to the convergence-related variables in the above equation are used to generate new components using crossover and mutation operators. Then, the newly generated components are compared with x. i,n The components that do not participate in crossover and mutation are combined to generate new individuals v i , the new individual v i Add it to the new population newpop, and repeat the above steps until the number of individuals in newpop is the same as pop n The number of individuals in is equal;

[0091] S443: Based on the elite retention strategy, newpop and pop n Fusion, and non-dominated sorting, to obtain the non-dominated rank of each individual in the fused population;

[0092] S444: Score each individual based on the sorted non-dominated level and the target value. The scoring formula is as follows:

[0093] score = rankNum * sum (loss)

[0094] Where rankNum is the non-dominated rank, and sum(loss) represents the sum of the target values ​​of the individual on all objective functions. n The scores of individuals in the list are compared in turn. If the score of an individual in newpop is lower, pop is replaced. n After all individuals are compared and replaced, let n = n + 1;

[0095] S445: Repeat steps S442 to S444 until the preset number of iterations is reached, and output pop n .

[0096] The specific optimization steps for sensitivity-related variables are as follows:

[0097] S445: Input the current nth generation population pop n , and diversity-related variables, and construct an empty set p;

[0098] S446: Based on the binary tournament method from pop n Select an individual x from i,n , extract x i,n The decision variable components belonging to the diversity-related variables in the above equation are used to generate new components using SBX and PM operators, and then the newly generated components are compared with x. i,n The components that do not participate in crossover and mutation are combined to generate new individuals v i , the new individual v i Add it to the new population newpop, and repeat the above steps until the number of individuals in newpop is the same as pop n The number of individuals in is equal;

[0099] S447: Newpop and pop n Merge into a new population R and generate an m-row unit matrix I(m,m), where m is the number of targets of the optimization problem and each row of I(m,m) corresponds to a reference vector of a target.

[0100] S448: Calculate and normalize the target vector for each individual in R, based on the formula x i ′=x i +δ i ×(u i -l i )

[0101] (u i represents the upper bound of the i-th component of the decision variable, l i represents the lower bound of the i-th component of the decision variable, δ i =(x i -l i ) / (u i -l i ), calculate the angle between the target vector of all individuals in R and each reference vector, find the individual with the smallest angle with each reference vector, remove them from R and add them to p;

[0102] S449: Repeat step S448 to calculate the minimum angle between the target vectors of the remaining individuals in R and the target vectors of all individuals in p, find the individual with the largest angle, remove it from R and add it to p;

[0103] S4410: Repeat step S449 until the number of individuals in p is equal to that in pop nThe number of individuals is the same, let pop n =p, and output pop n . Specific embodiment 2:

[0105] In order to verify the effectiveness of the present method, in this embodiment, the blood pressure waveforms of four arteries, namely the radial artery, splenic artery, internal iliac artery and anterior tibial artery, are taken as targets at the same time, and modeling and simulation are performed under the four normal target conditions and the four femoral artery stenosis condition respectively.

[0106] As attached Figure 5 As shown in the figure, under normal conditions, the radial artery, splenic artery, internal iliac artery and anterior tibial artery modeling simulation and actual measurement comparison results are as follows: from the morphology of the blood pressure waveform, the blood pressure waveforms of the four arteries are well matched as a whole, and the differences between the simulated values ​​and the measured values ​​of systolic and diastolic pressure are relatively small.

[0107] As attached Figure 6 As shown in the figure, the modeling simulation and measured comparison results of the radial artery, splenic artery, internal iliac artery and anterior tibial artery in the sixth stenosis stage of the four target femoral arteries. In this case, the systolic pressure of the anterior tibial artery is around 72 mmHg, which is about 16 mmHg lower than that in normal conditions. The blood pressure waveform simulated by the model can well reflect this situation.

[0108] Although the present invention has been disclosed as above in terms of preferred embodiments, they are not intended to limit the present invention. Anyone skilled in the art can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection defined by the claims of this application.

Claims

1. A personalized cardiovascular system modeling method based on multi-objective optimization, characterized by: The following steps are involved: Step 1: Based on the anatomical structure of the human cardiovascular system, a zero-dimensional cardiovascular system simulation model including the main arteries, arterioles, capillaries and veins of the human body is constructed; Step 2: Building a cardiovascular in vitro simulated circulatory system capable of collecting hemodynamic data based on the zero-dimensional cardiovascular system simulation model; the cardiovascular simulated circulatory system simulates the ventricles, atria, arterial vascular trees, compliance devices, and heart valve structures that constitute the cardiovascular system, and can simulate the work of the cardiovascular system through the reciprocating motion of the servo motor and its control system; Step 3: Determine the objective function and decision variables of the multi-objective algorithm; The objective function is constructed by using the root mean square error, systolic pressure difference, diastolic pressure difference and dicrotic wave error physiological information between the simulated blood pressure waveform simulated by the zero-dimensional cardiovascular system simulation model and the measured blood pressure waveform collected by the cardiovascular in vitro simulation circulation system; the blood viscosity resistance, vascular compliance and blood flow inertia of the model are used as the decision variables of the model; Step 4: Use a multi-objective evolutionary algorithm to adjust the parameters of the zero-dimensional cardiovascular system simulation model to obtain the optimal parameters, and then obtain a personalized cardiovascular zero-dimensional model with optimal parameters.

2. The personalized cardiovascular system modeling method based on multi-objective optimization according to claim 1, characterized in that: In step one, the constructed zero-dimensional cardiovascular system simulation model contains 166 blood vessels, of which the blood vessels numbered 1-71 are the main medium and large arteries in various parts of the human body, all of which are bifurcated structures; the blood vessels numbered 72-73 are two connected venous structures, and the venous structure is simplified into a series structure of two non-bifurcated vascular segments; the blood vessels numbered 74-76 represent the pulmonary artery, pulmonary arterioles and capillaries and pulmonary veins, respectively, and do not contain bifurcated structures; the blood vessels numbered 77-166 represent arterioles and capillaries; in the in vitro simulated circulatory system constructed in step two, the arterioles and capillaries are connected to the ends of the medium and large arteries in a three-section series manner, and are uniformly aggregated and flow into the venous structure numbered 72.

3. The personalized cardiovascular system modeling method based on multi-objective optimization according to claim 2, characterized in that: In step 2, the cardiovascular in vitro simulation circulation system can simulate and collect different hemodynamic states under normal cardiovascular conditions and various stenosis states, specifically: first measure the blood pressure waveforms of the radial artery, splenic artery, internal iliac artery and anterior tibial artery under normal conditions through the cardiovascular simulation circulation system; then change the stenosis degree of the femoral artery of the cardiovascular simulation circulation system multiple times, and measure the blood pressure waveforms of the above four arteries; the various stenosis states are different degrees of stenosis of the femoral artery.

4. The personalized cardiovascular system modeling method based on multi-objective optimization according to claim 3, characterized in that: The objective function constructed in step 3 is as follows: loss1=RMSE (1) loss2=∆SBP+∆DBP (2) loss3=∆P1+∆P2 (3) loss4=ΔT1+ΔT2+ΔT3+ΔT4 (4) lossall1=a1*loss1+a2*loss2+a3*loss3+a4*loss4 (5) In the above formula, loss1 is the overall error RMSE between the measured blood pressure waveform collected by the cardiovascular in vitro simulation circulation system and the simulated blood pressure waveform simulated by the zero-dimensional cardiovascular system simulation model; loss2 is the systolic pressure ∆SBP and diastolic pressure error ∆DBP between the measured blood pressure waveform and the model simulated blood pressure waveform; loss3 is the error of the replay wave part of the measured blood pressure waveform and the model simulated blood pressure waveform, where ∆P1 is the difference between the blood pressures of the two at the time of the replay notch; ∆P2 is the difference between the maximum values ​​of the dicrotic blood pressure of the two; loss4 is the weighted sum of the errors at each moment, where ∆T1 is the difference between the corresponding moments of the systolic pressure of the two, ∆T2 is the difference between the corresponding moments of the diastolic pressure of the two, ∆T3 is the difference between the corresponding moments of the maximum values ​​of the dicrotic blood pressure of the two; ∆T4 is the difference between the corresponding moments of the replay notch; loss all1 is the weighted sum of the errors between the measured blood pressure waveform and the simulated blood pressure waveform, where a1, a2, a3 and a4 are loss all1 Weighting coefficients for different error terms; The specific decision variables are: the blood viscosity resistance, vascular compliance, and blood flow inertia of the 166 blood vessels in the zero-dimensional cardiovascular system simulation model constitute the decision variable X, as shown in the following formula: (6) In the above formula, Ri represents the vascular blood viscosity resistance of each medium and large artery and the small arteries and capillaries at the end of the model, where 1≤i≤166; Cj represents the vascular compliance of each medium and large artery, small artery and capillary, where 1≤j≤166; Lk represents the blood flow inertia of each medium and large artery, 1≤k≤76.

5. The personalized cardiovascular system modeling method based on multi-objective optimization according to claim 4, characterized in that: In step 4, a multi-objective evolutionary algorithm is used to adjust the parameters of the zero-dimensional cardiovascular system simulation model to obtain the optimal parameters, which specifically includes the following steps: S41: randomly generate initial values ​​of each decision variable in the decision space; according to the different parameter ranges of each component in the decision variable, use direct uniform distribution or logarithmic transformation random generation method respectively; the direct uniform distribution random generation method is used for larger arteries such as the aorta, and the logarithmic random generation method is used for the parameters of the components whose parameter ranges span more than two orders of magnitude; S42: randomly generate the initial value of each decision variable as an individual in the initial population, form a set, and form an initialized population; S43: Analyze the type attributes of each decision variable; the type attributes specifically include convergence-related, diversity-related and sensitivity-related type attributes; S44: Optimize the decision variables of convergence-related attributes, diversity-related attributes, and sensitivity-related attributes respectively; S45: Based on the matching degree between the simulated waveform and the measured waveform, the best individual in the population is selected as a feasible solution to the parameter estimation problem in the cardiovascular in vitro simulated circulation system; specifically, a preset number of decision variables closest to the origin of the decision space are screened out through the Euclidean distance, and the decision variable with the best waveform matching degree is selected as the optimal solution.