阅读说明：本技术 一种黏菌优化算法的心血管疾病的识别模型的建立方法 (Method for establishing cardiovascular disease recognition model by using slime mold optimization algorithm ) 是由 郭睿 颜建军 朱光耀 张春柯 王忆勤 燕海霞 武文杰 于 2021-09-08 设计创作，主要内容包括：本发明公开了一种黏菌优化算法的心血管疾病的识别模型的建立方法,采集腕部压力脉搏波和指端容积脉搏波；建立指端微循环的容积脉搏血流模型；将腕部压力脉搏波和指端容积脉搏波分别作为容积脉搏血流模型的实际输入和期望输出,然后运用黏菌优化算法降低模型参数的不确定性,估计得得到容积脉搏血流模型的最终参数,实现了微循环生理状态信息的提取；然后基于机器学习算法建立心血管疾病的识别模型,以此判断心血管的健康状态。本发明能够获得更高的准确率,取得了较好的识别效果。(The invention discloses a method for establishing a cardiovascular disease identification model by using a slime optimization algorithm, which comprises the steps of collecting wrist pressure pulse waves and finger tip volume pulse waves; establishing a volume pulse blood flow model of finger tip microcirculation; the wrist pressure pulse wave and the finger tip volume pulse wave are respectively used as the actual input and the expected output of the volume pulse blood flow model, then the uncertainty of the model parameters is reduced by applying a slime optimization algorithm, the final parameters of the volume pulse blood flow model are obtained by estimation, and the extraction of the microcirculation physiological state information is realized; then, a cardiovascular disease recognition model is established based on a machine learning algorithm, so that the health state of the cardiovascular disease is judged. The invention can obtain higher accuracy and better identification effect.)

1. A method for establishing a cardiovascular disease identification model of a slime optimization algorithm is characterized by comprising the following steps:

step 1, collecting wrist pressure pulse waves and finger tip volume pulse waves;

step 2, establishing a volume pulse blood flow model of finger tip microcirculation;

step 3, taking wrist pressure pulse waves and finger tip volume pulse waves as actual input and expected output of the volume pulse blood flow model respectively, reducing uncertainty of model parameters by using a slime optimization algorithm, and estimating to obtain final parameters of the volume pulse blood flow model, thereby realizing extraction of microcirculation physiological state information;

and 4, establishing a cardiovascular disease identification model based on a machine learning algorithm so as to judge the cardiovascular health state.

2. The method for building an identification model of cardiovascular diseases by using myxomycete optimization algorithm according to claim 1, wherein the volume pulse blood flow model of finger tip microcirculation is built by building an electric network model of cardiovascular simulation,

the inductance L represents the degree of difficulty of the change in the flow of blood in the microcirculation, and represents the inertia of the flow of blood in the arterioles; the capacitance C represents the rate of change of the total volume of blood vessels in the microcirculation with the blood pressure P, and represents the overall compliance of the microcirculation; resistance R represents the total resistance to which blood is subjected when flowing in circulation;

from this model, the following mathematical expression can be established:

the model equation obtained by the arrangement is as follows:

to obtain the parameter values of the model, the optimal R, L, C value is solved within the set parameter range by applying the slime optimization algorithm.

3. The method for building a cardiovascular disease identification model by using a slime optimization algorithm according to claim 2, wherein the method for solving the optimal R, L, C value in the set parameter range by using the slime optimization algorithm comprises the following specific steps of:

step 1, setting an initial value of a population;

step 2, calculating the current fitness value and sequencing;

step 3, updating the positions of the population;

step 4, calculating the fitness value again, and updating the optimal position in the population;

and step 5, judging the conditions for finishing the optimizing process, and if the conditions are not met, executing the step 2 to the step 5 again.

4. The method for building a cardiovascular disease recognition model by a slime optimization algorithm according to claim 3, wherein the formula for updating the location in step 3 is as follows:，

where the sum is the upper and lower bounds of the search range, respectively, t is the current iteration,in order to simulate the position with the highest concentration of the current odor in the process of foraging by the slime mold,in the range of，，For two randomly selected positions, W represents the weight of the position,the linear decrease from 1 to 0 is obtained,

in addition, the parametersThe expression of (a) is as follows:

in the formulaIs composed ofThe degree of fitness of (a) to (b),for the best fitness in all iterations,

parameter(s)The expression of (a) is as follows:

in the formula (I), the compound is shown in the specification,for the maximum number of iterations, the formula for W is as follows:

wherein r is [0, 1]]The random value of the (c) bit of the (c),、respectively the best fitness and the worst fitness in the current iteration process,is composed ofThe top 50% of the population in the ranking.

5. The method for building a cardiovascular disease recognition model based on a slime optimization algorithm according to claim 1, wherein the step 4 is to build the cardiovascular disease recognition model based on a machine learning algorithm by a Random Forest classification algorithm (RF), wherein the structure of the cardiovascular disease recognition model is substantially the same as that of a regression algorithm in a conventional Random Forest classification algorithm, and the difference between the two algorithms is that:

(1) the Gini impure degree function is used to measure when the nodes of the decision tree are cut,

such as formulaAs shown in the drawings, the above-described,

the probability of the occurrence of the value of the target variable in the current node training sample is obtained, and the impurity degree is lower when the type of the node sample is more definite;

(2) and determining a final classification result by using a voting method, and taking the mode of each decision tree as the final classification result according to the classification result of each decision tree.

Technical Field

The invention relates to the technical field of blood pressure prediction methods, in particular to a method for establishing a cardiovascular disease identification model of a slime mold optimization algorithm.

Background

In the cardiovascular system, microcirculation provides an important site for the exchange of substances between blood and tissues, and is the only route by which blood flows from arteries into veins. According to clinical research, the blood flow change of finger tip microcirculation (or nail fold microcirculation) can reflect the state of important components in the cardiovascular system such as heart, artery and the like, and the microcirculation state has obvious relation with cardiovascular diseases.

The microcirculation is a complex system, the microcirculation states of different human bodies have larger uncertainty, and the established model can not completely simulate the structure of the microcirculation, so that a plurality of local optimal solutions can be generated when model parameters are estimated, and the identification model has larger identification errors. Therefore, the method for establishing the cardiovascular disease identification model by the slime optimization algorithm is improved.

Disclosure of Invention

In order to solve the technical problems, the invention provides the following technical scheme:

the method for establishing the cardiovascular disease identification model by using the slime optimization algorithm comprises the following steps:

step 1, collecting wrist pressure pulse waves and finger tip volume pulse waves;

step 2, establishing a volume pulse blood flow model of finger tip microcirculation;

step 3, taking wrist pressure pulse waves and finger tip volume pulse waves as actual input and expected output of the volume pulse blood flow model respectively, reducing uncertainty of model parameters by using a slime optimization algorithm, and estimating to obtain final parameters of the volume pulse blood flow model, thereby realizing extraction of microcirculation physiological state information;

and 4, establishing a cardiovascular disease identification model based on a machine learning algorithm so as to judge the cardiovascular health state.

3. As a preferred technical scheme of the invention, the method for establishing the volume pulse blood flow model of the finger tip microcirculation is to establish an electric network model of cardiovascular simulation,

the inductance L represents the degree of difficulty of the change in the flow of blood in the microcirculation, and represents the inertia of the flow of blood in the arterioles; the capacitance C represents the rate of change of the total volume of blood vessels in the microcirculation with the blood pressure P, and represents the overall compliance of the microcirculation; resistance R represents the total resistance to which blood is subjected when flowing in circulation;

from this model, the following mathematical expression can be established:

the model equation obtained by the arrangement is as follows:

to obtain the parameter values of the model, the optimal R, L, C value is solved within the set parameter range by applying the slime optimization algorithm.

As a preferred technical solution of the present invention, the method for solving the optimal R, L, C value in the set parameter range by using the slime optimization algorithm comprises the following specific steps of:

step 1, setting an initial value of a population;

step 2, calculating the current fitness value and sequencing;

step 3, updating the positions of the population;

step 4, calculating the fitness value again, and updating the optimal position in the population;

and step 5, judging the conditions for finishing the optimizing process, and if the conditions are not met, executing the step 2 to the step 5 again.

As a preferred embodiment of the present invention, the formula for updating the position in step 3 is as follows:

where the sum is the upper and lower bounds of the search range, respectively, t is the current iteration,in order to simulate the position with the highest concentration of the current odor in the process of foraging by the slime mold,in the range of [ -a, a [)]， For two randomly selected positions, W represents the weight of the position,linearly decreasing from 1 to 0. Further, the expression of the parameter p is as follows:

p＝tanh|S(i)-DF|

wherein S (i) isDF is the best fitness in all iterations.The expression for parameter a is as follows:

in the formula, maxT is the maximum number of iterations, and the formula of W is as follows:

in the formula, r is a random value in [0, 1], bF and wF are respectively the best fitness and the worst fitness in the current iteration process, and condition is the top 50% of the population in the ranking of S (i).

As a preferred technical solution of the present invention, the method for establishing the cardiovascular disease identification model based on the machine learning algorithm in step 4 is to establish the cardiovascular disease identification model by a Random Forest classification algorithm (RF), wherein the structure of the cardiovascular disease identification model is substantially consistent with that of a traditional classification algorithm and that of a regression algorithm, and the difference between the two algorithms is as follows:

(1) the Gini impure degree function is used to measure when the nodes of the decision tree are cut,

such as formulaAs shown in the drawings, the above-described,

p_mkthe probability of the occurrence of the value of the target variable in the current node training sample is obtained, and the impurity degree is lower when the type of the node sample is more definite;

(2) and determining a final classification result by using a voting method, and taking the mode of each decision tree as the final classification result according to the classification result of each decision tree.

The invention has the beneficial effects that: the method for establishing the cardiovascular disease identification model by the slime mold optimization algorithm comprises the steps of firstly collecting wrist pressure pulse waves and finger tip volume pulse waves; establishing a volume pulse blood flow model of finger tip microcirculation; the wrist pressure pulse wave and the finger tip volume pulse wave are respectively used as the actual input and the expected output of the volume pulse blood flow model, then the uncertainty of the model parameters is reduced by applying a slime optimization algorithm, the final parameters of the volume pulse blood flow model are obtained by estimation, and the extraction of the microcirculation physiological state information is realized; then, a cardiovascular disease recognition model is established based on a machine learning algorithm, so that the health state of the cardiovascular disease is judged. The invention can obtain higher accuracy and better identification effect.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings;

FIG. 1 is a flow chart of the R, L, C parameter estimation flow chart of the present invention;

FIG. 2 is a flow chart of the cardiovascular disease identification model of the present invention;

FIG. 3 is a graph of a volumetric pulse blood flow model electrical network of the microcirculation of the present invention;

FIG. 4 is a diagram of a method for calibrating pulse waves;

FIG. 5 is a diagram of a calibration method of the pressure pulse wave;

FIG. 6 is Q after calibration_mGraph against CO;

FIG. 7 is a graph comparing the simulated volume pulse wave with the original volume pulse wave;

FIG. 8 is a graph of predicted results for a cardiovascular disease identification model;

FIG. 9 is a graph of the results of a cardiovascular disease classifier comparison experiment.

Detailed Description

The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.

Example (b): the invention relates to a method for establishing a cardiovascular disease identification model by using a slime optimization algorithm, which comprises the following steps:

step 1, collecting wrist pressure pulse waves and finger tip volume pulse waves;

step 2, establishing a volume pulse blood flow model of finger tip microcirculation;

step 3, taking wrist pressure pulse waves and finger tip volume pulse waves as actual input and expected output of the volume pulse blood flow model respectively, reducing uncertainty of model parameters by using a slime optimization algorithm, and estimating to obtain final parameters of the volume pulse blood flow model, thereby realizing extraction of microcirculation physiological state information;

and 4, establishing a cardiovascular disease identification model based on a machine learning algorithm so as to judge the cardiovascular health state.

The method for establishing the volume pulse blood flow model of finger tip microcirculation is to establish an electric network model of cardiovascular simulation, as shown in figure 3,

the inductance L represents the degree of difficulty of the change in the flow of blood in the microcirculation, and represents the inertia of the flow of blood in the arterioles; the capacitance C represents the rate of change of the total volume of blood vessels in the microcirculation with the blood pressure P, and represents the overall compliance of the microcirculation; resistance R represents the total resistance to which blood is subjected when flowing in circulation;

from this model, the following mathematical expression can be established:

the model equation obtained by the arrangement is as follows:

to obtain the parameter values of the model, the optimal R, L, C value is solved within the set parameter range by applying the slime optimization algorithm.

As shown in fig. 1, the method for solving the optimal R, L, C value in the set parameter range by applying the slime optimization algorithm includes the following specific steps of finding the position of R, L, C value:

step 1, setting an initial value of a population;

step 2, calculating the current fitness value and sequencing;

step 3, updating the positions of the population;

step 4, calculating the fitness value again, and updating the optimal position in the population;

and step 5, judging the conditions for finishing the optimizing process, and if the conditions are not met, executing the step 2 to the step 5 again.

The formula for updating the position in step 3 is as follows:

where the sum is the upper and lower bounds of the search range, respectively, t is the current iteration,in order to simulate the position with the highest concentration of the current odor in the process of foraging by the slime mold,in the range of [ -a, a [)]， For two randomly selected positions, W represents the weight of the position,linearly decreasing from 1 to 0. Further, the expression of the parameter p is as follows:

p＝tanh|S(i)-DF|

wherein S (i) isDF is the best fitness in all iterations. The expression for parameter a is as follows:

in the formula, maxT is the maximum number of iterations, and the formula of W is as follows:

in the formula, r is a random value in [0, 1], bF and wF are respectively the best fitness and the worst fitness in the current iteration process, and condition is the top 50% of the population in the ranking of S (i).

The method for establishing the cardiovascular disease identification model based on the machine learning algorithm in the step 4 is to establish the cardiovascular disease identification model through a Random Forest classification algorithm (RF), wherein the structure of the cardiovascular disease identification model is basically consistent with that of a classification algorithm and a regression algorithm of a traditional Random Forest classification algorithm, and the difference between the two algorithms is as follows:

(1) the Gini impure degree function is used to measure when the nodes of the decision tree are cut,

such as formulaAs shown in the drawings, the above-described,

p_mkthe probability of the occurrence of the value of the target variable in the current node training sample is obtained, and the impurity degree is lower when the type of the node sample is more definite;

(2) and determining a final classification result by using a voting method, and taking the mode of each decision tree as the final classification result according to the classification result of each decision tree.

The signals acquired by the pulse sensor can be used to approximate pulse signals representing the variation of parameters such as pressure and volume in blood vessels along with time, but the signals are represented by the variation of voltage quantity, and the amplitude of the signals does not have specific physiological significance. Therefore, these curves need to be calibrated before they can be used in the parameter estimation of the cardiovascular simulation model.

Fig. 4 and 5 show signals of a pressure pulse wave and a volume pulse wave, each of which includes a direct current component and a pulsating component. The signal of the pressure pulse wave comes from the variation of the blood pressure inside the arterial vessel, so the curve can be calibrated using the blood pressure values [46 ]. In the pulse period of the ventricle, the maximum pressure in the artery is the systolic pressure, the minimum pressure is the diastolic pressure, and the maximum pressure and the minimum pressure respectively correspond to the maximum value and the minimum value in the pulse wave, so that a calibration formula of the original single-period pressure pulse wave can be obtained:

in the formula, M_s、M_dThe blood ejected from left ventricle is flowed into microcirculation of human body through artery under the action of pulse wave pressure and blood perfusion, and can implement material exchange with histiocyte in microcirculation, and the volume pulse wave of finger end microcirculation can reflect the change of blood volume in the micro-pulse and capillary vessel along with the pulsation of heart, so that the average value Q of volume pulse wave can be obtained_mThe amount of blood output by the left ventricle, i.e., cardiac output, is normalized. Peripheral vascular resistance R (PRU), cardiac output CO (mL/s) and P according to hemodynamic definition_mThe following relationship exists between mean arterial pressure (mmHg):

in the formula, P_mCalculable, Q_m(mL/s) is the mean pulse blood flow, SV (mL/beat) is the stroke volume, and the pulse wave period, and the calculation formula is as follows

From this, it can be derived:

in the formula, K, K' represents waveform characteristics of the pressure pulse wave and the volume pulse wave.

Under the assumption of linearization, the direct current components and the pulsating components of the pressure pulse wave and the volume pulse wave in the above formula are respectively corresponding, so the maximum value and the minimum value of the volume pulse wave are calibrated as follows,

the calibration formula of the obtained volume pulse wave is consistent with the calibration method of the pressure pulse wave and is as follows

The cardiovascular disease identification model selects the samples of hypertension and coronary heart disease which are more common and have higher risk as target objects. As shown in FIG. 2, in the process of establishing the identification model, firstly, the features are extracted from the pulse wave samples, and then the estimated R, L, C model parameters, as well as the systolic pressure Ps, the diastolic pressure Pd, the average pressure Pm, the CO cardiac output, and Q are used_maxMaximum blood flow sum Q_minAnd establishing a characteristic set by the minimum blood flow, and finally establishing an identification model of health, hypertension and coronary heart disease based on a random forest algorithm.

In the experiment, the pulse wave is firstly calibrated, and the average blood flow calculated by the volume pulse wave after calibration is compared with the cardiac output theory calculated by the pressure pulse wave after calibration. As shown in FIG. 6, the two methods have good consistency, and the average absolute error is 11.10. + -. 8.02mL/s, which indicates that the calibration method has certain accuracy.

As shown in fig. 7, in order to verify the effect of performing the volumetric pulse wave blood flow model uncertainty quantification by using the myxomycete optimization algorithm, the difference between the simulated volumetric pulse wave output by the model after parameter optimization and the actually acquired original volumetric pulse wave is compared. It can be seen that the similarity of the two waveforms is high, which indicates that the slime optimization algorithm searches a better position in R, L, C parameter space, so that the estimated model parameter value can substantially represent the actual physiological state of the microcirculation. The RF, SVM, KNN and DT algorithms are widely applied to classification and identification of pulse waves, and in order to verify the classification performance of the RF, the algorithms are respectively used for establishing a cardiovascular disease identification model for comparative analysis. In the experiment, the performance of each classifier is tested by adopting a 3-fold cross inspection method, and the average accuracy, the average accuracy and the average recall rate of 3 experiments are calculated, and the result is shown in fig. 9. The average RF accuracy is improved by 2.8% compared with suboptimal KNN and reaches 91.96%; the average accuracy is improved by 3 percent and reaches 92.13 percent; the average recall rate is improved by 2.49 percent and reaches 92.05 percent. The results show that the RF has good classification performance in four classification algorithms and has better distinguishing capability on cardiovascular parameter characteristics with more cross overlapping in space, so that the RF algorithm is finally selected to establish a cardiovascular disease identification model.

As shown in fig. 8, the prediction result of the cardiovascular disease identification model established on the microcirculation state parameter feature set by using the RF algorithm shows that the identification accuracy of each type of sample reaches more than 88%. The identification accuracy rate of the coronary heart disease sample is 95.51% at the highest, the average identification accuracy rate of the healthy sample is 92.11%, and the average identification accuracy rate of the hypertension sample is 88.55%. The reason why the identification accuracy of the hypertension sample is low is that: the pulse wave and the parameters of the cardiovascular model of a healthy sample with partial normal high blood pressure may show similar characteristics, while patients with coronary heart disease usually have a history of hypertension and therefore may also show similar characteristics. On the whole, the established model has better distinguishing capability on the characteristic parameters under different cardiovascular health states, and can realize the effective identification of hypertension and coronary heart disease.

Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.