Carotid artery pulse wave velocity prediction method and equipment based on machine learning and medium

Abstract

The invention discloses a carotid artery pulse wave velocity prediction method and device based on machine learning and a medium, and the method comprises the steps: carrying out the time domain feature analysis and correlation analysis of an original pulse wave data set of each test sample, and obtaining a plurality of key time domain features of each carotid artery position of each test sample; obtaining a total regression coefficient of the key time domain features of each carotid artery position by adopting an LASSO regression model; then, performing permutation and combination on the carotid artery positions to obtain a plurality of key time domain features under each carotid artery position combination of each test sample; constructing a test sample feature set of each carotid artery position combination, training a machine learning model according to the test sample feature set, and optimizing to obtain a carotid artery pulse wave velocity prediction model and a target carotid artery position combination; and performing carotid artery pulse wave velocity prediction on a to-be-predicted sample by using the carotid artery pulse wave velocity prediction model to obtain a predicted pulse wave velocity. The detection efficiency and the detection accuracy of the carotid artery pulse wave velocity are improved.

Description

Technical Field

[0001] The present application relates to the field of pulse wave velocity measurement, and particularly to a carotid artery pulse wave velocity prediction method, device and medium based on machine learning. Background Art

[0002] Cardiovascular Disease (CVD) is one of the main causes of death globally, and the number of deaths caused by CVD accounts for 33% of the total global deaths. Atherosclerosis (AS) can increase the afterload of the left ventricle, reduce coronary artery perfusion, and trigger cardiovascular diseases such as coronary artery disease. It is an important marker of Cardiovascular Disease (CVD). Due to its anatomical structure and hemodynamic characteristics, the carotid artery is one of the earliest affected sites, and its intima-media thickness can provide important information about AS. Therefore, Carotid Atherosclerosis (CAS) is often regarded as a warning signal for CVD.

[0003] Pulse Wave Velocity (PWV) is the conduction velocity of the pressure wave generated by each heartbeat along the arterial wall. It is often considered the simplest, non-invasive, and reliable method for quantitatively evaluating arteriosclerosis. The value of PWV is inversely proportional to vascular elasticity. The larger the value, the greater the degree of vascular sclerosis, and vice versa, the better the vascular elasticity. Most traditional PWV measurements are global, such as Carotid-Femoral Pulse Wave Velocity (CF-PWV) and Brachial-Ankle Pulse Wave Velocity (BA-PWV). Since global PWV is easily affected by physiological and pathological factors and is not sensitive enough to slight arterial elasticity changes, its application in early detection of arterial stiffness changes is limited. Local PWV can not only accurately quantify the elasticity of a single artery but has also been proven to be a strong independent predictor of cardiovascular disease mortality. Therefore, local carotid PWV is of great significance in the clinical early detection of arterial stiffness and prevention of cardiovascular diseases. Currently, the methods for measuring local carotid PWV mainly include the Transit-Time (TT) method, the circular method, the impedance analysis method, etc. Among them, the TT method estimates the pulse wave conduction time by simultaneously acquiring the pulse waves at two or more positions on a section of blood vessel, and then obtains the carotid artery pulse wave velocity (local PWV) based on the time and distance relationship. This method only involves pulse wave detection and is widely used clinically. However, pulse wave detection introduces errors due to operation experience, respiration, reflected waves, etc., resulting in a complex detection process and the inability to accurately and automatically detect the carotid artery pulse wave velocity. Summary of the Invention

[0004] The purpose of this application is to provide a method, device and medium for predicting carotid artery pulse wave velocity based on machine learning, which can improve the detection efficiency and accuracy of carotid artery pulse wave velocity.

[0005] To achieve the above object, this application provides the following solutions:

[0006] In the first aspect, this application provides a method for predicting carotid artery pulse wave velocity based on machine learning. The method for predicting carotid artery pulse wave velocity based on machine learning includes: obtaining carotid artery data of multiple test samples; the carotid artery data includes: clinical parameters and a group of original pulse wave data; the clinical parameters include: age, heart rate, stroke volume and cardiac output; the group of original pulse wave data includes pulse wave data at multiple carotid artery positions; the pulse wave data is a blood vessel pressure change curve within one cardiac cycle; performing time-domain feature analysis and correlation analysis on the group of original pulse wave data of each test sample to obtain multiple key time-domain features at each carotid artery position of each test sample; obtaining the total regression coefficient of the key time-domain features at each carotid artery position by using the LASSO regression model according to the multiple key time-domain features at each carotid artery position of each test sample; obtaining multiple key time-domain features under each combination of carotid artery positions of each test sample by using the total regression coefficient of the key time-domain features at each carotid artery position in a permutation and combination manner for the carotid artery positions; constructing a test sample feature set for each combination of carotid artery positions by using the multiple key time-domain features under each combination of carotid artery positions of all test samples and the clinical parameters of all test samples; training a machine learning model respectively according to the test sample feature set of each combination of carotid artery positions, and obtaining a carotid artery pulse wave velocity prediction model and a target combination of carotid artery positions after optimization; the target combination of carotid artery positions is the combination of carotid artery positions corresponding to the carotid artery pulse wave velocity prediction model; obtaining multiple key time-domain features under the target combination of carotid artery positions of the sample to be predicted, and using the carotid artery pulse wave velocity prediction model to predict the carotid artery pulse wave velocity of the sample to be predicted to obtain the predicted pulse wave velocity.

[0007] In the second aspect, this application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the computer program to implement the above-mentioned method for predicting carotid artery pulse wave velocity based on machine learning.

[0008] In the third aspect, this application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the above-mentioned method for predicting carotid artery pulse wave velocity based on machine learning.

[0009] According to the specific embodiments provided by this application, the following technical effects are disclosed:

[0010] In this application, time-domain feature analysis and correlation analysis are performed on the original pulse wave data set of each test sample to obtain multiple key time-domain features at each carotid artery position of each test sample; the total regression coefficients of the key time-domain features at each carotid artery position are obtained by using the LASSO regression model according to the multiple key time-domain features at each carotid artery position of each test sample; based on the total regression coefficients of the key time-domain features at each carotid artery position, a permutation and combination method is used for the carotid artery positions to obtain multiple key time-domain features under each combination of carotid artery positions of each test sample. By performing feature extraction and feature screening on the pulse wave data, the accuracy of the time-domain features of the pulse wave data is improved. Using the multiple key time-domain features under each combination of carotid artery positions of all test samples and the clinical parameters of all test samples, a test sample feature set for each combination of carotid artery positions is constructed; machine learning models are trained respectively according to the test sample feature sets of each combination of carotid artery positions, and after optimization, a carotid artery pulse wave velocity prediction model and a target combination of carotid artery positions are obtained; multiple key time-domain features under the target combination of carotid artery positions of the sample to be predicted are obtained, and the carotid artery pulse wave velocity prediction model is used to predict the carotid artery pulse wave velocity of the sample to be predicted, and the predicted pulse wave velocity is obtained. By training machine learning models according to the test sample feature sets of each combination of carotid artery positions and optimizing among multiple trained machine learning models to obtain the best trained machine learning model as the carotid artery pulse wave velocity prediction model, the detection efficiency and detection accuracy of the carotid artery pulse wave velocity are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0012] Figure 1 It is a schematic flow chart of a method for predicting carotid artery pulse wave velocity based on machine learning provided by an embodiment of the present application.

[0013] Figure 2 It is a schematic diagram of the carotid artery position provided by an embodiment of the present application.

[0014] Figure 3 It is a schematic flow chart of constructing a test sample feature set for each combination of carotid artery positions provided by an embodiment of the present application.

[0015] Figure 4 It is a schematic flow chart of machine learning model training provided by an embodiment of the present application.

[0016] Figure 5 This is a schematic structural diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0017] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0018] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0019] Embodiment 1, as Figure 1 shown, this embodiment provides a carotid pulse wave velocity prediction method based on machine learning. The carotid pulse wave velocity prediction method based on machine learning includes the following steps.

[0020] S1. Obtain carotid data of multiple test samples; the carotid data includes: clinical parameters and an original pulse wave data group; the clinical parameters include: age (Age), heart rate (HR), stroke volume (SV), and cardiac output (CO); the original pulse wave data group includes pulse wave data at multiple carotid positions; the pulse wave data is a blood vessel pressure change curve within one cardiac cycle. The carotid positions are as Figure 2 shown.

[0021] Further, the method for obtaining the carotid data of multiple test samples is to obtain it through simulation of a pulse wave database.

[0022] Optionally, the number of carotid positions is 13.

[0023] In the actual application process, the acquisition process of carotid artery data of multiple test samples is as follows: First, a method designed and verified by Peter Charlton et al. for simulating the pulse waves of healthy adults of different ages was used to create a pulse wave database containing 4,374 virtual subjects for computer evaluation of hemodynamic and pulse wave metrics. This method was used to simulate the pulse waves locally propagated at different positions of the carotid artery under different human hemodynamic parameter conditions. First, the hemodynamic parameters of healthy adults aged 25 to 75 were determined, and a one-dimensional model (1-D) was used to simulate the pulse waves of different arteries during the aging process of healthy adults with hemodynamic parameters at different ages. Finally, multiple positions were segmented from the left common carotid artery of the human body. Since the scanning probe can detect the pulse wave signals at 13 positions on the carotid artery simultaneously when clinically collecting carotid artery pulse waves. For the sake of consistency, in this embodiment, the left common carotid artery segment was divided into 13 positions, and finally, the pulse waves at 13 positions on 4,374 carotid arteries were successfully simulated.

[0024] S2. As Figure 3 shown, perform time-domain feature analysis and correlation analysis on the original pulse wave data group of each test sample to obtain multiple key time-domain features of each carotid artery position of each test sample, which specifically includes the following steps.

[0025] S21. Perform time-domain feature analysis on the original pulse wave data group of each test sample to obtain multiple time-domain features of each carotid artery position of each test sample.

[0026] S22. Perform correlation analysis on the multiple time-domain features of each carotid artery position of each test sample with the theoretical pulse wave velocity to obtain the correlation coefficient of each time-domain feature of each carotid artery position.

[0027] S23. Screen out the time-domain features corresponding to the average absolute value of the correlation coefficient greater than the preset value among the correlation coefficients of all time-domain features of all carotid artery positions as the key time-domain features, and obtain multiple key time-domain features of each carotid artery position of each test sample.

[0028] Optionally, the method of correlation analysis is the Pearson correlation coefficient method.

[0029] Optionally, the preset value is 0.2.

[0030] In the actual application process, the change of pulse wave pressure is an important basis for evaluating the cardiovascular physiological state of the human body, and the most significant change during propagation is the change of characteristic points on the pulse wave. Time-domain analysis is to analyze the dynamic characteristics (time-domain features) of the pulse wave waveform information in the time direction.

[0031] S3. As Figure 3As shown in the figure, the total regression coefficients of the key time-domain features of each carotid artery position are obtained by using the LASSO regression model based on the multiple key time-domain features of each carotid artery position of each test sample, and the specific steps are as follows.

[0032] S31. Normalize each key time-domain feature of each carotid artery position of each test sample to obtain the normalized key time-domain feature of each carotid artery position of each test sample.

[0033] S32. Based on all the key time-domain features of all carotid artery positions of all test samples after normalization, and the theoretical pulse wave velocity, use the LASSO regression model to calculate the absolute value of the regression coefficient of each key time-domain feature of each carotid artery position.

[0034] S33. Sum the absolute values of the regression coefficients of each key time-domain feature of each carotid artery position to obtain the total regression coefficient of the key time-domain feature of each carotid artery position.

[0035] In the actual application process, the LASSO regression model has the characteristic of reducing the regression coefficient of the key time-domain feature to zero, and the absolute value of the regression coefficient can reflect the influence degree of the key time-domain feature on the theoretical pulse wave velocity.

[0036] S4. As Figure 3 As shown in the figure, based on the total regression coefficient of the key time-domain feature of each carotid artery position, use the permutation and combination method for the carotid artery positions to obtain multiple key time-domain features under each combination of carotid artery positions of each test sample, and the specific steps are as follows.

[0037] S41. Sort the carotid artery positions in descending order according to the total regression coefficient of the key time-domain feature to obtain the sorted carotid artery positions.

[0038] In the actual application process, by sorting multiple carotid artery positions, the magnitude relationship of the contribution degrees of different carotid artery positions to the theoretical pulse wave velocity can be determined.

[0039] S42. Use the cumulative addition method to perform permutation and combination on the sorted carotid artery positions to obtain multiple combinations of carotid artery positions.

[0040] Step S42 specifically includes: taking the first carotid artery position in the sorted carotid artery positions as a combination of carotid artery positions, and successively adding one carotid artery position backward each time as a new combination of carotid artery positions until all carotid artery positions are added up to obtain multiple combinations of carotid artery positions; the number of combinations of carotid artery positions is the same as the number of carotid artery positions.

[0041] S43. Select multiple key time-domain features for each combination of carotid artery positions of each test sample according to the corresponding combination of carotid artery positions.

[0042] S5. As Figure 3 shown, use the multiple key time-domain features for each combination of carotid artery positions of all test samples and the clinical parameters of all test samples to construct a test sample feature set for each combination of carotid artery positions.

[0043] S6. As Figure 4 Train machine learning models respectively according to the test sample feature sets for each combination of carotid artery positions, and obtain the carotid pulse wave velocity prediction model and the target carotid artery position combination after optimization; the target carotid artery position combination is the carotid artery position combination corresponding to the carotid pulse wave velocity prediction model; the machine learning model includes at least any one of a multiple linear regression model, a Bayesian ridge regression model, a k-nearest neighbor regression model, a support vector regression model, and a convolutional neural network.

[0044] Step S6 specifically includes the following steps.

[0045] S61. Divide the key time-domain test sample feature set for each combination of carotid artery positions into a training set, a validation set, and a test set.

[0046] S62. Use the training set to train the machine learning model with the pulse wave velocity as the label and the minimum loss function as the goal.

[0047] S63. Use the validation set to select the optimal parameters of the machine learning model to obtain the trained machine learning model.

[0048] S64. Use the test set to evaluate the performance of each trained machine learning model to obtain the normalized root mean square error of the trained machine learning model.

[0049] S65. Obtain the normalized root mean square error of the trained machine learning model for each combination of carotid artery positions in the above manner (steps S61 - S64).

[0050] S65. Select the trained machine learning model with the smallest root mean square error among the multiple trained machine learning models as the carotid pulse wave velocity prediction model, and obtain the carotid artery position combination corresponding to the carotid pulse wave velocity prediction model as the target carotid artery position combination.

[0051] In the actual application process, this embodiment first obtains 13 feature combination sets (the key time-domain test sample feature set of carotid artery position combination). For each set of feature combination sets, they are successively divided into a training set, a validation set, and a test set. According to the following evaluation indicators, the model parameters are adjusted, and at the same time, the estimation accuracies under different feature combination sets are compared to find the machine learning model for estimating the local PWV (carotid artery pulse wave velocity) under the optimal position feature combination.

[0052] This embodiment introduces five traditional machine learning models, specifically including:

[0053] 1) Multiple Linear Regression (MLR) is a method in statistics used to analyze the relationship between multiple independent variables and a dependent variable and fit the relationship between them with a linear equation.

[0054] 2) Bayesian linear regression (BRR) is a regression method based on Bayesian inference, which combines the ideas of ridge regression and Bayesian statistics.

[0055] 3) K-Nearest Neighbor (KNN) is a non-parametric regression method that does not require assumptions about the data. Its core lies in how to measure the similarity between different data points.

[0056] 4) Support Vector Regression (SVR) is a non-parametric supervised machine learning algorithm.

[0057] 5) Convolutional Neural Networks (CNN) is a deep learning model. Its core idea is to extract features through operations such as convolution and pooling, map the input data to a high-dimensional feature space, and then classify or regress the features through a fully connected layer. This embodiment designs a one-dimensional convolutional neural network (CNN1DModel) to handle the estimation problem of local PWV.

[0058] The training and testing of each model both use the root mean square error (normalized root mean square error) (NRMSE) or the coefficient of determination (R 2 ) as the evaluation indicator, and the specific formula is as follows.

[0059]

[0060] In the formula, pre_lpwv i represents the carotid artery pulse wave velocity estimated for the i-th subject based on the time-domain features of the pulse wave, and the_lpwv iIt represents the theoretical carotid artery pulse wave velocity calculated based on carotid artery parameters for the i-th subject, and n represents the number of subjects. h is the wall thickness, R is the average radius of the entire carotid artery, and k 1 、k 2 and k 3 are empirical constants. k 1 is usually set to 3×106 g·s -2 .cm -1 , k 2 is set to -13.5×106 g·s -2 .cm -1 , k 3 is set according to the age and average blood pressure of each subject.

[0061] S7. Obtain multiple key time-domain features under the target carotid artery position combination of the sample to be predicted, and use the carotid artery pulse wave velocity prediction model to predict the carotid artery pulse wave velocity of the sample to be predicted, obtaining the predicted pulse wave velocity.

[0062] The technical effects of this application are as follows.

[0063] This application proposes a method for predicting carotid artery pulse wave velocity based on machine learning. First, establish a simulation model of local carotid artery pulse wave propagation with different hemodynamic parameters as the research data set (carotid artery data of multiple test samples); then, use the pulse wave feature points and hemodynamic parameters at different positions (multiple key time-domain features at each carotid artery position of each test sample) as the feature set, and use the correlation coefficient method to screen out the key features at each position, and then use LASSO regression to analyze the total contribution degree of the features at each position (the total regression coefficient of the key time-domain features at each carotid artery position). In the order from high to low contribution degree, gradually accumulate and select the features and hemodynamic parameters at each position as the model input (the feature set of test samples for each carotid artery position combination). Finally, use multiple linear regression, Bayesian ridge regression, k-nearest neighbor regression, support vector regression, and convolutional neural network algorithms for training and testing, and use the normalized root mean square error (NRMSE) or coefficient of determination (R 2 ) as the evaluation index to evaluate the performance of each model in detecting local PWV. This application has successfully achieved high-precision automatic estimation of local pulse wave velocity, simplified the detection process of carotid artery pulse wave velocity, improved the detection efficiency and accuracy of carotid artery pulse wave velocity, and can provide a theoretical basis and technical support for early quantitative evaluation of local carotid artery vascular elasticity in clinical practice, which is of great significance for the prevention and early screening of clinical cardiovascular diseases.

[0064] Example 2, this application also provides a computer device, which can be a server or a terminal, and its internal structure diagram can be as Figure 5As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store processed data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements a method for predicting carotid pulse wave velocity based on machine learning.

[0065] Those skilled in the art can understand that Figure 5 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0066] Embodiment 3, the present application also provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the above-mentioned various methods are implemented.

[0067] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memories can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0068] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0069] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for predicting carotid artery pulse wave velocity based on machine learning, characterized in that: The carotid artery pulse wave velocity prediction method based on machine learning includes: Acquire carotid artery data of multiple test samples; the carotid artery data includes: clinical parameters and original pulse wave data group; the clinical parameters include: age, heart rate, stroke volume and cardiac output; the original pulse wave data group includes pulse wave data of multiple carotid artery positions; the pulse wave data is a vascular pressure change curve within a cardiac cycle; Performing time domain feature analysis and correlation analysis on the original pulse wave data set of each test sample to obtain multiple key time domain features of each carotid artery position of each test sample; According to multiple key time domain features of each carotid artery position of each test sample, a LASSO regression model is used to obtain the total regression coefficient of the key time domain features of each carotid artery position; Based on the total regression coefficient of the key time domain features of each carotid artery position, the carotid artery positions are permuted and combined to obtain multiple key time domain features of each carotid artery position combination of each test sample; A test sample feature set for each carotid artery position combination is constructed using multiple key time domain features of all test samples under each carotid artery position combination and clinical parameters of all test samples; The machine learning model is trained according to the test sample feature set of each carotid artery position combination, and the carotid artery pulse wave velocity prediction model and the target carotid artery position combination are obtained after optimization; the target carotid artery position combination is the carotid artery position combination corresponding to the carotid artery pulse wave velocity prediction model; A plurality of key time domain features under the target carotid artery position combination of the sample to be predicted are obtained, and the carotid artery pulse wave velocity prediction model is used to predict the carotid artery pulse wave velocity of the sample to be predicted to obtain the predicted pulse wave velocity.

2. The method for predicting carotid artery pulse wave velocity based on machine learning according to claim 1, characterized in that: The original pulse wave data set of each test sample is subjected to time domain feature analysis and correlation analysis to obtain multiple key time domain features of each carotid artery position of each test sample, including: Performing time domain feature analysis on the original pulse wave data group of each test sample to obtain multiple time domain features of each carotid artery position of each test sample; A correlation analysis is performed between multiple time domain features of each carotid artery position of each test sample and the theoretical pulse wave velocity to obtain a correlation coefficient of each time domain feature of each carotid artery position; The time domain features corresponding to the time domain features whose average absolute values ​​of the correlation coefficients are greater than the preset values ​​are selected from the correlation coefficients of all the time domain features at all the carotid artery positions as the key time domain features, and multiple key time domain features are obtained for each carotid artery position of each test sample.

3. The method for predicting carotid artery pulse wave velocity based on machine learning according to claim 1, characterized in that: According to multiple key time domain features of each carotid artery position of each test sample, the LASSO regression model is used to obtain the total regression coefficient of the key time domain features of each carotid artery position, including: Normalizing each key time domain feature of each carotid artery position of each test sample to obtain each key time domain feature of each carotid artery position of each test sample after normalization; Based on all the key time domain features of all the carotid artery positions of all the test samples after normalization, and the theoretical pulse wave velocity, the absolute value of the regression coefficient of each key time domain feature of each carotid artery position is calculated using the LASSO regression model; The absolute values ​​of the regression coefficients of each key time domain feature at each carotid artery position are summed to obtain the total regression coefficient of the key time domain feature at each carotid artery position.

4. The method for predicting carotid artery pulse wave velocity based on machine learning according to claim 1, characterized in that: Based on the total regression coefficient of the key time domain features of each carotid artery position, the carotid artery positions are permuted and combined to obtain multiple key time domain features of each carotid artery position combination of each test sample, including: The carotid artery positions are sorted from large to small according to the total regression coefficient of the key time domain features to obtain the sorted carotid artery positions; The sorted carotid artery positions are arranged and combined using the accumulation method to obtain multiple carotid artery position combinations; According to each carotid artery position combination, multiple key time domain features under each carotid artery position combination of each test sample are selected.

5. The method for predicting carotid artery pulse wave velocity based on machine learning according to claim 4, characterized in that: The sorted carotid artery positions are arranged and combined using the accumulation method to obtain a variety of carotid artery position combinations, including: The first carotid artery position among the sorted carotid artery positions is taken as a carotid artery position combination, and is accumulated backward one by one, each time a carotid artery position is accumulated as a new carotid artery position combination, until all carotid artery positions are accumulated, thereby obtaining a plurality of carotid artery position combinations; the number of the carotid artery position combinations is the same as the number of carotid artery positions.

6. The method for predicting carotid artery pulse wave velocity based on machine learning according to claim 1, characterized in that: The machine learning model is trained according to the test sample feature set of each carotid artery position combination, and the carotid artery pulse wave velocity prediction model and the target carotid artery position combination are obtained after optimization, including: The key time-domain test sample feature set for each carotid artery position combination is divided into a training set, a validation set, and a test set; The pulse wave velocity is used as a label and the loss function is minimized as the goal. The training set is used to train the machine learning model. Use the validation set to select the optimal parameters of the machine learning model and obtain a trained machine learning model; Use the test set to evaluate the model performance of each trained machine learning model and obtain the normalized root mean square error of the trained machine learning model; Obtain the normalized root mean square error of the trained machine learning model for each carotid artery position combination in the above manner; A trained machine learning model with the smallest root mean square error is selected from multiple trained machine learning models as a carotid artery pulse wave velocity prediction model, and a carotid artery position combination corresponding to the carotid artery pulse wave velocity prediction model is obtained as a target carotid artery position combination.

7. The method for predicting carotid artery pulse wave velocity based on machine learning according to claim 1, characterized in that: The method for acquiring the carotid artery data of the plurality of test samples is to obtain the data through pulse wave database simulation.

8. The method for predicting carotid artery pulse wave velocity based on machine learning according to claim 1, characterized in that: The machine learning model includes at least: any one of a multivariate linear regression model, a Bayesian ridge regression model, a nearest neighbor regression model, a support vector regression model and a convolutional neural network.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the carotid artery pulse wave velocity prediction method based on machine learning as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting carotid artery pulse wave velocity based on machine learning described in any one of claims 1 to 8 is implemented.

Images