A method and electronic device for determining carotid-femoral artery pulse wave velocity

Abstract

The embodiment of the present application provides a method and electronic device for determining the carotid-femoral pulse wave velocity, including: obtaining a user feature and a second feature, then determining the distance between the carotid artery and the femoral artery of the user to be measured based on the user feature, and determining the pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured based on the second feature, and then determining the carotid-femoral pulse wave velocity value of the user to be measured according to the distance and the pulse wave propagation time. Among them, the user feature is obtained based on the user information of the user to be measured, the second feature represents the feature associated with the pulse wave propagation time, the second feature is obtained based on the first feature and the user feature, and the first feature is obtained based on the photoelectric volumetric pulse wave PPG signal, and the PPG signal is collected from the user to be measured. In this way, the detection of arteriosclerosis is achieved, and the accuracy of the carotid-femoral pulse wave velocity value can be improved, thereby improving the accuracy of the arteriosclerosis detection result.

Description

Technical Field

[0001] The present application relates to the technical field of carotid-femoral pulse wave velocity detection, and in particular to a method for determining carotid-femoral pulse wave velocity and an electronic device. Background Art

[0002] Carotid-femoral Pulse Wave Velocity (cfPWV) is one of the key parameters in the study of vascular physiology and cardiovascular disease. cfPWV can reflect the degree of arteriosclerosis, thereby helping to assess patients' cardiovascular health status and predict the risk of future cardiovascular disease.

[0003] In the related art, a measuring device (such as an arteriosclerosis detector, etc.) can be used to measure the pulse wave propagation time difference between the carotid artery and the femoral artery, and then the cfPWV value can be determined based on the pulse wave propagation time difference. However, on the one hand, the price of the measuring device is relatively high, which increases the measurement cost of cfPWV; on the other hand, the measuring device is difficult to carry, which increases the difficulty of detection. Summary of the invention

[0004] The method and electronic device for determining the carotid-femoral pulse wave velocity provided in the present application solve the problems of high detection cost and difficulty of cfPWV.

[0005] In order to achieve the above objectives, this application adopts the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a method for determining a carotid-femoral pulse wave velocity, comprising: obtaining a user feature and a second feature, then determining the distance between the carotid artery and the femoral artery of the user to be measured based on the user feature, and determining the pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured based on the second feature, and then determining the carotid-femoral pulse wave velocity cfPWV value of the user to be measured according to the distance and the pulse wave propagation time. The user feature is obtained based on the user information of the user to be measured, the second feature represents a feature associated with the pulse wave propagation time, the second feature is obtained based on the first feature and the user feature, and the first feature is obtained based on a photoplethysmogram (PPG) signal, and the PPG signal is collected from the user to be measured.

[0007] In this way, by determining the distance between the carotid artery and the femoral artery of the user to be measured through the user characteristics obtained based on the user information, and determining the pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured through the second feature obtained based on the PPG signal, the cfPWV value of the user to be measured can be determined according to the distance and the pulse wave propagation time to reflect the degree of arteriosclerosis. The present application can use a device with a PPG signal acquisition function to realize the acquisition of PPG signals, such as watches, bracelets, etc., which reduces hardware costs and is easy to carry, and reduces the difficulty of detection; at the same time, by respectively determining the distance and the pulse wave propagation time, and then determining the cfPWV value, it is different from the related art that directly determines the cfPWV value, and the accuracy of the cfPWV value can be improved.

[0008] In a possible implementation, the second feature is obtained in the following manner: determining the first feature corresponding to the PPG signal, the first feature including time domain features, frequency domain features and time-frequency domain features, and then determining the second feature based on the time domain features, frequency domain features, time-frequency domain features and user features. It should be understood that since the PPG signal can reflect the flow of blood in the blood vessels of the user to be measured, and the time domain features, frequency domain features and time-frequency domain features can reflect the changes in the PPG signal, processing the PPG signal into time domain features, frequency domain features and time-frequency domain features is conducive to determining the pulse wave propagation time and improving the accuracy of the pulse wave propagation time.

[0009] In a possible implementation, the second feature is determined based on time domain features, frequency domain features, time-frequency domain features and user features, including: determining the correlation value between each feature in the time domain features and the pulse wave propagation time, determining the correlation value between each feature in the frequency domain features and the pulse wave propagation time, determining the correlation value between each feature in the time-frequency domain features and the pulse wave propagation time, and determining the correlation value between each feature in the user features and the pulse wave propagation time, to obtain n correlation values, and then sorting the n correlation values ​​in descending order to obtain sorted n correlation values, where n is an integer greater than or equal to 1, and taking the features corresponding to the first m correlation values ​​among the sorted n correlation values ​​as the second feature, where m is a positive integer less than n.

[0010] In order to ensure the accuracy of the pulse wave propagation time, it is necessary to filter the features with a smaller correlation degree, so as to ensure the accuracy of the pulse wave propagation time determined based on the second feature.

[0011] It should be noted that if the first feature includes time domain features, frequency domain features, time-frequency domain features and user features, then the time domain features, frequency domain features, time-frequency domain features and user features each include at least one feature, so n is an integer greater than or equal to 1; if the first feature includes time domain features and user features, then n is an integer greater than or equal to 2. It should be understood that the minimum value of n is related to the number of feature types included in the first feature, and the feature type means the attribute corresponding to the feature, such as time domain attributes, frequency domain attributes, time-frequency domain attributes, user attributes, etc.

[0012] In a possible implementation, the second feature includes k features, k is a positive integer, and after obtaining the second feature, it also includes: determining the contribution of each feature in the second feature to the time prediction model to obtain k contributions, sorting the k contributions in order from large to small, and taking the features corresponding to the first q contributions of the sorted k contributions as the third feature, and further determining the pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured according to the third feature. Among them, the time prediction model is used to predict the pulse wave propagation time, and q is a positive integer less than or equal to k. It should be understood that in the prediction process of the time prediction model, not all second features contribute equally to the prediction of the pulse wave propagation time. Redundant and irrelevant features will increase the complexity of the model, may lead to overfitting, and reduce the generalization ability of the model. Therefore, it is necessary to perform feature selection on the second feature, which can remove these unnecessary features, simplify the time prediction model, and improve the robustness, accuracy and generalization of the time prediction model, as well as the accuracy of the pulse wave propagation time.

[0013] In a possible implementation, determining the time domain features, frequency domain features, and time-frequency domain features corresponding to the PPG signal includes: filtering the PPG signal to obtain a processed PPG signal, and extracting features from the processed PPG signal to obtain time domain features, frequency domain features, and time-frequency domain features.

[0014] Among them, by filtering and then extracting features from the PPG signal, the PPG signal can be made smoother, the noise can be reduced, and the extracted features can be made more reliable.

[0015] In a possible implementation, determining the distance between the carotid artery and the femoral artery of the user to be measured based on the user characteristics includes: determining the distance between the carotid artery and the femoral artery of the user to be measured based on the user characteristics through a distance prediction model; the distance prediction model is obtained based on training of a linear fitting model.

[0016] In one possible implementation, determining the pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured based on the second feature includes: determining the pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured based on the second feature through a time prediction model; the time prediction model is obtained by training a linear fitting model.

[0017] In a second aspect, an embodiment of the present application provides an electronic device, including: a processor and a memory;

[0018] Wherein, one or more computer programs are stored in the memory, and the one or more computer programs include instructions; when the instructions are executed by the processor, the electronic device executes the carotid-femoral pulse wave velocity determination method of the first aspect.

[0019] In a third aspect, an embodiment of the present application provides a computer storage medium, including computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the method for determining the carotid-femoral pulse wave velocity as in the first aspect.

[0020] In a fourth aspect, an embodiment of the present application provides a computer program product. When the computer program product is run on a computer, the computer executes the method for determining the carotid-femoral pulse wave velocity as in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic flow chart of a method for determining carotid-femoral pulse wave velocity provided in an embodiment of the present application;

[0022] Figure 2 A schematic diagram of an interface for measuring cfPWV of a wearable device provided in an embodiment of the present application;

[0023] Figure 3 A schematic diagram of a pulse wave waveform diagram corresponding to a PPG signal provided in an embodiment of the present application;

[0024] Figure 4 A schematic diagram of a carotid-femoral pulse wave velocity determination process provided in an embodiment of the present application;

[0025] Figure 5 A schematic diagram of an interactive process of a method for determining carotid-femoral pulse wave velocity provided in an embodiment of the present application;

[0026] Figure 6 A schematic diagram of the distribution of cfPWV values ​​obtained by a different determination method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to facilitate understanding of the contents of the following embodiments, the nouns or terms that may appear in the following embodiments are first explained:

[0028] The carotid artery is an artery in the human neck. In the human body, the carotid artery is divided into the left carotid artery and the right carotid artery. These two arteries are located in the front of the left and right sides of the neck and are large blood vessels. The carotid artery auscultation position is the carotid bifurcation auscultation area, which is located at the level of the thyroid cartilage or the level of the fourth cervical vertebra, that is, the lower back of the head, near the supraclavicular fossa.

[0029] The femoral artery is the main trunk of the lower limb artery. It is a direct continuation of the external iliac artery. The femoral artery is located in the lower limb. It extends downward from the back of the midpoint of the inguinal ligament, passes through the femoral triangle area, and continues to extend downward, eventually becoming the popliteal artery. The femoral artery is located deep to the midpoint of the inguinal ligament and enters the femoral triangle area. In the femoral triangle, the femoral artery is first located on the lateral side of the femoral vein, and then gradually moves from the outside to the front of the femoral vein. Because the femoral artery is superficial, its pulsation can be felt by hand, especially at the midpoint of the groin. Therefore, the auscultation position of the femoral artery is at the midpoint of the groin.

[0030] The pulse wave is formed by the heart's pulsation (vibration) propagating along the arteries and blood flow to the periphery. The pulse wave is caused by the blood being pushed into the arterial system when the heart contracts, resulting in periodic changes in the expansion and contraction of the arterial wall.

[0031] cfPWV (carotid-femoral pulse wave velocity) is an important measure of arterial stiffness and is calculated by measuring the time it takes for the pulse wave to travel from the carotid artery to the femoral artery (i.e., pulse wave propagation time, or pulse wave propagation time difference, PWTT) and dividing it by the actual distance between these two points. Higher cfPWV values ​​generally indicate harder or less elastic arterial walls and are an important predictor of cardiovascular disease risk.

[0032] ECG (electrocardiogram) refers to the detection of signals generated by the heart's electrical activity through electrodes.

[0033] PPG (Photoplethysmography) refers to the use of a photoelectric sensor to measure changes in light absorption caused by blood flowing in blood vessels, thereby reflecting the pulse wave.

[0034] The main wave height refers to the vertical distance of the highest point in the pulse wave waveform, which usually represents the peak value of the arterial pressure wave. The main wave height is closely related to the systolic pressure of arterial blood pressure and is one of the important indicators for evaluating arterial pressure.

[0035] The peak time refers to the time interval from the start of the pulse wave to the highest point of the waveform. The peak time reflects the speed of arterial pressure wave propagation and the elasticity of the arterial wall. When the elasticity of the arterial wall decreases, the peak time may be prolonged.

[0036] The cardiac cycle refers to the time required for the heart to complete one contraction and relaxation, including the systolic period and the diastolic period. The cardiac cycle is one of the basic parameters for evaluating cardiac function and is inversely proportional to the heart rate. When the heart rate increases, the cardiac cycle shortens; when the heart rate slows down, the cardiac cycle lengthens.

[0037] Normalized time to peak is the ratio of the time to peak to the height of the main wave. Normalized time to peak is used to assess the elasticity of the arterial wall and the resistance to blood flow. The smaller the ratio, the more elastic the arterial wall and the lower the resistance to blood flow.

[0038] Ejection time refers to the time interval from the start of left ventricle ejection to the end of ejection, which usually corresponds to the QRS complex (ventricular depolarization complex) in the electrocardiogram. Ejection time is one of the important indicators for evaluating cardiac contractile function. Prolonged ejection time may indicate weakened cardiac contractile function.

[0039] The reflection index refers to the ratio of the height of the reflected wave to the height of the main wave in the pulse waveform. The reflection index reflects the degree of reflection of the arterial wave and the elasticity of the arterial wall. An increase in the reflection index may indicate a decrease in the elasticity of the arterial wall and an increase in the reflected wave.

[0040] The stiffness index refers to a parameter related to arteriosclerosis and is used to assess the degree of arteriosclerosis.

[0041] SEVR (Subendocardial Viability Ratio) refers to the ratio of the right area to the left area of ​​the Dicrotic Notch in the pulse waveform. SEVR is used to evaluate the enhancement of arterial reflection waves and is an important indicator for evaluating arterial function and sclerosis. An increase in SEVR may indicate enhanced reflection waves and decreased arterial function. The Dicrotic Notch is a characteristic point on the arterial pulse wave, also known as the notch of the dicrotic wave.

[0042] SDNN (Standard Deviation of Normal to Normal Intervals) is an important indicator in the time domain analysis of heart rate variability. SDNN represents the overall standard deviation of the sinus beat RR interval. The normal value range of SDNN is 141ms±39ms. This indicator can reflect the overall level of heart rate variability. The larger the SDNN value, the higher the heart rate variability and the better the cardiac autonomic nervous system regulation function; conversely, the smaller the SDNN value, it may indicate that the cardiac autonomic nervous system regulation function is impaired.

[0043] RMSSD (Root Mean Square of Successive Differences) is another time domain indicator used to evaluate heart rate variability. RMSSD reflects the changes in heart beat intervals. The normal range of RMSSD varies with individual differences, but it is generally considered to be between 20-50 milliseconds. Some studies also believe that its normal range is (27±12) ms. A higher RMSSD value may indicate better cardiac autonomic nervous regulation, while a lower value may be related to heart disease or autonomic nervous dysfunction. RMSSD is an important value in heart rate variability analysis, and is used together with indicators such as SDNN to evaluate abnormal heart rate variability.

[0044] Daubechies wavelet (dbN) is an orthogonal wavelet, usually expressed as dbN, where N is the sequence number of the wavelet (for example, db1, db2, etc.). The larger N is, the higher the vanishing moments of the wavelet and the stronger the localization ability in the frequency domain, but it also increases the computational complexity.

[0045] Symlets wavelet (symN) is an orthogonal compactly supported wavelet function obtained by improving the Daubechies wavelet, usually expressed as symN (N=2, 3, ..., 8).

[0046] Coiflets wavelet (coifN) is a discrete wavelet, usually represented as coifN (N=1, 2, ..., 5).

[0047] Meyr wavelet is a relatively uncommon wavelet basis, and its definition and properties may vary in different literature or applications. Generally speaking, Meyr wavelet may refer to a wavelet function with specific mathematical expressions and properties.

[0048] BMI (Body Mass Index) is an internationally accepted indicator used to assess whether a person's weight is within a healthy range. The calculation formula for BMI is weight (in kilograms) divided by the square of height (in meters), that is, BMI = weight (kg) / (height (m) × height (m)).

[0049] Mean Absolute Error (MAE) is an indicator used to measure the prediction accuracy of a prediction model on continuous data. It measures the average absolute difference between the predicted value and the true value, indicating the average degree of deviation between the predicted value and the true value, and is one of the commonly used performance evaluation indicators in regression tasks.

[0050] In the related art, the measuring equipment generally collects the pulse wave of the patient's carotid artery and femoral artery at the same time to obtain the pulse wave propagation time difference, and manually measures the distance between the patient's carotid artery and femoral artery, and then determines the cfPWV value based on the pulse wave propagation time difference and the measured distance. The higher the cfPWV value, the harder the arterial wall is and the more serious the degree of arteriosclerosis is.

[0051] However, the high price of measuring equipment (such as arteriosclerosis detectors) increases the cost of measuring cfPWV; and because the measuring equipment is large and difficult for patients to carry, the difficulty of detection is increased.

[0052] See also Figure 1 In order to solve the above technical problems, the present application provides a method for determining carotid-femoral pulse wave velocity, which may include:

[0053] S101: Acquire user information and a PPG signal of a user to be measured.

[0054] The user to be measured refers to the user whose cfPWV is to be measured. User information refers to information related to the user to be measured, and the user information can be obtained by filling in the relevant information of the user to be measured. The PPG signal can be collected from the user to be measured by a wearable device or a device with a PPG signal collection function.

[0055] It should be noted that in the embodiments of the present application, the user to be measured needs to wear a wearable device (such as a watch, a bracelet, a wearable device with a PPG signal collection function, etc.) or a device with a PPG signal collection function (such as a medical device, etc.) to obtain the PPG signal. For the sake of convenience, in the following embodiments, the wearable device or the device with a PPG signal collection function can be collectively referred to as a collection device.

[0056] It should be understood that the user to be measured or the auxiliary testing personnel (auxiliary testing personnel refers to the personnel who help the user to be measured to perform arteriosclerosis testing, such as doctors, etc.) can fill in the information through the information filling interface displayed by the collection device, or fill in the information on the information filling interface of the terminal device associated with the collection device. The terminal device means a device that has the ability to communicate with the collection device, such as a mobile phone, tablet, etc., and is not specifically limited here.

[0057] In a possible implementation, only the height and age in the user information may be filled in, and the information other than the height and age may be fully filled in, selectively filled in, or left blank, which is not specifically limited here.

[0058] It should be understood that in terms of physiology, the pulse wave propagation velocity is closely related to vascular arteriosclerosis. Different degrees of arteriosclerosis will change the propagation of hemodynamics, thereby affecting the waveform characteristics of the PPG signal. Therefore, the collected PPG signal can provide relevant information for determining the cfPWV value.

[0059] It should be understood that if the acquisition device is a wearable device, a single measurement or a periodic measurement can be performed through the wearable device. If the measurement mode of the wearable device is a single measurement mode, since a single measurement will not have a small impact on the power consumption of the wearable device, the sampling rate of the PPG signal can be set to a higher frequency, such as 100 Hz. If the measurement mode of the wearable device is a periodic measurement mode, in order to reduce the power consumption of the wearable device, the sampling rate of the PPG signal can be set to a lower frequency, such as 25 Hz.

[0060] It should be understood that if the user to be measured uses a periodic measurement mode, the wearable device will perform periodic measurements based on the period determined by the user to be measured, thereby achieving continuous monitoring of the cfPWV value of the user to be measured.

[0061] In some possible implementations, the system can also detect the wearing condition of the acquisition device. If the user does not wear the acquisition device correctly, resulting in a poor wearing condition, the system of the acquisition device can prompt the user to be measured or the auxiliary detection personnel to adjust the wearing position of the detection device through voice, pop-up display, etc.

[0062] In the embodiment of the present application, the time length of the PPG signal collected by the system of the collection device can be determined according to actual needs, such as 30 seconds, 40 seconds or other time lengths, which are not specifically limited here.

[0063] S102: Determine a user feature according to the user information, and determine a first feature according to the PPG signal.

[0064] User characteristics may include personal information characteristics related to the user to be measured. Personal information characteristics refer to the characteristics of the user to be measured, such as personal information characteristics may include but are not limited to height, weight, age, gender, BMI (Body Mass Index, i.e. body mass index), etc.

[0065] In some possible implementations, user characteristics may include personal information characteristics and health characteristics. Health characteristics refer to characteristics related to the health of the user to be measured, such as health characteristics may include but are not limited to the user's history of hypertension, history of diabetes, medication use, smoking habits, drinking habits, etc.

[0066] The first feature refers to a feature obtained by processing the PPG signal, such as a time domain feature, a frequency domain feature, and a time-frequency domain feature.

[0067] Time domain features refer to the physiological features, waveform features, and heart rate features of the PPG signal. Figure 2 The pulse wave waveform diagram corresponding to the PPG signal shown, physiological characteristics include but are not limited to the main wave height (h1), peak time (t1), cardiac cycle (T), normalized peak time (t1 / h1), ejection time (t2), reflection index (h3 / h1), stiffness index (user height / t4) and SEVR (A2 / A1, where A2 and A1 are the areas on the right and left sides of the Jiangzhong Isthmus), etc.

[0068] The waveform features include the heights of the main wave of the pulse waveform corresponding to the PPG signal at 10%, 30%, 50%, 70% and 90%, the waveform width, the start time and the end time, and the k spectrum value. The k spectrum value is calculated by the following formula (1):

[0069] ;Formula (1)

[0070] in, is the amplitude of the measured PPG signal, T is the cardiac cycle, is the minimum amplitude of the PPG signal, is the maximum amplitude of the PPG signal.

[0071] The amplitude of the PPG signal refers to the difference between the maximum and minimum values ​​of the light intensity change within a single pulse cycle. This difference reflects the change in blood volume in the blood vessels during the contraction and relaxation of the heart. It should be understood that when the heart contracts, more blood is pumped into the blood vessels, causing the blood vessel volume to increase, thereby reducing the amount of transmitted or reflected light (depending on the type of PPG sensor), forming a peak (or trough, depending on how the signal is processed) in the PPG signal. Conversely, when the heart relaxes, the blood vessel volume decreases, and the amount of transmitted or reflected light increases, forming a trough (or peak) in the PPG signal.

[0072] Heart rate related features include heart rate and heart rate variability indicators. Heart rate variability indicators include but are not limited to SDNN and RMSSD.

[0073] Heart rate refers to the number of times the heart beats per minute and is one of the important indicators for assessing cardiovascular health. Heart rate can be determined by calculating the time interval between adjacent peaks in the PPG signal.

[0074] The heart rate variability index refers to the value of the change in the difference between successive heartbeat cycles. The level of the heart rate variability index can reflect the regulatory ability of the autonomic nervous system, and further reflect the physiological state of the user to be measured, such as the emotional level and sleep quality. The heart rate variability index can be obtained through the PPG signal.

[0075] Frequency domain features refer to the signal energy ratio of PPG signals in different frequency domains, such as the signal energy ratio of PPG signals in the frequency ranges of 0.5-0.8Hz, 0.8-1.3Hz, 1.3-1.8Hz and 1.8-2.5Hz. Among them, energy is used to reflect the strength or "activity" of PPG signals in different frequency ranges, and signal energy ratio means the proportion of signals in a frequency range to the entire PPG signal.

[0076] As a possible implementation method, the signal energy ratio can be obtained by:

[0077] A1: Preprocess the collected PGG signal.

[0078] The preprocessing may include but is not limited to denoising, filtering, etc. By preprocessing the PPG signal, the quality of the PPG signal can be improved.

[0079] A2: The PPG signal is converted from the time domain to the frequency domain through Fourier transform (such as fast Fourier transform FFT) to obtain the spectrum corresponding to the PPG signal.

[0080] The spectrum is the representation of the PPG signal in the frequency domain, and the spectrum describes the distribution of different frequency components in the PPG signal.

[0081] A3: Calculate the sum of the squares of the spectrum in each frequency range to obtain the energy corresponding to each frequency domain range.

[0082] As an example, assuming that the energy proportion in the range of 0.5-0.8 Hz is calculated, the part of the spectrum corresponding to 0.5-0.8 Hz can be found, and then the sum of the squares of this part of the spectrum can be calculated.

[0083] A4: Calculate the sum of the squares of the entire spectrum to get the total energy of the PPG signal.

[0084] A5: Divide the energy in each frequency range by the total energy to obtain the signal energy ratio corresponding to each frequency domain range.

[0085] The time-frequency domain features refer to the parameters extracted from the PPG signal after wavelet transformation, such as mean, standard deviation, energy, entropy, etc. In the embodiment of the present application, the PPG signal can be subjected to wavelet transformation using wavelets such as Daubechies, Symlets, Coiflets, and Meyr.

[0086] S103: Determine a second feature based on the user feature and the first feature.

[0087] The second feature refers to the feature of the first feature that has a higher correlation with the pulse wave propagation time. The second feature can be determined from time domain features, frequency domain features, time-frequency domain features, and user features.

[0088] As an example, assuming that the first feature includes feature 1, feature 2 and feature 3, the user features include feature 4 and feature 5, the correlation between feature 1 and pulse wave propagation time is 10%, the correlation between feature 2 and pulse wave propagation time is 50%, the correlation between feature 3 and pulse wave propagation time is 90%, the correlation between feature 4 and pulse wave propagation time is 70%, and the correlation between feature 5 and pulse wave propagation time is 80%, the feature with a correlation greater than 50% with the pulse wave propagation time is taken as the second feature, then the second features are feature 3, feature 4 and feature 5.

[0089] In a possible implementation, the second feature may be all features in the first feature and all features in the user feature.

[0090] It should be understood that the embodiment of the present application can improve the comprehensiveness and accuracy of feature extraction by fully exploiting the multi-dimensional features (time domain features, frequency domain features and time-frequency domain features) of the PPG signal and then determining the second feature based on the multi-dimensional features and user characteristics.

[0091] S104: Determine the distance between the carotid artery and the femoral artery of the user to be measured based on the user characteristics by using the distance prediction model.

[0092] The distance prediction model refers to a model for predicting the distance between the carotid artery and the femoral artery of a user to be measured. The distance prediction model can be obtained based on a linear fitting model, a transfer model, a neural network model or other model training.

[0093] In a possible implementation, the distance prediction model can be trained based on a linear fitting model. The linear fitting model is used to describe the relationship between two or more variables through a linear equation. In the linear fitting model, a linear combination of one or more independent variables (user features) is used to predict or explain the change of the dependent variable (the distance between the carotid artery and the femoral artery). The core of the linear fitting model is to assume that there is a linear relationship between the dependent variable and the independent variable, that is, the relationship between them can be approximated by a straight line (in two-dimensional space) or a hyperplane (in multidimensional space).

[0094] It should be understood that since the linear fitting model has a relatively fast prediction speed, it can quickly predict new input data. At the same time, the linear fitting model occupies a small amount of memory and has fewer computing resources, which puts less burden on wearable devices.

[0095] In a possible implementation, since height is highly correlated with the distance (between the carotid artery and the femoral artery), the distance between the carotid artery and the femoral artery can be obtained by inputting height into a distance prediction model.

[0096] In another possible implementation, in order to improve the accuracy of distance prediction, user characteristics may include height, weight, gender, etc. The height, weight, and gender are input into the distance prediction model to obtain the distance between the carotid artery and the femoral artery.

[0097] In a possible implementation, the manually measured distance between the carotid artery and the femoral artery of the user to be measured can also be input into the distance prediction model as a user feature. The distance prediction model will calibrate the input distance. If the input distance differs from the distance predicted by the distance prediction model within a preset range (for example, 10%), the input distance can be used as the output distance; otherwise, the distance predicted by the distance prediction model is used as the output distance.

[0098] It should be understood that if the distance predicted by the distance prediction model differs from the manually measured distance within a preset range, it can be considered that the manually measured distance is more accurate. In this case, it is more accurate to use the manually measured distance as the distance between the carotid artery and the femoral artery, thereby improving the accuracy of cfPWV; if the distance predicted by the distance prediction model differs from the manually measured distance within a preset range, it can be considered that there is a problem with the manual measurement and the distance of the distance prediction model is more accurate. In this case, it is more accurate to use the distance predicted by the distance prediction model as the distance between the carotid artery and the femoral artery, thereby improving the accuracy of cfPWV.

[0099] S105: Determine the pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured based on the second feature through the time prediction model.

[0100] The time prediction model refers to a model for predicting the pulse wave propagation time between the carotid artery and the femoral artery. The time prediction model can be obtained based on a linear fitting model, a transfer model or other model training.

[0101] In a possible implementation, the second feature may include a time domain feature and an age feature in the user feature, and the time prediction model may predict the pulse wave propagation time according to the time domain feature and the age feature.

[0102] It should be understood that since the time domain characteristics can directly reflect the changes of PPG signals over time, the dynamic process of heart beats and blood flow can be reflected by calculating the number of peaks, waveform shape, time interval and other parameters; and the pulse wave propagation time is based on the measurement on the time axis, that is, the time difference between the pulse wave of the carotid artery and the pulse wave of the femoral artery. Therefore, the time domain characteristics of the PPG signal can be highly correlated with the pulse wave propagation time.

[0103] Similarly, as we age, the blood vessel wall undergoes a natural aging process. The elastic tissue of the blood vessel wall (such as elastic fibers) gradually decreases and is replaced by inelastic fibers (such as collagen fibers), causing the blood vessel wall to gradually become harder, more brittle, and less elastic. This change is physiological and inevitable. Therefore, there is a high correlation between hardening or reduced elasticity of the arterial wall and age characteristics.

[0104] In another possible implementation, the pulse wave propagation time may be predicted by the third feature.

[0105] The third feature refers to the feature selected from the second feature. It should be understood that although the second feature refers to a feature with a high correlation with the pulse wave propagation time, in the prediction process of the time prediction model, not all second features contribute equally to the prediction of the pulse wave propagation time. Some features may be highly correlated with the pulse wave propagation time (i.e., features with a large contribution), while other features may contribute less to the pulse wave propagation time or even be redundant (hereinafter referred to as such features as features with a small contribution). Redundant and irrelevant features will increase the complexity of the model, may lead to overfitting, and reduce the generalization ability of the model. Therefore, it is necessary to perform feature selection on the second feature, which can remove these unnecessary features, simplify the time prediction model, and improve the robustness, accuracy and generalization of the time prediction model, as well as the accuracy of the pulse wave propagation time. Feature selection can delete features with a small contribution in the second feature by filtering, packaging, embedding, etc.

[0106] The filtering method selects features based on statistical tests or correlation analysis, such as chi-square test, ANOVA, mutual information, etc. This method does not depend on a specific machine learning algorithm. The packaging method uses the target machine learning algorithm to evaluate the quality of feature subsets and select the optimal feature combination through iteration. This method has a high computational cost, but it can usually find a more accurate feature subset. The embedding method automatically selects features during model training, such as tree-based methods (random forest, gradient boosting machine, etc.) to evaluate the importance of features.

[0107] As an example, assuming that the second feature includes time domain features (A, B, C), frequency domain features (D, E), time-frequency domain features (H, I, G, K, L) and user features (M, N, O, P), and these 14 features are filtered by the filtering method, and the features with smaller contributions are determined to be D, E, H, I, G, O, and P. Then the third features include A, B, C, K, L, M, and N.

[0108] It should be noted that although both the distance prediction model and the time prediction model can be trained based on the same initial model (such as a linear fitting model), the distance prediction model and the time prediction model are two independent models. The two models predict the distance between the carotid artery and the femoral artery and the pulse wave propagation time, respectively, which can reduce the model's sensitivity to noise and outliers.

[0109] It should be noted that step 104 and step S105 of the embodiment of the present application can be predicted at the same time, or step S105 can be performed first and then step S104, which is not specifically limited here.

[0110] S106: Determine the cfPWV value according to the distance between the carotid artery and the femoral artery and the pulse wave propagation time through the cfPWV prediction model.

[0111] The cfPWV prediction model refers to a model used to predict the cfPWV value of the user to be measured. In a possible implementation, the cfPWV prediction model may be the following formula (2):

[0112] ;Formula (2)

[0113] in, represents the coefficient, represents the distance between the carotid artery and the femoral artery, represents the pulse wave propagation time between the carotid artery and the femoral artery. The value range of k is [0.8, 1.2], which can be selected according to actual needs, such as 0.8, 0.9, 1.2, etc., and is not specifically limited here.

[0114] It should be noted that if the cfPWV value is predicted directly based on user features and the second feature, the sensitivity of the cfPWV prediction model to noise and outliers will increase, thereby reducing the robustness of the cfPWV model. Therefore, the distance between the carotid artery and the femoral artery and the pulse wave propagation time are determined by the distance prediction model and the time prediction model respectively, and then the cfPWV value is predicted according to the distance and the pulse wave propagation time by the cfPWV model, which can ensure the accuracy and stability of the predicted cfPWV value.

[0115] It should be noted that the distance prediction model, time prediction model and cfPWV prediction model in the embodiments of the present application can all adopt the training methods involved in the relevant technologies, which will not be described in detail here.

[0116] It should be understood that there is a close correlation between arteriosclerosis and cfPWV. As a non-invasive indicator, cfPWV can be used to assess the degree of arteriosclerosis. cfPWV is calculated by measuring the distance between the carotid artery and the femoral artery and the pulse wave propagation time. The more severe the arteriosclerosis, the less distensibility and elasticity of the blood vessels, the higher the stiffness, and the higher the cfPWV value.

[0117] In the method for determining the carotid-femoral pulse wave velocity provided in the embodiment of the present application, the acquisition device can be a wearable device. Without increasing the measurement cost of cfPWV, the user can carry it with him / her, thereby reducing the difficulty of detection. Moreover, by filling in the user information and acquiring the PPG signal, the user characteristics and the second characteristics can be determined. Then, based on the user characteristics and the second characteristics, the distance between the carotid artery and the femoral artery and the pulse wave propagation time can be predicted respectively. The method is suitable for wearable devices that only support PPG signal acquisition, thereby improving the compatibility of wearable devices. Moreover, while ensuring the accuracy of cfPWV, the complexity of detection is reduced.

[0118] Based on the method for determining the carotid-femoral pulse wave velocity provided in the above embodiment, in order to further illustrate the application process of the method for determining the carotid-femoral pulse wave velocity, as an example, Figure 3 As shown, the acquisition device takes a watch as an example. The watch has a PPG signal acquisition function. After the user to be measured wears the watch, he opens the user interface for cfPWV measurement (such as Figure 3 Then, after the user clicks the "Start" button, he can choose to perform a single measurement or a periodic measurement (as shown in a in the figure). Figure 3 Assume further that the user selects single measurement, the system will also prompt the user to choose whether to manually enter the distance or automatically predict it (as shown in b in the figure). Figure 3 Assuming that the user selects automatic prediction, the user can click the "Confirm" button to collect the PPG signal, or click the "Cancel" button to cancel the collection of the PPG signal (as shown in Figure 3 Assume that the user clicks the "Confirm" button, the system starts collecting PPG signals (as shown in d in Figure 3 After the PPG signal is collected, the watch will automatically predict the cfPWV value based on the pre-filled user information and PPG signal, and display the predicted cfPWV value (as shown in Figure 5-1). Figure 3 ).

[0119] Based on the method for determining the carotid-femoral pulse wave velocity provided in the above embodiment, see Figure 4 The embodiment of the present application also provides another carotid-femoral pulse wave velocity detection process, which can be:

[0120] Step 1: Get user information and PPG signal.

[0121] The user information is obtained based on the user information filled in by the user to be measured. The user information includes personal information characteristics and health characteristics. The PPG signal is a signal collected by a wearable device or a device with a PPG signal collection function.

[0122] Step 2: Obtain user features based on user information, and perform filtering and feature extraction on the PPG signal to obtain time domain features, frequency domain features, and time-frequency domain features.

[0123] Among them, the time domain feature, the frequency domain feature and the time-frequency domain feature are the first features of the above embodiment.

[0124] The filtering process includes three processing steps: removing baseline drift, extracting stable segments and signal smoothing.

[0125] Baseline drift removal refers to removing a slowly changing background trend from the PPG signal, which is usually caused by instrument drift, environmental changes, or physiological processes. Baseline drift will cause the PPG signal to deviate from its true value, affecting subsequent PPG signal analysis and processing. Methods for removing baseline drift include but are not limited to: wavelet transform, filtering technology, smoothing prior method, or MATLAB built-in function, etc.

[0126] Wavelet transform is mainly through multi-scale decomposition process. The baseline trend of the signal can be observed in the decomposed low-frequency coefficients, and this baseline trend can be subtracted from the original signal. Filtering technology is to remove low-frequency baseline drift components through low-pass filters, median filters, etc., and retain high-frequency valid signals. The smoothing prior method uses the prior knowledge of the smoothness of the signal to remove the baseline drift through the optimization algorithm. MATLAB built-in functions are used directly to remove the linear trend term of the signal through functions such as detrend.

[0127] The stationary segment extraction process refers to identifying and separating the segments whose statistical characteristics (such as mean, variance, etc.) do not change over time from the PPG signal. The PPG signal after stationary segment processing is very important for subsequent signal analysis and processing, which can improve predictability and analyzability.

[0128] Signal smoothing refers to processing the PPG signal through a certain algorithm or filter to reduce random noise or high-frequency interference in the PPG signal and make the PPG signal smoother. Signal smoothing helps to improve the signal-to-noise ratio and readability of the PPG signal, providing a better basis for subsequent signal analysis and processing. Signal smoothing methods include but are not limited to: sliding average method, low-pass filtering, median filtering or other filtering methods.

[0129] The sliding average method is to average the values ​​of a certain point in the signal and its surrounding points, and use the average value as the smoothing value of the point. Low-pass filtering is to remove the high-frequency components in the signal through a low-pass filter and retain the effective low-frequency signal. Median filtering is to take the median of the surrounding points for each point in the signal as the smoothing value of the point, which is suitable for removing noise, etc. Other filtering methods can be Gaussian filtering, Wiener filtering, etc., which can also be used for signal smoothing.

[0130] The feature extraction process of time domain features, frequency domain features and time-frequency domain features can refer to the acquisition process of time domain features, frequency domain features and time-frequency domain features in the above embodiment, which will not be repeated here.

[0131] Step 3: Use the distance prediction model to predict the distance between the carotid artery and the femoral artery based on user characteristics to obtain the model predicted distance L1.

[0132] Step 3 of the embodiment of the present application is the same as the distance prediction process of step S104 in the above embodiment. For relevant explanations, please refer to the explanations of the above step S104, which will not be repeated here.

[0133] Step 4: Determine whether there is a manually input distance L2; if not, use the model predicted distance L1 as the output distance L; if so, determine whether the difference between the manually input distance L2 and the model predicted distance L1 is less than 10%; if greater than or equal to 10%, use the model predicted distance L1 as the output distance L; if less than 10%, use the manually input distance L2 as the output distance L.

[0134] The manually input distance L2 is the distance manually added by the user to be measured, the model predicted distance L1 is the distance predicted by the distance prediction model based on user characteristics, and the output distance L is the distance output by the final distance prediction model.

[0135] Step 5: Determine the second feature (or feature vector) based on the user feature, the time domain feature, the frequency domain feature, and the time-frequency domain feature.

[0136] The method for determining the second feature can refer to the method for determining the second feature in the above embodiment, which will not be repeated here.

[0137] Step 6: Determine the pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured based on the second feature through the time prediction model .

[0138] The implementation process of step 6 in the embodiment of the present application is similar to the implementation process of step S105 in the above embodiment. For relevant explanations, please refer to the explanation of step S105, which will not be repeated here.

[0139] Step 7: Based on the output distance L and pulse wave propagation time The cfPWV value of the user to be measured is determined.

[0140] The implementation process of step 7 in the embodiment of the present application is similar to the implementation process of step S106 in the above embodiment. For relevant explanations, please refer to the explanation of step S106, which will not be repeated here.

[0141] Based on the method for determining the carotid-femoral pulse wave velocity provided in the above embodiment, see Figure 5 The embodiment of the present application also provides an interactive process of a method for determining a carotid-femoral pulse wave velocity. The acquisition device takes a wearable device as an example. The process may include:

[0142] S501: The user selects single measurement or periodic measurement through the measurement interface of the wearable device.

[0143] The user may be a user to be measured (such as a patient) or a user who assists the user to be measured in measuring (such as a doctor). The measurement interface refers to the interface of the application for measuring the cfPWV value displayed by the wearable device.

[0144] S502: The application triggers cfPWV monitoring in response to the user's operation of triggering the user interface.

[0145] The application is an application deployed on a wearable device, and the application is used to measure the cfPWV value. The user interface is an interface for the user to interact with the application.

[0146] S503: The system responds to the action of triggering cfPWV monitoring by the application and issues a collection instruction.

[0147] The system is the operating system of the wearable device, which can receive the cfPWV monitoring action triggered by the application and send collection instructions to the wearable device. The collection instruction means the instruction to collect PPG signals.

[0148] S504: The PPG collection indicator light of the device lights up, and the PPG signal of the user to be measured is collected.

[0149] S505: The algorithm module detects the wearing condition of the device, and reports to the system if the device is not worn or is not worn properly.

[0150] The algorithm module is a module set in the system. The algorithm module includes a variety of different algorithms, such as algorithms for detecting the wearing status of wearable devices, feature extraction algorithms, parsing algorithms, distance prediction models, time prediction models, cfPWV prediction models, etc., which are not specifically limited here.

[0151] S506: The algorithm module analyzes the collected PPG signal.

[0152] S507: The algorithm module caches the parsed PPG signal for 30 seconds.

[0153] Among them, the embodiment of the present application takes caching the PPG signal for "30 seconds" as an example, and the length of the cache time can be determined according to actual needs.

[0154] S508: The algorithm module extracts features from the cached PPG signal.

[0155] S509: The algorithm module calculates the distance between the carotid artery and the femoral artery and the pulse wave propagation time based on the user feature and the second feature.

[0156] The algorithm module determines the distance between the carotid artery and the femoral artery and the pulse wave propagation time. For details, see the process of determining the distance in step S104 and the process of determining the pulse wave propagation time in step S105 of the above embodiment, which will not be repeated here.

[0157] The user characteristics are obtained based on the user information manually filled in by the user. If the user manually fills in the distance between the carotid artery and the femoral artery, the user interface of the application will display the measurement point and precautions.

[0158] S510: Send the calculated cfPWV value to the system.

[0159] S511: The system stores the cfPWV value and displays the cfPWV value and pulse wave propagation time through the user interface of the application.

[0160] Among them, after displaying the cfPWV value and pulse wave propagation time, the PPG collection indicator light of the device goes out, and the measurement ends.

[0161] Based on the method for determining the carotid-femoral pulse wave velocity provided in the above embodiment, see Figure 6 , Figure 6 Method 1 in the above shows the cfPWV values ​​directly measured using the random forest model in the related art (e.g. Figure 6 point shown in method 1) Figure 6Method 2 in the present application shows the cfPWV value obtained by using the predicted distance and the predicted pulse wave propagation time (such as Figure 6 point shown in method 2) Figure 6 Method 3 in the present application shows the cfPWV value obtained by using the manually measured distance and the predicted pulse wave propagation time (such as Figure 6 point shown in method 3).

[0162] at the same time, Figure 6 The cfPWV value corresponding to method 1 in the figure also has a reference line y1. Figure 6 The cfPWV value corresponding to method 2 in the figure also has a reference line y2. Figure 6 In the cfPWV value graph corresponding to the method 3 in the embodiment, a reference line y3 is also provided, and these three reference lines are used to indicate that the cfPWV value predicted by the method 1, the method 2 or the method 3 is equal to the actual value. It should be understood that the more points in the cfPWV value graph are close to the reference line, the more accurate the method is.

[0163] Combination Figure 6 It can be seen from the cfPWV value graph corresponding to the three methods shown in that the cfPWV values ​​predicted by methods 2 and 3 are more accurate than the cfPWV values ​​directly measured by method 1, and the cfPWV value obtained by method 3 is more accurate than the cfPWV value obtained by method 2.

[0164] In addition, the mean absolute error (MAE) of method 1 is 0.8751, the mean absolute error (MAE) of method 2 is 0.7974, and the mean absolute error (MAE) of method 3 is 0.7505.

[0165] It should be understood that MAE represents the average of the absolute values ​​of the differences between the predicted value and the actual value. The smaller the MAE, the smaller the deviation between the predicted cfPWV value and the actual cfPWV value, and the more accurate the cfPWV value obtained in this way. Compared with directly using the random forest model to measure the cfPWV value, the embodiment of the present application determines the cfPWV value with higher accuracy by respectively predicting the distance between the carotid artery and the femoral artery and the pulse wave propagation time; at the same time, if the user can accurately measure the distance between the carotid artery and the femoral artery, the cfPWV value can be determined by using the manually entered distance and the predicted pulse wave propagation time, which can further improve the measurement accuracy of cfPWV.

[0166] It should be noted that in the above embodiments, the method for determining the carotid-femoral pulse wave velocity can be implemented by different devices:

[0167] The method for determining the carotid-femoral pulse wave velocity can be implemented through a wearable device. The wearable device can be a wearable device with a PPG signal acquisition function, such as a watch, a bracelet, etc.

[0168] The method for determining the carotid-femoral pulse wave velocity can also be implemented by a device with computing capabilities and PPG signal acquisition functions, such as medical equipment and the like.

[0169] The method for determining the carotid-femoral pulse wave velocity can also be implemented by combining an acquisition device and a terminal device. The acquisition device refers to a device with a PPG signal acquisition function, and the terminal device refers to a device with computing capabilities, such as a watch for the acquisition device and a mobile phone for the terminal device, or a medical device for the acquisition device and a computer for the terminal device. The acquisition device is used to acquire PPG signals, and the terminal device is used to determine the cfPWV value based on the PPG signals and user information.

[0170] The technical solution of this embodiment, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment. The aforementioned storage medium includes: flash memory, mobile hard disk, read-only memory, random access memory, magnetic disk or optical disk, etc., various media that can store program codes.

[0171] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for determining carotid-femoral pulse wave velocity, characterized in that: include: Acquire a user feature and a second feature; the user feature is obtained based on user information of the user to be measured; the second feature represents a feature associated with pulse wave propagation time; the second feature is obtained based on the first feature and the user feature; the first feature is obtained based on a photoplethysmogram (PPG) signal; the PPG signal is obtained by collecting the user to be measured; the second feature includes k features, where k is a positive integer; Determine the contribution of each feature in the second feature to the time prediction model to obtain k contributions; the time prediction model is used to predict the pulse wave propagation time; Sort the k contributions in descending order; The features corresponding to the first q contribution degrees among the sorted k contribution degrees are taken as the third feature; q is a positive integer less than or equal to k; determining a distance between the carotid artery and the femoral artery of the user to be measured based on the user characteristic, and determining a pulse wave propagation time between the carotid artery and the femoral artery of the user to be measured based on the third characteristic; The carotid-femoral pulse wave velocity value of the user to be measured is determined according to the distance and the pulse wave propagation time.

2. The method according to claim 1, characterized in that The second feature is obtained by: Determine the first feature corresponding to the PPG signal, the first feature comprising a time domain feature, a frequency domain feature, and a time-frequency domain feature; The second feature is determined according to the time domain feature, the frequency domain feature, the time-frequency domain feature and the user feature.

3. The method according to claim 2, characterized in that The determining the second feature according to the time domain feature, the frequency domain feature, the time-frequency domain feature and the user feature includes: Determine the correlation value between each feature in the time domain feature and the pulse wave propagation time, determine the correlation value between each feature in the frequency domain feature and the pulse wave propagation time, determine the correlation value between each feature in the time-frequency domain feature and the pulse wave propagation time, and determine the correlation value between each feature in the user feature and the pulse wave propagation time, and obtain n correlation values; wherein n is an integer greater than or equal to 1; Sorting the n association values ​​in descending order to obtain sorted n association values; The features corresponding to the first m associated values ​​among the sorted n associated values ​​are taken as the second features; m is a positive integer less than n.

4. The method according to claim 2, characterized in that: The determining the first feature corresponding to the PPG signal includes: Performing filtering processing on the PPG signal to obtain a processed PPG signal; Perform feature extraction on the processed PPG signal to obtain the time domain feature, the frequency domain feature and the time-frequency domain feature.

5. The method according to any one of claims 1 to 3, characterized in that: The determining the distance between the carotid artery and the femoral artery of the user to be measured based on the user characteristics includes: The distance between the carotid artery and the femoral artery of the user to be measured is determined based on the user characteristics through a distance prediction model; the distance prediction model is obtained based on training of a linear fitting model.

6. The method according to any one of claims 1 to 3, characterized in that: The time prediction model is obtained by training a linear fitting model.

7. An electronic device, characterized in that: include: Processor and memory; Wherein, one or more computer programs are stored in the memory, and the one or more computer programs include instructions; when the instructions are executed by the processor, the electronic device executes the carotid-femoral pulse wave velocity determination method according to any one of claims 1-6.

8. A computer storage medium, characterized in that: The method comprises computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the method for determining the carotid-femoral pulse wave velocity as described in any one of claims 1 to 6.

9. A computer program product, characterized in that When the computer program product runs on a computer, the computer executes the method for determining the carotid-femoral pulse wave velocity according to any one of claims 1 to 6.

Images