Wrist blood pressure measuring method and device based on pressure oscillation amplitude mapping and medium

Abstract

The invention discloses a wrist blood pressure measuring method and device based on pressure oscillation amplitude mapping and a medium. The method comprises the following steps: constructing a wrist-upper arm blood pressure oscillation wave data set, establishing an oscillation wave mapping measurement model, performing wrist blood pressure measurement based on pressure oscillation amplitude mapping, automatically identifying a maximum amplitude point based on upper arm blood pressure oscillation wave data obtained by mapping, calculating systolic pressure and diastolic pressure, and outputting a blood pressure estimation result. The intelligent wearable device for blood pressure measurement comprises a pressure acquisition module and a calculation module. The oscillatory waves collected at the wrist are mapped to the oscillatory waves of the upper arm, so that the convenient wrist blood pressure measurement intelligent wearable device reaches the clinical approval precision close to that of an upper arm type blood pressure measurement device, and the result of blood pressure measurement conducted by a patient through the blood pressure measurement intelligent wearable device has higher clinical reference value.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of blood pressure measurement, in particular to a wrist blood pressure measurement method and device based on pressure oscillation amplitude mapping, and a medium. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] Blood pressure refers to the lateral pressure generated by blood on the wall of a unit area of blood vessels. As a key physiological parameter of the human body, blood pressure is not only an important indicator of cardiovascular health, but also an important basis for assisting clinical diagnosis and treatment. Therefore, rapid and accurate blood pressure measurement is crucial.

[0004] Currently, blood pressure measurement methods are mainly divided into two categories: invasive and non-invasive. Invasive blood pressure measurement involves directly inserting an arterial catheter into the artery to obtain continuous blood pressure readings, which is the "gold standard" for blood pressure measurement and is mainly used for critically ill patients with unstable hemodynamics who require close monitoring. In non-invasive blood pressure measurement, the Korotkoff sound method and the oscillometric method are the most common methods. The Korotkoff sound method relies on a stethoscope to capture Korotkoff sounds, but this method has significant individual differences and cannot provide continuous blood pressure data. The oscillometric method estimates blood pressure values through oscillatory waves generated during the inflation and deflation of a cuff.

[0005] In the existing process of wrist blood pressure measurement by airbag inflation, most measurements are based on the principle of the oscillometric method. The measurement process is as follows: the air pump inflates the airbag, causing the airbag to expand under pressure. During the inflation and pressure increase process, the radial artery of the wrist is compressed, and the pressure sensor extracts the pulse oscillation wave signal. The blood pressure is calculated based on the characteristics of the extracted pulse wave oscillation signal. Due to the poor accuracy of wrist blood pressure measurement, in order to obtain more accurate blood pressure measurement values, blood pressure needs to be calibrated and compensated. The existing technology uses sensors on the photoplethysmography (PPG), electrocardiogram (ECG), and other wrist devices to assist blood pressure calculation, thereby calibrating and compensating the blood pressure values.

[0006] Therefore, how to solve the problem of existing wrist blood pressure meters needing to rely on multiple sensors for calibration and compensation during blood pressure measurement, and only calibrating and compensating blood pressure values, has become the subject of the present application. SUMMARY

[0007] The application aims to provide a wrist blood pressure measurement method and device based on pressure oscillation amplitude mapping, and a medium.

[0008] To achieve the above-mentioned purpose, the first aspect of the application provides a wrist blood pressure measurement method based on pressure oscillation amplitude mapping, which comprises the following steps:

[0009] A wrist-upper arm blood pressure oscillation wave dataset is constructed, wrist blood pressure oscillation wave data and paired upper arm blood pressure oscillation wave data of the brachial artery of a subject are actually collected multiple times, artificial generated wrist and upper arm paired oscillation wave data is obtained by using a conditional variational autoencoder for expansion, and the actually collected and artificial generated wrist and upper arm paired oscillation wave data constitutes the wrist-upper arm blood pressure oscillation wave dataset.

[0010] An oscillation wave mapping measurement model is established, a difference sequence of the wrist oscillation wave minus the upper arm pressure oscillation wave is extracted as a deviation label according to the wrist-upper arm blood pressure oscillation wave dataset, hemodynamic features of the wrist oscillation wave are extracted, the hemodynamic features are spliced into a hemodynamic feature vector, the hemodynamic feature vector is put into a deviation calibration sequence generator for deep learning, and a deviation calibration sequence is generated point by point, so as to obtain an oscillation wave mapping measurement relationship between the wrist blood pressure oscillation wave data and the upper arm blood pressure oscillation wave data, and the oscillation wave mapping measurement model is established.

[0011] Wrist blood pressure measurement is performed based on pressure oscillation amplitude mapping, a gas bag is inflated and pressurized, wrist blood pressure oscillation wave data of a measurer is collected, a hemodynamic feature vector is extracted from the wrist blood pressure oscillation wave data of the current measurer and input into the oscillation wave mapping measurement model, oscillation amplitude differences between the radial artery and the brachial artery of the measurer under different pressures are inferred according to the oscillation wave mapping measurement relationship, the pressure oscillation amplitude at the radial artery is mapped to the brachial artery, a standard oscillographic algorithm is used, the maximum amplitude point is automatically identified based on the mapped upper arm blood pressure oscillation wave data, systolic pressure and diastolic pressure are calculated, and a blood pressure estimation result is output.

[0012] The second aspect of the application provides a blood pressure measurement intelligent wearable device, which applies the wrist blood pressure measurement method based on pressure oscillation amplitude mapping as described in the first aspect, and comprises a pressure collection module and a calculation module.

[0013] The pressure collection module is used to actually collect wrist blood pressure oscillation wave data of a measurer.

[0014] The calculation module has a built-in oscillation wave mapping measurement model. The calculation module is used to extract hemodynamic feature vectors from the actual collected wrist blood pressure oscillation wave data, input them into the oscillation wave mapping measurement model, and infer the difference in oscillation amplitude between the radial artery and the brachial artery of the individual under different pressures based on the oscillation wave mapping measurement relationship. The pressure oscillation amplitude at the radial artery is mapped to the brachial artery. Using a standard oscilloscope algorithm, the maximum amplitude point is automatically identified based on the mapped upper arm blood pressure oscillation wave data, and systolic and diastolic blood pressure are calculated, and the blood pressure estimation result is output.

[0015] A third aspect of the invention provides a readable storage medium storing a control program, which, when executed by a computing module, causes the computing module to perform the steps of the method described in the first aspect of the invention.

[0016] The relevant content of this invention is explained as follows:

[0017] 1.The wrist blood pressure measurement method based on pressure oscillation amplitude mapping, the blood pressure measurement intelligent wearable device and the readable storage medium are innovative designs aiming at the problem that the existing wrist blood pressure meter needs to use multiple sensors for calibration compensation in the blood pressure measurement process and only calibrates and compensates for blood pressure values, are suitable for blood pressure measurement through airbag pressurization at the wrist, do not need additional sensor devices, only use the pressure oscillation wave signal extracted by the pressure sensor, and use the deep learning method to map the oscillation wave collected at the wrist to the upper arm oscillation wave, thereby improving the blood pressure measurement accuracy. In the wrist blood pressure measurement method based on pressure oscillation amplitude mapping, based on the principle of hemodynamics, a deep learning driven wrist-upper arm blood pressure oscillation mapping measurement model establishment and application framework is designed. Firstly, a real collection combined with an artificial enhancement generation method is used to expand the wrist-upper arm blood pressure oscillation wave dataset, so as to provide a scientific, reasonable, accurate, reliable and efficient data basis for the subsequent establishment of the oscillation wave mapping measurement model, thereby improving the accuracy and practicability of the oscillation wave mapping measurement model in wrist blood pressure measurement. Then, in the process of establishing the oscillation wave mapping measurement model, the deep learning method is used to combine the features on the oscillation wave with the physiological parameter blood vessel elasticity, the difference sequence of the wrist oscillation wave minus the upper arm pressure oscillation wave is extracted as a deviation label according to the wrist-upper arm blood pressure oscillation wave dataset, the hemodynamic features of the wrist oscillation wave are extracted, then the hemodynamic features are spliced into a hemodynamic feature vector, which is put into a deviation calibration sequence generator for deep learning, and a deviation calibration sequence is generated point by point, so as to obtain the oscillation wave mapping measurement relationship between the wrist blood pressure oscillation wave data and the upper arm blood pressure oscillation wave data, and establish the oscillation wave mapping measurement model. In this step, through high-order feature extraction on the pressure oscillation wave, the deep learning model is used to achieve the effect of re-constructing the oscillation wave at the brachial artery. Finally, in the step of wrist blood pressure measurement based on pressure oscillation amplitude mapping, only the features of the display blood vessels and physiological characteristics in the individual oscillation wave and the wrist and upper arm oscillation wave deviation of the individual are used, which not only solves the gap problem between the wrist oscillometric pressure oscillation signal and the upper arm pressure oscillation signal, but also significantly improves the accuracy and practicability of non-invasive portable blood pressure monitoring through deep learning technology.

[0018] 2. In the first aspect of the above technical solution, in the step of constructing the wrist-upper arm blood pressure oscillatory wave dataset, wherein the conditional variational autoencoder comprises an encoder and a decoder, the upper arm blood pressure oscillatory wave data is input as an input variable, the feature vector obtained after the time series embedding of the wrist blood pressure oscillatory wave data is input as a conditional variable into the encoder, the input variable is mapped into a latent variable under the restriction of the conditional variable, and the decoder obtains the augmented generated upper arm paired oscillatory wave data according to the latent variable and the conditional variable; the augmented generated upper arm paired oscillatory wave data is paired into a group with the corresponding actually collected wrist blood pressure oscillatory wave data, and artificial generated wrist and upper arm paired oscillatory wave data is obtained. Through the implementation of this step, more matched and more real and large amount of artificial generated wrist and upper arm paired oscillatory wave data is obtained, so as to further provide more abundant and accurately matched wrist-upper arm blood pressure oscillatory wave dataset, thereby further improving the establishment of a more optimized oscillatory wave mapping measurement model and more accurate measurement results.

[0019] 3. In the first aspect of the above technical solution, in the step of extracting the hemodynamic features of the wrist oscillatory wave, the features of each pulse corresponding oscillatory wave and the features of the oscillatory wave envelope constructed by all oscillatory waves are extracted.

[0020] The features of each pulse corresponding oscillatory wave include but are not limited to the reflection peak-to-main peak amplitude ratio, the reflection peak occurrence time, and the reflection wave significance.

[0021] The features of the oscillatory wave envelope include but are not limited to the maximum amplitude of the envelope and the corresponding pressure and time, the area covered by the oscillatory envelope and the area ratio before and after the maximum amplitude point, and the maximum slope when the envelope rises and falls.

[0022] Through the implementation of the above method, the vascular elasticity information is obtained through the features of the oscillatory wave and the oscillatory wave envelope, and the pressure wave propagation speed and the reflection wave amplitude of the blood vessel under different external pressures are further derived, so that the establishment and reasoning of the oscillatory wave mapping measurement relationship between the wrist blood pressure oscillatory wave data and the upper arm blood pressure oscillatory wave data are more reasonable.

[0023] 4. In the first aspect of the above technical solution, in the step of processing the PPG signal in the window, in the step of augmenting the artificial generated wrist and upper arm paired oscillatory wave data by using the conditional variational autoencoder, the following process is included:

[0024] First, a time series embedding tool is used to encode the wrist blood pressure oscillatory wave data into a feature vector, and the feature vector is input as a conditional variable into the conditional variational autoencoder CVAE model. Relying on the unsupervised training method, the network is reconstructed to complete the training of the CVAE network by inputting the signal.

[0025] After the CVAE network is trained, the embedding vector is extracted from the real wrist blood pressure oscillatory wave data, the vector and the randomly sampled normal distribution hidden vector are spliced, and input into the decoder of the CVAE model to obtain a large amount of artificially generated upper arm paired oscillatory wave data; the artificially obtained upper arm oscillatory wave and the corresponding real collected wrist oscillatory wave are paired into a group and stored in the database.

[0026] The implementation of the above method makes the process of expanding the wrist and upper arm paired oscillatory wave data more efficient and controllable, and can create training data matched with the relevant data of multiple individual subjects, thereby further improving the accuracy of the oscillatory wave mapping measurement model.

[0027] 5. In the first aspect of the above technical solution, the objective function L of the conditional variational autoencoder CVAE model is CVAE :

[0028] ;

[0029] wherein x represents an input variable, the upper arm pressure oscillatory wave signal is taken as the input variable x, c represents a conditional variable, the feature vector obtained by time embedding of the wrist blood pressure oscillatory wave data is taken as the conditional variable c, z represents a hidden variable, the upper arm pressure oscillatory wave signal generated by different sampling from the normal distribution sampling is taken as the hidden variable z, represents the encoder of the CVAE model, represents the decoder of the CVAE model, represents the prior distribution of the CVAE model, and D KL is a KL operator.

[0030] Further confirmation of the objective function L of the conditional variational autoencoder CVAE model CVAE ensures the reliability and interpretability of the encoding vector.

[0031] 6. In the first aspect of the above technical solution, the wrist-upper arm oscillatory wave acquisition module is used to acquire and enhance the obtained wrist-upper arm blood pressure oscillatory wave data set, and the deviation calibration sequence obtained by subtracting the upper arm pressure oscillatory wave from the wrist oscillatory wave is taken as the oscillatory wave mapping label of the wrist-upper arm oscillatory wave mapping measurement module.

[0032] The wrist blood pressure measurement based on the oscillatory wave mapping measurement model and the pressure oscillatory amplitude mapping is performed by using the wrist-upper arm oscillatory wave mapping measurement module.

[0033] In the process of taking the feature vector obtained by time series embedding of the wrist blood pressure oscillatory wave data as the conditional variable, the time series embedding adopted includes: feeding the time series signal of the wrist blood pressure oscillatory wave data into the encoding layer of a trained variational autoencoder, and extracting the encoding vector as the feature vector, the variational autoencoder has a VAE model, the objective function L of the VAE model is VAE :

[0034] ;

[0035] wherein, represents the encoder of the VAE model, represents the decoder of the VAE model, represents the prior distribution of the VAE model.

[0036] The design uses time series embedding in deep learning technology to obtain more reliable and effective feature vectors.

[0037] 7. In the first aspect of the above technical solution, the step of putting the bias calibration sequence generator into deep learning is performed in a forced learning manner, and the decoder uses the actually collected upper arm blood pressure oscillatory wave data as input at each step of generation;

[0038] In the step of inferring the oscillatory amplitude difference between the radial artery and the brachial artery of the measurer individual under different pressures according to the oscillatory wave mapping measurement relationship, the wrist blood pressure oscillatory wave data of the measurer is used to extract the hemodynamic feature vector of the measurer individual, and the hemodynamic feature vector is input into the oscillatory wave mapping measurement model. After the oscillatory wave mapping measurement relationship between the wrist blood pressure oscillatory wave data and the upper arm blood pressure oscillatory wave data of the measurer individual is extracted by inference, the oscillatory wave mapping measurement relationship of the measurer individual contains the oscillatory amplitude difference under different pressures. The blood pressure oscillatory wave envelope at the wrist in the wrist blood pressure oscillatory wave data is envelope reconstructed to reconstruct the blood pressure oscillatory wave envelope at the upper arm, and the pressure oscillation amplitude at the radial artery is mapped to the brachial artery.

[0039] Through the application of different learning methods in the deep learning and inference process, the efficiency and stability of the oscillatory wave mapping measurement model training are improved, and the accuracy in the inference process is improved.

[0040] 8. In the second aspect of the above technical solution, the blood pressure measurement intelligent wearable device is a device worn on the wrist of a user, and the blood pressure measurement intelligent wearable device comprises a shell, and a pressure collection module comprises a gas pump, a gas bag and a pressure sensor mounted on the shell; a mainboard is mounted in the shell, and the calculation module is mounted on the mainboard. Through the above further design of the blood pressure measurement intelligent wearable device, the structure of the wearable device specially used for wrist blood pressure measurement is optimized, and in combination with the implementation of the method, more accurate and reliable blood pressure measurement can be completed without the cooperation of other sensors (such as ECG signal sensors, PPG signal sensors and the like).

[0041] Due to the use of the above scheme, the present application has the following advantages and effects compared with the prior art:

[0042] 1. Through the implementation of the technical scheme of the present application, the wrist blood pressure measurement method based on pressure oscillation amplitude mapping, the blood pressure measurement intelligent wearable device and the readable storage medium are designed to solve the problem that the existing wrist blood pressure meter needs to use multiple sensors for calibration and compensation during blood pressure measurement and only calibrates and compensates the blood pressure value. The wrist blood pressure measurement method is suitable for blood pressure measurement by gas bag pressurization at the wrist, does not require additional sensor devices, and only uses the pressure oscillation wave signal extracted by the pressure sensor. Through the method of deep learning, the wrist oscillation wave collected is mapped to the upper arm oscillation wave, and the blood pressure measurement accuracy is improved.

[0043] 2. Through the implementation of the technical scheme of the present application, in the wrist blood pressure measurement method based on pressure oscillation amplitude mapping, a deep learning driven wrist-upper arm blood pressure oscillation mapping measurement model establishment and application framework is designed based on the principle of hemodynamics. First, a real collection combined with an artificial enhancement generation method is used to expand the wrist-upper arm blood pressure oscillation wave data set, to provide a scientific, reasonable, accurate, reliable and efficient data basis for the subsequent establishment of the oscillation wave mapping measurement model, so as to improve the accuracy and practicability of the oscillation wave mapping measurement model in wrist blood pressure measurement.

[0044] 3. Through the implementation of the technical scheme of the present application, in the process of establishing an oscillatory wave mapping measurement model, the method of deep learning is used to combine the features on the oscillatory wave with the physiological parameter of blood vessel elasticity, the difference sequence of the wrist oscillatory wave minus the upper arm pressure oscillatory wave is extracted as a deviation label according to the wrist-upper arm blood pressure oscillatory wave data set, the hemodynamic features of the wrist oscillatory wave are extracted, then the hemodynamic features are spliced into a hemodynamic feature vector, which is put into a deviation calibration sequence generator for deep learning, and a deviation calibration sequence is generated point by point to obtain the oscillatory wave mapping measurement relationship between the wrist blood pressure oscillatory wave data and the upper arm blood pressure oscillatory wave data, and the oscillatory wave mapping measurement model is established. In this step, through high-order feature extraction on the pressure oscillatory wave and with the help of the deep learning model, the effect of re-constructing the oscillatory wave at the brachial artery is achieved.

[0045] 4. Through the implementation of the technical scheme of the present application, in the step of wrist blood pressure measurement based on pressure oscillation amplitude mapping, only the feature set of the displayed blood vessels and physiological characteristics in the individual oscillatory wave and the predicted wrist and upper arm oscillatory wave deviation of the individual are used, which not only solves the gap problem between the wrist oscillographic pressure oscillation signal and the upper arm pressure oscillation signal, but also significantly improves the accuracy and practicability of non-invasive portable blood pressure monitoring through deep learning technology.

[0046] 5. In summary, the present application bridges the deviation of the wrist and upper arm type airbag pressure oscillation wave signals in the oscillographic pressure measurement process, aligns the two signals, maps the oscillatory wave collected at the wrist to the upper arm oscillatory wave, so that the convenient wrist blood pressure measurement intelligent wearable device can achieve the clinical recognized accuracy close to the upper arm type blood pressure measurement device, and the result of blood pressure measurement by the patient using the blood pressure measurement intelligent wearable device has more clinical reference value. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 It is a flowchart of the method of the embodiment of the present application;

[0048] Figure 2 It is a schematic diagram of the hardware module composition of the device of the embodiment of the present application;

[0049] Figure 3 It is a schematic diagram of the structure of the device of the embodiment of the present application;

[0050] Figure 4 It is a structure sectional view of the device of the embodiment of the present application;

[0051] Figure 5 It is a schematic diagram of the establishment and application framework of the wrist-upper arm blood pressure oscillatory wave mapping measurement model in the embodiment of the present application;

[0052] Figure 6 It is a schematic diagram of the CVAE model of the embodiment of the present application;

[0053] Figure 7 Fig. 1 is a schematic diagram of envelope curve generation by feature extraction of oscillation waveform graph of embodiment of the present application;

[0054] Figure 8 Fig. 2 is a schematic diagram of feature point extraction of embodiment of the present application;

[0055] Figure 9 Fig. 3 is a schematic diagram of wrist oscillation amplitude envelope curve of embodiment of the present application;

[0056] Figure 10 Fig. 4 is a schematic diagram of application of subject 1 for blood pressure measurement by using embodiment of the present application;

[0057] Figure 11 Fig. 5 is a schematic diagram of application of subject 2 for blood pressure measurement by using embodiment of the present application;

[0058] Figure 12 Fig. 6 is a schematic diagram of application of subject 3 for blood pressure measurement by using embodiment of the present application;

[0059] Figure 13 Fig. 7 is a schematic diagram of application of subject 4 for blood pressure measurement by using embodiment of the present application.

[0060] The parts of the above figures are shown as follows:

[0061] 1. housing;

[0062] 2. pressure collection module;

[0063] 21. air pump; 22. air bag; 23. pressure sensor;

[0064] 3. calculation module. DETAILED DESCRIPTION

[0065] In order to make the above objectives, features and advantages of the present application more apparent, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. In the following description, a lot of specific details are set forth in order to fully understand the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the present application, so the present application is not limited by the specific embodiments disclosed below.

[0066] The terms used herein are only for describing specific embodiments, and are not intended to limit the present application. The singular forms such as "a", "this", "this", "this" and "the" as used herein also include the plural forms.

[0067] The terms "first", "second", and the like as used herein do not imply a particular order or sequence, but are used to distinguish one component from another. They are not intended to mean "one", "the" or "at least one".

[0068] The terms "connected" or "positioned" as used herein can mean that two or more components or devices are in direct physical contact with each other, or are indirectly in physical contact with each other, or can mean that two or more components or devices are in operative or communicative connection.

[0069] The terms "comprising", "including", "containing", and the like as used herein are open-ended terms that are intended to mean "including, but not limited to".

[0070] The terms used herein are generally used in the same meaning as those commonly used in the art, unless otherwise specified in the context of the present specification. Certain terms used to describe the present application are discussed below or elsewhere in the specification, to provide additional guidance to those skilled in the art in understanding the description of the present application.

[0071] The present application aims to achieve the clinical recognition accuracy of the convenient wrist blood pressure measurement intelligent wearable device close to the upper arm blood pressure measurement device, and to solve the problem that the existing wrist blood pressure meter needs to use multiple sensors for calibration compensation during blood pressure measurement and only calibrates and compensates for blood pressure values. A wrist blood pressure measurement method based on pressure oscillation amplitude mapping, a blood pressure measurement intelligent wearable device, and a readable storage medium are designed, which are suitable for blood pressure measurement at the wrist through airbag pressurization, do not require additional sensor devices, and only use the pressure oscillation wave signal extracted by the pressure sensor. Through the method of deep learning, the oscillation wave collected at the wrist is mapped to the upper arm oscillation wave, and the blood pressure measurement accuracy is improved.

[0072] To achieve the purpose of the present application, the applicant has studied the physical and physiological background of the present application: according to the amplification effect of pressure wave, when the arterial pressure wave propagates from the central artery to the peripheral blood vessels, it encounters the end arterioles and capillaries, and the reflected wave is superimposed to amplify the pressure wave. The systolic pressure of the radial artery is usually slightly higher than that of the brachial artery, with a difference of about 5-30 mmHg. This difference is related to physiological parameters such as vascular elasticity and peripheral resistance. The diastolic pressure difference is smaller, usually close to or slightly lower than the brachial artery (the difference is generally <10 mmHg), because the diastolic pressure wave attenuates quickly. Therefore, according to this physiological and physical characteristic, the embodiment of the present application extracts the characteristics in the pressure oscillation wave, uses the method of deep learning, and derives the amplitude ratio of the reflected wave under different pressures, i.e. the difference in oscillation wave amplitude between the radial artery and the brachial artery, according to the principle that "the compliance of blood vessels with different elasticity is different, and the pressure wave conduction speed and reflected wave amplitude are different".

[0073] In an embodiment, the application provides a wrist blood pressure measurement method based on pressure oscillation amplitude mapping, which comprises the following steps:

[0074] A wrist-upper arm blood pressure oscillation wave dataset is constructed, wrist blood pressure oscillation wave data and paired upper arm blood pressure oscillation wave data of the brachial artery of a subject are collected multiple times, and a conditional variational autoencoder is used to expand to obtain artificially generated wrist and upper arm paired oscillation wave data. The real collected and artificially generated wrist and upper arm paired oscillation wave data constitute the wrist-upper arm blood pressure oscillation wave dataset.

[0075] An oscillation wave mapping measurement model is established, a difference sequence of the wrist oscillation wave minus the upper arm pressure oscillation wave is extracted as a deviation label according to the wrist-upper arm blood pressure oscillation wave dataset, hemodynamic features of the wrist oscillation wave are extracted, the hemodynamic features are spliced into a hemodynamic feature vector, the hemodynamic feature vector is put into a deviation calibration sequence generator for deep learning, and a deviation calibration sequence is generated point by point to obtain an oscillation wave mapping measurement relationship between the wrist blood pressure oscillation wave data and the upper arm blood pressure oscillation wave data, and the oscillation wave mapping measurement model is established.

[0076] The wrist blood pressure is measured based on the pressure oscillation amplitude mapping, the air bag is inflated and pressurized, the wrist blood pressure oscillation wave data of the measurer is collected, the hemodynamic feature vector is extracted from the wrist blood pressure oscillation wave data of the current measurer and input into the oscillation wave mapping measurement model, the oscillation amplitude difference between the radial artery and the brachial artery of the measurer under different pressures is inferred according to the oscillation wave mapping measurement relationship, the pressure oscillation amplitude at the radial artery is mapped to the brachial artery, the standard oscillographic algorithm is used, the maximum amplitude point is automatically identified based on the mapped upper arm blood pressure oscillation wave data, the systolic pressure and diastolic pressure are calculated, and the blood pressure estimation result is output.

[0077] In the wrist blood pressure measurement method based on the pressure oscillation amplitude mapping, a deep learning driven wrist-upper arm blood pressure oscillation wave mapping measurement model establishment and application framework is designed based on the hemodynamic principle. The wrist-upper arm blood pressure oscillation wave dataset is expanded by real collection combined with artificial enhancement generation, which provides a scientific, reasonable, accurate, reliable and efficient data basis for the subsequent establishment of the oscillation wave mapping measurement model, so as to improve the accuracy and practicability of the oscillation wave mapping measurement model in wrist blood pressure measurement.

[0078] In the wrist blood pressure measurement based on the pressure oscillation amplitude mapping step described above, only the feature set of the displayed blood vessels and physiological characteristics in the individual oscillation wave and the prediction of the wrist and upper arm oscillation wave deviation of the individual are used to not only solve the gap problem between the wrist oscillation signal and the upper arm oscillation signal, but also significantly improve the accuracy and practicability of non-invasive portable blood pressure monitoring through deep learning technology.

[0079] In one of the embodiments of the present application, in the step of constructing the wrist-upper arm blood pressure oscillation wave data set, the conditional variational autoencoder includes an encoder and a decoder, the upper arm blood pressure oscillation wave data is input as an input variable, and the feature vector obtained after the wrist blood pressure oscillation wave data is time embedded is input as a conditional variable into the encoder, the input variable is mapped into a latent variable under the restriction of a specific conditional variable, and the decoder obtains the augmented generated upper arm paired oscillation wave data according to the latent variable and the conditional variable. The augmented generated upper arm paired oscillation wave data is paired into a group with the corresponding actually collected wrist blood pressure oscillation wave data to obtain the artificially generated wrist and upper arm paired oscillation wave data. Through the implementation of this step, more matched and more real and large amount of artificially generated wrist and upper arm paired oscillation wave data are obtained, which further provides more abundant and accurately matched wrist-upper arm blood pressure oscillation wave data set, thereby further improving the establishment of a more optimized oscillation wave mapping measurement model and more accurate measurement results.

[0080] In another embodiment of the present application, in the step of extracting the hemodynamic features of the wrist oscillation wave, the features on the oscillation wave corresponding to each pulse and the features of the oscillation wave envelope constructed by all oscillation waves are extracted.

[0081] The features on the oscillation wave corresponding to each pulse include but are not limited to the reflection peak-to-main peak amplitude ratio, the reflection peak occurrence time, and the reflection wave significance.

[0082] The features of the oscillation wave envelope include but are not limited to the maximum amplitude of the envelope and the corresponding pressure and time, the area covered by the oscillation envelope and the area ratio before and after the maximum amplitude point, and the maximum slope when the envelope rises and falls.

[0083] Through the implementation of the above method, the vascular elasticity information is obtained through the features of the oscillation wave and the oscillation wave envelope, and the pressure wave propagation speed and the reflection wave amplitude of the blood vessels under different external pressures are further derived, so that the establishment and reasoning of the oscillation wave mapping measurement relationship between the wrist blood pressure oscillation wave data and the upper arm blood pressure oscillation wave data are more reasonable.

[0084] In another embodiment of the present invention, the step of processing the PPG signal within the window, specifically the step of expanding the artificially generated wrist and upper arm paired oscillatory wave data using a conditional variational autoencoder, includes the following process:

[0085] First, the wrist blood pressure oscillation data is encoded into a feature vector using a temporal embedding tool. This feature vector is then used as a conditional variable input to the conditional variational autoencoder (CVAE) model. Based on unsupervised training, the CVAE network is trained by reconstructing the input signal through the network.

[0086] After the CVAE network is trained, the embedding vector is extracted from real wrist blood pressure oscillation data. This vector is then concatenated with randomly sampled normal distribution latent vectors and input into the decoder of the CVAE model to obtain a large amount of artificially generated upper arm paired oscillation data. The artificially obtained upper arm oscillation data and their corresponding real-collected wrist oscillation data are then paired and grouped and stored in the database.

[0087] The implementation of the above methods makes the process of expanding wrist and upper arm paired oscillatory wave data more efficient and controllable, and can create training data that matches the relevant data of multiple individual subjects, thereby further improving the accuracy of the oscillatory wave mapping measurement model.

[0088] In one embodiment of the present invention, the objective function L of the conditional variational autoencoder (CVAE) model is... CVAE for:

[0089] ;

[0090] Where x represents the input variable, which is the upper arm pressure oscillation wave signal; c represents the condition variable, which is the feature vector obtained by temporal embedding of wrist blood pressure oscillation wave data; and z represents the latent variable, which is the upper arm pressure oscillation wave signal generated from different samples obtained from a normal distribution. The encoder representing the CVAE model. The decoder representing the CVAE model. D represents the prior distribution of the CVAE model. KL It is the KL operator.

[0091] Therefore, the objective function L of the conditional variational autoencoder CVAE model is... CVAE Further confirmation ensures the reliability and interpretability of the encoded vector.

[0092] In yet another embodiment of the present application, the wrist- upper arm type oscillatory wave acquisition module is used to acquire and enhance the obtained wrist- upper arm blood pressure oscillatory wave data set, and the deviation calibration sequence obtained by subtracting the upper arm pressure oscillatory wave from the wrist oscillatory wave is used as the oscillatory wave mapping label of the wrist- upper arm type oscillatory wave mapping measurement module;

[0093] The wrist- upper arm type oscillatory wave mapping measurement module is used to perform wrist blood pressure measurement based on the oscillatory wave mapping measurement model and the pressure oscillatory amplitude mapping.

[0094] In the process of using the time sequence embedding of the wrist blood pressure oscillatory wave data as the conditional variable, the time sequence embedding used includes feeding the time sequence signal of the wrist blood pressure oscillatory wave data into the encoding layer of a trained variational autoencoder, and extracting the encoding vector as the feature vector. The variational autoencoder uses a VAE model, and the objective function L of the VAE model is: VAE

[0095]

[0096] , which represents the encoder of the VAE model, , which represents the decoder of the VAE model, , which represents the prior distribution of the VAE model.

[0097] The design uses time sequence embedding in deep learning technology to obtain more reliable and effective feature vectors.

[0098] In another embodiment of the present application, the step of putting the deviation calibration sequence generator into deep learning is performed in a forced learning manner, and the decoder uses the actually acquired upper arm blood pressure oscillatory wave data as input at each generation step.

[0099] In the step of inferring the oscillatory amplitude difference between the radial artery and the brachial artery of the measurer individual at different pressures according to the oscillatory wave mapping measurement relationship, the wrist blood pressure oscillatory wave data of the measurer is used to extract the hemodynamic feature vector of the measurer individual, and the hemodynamic feature vector is input into the oscillatory wave mapping measurement model. The oscillatory wave mapping measurement relationship between the wrist blood pressure oscillatory wave data and the upper arm blood pressure oscillatory wave data of the measurer individual is extracted by inference, and the oscillatory wave mapping measurement relationship of the measurer individual contains the oscillatory amplitude difference at different pressures. The blood pressure oscillatory wave envelope at the wrist in the wrist blood pressure oscillatory wave data is envelope reconstructed to reconstruct the blood pressure oscillatory wave envelope at the upper arm, and the pressure oscillatory amplitude mapping from the radial artery to the brachial artery is completed.

[0100] ​​​Through the above application of deep learning and different learning methods in the inference process, the efficiency and stability of the oscillatory wave mapping measurement model training are improved, and the accuracy in the inference process is improved.

[0101] Embodiment two, the blood pressure measurement intelligent wearable device applies the wrist blood pressure measurement method based on the pressure oscillation amplitude mapping as described in embodiment one of the present application, the blood pressure measurement intelligent wearable device includes a pressure acquisition module 2, a calculation module 3;

[0102] The pressure acquisition module 2 is used to actually acquire the wrist blood pressure oscillation wave data of the measurer;

[0103] The calculation module 3 is internally provided with an oscillation wave mapping measurement model, and the calculation module is used to extract the hemodynamic feature vector of the actually acquired wrist blood pressure oscillation wave data, input the oscillation wave mapping measurement model, infer the oscillation amplitude difference between the radial artery and the brachial artery of the measurer under different pressures according to the oscillation wave mapping measurement relationship, map the pressure oscillation amplitude at the radial artery to the brachial artery, automatically identify the maximum amplitude point based on the mapping obtained upper arm blood pressure oscillation wave data by using a standard oscillograph algorithm, calculate the systolic pressure and diastolic pressure, and output the blood pressure estimation result.

[0104] In the specific embodiment of embodiment two of the present application, the blood pressure measurement intelligent wearable device is a device worn on the wrist of a user, the blood pressure measurement intelligent wearable device includes a shell 1, the pressure acquisition module 2 includes a gas pump 21, a gas bag 22 and a pressure sensor 23 installed on the shell 1; a mainboard is installed in the shell 1, and the calculation module 3 is installed on the mainboard. Through the above further design of the blood pressure measurement intelligent wearable device, the structure of the wearable device specially used for wrist blood pressure measurement is optimized, and combined with the implementation of the method, more accurate and reliable blood pressure measurement can be completed without the cooperation of other sensors (such as ECG signal sensors, PPG signal sensors, etc.).

[0105] Embodiment three, the readable storage medium stores a control program, and when the control program is executed by the calculation module, the calculation module executes the steps of the method as described in embodiment one of the present application.

[0106] The composition and working principle of the present application will be described below in combination with the hardware modules, measurement implementation process and main flow of the method in the embodiments of the present application.

[0107] Figure 2As shown in the figure, the hardware module structure of the blood pressure measurement intelligent wearable device is shown, mainly including a pressure collection module and a calculation module, the pressure collection module and the calculation module can be integrated into a wearable device, or the pressure collection module and the calculation module are separated, the collected signals of the pressure collection module are transmitted to the calculation module through a wired or wireless mode, and then calculation, output / display of results are performed.

[0108] Figure 3 、 Figure 4 As shown in the figure, the structure of the blood pressure measurement intelligent wearable device is shown when the blood pressure measurement intelligent wearable device is a sphygmomanometer watch, wherein the watch includes a shell 1, a pressure collection module 2 includes a gas pump 21, a gas bag 22 and a pressure sensor 23 installed on the shell 1; a mainboard is installed in the shell 1, and the calculation module 3 is installed on the mainboard. The sphygmomanometer watch is worn on the wrist of a user, the gas pump 21 works when blood pressure measurement is needed, the gas bag 22 is inflated, and then the wrist radial artery pressure signal is collected by the pressure sensor 23.

[0109] The wrist blood pressure measurement method based on the pressure oscillation amplitude mapping in the embodiment of the application is applied to the blood pressure measurement intelligent wearable device, and the actual working process can be referred to as follows:

[0110] 1. Start measurement;

[0111] 2. Inflating the gas bag;

[0112] 2. Collecting the wrist radial artery pressure oscillation wave;

[0113] 3. Real-time extracting the amplitude, corresponding pressure and high-order characteristic value on the waveform of each pressure oscillation wave;

[0114] 4. Ending the pressurization, storing all extracted oscillation wave characteristic matrices, and starting post-processing;

[0115] 5. Constructing the original oscillation envelope (the horizontal axis is pressure, and the vertical axis is oscillation amplitude);

[0116] 6. Using a deep learning model to calculate the pressure-amplitude difference mapping relationship between the radial artery and the brachial artery according to the characteristics;

[0117] 7. Subtracting the mapping relationship from the original oscillation envelope;

[0118] 8. Obtaining the processed oscillation envelope (equivalent to the oscillation envelope at the brachial artery);

[0119] 9. The maximum amplitude corresponds to the average pressure, and the systolic pressure and diastolic pressure are obtained by using the amplitude coefficient method.

[0120] Figure 5As shown in the figure, the establishment and application framework of the deep learning driven wrist-upper arm blood pressure oscillatory wave mapping measurement model designed in the embodiment of the present application is shown, and the goal of designing the framework is to cross the shift of the wrist and upper arm type cuff pressure oscillatory wave signals in the oscillometric pressure measurement process, and to realize the alignment of the two signals; the whole framework is divided into two modules: the first module is a wrist-upper arm type oscillatory wave acquisition module (the upper half of the block diagram), and the second module is a wrist-upper arm type oscillatory wave mapping measurement module based on the wrist oscillatory wave hemodynamic characteristics (the lower half of the block diagram).

[0121] As shown in the figure, the first module (wrist-upper arm type oscillatory wave acquisition module) is applied, and the goal is to obtain a large number of pairs of wrist-upper arm type blood pressure oscillatory waves. In order to achieve this goal, on the one hand, a large number of subjects are recruited, and the age span and blood pressure distribution interval are required to be wide. For these subjects, the wrist blood pressure oscillatory wave data and the corresponding upper arm type blood pressure oscillatory wave data are collected multiple times (generally more than three times), and the pairs of oscillatory wave data are stored in the database. Figure 5

[0122] Figure 6 As shown in the figure, the schematic diagram of the CVAE model is shown, in order to expand the wrist-upper arm blood pressure oscillatory wave data set, the present application generates artificial upper arm type oscillatory wave signals by using a conditional variational autoencoder (CVAE), and the objective function L of the conditional variational autoencoder CVAE model is CVAE :

[0123] ;

[0124] Wherein, x represents an input variable, the upper arm type pressure oscillatory wave signal is taken as the input variable x, c represents a conditional variable, the feature vector obtained by time series embedding of the wrist blood pressure oscillatory wave data is taken as the conditional variable c, z represents a hidden variable, the upper arm type pressure oscillatory wave signal generated by different sampling from the normal distribution sampling is taken as the hidden variable z, represents the encoder of the CVAE model, represents the decoder of the CVAE model, represents the prior distribution of the CVAE model, and D KL is a KL operator;

[0125] In the process of taking the feature vector obtained by time series embedding of the wrist blood pressure oscillatory wave data as the conditional variable, the time series embedding includes: feeding the time series signal of the wrist blood pressure oscillatory wave data into an encoding layer of a variational autoencoder which has been trained, and extracting the encoding vector as the feature vector, the variational autoencoder has a VAE model, and the objective function L of the VAE model is VAE :​

[0126] ;

[0127] wherein, represents an encoder of the VAE model, represents a decoder of the VAE model, represents a prior distribution of the VAE model.

[0128] Specifically, first, the wrist oscillatory wave signal is encoded into a feature vector by using a time series embedding tool. The feature vector is input into the CVAE architecture as a condition. The input and output of the architecture are both upper arm oscillatory wave signals. Through the unsupervised training mode, the training of the CVAE network is completed by making the network reconstruct the input signal. After the CVAE network is trained, the embedding vector of the real wrist oscillatory wave is extracted, the vector and the randomly sampled normal distribution hidden vector are spliced, and input into the decoder of the CVAE to obtain a large number of artificially generated upper arm oscillatory wave signals. The artificially obtained upper arm oscillatory wave and the corresponding real collected wrist oscillatory wave are paired into groups and stored in the database.

[0129] The final wrist-upper arm blood pressure oscillatory wave dataset contains: real collected wrist blood pressure oscillatory wave data, corresponding upper arm blood pressure oscillatory wave data, and artificially generated upper arm oscillatory wave data.

[0130] Generally, in the implementation process of the wrist-upper arm oscillatory wave collection module, the total amount of matched data in the generated wrist-upper arm blood pressure oscillatory wave dataset is not less than 3000 people (about 9000), and the actual collection and generation can each account for half, so that the final blood pressure distribution meets the normal distribution, and the blood pressure distribution can be from 90mmHg, every 20mmHg interval: less than 90, 90-110……, >170. As the amount of oscillatory wave data of the upper arm paired oscillatory wave data that needs to be collected and artificially generated is not limited by the present application, the total amount of data can be more than 100, more than 500, and more than 1000. The blood pressure distribution method is not limited to every 20mmHg interval, and can be more subdivided or more coarse.

[0131] In the case of Figure 5The first step of the application of the second module (wrist-to-upper arm oscillatory wave mapping module) is as follows. First, using the wrist-to-upper arm oscillatory wave data set in the first module, the wrist oscillatory wave minus the upper arm pressure oscillatory wave is extracted to obtain a deviation calibration sequence as the oscillatory wave mapping label to be learned by module two. Second, the hemodynamic features of the wrist oscillatory wave are extracted, including the features on the oscillatory wave corresponding to each pulse, such as the reflected peak-to-main peak amplitude ratio, the reflected peak occurrence time, the reflected wave prominence, and the like; and the features of the oscillatory wave envelope, such as the maximum amplitude of the envelope and the corresponding pressure and time, the area covered by the oscillatory envelope and the area ratio before and after the maximum amplitude point, the maximum slope when the envelope rises and falls, and the like.

[0132] wherein, Figure 7 An oscillatory wave waveform diagram is shown for generating an envelope curve through feature extraction, Figure 7 (a) is a time series oscillatory wave after detrending, and the peak point, trough point, and reflected wave point (such as Figure 7 (b)) on each oscillation are extracted. Figure 7 (c) is the peak-to-peak value of each wave after the peak-to-trough extraction, which constitutes the upper and lower envelope lines ( Figure 7 (c)), and the amplitude envelope formed only by the peak-to-peak value is Figure 7 (c) in the upper envelope-lower envelope. Figure 7 (d) is the amplitude scatter plot, and the envelope curve of (d) is obtained after spline interpolation.

[0133] Figure 8 A feature point extraction schematic diagram is shown, and the features extracted on the time series oscillatory wave waveform are as follows:

[0134] The features extracted on the time series oscillatory wave waveform are as follows:

[0135] 1. The peak and reflected peak occurrence time sequence: , ;

[0136] 2. The reflected peak-to-main peak amplitude ratio sequence ratio i , ;

[0137] 3. The reflected wave prominence sequence prominence i ,

[0138] (if amp i =0, prominence=0).

[0139] Figure 9 A wrist oscillatory amplitude envelope curve schematic diagram is shown, and the features extracted on the oscillatory wave envelope are as follows:

[0140] 1. The maximum amplitude of the envelope maxamp, and its corresponding pressure p maxamp and time t maxamp ;

[0141] 2. The area covered by the oscillatory envelope (the area of the oscillatory wave envelope (red line) between t start and t end ), the areas S right , S left before and after the maximum amplitude point, and the area ratio S right / S left ;

[0142] 3. The maximum slope of the envelope when rising and falling maxLiftSlope, maxDropSlope (both absolute values).

[0143] Through the implementation of the above embodiments, the shift of the wrist and upper arm type airbag pressure oscillatory wave signal exhibited in the oscillometric pressure measurement process is realized, the alignment of the two signals is achieved, the oscillatory wave collected at the wrist is mapped to the upper arm oscillatory wave, so that the wrist blood pressure measurement intelligent wearable device is convenient to achieve the clinical recognized accuracy close to the upper arm type blood pressure measurement device, the result of the blood pressure measurement of the patient using the blood pressure measurement intelligent wearable device is more clinically valuable, and the purpose of the present application is achieved.

[0144] In the application of the second module (wrist-upper arm type oscillatory wave mapping measurement module) as shown in Figure 5 , then these features are spliced into a hemodynamic feature vector, and thereafter put into a bias calibration sequence generator based on a recurrent neural network (RNN) as input. The bias calibration sequence generator is not limited to a specific RNN module, and long short-term memory network (LSMT), gated neural unit (GRU), etc. can be used. Next, by using the method of teacher forcing, the bias calibration sequence is used as a label for point-by-point regression learning. In the inference stage, according to a certain wrist blood pressure oscillatory wave data, the hemodynamic feature vector is extracted therefrom, and input into the bias calibration sequence generator, so as to generate a bias calibration sequence point by point. The wrist pressure oscillatory wave sequence is subtracted by the bias calibration sequence to obtain the final upper arm oscillatory wave sequence.

[0145] Finally, the maximum amplitude point (MAP) is automatically identified based on the mapping obtained oscillatory wave signal of the upper arm and systolic blood pressure (SBP) and diastolic blood pressure (DBP) are calculated using the standard oscillometric algorithm, and the accurate blood pressure estimation result is output. This method only uses the characteristics of the display blood vessels and physiological characteristics in the individual oscillatory wave and predicts the deviation of the individual wrist and upper arm oscillatory wave, not only solves the gap problem between the wrist oscillometric pressure oscillation signal and the upper arm oscillatory pressure oscillation signal, but also significantly improves the accuracy and practicability of non-invasive portable blood pressure monitoring through deep learning technology.

[0146] Based on the mapping relationship extracted by the deep learning architecture described above, the oscillatory amplitude difference under different pressures is directly subtracted from the oscillatory amplitude collected at the wrist, that is, the oscillatory amplitude at the radial artery is mapped to the brachial artery, and the measurement value at the brachial artery is directly obtained by the amplitude coefficient method in the oscillometric method principle.

[0147] As shown in Figures 10 to 13 , 4 subjects are shown as measurers, and the application of the wrist blood pressure measurement method based on pressure oscillation amplitude mapping of the embodiment of the application is adopted. The blue curve is the wrist oscillatory wave envelope (the original envelope collected by the watch), the red curve is the amplitude mapping of the wrist_upper arm oscillatory wave calculated by the model, and the yellow curve is obtained after the blue curve is subtracted from the red curve, that is, the constructed oscillatory wave envelope at the upper arm. Figures 10 to 13 4 subjects under different blood pressure distributions can use this method for envelope reconstruction.

[0148] Through the implementation of the above embodiments, the shift of the wrist and upper arm type cuff pressure oscillation signal in the oscillometric pressure measurement process is crossed, the alignment of the two signals is realized, the oscillatory wave collected at the wrist is mapped to the upper arm oscillatory wave, so that the convenient wrist blood pressure measurement intelligent wearable device can achieve the clinical recognized accuracy close to the upper arm type blood pressure measurement device, the result of the blood pressure measurement by the patient using the blood pressure measurement intelligent wearable device has more clinical reference value, and the purpose of the present application is achieved.

[0149] The above embodiments are only for illustrating the technical concept and characteristics of the present application, the purpose is to enable those skilled in the art to understand the content of the present application and implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made according to the spirit and essence of the present application shall be covered within the protection scope of the present application.

Claims

1. A wrist-based blood pressure measurement method based on pressure oscillation amplitude mapping, characterized in that, The wrist blood pressure measurement method includes: A wrist-upper arm blood pressure oscillation dataset was constructed. Multiple real-world data collections were conducted on the wrist blood pressure oscillation data at the radial artery and the paired upper arm blood pressure oscillation data at the brachial artery. Artificially generated wrist and upper arm paired oscillation data were obtained by expanding the dataset using a conditional variational autoencoder. The real-world data collections and the artificially generated wrist and upper arm paired oscillation data together constituted the wrist-upper arm blood pressure oscillation dataset. An oscillation wave mapping measurement model was established. The difference sequence between the wrist oscillation wave and the upper arm pressure oscillation wave was extracted from the wrist-upper arm blood pressure oscillation wave dataset as a deviation label. The hemodynamic features of the wrist oscillation wave were extracted and concatenated into a hemodynamic feature vector. This vector was then fed into a deviation calibration sequence generator for deep learning, and a deviation calibration sequence was generated point by point to obtain the oscillation wave mapping measurement relationship between the wrist blood pressure oscillation wave data and the upper arm blood pressure oscillation wave data. Based on this, an oscillation wave mapping measurement model was established. Wrist blood pressure measurement is performed based on pressure oscillation amplitude mapping. An air bladder is inflated and pressurized to collect wrist blood pressure oscillation data of the measurer. Hemodynamic feature vectors are extracted from the current measurer's wrist blood pressure oscillation data and input into the oscillation wave mapping measurement model. Based on the oscillation wave mapping measurement relationship, the difference in oscillation amplitude between the radial artery and brachial artery of the individual measurer under different pressures is inferred. The pressure oscillation amplitude at the radial artery is mapped to the brachial artery. Using a standard oscilloscope algorithm, the maximum amplitude point is automatically identified based on the mapped upper arm blood pressure oscillation data, and systolic and diastolic blood pressure are calculated, outputting the blood pressure estimation result.

2. The wrist blood pressure measurement method based on pressure oscillation amplitude mapping according to claim 1, characterized in that: In the step of constructing the wrist-upper arm blood pressure oscillation dataset, the conditional variational autoencoder includes an encoder and a decoder. The upper arm blood pressure oscillation data is used as the input variable, and the feature vector obtained by temporally embedding the wrist blood pressure oscillation data is used as the condition variable and input to the encoder. The input variable is mapped to the latent variable under the constraint of the condition variable. The decoder obtains the expanded upper arm paired oscillation data according to the latent variable and the condition variable. The expanded upper arm paired oscillation data is paired with its corresponding real-collected wrist blood pressure oscillation data to form a group, thus obtaining the artificially generated wrist and upper arm paired oscillation data.

3. The wrist blood pressure measurement method based on pressure oscillation amplitude mapping according to claim 1, characterized in that: The step of extracting the hemodynamic features of wrist oscillation waves includes extracting features on the oscillation wave corresponding to each pulse and features of the oscillation wave envelope constructed from all oscillation waves. The characteristics of the oscillatory wave corresponding to each pulse include, but are not limited to, the ratio of the amplitude of the reflected wave peak to the main peak, the time of the appearance of the reflected wave peak, and the significance of the reflected wave. The characteristics of the oscillating wave envelope include, but are not limited to, the maximum amplitude and its corresponding pressure and time, the area covered by the oscillating envelope and the area ratio before and after the maximum amplitude point, and the maximum slope of the envelope during its rise and fall.

4. The wrist blood pressure measurement method based on pressure oscillation amplitude mapping according to claim 1, characterized in that, The steps for obtaining artificially generated paired wrist and upper arm oscillatory wave data by augmenting with a conditional variational autoencoder include the following: First, the wrist blood pressure oscillation data is encoded into a feature vector using a temporal embedding tool. This feature vector is then used as a conditional variable input to the conditional variational autoencoder (CVAE) model. Based on unsupervised training, the CVAE network is trained by reconstructing the input signal through the network. After the CVAE network is trained, the embedding vector is extracted from real wrist blood pressure oscillation data. This vector is then concatenated with randomly sampled normal distribution latent vectors and input into the decoder of the CVAE model to obtain a large amount of artificially generated upper arm paired oscillation data. The artificially obtained upper arm oscillation data and their corresponding real-collected wrist oscillation data are then paired and grouped and stored in the database.

5. The wrist blood pressure measurement method based on pressure oscillation amplitude mapping according to claim 4, characterized in that: The objective function L of the conditional variational autoencoder (CVAE) model CVAE for: ； Where x represents the input variable, which is the upper arm pressure oscillation wave signal; c represents the condition variable, which is the feature vector obtained by temporal embedding of wrist blood pressure oscillation wave data; and z represents the latent variable, which is the upper arm pressure oscillation wave signal generated from different samples obtained from a normal distribution. The encoder representing the CVAE model. The decoder representing the CVAE model. D represents the prior distribution of the CVAE model. KL It is the KL operator.

6. The wrist blood pressure measurement method based on pressure oscillation amplitude mapping according to claim 5, characterized in that: The wrist-upper arm oscillatory wave dataset was acquired and enhanced using a wrist-upper arm oscillatory wave acquisition module. The deviation calibration sequence obtained by subtracting the upper arm pressure oscillatory wave from the wrist oscillatory wave was used as the oscillatory wave mapping label of the wrist-upper arm oscillatory wave mapping measurement module. Wrist blood pressure measurement based on oscillation wave mapping measurement model and pressure oscillation amplitude mapping was performed using a wrist-upper arm oscillation wave mapping measurement module. In the process of using the feature vector obtained by temporal embedding of wrist blood pressure oscillation data as a condition variable, the temporal embedding method includes: feeding the temporal signal of the wrist blood pressure oscillation data into the encoding layer of a pre-trained variational autoencoder (VAE), extracting the encoded vector as the feature vector, and the VAE model having an objective function L. VAE for: ； in, This represents the encoder of the VAE model. This represents the decoder of the VAE model. This represents the prior distribution of the VAE model.

7. The wrist blood pressure measurement method based on pressure oscillation amplitude mapping according to claim 6, characterized in that: In the step of deep learning in the deviation calibration sequence generator, a forced learning method is adopted, and the decoder uses real upper arm blood pressure oscillation data as input when generating each step; In the step of inferring the difference in oscillation amplitude between the radial and brachial arteries of an individual under different pressures based on the oscillation wave mapping measurement relationship, a non-forced learning method is adopted. The hemodynamic feature vector of the individual is extracted by collecting wrist blood pressure oscillation wave data. After the hemodynamic feature vector is input into the oscillation wave mapping measurement model, the oscillation wave mapping measurement relationship between the wrist blood pressure oscillation wave data and the upper arm blood pressure oscillation wave data of the individual is extracted through inference. The oscillation wave mapping measurement relationship of the individual includes the difference in oscillation amplitude under different pressures. The blood pressure oscillation wave envelope at the wrist in the wrist blood pressure oscillation wave data is reconstructed to reconstruct the blood pressure oscillation wave envelope at the upper arm, thus completing the mapping of the pressure oscillation amplitude at the radial artery to the brachial artery.

8. A smart wearable device for measuring blood pressure, characterized in that, The intelligent wearable device for blood pressure measurement uses the wrist-based blood pressure measurement method based on pressure oscillation amplitude mapping as described in any one of claims 1 to 7, and the intelligent wearable device for blood pressure measurement includes a pressure acquisition module and a calculation module; The pressure acquisition module is used to actually collect wrist blood pressure oscillation data from the person being measured. The calculation module has a built-in oscillation wave mapping measurement model. The calculation module is used to extract hemodynamic feature vectors from the actual collected wrist blood pressure oscillation wave data, input them into the oscillation wave mapping measurement model, and infer the difference in oscillation amplitude between the radial artery and the brachial artery of the individual under different pressures based on the oscillation wave mapping measurement relationship. The pressure oscillation amplitude at the radial artery is mapped to the brachial artery. Using a standard oscilloscope algorithm, the maximum amplitude point is automatically identified based on the mapped upper arm blood pressure oscillation wave data, and systolic and diastolic blood pressure are calculated, and the blood pressure estimation result is output.

9. The intelligent wearable device for measuring blood pressure according to claim 8, characterized in that, The blood pressure measurement smart wearable device is a device worn on the user's wrist. The blood pressure measurement smart wearable device includes a housing, and the pressure acquisition module includes an air pump, an air bag, and a pressure sensor installed on the housing; a motherboard is installed inside the housing, and the calculation module is installed on the motherboard.

10. A readable storage medium, characterized in that: The readable storage medium stores a control program, which, when executed by the computing module, causes the computing module to perform the steps of the method as described in any one of claims 1 to 7.

Images