A PPG baseline calibration method for wearable continuous blood pressure monitoring devices

Abstract

The application provides a PPG baseline calibration method for a wearable continuous blood pressure monitoring device, which comprises the following steps: decoupling the relationship between contact pressure and infrared PPG baseline; since pressure is a scalar, the contact pressure change of the measured position can be equivalent to the opposite change of the hydrostatic pressure inside the artery; collecting infrared PPG waveforms under different hydrostatic pressures through two different height postures to model the relationship between contact pressure and PPG waveform, MLP modeling and the setting of a dual-wavelength calibrator; without affecting the extraction of blood pressure related features; enabling the existing blood pressure monitoring algorithm to achieve the A-class standard of BHS in long-term continuous monitoring; the method can be directly deployed in the existing blood pressure monitoring system without modifying the original blood pressure algorithm or model structure; the problem that PPG baseline drift occurs due to long-term wearing of the wearable device and thus continuous and accurate blood pressure monitoring cannot be achieved is solved; and more accurate continuous blood pressure monitoring can be achieved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of non-invasive continuous blood pressure monitoring, and particularly relates to a PPG baseline calibration method for a wearable continuous blood pressure monitoring device. BACKGROUND

[0002] As a commonly used physiological sensor, a photoplethysmography (PPG) sensor can use a multi-wavelength LED to monitor physiological signals such as heart rate, blood oxygen, and blood pressure; however, the PPG sensor is sensitive to contact pressure, that is, as the wearing time increases, the pressure between the sensor and the skin will change due to local slight swelling; although such swelling does not affect the usability, it will cause a significant drift of the baseline of the PPG (i.e., the direct current component of the PPG); the baseline of the PPG is highly related to blood pressure, and such drift will seriously affect the continuous blood pressure monitoring performance of the wearable device.

[0003] In order to overcome the above-mentioned deficiencies, the prior art improves the overall working scheme of the PPG through some designs, for example:

[0004] ① Some blood pressure monitoring works use filtering methods to eliminate short-term baseline drift in PPG data, such as using wavelet transform to eliminate short-term baseline drift in PPG data or using least square fitting baseline to suppress short-term baseline drift.

[0005] This method cannot solve the problem of long-term wearing baseline drift, and will affect the PPG morphological features, thereby affecting the extraction of blood pressure related features.

[0006] ② Some blood pressure monitoring works adjust the tightness of the wristband to maintain the optimal contact pressure between the PPG sensor and the skin.

[0007] This method is not suitable for all wearable devices, such as smart rings, and manual adjustment has individual difference errors, making the result more inaccurate. SUMMARY

[0008] In order to solve the above-mentioned technical problems, the application provides a PPG baseline calibration method for a wearable continuous blood pressure monitoring device, which solves the problem of PPG baseline drift caused by long-term wearing of the wearable device, can be applied to wearable devices with PPG sensors, and realizes non-invasive, continuous, and comfortable blood pressure monitoring.

[0009] A PPG baseline calibration method for a wearable continuous blood pressure monitoring device, comprising:

[0010] Step one, decoupling the relationship between contact pressure and infrared PPG baseline;

[0011] Since pressure is a scalar, the change of contact pressure of the measured position can be equivalent to the opposite change of hydrostatic pressure inside the artery; therefore, the infrared PPG signals under different hydrostatic pressures are collected by two different height postures to model the relationship between contact pressure and PPG baseline;

[0012] However, the fluctuation of infrared PPG baseline is also related to arterial blood pressure, so we need to decouple the influence of contact pressure on infrared PPG baseline;

[0013] As an example, the two different height postures are: the finger is placed at the level of the heart and below the heart.

[0014] As an example, the position below the heart is preferably the position at the level of the navel.

[0015] Through a large number of experiments, it is found that the green light PPG baseline is almost only affected by the contact pressure; because the green light PPG can only reach the shallow capillary blood vessels of the skin, the green light PPG baseline only contains the blood pressure information of the capillary blood vessels; and the capillary blood pressure does not fluctuate with the fluctuation of arterial pressure, which means that the green light PPG baseline does not change due to the change of arterial pressure, that is, the influence of contact pressure on infrared PPG baseline can be decoupled through the change of green light PPG baseline;

[0016] Step two, MLP modeling;

[0017] A dual-wavelength calibrator is constructed by using the multilayer perception (MLP) in machine learning to model the influence of contact pressure on the green light PPG baseline and the infrared PPG baseline, so as to eliminate the influence of contact pressure on the infrared PPG baseline;

[0018] Step three, setting of the dual-wavelength baseline calibrator;

[0019] The dual-wavelength baseline calibrator includes a 1-layer MLP for input modeling and a 3-layer MLP for calibration, which is designed as (formula 1):

[0020]

[0021] Wherein: represents the calibrated baseline value of PPG infrared light;

[0022] The model input includes static information and dynamic information;

[0023] The static information includes the baseline and amplitude of dual-wavelength PPG at different heights: And ;

[0024] The dynamic information includes the baseline and amplitude of the current PPG: and ;

[0025] During model training, the measured diastolic pressure is brought into the blood pressure algorithm (formula 2) by inverse regression, that is, the calibrated infrared PPG direct current component true value is obtained for supervised learning;

[0026]

[0027] Wherein: and DC and diastolic time T are personalized proportion coefficients of infrared PPG direct current component, and DBP and PP are personalized bias coefficients of diastolic pressure (DBP) and pulse pressure (PP), which can be solved by 2 blood pressure measurements;

[0028] The beneficial effects of the present application are:

[0029] ① Without changing the PPG original waveform, that is, without affecting the extraction of blood pressure related features;

[0030] ② The existing blood pressure monitoring algorithm can achieve the A class standard of BHS (British Hypertension Society standard) in long-term continuous monitoring; in addition, the method can be directly deployed in the existing blood pressure monitoring system without modifying the original blood pressure algorithm or model structure.

[0031] ③ The problem of PPG baseline drift caused by long-term wearing of wearable devices, which cannot continuously and accurately monitor blood pressure, is solved.

[0032] ④ The present application is suitable for various wearable devices and can realize more accurate continuous blood pressure monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 It is the principle design drawing of the PPG baseline calibration method for wearable continuous blood pressure monitoring device of the present application. DETAILED DESCRIPTION

[0034] Below, with reference to Figure 1 , a PPG baseline calibration method for wearable continuous blood pressure monitoring device, comprising:

[0035] Step one, decoupling the relationship between contact pressure and infrared PPG baseline;

[0036] Since pressure is a scalar, the change of contact pressure at the measured position can be equivalent to the opposite change of hydrostatic pressure inside the artery; therefore, infrared PPG signals under different hydrostatic pressures are collected through two different height postures to model the relationship between contact pressure and PPG baseline;

[0037] However, the fluctuation of the infrared PPG baseline is also related to the arterial blood pressure, so we need to decouple the effect of contact pressure on the infrared PPG baseline.

[0038] As an example, the two different height postures are: the finger is placed at the level of the heart and below the heart.

[0039] As an example, the position below the heart is preferably the position at the level of the navel.

[0040] Through a large number of experiments, it is found that the green light PPG is almost only affected by the contact pressure; because the green light PPG can only reach the shallow layer of the skin capillary, the green light PPG only contains the blood pressure information of the capillary; and the capillary blood pressure does not fluctuate with the fluctuation of the arterial pressure, which means that the green light PPG baseline does not change with the change of the arterial pressure, that is, the effect of the contact pressure on the infrared PPG baseline can be decoupled through the change of the green light PPG baseline.

[0041] Step two, MLP modeling;

[0042] A double-wavelength calibrator is constructed by using the multilayer perception (MLP) in machine learning, which is used to model the effect of contact pressure on the green light PPG baseline and the infrared PPG baseline, so as to eliminate the effect of contact pressure on the infrared PPG baseline.

[0043] Step three, setting of the double-wavelength baseline calibrator;

[0044] The double-wavelength baseline calibrator includes a 1-layer MLP for input modeling and a 3-layer MLP for calibration, which is designed as (formula 1):

[0045]

[0046] Wherein: represents the calibrated baseline value of the PPG infrared light;

[0047] The model input includes static information and dynamic information;

[0048] The static information includes the baseline and amplitude of the double-wavelength PPG at different heights: and ;

[0049] The dynamic information includes the baseline and amplitude of the current PPG: and ;

[0050] During model training, the measured diastolic pressure is brought into the blood pressure algorithm (formula 2) by inverse regression, so as to obtain the calibrated infrared PPG direct current component true value for supervised learning.

[0051]

[0052] wherein: and are personalized proportional coefficients of the infrared PPG direct current component DC and the diastolic time T, respectively, and are personalized bias coefficients of the diastolic blood pressure (DBP) and the pulse pressure (PP), respectively, which can be solved by 2 blood pressure measurements;

[0053] In order to better illustrate the design principle of the present application, the principle features of the present application will be briefly introduced in combination with the drawings in the description Figure 1 Brief introduction, again comb the principle features of the present application:

[0054] Example 1:

[0055] Firstly, the user will be asked to do two simple gestures after wearing the wearable device, that is, to raise and lower the hands to two different heights;

[0056] At this time, the PPG sensor in the ring will alternately emit green light and infrared light, and collect PPG data under different wavelengths through the light sensor;

[0057] Secondly, static information of green light and infrared light PPG data at different heights is extracted from the PPG data of the two simple gestures;

[0058] In daily blood pressure monitoring, the PPG sensor in the ring will collect offset green light and infrared light PPG data (dynamic information) due to the influence of contact pressure.

[0059] Finally, in order to obtain calibrated infrared PPG data, the static information obtained by wearing for the first time is input into the trained dual-wavelength PPG baseline calibrator together with real-time dynamic information.

[0060] Example 2:

[0061] Comparison results of the present application and prior art in measuring blood pressure:

[0062] Table 1: Comparison of experimental results (AAMI standard)

[0063]

[0064] Table 2: Comparison of experimental results (BHS standard)

[0065]

[0066] Wherein: AAMI is the international standard for evaluating electronic sphygmomanometer issued by American Medical Instrument Promotion Association, and BHS is the standard of British Hypertension Society;

[0067] ME is the mean error (formula 3), STD is the standard deviation (formula 4), and BP algorithm is formula 5.

[0068]

[0069]

[0070] Wherein: is the algorithm estimated blood pressure, is the true measured blood pressure, is the sample number.

[0071] Table 1 and Table 2 compare the ME and STD of the blood pressure algorithm without using the dual-wavelength baseline calibrator and the blood pressure algorithm after using the dual-wavelength baseline calibrator; it can be observed that, without using the dual-wavelength baseline calibrator, the ME and STD are obviously increased; although the prior art without using the dual-wavelength baseline calibrator, the mean error ME and the standard deviation STD meet the AAMI standard, but its SBP and DBP do not meet the BHS A-class standard.

[0072] The test results prove the effectiveness of the dual-wavelength baseline calibrator of the application.

[0073] In order to better illustrate the architecture scheme of the application, the following is briefly described through knowledge introduction:

[0074] ① Characteristics of green PPG and infrared PPG: PPGs of different wavelengths have different penetration depths, thus containing different blood pressure information;

[0075] Specifically, green PPG (typical wavelength 540 nm) can only reach the superficial capillary vessels of the skin, while infrared PPG (typical wavelength 940 nm) can penetrate deep into the arteries; therefore, green PPG is commonly used to monitor capillary blood pressure, while infrared PPG is commonly used to monitor arterial blood pressure.

[0076] ② PPG baseline and diastolic blood pressure (DBP): the Frank-Starling Law indicates that the cardiac output (representing the blood volume in the arteries) increases with the increase of diastolic blood pressure;

[0077] This indicates that the baseline of infrared PPG, which reflects the end-diastolic arterial blood volume, will also rise with the increase of diastolic blood pressure. Therefore, the baseline of infrared PPG can be used to estimate diastolic blood pressure.

[0078] ③ Hydrostatic pressure: the flow of blood is affected by gravity, thus producing a hydrostatic pressure effect, i.e. the actual pressure of the artery at the measured position is the arterial pressure + hydrostatic pressure;

[0079] Hydrostatic pressure In relation to the relative height of the measured position from the heart, it can be expressed as: wherein is the height offset of the measurement site relative to the heart, represents the density of the blood, represents the acceleration of gravity. Generally, a height offset of 1 centimeter will result in a change in arterial pressure of about 0.775 mmHg.

[0080] The present application does not change the PPG original waveform, i.e. does not affect the extraction of blood pressure related features; can make the existing blood pressure monitoring algorithm reach the A class standard of BHS (British Hypertension Society standard) in long-term continuous monitoring; in addition, the method can be directly deployed in the existing blood pressure monitoring system without modifying the original blood pressure algorithm or model structure; solves the problem that PPG baseline drift caused by long-term wearing of wearable devices, so that blood pressure cannot be continuously and accurately monitored; the present application is suitable for various wearable devices, and more accurate continuous blood pressure monitoring can be achieved.

[0081] The above only describes the preferred embodiments of the present application, and it should be understood that the above description of the embodiments is only used to help understand the method of the present application and its core idea, and does not limit the protection scope of the present application, and any modification, equivalent replacement, etc. within the idea and principle of the present application should be included in the protection scope of the present application.

Claims

1. A PPG baseline calibration method for wearable continuous blood pressure monitoring devices, characterized in that, Comprise: Step one, decoupling the relationship between contact pressure and infrared PPG baseline; Infrared PPG signals under different hydrostatic pressures are collected through two different height postures to model the relationship between contact pressure and PPG baseline; Through the change of green light PPG baseline, the influence of contact pressure on infrared PPG baseline is decoupled; Step two, MLP modeling; A dual-wavelength calibrator is built by using multi-layer perception mechanism in machine learning to model the influence of contact pressure on green light PPG baseline and infrared PPG baseline, so as to eliminate the influence of contact pressure on infrared PPG baseline; Step three, setting of dual-wavelength baseline calibrator; The dual-wavelength baseline calibrator comprises a 1-layer MLP for input modeling and a 3-layer MLP for calibration, which is designed as: ， wherein: represents the calibrated baseline value of the PPG infrared light; The model input contains static information and dynamic information; The static information includes: baseline and amplitude of dual-wavelength PPG at different heights: and ; The dynamic information includes: baseline and amplitude of current PPG: and ; During model training, the measured diastolic pressure is brought into the blood pressure algorithm formula 2 by inverse regression, so as to obtain the calibrated infrared PPG direct current component true value for supervised learning; ， wherein: and are personalized proportionality coefficients for the infrared PPG direct current component DC and the diastolic time T, respectively, and are personalized bias coefficients for the diastolic blood pressure and the pulse pressure, respectively, which can be solved from 2 blood pressure measurements.

2. The PPG baseline calibration method for wearable continuous blood pressure monitoring device according to claim 1, wherein, The two different height postures are: the finger is placed at the level of the heart and below the heart.

3. The PPG baseline calibration method for wearable continuous blood pressure monitoring device of claim 2, wherein, The position below the heart is the position at the level of the navel.

4. The PPG baseline calibration method for wearable continuous blood pressure monitoring device of claim 1, wherein, Green light PPG is almost only affected by contact pressure; because green light PPG can only reach the superficial capillary vessels of the skin, green light PPG only contains the blood pressure information of capillary vessels; and capillary blood pressure does not fluctuate with the fluctuation of arterial pressure, which means that the green light PPG baseline does not change with the change of arterial pressure.

Images