CN116981398A

CN116981398A - Blood pressure prediction method, blood pressure prediction device, and computer program

Info

Publication number: CN116981398A
Application number: CN202080107676.4A
Authority: CN
Inventors: 佐藤正平
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2020-12-04
Filing date: 2020-12-04
Publication date: 2023-10-31
Also published as: WO2022116160A1

Abstract

A blood pressure prediction method, apparatus and computer program are provided that can improve the accuracy of plethysmogram-based blood pressure measurements. The method comprises the following steps: measuring a pulse waveform of the volume pulse wave; converting the measured pulse waveform using an adjustable filter; calculating at least one index for the converted pulse waveform, the at least one index indicating a characteristic of a pressure pulse wave; evaluating a correlation between the calculated index and the corresponding blood pressure to adjust the tunable filter; outputting the converted pulse waveform filtered using the tunable filter; and predicting blood pressure based on the output pulse waveform.

Description

Blood pressure prediction method, blood pressure prediction device, and computer program

Technical Field

The present invention relates to a blood pressure prediction method, device and computer program, in particular to a sleeveless blood pressure measurement based on pulse wave analysis (Pulse Wave Analysis, PWA).

Background

Blood pressure is measured continuously to monitor blood pressure fluctuations on the day or every day, helping to grasp the risk of cardiovascular disease and chronic kidney disease. Blood pressure has been measured in the past, typically using dynamic blood pressure monitoring (Ambulatory Blood Pressure Monitoring, ABPM) by analyzing oscillations in cuff pressure measurements. However, this approach requires that the cuff be attached to the user, which is inconvenient in many cases, making the user experience poor.

Therefore, a method of measuring blood pressure using a sleeveless band has attracted attention. Wave pulse wave analysis (Pulse Wave Analysis, PWA) based on volume pulse waves is a known method of sleeveless blood pressure measurement. The PWA-based sleeveless blood pressure measurement has the following advantages: the measuring device is easy to configure and can be implemented and applied more flexibly using a single Photoplethysmography (PPG) sensor. For example, PWA-based cuff-less blood pressure measurements may be easily implemented in wearable terminals such as smartwatches, headphones, smartphones, pulse oximeters, and the like.

The PWA-based sleeveless blood pressure measurement exploits the relationship between blood pressure and pulse waveform shape changes caused by differences in pulse wave reflection at the branch from the aorta to the femoral artery, arterial compliance, and vascular resistance. In practice, such measurements use waveforms obtained by pulse wave decomposition, pulse waveforms and their differential or integral waveforms. From these waveforms, characteristic points such as peak and zero-crossing points, corresponding wave heights or integral values of the waveforms, root Mean Square (RMS), and the like are extracted as features. In addition, the amplitude and the section average value may be extracted as features. The level of blood pressure is then determined by regression or classifier based on the features related to the blood pressure altitude.

In the early stages of PWA studies, pulse wave analysis of aortic pressure waveforms using invasive blood pressure monitors was directly applied to PPG. However, since the volume pulse wave is different from the pressure waveform in characteristics, the principle of calculating the blood pressure from the pressure waveform in characteristics cannot be directly applied to the volume pulse wave. For example, parameters such as the ratio of the wave height to the integral of the pressure waveform should have a very high correlation with blood pressure. On the other hand, if a similar index is calculated using a plethysmogram, the correlation with blood pressure is not high.

For this reason, a method of predicting blood pressure from a waveform of PPG using machine learning is being performed. However, such machine learning requires high computational complexity.

Disclosure of Invention

The present invention aims to provide a blood pressure prediction method, a blood pressure prediction device and a computer program that can improve the accuracy of plethysmogram-based blood pressure measurement by a simple, non-invasive technique.

In a first aspect, there is provided a method of predicting blood pressure, the method comprising: measuring a pulse waveform of the volume pulse wave; converting the measured pulse waveform using an adjustable filter; calculating at least one index for the converted pulse waveform, the at least one index indicating a characteristic of a pressure pulse wave; evaluating a correlation between the calculated index and the corresponding blood pressure to adjust the tunable filter; outputting the converted pulse waveform filtered using the tunable filter; and predicting blood pressure based on the output pulse waveform.

According to this implementation, the measured pulse waveform is converted using an adjustable filter such that the measured pulse waveform approximates a pressure waveform. In addition, when blood pressure prediction is applied to a group close to "waveforms and blood pressure data sets prepared in advance", a filter adjusted or optimized for the data sets may be used. Thus, the accuracy of the plethysmogram-based blood pressure measurement is improved.

With reference to one possible implementation manner of the first aspect, the method further includes: calculating the at least one index finger for the measured pulse waveform, wherein the evaluating step comprises: the correlation between the index calculated before and after the conversion step and the corresponding blood pressure is evaluated.

According to this implementation, the tunable filter may be adjusted by comparing the correlation of the indicator after the filtering and the blood pressure with the correlation of the indicator before the filtering and the blood pressure.

With reference to one possible implementation manner of the first aspect, the feature is selected from a group consisting of a peak product ratio, a systolic area and an enhancement index of the pulse waveform.

According to this implementation, the above features may be used to evaluate the correlation of the converted waveform with blood pressure.

With reference to one possible implementation manner of the first aspect, the method further includes: an indicator is selected from the at least one indicator based on the evaluation, wherein the predicting step predicts blood pressure based on the output pulse waveform and the selected indicator.

According to this implementation, when blood pressure prediction is applied to a group close to the "pre-prepared waveform and blood pressure dataset", a filter adjusted or optimized for the dataset may be used.

With reference to one possible implementation manner of the first aspect, the tunable filter includes at least one of the following: a variable roll-off filter, an equalization filter with adjustable Q value, or a variable phase shifter.

According to such an implementation, the above-described filter may be used to approximate the input waveform to the pressure waveform.

With reference to one possible implementation manner of the first aspect, the method further includes: inverse filtering the measured pulse waveform using an inverse filter, the inverse filter based on a filter configuration in a measurement circuit that has measured the pulse waveform,

wherein the filtering step is applied to the inverse filtered pulse waveform.

According to this implementation, the raw measurement signal may be obtained by inverse filtering the input waveform.

With reference to one possible implementation manner of the first aspect, the filter in the measurement circuit is a high-pass filter.

According to this implementation, the low frequency region of the input waveform of the volume pulse wave can be emphasized.

With reference to one possible implementation manner of the first aspect, the tunable filter is a variable roll-off filter, and a roll-off of the variable roll-off filter is less than ±6dB/oct.

According to this implementation, the roll-off of the variable roll-off filter is less than +6dB/oct. Thus, an appropriate roll-off of the variable roll-off filter can be identified.

According to this implementation, the roll-off of the tunable filter may be continuously changed.

With reference to one possible implementation manner of the first aspect, the method further includes: applying a blood circulation model to the converted pulse waveform.

According to this implementation, the waveform of the peripheral blood flow can be converted into waveforms of the blood flow in the heart and the aorta by using a blood circulation model.

With reference to one possible implementation manner of the first aspect, the evaluating step includes: obtaining a plurality of predicted blood pressure values using the plurality of calibration data; and averaging the plurality of predicted blood pressure values.

According to this implementation, errors in calibration data caused by reference blood pressure monitored using the cuff can be reduced.

With reference to one possible implementation manner of the first aspect, the evaluating step further includes: obtaining a plurality of observation data; a plurality of predicted blood pressure values is calculated from the observed data using the plurality of calibration data.

According to such an implementation, the accuracy of the blood pressure measurement may be improved based on the plurality of calibration data.

With reference to a possible implementation manner of the first aspect, the plurality of predicted blood pressure values are associated with the feature.

According to this implementation, the level of the blood pressure is determined based on the characteristics.

With reference to a possible implementation manner of the first aspect, the plurality of predicted blood pressure values includes at least two parts, wherein the averaging the plurality of predicted blood pressure values includes averaging the at least two parts.

In a second aspect, there is provided a blood pressure predicting apparatus including: a measuring circuit for measuring a pulse waveform of the volume pulse wave; a conversion unit for converting the measured pulse waveform using an adjustable filter; an index calculation unit for calculating at least one index for the converted pulse waveform, the at least one index indicating a characteristic of a pressure pulse wave; an evaluation unit for evaluating a correlation between the calculated index and the corresponding blood pressure to adjust the adjustable filter; an output unit outputting the converted pulse waveform filtered using the tunable filter; and a prediction unit for predicting blood pressure based on the output pulse waveform.

With reference to one possible implementation manner of the second aspect, the indicator calculating unit further calculates the at least one indicator for the measured pulse waveform,

the evaluation unit evaluates the correlation between the index calculated before and after the conversion and the corresponding blood pressure.

With reference to one possible implementation manner of the second aspect, the feature is selected from the group consisting of a peak product ratio, a systolic area and an enhancement index of the pulse waveform.

With reference to one possible implementation manner of the second aspect, the tunable filter includes at least one of the following: a variable roll-off filter, an equalization filter with adjustable Q value, or a variable phase shifter.

With reference to one possible implementation manner of the second aspect, the apparatus further includes an inverse filter for performing inverse filtering on the measured pulse waveform, wherein the inverse filter is based on a filter configuration in a measurement circuit that has measured the pulse waveform.

With reference to one possible implementation manner of the second aspect, the filter in the measurement circuit is a high-pass filter.

With reference to one possible implementation manner of the second aspect, the tunable filter is a variable roll-off filter, and a roll-off of the variable roll-off filter is less than ±6dB/oct.

With reference to a possible implementation manner of the second aspect, the apparatus further includes a blood circulation model for applying to the converted pulse waveform.

With reference to one possible implementation manner of the second aspect, the evaluation unit is further configured to: obtaining a plurality of predicted blood pressure values using the plurality of calibration data; and averaging the plurality of predicted blood pressure values.

With reference to one possible implementation manner of the second aspect, the evaluation unit is further configured to: obtaining a plurality of observation data;

a plurality of predicted blood pressure values is calculated from the observed data using the plurality of calibration data.

With reference to a possible implementation manner of the second aspect, the plurality of predicted blood pressure values are associated with the feature.

With reference to the second aspect, in a possible implementation manner, the plurality of predicted blood pressure values includes at least two parts, and the averaging the plurality of predicted blood pressure values includes averaging the at least two parts.

In a third aspect, a computer program product is provided, the computer program product comprising computer executable instructions for storage on a non-transitory computer readable medium, which when executed by a processor, cause the processor to perform the above method.

In a fourth aspect, there is provided a wearable device comprising: a light emitting source for illuminating light to a human body part wearing the wearable device; a light receiving element for receiving the transmitted light or the reflected light; and the measuring circuit is used for converting the transmitted light or the reflected light into pulse waveforms.

According to this aspect, the measurement circuit converts the transmitted light or the reflected light into a pulse waveform. Thus, the converted pulse waveform may be displayed on the wearable device itself or other devices connected to the wearable device.

With reference to one possible implementation manner of the fourth aspect, the wearable device includes at least one of the following: smart watches or headphones.

According to such an implementation, the transmitted light or the emitted light may be converted in the smart watch or headset.

With reference to one possible implementation manner of the fourth aspect, the pulse waveform may be displayed on the smart watch.

According to such an implementation, the user may view the converted pulse waveform on the smart watch.

With reference to one possible implementation manner of the fourth aspect, the earphone is connected to a smart phone, and the pulse waveform may be displayed on the smart phone.

According to such an implementation, the user may view the converted pulse waveform on the smartphone.

With reference to one possible implementation manner of the fourth aspect, the earphone is connected to a smart watch, and the pulse waveform may be displayed on the smart watch.

With reference to one possible implementation manner of the fourth aspect, the pulse waveform is similar to an arterial pressure waveform or an aortic pressure waveform.

According to such an implementation, an arterial pressure waveform or an aortic pressure waveform may be obtained by using the wearable device.

Drawings

FIG. 1 is a diagram illustrating a sensing unit provided by one embodiment;

FIG. 2 is a block diagram schematically illustrating a hardware configuration of a system for predicting blood pressure provided by one embodiment;

fig. 3 is a block diagram showing a functional configuration of the blood pressure prediction apparatus;

FIG. 4 is a flowchart showing a procedure of the blood pressure prediction method;

fig. 5 is a diagram showing one example of an input waveform of an input volume pulse wave;

fig. 6 is a diagram showing one example of a 1-high-pass filter included in the measurement circuit;

Fig. 7 is a diagram showing the cut-off frequency characteristic of the tunable filter;

fig. 8 is a diagram showing an equivalent circuit of an exemplary blood circulation model;

fig. 9 (a) and 9 (b) are diagrams showing examples of output waveforms;

fig. 10 is a diagram showing one example of a pressure waveform;

FIG. 11 is a diagram illustrating a method of assessing systolic area;

FIG. 12 is a diagram illustrating an enhancement index based evaluation method;

FIG. 13 is a table showing one example of the relationship among observation data, calibration data, and prediction data;

fig. 14 is a graph of data in the table shown in fig. 13.

Detailed Description

In order to enable those skilled in the art to better understand the technical solutions of the present invention, the following description will make clear and complete descriptions of the technical solutions of the embodiments of the present invention with reference to the accompanying drawings. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to fall within the scope of the invention.

The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments of the invention only and is not intended to be limiting of the invention.

In the present invention, the waveform signal of the volume pulse wave measured by the PPG is subjected to a certain operation so as to be closer to the pressure waveform, and the feature more directly related to the blood pressure is extracted. In the present embodiment, the integral differentiation process is performed on the original waveform, and then the inverse filter and the tunable filter are applied to the measured pulse waveform. The inverse filter compensates for attenuation and phase distortion already contained in the low frequency region of the source signal by the PPG measurement circuit. In addition, the compensated signal is processed by a tunable filter. A variable roll-off filter may be used as the tunable filter. The variable roll-off filter is used to adjust the roll-off, which indicates the attenuation (slope) of the pass characteristic of the frequency band. The volume pulse waveform is brought close to the pressure pulse waveform (pressure waveform) by adjusting the roll-off so that the correlation coefficient between the characteristics of the converted waveform and the blood pressure is maximized according to the filtering by the tunable filter. For example, the ratio of the wave height of the pulse waveform to the corresponding integrated amount, systolic pressure (systolic blood pressure, SBP), diastolic pressure (diastolic blood pressure, DBP), mean blood pressure (mean blood pressure, MBP), pulse Pressure (PP), and the like may be employed as the characteristic.

The above method is particularly useful for improving the prediction accuracy of SBP or PP. By combining this approach with a calibration approach that improves the accuracy of the DBP/MBP prediction, the prediction errors of SBP and DBP can be suppressed simultaneously. This is particularly effective for predicting DBP/MBP.

Fig. 1 is a diagram illustrating a sensing unit provided by one embodiment. The sensing unit 103 is configured as a pulse oximeter and is attached to the fingertip of the user's finger 102. As described above, photoplethysmography (PPG) is mainly used for measuring blood pressure by a sensor worn on a fingertip. In PPG, a pulse wave signal measured at a fingertip is output to a blood pressure predicting device through a cable 104. The blood pressure prediction device is used for predicting blood pressure based on the value of the signal. Such pulse wave measurement has the advantage of being simpler and non-invasive than the cuff pressure method. It should be noted that the fingertip pulse wave sensor shown in fig. 1 is only one example, and the present invention can be applied to a sensor that measures an ear pulse wave, such as using headphones.

Fig. 2 is a block diagram schematically illustrating a hardware configuration of a system for predicting blood pressure provided by some embodiments. The system 200 is configured by connecting the sensing unit 103 with the blood pressure predicting device 202 by a cable 104. The sensing unit 103 includes a sensor 208 and a measurement circuit 211. The sensor 208 includes: a light emitting diode (light emitting diode, LED) 209, the light emitting diode 209 being a light emitting element for illuminating a fingertip with a certain amount of light; a Photodiode (PD) 210, the photodiode 210 being a light receiving element for receiving transmitted light or reflected light. The measurement circuit 211 converts a change in the amount of light received by the PD 210 into a signal value, and includes a filter 212 and an analog-to-digital converter (ADC) 213. The filter 212 is used to filter an analog signal indicating a change in the amount of light. The ADC 213 is used to convert the filtered analog signal into a digital signal of a pulse waveform.

The blood pressure predicting apparatus 202 is configured such that the controller 203, the input unit 206, the memory 207, the storage device 204, and the display 205 are connected to communicate with each other. The memory 207 may be composed of cache memory, RAM, or the like. The memory 207 includes a read area for a computer program of the controller 203 and a write area serving as a work area for writing processing data of the computer program.

The storage device 204 is a computer-readable storage medium that can be configured with a Hard Disk Drive (HDD), a solid state Drive (Solid State Drive, SSD), or the like. The storage device 204 is an Operating System (OS) for controlling the blood pressure predicting apparatus 202, a utility program, a device driver for Operating a peripheral device, an application program for a specific task, and the like. The controller 203 is configured to execute the information processing according to the present embodiment by loading a computer program stored in the storage device 204 into the memory 207 and executing the computer program. The controller 203 may be composed of a central processing unit (central processing unit, CPU), a microprocessor (Microprocessor Unit, MPU), and the like.

The input unit 206 is a User Interface (UI) for a User to input information to the controller 203, and may be a keyboard, a mouse, a touch screen, a touch pad, or the like. The display 205 is used to display information input from the controller 203, and may be configured of a liquid crystal display, an organic Electroluminescence (EL) display, or the like.

In some embodiments, the system 200 may be configured as a wearable device. In this case, the LED may be a light emitting source for irradiating light to a human body part wearing the wearable device. In addition, the PD 210 may be a light receiving element for receiving transmitted light or reflected light. Further, the measurement circuit 211 and the blood pressure predicting device 202 may be measurement circuits for converting transmitted light or reflected light into pulse waveforms (pressure pulse waveforms).

Here, the wearable device may be at least one of: smart watches or headphones.

If the wearable device is the smart watch, the pulse waveform may be displayed on the smart watch.

If the wearable device is the headset, the headset may be connected to a smartphone and the pulse waveform may be displayed on the smartphone.

In addition, the earphone may be connected to a smart watch, and the pulse waveform may be displayed on the smart watch.

Furthermore, the pulse waveform converted by the measurement circuit may be similar to an arterial pressure waveform or an aortic pressure waveform.

Fig. 3 is a block diagram showing a functional configuration of the blood pressure prediction apparatus 202 provided by some embodiments. The blood pressure prediction device 202 includes an inverse filter 301, a volume pressure conversion unit 302, an index calculation unit 303, a cyclic model application unit 304, an evaluation unit 305, an output unit 306, a prediction unit 307, and a display 205. The inverse filter 301 is used to inverse filter the input pulse waveform. The volume-pressure converting unit 302 is configured to apply an adjustable filter to the inversely filtered pulse waveform to convert the volume pulse wave into an approximate pressure waveform. The index calculation unit 303 is configured to calculate an index for an original input waveform or a filtered pulse waveform. The indicator is indicative of a characteristic of the pressure pulse wave. The circulation model application unit 304 is configured to apply a blood circulation model to the converted pressure waveform. The evaluation unit 305 is arranged to evaluate the correlation between the calculated index and the corresponding blood pressure and to adjust the adjustable filter based on this evaluation. The output unit 306 is used for outputting the pulse waveform filtered using the tunable filter. The prediction unit 307 is used to predict the blood pressure from the converted pressure waveform.

Next, with reference to fig. 4, a procedure of the blood pressure prediction method performed by the blood pressure prediction apparatus 202 will be described.

(1) An index of the pressure pulse is calculated.

In step S401, the index calculation unit 303 calculates an index of the pressure waveform with respect to the input waveform. The indicator is indicative of a characteristic of the pressure pulse wave. In one embodiment, a peak product ratio, a systolic area, and/or an enhancement index (Augmentation Index, AI or AIx) may be used as the index. These parameters will be described in detail below.

(2) Inverse filtering

In step S402, the digital value of the input waveform of the volume pulse wave is inverse-filtered. Fig. 5 is a diagram showing one example of waveforms of input volume pulse waves, in which the horizontal axis represents time and the vertical axis represents gain (voltage). Before the ADC 213 samples the output signal (voltage) of the PD 210, the measurement circuit 211 is provided with a filter 212 and a capacitor, and the volume pulse waveform is affected by them. When these responses are known, the inverse filter may be configured based on the filter in the measurement circuit 211. In this case, an inverse filter is applied to the input waveform of the volume pulse wave caused by the measurement circuit 211 to restore the original pulse waveform. In one embodiment, an inverse filter of the high pass filter is preferably used in the measurement circuit 211, which affects the low frequency region of the input waveform of the volume pulse wave.

Fig. 6 shows an example of a 1-high pass filter as an example of the filter 212 included in the measurement circuit 211. The high pass filter shown in the figure comprises an operational amplifier, a capacitor C, resistors R1 and R2. The passband gain of this circuit is denoted-R ₂ /R ₁ The cut-off frequency fc is expressed as fc=1/2 pi R ₁ C. In addition, the transfer function is expressed as

The transfer function H (z) is frequency converted using a bilinear z-transform and β is calculated by predistortion. In the example of fig. 6, β is determined based on the cut-off frequency of the analog filter of the measurement circuit 211 and the sampling frequency of the ADC 213, and an inverse filter is applied to the input signal. According to the transfer function H (z), the inverse filter is represented as

In one embodiment, the measurement circuit 211 may include a low pass filter or notch filter for removing noise.

(3) Low band compensation

In step S403, the inversely filtered pulse waveform is compensated by the volume-pressure converting unit 302 to be closer to the pressure waveform, and the feature more directly related to the blood pressure is extracted. A method of performing convolution on a specific transfer function is considered as a method of performing waveform conversion so that the characteristic of the volume pulse wave is related to blood pressure. This is the same as applying an inverse model or an inverse filter to the pressure volume model. While such pressure-volume models are known as pressure-volume curves of the heart, there is no such model that is suitable for use in the peripheral circulation. The pressure-volume curve may be applied after converting the waveform of the peripheral blood flow into waveforms of the blood flow in the heart and the aorta using a blood circulation model. However, accurate identification of blood circulation models requires invasive measurements. In addition, the pressure-volume relationship of the heart is affected by arterial compliance.

In this embodiment, instead of performing fine tuning using a complex model, a tunable filter is applied to the inversely filtered pulse waveform. Specifically, a tunable filter is applied to the low frequency component.

A variable roll-off filter may be used as the tunable filter. One example of a variable roll-off filter is a 1/f filter (-3 dB/oct) for generating colored noise. The 1/f filter may be implemented as an FIR filter or as a cascaded 1 pole IIR filter (Paul Kellett method). The 1/f filter can be extended to a so-called 1/f ^exponent A filter. For example, 1/f ^α The filter is a tunable filter, and the attenuation (roll-off) of the pass characteristics of the frequency band can be adjusted by adjusting the value of the α -order. Thus, α may be referred to as a slope parameter. When α=1, this filter is called an inverse f filter. A lower roll-off will result in a slower filter attenuation (more signal passing out of frequency range), rollA higher drop will result in a steeper filter attenuation. This method is useful for correcting integral-based features such as peak-to-product ratio and systolic area. The slope of the filter is less than + -6dB/oct (+ -20 dB/decade), the cut-off frequency is 0Hz or very low.

In PWA, not only the original plethysmographic waveform but also the differential or integral waveform is analyzed. In this case, the volume-pressure conversion unit 302 has a differentiator and an integrator. The slope of the differentiator is +6dB/oct and the slope of the integrator is-6 dB/oct. The tunable filter is used to search for values between these slopes (-6 dB/oct to +6 dB/oct). By combining the filter with a differentiator and integrator, it is possible to search for example in the range of 6 to 12dB/oct or-6 to 12 dB/oct. For example, to achieve-3 dB/oct, a differentiator (-6 dB/oct) and a 1/f filter (+3 dB/oct) are used in combination.

Fig. 7 is a diagram showing the cut-off frequency characteristic of the tunable filter provided in the present embodiment. In the figure, the horizontal axis represents the angular velocity ω (rad/s) of frequency, and the vertical axis represents the gain a (ω) (dB). Each of the curves 701 to 705 shows the gain of a tunable filter having different characteristics, and numerals 1 to 5 in the vicinity of each curve show the order of the filter. If the filter order is large, the roll-off becomes steep. In the present embodiment, the optimum value of the order is searched from the range between the order 0 and the order 1, as indicated by a broken line 706.

(4) Using blood pressure circulation models

The pressure waveform is approximately restored by step S403. Alternatively, step S404 may be optionally performed by the cyclic model application unit 304, and a blood circulation model such as Windkessel model is applied to the waveform for conversion. However, the cyclic model must be identified by another method. It has been found that this procedure can be omitted for PPG measured in the outer ear.

Fig. 8 is a diagram showing an equivalent circuit of an exemplary blood circulation model. How blood pressure is conducted from the center of the heart to the periphery of the body can be considered by replacing the model with an LRC circuit of the circuit shown in fig. 8. This is called a model expressing the vascular system. In the case of the LRC circuit, The voltage V corresponds to the blood pressure immediately after exiting the Left Ventricle (LV) (i.e., central blood pressure). Further, the inductance L corresponds to the elastic characteristic of the artery, the resistance R1 corresponds to the blood pressure resistance of the artery, the resistance R2 corresponds to the peripheral vascular resistance, the capacitance C1 corresponds to the compliance of the arterial vessel, and the capacitance C2 corresponds to the compliance of the peripheral vessel. The LRC circuit in fig. 8 can be considered as a combination of the aortic circuit on the left and the peripheral system circuit on the right. In this circuit, an aortic side circuit G ₁ Transfer function of (S) and peripheral circuit G ₂ The transfer function of (S) can be expressed as follows:

transfer function G converted by z-conversion ₁ (Z) and G ₂ (Z) is represented as follows:

where T is the sampling period, a=c ₁ R ₁ ，b＝LC ₂ ，c＝L/R ₂ 。

The waveform recovered in step S403 approximates the peripheral blood pressure waveform. The voltage waveform of the LRC circuit can be obtained using these two transfer functions. Thus, the central blood pressure waveform can be back calculated from the peripheral blood pressure waveform.

Fig. 9 (a) shows output waveforms generated by performing steps S402 to S404 on the input waveforms shown in fig. 5. Fig. 9 (b) shows another example of an output waveform. These output waveforms are the result of converting the waveform of the volume pulse wave to approximate the pressure waveform.

(5) Post-conversion calculation index

In step S405, the index calculation unit 303 recalculates the index calculated in step S401 for the converted waveform.

(6) Evaluating characteristics of pressure waveforms

Next, the process advances to step S406, and the evaluation unit 305 evaluates the characteristics of the converted waveform. In this evaluation, the correlation between the index of the converted pulse waveform and the corresponding blood pressure is compared with the correlation between the index obtained in step S401 and the corresponding blood pressure to determine that the converted pulse waveform is usable for predicting the blood pressure. This evaluation is performed on the index of systolic (systolic blood pressure, SBP), diastolic (diastolic blood pressure, DBP), mean arterial (mean arterial pressure, MAP) and Pulse Pressure (PP) that is considered to have the highest correlation with the target feature. The positions of these values in the pulse waveform are shown in fig. 10 and 11.

If a pressure waveform is obtained in step S403 or S404, some of the indices obtained therefrom should be closely related to blood pressure. Examples of these indicators are shown below.

(i) Peak product ratio

Fig. 10 is a diagram illustrating an exemplary pressure waveform. The ratio of the blood pressure height at the peak of the pressure waveform to the integral ≡pdt corresponds to the ratio of the pulse pressure h to the average arterial pressure Pa. The mean arterial pressure Pa of the blood pressure P from time t1 to t2 is expressed as

Therefore, the integral amount of the blood pressure P from time t1 to time t2 is equal to the area of a rectangle having a width t2-t1 and a height Pa, as shown on the right side of fig. 10. The area is represented by the following formula:

since the peak product ratio is expressed as ≡pdt/h, if one of ≡pdt and h is known, the other can be calculated. In the case where the pulse pressure h increases due to reflection and superposition of pulse waves in the aorta, this feature also increases. Thus, the peak product ratio can be used as a feature to evaluate whether the correlation with blood pressure is high.

(ii) Area of systolic phase

The amount of integration of the pressure waveform from the beginning of contraction to the dicrotic notch is proportional to stroke volume and pulse pressure and is related to preload and myocardial contractility. Fig. 11 is a diagram illustrating an evaluation method of the systolic area. The systolic area shown in the figure corresponds to the integral of the pressure waveform from the beginning of systole to the dicrotic notch. The systolic area is proportional to stroke volume and pulse pressure. Thus, the systolic area can be evaluated as a feature.

(iii) Enhancement index (Augmentation Index, AI or AIx)

Fig. 12 is a diagram illustrating an enhancement index-based evaluation method. The left heart contracts and ejects blood to expand the aortic wall and propagate in the form of pulse waves. The pulse wave is observed as a single wave in which traveling waves propagate directly in the lower aortic portion and reflected waves overlap each other. If the aortic compliance is low, the time required for the reflected wave to reach the periphery is short, the wave height of the superimposed pressure pulse wave becomes high, and the pulse pressure increases. In fig. 12, P1 represents a wave height P1 caused by only the ejection wave, and P2 represents a wave height increased by the influence of the reflected wave. The difference between the wave height P2 and the wave height P1 is called the boost pressure (Augmentation Pressure, AP). In addition, the ratio of AP to wave height PP, AP/PP, is referred to as the enhancement index (Augmentation Index, AI or AIx). AIx is an indicator of arterial compliance and peripheral vascular tone, and is related to mean arterial blood pressure. Therefore, AI or AIx can be used as a feature for evaluation.

It should be noted that the above parameters are only examples, and all parameters called pressure pulse wave indicators can be used for evaluation. In addition, combinations of the above parameters may also be used for evaluation.

Based on the above parameters, the correlation between the waveform and the blood pressure is determined from the evaluation result of the features obtained in step S407. When it is determined that feedback to the tunable filter is required (correlation is low), the process advances again to step S402. When it is finally determined in step S407 that feedback to the tunable filter is not necessary, the slope of the tunable filter at which the correlation is maximum is selected as the optimal slope. The slope is also applicable to other data. Furthermore, a corresponding index is selected.

(7) Predicting blood pressure

In step S408, the output unit 306 outputs a waveform using the filter whose slope was selected in step S407. Then, the prediction unit 307 executes PWA by using the output waveform and the index selected in step S407 to predict the blood pressure based on the output pressure waveform. For example, the blood pressure may be predicted by using the correspondence between the pressure waveform stored in the blood pressure predicting device 202 and the blood pressure.

In one embodiment, the raw waveforms or intermediate data generated in the above process may optionally be used for prediction.

The above method compensates the integrated amount of the waveform by compensating the low frequency component so as to be close to the pressure waveform. Here, the integral of the pressure pulse corresponds to the mean arterial pressure. Therefore, if the mean arterial pressure can be obtained with high accuracy, the systolic pressure and the pulse pressure can be predicted based on the above-described features and the mean arterial pressure. In order to accurately obtain the mean arterial pressure without using the pulse wave velocity, the PWA needs to be applied using characteristics such as the propagation time difference of the ejection wave and the reflected wave. But in this case its performance is limited.

In this embodiment, the predicting includes:

obtaining a plurality of predicted blood pressure values using the plurality of calibration data;

and averaging the plurality of predicted blood pressure values.

In the case of a system requiring calibration, the values measured by PWA may be calibrated using different calibration values and an estimate calculated for the calibration values. Then, an average value of the plurality of predicted blood pressure values is obtained. The method can improve the accuracy of estimating the mean arterial pressure.

That is, calibration is performed twice or more, a plurality of prediction results are obtained by PWA, and the average value of these results is outputted as a final result. This technique can effectively reduce errors in calibration data caused by reference blood pressure monitored using a cuff. Hereinafter, this calibration method is referred to as "integrated calibration".

Next, the method of integrated calibration will be described with reference to fig. 13 and 14. Fig. 13 shows an example of the relationship between observation data, calibration data, and prediction data provided by the present embodiment. The Blood Pressure (BP) of the observation Data data_a is referred to as bp_a, and the BP of the observation Data data_b is referred to as bp_b. The BP of the calibration Data data_a is referred to as bp_a, the BP of the calibration Data data_b is referred to as bp_b, and the BP of the calibration Data data_c is referred to as bp_c. In the example shown in fig. 13, a plurality of calibration data are used to obtain a plurality of prediction data. Each data is associated with the feature.

The calibration data is used to calculate prediction data from the observation data. BP predicted from the observation Data data_a using the calibration Data data_A is BP_c+A. BP predicted from the observation Data data_a using the calibration Data data_B is BP_d+B. BP predicted from the observation Data data_a using the calibration Data data_C is BP_e+C. The average value of the predicted BP is expressed as the average value of bp_c+ A, BP _d+b and bp_e+c, or "average value (c+ A, d + B, e +c)". Each data is associated with the feature.

Similarly, BP predicted from the observation Data data_b using the calibration Data data_a is bp_f+a. BP predicted from the observation Data data_b using the calibration Data data_B is BP_g+B. BP predicted from the observation Data data_b using the calibration Data data_C is BP_h+C. The average value of the predicted BP is expressed as the average value of bp_f+ A, BP _g+b and bp_h+c, or "average value (f+ A, g + B, h +c)". Each data is associated with the feature.

Fig. 14 is a graph of data in the table shown in fig. 13. In fig. 14, the horizontal axis represents BP, and the vertical axis represents features.

Using the integrated calibration provided in this example, blood pressure was predicted for 9 healthy subjects and compared to the measured values. Tables 1 and 2 show the average absolute error of the SBP and DBP predictions. In table 1, SBP and DBP are predicted by linearly combining the value of the mean arterial pressure obtained by PWA using the integrated calibration and the peak product ratio. In this table, "motionless" means a state in which the subject is sitting, while "reduced" means a state after deep breathing in a sleeping position, and "increased" means a state after mental stress test. In addition, "population" means an average of the above three values. The result error of table 2 is smaller than the case of using the normal PWA shown in table 1, which indicates that the measurement accuracy is improved.

TABLE 1

Conventional method	SBP	DBP
			Immobilized type	4.8	4.9
Reduction of	9.5	7.3
			Increase in	9.2	6.6
Overall (L)	7.8	6.3

TABLE 2

Example 1+2	SBP	DBP
			Immobilized type	4.9	2.8
Reduction of	6.2	3.9
			Increase in	5.7	4.7
Overall (L)	5.6	3.7

In another embodiment, step S402 may be omitted if the measurement circuit does not have a high pass filter. This embodiment is useful for measurement methods or devices having a specific measurement circuit such as a delta-sigma ADC.

In another embodiment, the tunable filter is not limited to a slope variable filter. For example, an equalization filter with an adjustable Q value (quality factor) may be used instead of the slope variable filter. Furthermore, a phase shifter may be used to change only the phase of the input wave. For example, a hilbert transform or an all-pass filter may be used as the tunable filter.

According to the above embodiment, it is not necessary to search for complicated parameters to predict blood pressure. In addition, since embodiments of the present invention may be used for PWA-based sleeveless blood pressure measurement, the present invention is applicable to various PPG sensor modules such as wristwatches, fingerclips, earplugs, video PPG, and the like.

In addition, no invasive parameters are required, at least for in-ear PPG and finger pulse wave sensors.

Furthermore, after tuning an active filter, the peak-to-product ratio is strongly correlated with BP, and the slope parameter of the filter can also be used as a parameter for predicting BP in general.

Furthermore, after adjusting the slope parameters, the same filter may be applied to a similar group of subjects without adjustment for each individual. The subject group may also be classified based on blood pressure.

Still further, better blood pressure predictions may be provided for calibration and calibration-free predictions without the need for personal parameters such as height, weight, etc.

Still further, the present invention may significantly improve the accuracy of predictions when combined with integrated calibration.

An embodiment of the invention also provides a computer program product. All or part of the methods described in the above method embodiments may be implemented by software, hardware, firmware, or any combination thereof. When the methods are implemented in software, all or part of the methods may be implemented in the form of a computer program product. The computer program product may include one or more computer instructions. When loaded and executed on a computer, generates all or part of a program or function according to the method embodiments described above.

The above description is only a specific implementation of the present invention and is not intended to limit the scope of the present invention. Any changes or substitutions that would be readily apparent to one of ordinary skill in the art within the scope of the disclosed technology are intended to be encompassed within the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.

Claims

1. A method of predicting blood pressure, comprising:

measuring a pulse waveform of the volume pulse wave;

converting the measured pulse waveform using an adjustable filter;

Calculating at least one index for the converted pulse waveform, the at least one index indicating a characteristic of a pressure pulse wave;

evaluating a correlation between the calculated index and the corresponding blood pressure to adjust the tunable filter;

outputting the converted pulse waveform filtered using the tunable filter;

and predicting blood pressure based on the output pulse waveform.

2. The method as recited in claim 1, further comprising:

calculating the at least one indicator for the measured pulse waveform, wherein the evaluating step comprises: the correlation between the index calculated before and after the conversion step and the corresponding blood pressure is evaluated.

3. The method of claim 1 or 2, wherein the characteristic is selected from the group consisting of peak-to-product ratio of pulse waveform, systolic area, and enhancement index.

4. A method according to any one of claims 1 to 3, further comprising:

an indicator is selected from the at least one indicator based on the evaluation, wherein the predicting step predicts blood pressure based on the output pulse waveform and the selected indicator.

5. The method according to any one of claims 1 to 4, wherein the tunable filter comprises at least one of: a variable roll-off filter, an equalization filter with adjustable Q value, or a variable phase shifter.

6. The method according to any one of claims 1 to 5, further comprising: inverse filtering the measured pulse waveform using an inverse filter, the inverse filter based on a filter configuration in a measurement circuit that has measured the pulse waveform,

wherein the filtering step is applied to the inverse filtered pulse waveform.

7. The method of claim 6, wherein the filter in the measurement circuit is a high pass filter.

8. The method of claim 1, wherein the tunable filter is a variable roll-off filter having a roll-off of less than ± 6dB/oct.

9. The method according to any one of claims 1 to 8, further comprising: applying a blood circulation model to the converted pulse waveform.

10. The method according to any one of claims 1 to 9, wherein the evaluating step comprises:

and averaging the plurality of predicted blood pressure values.

11. The method of claim 10, wherein the evaluating step further comprises:

obtaining a plurality of observation data;

12. The method of claim 11, wherein the plurality of predicted blood pressure values are associated with the characteristic.

13. The method of any one of claims 10 to 12, wherein the plurality of predicted blood pressure values comprises at least two portions, and wherein the averaging the plurality of predicted blood pressure values comprises averaging the at least two portions.

14. A blood pressure prediction device, comprising:

a measuring circuit for measuring a pulse waveform of the volume pulse wave;

a conversion unit for converting the measured pulse waveform using an adjustable filter;

an index calculation unit for calculating at least one index for the converted pulse waveform, the at least one index indicating a characteristic of a pressure pulse wave;

an evaluation unit for evaluating a correlation between the calculated index and the corresponding blood pressure to adjust the adjustable filter;

An output unit for outputting the converted pulse waveform filtered using the tunable filter;

and a prediction unit for predicting blood pressure based on the output pulse waveform.

15. The apparatus according to claim 14, wherein the index calculation unit further calculates the at least one index for the measured pulse waveform, wherein the evaluation unit evaluates the correlation between the index calculated before and after the conversion and the respective blood pressure.

16. The device of claim 14 or 15, wherein the characteristic is selected from the group consisting of peak-to-product ratio of pulse waveform, systolic area, and enhancement index.

17. The apparatus according to any one of claims 14 to 16, wherein the evaluation unit selects an index from the at least one index based on the evaluation, and wherein the prediction unit predicts the blood pressure based on the output pulse waveform and the selected index.

18. The apparatus according to any one of claims 11 to 14, wherein the tunable filter comprises at least one of: a variable roll-off filter, an equalization filter with adjustable Q value, or a variable phase shifter.

19. The apparatus of any one of claims 14 to 18, further comprising an inverse filter for inverse filtering the measured pulse waveform, wherein the inverse filter is based on a filter configuration in a measurement circuit that has measured the pulse waveform.

20. The apparatus of claim 16, wherein the filter in the measurement circuit is a high pass filter.

21. The apparatus of claim 14, wherein the tunable filter is a variable roll-off filter having a roll-off of less than ± 6dB/oct.

22. The apparatus of any one of claims 14 to 21, further comprising a blood circulation model for application to the converted pulse waveform.

23. The apparatus according to any one of claims 14 to 22, wherein the evaluation unit is further configured to:

and averaging the plurality of predicted blood pressure values.

24. The apparatus of claim 23, wherein the evaluation unit is further configured to:

obtaining a plurality of observation data;

25. The apparatus of claim 24, wherein the plurality of predicted blood pressure values are associated with the characteristic.

26. The apparatus of any one of claims 23 to 25, wherein the plurality of predicted blood pressure values comprises at least two portions, wherein the averaging the plurality of predicted blood pressure values comprises averaging the at least two portions.

27. A computer program product comprising computer executable instructions for storage on a non-transitory computer readable medium, which when executed by a processor cause the processor to perform the method according to any one of claims 1 to 10.

28. A wearable device, comprising:

a light emitting source for illuminating light to a human body part wearing the wearable device;

a light receiving element for receiving the transmitted light or the reflected light;

and the measuring circuit is used for converting the transmitted light or the reflected light into pulse waveforms.

29. The wearable device of claim 28, wherein the wearable device comprises at least one of: smart watches or headphones.

30. The wearable device of claim 29, wherein the pulse waveform is displayable on the smart watch.

31. The wearable device of claim 29, wherein the headset is connected to a smartphone and the pulse waveform is displayable on the smartphone.

32. The wearable device of claim 29, wherein the headset is connected to a smart watch and the pulse waveform can be displayed on the smart watch.

33. The wearable apparatus of any of claims 28-32, wherein the pulse waveform is similar to an arterial pressure waveform or an aortic pressure waveform.