Blood pressure estimation device, computer program, and recording medium

Abstract

To estimate a blood pressure from one physiological index such as a cardiac rate with a high degree of precision.SOLUTION: In setting a quaternary polynomial, a boundary frequency for distinguishing between a non-periodic component and a periodic component is extracted from a double logarithm graph, and a frequency gradient is acquired from an approximate curve indicated by a quaternary polynomial set in a low frequency band or a high frequency band with the boundary frequency in between. By performing settings of the approximate curve indicated by the quaternary polynomial in each band of the non-periodic component and the periodic component with the boundary in between in the double logarithm graph, the frequency gradient can be obtained in the form matched to a wave motion in a blood vessel, thereby improving precision of blood pressure value estimation.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a technique for estimating blood pressure without using a cuff, using at least one of biosignal data of heart rate and an acoustic pulse wave caused by fluctuations in left ventricular pressure obtained from the body surface. [Background technology]

[0002] The applicant has disclosed in Patent Document 1 a blood pressure estimation device that analyzes biosignal data obtained from a biosignal detection sensor using three-dimensional knitted fabric, captures vibration information within the body caused by blood flow fluctuations from the ventricular filling phase to the isovolumic contraction phase, determines an index showing the fluctuations by applying the Lorenz plot method, and estimates blood pressure values ​​from the index (Patent Document 1).

[0003] Specifically, the technology captures biosignals (acoustic pulse waves (APW)) from the surface of the body on the back using a biosignal detection sensor, filters the time-series waveform in a specified frequency band to obtain a filtered waveform that makes the cardiac cycle apparent, compares this filtered waveform with electrocardiogram waveform data obtained from an electrocardiograph measured at the same time, and identifies the waveform components from the ventricular filling period to the isovolumic contraction period. After that, the Lorentz plot method is applied, and using the slope of a large number of point clouds plotted at a specified measurement time as a reference, a new time series is constructed for the angle difference with the slope of the point clouds at a shorter measurement time. The time-series waveform is frequency-analyzed, the frequency analysis results are displayed on a double logarithmic axis, a regression line is found for the range from VLF to LF in the analyzed waveform, the slope of the regression line is defined as the fractal slope, and blood pressure is estimated from the fractal slope.

[0004] Patent Document 2 focuses on the fact that pulse wave velocity (PWV) depends on blood pressure and the elastic properties of arteries, and discloses a technique for measuring blood pressure assuming that blood pressure is proportional to PWV. It also describes that PWV is calculated from PAT (Pulse Arrival Time). Since PWV is inversely proportional to the derivative of PAT, blood pressure is estimated by calculating the derivative value of PAT. It also describes that heart rate and oxygen saturation are measured by an optical sensor used in mobile computers such as mobile phones.

[0005] Both Patent Documents 1 and 2 have in common the point that they are techniques capable of estimating blood pressure without using a cuff-based blood pressure monitor, i.e., an upper arm blood pressure monitor. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] JP 2019-122502 A [Patent Document 2] US2015 / 0320328A1 publication Summary of the Invention [Problem to be solved by the invention]

[0007] By the way, the complex control for maintaining blood pressure homeostasis has fluctuations due to energy fluctuations. Heart rate (HR) is derived from the RRI of the electrocardiogram (ECG), and heart rate variability (HRV) has fluctuations, and the fluctuation index is calculated from the results of frequency analysis of heart rate variability. Heart rate variability is reflected in afterload and baroreflex, and arterial pressure fluctuates due to afterload fluctuations and baroreflex.

[0008] In Patent Document 1, the fractal slope is used as an index showing this fluctuation, but the fractal slope is an index showing the characteristics of the analysis waveform only by a regression line. Therefore, it is difficult to say that it fully reflects the characteristics of blood pressure of a complex control structure having a complex control mechanism as described above.

[0009] In addition, as in Patent Document 2, blood pressure is calculated by detecting the pulse wave velocity. Although the pulse wave velocity is said to be correlated with blood pressure, the degree of correlation is not so high. Therefore, even if the pulse wave velocity or the pulse wave propagation delay time used as a substitute for the pulse wave velocity is simply estimated in comparison with the blood pressure, there are concerns about the accuracy.

[0010] The present invention has been made in consideration of the above, and an object of the present invention is to provide a technology that can estimate blood pressure with higher accuracy from biosignal data of either heart rate or acoustic pulse wave caused by left ventricular pressure fluctuation obtained from the body surface, without using a cuff. [Means for solving the problem]

[0011] In order to solve the above problems, the cardiovascular system state estimating device of the present invention comprises: A biosignal receiving unit that receives at least one biosignal data of a heart rate and an acoustic pulse wave caused by a left ventricular pressure fluctuation obtained from a body surface; a time-series fluctuation data calculation unit for calculating time-series fluctuation data reflecting time-series fluctuation information of the biological signal data; a frequency analysis unit that performs frequency analysis on the time-series fluctuation data and displays the frequency analysis result in a double logarithmic graph; a blood pressure estimation equation constructing unit that constructs a blood pressure estimation equation using the biosignal data of a plurality of subjects, the log-log graph, and actual blood pressure values ​​measured by an upper arm blood pressure monitor; an estimated blood pressure value calculation unit that applies the biological signal data of the subject to be estimated to the blood pressure estimation formula to obtain an estimated blood pressure value of the subject to be estimated; having The blood pressure estimation equation constructing unit a first boundary frequency calculation unit that calculates a boundary frequency between aperiodic components and periodic components while displaying a power spectrum of a predetermined frequency band in a range from VLF to LF on the log-log graph; a first analysis target frequency band selection unit that selects, from the log-log graph, a low frequency band or a high frequency band that sandwiches the boundary frequency as an analysis target frequency band; a first frequency gradient calculation unit that fits an approximation curve represented by a fourth-order polynomial by curve fitting to the display of the power spectrum in the analysis target frequency band, and calculates a frequency gradient of the approximation curve represented by the fourth-order polynomial; an estimation equation calculation unit that calculates an equation showing a correlation between the actual blood pressure value and the frequency gradient from a response surface obtained by plotting the biological signal data, the actual blood pressure value, and the frequency gradient on the coordinate axes of a three-dimensional coordinate system, as the blood pressure estimation equation; has.

[0012] The first boundary frequency calculation unit is preferably configured to apply an approximation curve represented by a third-order polynomial by curve fitting to a power spectrum display of a predetermined frequency band in a range from VLF to LF in the double logarithmic graph, and to determine the frequency of an inflection point of the approximation curve represented by the third-order polynomial as the boundary frequency between the aperiodic component and the periodic component.

[0013] The first analysis target frequency band selection unit is If the measured blood pressure value is classified as high blood pressure or high blood pressure, the high frequency band is selected as the band to be analyzed; When the actual measured blood pressure value is less than a blood pressure value classified as high blood pressure, the low frequency band is selected as the band to be analyzed. It is preferable to configure as follows.

[0014] The first frequency gradient calculation unit is a first approximation curve setting unit that obtains an approximation curve represented by the fourth-order polynomial by curve fitting for the analysis target frequency band; a first inflection point (stationary point) determination unit that determines whether or not there is an inflection point (stationary point) that also serves as a stationary point among two inflection points of the approximation curve represented by the fourth-order polynomial; a first gradient determination unit that determines, when the inflection point (stationary point) does not exist, a gradient of a chord between two inflection points in the fourth-order polynomial in the frequency band to be analyzed as the frequency gradient, and that determines, when the inflection point (stationary point) exists, a gradient of a tangent to one of the two inflection points as the frequency gradient; It is preferred that the compound has the formula:

[0015] The estimated blood pressure value calculation unit a second boundary frequency calculation unit that calculates a boundary frequency between aperiodic components and periodic components while displaying a power spectrum of a predetermined frequency band in a range from VLF to LF for a double logarithmic graph obtained by the time-series fluctuation data calculation unit and the frequency analysis unit using the biosignal data of the subject; a second analysis target frequency band selection unit that selects a low frequency band or a high frequency band that sandwiches the boundary frequency in the double logarithmic graph of the estimation target person as an analysis target frequency band; a second frequency gradient calculation unit that fits an approximation curve represented by a fourth-order polynomial by curve fitting to the display of the power spectrum in the analysis target frequency band, and obtains a frequency gradient of the approximation curve represented by the fourth-order polynomial as a frequency gradient for estimation; an estimation unit that substitutes the frequency gradient for estimation obtained by the second frequency gradient calculation unit into the blood pressure estimation equation to obtain an estimated blood pressure value of the estimation subject; It is preferred that the compound has the formula:

[0016] The second boundary frequency calculation unit is an approximation curve represented by a third-order polynomial is fitted by curve fitting to a power spectrum display of a predetermined frequency band in a range from VLF to LF in the double logarithmic graph, and a frequency of an inflection point of the approximation curve represented by the third-order polynomial is determined as a boundary frequency between the non-periodic component and the periodic component; The second analysis target frequency band selection unit is In displaying the power spectrum of a predetermined frequency band ranging from VLF to LF including the boundary frequency in the double logarithmic graph, the slope index FS of the regression line of the entire frequency band and the slope index FS of the regression line of the frequency band in the range above the boundary frequency are BF Ratio to:FS BF a slope ratio calculation unit for calculating a slope ratio of the regression line, which is / FS; The regression line slope ratio (FS BF A provisional estimated blood pressure value calculation unit that calculates a provisional estimated blood pressure value using a correlation diagram between the actual blood pressure measurement value and the blood pressure measurement value. having If the provisional estimated blood pressure value is classified as high blood pressure or high blood pressure, the high frequency band is selected as an analysis target band; When the provisional estimated blood pressure value is less than a blood pressure value classified as a high blood pressure, the low frequency band is selected as an analysis target band, The second frequency gradient calculation unit is a second approximation curve setting unit that obtains an approximation curve represented by the fourth-order polynomial by curve fitting for the analysis target frequency band; a second inflection point (stationary point) determination unit that determines whether or not there is an inflection point (stationary point) that also serves as a stationary point among two inflection points of the approximation curve represented by the fourth-order polynomial; a second gradient determination unit that determines, when the inflection point (stationary point) does not exist, a gradient of a chord between two inflection points in the fourth-order polynomial in the frequency band to be analyzed as the frequency gradient, and that determines, when the inflection point (stationary point) exists, a gradient of a tangent to one of the two inflection points as the frequency gradient; It is preferred that the compound has the formula:

[0017] The blood pressure estimation equation constructing unit and a general shape characteristic determining unit for applying a regression line or a regression curve having a high coefficient of determination to a power spectrum display of a predetermined frequency band in the range from VLF to LF including the boundary frequency in the double logarithmic graph, and determining a general shape characteristic of the power spectrum display. It is preferred that the compound has the formula: The estimated blood pressure value calculation unit and a general shape characteristic determining unit for applying a regression line or a regression curve having a high coefficient of determination to a power spectrum display of a predetermined frequency band in the range from VLF to LF including the boundary frequency in the double logarithmic graph, and determining a general shape characteristic of the power spectrum display. It is preferred that the compound has the formula:

[0018] The time series fluctuation data in the time series fluctuation data calculation unit is It is preferable that the Lorentz plot method is applied to the biosignal data, a reference slope is obtained from the Lorentz plot in a specified reference time range, and a relative slope, which is the difference between the slope of each Lorentz plot created at each specified time shorter than the reference time range and the reference slope, is subjected to moving average processing, and the obtained relative slope is time series data. The heart rate may be used as the vital sign data, and the estimated blood pressure value may be a systolic blood pressure. The acoustic pulse wave may be used as the biological signal data, and the estimated blood pressure value may be a systolic blood pressure or a diastolic blood pressure.

[0019] The present invention also provides a computer program for causing a computer to function as a blood pressure estimation device, comprising: receiving at least one of biosignal data of a heart rate and an acoustic pulse wave caused by a change in left ventricular pressure obtained from a body surface; A step of obtaining time-series fluctuation data reflecting time-series fluctuation information of the biosignal data; a step of performing a frequency analysis on the time-series fluctuation data and displaying a result of the frequency analysis in a double logarithmic graph; A step of constructing a blood pressure estimation equation using the biosignal data of a plurality of subjects, the log-log graph, and actual blood pressure values ​​measured by an upper arm blood pressure monitor; applying the biological signal data of the subject to be estimated to the blood pressure estimation formula to obtain an estimated blood pressure value of the subject to be estimated; and causing the computer to execute the steps of: In the step of constructing the blood pressure estimation equation, In the double logarithmic graph, a boundary frequency between a non-periodic component and a periodic component is obtained while displaying a power spectrum of a predetermined frequency band in a range from VLF to LF; In the log-log graph, a low frequency band or a high frequency band sandwiching the boundary frequency is selected as a frequency band to be analyzed; an approximation curve represented by a fourth-order polynomial is fitted to the power spectrum display in the analysis target frequency band by curve fitting, and a frequency gradient of the approximation curve represented by the fourth-order polynomial is obtained; A step of obtaining an equation showing a correlation between the actual blood pressure value and the frequency gradient as the blood pressure estimation equation from a response surface obtained by plotting the biological signal data, the actual blood pressure value, and the frequency gradient on the coordinate axes of a three-dimensional coordinate system. A computer program for causing the computer to execute the to provide.

[0020] In the step of determining the boundary frequency, it is preferable to apply an approximation curve represented by a third-order polynomial by curve fitting to the power spectrum display of a predetermined frequency band in the range from VLF to LF in the double logarithmic graph, and determine the frequency of an inflection point of the approximation curve represented by the third-order polynomial as the boundary frequency between the nonperiodic component and the periodic component. In the step of selecting a frequency band to be analyzed, If the measured blood pressure value is classified as high blood pressure or high blood pressure, the high frequency band is selected as the band to be analyzed; When the actual measured blood pressure value is below a blood pressure value classified as high blood pressure, it is preferable to select the low frequency band as the band to be analyzed. In the procedure for calculating the frequency gradient of the approximation curve represented by the fourth-order polynomial, An approximation curve represented by the fourth-order polynomial is obtained by curve fitting for the frequency band to be analyzed; determining whether or not there is an inflection point (stationary point) that also serves as a stationary point in the approximation curve represented by the fourth-order polynomial; a step of determining, when the inflection point (stationary point) does not exist, a gradient of a chord between two inflection points in the fourth-order polynomial in the frequency band to be analyzed as the frequency gradient, and, when the inflection point (stationary point) exists, determining a gradient of a tangent to the inflection point (stationary point) as the frequency gradient. It is preferable to have the computer execute the following:

[0021] In the step of calculating the estimated blood pressure value, determining a boundary frequency between aperiodic components and periodic components while displaying a power spectrum of a predetermined frequency band in a range from VLF to LF for a double logarithmic graph obtained using the biosignal data of the subject; In the log-log graph of the estimated subject, a low frequency band or a high frequency band sandwiching the boundary frequency is selected as an analysis target frequency band; an approximation curve represented by a fourth-order polynomial is fitted to the display of the power spectrum in the analysis target frequency band by curve fitting, and a frequency gradient of the approximation curve represented by the fourth-order polynomial is obtained as a frequency gradient for estimation; It is preferable that the frequency gradient for estimation obtained by the second frequency gradient calculation unit is substituted into the blood pressure estimation equation to obtain the estimated blood pressure value of the estimation subject.

[0022] an approximation curve represented by a third-order polynomial is fitted by curve fitting to a power spectrum display of a predetermined frequency band ranging from VLF to LF in a double logarithmic graph obtained using the biosignal data of the subject, and a frequency of an inflection point of the approximation curve represented by the third-order polynomial is determined as a boundary frequency between the non-periodic component and the periodic component; In displaying the power spectrum of a predetermined frequency band ranging from VLF to LF including the boundary frequency in the double logarithmic graph, the slope index FS of the regression line of the entire frequency band and the slope index FS of the regression line of the frequency band in the range above the boundary frequency are BF Ratio to:FS BF Calculate the regression line slope ratio, which is / FS. The regression line slope ratio (FS BF A provisional estimated blood pressure value is calculated using a correlation diagram between the actual blood pressure measurement and the FS. If the provisional estimated blood pressure value is classified as high blood pressure or high blood pressure, the high frequency band is selected as an analysis target band; If the provisional estimated blood pressure value is less than a blood pressure value classified as a high blood pressure, the low frequency band is selected as an analysis target band; An approximation curve represented by the fourth-order polynomial is obtained by curve fitting for the frequency band to be analyzed; determining whether or not there is an inflection point (stationary point) that also serves as a stationary point in the approximation curve represented by the fourth-order polynomial; a step of determining, when the inflection point (stationary point) does not exist, a gradient of a chord between two inflection points in the fourth-order polynomial in the frequency band to be analyzed as the frequency gradient, and, when the inflection point (stationary point) exists, determining a gradient of a tangent to the inflection point (stationary point) as the frequency gradient. It is preferable that the computer is configured to execute the above steps.

[0023] The present invention also provides a computer-readable recording medium on which the computer program is recorded. Effect of the Invention

[0024] According to the present invention, the time series data of the fluctuation of at least one of the biosignal data of the heart rate and the acoustic pulse wave caused by the fluctuation of the left ventricular pressure obtained from the body surface is frequency-analyzed and displayed on a double logarithmic graph, and the result is approximated by an approximation curve represented by a fourth-order polynomial, and the slope obtained from the approximation curve represented by the fourth-order polynomial is calculated as the frequency gradient. The approximation curve represented by the fourth-order polynomial is considered to contain information on stabilization of blood pressure involving the sympathetic nervous system via the 1 / f fluctuation, which is a characteristic of biological validity, and the frequency gradient, which is the slope of the approximation curve represented by the fourth-order polynomial, contains information on stabilizing blood pressure and is blood pressure dependent. The present invention divides this frequency gradient into the slope between the inflection points in the approximation curve represented by the fourth-order polynomial or the slope of the tangent to the inflection point, and employs either of them. As a result, when the actual blood pressure value is near normal blood pressure, the integrated result of multiple controls tends to result in the adoption of the chord gradient connecting the two inflection points, whereas in the case of high blood pressure and low blood pressure, only one control factor is enhanced and only one inflection point appears, so the tangent gradient at the inflection point tends to be adopted.

[0025] As a result, the present invention can estimate blood pressure without using an upper arm blood pressure monitor using a cuff. In the present invention, when setting the above fourth-order polynomial, a boundary frequency that distinguishes between aperiodic components and periodic components is extracted from a double logarithmic graph, and the above frequency gradient is obtained from an approximation curve represented by a fourth-order polynomial that is set in the low-frequency band or high-frequency band that sandwiches the boundary frequency.

[0026] The waves in blood vessels are a superposition of waves generated by various factors. The low frequency component (LF component) of short-term heart rate fluctuation, which is a neural regulation, occurs when the Mayer wave of blood pressure fluctuation is reflected in the heart rate by the baroreceptor reflex, and is usually distributed in the frequency band of 0.06 to 0.11 Hz, with the periodic component accounting for the posture. The very low frequency component (VLF component) defined as 0.003 to 0.04 Hz, which is a long-term heart rate fluctuation, contains information that reflects the regulation of the effective blood volume of the circulatory regulation system and the health of the entire organism, but the non-periodic component accounts for the posture. Therefore, by setting the approximation curve represented by a fourth-order polynomial for each band on either side of the boundary between the non-periodic component and the periodic component while displaying the double logarithmic graph, the frequency gradient can be obtained in a form that matches the waves in blood vessels, and the estimation accuracy of blood pressure values ​​can be improved. [Brief description of the drawings]

[0027] [Figure 1] FIG. 1 is a block diagram showing a configuration of a blood pressure estimation device according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a block diagram for explaining the configuration of the blood pressure estimation equation constructing unit. [Diagram 3] FIG. 3 is a block diagram for explaining the configuration of the estimated blood pressure value calculation unit. [Figure 4] FIG. 4 is a flowchart showing the process of frequency analyzing time-series fluctuation data (Δθi function) showing fluctuations in heart rate (instantaneous heart rate) and obtaining a power spectrum (spectral diagram) that is the result of the frequency analysis. [Diagram 5] FIG. 5 is a diagram showing data output in each step of FIG. 5 in accordance with the flowchart of FIG. [Figure 6]FIG. 6 is a flow chart showing the steps for obtaining the Δθi function using the APW. [Figure 7] FIG. 7 is a diagram showing data output in each step of FIG. 6 in accordance with the flowchart of FIG. [Figure 8] FIG. 8 is a flow chart showing the steps for determining the boundary frequency (BF). [Figure 9] FIG. 9 is a diagram showing data output in each step of FIG. 8 in accordance with the flowchart of FIG. [Figure 10] 10(a) to (c) are two-dimensional scatter plots obtained from a total of 171 healthy subjects and outpatient cardiology patients (260 cases), and FIG. 10(d) is a three-dimensional response surface plot. [Figure 11] FIG. 11 is a flow chart showing the process of determining the frequency band to be analyzed by obtaining the regression line slope ratio. [Figure 12] FIG. 12(a) is a diagram showing the data output at each step of FIG. 11 in accordance with the flowchart of FIG. 11, and FIG. 12(b) is a correlation diagram between the regression line slope ratio and the actual measured value of systolic blood pressure. [Figure 13] FIG. 13 is a block diagram for explaining the configuration of the general shape characteristics determining unit. [Figure 14] FIG. 14 is a diagram showing an example of an analysis of GI or GT, which is a frequency gradient. [Figure 15] FIG. 15 is a flow chart showing the steps of determining the frequency gradient, GI or GT. [Figure 16] Figures 16(a) to (e) are an example of analysis of a subject whose measured systolic blood pressure was 76 mmHg, where (a) is the Lorentz plot, (b) is the Δθi function, (c) is a plot of the power spectrum, (d) is a graph showing the approximate curve of a fourth-order polynomial, and (e) is a graph showing the frequency gradient. [Figure 17]Figures 17(a) to (e) are an example of analysis of a subject whose measured systolic blood pressure was 101 mmHg, where (a) is a Lorentz plot, (b) is the Δθi function, (c) is a plot of the power spectrum, (d) is a graph showing the approximate curve of a fourth-order polynomial, and (e) is a graph showing the frequency gradient. [Figure 18] Figures 18(a) to (e) are an example of analysis of a subject whose measured systolic blood pressure was 116 mmHg, where (a) is the Lorentz plot, (b) is the Δθi function, (c) is a plot of the power spectrum, (d) is a graph showing the approximate curve of a fourth-order polynomial, and (e) is a graph showing the frequency gradient. [Figure 19] Figures 19(a) to (e) are an example of analysis of a subject whose measured systolic blood pressure was 127 mmHg, where (a) is the Lorentz plot, (b) is the Δθi function, (c) is a plot of the power spectrum, (d) is a graph showing the approximate curve of a fourth-order polynomial, and (e) is a graph showing the frequency gradient. [Figure 20] Figures 20(a) to (e) are an example of analysis of a subject whose measured systolic blood pressure was 144 mmHg, where (a) is the Lorentz plot, (b) is the Δθi function, (c) is a plot of the power spectrum, (d) is a graph showing the approximate curve of a fourth-order polynomial, and (e) is a graph showing the frequency gradient. [Figure 21] Figures 21(a) to (e) are an example of analysis of a subject whose measured systolic blood pressure was 162 mmHg, where (a) is the Lorentz plot, (b) is the Δθi function, (c) is a plot of the power spectrum, (d) is a graph showing the approximate curve of a fourth-order polynomial, and (e) is a graph showing the frequency gradient. [Figure 22] FIG. 22 shows the results of a six-class analysis of variance that indicates the general characteristics of the log-log spectrum graph. [Figure 23] Figure 23(a) is a graph showing the correlation between the inflection points of the approximation curve represented by a fourth-order polynomial and the actual measured value of systolic blood pressure, and Figure 23(b) is a graph showing the correlation between the frequency of the inflection points used to calculate the frequency gradient and the frequency of the actual measured value of systolic blood pressure. [Figure 24]Figure 24(a) shows the correlation between estimated blood pressure (eSBP) calculated using instantaneous heart rate data from 171 patients (260 cases) and actual systolic blood pressure (SBP), and Figure 24(b) is a cross-tabulation table showing the relationship between actual systolic blood pressure and GI or GT. [Diagram 25] Figure 25 shows the correlation between estimated blood pressure (eSBP) and actual systolic blood pressure (SBP) for 22 subjects different from the 171 subjects (260 cases), and Figure 25(b) is a cross-tabulation table. [Figure 26] FIG. 26 is a diagram for explaining application to left ventricular pressure waveforms and the reason why normal blood pressure often results in GI and hypertension and hypotension often results in GT. [Figure 27] FIG. 27(a), like FIG. 26, is a diagram to explain why GI is often observed in cases of normal blood pressure and GT is often observed in cases of high blood pressure and low blood pressure, and FIG. 27(b) is an image diagram of heart rate fluctuations. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0028] (Blood pressure estimation device) 1 shows a schematic configuration of one embodiment of a blood pressure estimation device 1. The blood pressure estimation device 1 of this embodiment is composed of a computer (the type of computer is not limited, and includes a personal computer, a microcomputer, a mobile information terminal such as a smartphone, and the like), and receives biological signal data and the like and executes a predetermined process.

[0029] Specifically, the blood pressure estimation device 1 has a computer program stored in a storage unit (including a recording medium such as a hard disk built into the blood pressure estimation device 1 composed of a computer, various removable recording media, and recording media of other computers connected by communication means) that executes procedures that function as the biosignal receiving unit 11, the time-series fluctuation data calculating unit 12, the frequency analyzing unit 13, the blood pressure estimation equation constructing unit 14, and the estimated blood pressure value calculating unit 15. The blood pressure estimation device 1 can also be realized using an electronic circuit having one or more storage circuits in which the above computer programs are incorporated. That is, the biosignal receiving unit 11 can be configured as a transmitting / receiving circuit of a computer, and the time-series fluctuation data calculating unit 12, the frequency analyzing unit 13, the blood pressure estimation equation constructing unit 14, and the estimated blood pressure value calculating unit 15 can be configured as a processing circuit that calculates time-series fluctuation data from the biosignal data received in the transmitting / receiving circuit, performs frequency analysis, constructs a blood pressure estimation equation, and uses the blood pressure estimation equation to obtain an estimated blood pressure value.

[0030] The computer program can be provided by storing it in a recording medium. The recording medium storing the computer program may be a non-transient recording medium. There is no particular limitation on the non-transient recording medium, and examples of the non-transient recording medium include a flexible disk, a hard disk, a CD-ROM, an MO (magneto-optical disk), a DVD-ROM, and a memory card. It is also possible to transmit the computer program to a computer via a communication line and install it therein.

[0031] The biosignal receiving unit 11 receives at least one of biosignal data of the heart rate and the acoustic pulse wave caused by the left ventricular pressure fluctuation obtained from the body surface. The heart rate and the acoustic pulse wave are physiological indexes of the cardiovascular system, and their fluctuations affect the effective blood volume and become blood pressure control factors. The heart rate can be an instantaneous heart rate calculated from the RR interval (60 / Δt / min: δt=RRI) of an electrocardiogram (ECG) obtained from an electrocardiograph. The instantaneous heart rate may be calculated in an electrocardiograph and the biosignal receiving unit 11 may simply receive the data, or the biosignal receiving unit 11 may receive only the RRI and calculate the instantaneous heart rate. The biosignal receiving unit 11 of this embodiment may have either of these functions. In addition, a wristwatch type electrocardiograph, for example, a smart watch, may be used. The acoustic pulse wave (APW) is sound and vibration information collected from the back, and an acoustic pulse wave detection sensor used is used in a drowsy driving warning device (Sleep Buster (registered trademark)) manufactured by Delta Tooling Co., Ltd. In addition, the acoustic pulse wave (APW) uses data (APW20) that is obtained by filtering the received data at 10 to 30 Hz.

[0032] The time-series fluctuation data calculation unit 12 obtains time-series data (time-series fluctuation data) of fluctuations (fluctuations) of the biosignal data received by the biosignal receiving unit 11. The time-series fluctuation data is data constructed by comparing biosignal data at a specific time with biosignal data at the next specific time, and plotting data showing the fluctuation between the two. The fluctuation is based on the time phase of atrial contraction-ventricular contraction by ECG, i.e., the R wave, and in this embodiment, the fluctuation is converted into angle information by the Lorentz plot method, and a function with the angle as an argument ("Δθi function") is obtained.

[0033] That is, the time series fluctuation data calculation unit 12 obtains a reference slope from the Lorentz plot in the reference time range, performs moving average processing on the relative slope (characteristic of fluctuation), which is the difference between the slope of each Lorentz plot created at each specified time shorter than the reference time range and the reference slope, and obtains the time series data of the obtained relative slope as time series fluctuation data of the biosignal data.

[0034] For example, as shown in Fig. 4 and Fig. 5, after receiving the instantaneous heart rate (HR) from the RRI of the electrocardiogram (S401, S501), the horizontal axis is set to the i+1th value and the vertical axis is set to the i-th heart rate (instantaneous heart rate) value, and a Lorenz plot for a predetermined reference range time (e.g., 360 seconds) is created, and the slope of the regression line (reference slope) of the plotted points is obtained (S402, S502-1). The reference range time is preferably 180 seconds or more, and more preferably 360 seconds. The time series data of physiological indices includes information on the amount of fluctuation, the periodicity of the fluctuation, and the manner and pattern of the fluctuation.

[0035] As mentioned above, the waves in blood vessels are a superposition of waves generated by various factors, and the LF component is by definition 0.04 to 0.15 Hz, but is usually distributed in the frequency band of 0.06 to 0.11 Hz. On the other hand, the very low frequency component (VLF component) defined as 0.003 to 0.04 Hz, which is the long-term heart rate fluctuation, contains information that reflects the regulation of the effective blood volume of the circulatory regulatory system and the health of the entire organism. To capture the frequency component of 0.003 Hz, a minimum measurement time of 1 / 0.003 = 333.4 seconds is required. If the measurement time is 360 seconds, frequency components of 0.0028 Hz or higher can be captured.

[0036] For a time shorter than the reference range time (for example, when the reference range time is 360 seconds, a measurement time shorter than that (for example, 30 seconds)), the Lorentz plot method is applied to similarly plot, and the slope of the regression line of the point cloud plotted in this short time window (short-time point cloud slope) is obtained, and the deviation angle between this short-time point cloud slope (θi) and the reference slope (θ0) is obtained as the relative slope (Δθi) (S502-2 in FIG. 5). As described above, in this embodiment, the pulsation periods one beat before and one beat after are taken on the vertical and horizontal axes, respectively, and when the Lorentz plot falls on a 45-degree straight line drawn from the origin on the graph, this situation indicates that the fluctuation for each beat is zero. Therefore, the deviation angle (relative slope (Δθi)) indicates fluctuation information of the instantaneous heart rate.

[0037] The short-time point cloud slope (θi) is sequentially obtained for each time window by moving average processing. Since heart rate fluctuations respond more quickly to stimulation of the parasympathetic nervous system than to stimulation of the sympathetic nervous system, the time resolution is determined taking into consideration the heart rate regulation function and respiratory regulation by the parasympathetic nervous system, and in this embodiment, the time resolution is set to 3 seconds, and the time window of the moving average is set to 30 seconds. Then, the short-time point cloud slope (θi) and relative slope (Δθi) are obtained for each time window, and a time series of the relative slope (Δθi) is created (S403 in FIG. 4, S503 in FIG. 5). This time series becomes time series data (time series fluctuation data) showing the fluctuation of the heart rate (instantaneous heart rate), that is, the above-mentioned "Δθi function". Note that 30 seconds is twice the average time of 15 seconds, which is the time in which the effect of enhancing and suppressing the sympathetic nervous system is reflected.

[0038] On the other hand, when the biological signal data is APW20, three extreme values ​​(a0 to a2) between atrial contraction and ventricular contraction are extracted from the received APW20 data, and two inter-extreme value amplitudes (a2-a0, a1-a0) are found (S601 to S602 in FIG. 6, S701 to S702 in FIG. 7). These values ​​are plotted on the horizontal and vertical axes to create a Lorentz plot, and the reference slope (S603 in FIG. 6, S703 in FIG. 7) is found as described above. Furthermore, the relative slope (S604 in FIG. 6, S704 in FIG. 7) is found, and a Δθi function is created (S605 in FIG. 6, S705 in FIG. 7).

[0039] In both cases, it is preferable to apply a low-pass filter of 0.08 Hz to the Δθi function. According to the above moving average process, the total measurement time is 360 seconds, the window is 30 seconds, and the slide is 3 seconds, so a point group of 110 points is plotted. Therefore, the frequency resolution (Δf) is (1 / 3) / 110=0.003 Hz. However, by applying the additive average of three points (0 Hz, 0.003 Hz, 0.006 Hz) on the frequency axis, the minimum value of the effective frequency band is 0.006 Hz. In addition, as described later, "non-perturbation" in which non-periodic components are predominant tends to produce white noise, and white noise makes the estimation of correlation inaccurate. Therefore, it is preferable to set a high-pass filter at 0.006 Hz, taking into consideration the separation of non-periodic components and white noise in the frequency band less than 0.02 Hz from the measured biosignal. Also, if the waveform is defined as being identified at five points, the time required for the waveform is 12 seconds. Therefore, 1 / 12 seconds = 0.08 Hz is the cutoff frequency, and it can be said that the waveform with a frequency band of 0.08 Hz or more is unreliable. Therefore, it is preferable to apply a low-pass filter of 0.08 Hz in this way. This makes it possible to remove respiratory fluctuation components in the HF band of 0.15 Hz or more.

[0040] The frequency analysis unit 13 is a means for frequency analysis (S404 in FIG. 4, S504 in FIG. 5) of the Δθi function (applied with a 0.08 Hz low-pass filter), which is the time series data of the fluctuation of the biosignal data obtained from the time series fluctuation data calculation unit 12, and displaying it on a double logarithmic axis (S405 in FIG. 4, S505 in FIG. 5), and capturing the way in which the biosignal data fluctuates. S505 in FIG. 5 shows an example of a double logarithmic graph display of the power spectrum (shown as a plotted group of points) as the frequency analysis result. The blood pressure estimation equation construction unit 14 and the estimated blood pressure value calculation unit 15 perform predetermined processing on the double logarithmic graph display (analysis waveform) obtained by the frequency analysis unit 13.

[0041] The blood pressure estimation equation constructing unit 14 constructs a blood pressure estimation equation using data of multiple subjects. Specifically, as shown in Fig. 2, when measuring the biosignals of the heart rates or acoustic pulse waves of multiple subjects, the blood pressure estimation equation is constructed using actual blood pressure values ​​measured by a cuff-based blood pressure monitor (upper arm blood pressure monitor) and a frequency gradient calculated by a first frequency gradient calculation unit 143 described later. The blood pressure estimation equation constructing unit 14 includes a first boundary frequency calculation unit 141, a first analysis target frequency band selection unit 142, a first frequency gradient calculation unit 143, and an estimation equation calculation unit 144.

[0042] The first boundary frequency calculation unit 141 determines the boundary frequency (BF) between aperiodic components and periodic components for a power spectrum display (analysis waveform) of a double logarithmic graph in a predetermined frequency band ranging from VLF to LF.

[0043] The very low frequency component (VLF component) defined as 0.003 to 0.04 Hz, which is the long-term heart rate fluctuation, is dominated by non-periodic components. The non-periodic components contain information about the "three basic principles of circulatory function." The non-periodic components are expressed as 1 / f n The non-periodic components are fractals with a power spectrum of the same shape, and have self-similarity, meaning that similar shapes appear even when the time scale is changed. Non-periodic components do not have a constant frequency, so no peaks form in the power spectrum.

[0044] In addition, the sympathetic nervous system, mediated by the baroreflex, is mainly involved in the stabilization of blood pressure. The baroreflex acts at frequencies below 0.1 Hz to suppress blood pressure fluctuations. The baroreflex controls blood pressure by changing cardiac function, heart rate, vascular resistance, and effective blood volume, of which the regulation of effective blood volume is the most powerful factor controlling blood pressure. The Mayer wave of blood pressure fluctuations is reflected in heart rate by the baroreceptor reflex. It is known that in subjects with normal blood pressure and high normal blood pressure, the parasympathetic nervous system is dominant in the heart rate regulation function, while in subjects with high blood pressure and hypertension, the sympathetic and sympathetic nervous systems are dominant in the heart rate regulation function, or the sympathetic nervous system is dominant in the heart rate regulation function. The LF component is mediated by both the cardiac parasympathetic and sympathetic nerves.

[0045] On the other hand, in any tissue of the body, blood flow increases with a sudden increase in arterial pressure, but even if arterial pressure remains elevated, it is known that blood flow returns to a nearly normal level in most tissues within 60 seconds (see Guyton and Hall Textbook of Medical Physiology Thirteen edition, John E. Hall, pp. 206-207). From this, when the mechanism of acute autoregulation is examined from the viewpoint of frequency, the fundamental frequency of arterial pressure fluctuation in 60 seconds is 0.02 Hz, since 1 / 60 ≒ 0.02 Hz. It is considered that periodic components of multiples of 0.02 Hz are generated due to the nonlinearity of the mechanisms of acute autoregulation. The frequency of the LF component, which is usually defined as the range of 0.04 to 0.15 Hz in pulse wave measurement, is actually distributed in the frequency band of 0.06 to 0.11 Hz, as mentioned above. Therefore, at frequencies below 0.02 Hz, non-periodic components dominate, at frequencies above 0.06 Hz, periodic components dominate, and between 0.02 and 0.06 Hz, periodic and non-periodic components are mixed. In addition, from the acute curve characteristics of the acute autoregulation mechanisms, blood pressure fluctuates linearly with respect to blood flow rate when the blood pressure is in the range of 80 to 180 mmHg, but fluctuates non-linearly when the blood pressure is below 80 mmHg or above 180 mmHg.

[0046] Here, we will explain the phenomenon in which periodic components are superimposed on non-periodic components using the terms "non-perturbation" and "perturbation." Frequencies of 0.006 Hz or higher are defined as "perturbation," a short-term fluctuating periodic motion in which multiple periodic components appear randomly in both ± directions, while frequencies below 0.006 Hz are defined as "non-perturbation," a baseline-wandering periodic motion that is a bias in a fixed direction that indicates the direction of movement. The boundary frequency (BF) between the non-periodic and periodic components in "non-perturbation" occurs between or near 0.02 to 0.06 Hz, and this BF is 1 / f of the non-periodic components. nThis determines the fluctuation coefficient and also determines the baseline shape of the periodic component fluctuation.

[0047] The first boundary frequency calculation unit 141 finds the boundary frequency (BF) between the non-periodic and periodic components contained in the fluctuation of the autoregulation mechanism. As described above, the fluctuation according to the blood pressure control factor differs depending on the type of blood pressure (normal blood pressure, hypertension, etc.), but by finding the boundary frequency, it is possible to analyze an appropriate frequency band according to the type of blood pressure.

[0048] The results of the manifestation of the blood pressure control function are reflected in the non-periodic and periodic components contained in the acute autoregulation mechanisms, and cause the boundary frequency (BF) to fluctuate. For example, in acute autoregulation, 0.02 Hz is the fundamental frequency unit, but the periodic vasomotion of local blood flow regulation repeats periodic opening and closing several times per minute, fluctuating in the frequency band of 0.02 to 0.05 Hz. From this, it is thought that the BF with the inflection point exists near 0.02 Hz. Therefore, in order to identify the inflection point frequency, an approximation curve expressed by a third-order polynomial (an approximation curve expressed by a third-order polynomial by regression analysis) is created by curve fitting the power spectrum below 0.06 Hz, and the inflection point is calculated by second-order differentiation to obtain the boundary frequency (BF).

[0049] Specifically, as shown in FIG. 8 and FIG. 9, the first boundary frequency calculation unit 141 displays the power spectrum of a double logarithmic graph (analysis waveform = spectrum diagram) (S801, S901), creates an approximation curve represented by a third-order polynomial in the frequency band of 0.006 to 0.06 Hz (S802, S902), or creates an approximation curve represented by a third-order polynomial in the frequency band of 0.01 to 0.06 Hz in order to separate white noise if it is included, extracts the inflection point (S803, S903), and determines the frequency of the inflection point as the boundary frequency (BF) (S804, S904).

[0050] The first analysis target frequency band selection unit 142 selects either the low frequency band or the high frequency band sandwiching the boundary frequency (BF) obtained by the first boundary frequency calculation unit 141 as the analysis target frequency band. The blood pressure estimation equation construction unit 14 constructs a blood pressure estimation equation using the actual blood pressure measurement value as described above in addition to the biosignal data of the heart rate or acoustic pulse wave. When measuring the biosignal data of the heart rate or acoustic pulse wave of multiple subjects, the actual blood pressure measurement value is measured using an upper arm blood pressure monitor, and the actual blood pressure measurement value is known information. Therefore, the analysis frequency band is determined based on the actual blood pressure measurement value. The actual blood pressure measurement value is input to the blood pressure estimation equation construction unit 14 by any input means (keyboard, mouse, etc.) or is input via a communication line from the upper arm blood pressure monitor or a computer that stores the data.

[0051] In this embodiment, in the case of normal blood pressure or high-normal blood pressure (i.e., in the case of blood pressure values ​​below those classified as high blood pressure), a low-frequency band (0.006 Hz (or 0.01 Hz) to BF), in which the effect of non-periodic components is more reflected, is selected as the band to be analyzed. When the actual blood pressure value is classified as high blood pressure or high blood pressure, the sympathetic nerve activity is relatively increased compared to cases below normal blood pressure, and a high-frequency band (BF to 0.06 Hz band), in which the effect of periodic components is more reflected, is selected as the band to be analyzed.

[0052] In this specification, blood pressure classification is based on the classification defined in the "Guidelines for the Treatment of Hypertension 2019 (JSH2019)" (issued by the Japanese Society of Hypertension), and for systolic blood pressure, it is defined as "normal blood pressure: less than 120 mmHg", "high normal blood pressure: 120-129 mmHg", "high blood pressure: 130-139 mmHg", and "hypertension: 140 or higher (stage I hypertension: 140-159, stage II hypertension: 160-179, stage III hypertension: 180 or higher)". However, in this specification, systolic blood pressure of 100 mmHg or lower is classified as "low blood pressure" according to the WHO criteria. Note that the above "normal blood pressure: less than 120 mmHg" and "high normal blood pressure: 120-129 mmHg" in the JSH 2019 version correspond to what was defined as "optimal blood pressure: less than 120 mmHg" and "normal blood pressure: 120-129 mmHg" in the "Hypertension Treatment Guidelines 2014 (JSH2014) version."

[0053] The first frequency gradient calculation unit 143 calculates a frequency gradient from an analysis target frequency band corresponding to the actual measurement value of blood pressure selected by the first analysis target frequency band selection unit 142 in the double logarithmic graph calculated by the frequency analysis unit 13. The first frequency gradient calculation unit 143 includes a first approximation curve setting unit 1431, a first inflection point (stationary point) determination unit 1432, and a first gradient determination unit 1433, as shown in FIG.

[0054] The first approximate curve setting unit 1431 applies an approximate curve represented by a fourth-order polynomial by curve fitting to the display consisting of a group of points on a log-log graph of the analysis target frequency band selected by the first analysis target frequency band selecting unit 142. For example, as the approximate curve represented by a fourth-order polynomial, an approximate curve represented by a fourth-order polynomial is created by regression analysis.

[0055] The first inflection point (stationary point) determination unit 1432 determines whether or not there is a point in the approximate curve represented by the fourth-order polynomial that is both an inflection point and a stationary point, that is, a point at which the curve changes from concave (upward convex) to convex (downward convex) or vice versa, and whose differential coefficient indicates a value close to 0 (referred to as an "inflection point (stationary point)" in this specification). The first approximate curve setting unit 1431 draws an approximate curve of the fourth-order polynomial so that there are two inflection points. It is determined whether or not the slope of the tangent represented by the differential coefficient of the two inflection points of this approximate curve is close to 0. The "value close to 0" varies depending on the type of biosignal, that is, whether it is heart rate (instantaneous heart rate) or APW, but based on an analysis of data from 260 cases of 171 people, it is preferable to set it to "within 0±0.2". From the experimental example described later, it was shown that the estimated blood pressure value shows a value closer to the actual measured blood pressure value when either the gradient of the chord between two inflection points (Gradient of Inflection: GI) or the gradient of the tangent to the inflection point (stationary point) is used as the frequency gradient depending on the presence or absence of an inflection point (stationary point). For this reason, the presence or absence of an inflection point (stationary point) is judged.

[0056] If the first inflection point (stationary point) judging unit 1432 judges that there is no inflection point (stationary point), the first gradient deciding unit 1433 decides that the gradient of the chord between the above two inflection points (GI) is the frequency gradient, and if the first inflection point (stationary point) judging unit 1432 judges that there is an inflection point (stationary point), the first gradient deciding unit 1433 adopts the gradient of the tangent (GT) of the inflection point on the low frequency side of the two inflection points (including the inflection point (stationary point)) as the frequency gradient depending on whether the actual blood pressure measurement value is 130 mmHg or more, if it is less than 130 mmHg, and adopts the gradient of the tangent (GT) of the inflection point on the high frequency side as the frequency gradient if it is 130 mmHg or more. This is based on the results of an experiment using data from 260 cases of 171 people.

[0057] The estimation equation calculation unit 144 uses the heart rate or APW20, which are biosignal data, the actual measured value of blood pressure, and the frequency gradient (GI or GT) calculated by the first frequency gradient calculation unit 143 to calculate a response surface of a three-dimensional coordinate system with these data as its coordinate axes, and from the response surface, calculates an equation showing the correlation between the actual measured value of blood pressure and the frequency gradient (GI or GT) as a blood pressure estimation equation.

[0058] Figures 10(a)-(c) are two-dimensional scatter plots obtained from a total of 171 healthy subjects and outpatients at a cardiology clinic (260 cases). A graph showing the relationship between the heart rate, which is biosignal data, and the actual blood pressure measured using an upper arm blood pressure monitor (upper arm systolic blood pressure) (Figure 10(a)) and a graph showing the relationship between the heart rate and frequency gradient (GI or GT) (Figure 10(b)) were superimposed on the axis representing the coordinates common to both graphs, which represent the heart rate, and the correlation between the frequency gradient (GI or GT) and the actual blood pressure measured (upper arm systolic blood pressure) at the intersection of the two graphs was obtained. Figure 10(c) shows that graph.

[0059] FIG. 10(d) is a three-dimensional response surface combining FIGS. 10(a) to (c). This response surface model is expressed by the following equation.

[0060]

number

[0061] FIG. 10(c) corresponds to a mirror transformation of this three-dimensional response surface, and from FIG. 10(c), the correlation between the actual blood pressure measurement value (brachial systolic blood pressure) and the frequency gradient (GI or GT) is obtained as a linear regression equation: y=27.398x+136.04, which is the blood pressure estimation equation of this embodiment. The coefficient of determination of this linear regression equation R 2 =0.8663.

[0062] The above response surface model means that arterial pressure is expressed by the frequency gradient (GI or GT), which is an index of fluctuation in heart rate and circulation. This is derived as an equation expressing the left ventricular pressure waveform with intracardiac blood flow velocity as a parameter (an equation for calculating the area enclosed by the ventricular pressure-volume loop).

[0063] The left ventricular pressure waveform is drawn using an approximate equation obtained according to Rayleigh for the two-dimensional wave generated on the xy plane using a solution method based on series expansion of complex potential. Note that ε is the ejection fraction, and h is an index (m) that replaces the mechanical energy of blood flow with the height of potential energy. If a is the left ventricular pressure (mmHg) and c (m / sec) is the blood flow velocity, then The approximate equation for the pressure waveform η(x, t) is Represented as TIFF2025068195000003.tif12123.

[0064] The intracardiac blood flow velocity c is TIFF2025068195000004.tif12123The mechanical energy of blood flow, h, is Represented as TIFF2025068195000005.tif12123, If you say TIFF2025068195000006.tif12123, The result is TIFF2025068195000007.tif12123.

[0065] If we integrate equation (4) from -∞ to +∞ with respect to x, the area of ​​the waveform represented by equation (4) is equal to the area of ​​a rectangle with width 2 / κ and left ventricular pressure a, which is the mechanical work performed by each phase of the cardiac cycle. TIFF2025068195000008.tif13142

[0066] Therefore, if the left ventricular pressure a is expressed in terms of κ and mechanical energy h, The result is TIFF2025068195000009.tif14142.

[0067] When the wave height is small, The result is TIFF2025068195000010.tif14142.

[0068] These equations explain the Guyton model and suggest that the left ventricular pressure waveform is expressed as an exponential function of the intracardiac blood flow velocity c. The ventricular pressure value a is expressed as f(x,y), and the heart rate-dependent κ 2 heart rate x i h is an index of cardiac contractility related to the maximum rate of ventricular volume change and intraventricular pressure rise. 3 The gradient of the energy fluctuation is expressed as the circulation fluctuation index GI or GT, i.e., y i Then, the response surface representing the control state of arterial pressure is expressed by the following equation:

[0069] The relationship TIFF2025068195000011.tif14145 holds, and arterial pressure is determined by a two-variable, nth-order polynomial of heart rate and circulatory fluctuation index GI or GT.

[0070] The estimated blood pressure value calculation unit 15 applies the biosignal data (heart rate or APW20) of the estimation subject to the above-mentioned blood pressure estimation equation obtained by the estimation equation calculation unit 144 of the blood pressure estimation equation construction unit 14, and calculates the blood pressure (estimated blood pressure value: eSBP) of the estimation subject at the time of measuring the biosignal data.

[0071] As shown in FIG. 3, the estimated blood pressure value calculation unit 15 has a second frequency gradient calculation unit 153, which calculates a frequency gradient (GI or GT) for estimation. In the blood pressure estimation equation construction unit 14, since the blood pressure is known information, the analysis target frequency band required to calculate the frequency gradient can be set to a high frequency band or a low frequency band sandwiching a boundary frequency (BF) based on the actual measurement value of the blood pressure. However, in the estimation blood pressure value calculation unit 15, since the blood pressure is unknown, the analysis target frequency band cannot be set in the same way as in the blood pressure estimation equation construction unit 14. Therefore, the estimated blood pressure value calculation unit 15 has a second boundary frequency calculation unit 151, a second analysis target frequency band selection unit 152, and an estimation unit 154 in addition to the second frequency gradient calculation unit 153. In addition, the second analysis target frequency band selection unit 152 has a slope ratio calculation unit 1521 and the like that calculates a selection index for the analysis target frequency band instead of the actual measurement value of the blood pressure.

[0072] The second boundary frequency calculation unit 151 has the same configuration as the above-mentioned first boundary frequency calculation unit 141, and obtains the boundary frequency (BF) between aperiodic components and periodic components for the power spectrum display (analysis waveform) of a double logarithmic graph in a predetermined frequency band ranging from VLF to LF. Specifically, for the power spectrum display (analysis waveform) of a double logarithmic graph, a third-order polynomial is created in a frequency band of 0.006 to 0.06 Hz, or, if white noise is included, a third-order polynomial is created in a frequency band of 0.01 to 0.06 Hz to separate it, and the frequency of the inflection point is obtained as the boundary frequency (BF) (see Figs. 8 and 9).

[0073] Once the boundary frequency (BF) is obtained (S1101 in FIG. 11), the second analysis target frequency band selection unit 152 selects whether the low frequency band or the high frequency band sandwiching the boundary frequency (BF) is to be analyzed. The slope ratio calculation unit 1521 obtains a selection index for this analysis target frequency band. Specifically, as shown in FIG. 11 and FIG. 12, in a power spectrum display of a log-log graph of a predetermined frequency band ranging from VLF to LF including the boundary frequency (BF), the slope ratio calculation unit 1521 calculates a selection index for this analysis target frequency band. Specifically, as shown in FIG. 11 and FIG. 12, the slope ratio calculation unit 1521 calculates a selection index for the analysis target frequency band by comparing the slope index (FS) of the regression line of the entire frequency band (0.006 to 0.08 Hz in this embodiment) in which the effects of the non-periodic and periodic components are reflected, and the slope index FS of the regression line of the frequency band in the range above the boundary frequency (BF) (BF to 0.08 Hz in this embodiment) in which the effects of the periodic components are more reflected. BF Ratio to:FS BF The regression line slope ratio, which is / FS, is calculated (S1102, S1103). However, white noise and the power spectrum of the sudden change part of the power spectrum affected by the low-pass filter are excluded.

[0074] The regression line from 0.006 to 0.08 Hz represents fluctuations reflecting the health of the entire organism, including the blood pressure homeostasis function, while the regression line from the boundary frequency (BF) to 0.08 Hz represents fluctuations in neural regulation. In the case of normal blood pressure and high-normal blood pressure, the regression line from 0.006 Hz (or 0.01 Hz) to BF and the regression line from BF to 0.08 Hz are both similar to the regression line from 0.006 to 0.08 Hz, and the regression line slope ratio (FS BF In the case of high blood pressure or hypertension, the power spectrum in the BF to 0.08 Hz band is enhanced and has a slope close to 0 or a positive slope, and the regression line slope ratio (FS BF / FS) is negative. For example, in the example shown in S1202-1 of FIG. 12, FS=-1.0325, and in the example shown in S1202-2, FS BF =-1.0641. In this case, the regression line slope ratio (FS Ratio) (FS BF / FS) = 0.97 (S1203).

[0075] Regression line slope ratio (FS BFThe correlation between the actual measurement of blood pressure (systolic blood pressure in this embodiment) and the mean blood pressure (Frequency Response FS) is measured in advance and stored in the storage unit (DB). Figure 12(b) shows the correlation diagram, which was created using data from 171 people (260 cases). The coefficient of determination R 2 = 0.509, and the regression line slope ratio (FS BF / FS) showed a high correlation. Therefore, the second analysis target frequency band selection unit 152 is provided with a regression line slope ratio (FS BF A provisional estimated value calculation unit 1522 is set to calculate a provisional estimated blood pressure value from the correlation diagram shown in FIG. 12(b) using the regression line slope ratio (FS BF A provisional estimated blood pressure value is calculated using the frequency band (Frequency (Frequency) / FS) (S1104 in FIG. 11). If the provisional estimated blood pressure value is classified as high blood pressure or high blood pressure, the second analysis target frequency band selection unit 152 selects a high frequency band including a boundary frequency (BF) as the analysis target band, and if the provisional estimated blood pressure value is below a blood pressure value classified as high blood pressure, the second analysis target frequency band selection unit 152 selects a low frequency band including a boundary frequency (BF) as the analysis target band (S1105 in FIG. 11).

[0076] Similar to the first frequency gradient calculation section 143, the second frequency gradient calculation section 153 obtains a frequency gradient from the analysis target frequency band corresponding to the actual measurement value of the blood pressure selected by the second analysis target frequency band selection section 152. The second frequency gradient calculation section 153 has a second approximation curve setting section 1531, a second inflection point (stationary point) determination section 1532, and a second gradient determination section 1533 (see FIG. 3).

[0077] The second approximation curve setting unit 1531, the second inflection point (stationary point) determining unit 1532, and the second gradient determining unit 1533 have the same functions as the first approximation curve setting unit 1431, the first inflection point (stationary point) determining unit 1432, and the first gradient determining unit 1433, respectively. That is, the second approximation curve setting unit 1531 applies an approximation curve represented by a fourth-order polynomial by curve fitting to a display consisting of a group of points on a log-log graph of the analysis target frequency band selected by the first analysis target frequency band selecting unit 142, and creates an approximation curve of the fourth-order polynomial by regression analysis, for example, and the second inflection point (stationary point) determining unit 1532 determines the presence or absence of an inflection point (stationary point) that is both an inflection point and a stationary point in the approximation curve represented by the fourth-order polynomial. If the second inflection point (stationary point) judgment unit 1532 judges that an inflection point (stationary point) does not exist, the second gradient determination unit 1533 determines the gradient (GI) of the chord between two inflection points of the approximation curve represented by the fourth-order polynomial as the frequency gradient, and if the second inflection point (stationary point) judgment unit 1532 judges that an inflection point (stationary point) exists, the second gradient determination unit 1533 determines the gradient (GT) of the tangent to one of the two inflection points as the frequency gradient. Which of the two inflection points is to be selected is determined by the provisional estimate calculation unit 1522 based on the regression line gradient ratio (FS BF / FS) is judged in the same manner as in the first inflection point (stationary point) judgment unit 1432, depending on whether the provisional estimated blood pressure value using is 130 mmHg or more. If the provisional estimated blood pressure value is less than 130 mmHg, the slope (GT) of the tangent to the inflection point on the low frequency side is adopted, and if it is 130 mmHg or more, the slope (GT) of the tangent on the high frequency side is adopted.

[0078] The estimation unit 154 substitutes the frequency gradient for estimation obtained by the second gradient determination unit 1533 of the second frequency gradient calculation unit 153 into the above blood pressure estimation formula. This makes it possible to obtain an estimated blood pressure value of the estimation subject only by measuring the heart rate, which is biosignal data, without using an upper arm blood pressure monitor.

[0079] Here, if the actual blood pressure value has been measured in advance, the frequency band to be analyzed can be selected based on the actual measurement value as described above. However, if an estimated blood pressure value is to be obtained, it is necessary to select the frequency band to be analyzed based on some criterion on a double logarithmic graph. BF / FS) is calculated to select the frequency band to be analyzed, but a method that can be used separately from this or together with the regression line slope ratio can be provided in a configuration that includes an outline feature determination unit 160 as shown in FIG. 13.

[0080] The general shape characteristic determination unit 160 determines the general shape characteristic of the power spectrum display (analysis waveform) of the double logarithmic graph for the same frequency band as the first boundary frequency calculation unit 141 and the second boundary frequency calculation unit 151, that is, for the frequency band of 0.006 to 0.06 Hz (0.01 to 0.06 Hz to separate white noise if it is included). The method is to fit a regression line or approximation curve expressed by a first to fourth degree polynomial to the power spectrum display. As a result, it is divided into a first and second degree polynomial type, a third degree polynomial type, and a fourth degree polynomial type. When the shape of each graph was compared with the actual measured blood pressure values ​​for 171 people (260 cases), the first and second degree polynomial types were mainly non-periodic components and were often low blood pressure, the third degree polynomial type was often high blood pressure or high blood pressure, and the fourth degree polynomial type was often normal blood pressure (excluding 100 mmHg or less) or high normal blood pressure. From these results, by judging the power spectrum display (analysis waveform) of the double logarithmic graph, it is possible to broadly classify blood pressure into three types based on its general shape: low blood pressure, normal blood pressure / high normal blood pressure, and high blood pressure / high blood pressure, even if the actual blood pressure value is unknown.

[0081] When the general feature is determined by the general feature determination section 160, the first analysis target frequency band selection section 142 or the second analysis target frequency band selection section 152 selects the analysis target frequency band with reference to the determination result. That is, in the case of a first or second degree polynomial type, the analysis target frequency band is set to a range of 0.006 to BF, in the case of a third degree polynomial type, the analysis target frequency band is set to a range of BF to 0.06 Hz, and in the case of a fourth degree polynomial type, the analysis target frequency band is set to a range of 0.006 to BF.

[0082] However, since the first analysis target frequency band selection section 142 has the actual measured value of the blood pressure known in advance, it does not follow this judgment, but accumulates it as data for improving the accuracy of the judgment result of the general shape characteristics judgment section 160.

[0083] On the other hand, since the blood pressure is at an unknown stage in the second analysis target frequency band selection unit 152, the regression line slope ratio (FS BF / FS) is used to obtain a provisional estimated blood pressure value. Therefore, the judgment result of the general feature judgment unit 160 can be used as a confirmation index of whether the provisional estimated blood pressure value is appropriate. If the provisional estimated blood pressure value and the judgment result of the general feature of the display of the power spectrum correspond to each other, the provisional estimated blood pressure value can be said to be appropriate. If they do not correspond to each other, the method of setting the regression line or the regression curve when judging the regression line slope ratio or the general feature can be reviewed (e.g., whether white noise to be discarded is included in the display of the power spectrum) and recalculation can be performed. If the provisional estimated blood pressure value and the judgment result of the general feature of the display of the power spectrum do not correspond to each other, the frequency gradient can be obtained according to either the provisional estimated blood pressure value or the judgment result of the general feature, or the frequency gradient can be obtained from each of the judgment results.

[0084] The subsequent steps are as described above, and in the blood pressure estimation equation constructing unit 14, the first inflection point (stationary point) determining unit 1432 determines whether or not there is an inflection point (stationary point), and the first gradient determining unit 1433 adopts either the GI or GT frequency gradient. In the estimated blood pressure value calculating unit 15, the second inflection point (stationary point) determining unit 1532 determines whether or not there is an inflection point (stationary point), and the second gradient determining unit 1533 adopts either the GI or GT frequency gradient.

[0085] FIG. 14 shows an example of frequency gradient: GI or GT for six subjects whose actual systolic blood pressure measurements are as follows (from the left side of the figure): 1-1: 76 mmHg (hypotension), 1-2: 101 mmHg (normal blood pressure), 2-3: 144 mmHg (stage I hypertension), 3: 162 mmHg (stage II hypertension), 2-2: 127 mmHg (high normal blood pressure), and 2-1: 116 mmHg (normal blood pressure). As shown in FIG. 15, regarding the power spectrum display (analysis waveform) of a double logarithmic graph obtained by frequency analysis of instantaneous heart rate data of an electrocardiogram waveform, after determining a boundary frequency (BF), the outline feature is determined by the outline feature determination unit 160 (S1501), and according to the boundary frequency (BF), the polynomial type is divided into a linear / quadratic polynomial type, a cubic polynomial type, and a quartic polynomial type (S1502). An approximation curve represented by a quartic polynomial is set for the frequency band to be analyzed according to each polynomial type (S1503), the presence or absence of an inflection point (stationary point) is determined (S1504), and GI or GT is obtained as the frequency gradient.

[0086] Figures 16 to 21 show the process of analyzing the instantaneous heart rate data of the above six subjects using Lorentz plot analysis ((a) of each figure), obtaining time series fluctuation data (Δθi function) ((b) of each figure), performing frequency analysis to obtain a power spectrum display (analysis waveform) of a double logarithmic graph, determining the general shape characteristics ((c) of each figure), determining the boundary frequency (BF), setting a fourth-order polynomial ((d) of each figure), and obtaining the frequency gradient (GI or GT) ((e) of each figure). As a result, the frequency gradients of the six subjects were as follows:

[0087] Frequency gradient of subjects with (1-1: low blood pressure): GT = -2.39 Frequency gradient of subjects (1-2: normal blood pressure): GI=-1.16 Frequency gradient of subjects (2-1: normal blood pressure): GI=-1.13 Frequency gradient of subjects with (2-2: high normal blood pressure): GT = -0.35 Frequency gradient of subjects with (2-3: Grade I hypertension): GT = 0.13 Frequency gradient of subjects with (3: II hypertension): GI=0.68

[0088] From Figures 16 to 21, the fluctuation phenomenon could be observed from the Lorentz plots. This fluctuation (CVI) is age-dependent. In hypotensive and hypertensive subjects, the fluctuations in the Lorentz plots were small, while in normotensive subjects, the fluctuations were both large and small. In normotensive subjects, the fluctuations in the Lorentz plots were small, but the baseline fluctuations in the θi time series data appeared at 0.01 Hz, and in normotensive subjects, the fluctuations in the Lorentz plots were somewhat large, but the baseline fluctuations in the θi time series data appeared at 0.02 Hz. In the former, the heart rate fluctuations were small, and the fluctuations relative to the average heart rate were also small. In the latter, the heart rate fluctuations and the fluctuations relative to the average heart rate were both large. These are thought to suggest the activity of nervous control. The maximum value of the power spectrum appeared at 0.01 to 0.02 Hz, suggesting the involvement of acute autoregulation mechanisms.

[0089] From the θi time series data, the baseline fluctuation (regression curve), fluctuation amplitude, and periodic fluctuation could be captured, and these non-periodic and periodic components were distinguished by boundary frequency (BF). When multiple waves that are considered to be harmonics of the periodic component occurred in the θi time series data, a peak appeared in the spectrogram.

[0090] The regression curve of the spectrum in the frequency band of 0.00-0.06 Hz, which shows the general characteristics of the graph of the double logarithmic spectrum, was expressed by a third-order polynomial in hypertensive subjects. A second peak appeared on the spectrum at frequencies higher than 0.06 Hz. The regression line of low blood pressure was expressed by a first- or second-order polynomial, and the regression curve of normal blood pressure subjects was expressed by a fourth-order polynomial. At higher blood pressures, an inflection point (stationary point) appeared on the fourth-order polynomial. The boundary frequency (BF) appeared near 0.02 Hz.

[0091] Based on the power spectrum display (analysis waveform) of the double logarithmic graph, the blood pressure was divided into the above six categories: (1-1: low blood pressure), (1-2: normal blood pressure), (2-1: normal blood pressure), (2-2: high normal blood pressure), (2-3: stage I hypertension), and (3: stage II hypertension). When this was verified for 171 people (260 cases), as shown in Figure 22, the analysis of variance showed p=2.0×10 -56 It was found that the above six classifications are significant in determining systolic blood pressure.

[0092] 23(a) is a diagram (260 cases, 519 points) showing the correlation between the inflection points of the approximation curve represented by the fourth-order polynomial obtained by the first approximation curve setting unit 1431 of the first frequency gradient calculation unit 143 and the actual measured value of the systolic blood pressure. 2 =0.224. FIG. 23(b) is a diagram (260 cases, 260 points) showing the correlation between the frequency of the inflection point used in the calculation of GI or GT, which is the frequency gradient determined by the first gradient determination unit 1433, and the frequency of the actual measured value of the systolic blood pressure. The coefficient of determination R 2 =0.5218. Both are dependent on the systolic blood pressure, and in particular, Figure 23(b) shows a high correlation.

[0093] 24(a) is a diagram showing the correlation between the estimated blood pressure value (eSBP) calculated by the estimated blood pressure value calculation unit 15 using data on the instantaneous heart rate of 171 people (260 cases) and the actual measured value of the systolic blood pressure (SBP). It shows a high coefficient of determination, indicating that the accuracy of the estimated blood pressure value obtained in this embodiment is high.

[0094] FIG. 24(b) is a cross-tabulation table showing the relationship between the systolic blood pressure of 171 people (260 cases) and whether the systolic gradient determined by the first gradient determining unit 1533 was GI or GT.

[0095] In the case of systolic blood pressure less than 100mmHg classified as low blood pressure, and systolic blood pressure 130mmHg or more classified as high blood pressure and high blood pressure, GT was often output as the frequency gradient, and in the range of normal blood pressure and high normal blood pressure of 100 to 129mmHg, GI was often output as the frequency gradient. From this, it can be seen that 100mmHg and 130mmHg are the boundaries where GT switches to GI and GI switches to GT, and that systolic blood pressure (SBP) and frequency gradient (GI or GT) are significantly related. Regarding the classification of systolic blood pressure (SBP) in the above table, an attempt was made to evaluate whether a significant difference was obtained between GI and GT using Cramer's correlation coefficient, and the correlation coefficient was 0.558, indicating a moderate to slightly strong correlation. Cramer's correlation coefficient is an index that shows the correlation between items on the front and back of a cross-tabulation table, with 0.1 to 0.25 indicating a weak correlation, 0.25 to 0.5 indicating a moderate correlation, and anything over 0.5 indicating a strong correlation.

[0096] Figures 25(a) and (b) show the results of verification of the estimation accuracy of the estimated blood pressure (eSBP) calculated from the instantaneous heart rate data using the above blood pressure estimation formula in the estimated blood pressure calculation unit 15 and the actual brachial systolic blood pressure (SBP) measured at the same time for 22 subjects in their 30s to 70s, different from the 171 subjects (260 cases) used to construct the blood pressure estimation formula in the blood pressure estimation formula construction unit 14. The coefficient of determination R 2 = 0.9387. Cramer's correlation coefficient was 1, with no false negatives or false positives.

[0097] From the above, many people with normal blood pressure have GI fluctuations, while many people with GT fluctuations have abnormal blood pressure, with the boundary being 100mmHg and 130mmHg in systolic blood pressure. The reason for this is not clear at present, but the inventor believes that the following hypothesis regarding non-perturbation and perturbation could be one basis. This is related to the regulatory function of the autonomic nervous system and acute autoregulation mechanisms, and will be explained based on Figure 26. Figure 26 is a repurposed double-well potential system of the stochastic resonance phenomenon, where GI is a bistable system that shows a steady state, and GT is a system due to modulation of the control mechanism, and both can be described by a fourth-order polynomial.

[0098] Humans have eight arterial pressure defense mechanisms, which can be divided into groups that react quickly or act as intermediates on a second-to-minute, minute-to-hour, or day-to-month-to-year basis, and groups that regulate pressure for a long period of time. The baroreceptor feedback mechanism, central nervous system ischemia mechanism, and chemoreceptor mechanism are among those that operate on a second-to-minute or minute-to-minute basis, and in response to a drop in arterial pressure, these neural mechanisms work together to increase venous return, increase heart rate and cardiac contractility, inhibit arterial blood flow by constricting peripheral arterioles, and normalize arterial pressure. In response to a sudden increase in arterial pressure, they work in the opposite direction to lower pressure to the normal level.

[0099] Mechanisms that operate on a scale of minutes or hours include the renin-angiotensin vasoconstriction system, which increases arterial pressure; the vascular stress-relaxation system, which responds to an excessive increase in intravascular pressure by causing blood vessels to remain stretched for several minutes to several hours, resulting in a drop in arterial pressure and a return to normal; and the capillary fluid shift mechanism, which decreases or increases the amount of blood in the circulatory system in response to fluctuations in capillary pressure, causing pressure to decrease or increase. These mechanisms are active for periods ranging from 30 minutes to several hours.

[0100] The kidneys play a major role in the long-term control of arterial pressure, which operates on a scale of days, months, or years, and control is performed to support arterial pressure through the interaction of the renin-angiotensin system, aldosterone, and the renal-humoral system. Arterial pressure control begins as life-sustaining control through neural control, then transitions to an intermediate pressure control mechanism, and finally stabilizes with a long-term arterial pressure control mechanism through the renal-humoral system.

[0101] Stabilization of blood pressure homeostasis is achieved by the 1 / f n The non-periodic components of the acute autoregulation mechanism, the Frank-Starling mechanisms and the Renin-Angiotensin-Aldosterone mechanisms, support the non-perturbation of acute autoregulation mechanisms and constitute the field system of blood pressure control. And the neural and various regulatory factors constitute the perturbation system related to inhibition and enhancement.

[0102] As shown in Figure 27(a), the model diagrams for normal blood pressure are "1-2:GI" and "2-1:GI", and the model diagram for high-normal blood pressure is "2-2:GI". It is considered to be a bistable system consisting of a system in which non-perturbation of the Acute Autoregulation Mechanisms acts and a system of perturbation due to neural and various regulatory factors with different time constants. In a bistable system, non-perturbation due to the Renin-Angiotensin-Aldosterone Mechanisms is reflected in the spectrum of the Acute Autoregulation Mechanisms in the frequency band below 0.04 Hz, and perturbation due to sympathetic and parasympathetic nervous activity is reflected in the 0.006 to 0.08 Hz range, and short- to long-term pressure control acts in an integrated manner. Regulation of the nervous system is relatively fluctuating. Both suppression and enhancement states are alternately repeated, and the 1 / f n The spectrum is dispersed linearly symmetrically with respect to the regression line due to type fluctuation.

[0103] The model diagram for high blood pressure and hypertension is "2-3:GT" or "3:GT", and the 1 / f nIn this system, which has a spectrum where neural regulatory factors strongly act on the fluctuation of the type, an inflection point (stationary point) is thought to appear in the frequency band close to 0.04 Hz. The control of the increase and decrease in blood pressure in arterial pressure fluctuations is antagonistic, and the blood pressure often remains stable at a high level. Neural control is activated, and the 1 / f n The idea is that the neural mechanisms modify the arterial blood pressure control mechanism when superimposed on the regression line due to type fluctuations.

[0104] The model diagram for low blood pressure is "1-1:GT", where neural control acts to suppress and often results in a stable low state. Control is performed so that the arterial pressure does not drop below its current level. The perturbation is small, and the entire system of Acute Autoregulation Mechanisms is unperturbed by the fluctuations in control of the Renin-Angiotensin-Aldosterone Mechanisms. The inflection point is near 0.02 Hz, and the unperturbation is 1 / f n It becomes closer to fluctuation.

[0105] It has been pointed out that the state of each frequency component of VLF, LF, and HF is related to the prognosis of cardiovascular disease, but since the above frequency gradient (GI or GT) is blood pressure dependent, it can be a cardiocirculatory system control factor related to the homeostasis of blood pressure as a factor independent of VLF, LF, and HF. Figure 27(b) is an image showing blood pressure fluctuations in a time series using the analysis results and measurement data.

[0106] In addition, when the instantaneous heart rate is used as the biosignal data, the estimated blood pressure value is the systolic blood pressure, but the above APW20 includes the apical beat, heart sound information, and atrial information (cardiac contractility information) that reflect the cardiac function, and by using an electrocardiogram in combination to identify the time phase, it is possible to estimate not only the systolic blood pressure but also the diastolic blood pressure. Therefore, if the frequency gradient (GI or GT) is obtained using the APW20 as the biosignal data in the same process as above, it is thought that it is possible to individually evaluate the systolic blood pressure, diastolic blood pressure, and even various control functions that maintain human homeostasis. [Explanation of symbols]

[0107] 1. Blood pressure estimation device 11 Biological signal receiving unit 12 Time series fluctuation data calculation section 13 Frequency analysis section 14 Blood pressure estimation formula construction section 141 1st boundary frequency calculation section 142 First analysis target frequency band selection unit 143 First frequency gradient calculation unit 1431 First approximate curve setting section 1432 First inflection point (stationary point) determination section 1433 First Gradient Determination Unit 15. Estimated blood pressure calculation unit 151 2nd boundary frequency calculation section 152 Second analysis target frequency band selection unit 1521 Slope ratio calculation unit 1522 Provisional Estimated Value Calculation Unit 153 Second frequency gradient calculation unit 1531 Second approximate curve setting section 1532 Second inflection point (stationary point) determination section 1533 Second Gradient Determination Unit 154 Estimation Department 160 General shape feature determination unit

Claims

1. A biosignal receiving unit that receives at least one biosignal data of a heart rate and an acoustic pulse wave caused by a left ventricular pressure fluctuation obtained from a body surface; a time-series fluctuation data calculation unit for calculating time-series fluctuation data reflecting time-series fluctuation information of the biological signal data; a frequency analysis unit that performs frequency analysis on the time-series fluctuation data and displays the frequency analysis result in a double logarithmic graph; a blood pressure estimation equation constructing unit that constructs a blood pressure estimation equation using the biosignal data of a plurality of subjects, the log-log graph, and actual blood pressure values ​​measured by an upper arm blood pressure monitor; an estimated blood pressure value calculation unit that applies the biological signal data of the subject to be estimated to the blood pressure estimation formula to obtain an estimated blood pressure value of the subject to be estimated; having The blood pressure estimation equation constructing unit a first boundary frequency calculation unit that calculates a boundary frequency between a non-periodic component and a periodic component while displaying a power spectrum of a predetermined frequency band in a range from VLF to LF in the double logarithmic graph; a first analysis target frequency band selection unit that selects, from the log-log graph, a low frequency band or a high frequency band that sandwiches the boundary frequency as an analysis target frequency band; a first frequency gradient calculation unit that fits an approximation curve represented by a fourth-order polynomial by curve fitting to the display of the power spectrum in the analysis target frequency band, and calculates a frequency gradient of the approximation curve represented by the fourth-order polynomial; an estimation equation calculation unit that calculates an equation showing a correlation between the actual blood pressure value and the frequency gradient from a response surface obtained by plotting the biological signal data, the actual blood pressure value, and the frequency gradient on the coordinate axes of a three-dimensional coordinate system, as the blood pressure estimation equation; A blood pressure estimation device having the above structure.

2. The first boundary frequency calculation unit applies an approximation curve represented by a third-order polynomial by curve fitting to a power spectrum display of a predetermined frequency band ranging from VLF to LF in the double logarithmic graph, and obtains a frequency of an inflection point of the approximation curve represented by the third-order polynomial as a boundary frequency between the non-periodic component and the periodic component. The blood pressure estimation device according to claim 1.

3. The first analysis target frequency band selection unit is If the measured blood pressure value is classified as high blood pressure or high blood pressure, the high frequency band is selected as the band to be analyzed; When the actual measured blood pressure value is less than a blood pressure value classified as high blood pressure, the low frequency band is selected as the band to be analyzed. The blood pressure estimation device according to claim 2.

4. The first frequency gradient calculation unit is a first approximation curve setting unit that determines an approximation curve represented by the fourth-order polynomial by curve fitting for the analysis target frequency band; a first inflection point (stationary point) determination unit that determines whether or not an inflection point (stationary point) that also serves as a stationary point is present among two inflection points of the approximation curve represented by the fourth-order polynomial; a first gradient determination unit that determines, when the inflection point (stationary point) does not exist, a gradient of a chord between two inflection points in the fourth-order polynomial in the frequency band to be analyzed as the frequency gradient, and that determines, when the inflection point (stationary point) exists, a gradient of a tangent to one of the two inflection points as the frequency gradient; 3. The blood pressure estimation device according to claim 2, further comprising:

5. The estimated blood pressure value calculation unit a second boundary frequency calculation unit that calculates a boundary frequency between a non-periodic component and a periodic component while displaying a power spectrum of a predetermined frequency band in a range from VLF to LF for a double logarithmic graph obtained by the time-series fluctuation data calculation unit and the frequency analysis unit using the biological signal data of the subject; a second analysis target frequency band selection unit that selects a low frequency band or a high frequency band that sandwiches the boundary frequency in the double logarithmic graph of the estimation target person as an analysis target frequency band; a second frequency gradient calculation unit that fits an approximation curve represented by a fourth-order polynomial by curve fitting to the display of the power spectrum in the analysis target frequency band, and obtains a frequency gradient of the approximation curve represented by the fourth-order polynomial as a frequency gradient for estimation; an estimation unit that calculates an estimated blood pressure value of the estimation subject by substituting the frequency gradient for estimation obtained by the second frequency gradient calculation unit into the blood pressure estimation formula; have The blood pressure estimation device according to claim 1.

6. The second boundary frequency calculation unit is an approximation curve represented by a third-order polynomial is fitted by curve fitting to a power spectrum display of a predetermined frequency band in a range from VLF to LF in the double logarithmic graph, and a frequency of an inflection point of the approximation curve represented by the third-order polynomial is determined as a boundary frequency between the non-periodic component and the periodic component; The second analysis target frequency band selection unit is In displaying the power spectrum of a predetermined frequency band ranging from VLF to LF including the boundary frequency in the double logarithmic graph, the slope index FS of the regression line of the entire frequency band and the slope index FS of the regression line of the frequency band in the range above the boundary frequency are ＢＦ Ratio to FS ＢＦ A slope ratio calculation unit that calculates a regression line slope ratio, which is / FS; The regression line slope ratio (FS ＢＦ A provisional estimated blood pressure value calculation unit that calculates a provisional estimated blood pressure value using a correlation diagram between the actual blood pressure measurement value and the FS. having If the provisional estimated blood pressure value is classified as high blood pressure or high blood pressure, the high frequency band is selected as an analysis target band; When the provisional estimated blood pressure value is less than a blood pressure value classified as a high blood pressure, the low frequency band is selected as an analysis target band, The second frequency gradient calculation unit is a second approximation curve setting unit that determines an approximation curve represented by the fourth-order polynomial by curve fitting for the analysis target frequency band; a second inflection point (stationary point) determination unit that determines whether or not an inflection point (stationary point) that also serves as a stationary point is present among two inflection points of the approximation curve represented by the fourth-order polynomial; a second gradient determination unit that determines, when the inflection point (stationary point) does not exist, a gradient of a chord between two inflection points in the fourth-order polynomial in the frequency band to be analyzed as the frequency gradient, and that determines, when the inflection point (stationary point) exists, a gradient of a tangent to either of the two inflection points as the frequency gradient; have The blood pressure estimation device according to claim 5.

7. The blood pressure estimation equation constructing unit and a general shape characteristic determining unit that applies a regression line or a regression curve having a high coefficient of determination to a power spectrum display of a predetermined frequency band in the range from VLF to LF including the boundary frequency in the double logarithmic graph, and determines a general shape characteristic of the power spectrum display. have The blood pressure estimation device according to claim 1.

8. The estimated blood pressure value calculation unit and a general shape characteristic determining unit that applies a regression line or a regression curve having a high coefficient of determination to a power spectrum display of a predetermined frequency band in the range from VLF to LF including the boundary frequency in the double logarithmic graph, and determines a general shape characteristic of the power spectrum display. have The blood pressure estimation device according to claim 6.

9. The time series fluctuation data in the time series fluctuation data calculation unit is The Lorentz plot method is applied to the biosignal data, a reference slope is obtained from the Lorentz plot in a predetermined reference time range, and a relative slope, which is the difference between the slope of each Lorentz plot created at each predetermined time shorter than the reference time range and the reference slope, is subjected to moving average processing, and the obtained relative slope is time series data. The blood pressure estimation device according to claim 1.

10. The heart rate is used as the vital sign data, and the estimated blood pressure value is a systolic blood pressure. The blood pressure estimation device according to claim 1.

11. The acoustic pulse wave is used as the biological signal data, and the estimated blood pressure value is a systolic blood pressure or a diastolic blood pressure. The blood pressure estimation device according to claim 1.

12. A computer program for causing a computer to function as a blood pressure estimation device, comprising: receiving at least one of biosignal data of a heart rate and an acoustic pulse wave caused by a change in left ventricular pressure obtained from a body surface; A step of obtaining time-series fluctuation data reflecting time-series fluctuation information of the biosignal data; a step of performing a frequency analysis on the time-series fluctuation data and displaying a result of the frequency analysis in a double logarithmic graph; A step of constructing a blood pressure estimation equation using the biosignal data of a plurality of subjects, the log-log graph, and actual blood pressure values ​​measured by an upper arm blood pressure monitor; applying the biological signal data of the subject to be estimated to the blood pressure estimation formula to obtain an estimated blood pressure value of the subject to be estimated; and causing the computer to execute the steps of: In the step of constructing the blood pressure estimation equation, In the double logarithmic graph, a boundary frequency between a non-periodic component and a periodic component is obtained while displaying a power spectrum of a predetermined frequency band in a range from VLF to LF; In the log-log graph, a low frequency band or a high frequency band sandwiching the boundary frequency is selected as a frequency band to be analyzed; an approximation curve represented by a fourth-order polynomial is fitted to the power spectrum display in the analysis target frequency band by curve fitting, and a frequency gradient of the approximation curve represented by the fourth-order polynomial is obtained; A step of obtaining an equation showing a correlation between the actual blood pressure value and the frequency gradient as the blood pressure estimation equation from a response surface obtained by plotting the biological signal data, the actual blood pressure value, and the frequency gradient on the coordinate axes of a three-dimensional coordinate system. A computer program that causes the computer to execute the above.

13. In the step of determining the boundary frequency, an approximation curve represented by a third-order polynomial is fitted by curve fitting to a power spectrum display of a predetermined frequency band in a range from VLF to LF in the double logarithmic graph, and the frequency of an inflection point of the approximation curve represented by the third-order polynomial is determined as the boundary frequency between the non-periodic component and the periodic component.

13. A computer program according to claim 12.

14. In the step of selecting a frequency band to be analyzed, If the measured blood pressure value is classified as high blood pressure or high blood pressure, the high frequency band is selected as the band to be analyzed; When the actual measured blood pressure value is less than a blood pressure value classified as high blood pressure, the low frequency band is selected as the band to be analyzed.

13. A computer program according to claim 12.

15. In the procedure for calculating the frequency gradient of the approximation curve represented by the fourth-order polynomial, An approximation curve represented by the fourth-order polynomial is obtained by curve fitting for the frequency band to be analyzed; determining whether or not there is an inflection point (stationary point) that also serves as a stationary point in the approximation curve represented by the fourth-order polynomial; a step of determining, when the inflection point (stationary point) does not exist, a gradient of a chord between two inflection points in the fourth-order polynomial in the frequency band to be analyzed as the frequency gradient, and, when the inflection point (stationary point) exists, determining a gradient of a tangent to the inflection point (stationary point) as the frequency gradient. The computer program product according to claim 14, which causes the computer to execute the steps of:

16. In the step of calculating the estimated blood pressure value, determining a boundary frequency between aperiodic components and periodic components while displaying a power spectrum of a predetermined frequency band ranging from VLF to LF for a double logarithmic graph obtained using the biological signal data of the subject; In the log-log graph of the estimated subject, a low frequency band or a high frequency band sandwiching the boundary frequency is selected as an analysis target frequency band; an approximation curve represented by a fourth-order polynomial is fitted to the display of the power spectrum in the analysis target frequency band by curve fitting, and a frequency gradient of the approximation curve represented by the fourth-order polynomial is obtained as a frequency gradient for estimation; The frequency gradient for estimation obtained by the second frequency gradient calculation unit is substituted into the blood pressure estimation formula to obtain an estimated blood pressure value of the estimation subject.

13. A computer program according to claim 12.

17. an approximation curve represented by a third-order polynomial is fitted by curve fitting to a power spectrum display of a predetermined frequency band ranging from VLF to LF in a double logarithmic graph obtained using the biosignal data of the subject, and a frequency of an inflection point of the approximation curve represented by the third-order polynomial is determined as a boundary frequency between the nonperiodic component and the periodic component; In displaying the power spectrum of a predetermined frequency band ranging from VLF to LF including the boundary frequency in the double logarithmic graph, the slope index FS of the regression line of the entire frequency band and the slope index FS of the regression line of the frequency band in the range above the boundary frequency are ＢＦ Ratio to FS ＢＦ / FS, calculate the regression line slope ratio, The regression line slope ratio (FS ＢＦ / FS) and the actual measured blood pressure value to obtain a provisional estimated blood pressure value, If the provisional estimated blood pressure value is classified as high blood pressure or high blood pressure, the high frequency band is selected as an analysis target band; If the provisional estimated blood pressure value is less than a blood pressure value classified as a high blood pressure, the low frequency band is selected as an analysis target band; An approximation curve represented by the fourth-order polynomial is obtained by curve fitting for the frequency band to be analyzed; determining whether or not there is an inflection point (stationary point) that also serves as a stationary point in the approximation curve represented by the fourth-order polynomial; a step of determining, when the inflection point (stationary point) does not exist, a gradient of a chord between two inflection points in the fourth-order polynomial in the frequency band to be analyzed as the frequency gradient, and, when the inflection point (stationary point) exists, determining a gradient of a tangent to the inflection point (stationary point) as the frequency gradient.

17. The computer program product according to claim 16, which causes the computer to execute the steps:

18. A computer-readable recording medium on which the computer program according to claim 11 is recorded.

Images