Hemodynamic index calculation device, computer program, and recording medium

Abstract

The present invention relates to estimating blood pressure with higher accuracy.　By using instantaneous heart rate time-series data, residual information including a frequency transfer function and a residual function can be applied to a log-log spectrum diagram with frequency on the horizontal axis and power spectrum on the vertical axis, and a local approximation curve for determining a frequency gradient can be set to an appropriate frequency band for each data. Therefore, the accuracy of the frequency gradients determined as first and second hemodynamic indices from the respective local approximation curves is improved, and the estimation accuracy of the blood pressure estimated using the frequency gradients is also improved.

Description

Circulatory dynamics index calculation device, computer program, and recording medium

[0001] This invention relates to a technique for estimating blood pressure based on information regarding heart rate.

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

[0003] Specifically, this technique involves capturing biological signals (acoustic pulse waves: APW) from the body surface of the back using a biosignal detection sensor, filtering the time-series waveform within a predetermined frequency band to obtain a filtered waveform that reveals the cardiac cycle, comparing this filtered waveform with electrocardiogram waveform data obtained from an electrocardiograph simultaneously, identifying waveform components from the ventricular filling phase to the isovolumetric contraction phase, then applying the Lorentz plot method to construct a new time series based on the angle difference between the slope of a large number of points plotted over a predetermined measurement time and the slope of a point group over a shorter measurement time, performing frequency analysis on this time-series waveform, displaying the frequency analysis results on a log-log scale, finding a regression line for the range from VLF to LF in the analyzed waveform, defining the slope of this regression line as the fractal slope, and estimating blood pressure from the fractal slope.

[0004] Patent Document 1 uses the fractal slope described above as an indicator of fluctuations in heart rate variability. However, arterial pressure is not regulated by a single pressure control system, but rather by an integrated, multifaceted system that is interconnected. In contrast, the fractal slope is an indicator that shows the characteristics of the analyzed waveform with only one regression line, and it cannot be said to adequately reflect the characteristics of the arterial pressure control system.

[0005] Therefore, the applicant has proposed in Patent Document 2 a technique in which heart rate, acoustic pulse wave, or pulse wave propagation delay time is used, the time waveform data of fluctuations obtained from these measurement data is frequency-analyzed and displayed on a log-log graph, the point cloud of power spectra displayed in the spectral diagram is roughly represented by a quartic function graph, and the slope between inflection points or the slope of the tangent line to the inflection point is used as an index of circulatory dynamics fluctuation.

[0006] Japanese Patent Publication No. 2019-122502, WO2024 / 162226A1

[0007] Guyton Physiology, 13th edition, published March 20, 2018, author John E. Hall, supervised translation by Yoshihiro Ishikawa, et al., publisher: Elsevier Japan Co., Ltd. Kuniaki Otsuka, Shigeko Nakajima, Nagatoku Kikuchi, Circadian variability and 1 / f fluctuation of heart rhythm, BME, 8, 17-21; (1994) Mirei Asano, Relationship between Lyapunov index of heart rate variability and blood pressure and autonomic nervous system indicators, Journal of the Japanese Society of Nursing Science and Technology, 9, 9-20, (2021-2022)

[0008] According to the technology described in Patent Document 2, the characteristics of physiological indicators related to complexity control can be captured more accurately than the technology described in Patent Document 1, which represents them solely by a regression line. As a result, the accuracy of the blood pressure estimate obtained using the resulting hemodynamic fluctuation indicator is also improved. However, higher accuracy in blood pressure estimation is desirable.

[0009] In view of the above, the present invention aims to provide a technology that can determine blood pressure (particularly systolic blood pressure) with higher accuracy.

[0010] To solve these problems, the present invention focuses on the following points. First, as described above, arterial pressure is not regulated by a single pressure control system, but rather by an integrated, multifaceted system of interconnected factors. It transitions from a short-term blood pressure regulation mechanism, which is a nervous regulatory mechanism, to an intermediate pressure control mechanism, and is stabilized by a long-term blood pressure regulation mechanism by the renin-angiotensin system, i.e., the renal-hydraulic pressure control mechanism. In other words, the mechanism of long-term stabilization involves a wide variety of interactions between factors with various regulatory capabilities, including the renal-hydraulic pressure control mechanism and nervous regulatory mechanisms. Among these, the inventors focused on the concept of a non-nerve-dependent intermediate pressure control mechanism (IPCM) mentioned in Non-Patent Document 1.

[0011] Regulatory mechanisms by the renal-hydraulic pressure control system primarily function in the frequency band of 0.006–0.017 Hz, while regulatory mechanisms by the nervous system primarily function in the frequency band of 0.017–0.06 Hz. Intermediate pressure control mechanisms act on both the renal-hydraulic pressure control system and the nervous system. That is, the combined frequency band of both, 0.006–0.06 Hz, is the main frequency band related to arterial pressure control, but intermediate pressure control mechanisms influence each of the above-mentioned regulations across this broad frequency band (global frequency band).

[0012] Incidentally, regarding Guyton's circulatory system model (Guyton model) shown in Non-Patent Document 1, there is the finding that "the endocrine and nervous systems are modified by fluctuations according to probability distributions, which are reflected in heart rate variability according to their respective reaction times (see Non-Patent Document 2), and the fluctuations in heart rate variability are represented by global and local fluctuations (see Non-Patent Document 3)." The renal-body fluid pressure control mechanism and the nervous regulatory mechanism, which are responsible for the action and compensation of the relay pressure control mechanism, are thought to cause fluctuations in the fluctuations of the relay pressure control mechanism (global fluctuations) across the entire global frequency band due to the difference in the degree of excitation of the two. It is thought that by detecting the degree of excitation of each, the fluctuations of the relay pressure control mechanism to each (local fluctuations) can be grasped. Therefore, it is important to distinguish between relay pressure control mechanisms that operate in the frequency band where regulation by the renal-hydraulic pressure control mechanism occurs (referred to as "Long-Term Mechanisms for Arterial Pressure Regulation (LTM)" in this specification) and relay pressure control mechanisms that operate in the frequency band where regulation by the nervous regulatory mechanism occurs (referred to as "Short-Term Mechanisms for Arterial Pressure Regulation (STM)" in this specification) in order to accurately understand hemodynamics.

[0013] However, these two intermediate pressure control mechanisms are not clearly distinguishable by a certain frequency boundary; rather, there is a frequency band in which both are affected (referred to as the "Overlap Band (OLB)" in this specification), and they are loosely separated by this overlap band. Moreover, the frequency band in which the long-term blood pressure regulation mechanism operates, the frequency band in which the short-term blood pressure regulation mechanism operates, and the overlap band are not constant even for the same person, as there are individual differences depending on age, presence or absence of disease, physical condition, etc., and they vary depending on physical condition and measurement timing.

[0014] Therefore, the inventor focused on setting these frequency bands for each data set to be analyzed to improve the accuracy of understanding the circulatory dynamics of the person being measured for blood pressure estimation, thereby improving the accuracy of blood pressure estimation.

[0015] On the other hand, in Non-Patent Document 1, Guyton states regarding arterial pressure control by the nervous system that "any reflex pressure control mechanism will oscillate if the feedback is sufficiently strong and there is a delay between the excitation of baroreceptors and the subsequent pressure response. Vasomotor waves exhibit the characteristics of arterial pressure control by nerve reflexes and operate on the same principles as mechanical and electrical systems. If the feedback gain is large and there is a delay in the reaction time of the arterial pressure control mechanism, oscillations with lower frequencies and larger amplitudes than under normal conditions will occur."

[0016] A response in which the gain is large in a short time and the reaction time is delayed is an impulse response. The action and compensation of the arterial pressure control response performed by the long-term and short-term blood pressure regulatory mechanisms is a step response with a phase difference. The Laplace transform is an extension of the Fourier transform, and the derivative at the inflection point of the frequency transfer function, which is represented on a log-log axis with the time axis changed to the frequency axis, is the frequency ratio ω / ω ０ log, which is expressed on a logarithmic scale. １０ ω / ω ０ This qualitatively correlates with the frequency gradient on the Bode plot, which has a phase lag φ. The fluctuating frequency gradient on the Bode plot is considered to qualitatively coincide with the fluctuations in the damping ratio in the vibration model system due to the biological system and the applied force, and the biological system and the compensating force. The damping ratio is the ratio of the damping coefficient to the critical damping coefficient.

[0017] Furthermore, the Guyton model explains the above-mentioned non-neuronal relay pressure control mechanisms in terms of (1) the renin-angiotensin vasoconstriction system (RAVS) that operates on a minute or hourly basis, (2) the stress-relaxation mechanism of the vasculature (SRMV), and (3) the capillary fluid shift mechanism (CFSM). Of these, (1) RAVS has a clearly defined frequency, but the frequencies of (2) SRMV and (3) CFSM have not been clearly defined. Therefore, it cannot be said that the transient response of the relay pressure control mechanism involved in the short-term blood pressure regulation mechanism, which is a neuronal regulatory mechanism, and the long-term blood pressure regulation mechanism by the renal-hydraulic pressure regulation mechanism has not been sufficiently investigated to date.

[0018] Therefore, the inventors also focused on examining the transient response of these intermediate pressure control mechanisms by applying a vibration model based on Guyton's theory mentioned above.

[0019] That is, the present invention includes a frequency analysis unit that performs frequency analysis on time-series data related to instantaneous heart rate to obtain a spectrum diagram of a logarithmic graph with frequency on the horizontal axis and power spectrum on the vertical axis, a frequency transfer function calculation unit that obtains a frequency transfer function of the spectrum diagram in an arterial pressure-related frequency band related to arterial pressure control, and determines a global approximation curve set by the frequency transfer function as an index indicating the characteristics of a relay control mechanism that affects both renal-fluid pressure control and neural regulation included in the arterial pressure control, a residual information calculation unit that obtains residual information including a residual function indicating the relationship between the residual between the global approximation curve and the power spectrum and frequency, a frequency band setting unit that uses the residual information to obtain three frequency bands: a first effective blood pressure regulation band in which a long-term blood pressure regulation mechanism among the relay control mechanisms that affect renal-fluid pressure control operates effectively, a second effective blood pressure regulation band in which a short-term blood pressure regulation mechanism among the relay control mechanisms that affect neural regulation operates effectively, and an overlap band in which both the long-term blood pressure regulation mechanism and the short-term blood pressure regulation mechanism operate, and a frequency gradient calculation unit that obtains a first circulatory dynamics index in which the frequency gradient of a local approximation curve indicating the distribution state of the power spectrum in the first effective blood pressure regulation band reflects the excitability of the long-term blood pressure regulation mechanism, and obtains a second circulatory dynamics index in which the frequency gradient of a local approximation curve indicating the distribution state of the power spectrum in the second effective blood pressure regulation band reflects the excitability of the short-term blood pressure regulation mechanism.

[0020] The residual information calculation unit preferably includes a process of obtaining, as the residual information, the residual function indicating the distribution state of the plot of the residual with respect to the frequency, a residual average value approximation curve that is an approximation curve of the average value of the residual, and two residual probability distribution curves that approximate, respectively, a residual high-value plot group composed of plots of the residual distributed in a region where the value of the residual is higher than a predetermined value or a residual low-value plot group composed of plots of the residual distributed in a region where the value of the residual is lower than a predetermined value, with respect to the residual average value approximation curve.

[0021] The frequency band setting unit uses the residual information to obtain the upper limit frequency of the long-term blood pressure regulation mechanism and the lower limit frequency of the short-term blood pressure regulation mechanism, and identifies the overlap band based on the upper limit frequency of the long-term blood pressure regulation mechanism and the lower limit frequency of the short-term blood pressure regulation mechanism. It is preferable to identify the frequency band below the lower limit frequency as the first effective blood pressure regulation band and the frequency band above the upper limit frequency as the second effective blood pressure regulation band.

[0022] The frequency band setting unit preferably selects the plot of the residual that determines the upper limit frequency of the long-term blood pressure regulation mechanism and the plot of the residual that determines the lower limit frequency of the short-term regulation mechanism from the residual low-value plot group or the residual high-value plot group.

[0023] The time series data regarding the instantaneous heart rate reconfigures a two-dimensional delay coordinate system by applying a Lorenz plot to the time series data of the instantaneous heart rate, and the reference angle information (θ ０ ), which is the slope of the regression line of the ensemble average in the reference time range, and the individual angle information (θ ｉ ), which is the slope of the regression line of the individual ensemble average of the two-dimensional delay coordinate system in an individual time range shorter than the reference time range. The difference (θ ｉ－０ ) is preferably the time series data of the difference (θ ｉ－０ ) obtained sequentially by a time series data analysis unit that obtains the difference while moving the individual time range.

[0024] The frequency transfer function calculation unit preferably has a global best approximation curve selection unit that obtains two types of functions, a fourth-order function and a fifth-order function, as the frequency transfer function, and selects a global best approximation curve for analysis from the global approximation curves represented by the fourth-order function and the fifth-order function respectively.

[0025] The global best approximation curve preferably has an inflection point and intersection extraction unit that extracts inflection points and intersection points of the tangent lines.

[0026] The frequency band setting unit preferably selects, from among the residual plots belonging to the group of high residual plots or the group of low residual plots, candidates for the residual plot that determines the upper limit frequency of the long-term blood pressure regulation mechanism and the residual plot that determines the lower limit frequency of the short-term regulation mechanism, in order of the frequency closest to the frequency corresponding to the inflection point or the intersection of the tangent lines.

[0027] The frequency band setting unit preferably determines the first and second candidate bands as the first and second effective blood pressure regulation bands, respectively, with respect to the first candidate band of the first effective blood pressure regulation band and the second candidate band of the second effective blood pressure regulation band, which are defined using the candidate residual plots. If the condition is met that the modification frequency showing the peak amplitude of the second local residual function determined for the second candidate band exists within the second candidate band, and the modification frequency showing the peak amplitude of the first local residual function determined for the first candidate band exists outside the second candidate band, then it is preferable to determine the first and second candidate bands as the first effective blood pressure regulation band and the second effective blood pressure regulation band, respectively.

[0028] Furthermore, it is preferable to have a database construction unit that obtains the first and second hemodynamic indices with respect to time-series data of instantaneous heart rates of a predetermined number of subject groups, uses these as explanatory variables, and constructs a database of a three-dimensional response surface model with the measured systolic blood pressure data of each subject group as the objective variable; an estimation index calculation unit that obtains the first and second hemodynamic indices with respect to time-series data of instantaneous heart rates of subjects targeted for blood pressure estimation, and uses these as the first and second hemodynamic indices for estimation, respectively; and a systolic blood pressure estimation unit that applies the first and second hemodynamic indices for estimation to the three-dimensional response surface model constructed by the database construction unit to obtain an estimated value of the systolic blood pressure of the subjects targeted for blood pressure estimation.

[0029] Preferably, if the estimated systolic blood pressure falls within the range of blood pressure abnormalities, including hypertension, the system includes a factor investigation unit that uses time-series data of time periods containing abnormal signs and time-series data of time periods not containing such abnormal signs, and performs analysis on each using the frequency analysis unit, frequency transfer function calculation unit, residual information calculation unit, frequency band setting unit, and frequency gradient calculation unit to calculate and compare estimated systolic blood pressure values, thereby investigating the causes of the blood pressure abnormality.

[0030] Furthermore, the present invention relates to a computer program that causes a computer to function as a hemodynamic index calculation device, comprising: a procedure for frequency analysis of time-series data relating to instantaneous heart rate to obtain a spectral diagram of a log-log graph with frequency on the horizontal axis and power spectrum on the vertical axis; a procedure for obtaining the frequency transfer function of the spectral diagram in the arterial pressure-related frequency band related to arterial pressure control, and obtaining a global approximation curve set by the frequency transfer function as an index showing the characteristics of the relay pressure control mechanism that affects both renal-body fluid pressure control and neurological regulation included in the arterial pressure control; a procedure for obtaining residual information including a residual function showing the relationship between the residual between the global approximation curve and the power spectrum and frequency; and a procedure for obtaining three frequency bands using the residual information: a first effective blood pressure regulation band in which the long-term blood pressure regulation mechanism among the relay pressure control mechanisms that affect renal-body fluid pressure control is effectively in operation, a second effective blood pressure regulation band in which the short-term blood pressure regulation mechanism among the relay pressure control mechanisms that affect neurological regulation is effectively in operation, and an overlap band in which both the long-term blood pressure regulation mechanism and the short-term blood pressure regulation mechanism are in operation. The present invention provides a computer program that causes the computer to perform the following steps: determine the frequency gradient of a local approximation curve showing the distribution state of the power spectrum in the first effective blood pressure regulation band as a first circulatory dynamics index reflecting the degree of excitation of the long-term blood pressure regulation mechanism; and determine the frequency gradient of a local approximation curve showing the distribution state of the power spectrum in the second effective blood pressure regulation band as a second circulatory dynamics index reflecting the degree of excitation of the short-term blood pressure regulation mechanism.

[0031] As the residual information, it is preferable to execute a procedure including a process of obtaining two residual probability distribution curves that approximate, respectively, a residual high-value plot group composed of plots of the residual distributed in a region where the value of the residual is higher than a predetermined value or a residual low-value plot group composed of plots of the residual distributed in a region where the value of the residual is lower than a predetermined value, based on a residual function showing a distribution state of plots of the residual with respect to the frequency, a residual average value approximation curve that is an approximation curve of the average value of the residual, and the residual average value approximation curve.

[0032] It is preferable to execute a procedure of obtaining an upper limit frequency of the long-term blood pressure regulation mechanism and a lower limit frequency of the short-term blood pressure regulation mechanism using the residual information, specifying the overlap band by the upper limit frequency of the long-term blood pressure regulation mechanism and the lower limit frequency of the short-term blood pressure regulation mechanism, and specifying a frequency band below the lower limit frequency as the first effective blood pressure regulation band and a frequency band above the upper limit frequency as the second effective blood pressure regulation band.

[0033] It is preferable to execute a procedure of selecting plots of the residual that determine the upper limit frequency of the long-term blood pressure regulation mechanism and plots of the residual that determine the lower limit frequency of the short-term regulation mechanism from the residual low-value plot group or the residual high-value plot group.

[0034] The time series data regarding the instantaneous heart rate is time series data of a difference (θ ０ ), which is obtained by a time series data analysis unit that sequentially obtains a difference (θ ｉ ) between reference angle information (θ ｉ－０ ) that is a slope of a regression line of an ensemble average in a reference time range and individual angle information (θ ｉ－０ ) that is a slope of a regression line of an individual ensemble average of the two-dimensional delayed coordinate system in an individual time range shorter than the reference time range, while moving the individual time range.

[0035] It is preferable to obtain two types of frequency transfer functions, a quartic function and a quintic function, and then perform a procedure to select the best global approximation curve to be used for analysis from among the global approximation curves represented by the respective quartic and quintic functions.

[0036] It is preferable to perform a procedure to extract inflection points and tangent intersection points for the aforementioned global best approximation curve.

[0037] It is preferable to perform a procedure to select, from among the residual plots belonging to the group of high residual plots or the group of low residual plots, candidates for the residual plot that determines the upper limit frequency of the long-term blood pressure regulation mechanism and the residual plot that determines the lower limit frequency of the short-term regulation mechanism, in order of the frequency closest to the inflection point or the intersection of the tangent lines.

[0038] It is preferable to perform a procedure to determine the first candidate band of the first effective blood pressure regulation band and the second candidate band of the second effective blood pressure regulation band, respectively, with respect to the first candidate band of the first effective blood pressure regulation band and the second candidate band of the second effective blood pressure regulation band, respectively, using the candidate residual plots. If the modification frequency showing the peak amplitude value of the second local residual function obtained for the second candidate band exists within the second candidate band, and the modification frequency showing the peak amplitude value of the first local residual function obtained for the first candidate band exists outside the second candidate band, then the first candidate band and the second candidate band are determined as the first effective blood pressure regulation band and the second effective blood pressure regulation band, respectively.

[0039] Furthermore, it is preferable to perform the following steps: a procedure to obtain the first and second hemodynamic indicators with respect to time-series data of instantaneous heart rates of a predetermined number of subject groups, use these as explanatory variables, and construct a database of a three-dimensional response surface model with the measured systolic blood pressure data of each subject group as the dependent variable; a procedure to obtain the first and second hemodynamic indicators with respect to time-series data of instantaneous heart rates of subjects targeted for blood pressure estimation, and use these as the first and second hemodynamic indicators for estimation, respectively; and a procedure to apply the first and second hemodynamic indicators for estimation to the three-dimensional response surface model constructed by the database construction unit to obtain an estimated value of the systolic blood pressure of the subjects targeted for blood pressure estimation.

[0040] If the estimated systolic blood pressure falls within the range of blood pressure abnormalities, including hypertension, it is preferable to perform a procedure to investigate the cause of the blood pressure abnormality by using the time series data of the time period containing abnormal signs included in the time series data of instantaneous heart rate and the time series data of the time period not containing the abnormal signs, and performing analysis on each using the frequency analysis unit, the frequency transfer function calculation unit, the residual information calculation unit, the frequency band setting unit, and the frequency gradient calculation unit to calculate the estimated systolic blood pressure for each and compare them.

[0041] Furthermore, the present invention provides a computer-readable recording medium on which the above-mentioned computer program is recorded.

[0042] According to the present invention, by using time-series data of instantaneous heart rate and applying residual information including the frequency transfer function and residual function to the spectral diagram of a log-log graph with frequency on the horizontal axis and power spectrum on the vertical axis, a local approximation curve for determining the frequency gradient can be set to an appropriate frequency band for each data point. Consequently, the accuracy of the frequency gradients that serve as the first and second circulatory dynamics indices obtained from this local approximation curve is improved, and the accuracy of blood pressure estimation using these frequency gradients is also improved.

[0043] Figure 1 is a block diagram showing the overall configuration of a hemodynamic index calculation device according to one embodiment of the present invention. Figure 2 is a block diagram showing the configuration of the frequency analysis unit. Figure 3 is a flowchart showing the calculation process in the frequency analysis unit. Figure 4 is a diagram showing a corresponding diagram according to the flowchart in Figure 3. Figure 5 is a flowchart showing the calculation process in the frequency transfer function calculation unit. Figure 6 is a flowchart showing the process from identifying the upper limit of the long-term blood pressure regulation mechanism (LTM) and the lower limit of the short-term blood pressure regulation mechanism (STM) from the spectral diagram. Figure 7 is a diagram showing a corresponding diagram according to the flowchart in Figure 6. Figure 8 is a diagram for explaining the inflection point and the intersection of the tangent lines. Figure 9 is a flowchart showing the process from determining the first and second effective blood pressure regulation bands to determining the first and second hemodynamic indices. Figure 10 is a diagram showing a corresponding diagram according to the flowchart in Figure 9. Figure 11 is a diagram showing an example of a three-dimensional response surface model. Figure 12 is a diagram showing the results of Experimental Example 1. Figure 13 is a diagram showing the results of Experimental Example 2. Figures 14(A) and (B) show the reconstruction of heart rate variability over 360 seconds using a Lorentz plot, and Figure 14(C) shows the mean and standard deviation of the residuals. Figures 15(A) and (B) show 10 cases in which frequency gradients were extracted using the data selected in Figures 14(A) and (B), and Figure 15(C) shows the distribution of intersection points of frequency gradients on contour lines of systolic blood pressure. Figures 16(a) and (b) show how the residual function obtained by the residual information calculation unit 30 varies depending on the measured value of systolic blood pressure. Figure 17 shows various data from a case study that provides clues for treatment planning in hypertensive patients.

[0044] [Hemodynamic Index Calculation Device 1] Figure 1 shows a schematic configuration of the hemodynamic index calculation device 1 of this embodiment. The hemodynamic index calculation device 1 of this embodiment is composed of a computer (the type of computer is not limited and includes personal computers, microcomputers, mobile information terminals such as smartphones, etc.) and performs predetermined processing.

[0045] Specifically, the hemodynamic index calculation device 1 is configured such that a computer program is stored in a storage unit (including recording media such as a hard disk built into the hemodynamic index calculation device 1 which is composed of a computer, as well as various removable recording media and recording media of other computers connected by communication means). This computer program causes the device to execute procedures for the functions of a frequency analysis unit 10, a frequency transfer function calculation unit 20, a residual information calculation unit 30, a frequency band setting unit 40, a frequency gradient calculation unit 50, a database construction unit 60, an estimated index calculation unit 70, and a systolic blood pressure estimation unit 80. The hemodynamic index calculation device 1 can also be realized using an electronic circuit having one or more storage circuits into which the above computer program is incorporated. That is, it can be composed of a processing circuit that functions as a frequency analysis unit 10, a frequency transfer function calculation unit 20, a residual information calculation unit 30, a frequency band setting unit 40, a frequency gradient calculation unit 50, a database construction unit 60, an estimated index calculation unit 70, and a systolic blood pressure estimation unit 80.

[0046] Furthermore, computer programs can be provided by storing them on a recording medium. The recording medium on which the computer program is stored may be a non-transient recording medium. Non-transient recording media are not particularly limited, but examples include flexible disks, hard disks, CD-ROMs, MO (magneto-optical disks), DVD-ROMs, and memory cards. It is also possible to transmit and install computer programs to a computer via a communication line.

[0047] [Overall Image of This Embodiment] First, the overall function of the hemodynamic index calculation device 1 of this embodiment will be explained based on Figure 1. The frequency analysis unit 10 obtains time-series data of instantaneous heart rate using, for example, electrocardiogram waveform data. Next, this time-series data is analyzed and a spectral diagram is obtained, which is displayed as a log-log graph plotting the power spectrum against frequency.

[0048] The frequency transfer function calculation unit 20 uses the frequency transfer function to obtain a global approximation curve in the frequency band related to arterial pressure control (global frequency band) that serves as an indicator of the operating state of the entire relay pressure control mechanism from the obtained spectral diagram. Furthermore, the residual information calculation unit 30 uses residual information, including the residual function, to identify the upper limit of the long-term blood pressure control mechanism (LTM) and the lower limit of the short-term blood pressure control mechanism (STM).

[0049] Next, the frequency band setting unit 40 identifies the interval between the upper frequency limit of the long-term blood pressure regulation mechanism (LTM) and the lower frequency limit of the short-term blood pressure regulation mechanism (STM) as the overlap band (OLB). Furthermore, within the range of the global frequency band, the interval below the lower frequency limit of the short-term blood pressure regulation mechanism (STM) is determined as the first effective blood pressure regulation band (Lower side Band: LSB) of the long-term blood pressure regulation mechanism (LTM), and the interval above the upper frequency limit of the long-term blood pressure regulation mechanism (LTM) is determined as the second effective blood pressure regulation band (Upper side Band: USB).

[0050] The frequency gradient calculation unit 50 recalculates approximation curves (local approximation curves) for the power spectra in the first effective blood pressure regulation band (LSB) and the second effective blood pressure regulation band (USB), and calculates the respective frequency gradients as the first and second hemodynamic indices (GT of LTM, GT of STM).

[0051] The database construction unit 60 measures the systolic blood pressure of a subject using an upper arm blood pressure monitor when instantaneous heart rate data is obtained, collects this measured systolic blood pressure data and first and second hemodynamic indices (GT of LTM, GT of STM) from a predetermined number of subjects, and constructs a database including a three-dimensional response surface model showing their relationship. The estimation index calculation unit 70 determines the first and second hemodynamic indices for estimation for the subject to be estimated, and the systolic blood pressure estimation unit 80 applies these first and second hemodynamic indices for estimation to the three-dimensional response surface model to obtain an estimated value of the systolic blood pressure (eSBP) for the subject to be estimated. The following describes in detail each component that performs each of the above steps.

[0052] [Frequency Analysis Unit 10] As shown in Figure 2, the frequency analysis unit 10 includes a time-series data analysis unit 11 and a spectrum calculation unit 12. (Time-series data analysis unit 11) The time-series data analysis unit 11 includes a time-series data analysis unit 11 that analyzes time-series data of instantaneous heart rate. The time-series data analysis unit 11 reconstructs a two-dimensional delayed coordinate system using a Lorentz plot for the time-series data of instantaneous heart rate (HR = 60 / Δt) and obtains an attractor related to heart rate variability. The time-series data of instantaneous heart rate can be, for example, electrocardiogram data obtained from an electrocardiograph (ECG) (S1 in Figure 3, Figure 4(a)).

[0053] A Lorenz plot uses two points from time-series data of instantaneous heart rate (HR) and plots them while shifting their range. Specifically, the x-axis is HR ｉ＋１ (HR ｉ＋２ , HR ｉ＋３ ...), HR on the y-axis ｉ (HR ｉ＋１ , HR ｉ＋２ ...) are taken, and their intersection points are plotted sequentially. This is done for a reference time range (360 seconds in this embodiment), and the first-order approximation line in the Cartesian coordinate system calculated from the ensemble average of the plots in the reference time range is defined as the principal axis. The inclination of this principal axis with respect to the x-axis is defined as reference angle information (θ ０ ) is calculated as (S2 in Figure 3, Figure 4(b)).

[0054] The time series data analysis unit 11 then reconstructs a two-dimensional delayed coordinate system using a Lorentz plot for individual time ranges shorter than the reference time range (30 seconds in this embodiment), similar to the above, obtains a first-order approximate straight line (median) from the ensemble average of the plots, and uses the individual angle information (θ) to determine its slope with respect to the x-axis. ｉ ) is obtained as. Individual angle information (θ ｉ ) is obtained sequentially while moving individual time ranges (30 seconds) with a predetermined overlap rate (e.g., 90%) (S2 in Figure 3, Figure 4(c)). Next, the time series data analysis unit 11 obtains reference angle information (θ ０ ) and each individual angle information (θ ｉ ) Difference angle information (θ ｉ－０ ) are calculated sequentially, and further, the difference angle information (θ) is calculated. ｉ－０) Time series data (θ ｉ－０ The time-series data is obtained (S3 in Figure 3, Figure 4(d)) and reconstructed in dot product representation. That is, θ ｉ－０ The time-series data is displayed as the general angle between the principal axis and the midline, by converting it into the principal axis angle and the midline angle information over a 30-second period.

[0055] θ ｉ－０ As shown in Figure 4(d), the time-series data uses time on the horizontal axis and angular information of the increasing / decreasing trend on the vertical axis (a + indicates a decreasing trend in instantaneous heart rate and a suppressive trend in the sympathetic nervous system when it is above the main axis, and a - indicates a increasing trend in instantaneous heart rate and a hyperactive trend in the sympathetic nervous system when it is below the main axis), with a difference (θ) calculated every 3 seconds at a 90% overlap rate. ｉ－０ The graph will show the time series plotted by the difference (θ). ｉ－０ The plot of θ is preferably at 3-second intervals. This is the average time that reflects the effects of parasympathetic nervous system activation and deactivation, and 30 seconds is approximately twice the average time of 12.5 seconds (10-15 seconds) that reflects the effects of sympathetic nervous system activation and deactivation. 360 seconds is 1 / 4 of an ultradian cycle of fluctuations in alertness, drowsiness, and mental work capacity. ｉ－０ The spectral diagram described later, obtained using time-series data, shows the global fluctuations of the arterial pressure control mechanism (10 －３ Information regarding baseline fluctuations at the Hz level and local fluctuations (10 －２ This will reflect information regarding vibrations at the Hz level.

[0056] (Spectrum calculation unit 12) The spectrum calculation unit 12 calculates the θ obtained from the time series data analysis unit 11. ｉ－０ This method involves frequency analysis of time-series data and displaying it on a log-log graph with frequency on the horizontal axis and power spectrum on the vertical axis (Figures 3S4, S5, and 4(e)). Figure 4(e) is an example of a log-log graph (spectrum diagram) of the power spectrum as a result of frequency analysis, with numerous points plotted. Figure 4(f) shows RRI data, and Figure 4(g) shows the frequency analysis results. The two extreme values ​​in the spectrum diagram of Figure 4(e) are absent in Figure 4(g), and the Mayer wave, respiration, and heart rate are captured, indicating that these biological signals are maintained with 1 / f fluctuations.

[0057] In this embodiment, frequency analysis was performed with a sampling frequency of 1000 Hz and a time resolution of 0.001 seconds. Therefore, the calculated effective bandwidth is up to 1000 / 2.5 = 400 Hz. Spectral analysis was performed in 2 ｎ This becomes possible with the data. When performing spectral analysis using 360 seconds of time-series data, 2 ７ = 128 is close to this condition. In the moving average of the time series data analysis unit 11, as described above, the 30-second window is shifted by 3 seconds, and the first point becomes available for calculation after 30 seconds. The heart rate regulation function and respiratory regulation function are reflected in the moving average performed at 30-second intervals. Therefore, (360 - 30) / 3 + 1 = 111 data points are output over 360 seconds. In order to perform spectral calculation, 2 ７ Since 128 data points are needed, the missing 17 data points are obtained by folding the data. Assuming that the response to sympathetic nervous system stimulation is captured in 12.5 seconds, a low-pass filter is applied at 1 / 12.5 = 0.08 Hz. This results in 1 / 360 = 0.0028 Hz, but since the power spectrum near 0.0028 Hz contains drift and offset components, a high-pass filter is applied at 0.006 Hz, assuming 0.0028 Hz × 2 = 0.0056 Hz or higher.

[0058] [Frequency Transfer Function Calculation Unit 20] The frequency transfer function calculation unit 20 performs a procedure to find an approximate curve for the arterial pressure-related frequency band related to arterial pressure control. Specifically, as shown in Figure 5, it comprises a global approximate curve setting unit 21, a global best approximate curve selection unit 22, and an inflection point / intersection extraction unit 23. Before explaining these individual components, we will first explain the frequency modifiers that determine the frequency bands to be analyzed, such as the arterial pressure-related frequency band and the frequency bands of the long-term blood pressure regulation mechanism (LTM) and short-term blood pressure regulation mechanism (STM) mentioned above.

[0059] The frequencies that act as regulatory factors determining these frequency bands include the basic period of arterial pressure fluctuations, 60 seconds (1 / 60 ≈ 0.017 Hz), as described in Non-Patent Literature 1, and the recovery time of baroreceptor reflexes, 21 seconds (1 / 21 ≈ 0.0476 Hz), as described in Guyton's prior research. The basic period of arterial pressure fluctuations is 0.017 Hz when linearity is maintained, but when nonlinearity becomes strong in hypotension or hypertension conditions, frequency components of 0.034 Hz, 0.051 Hz, 0.068 Hz, 0.085 Hz, 0.102 Hz, etc., which are harmonic components of 0.017 Hz, are generated.

[0060] Considering the frequencies in order, we have the fundamental frequency of arterial pressure fluctuations at 0.017 Hz, the harmonic component of arterial pressure fluctuations at 0.034 Hz, the baroreceptor reflex at 0.0476 Hz, the harmonic component of arterial pressure fluctuations at 0.051 Hz, the LF component at 0.06 Hz, and the harmonic components of arterial pressure fluctuations at 0.068 Hz, 0.085 Hz, and 0.102 Hz. Although the frequency of the LF component is generally defined as 0.04 to 0.15 Hz, there is evidence that actual measurements mainly show a distribution between 0.06 and 0.11 Hz.

[0061] Therefore, the frequency band up to 0.06 Hz, which includes the fundamental frequency of arterial pressure fluctuations at 0.017 Hz, its harmonic components at 0.034 Hz ​​and 0.051 Hz, and 0.0476 Hz related to baroreceptor reflexes, was defined as the operating band of the short-term blood pressure regulation mechanism (STM).

[0062] On the other hand, frequencies below 0.017 Hz are the frequency band in which aldosterone functions, and are the operating band of the long-term blood pressure regulation mechanism (LTM). When arterial pressure decreases, the sympathetic nervous system is stimulated by the baroreceptor reflex, followed by an increase in aldosterone secretion within minutes, modifying the pressure control characteristics of the renal-humoral mechanism. Fluctuations in the chemical reaction of aldosterone in the renal-humoral mechanism appear in the frequency gradient of the frequency band below 0.017 Hz. To capture the modification by aldosterone secretion, the basic time range in the time-series data analysis unit 11 is set to 360 seconds (1 / 180 ≈ 0.0056 Hz ≥ 0.0028 Hz ≈ 1 / 360) as described above. Then, a high-pass filter is set to 0.006 Hz to capture fluctuations in the renal-humoral mechanism. Furthermore, as mentioned above, a low-pass filter is set to 0.08 Hz to remove respiratory fluctuations, and the time resolution is set to 3 seconds because heart rate responds more quickly to parasympathetic nervous system stimulation than to sympathetic nervous system stimulation.

[0063] Based on these considerations, fluctuations in the long-term blood pressure regulation mechanism (LTM) are examined primarily in the frequency band of 0.006 to 0.017 Hz, and fluctuations in the short-term blood pressure regulation mechanism (STM) are examined primarily in the frequency band of 0.017 to 0.06 Hz. The arterial pressure-related frequency band related to arterial pressure control, which is the subject of analysis in the frequency transfer function calculation unit 20, is set to the combined frequency band of 0.006 to 0.06 Hz, encompassing both the long-term blood pressure regulation mechanism (LTM) and the short-term blood pressure regulation mechanism (STM).

[0064] (Global Approximation Curve Setting Unit 21) The frequency transfer function calculation unit 20 has a global approximation curve setting unit 21 that determines the frequency transfer function in the arterial pressure-related frequency band in the spectrum diagram obtained by the spectrum calculation unit 12 of the frequency analysis unit 10 and sets a global approximation curve using this frequency transfer function. The global approximation curve setting unit 21 determines a frequency transfer function that shows an approximation curve (the approximation curve relating to the entire arterial pressure-related frequency band is called the "global approximation curve") for the frequency band of 0.006 to 0.06 Hz among the plots of the power spectrum shown in the spectrum diagram. That is, the frequency transfer function is a function that shows a curve that approximates the distribution of each plot of the power spectrum by curve fitting in the spectrum diagram where frequency is the horizontal axis and power spectrum is the vertical axis, and is expressed as a fourth-order or fifth-order function. This global approximation curve is a candidate index that shows the characteristics of the transient response by the relay pressure control mechanism that affects both renal-body fluid pressure control and neuronal regulation included in arterial pressure control.

[0065] (Global Best Approximation Curve Selection Unit 22) The frequency transfer function calculation unit 20 further includes a global best approximation curve selection unit 22 and an inflection point / intersection point extraction unit 23. Of these, the global best approximation curve selection unit 22 selects the best approximation curve (global best approximation curve) to be used for analysis from among two types of global approximation curves represented by a quartic function and a quintic function, respectively, as determined by the global approximation curve setting unit 21. In other words, the two types of global approximation curves are candidates for indicators that show the characteristics of the intermediate pressure control mechanism, and from among these, the one corresponding to the indicator is determined as the global best approximation curve. Since the global best approximation curve shows the characteristics of the intermediate pressure control mechanism (IPCM), it will be referred to as the "IPCM approximation curve" below and in the drawings as appropriate.

[0066] As selection criteria, it is preferable to use evaluation metrics related to the residuals between the global approximation curves represented by quartic and quintic functions and the spectral diagram (the difference along the vertical axis between the power spectrum plot and each approximation curve). As evaluation metrics related to residuals, at least one of the following can be used: "the standard deviation (SD) of the residuals obtained for each power spectrum plot in the spectral diagram, the mean value of the absolute values ​​of the residuals (Mean(abs)), the best approximation error (εn), which is the difference between the standard deviation (SD) of the residuals and the mean value of the absolute values ​​of the residuals (Mean(abs)) (SD - Mean(abs)), and the distribution shape of the residuals (whether it is a symmetrical distribution or a distribution that spreads to one side or the other)."

[0067] In this embodiment, the best approximation error (εn) is used as the selection criterion. Specifically, a global approximation curve of either a quartic or quintic function with a best approximation error εn → 0 (in this specification, εn is 0 or closer to 0, and this is referred to as the "latent factor (LF1)") is selected as the global best approximation curve (see S12 in Figure 6 and Figure 7(b)).

[0068] Here, we explain why the best approximation error εn was included in the selection criteria. Let g(x) be the polynomial approximation function that satisfies the normal distribution and is easy to handle numerically in the interval (a,b), and the specified selection {x i} i=1 n The value {gx} i} i=1 n We select an nth-degree polynomial function f(x) for this.

[0069] For a function g(x) that is continuous on the interval (a,b), the k-th degree polynomial f k Created using (x) Let be the best approximation error, and the nth-degree function f(x) is called the best approximation function. Note that εn→0 is equivalent to Weierstrass's theorem.

[0070] Furthermore, the fact that the random variable X follows a normal distribution means that the probability density of the best approximation function f(x) is as follows: This state satisfies the following criteria.

[0071] Based on the fact that a continuous function on a closed interval follows Weierstrass's theorem, a fundamental theorem of approximation theory, and a normal distribution, we select an nth-degree function f(x) and {x}. i} i=1 n From the residuals, which represent the error between the two points, we calculated the standard deviation (SD) of the residuals and the mean value of the absolute values ​​of the residuals (Mean(abs)). We defined SD - Mean(abs) as εn, and used the condition that SD is as small as εn as εn → 0 to select the best approximation curve.

[0072] Here, if the homeostatic function performs a response that gives a large gain to the arterial pressure control mechanism, resulting in a delay in reaction time, in a short time, it becomes an impulse response. If we define the response of arterial pressure control as a unit impulse, the excitation of the feedback function of the pressure control mechanism can be represented by a delta function. Using the delta function, we obtain a weight function for the unit impulse. If we call the Laplace transform of the weight function the transfer function, then the transfer function is the ratio of the Laplace transforms of the output and the input. The Laplace transform of the input (unit impulse) is 1, and the Laplace transform of the response and the transfer function are equal. The transfer function is, It will become.

[0073] Therefore, by displaying the frequency analysis results of the transient response on a log-log axis and calculating the frequency gradient, the frequency transfer function can be estimated. Conversely, by setting the frequency transfer function and determining the frequency gradient, the operating state of the transient response of the intermediate pressure control mechanism can be determined.

[0074] (Inflection Point / Intersection Extraction Unit 23) The inflection point / intersection extraction unit 23 extracts inflection points and intersections of tangents for the global best approximation curve. That is, it differentiates the selected global best approximation curve to calculate the inflection point (IP) and finds the intersection of tangents (IT) to the inflection point. In S12 and S13 of Figure 6, if it is determined that a quintic function has been selected as the global best approximation curve, then IP and IT will be IP1, IT1, IP2, IT2, and IP3 in order from the low frequency side (see Figure 7(b)). If it is determined that a quartic function has been selected as the global best approximation curve, then IP and IT will be IP1, IT1, and IP2 in order from the low frequency side. The method for finding the intersection IT of the tangents to the inflection point is as follows. As shown in Figure 8, the inflection points IP1, IP2, and IP3 are found in the global best approximation curve. IT1 is the intersection of the tangent line a to IP1 and the tangent line b to IP2, and IT2 is the intersection of the tangent line b to IP3 and the tangent line c to IP3.

[0075] Here, we will explain the significance of inflection points and intersection points. First, the gradient of y = f(x) changes, and the gradient of the curve at an inflection point becomes the gradient of the tangent line at that point. The average gradient between two inflection points on the function y = f(x) is expressed in the form of a difference coefficient. The gradient at the limit where this difference becomes infinitely small is the derivative, or differential coefficient, of the function y = f(x) at an inflection point. If we draw a figure with the gradient of the curve y = f(x) as the vertical coordinate y', we obtain the first derivative of the original curve y = f(x), the gradient curve y' = f'(x'). Furthermore, from the derivative of y' = f'(x'), we obtain the second derivative of the original curve y = f(x), y'' = f''(x). Since y'' = 0 at the inflection point of the curve y = f(x), the sign of f''(x) changes before and after the inflection point. The inflection point (IP) is the point where the two adjustment mechanisms, LTM and STM, intersect. Furthermore, the intersection point (IT) where the tangents to the inflection points intersect represents the frequency at which the sign of the gradient curve changes. IP and IT determine the modification frequency of the arterial pressure regulation mechanism and the shape of the curve that reflects its control characteristics. The graph of the quartic function has oblique symmetry and three extrema. In some cases, two extrema are close together, reducing the number of extrema; these points of proximity are also points where the curves defined by the two regulation mechanisms intersect.

[0076] [Residual Information Calculation Unit 30] The residual information calculation unit 30 obtains various information about residuals (residual information) in a frequency-residual correlation diagram with frequency on the horizontal axis and residuals on the vertical axis. As shown in Figure 7(b), the residual is the difference along the axis of the power spectrum between the global best approximation curve and each plot showing the power spectrum, and as shown in Figure 7(c), a diagram with frequency on the horizontal axis and residuals on the vertical axis is a frequency-residual correlation diagram. The residual information is used as an element to detect the frequency band containing information on the long-term blood pressure regulation mechanism (LTM), the frequency band containing information on the short-term blood pressure regulation mechanism (STM), and the overlap band (OLB) where the information of both overlaps, among the global best approximation curve obtained as an indicator of the relay pressure control mechanism in the global frequency band related to arterial pressure control. In this embodiment, the residual information includes the residual function, the residual mean approximation curve, and the residual probability distribution curve.

[0077] (Residual Function) The residual function, as shown in Figure 7(c), is obtained by curve fitting the approximation curve of the residual plot shown in the frequency-residual correlation diagram, and is represented as a third-, fourth-, or fifth-order function. The frequency that shows the maximum value of the residual function, which is the amplitude peak (referred to as the "modification frequency" in this specification), is the frequency at which the long-term blood pressure regulation mechanism (LTM) and the short-term blood pressure regulation mechanism (STM) are particularly active, and the value (magnitude of the amplitude) of this frequency indicates the intensity of their modification.

[0078] The residual information calculation unit 30 calculates three residual functions. The first is a residual function calculated in the global frequency band range of 0.006 to 0.06 Hz (see S13 in Figure 6 and Figure 7(c)), and this residual function indicates the operating state of the relay pressure control mechanism across the entire global frequency band related to arterial pressure control. The second is a residual function indicating the operating state of the long-term blood pressure regulation mechanism (LTM), and the third is a residual function indicating the operating state of the short-term blood pressure regulation mechanism (STM) (see S14 in Figure 6 and Figure 7(d)). However, the second and third residual functions are calculated again after the frequency bands in which the long-term blood pressure regulation mechanism (LTM) and the short-term blood pressure regulation mechanism (STM) operate are set by the frequency band setting unit 40, which will be described later.

[0079] (Mean residual approximation curve) The mean residual approximation curve is an approximation curve that shows the mean of all residuals (see Figure 7(c)).

[0080] (Residual Probability Distribution Curve) As shown in Figure 7(c), the residual probability distribution curve is an approximation curve obtained for each plot group, which is divided into a high residual plot group (high group) consisting of residuals distributed in a region where the residual value is above a predetermined level, and a low residual plot group (low group) consisting of residuals distributed in a region where the residual value is below a predetermined level, with the residual mean approximation curve as the reference point. Note that the region where the residual value is above a predetermined level or below a predetermined level, with the residual mean approximation curve as the reference point, can be set as, for example, the region excluding plots where the difference from the residual mean approximation curve is within ±10% or within ±5%. In this embodiment, the residual high-value plot group and the residual low-value plot group are composed of plots in the region excluding plots within ±5%.

[0081] [Frequency Band Setting Unit 40] The frequency band setting unit 40 has the function of determining three frequency bands: the first effective blood pressure regulation band (LSB) in which the long-term blood pressure regulation mechanism (LTM) operates effectively, the second effective blood pressure regulation band (USB) in which the short-term blood pressure regulation mechanism (STM) operates effectively, and the overlap band (OLB) (see Figure 10).

[0082] The frequency band setting unit 40 determines the upper limit of the long-term blood pressure regulation mechanism (LTM) and the lower limit of the short-term blood pressure regulation mechanism (STM), and uses these upper and lower limits to identify the first effective blood pressure regulation band (LSB), the second effective blood pressure regulation band (USB), and the overlap band (OLB).

[0083] First, the residual plots determining the upper frequency limit of the long-term blood pressure regulation mechanism (LTM) and the residual plots determining the lower frequency limit of the short-term regulation mechanism (STM) are selected from either the low residual plot group or the high residual plot group for which residual probability distribution curves have been obtained. The selection method is as follows.

[0084] A-1: Prerequisites (a): When the local residual function set in the first effective blood pressure regulation band (LSB), defined by the minimum value of the global frequency band at 0.006 Hz and the upper limit frequency of the long-term blood pressure regulation mechanism (LTM), is represented by a third-, fourth-, or fifth-order residual function, the upper limit frequency will be near a minimum value on the higher frequency side of the modifier frequency, which is the frequency of the peak value (maximum value) of the local residual function. (b): When the local residual function set in the second effective blood pressure regulation band (USB), defined by the maximum value of the global frequency band at 0.06 Hz and the lower limit frequency of the short-term blood pressure regulation mechanism (STM), is represented by a third-, fourth-, or fifth-order residual function, the lower limit frequency will be near a minimum value on the lower frequency side of the modifier frequency, which is the frequency of the peak value (maximum value) of the local residual function.

[0085] A-2: Selection of plot group (a): From A-1(a), the residual plot corresponding to the upper limit frequency of the long-term blood pressure regulation mechanism (LTM) is selected from the low residual plot group. (b)-1: From A-1(b), the residual plot corresponding to the lower limit frequency of the short-term blood pressure regulation mechanism (LTM) is selected from the low residual plot group. (b)-2: However, if it is determined that the modification frequency is closer to the second effective blood pressure regulation band (USB) than a predetermined threshold is set based on the distance along the frequency axis from USB to determine whether it is close or not, it is selected from the high residual plot group. This is a constraint based on the rate of change of curvature of each approximation curve near the boundary between the long-term blood pressure regulation mechanism (LTM) and the short-term regulation mechanism (STM).

[0086] A-3: Selection of candidate residual plots (a): From the residual plots belonging to the low residual plot group, select candidates corresponding to the upper limit frequency of the long-term blood pressure regulation mechanism (LTM) in order from the plots closest to the frequency corresponding to the inflection point (IP) or the intersection of the tangent lines (IT). (b)-1: In the case of A-2(b)-1, from the residual plots belonging to the low residual plot group, select candidates corresponding to the lower limit frequency of the short-term blood pressure regulation mechanism (LTM) in order from the plots closest to the frequency corresponding to the inflection point (IP) or the intersection of the tangent lines (IT). (b)-2: In the case of A-2(b)-2, from the residual plots belonging to the high residual plot group, select candidates corresponding to the lower limit frequency of the short-term blood pressure regulation mechanism (LTM) in order from the plots closest to the frequency corresponding to the inflection point (IP) or the intersection of the tangent lines (IT). Once candidate residual plots corresponding to the upper and lower limits have been selected in this manner, the three frequency bands are determined using them. Specifically, this is as follows.

[0087] B-1: Define the first candidate band for the first effective blood pressure regulation band (LSB) and the second candidate band for the second effective blood pressure regulation band (USB) using candidate residual plots. In defining the first and second candidate bands, a combination of candidates is used in which the upper frequency limit is higher than the lower frequency limit. B-2: For the first and second candidate bands, determine the first and second local residual functions, respectively. B-3: Check whether the condition is met that the modification frequency showing the peak value of the amplitude of the second local residual function is within the second candidate band, and the modification frequency showing the peak value of the amplitude of the first local residual function is outside the second candidate band. B-4: The Upper Limit of LTM and Lower Limit of STM used for the first and second candidate bands that satisfy the conditions of B-3 are finally adopted, and each candidate band is determined as the first effective blood pressure regulation band (LSB) and the second effective blood pressure regulation band (USB) (see Figure 7(d)). The interval between the upper and lower frequencies is then determined as the overlap band (OLB) (see S21 in Figure 9 and Figure 10(d)).

[0088] [Frequency Gradient Calculation Unit 50] The frequency gradient calculation unit 50 first uses the log-log graph display (spectrum diagram) of the power spectrum obtained by the spectrum calculation unit 12 to determine the first effective blood pressure regulation band (LSB) and the second effective blood pressure regulation band (USB) according to the residual information described above, and calculates an approximate curve (local approximate curve) again (S22 in Figure 9, Figures 10(b), (c)). In S23 (LF3), it is determined whether or not this local approximate curve can be expressed as a cubic or quartic function. If it cannot be expressed as a cubic or quartic function, the lower limit of the arterial pressure-related frequency band is changed to 0.009 Hz (L1) or 0.012 Hz (L2), or the upper limit of the arterial pressure-related frequency band is changed to 0.064 Hz (U1) (S29 in Figure 9), and the calculations of each step shown in Figures 2 to 7 are performed again (S30 in Figure 9).

[0089] If, in S23 of Figure 9, the local approximation curve is determined to be represented by a cubic or quartic function, this local approximation curve is adopted as the best approximation curve. The frequency gradient of the local approximation curve showing the distribution of the power spectrum in the first effective blood pressure regulation band (LSB) is determined as the first hemodynamic index (GT of LTM), and the frequency gradient of the local approximation curve showing the distribution of the power spectrum in the second effective blood pressure regulation band (USB) is determined as the second hemodynamic index (GT of STM) (S24 of Figure 9). The first hemodynamic index (GT of LTM) is an index that reflects the degree of excitation of the long-term blood pressure regulation mechanism (LTM), and the second hemodynamic index (GT of STM) is an index that reflects the degree of excitation of the short-term blood pressure regulation mechanism (STM).

[0090] Next, as S25 in Figure 9, it is checked whether the value of the second circulatory dynamics index (GT of STM), which is the frequency gradient of the local approximation curve in the second effective blood pressure regulation band (USB), is within the range of ±3 (S25 in Figure 9). If it is not within the range, the process returns to steps S29 and S30 in Figure 9 above. If it is within the range, the first-order approximation line is further determined from the local approximation curve in the first effective blood pressure regulation band (LSB) (S26 in Figure 9, Figure 10(d)). Next, it is determined whether the sign of this first-order approximation curve matches the sign of the frequency gradient of the first effective blood pressure regulation band (LSB) (S27:LF5 in Figure 9). If they do not match, the process returns to steps S9 and S30 in Figure 9. If they do match, the frequency gradients of the first effective blood pressure regulation band (LSB) and the second effective blood pressure regulation band (USB) that satisfy all the above conditions are ultimately determined to be the first circulatory dynamics index (GT of LTM) and the second circulatory dynamics index (GT of STM), respectively (S28 in Figure 9).

[0091] Furthermore, the frequency gradient is determined by the slope of the tangent line when the local approximation curves obtained from the log-log graph (spectral diagram) of the power spectrum obtained by the spectral calculation unit 12 are represented by a cubic function, and when a quartic function is selected, the slope of the double tangent line (the chord connecting the two inflection points) is adopted (see Figures 10(b) and (c)).

[0092] The hemodynamic index calculation device 1 of this embodiment has the function of estimating systolic blood pressure using a first hemodynamic index (GT of LTM) and a second hemodynamic index (GT of STM). Therefore, the hemodynamic index calculation device 1 has a computer program stored in its storage unit that causes it to execute procedures that function as a database construction unit 60, an estimation index calculation unit 70, and a systolic blood pressure estimation unit 80.

[0093] [Database Construction Unit 60] The database construction unit 60 creates a three-dimensional response surface model using a first hemodynamic index (GT of LTM) and a second hemodynamic index (GT of STM) obtained from time-series data of instantaneous heart rate for a predetermined number of subject groups as explanatory variables, and the measured systolic blood pressure data for each of the subject groups as the dependent variable (see Figure 11). As shown in (1) of Figure 11, the individual first hemodynamic index (e.g., GT of LTM = 0.7991), the second hemodynamic index (e.g., GT of STM = -0.0065), and the measured systolic blood pressure (SBP = 137 mmHg) measured by an upper arm blood pressure monitor are plotted on a three-dimensional response surface with GT of LTM, GT of STM, and SBP as the X, Y, and Z axes, respectively, as shown in (2). As a result, as shown in (3), in this example, a three-dimensional response surface model represented by a fourth-order polynomial is constructed. The database construction unit 60 stores the obtained three-dimensional response surface model as a database in the storage unit of the circulatory dynamics index calculation device 1.

[0094] [Estimation Index Calculation Unit 70] The estimation index calculation unit 70 obtains time-series data of the instantaneous heart rate of the person whose blood pressure is to be estimated (blood pressure estimation target subject), analyzes this data, and determines the first hemodynamic index (GT of LTM) and the second hemodynamic index (GT of STM) of the blood pressure estimation target subject, and uses these as the estimation first hemodynamic index (estimation GT of LTM) and estimation second hemodynamic index (estimation GT of STM), respectively. The first hemodynamic index (GT of LTM) and the second hemodynamic index (GT of STM) are obtained by the frequency analysis unit 10, frequency transfer function calculation unit 20, residual information calculation unit 30, frequency band setting unit 40, and frequency gradient calculation unit 50 described above.

[0095] [Systolic Blood Pressure Estimation Unit 80] As shown in Figure 1, the systolic blood pressure estimation unit 80 applies the first estimated circulatory dynamic index (estimated GT of LTM) and the second estimated circulatory dynamic index (estimated GT of STM) obtained by the estimation index calculation unit 70 to a three-dimensional response surface model constructed by the database construction unit 60 to estimate the systolic blood pressure of the subject for blood pressure estimation and outputs the estimated value.

[0096] In the above explanation, the time-series data of instantaneous heart rate analyzed by the time-series data analysis unit 11 is based on data obtained from an electrocardiograph (ECG). However, the system is not limited to this, as long as similar data can be obtained from other devices. For example, a smartwatch, which is a type of electrocardiograph, or other wearable devices can be used. Alternatively, a biosignal detection sensor incorporated in the product name "Sleep Buster" manufactured by Delta Tooling Co., Ltd. captures biosignals (acoustic pulse wave: APW) from the body surface on the back, and time-series waveform data showing a period approximating the R-R interval (RRI) can be used, obtained by filtering the time-series waveform in a predetermined frequency band.

[0097] (Experiment Example 1) The experiment involved subjects aged 20 to 90 years old in a resting seated position with natural breathing. A total of 137 subjects were included: a healthy control group (88 people) who were healthy and not receiving medical treatment, and a group of outpatients (49 people) with cardiovascular diseases such as heart disease, hypertension, gout, diabetes, and dyslipidemia, or their risk factors. The experiment (examination) was performed only once per subject using an electrocardiograph. For the 137 subjects (225 cases), heart rate data was collected for 360 seconds using an electrocardiograph while the subjects were in a resting seated position, and systolic blood pressure (SBP) was measured using an upper arm blood pressure monitor. Measurements were taken at three locations: Delta Touring Co., Ltd., Shiga University of Medical Science, and Fukui Prefectural Saiseikai Hospital. At Delta Touring Co., Ltd. and Shiga University of Medical Science, car seats were installed as experimental equipment. At Fukui Prefectural Saiseikai Hospital, a relaxation chair was used. The experimental equipment, experimenters, subjects, analysts, and seasons in which the experiments were conducted were all different. Delta Tooling Co., Ltd., Shiga University of Medical Science, and Fukui Saiseikai Hospital recorded the average of two measurements for upper arm systolic blood pressure. However, if there was a difference of 5 mmHg or more between the two measurements, measurements were taken up to three or four times, and the average of the two closest measurements was recorded. Delta Tooling Co., Ltd. conducted experiments with multiple measurements on the same subject. In cases where data with a blood pressure fluctuation range exceeding 20 mmHg were measured, the minimum, average, and maximum values ​​were clearly indicated and analyzed.

[0098] The three-dimensional response surface model shown in Figure 11 was created using data from these 137 subjects (225 cases).

[0099] (Experimental Example 2) A 67-year-old subject (121 measurements) whose blood pressure was maintained at normal to high levels through medication was given the cooperation of a blood pressure measurement experiment over one year. The conditions for collecting heart rate data using an electrocardiograph and measuring systolic blood pressure using an upper arm blood pressure monitor were the same as in Experimental Example 1.

[0100] Figure 12 shows the correlation between the estimated systolic blood pressure (eSBP) obtained by the three-dimensional response surface model in Figure 11 and the measured systolic blood pressure (SBP), using the data of the subjects in Experimental Example 1 as the first estimated circulatory dynamic index (estimated GT of LTM) and the second estimated circulatory dynamic index (estimated GT of STM). The coefficient of determination is R ２ The result was 0.7733.

[0101] Here, considering the relationship between GT of LTM and GT of STM in the three-dimensional response surface model of Figure 11, the coefficient of determination is R ２ = 0.0136, indicating no correlation. The coefficient of determination for the relationship between SBP and GT of LTM was R ２ = 0.2065, and the coefficient of determination in the relationship between SBP and GT of STM is R ２ The correlation was 0.3486, indicating only a moderate correlation in both cases. However, when the first circulatory dynamics index (GT of LTM) and the second circulatory dynamics index (GT of STM) were applied as explanatory variables to the three-dimensional response surface model, an estimate of systolic blood pressure with a high correlation to the measured value was obtained.

[0102] Figure 13 shows the correlation between the estimated systolic blood pressure (eSBP) obtained by the three-dimensional response surface model in Figure 11 and the measured systolic blood pressure (SBP), using the data of the subjects in Experimental Example 2 as the first estimated circulatory dynamic index (estimated GT of LTM) and the second estimated circulatory dynamic index (estimated GT of STM). The coefficient of determination is R ２The result was 0.9, indicating an extremely high correlation. Figures 12 and 13 show frequency gradient plots for the 10 cases (a) to (j) shown in Figure 14, which will be described below. Both plots were shown to have statistical properties. In addition, errors occurred in the data near normal blood pressure in both cases, but the errors in Figure 12 were a combination of random and systematic errors, while the errors in Figure 13 were random errors.

[0103] (Discussion 1) Figures 14(A) and (B) are Lorentz plots reconstructing heart rate variability over 360 seconds. Figure 14(A) shows 6 cases selected from 225 cases involving 137 subjects, and Figure 14(B) shows 4 cases extracted from 121 cases involving a 67-year-old subject. These Lorentz plots allowed for classification into discrete and continuous sample spaces. Figure 14(C) shows that the sample space could be represented as discrete and continuous random variables based on the mean and standard deviation of the attractor structure. The gray points in Figure 14(C) represent the plots for all 225 cases. The Lorentz plots demonstrated the ability to capture extreme differences, such as the presence or absence of arrhythmias. Classification into normal heart rate (b)(d)(e)(g)(h)(i)(j) and abnormal heart rate (a)(c)(f) was possible.

[0104] (Discussion 2) Figures 15(A) and (B) show 10 cases in which an IPCM approximation curve was created from the data selected in Figures 14(A) and (B), and the frequency gradient was extracted from the best approximation curves of the local approximation curves for LTM and STM. By converting the dot product information into a frequency gradient via the IPCM approximation curve, residual function, and best approximation curve of the local approximation curve, it was suggested that each case may possess statistical properties. The 10 cases in which the frequency gradient was extracted from the best approximation curves of LTM and STM all had different graph shapes and became independent variables. The frequency gradients of LTM and STM, which became independent variables, were located on the IPCM approximation curve. As shown in Figure 15(C), when the intersection of the frequency gradients was observed via a systolic blood pressure contour plot (SBP contour plot), it was suggested that it may possess statistical properties.

[0105] (Discussion 3) Figure 16 shows how the residual function obtained by the residual information calculation unit 30 varies depending on the measured value of systolic blood pressure. Figure 16(a) shows the analysis results for 6 individuals and 6 cases extracted from 137 individuals and 225 cases. Figure 16(b) shows the results for 1 individual and 4 cases in which the relationship between systolic blood pressure and the residual function within a subject of Experimental Example 2 was analyzed. In Figure 16(a), the frequency of the Lower limit of STM of the OLB varied between 0.018Hz and 0.033Hz, and the frequency of the Upper limit of LTM varied between 0.03Hz and 0.045Hz. The OLB fell within a range of 0.012Hz. In Figure 16(b), the frequency of the Lower limit of STM of the OLB varied between 0.018Hz and 0.027Hz, and the frequency of the Upper limit of LTM varied between 0.026Hz and 0.03Hz. The OLB remained within a range of 0.009 Hz. As blood pressure increased, the OLB narrowed from 0.009 Hz to 0.003 Hz. Except for the hypertensive case in Figure 16(b), the modification frequencies in Figures 16(a) and (b) did not fluctuate significantly, and the range of residual fluctuations was similar. The modification frequencies occurred within 0.01–0.03 Hz for the LSB and within 0.03–0.06 Hz for the USB. In the hypertensive case in Figure 16(b), the modification frequencies for LSB and USB occurred within 0.02–0.03 Hz. The way the two residual functions fluctuate depending on blood pressure may reveal a regularity between arterial pressure and the statistical properties of the residuals. Furthermore, the relative fluctuations between the residual distribution curves may potentially capture the sensitivity of the arterial pressure control response. Therefore, the best approximation curve and frequency gradient calculated from the approximation curve and residual function can serve as indicators for tracking fluctuations in arterial pressure, and it was suggested that the modification frequency of the residual function outside the OLB, which arises from the mutual modification of LTM and STM, may be involved in arterial pressure control.

[0106] (Discussion 4) [Case study providing clues for treatment planning in hypertensive cases] A detailed study was conducted on a hypertensive subject. This subject was noted to have ventricular premature contractions (VPBs), and an analysis was performed using electrocardiogram RRI data in which VPBs regularly occurred between normal heartbeats, and an analysis of electrocardiogram RRI data in which VPBs were not present. The results are shown in Figure 17. In Figure 17, (Yes) indicates the analysis of data in the time period in which VPBs were present, and (No) indicates the analysis of data in the time period in which VPBs were not present. In the electrocardiogram RRI data in Figure 17 (1-1), VPBs appear where indicated by the arrows. (6) shows the results of examining IPCM approximation curves using fourth-order and fifth-order approximation curves for data that did not include VPBs. These were obtained by re-examining the target frequency band for analysis because the arterial pressure-related frequency band did not satisfy each latent factor. The fifth-order approximation curve, which showed the smallest εn of 0.0221, was adopted. In Figure 17, (1-1) and (1-2) show electrocardiogram RRI data, and (2-1) and (2-2) show Lorentz plots. (3) is a figure comparing time-series data for θi-0, and (4) is a figure comparing IPCM approximation curves. (5)(a) is a figure showing the residual mean approximation curve, residual probability distribution curve, and local residual functions in LSB and USB. (5)(b) is a figure showing the best approximation curve of the local approximation curve obtained in LSB and USB, and the frequency gradient (GT of LTM, GT of STM) obtained therefrom.

[0107] By removing the variability in HRV and analyzing the data, the GT of LTM changed from 1.048 to -0.3616, the GT of STM changed from 1.683 to -0.2775, and e SBP changed from 177 mmHg (SBP: 168 mmHg) to 129 mmHg. In the θi-0 time series data with VPB, the reference level is reset to hypertension, which suggests a case of secondary hypertension. That is, the presence or absence of VPB is reflected in the gain of the power spectrum, causing a difference in OLB and resulting in different values ​​for the frequency gradient. Even without VPB, the variance from the residual probability distribution curve and the residual approximation curve is large, but the power spectrum of the LSB in the approximation curve (6) becomes linear, the width of the OLB narrows, and differences in statistical properties occur. The white noise treatment of the LSB increases the gain of the residual distribution curve and flattens the frequency gradient, suggesting the possibility of renal dysfunction as a cause of hypertension.

[0108] Since chronic hypertension is thought to be caused by microvascular and tubulointerstitial damage in the kidneys, which inhibits sodium excretion, and by hormonal factors such as abnormalities in the renin-angiotensin system, this embodiment, which includes LSB as a target for investigation, is effective in investigating the causes of diseases such as hypertension. Specifically, by analyzing and comparing instantaneous heart rate data from time periods showing abnormal signs in electrocardiogram waveform data with instantaneous heart rate data from time periods without abnormal signs, a difference in the estimated systolic blood pressure can be found, which can be used as one piece of data to investigate the causes of blood pressure abnormalities such as hypertension.

[0109] Therefore, it is preferable that the hemodynamic index calculation device 1 has a configuration that, when the estimated value of systolic blood pressure falls within the range of blood pressure abnormalities including hypertension, uses time-series data of time periods containing abnormal signs and time-series data of time periods not containing abnormal signs included in the time-series data of instantaneous heart rate, performs analysis on each using the frequency analysis unit, the frequency transfer function calculation unit, the residual information calculation unit, the frequency band setting unit, and the frequency gradient calculation unit, calculates the estimated value of systolic blood pressure for each, compares them, and has a factor investigation unit that searches for the cause of the blood pressure abnormality.

[0110] 1. Circulatory Dynamics Index Calculation Device 10. Frequency Analysis Unit 20. Frequency Transfer Function Calculation Unit 30. Residual Information Calculation Unit 40. Frequency Band Setting Unit 50. Frequency Gradient Calculation Unit 60. Database Construction Unit 70. Estimation Index Calculation Unit 80. Systolic Blood Pressure Estimation Unit

Claims

1. A frequency analysis unit that performs frequency analysis on time-series data relating to instantaneous heart rate and obtains a spectral diagram of a log-log graph with frequency on the horizontal axis and power spectrum on the vertical axis; a frequency transfer function calculation unit that obtains the frequency transfer function of the spectral diagram in the arterial pressure-related frequency band related to arterial pressure control and obtains a global approximation curve set by the frequency transfer function as an indicator showing the characteristics of the relay pressure control mechanism that affects both renal-body fluid pressure control and neurological regulation included in the arterial pressure control; a residual information calculation unit that obtains residual information including a residual function showing the relationship between the residual between the global approximation curve and the power spectrum and frequency; a frequency band setting unit that uses the residual information to obtain three frequency bands: a first effective blood pressure regulation band in which the long-term blood pressure regulation mechanism among the relay pressure control mechanisms that affect renal-body fluid pressure control is effectively in operation; a second effective blood pressure regulation band in which the short-term blood pressure regulation mechanism among the relay pressure control mechanisms that affect neurological regulation is effectively in operation; and an overlap band in which both the long-term blood pressure regulation mechanism and the short-term blood pressure regulation mechanism are in operation. A hemodynamic index calculation device having a frequency gradient calculation unit that calculates the frequency gradient of a local approximation curve showing the distribution state of the power spectrum in the first effective blood pressure regulation band as a first hemodynamic index reflecting the degree of excitation of the long-term blood pressure regulation mechanism, and calculates the frequency gradient of a local approximation curve showing the distribution state of the power spectrum in the second effective blood pressure regulation band as a second hemodynamic index reflecting the degree of excitation of the short-term blood pressure regulation mechanism.

2. The circulatory dynamics index calculation device according to claim 1, wherein the residual information calculation unit includes a process to obtain, as residual information, a residual function that shows the distribution state of residual plots with respect to the frequency, a residual mean approximation curve which is an approximation curve of the mean value of the residuals, and two residual probability distribution curves that approximate a group of residual high value plots consisting of residual plots distributed in a region where the residual values ​​are higher than or equal to a predetermined value, or a group of residual low value plots consisting of residual plots distributed in a region where the residual values ​​are lower than or equal to a predetermined value, based on the residual mean approximation curve.

3. The hemodynamic index calculation device according to claim 1, wherein the frequency band setting unit uses the residual information to determine the upper limit frequency of the long-term blood pressure regulation mechanism and the lower limit frequency of the short-term blood pressure regulation mechanism, identifies the overlap band based on the upper limit frequency of the long-term blood pressure regulation mechanism and the lower limit frequency of the short-term blood pressure regulation mechanism, identifies the frequency band below the lower limit frequency as the first effective blood pressure regulation band, and identifies the frequency band above the upper limit frequency as the second effective blood pressure regulation band.

4. The hemodynamic index calculation device according to claim 3, wherein the frequency band setting unit selects a residual plot for determining the upper limit frequency of the long-term blood pressure regulation mechanism and a residual plot for determining the lower limit frequency of the short-term regulation mechanism from the group of low residual plots or the group of high residual plots.

5. The time series data relating to the instantaneous heart rate is obtained by applying a Lorentz plot to the time series data of the instantaneous heart rate to reconstruct a two-dimensional delayed coordinate system, and the reference angle information (θ) is the slope of the regression line of the ensemble mean in the reference time range. ０ ) and individual angular information (θ), which is the slope of the regression line of the individual ensemble average of the two-dimensional delayed coordinate system in an individual time range shorter than the reference time range. ｉ ) difference (θ ｉ－０ The difference (θ) obtained by the time series data analysis unit, which sequentially determines the difference (θ) while moving the individual time ranges, is obtained by the time series data analysis unit. ｉ－０ A circulatory dynamics index calculation device according to claim 1, which is time-series data of ).

6. The circulatory dynamics index calculation device according to claim 1, wherein the frequency transfer function calculation unit determines two types of frequency transfer functions, a quartic function and a quintic function, and has a global best approximation curve selection unit that selects the global best approximation curve to be used for analysis from among the global approximation curves represented by the respective quartic and quintic functions.

7. The circulatory dynamics index calculation device according to claim 6, further comprising an inflection point / intersection extraction unit for extracting inflection points and tangent intersection points for the global best approximation curve.

8. The hemodynamic index calculation device according to claim 7, wherein the frequency band setting unit selects from among the residual plots belonging to the group of high residual plots or the group of low residual plots, in order of proximity to the frequency corresponding to the inflection point or the intersection of the tangents, as candidates for the residual plot that determines the upper limit frequency of the long-term blood pressure regulation mechanism and the residual plot that determines the lower limit frequency of the short-term regulation mechanism.

9. The hemodynamic index calculation device according to claim 8, wherein the frequency band setting unit determines a first local residual function and a second local residual function with respect to a first candidate band of the first effective blood pressure regulation band and a second candidate band of the second effective blood pressure regulation band, respectively, with respect to the candidate residual plots, and if the condition is met that a modification frequency indicating the peak value of the amplitude of the second local residual function determined with respect to the second candidate band exists within the second candidate band and a modification frequency indicating the peak value of the amplitude of the first local residual function determined with respect to the first candidate band exists outside the second candidate band, the first candidate band and the second candidate band are determined as the first effective blood pressure regulation band and the second effective blood pressure regulation band, respectively.

10. The hemodynamic index calculation device according to claim 1, further comprising: a database construction unit that obtains the first hemodynamic index and the second hemodynamic index with respect to time-series data of instantaneous heart rate of a predetermined number of subject groups, uses these as explanatory variables, and constructs a database of a three-dimensional response surface model with the measured systolic blood pressure data of each subject group as the objective variable; an estimation index calculation unit that obtains the first hemodynamic index and the second hemodynamic index with respect to time-series data of instantaneous heart rate of subjects targeted for blood pressure estimation, and uses these as the first and second hemodynamic indexes for estimation, respectively; and a systolic blood pressure estimation unit that applies the first and second hemodynamic indexes for estimation to the three-dimensional response surface model constructed by the database construction unit to obtain an estimated value of the systolic blood pressure of the subjects targeted for blood pressure estimation.

11. The hemodynamic index calculation device according to claim 10, further comprising a factor investigation unit that, when the estimated systolic blood pressure falls within the range of blood pressure abnormalities including hypertension, uses time-series data of time periods containing abnormal signs included in the time-series data of instantaneous heart rate and time-series data of time periods not containing such abnormal signs, performs analysis on each using the frequency analysis unit, the frequency transfer function calculation unit, the residual information calculation unit, the frequency band setting unit, and the frequency gradient calculation unit, respectively, to calculate and compare the estimated systolic blood pressure values, thereby investigating the factors causing the blood pressure abnormality.

12. A computer program that causes a computer to function as a hemodynamic index calculation device, comprising: a procedure for frequency analysis of time-series data relating to instantaneous heart rate to obtain a spectral diagram of a log-log graph with frequency on the horizontal axis and power spectrum on the vertical axis; a procedure for obtaining the frequency transfer function of the spectral diagram in the arterial pressure-related frequency band related to arterial pressure control, and obtaining a global approximation curve set by the frequency transfer function as an index showing the characteristics of the relay pressure control mechanism that affects both renal-body fluid pressure control and neurological regulation included in the arterial pressure control; a procedure for obtaining residual information including a residual function showing the relationship between the residual between the global approximation curve and the power spectrum and frequency; and a procedure for obtaining three frequency bands using the residual information: a first effective blood pressure regulation band in which the long-term blood pressure regulation mechanism among the relay pressure control mechanisms that affect renal-body fluid pressure control is effectively in operation, a second effective blood pressure regulation band in which the short-term blood pressure regulation mechanism among the relay pressure control mechanisms that affect neurological regulation is effectively in operation, and an overlap band in which both the long-term blood pressure regulation mechanism and the short-term blood pressure regulation mechanism are in operation. A computer program that causes the computer to perform the following steps: determine the frequency gradient of a local approximation curve showing the distribution state of the power spectrum in the first effective blood pressure regulation band as a first circulatory dynamics index reflecting the degree of excitation of the long-term blood pressure regulation mechanism; and determine the frequency gradient of a local approximation curve showing the distribution state of the power spectrum in the second effective blood pressure regulation band as a second circulatory dynamics index reflecting the degree of excitation of the short-term blood pressure regulation mechanism.

13. The computer program according to claim 12, which causes the computer program to execute a procedure that includes a residual function showing the distribution state of residual plots with respect to the frequency, a residual mean approximation curve which is an approximation curve of the mean value of the residuals, and two residual probability distribution curves that approximate a group of residual high value plots consisting of residual plots distributed in a region where the residual values ​​are higher than or equal to a predetermined value, or a group of residual low value plots consisting of residual plots distributed in a region where the residual values ​​are lower than or equal to a predetermined value, based on the residual mean approximation curve.

14. A computer program according to claim 12, which uses the residual information to determine the upper limit frequency of the long-term blood pressure regulation mechanism and the lower limit frequency of the short-term blood pressure regulation mechanism, identifies the overlap band based on the upper limit frequency of the long-term blood pressure regulation mechanism and the lower limit frequency of the short-term blood pressure regulation mechanism, and identifies the frequency band below the lower limit frequency as the first effective blood pressure regulation band and the frequency band above the upper limit frequency as the second effective blood pressure regulation band.

15. The computer program according to claim 14, which causes the computer program to perform a procedure to select a residual plot that determines the upper limit frequency of the long-term blood pressure regulation mechanism and a residual plot that determines the lower limit frequency of the short-term regulation mechanism from the group of low residual plots or the group of high residual plots.

16. The time series data relating to instantaneous heart rate is obtained by applying a Lorentz plot to the time series data of instantaneous heart rate to reconstruct a two-dimensional delayed coordinate system, and the reference angle information (θ) is the slope of the regression line of the ensemble mean in the reference time range. ０ ) and individual angular information (θ), which is the slope of the regression line of the individual ensemble average of the two-dimensional delayed coordinate system in an individual time range shorter than the reference time range. ｉ ) difference (θ ｉ－０ The difference (θ) obtained by the time series data analysis unit, which sequentially determines the difference (θ) while moving the individual time ranges, is obtained by the time series data analysis unit. ｉ－０ The computer program according to claim 12, which is time-series data of ).

17. The computer program according to claim 12, which causes the program to obtain two types of frequency transfer functions, a quartic function and a quintic function, and to perform a procedure to select the best global approximation curve to be used for analysis from among the global approximation curves represented by the quartic function and the quintic function, respectively.

18. The computer program according to claim 17, which causes the computer program to perform a procedure for extracting inflection points and intersection points of tangents to the global best approximation curve.

19. The computer program according to claim 18, which causes the computer program to perform a procedure to select, from among the residual plots belonging to the group of high residual plots or the group of low residual plots, candidates for a residual plot that determines the upper limit frequency of the long-term blood pressure regulation mechanism and a residual plot that determines the lower limit frequency of the short-term regulation mechanism, in order of the frequency closest to the frequency corresponding to the inflection point or the intersection of the tangent lines.

20. The computer program according to claim 19, which causes the program to perform a procedure to determine the first candidate band of the first effective blood pressure regulation band and the second candidate band of the second effective blood pressure regulation band, respectively, with respect to the first candidate band of the first effective blood pressure regulation band and the second candidate band of the second effective blood pressure regulation band, with respect to the first candidate band of the first effective blood pressure regulation band and the second candidate band of the second effective blood pressure regulation band, respectively, if the condition is met that the modification frequency showing the peak value of the amplitude of the second local residual function obtained with respect to the second candidate band is within the second candidate band and the modification frequency showing the peak value of the amplitude of the first local residual function obtained with respect to the first candidate band is outside the second candidate band.

21. The computer program according to claim 12, further comprising: a procedure for obtaining the first hemodynamic index and the second hemodynamic index with respect to time-series data of instantaneous heart rate of a predetermined number of subject groups, using these as explanatory variables, and constructing a database of a three-dimensional response surface model with the measured systolic blood pressure data of each subject group as the dependent variable; a procedure for obtaining the first hemodynamic index and the second hemodynamic index with respect to time-series data of instantaneous heart rate of a subject to be blood pressure estimated, using these as the first hemodynamic index and the second hemodynamic index for estimation, respectively; and a procedure for applying the first hemodynamic index for estimation and the second hemodynamic index for estimation to the three-dimensional response surface model constructed by the database construction unit to obtain an estimated value of the systolic blood pressure of the subject to be blood pressure estimated.

22. The computer program according to claim 21, which, when the estimated systolic blood pressure falls within the range of blood pressure abnormalities including hypertension, uses time-series data of time periods containing abnormal signs and time-series data of time periods not containing such abnormal signs, and performs analysis on each using the frequency analysis unit, the frequency transfer function calculation unit, the residual information calculation unit, the frequency band setting unit, and the frequency gradient calculation unit to calculate and compare the estimated systolic blood pressure values ​​for each, thereby investigating the cause of the blood pressure abnormality.

23. A computer-readable recording medium on which a computer program according to any one of claims 12 to 22 is recorded.

Images