JP7221195B2 - Heart rate variability analyzer, method and program - Google Patents

Description

TECHNICAL FIELD The present invention relates to a heart rate variability analysis device, method and program, and more particularly to a heart rate variability analysis device, method and program capable of analyzing heart rate variability from electrocardiogram data and correctly evaluating the working condition of the autonomic nerves. .

It is known that the R-wave generation interval (RR interval) representing the start of ventricular contraction of heartbeat varies periodically, and the frequency component of the variation indicates the working condition of the autonomic nerves. This relationship can be used to evaluate the working condition of the autonomic nerves, and one method is to sequentially measure the RR intervals in the electrocardiographic data, frequency analyze the RR intervals (spectrum analysis), and change the is obtained, and the working condition of the autonomic nerve is evaluated from the frequency component.

FIG. 6 is an explanatory diagram of a conventional method for analyzing frequency components of RR interval fluctuations.

FIG. 1(a) shows electrocardiographic data EPG acquired by an electrocardiographic sensor or the like. Here, the R wave represents the start of ventricular contraction and appears with RR intervals RR1, RR2, RR3, .

In general, it is difficult to perform frequency analysis directly on data along the time axis. Therefore, when performing frequency analysis of RR interval fluctuations, time-series data is generated by replacing RR intervals RR1, RR2, RR3, . . . with amplitudes in the y-axis direction. frequency analysis.

4B shows time series data generated by replacing the RR intervals RR1, RR2, RR3, . . . with amplitudes A1, A2, A3, . . . in the y-axis direction. , A2, A3, . . . are equal to the RR intervals RR1, RR2, RR3, .

The figure (c) shows the power spectral density obtained by frequency analysis of the time-series data shown in the figure (b), the x-axis is the frequency (Hz), and the y-axis is the x-axis direction. Power spectral density (msec ² /Hz) at each frequency is shown. Note that the power spectral density is obtained by decomposing the power of the time-series data into each frequency component.
Here, LF is the sum of 0.04-0.15 Hz components, and HF is the sum of 0.15-0.40 Hz components. These values can be used as autonomic nerve indices.

In FIG. 6, time-series data is generated by replacing the RR interval in the x-axis direction with the amplitude in the y-axis direction as it is, but the reciprocal of the RR interval in the x-axis direction is replaced with the amplitude in the y-axis direction. or generate time-series data represented by a Beat function by setting constants in the x-axis direction at regular intervals and replacing the RR interval in the x-axis direction with the amplitude in the y-axis direction for each constant. It is also known to Hereinafter, the methods for generating the above time series data are referred to as Equal, 1/f fluctuation, and Beat Function, respectively.

FIG. 7 is an explanatory diagram showing a conventional method for generating time-series data, in which (a), (b), and (c) respectively show the Equa1, 1/f fluctuation and Beat Function methods. .
In Equa 1, as shown in FIG. 1(a), the RR interval (tn+1-tn) (msec) in the x-axis direction is changed to the R− Replaced by the amplitude in the y-axis direction in the R interval period. Therefore, the amplitude in the y-axis direction during the RR interval from tn (msec) to tn+1 (msec) in the x-axis direction is equal to the RR interval (tn+1-tn) (msec) in the x-axis direction.
In the 1/f fluctuation, as shown in FIG. 1(b), the reciprocal of the RR interval (tn+1-tn) (msec) in the x-axis direction, 1/(tn+1-tn) (msec ⁻¹ ) is Replace with the amplitude in the y-axis direction. Therefore, the amplitude in the y-axis direction in the RR interval period of tn (msec) to tn+1 (msec) in the x-axis direction is the reciprocal of the RR interval (tn+1-tn) (msec) in the x-axis direction 1/( tn+1-tn)(msec ⁻¹ ).
In the Beat Function, as shown in FIG. 1(c), the x-axis direction is constant at regular intervals, and the amplitude in the y-axis direction at constant n in the x-axis direction is the RR interval in the x-axis direction. Equal to (tn+1-tn) (msec).

In Non-Patent Document 1, an RRI time history is calculated from an electrocardiogram with the RR interval on the vertical axis and the time on the horizontal axis. It is described that the

Non-Patent Document 2 describes that between adjacent R-wave occurrence times, the reciprocal of the RR interval is taken as the instantaneous heart rate, and spectrum analysis is performed to capture heart rate fluctuations in real time. .

Non-Patent Document 3 describes spectral analysis of heart rate variability by regarding the RR interval as a function of the number of beats.

Transactions of the Japan Society of Mechanical Engineers Vol. 81, No. 832, 2015 p. 1-15 “Proposal of ride comfort control system by estimating occupant’s psychological state” BME Vol. 8, No. 10, 1994 p. 13-16 "1/f Spectrum Estimation of Heart Rate Fluctuation" Biophysics Vol. 28, No. 4, 1988 p. 32-36 "Heart rate variability and autonomic nerve function"

In the conventional analysis method of the frequency component of the RR interval fluctuation, as described above, the time series in which the RR interval in the x-axis direction in the electrocardiographic data is replaced with the numerical value as it is or the reciprocal thereof is replaced with the amplitude in the y-axis direction. Data is generated, and the time-series data generated thereby is frequency-analyzed to calculate the power spectral density, and the working condition of the autonomic nerves is evaluated from the power spectral density thus calculated.

However, since the stroke volume of the human heart is generally constant in a steady state, it is required that the sum of the power spectral densities in each beat of the heart be constant. If the sum of the power spectral densities is not constant, the sum of the energies of the time function will differ for each sampling time, so it will not be possible to correctly evaluate changes over time in the working condition of the autonomic nerves. That is, the total number of power spectral densities at each beat in the time series data obtained by replacing the value of the RR interval as it is or its reciprocal with the amplitude changes from time to time. The sum of the power spectral densities in the interval also changes, and the power spectral densities obtained by frequency analysis include this change. However, there is a problem that it is not possible to correctly evaluate changes over time.

An object of the present invention is to solve the above problems and to provide a heart rate variability analysis apparatus, method, and program capable of analyzing heart rate variability from electrocardiogram data and correctly evaluating the working condition of the autonomic nerves. be.

In order to solve the above problems, the present invention provides a heart rate variability analysis apparatus for analyzing heart rate variability from electrocardiographic data, comprising RR interval measuring means for sequentially measuring RR intervals in electrocardiographic data; - time-series data generation means for generating time-series data whose amplitude in the RR interval period is the reciprocal of the square root of each RR interval sequentially measured by the R-interval measurement means or a constant multiple thereof; It is characterized by comprising frequency analysis means for frequency-analyzing the time series data generated by the series data generation means and obtaining the power spectral density thereof.

Here, the total sum LF of predetermined low-frequency components and the total sum HF of predetermined high-frequency components in the power spectral density may be calculated, or LF/HF may be calculated in addition to these.

The present invention can be realized not only as a heart rate variability analysis device, but also as a heart rate variability analysis method having steps of processing of each part of the heart rate variability analysis apparatus, and as a computer program that causes a computer to implement the function of the processing. can also

In the present invention, the time-domain representation of the time-series data is optimized so that the sum of the power spectral densities per beat is equal over the entire time-series data. The sum of the power spectral densities of the count intervals can be equalized, under which the respective power spectral densities can be determined. Therefore, simply by comparing their power spectral densities, it becomes possible to correctly evaluate changes over time in the working condition of the autonomic nerves.

1 is a block diagram showing an embodiment of a heart rate variability analysis device according to the present invention; FIG. 2 is a flow chart showing the operation of the heart rate variability analysis device of FIG. 1; FIG. 4 is an explanatory diagram showing the relationship between the RR interval (x-axis) and the amplitude (y-axis) of electrocardiogram data for one beat; FIG. 3 shows time series data for a single beat count interval whose time domain representation has been optimized according to the present invention; FIG. 4 is an explanatory diagram showing the sum of power spectral densities for each beat in the present invention and the conventional method; FIG. 10 is an explanatory diagram of a general method of analyzing frequency components of RR interval fluctuations when a method of replacing the value of the RR interval as it is with an amplitude is used; FIG. 10 is an explanatory diagram showing a comparison of conventional time-series data generation methods;

The present invention will now be described with reference to the drawings.
FIG. 1 is a block diagram showing one embodiment of a heart rate variability analysis device according to the present invention.

A heart rate variability analysis device 10 of this embodiment includes an input/output interface 11 , a memory 12 , and a CPU 13 .

Electrocardiographic data from an electrocardiographic sensor or the like is input to the heart rate variability analysis device 10 through the input/output interface 11, and the power spectrum density of the heart rate variability analysis result in the heart rate variability analysis device 10 is output through the input/output interface 11. .

The memory 12 preliminarily stores a program for the heart rate variability analysis apparatus 10 to execute processing, and also appropriately holds data obtained by processing electrocardiographic data.

The CPU 13 executes processing according to programs stored in the memory 12 . Here, first, the RR intervals in the electrocardiographic data stored in the memory 12 are sequentially measured, and then the reciprocals of the square roots of the sequentially measured RR intervals or constant multiples thereof are calculated as the RR intervals. Time-series data is generated in terms of amplitude over a period, and the time-series data is subjected to frequency analysis to obtain its power spectral density.

As described above, the CPU 13 provides the RR interval measuring means 14 for sequentially measuring the RR intervals in the electrocardiographic data, the reciprocals of the square roots of the RR intervals sequentially measured by the RR interval measuring means 14, or Time-series data generation means 15 for generating time-series data with amplitude in the RR interval period as a constant multiple thereof, and frequency analysis of the time-series data generated by the time-series data generation means 15 It functions as frequency analysis means 16 for obtaining the power spectral density.

FIG. 2 is a flow chart showing the operation of the heart rate variability analysis device 10 of FIG.

First, in S21, electrocardiogram data is input to the heart rate variability analyzer 10 through the input/output interface 11, and stored in the memory 12 in S22. From S23 onward, the electrocardiographic data accumulated in the memory 12 are processed.

In S23, the RR intervals are sequentially calculated from the electrocardiographic data stored in the memory 12. In the next step, S24, the reciprocal of the square root of each RR interval sequentially calculated in S23 or a constant multiple thereof is calculated as the R - Generate time-series data by replacing the amplitude in the R interval period. In S24, the reciprocal of the square root or a constant multiple thereof is calculated from each RR interval measured in S23, and that value is given as the amplitude that continues during each RR interval.

In S25, frequency analysis is performed on the time-series data generated in S24. In this case, frequency analysis may be performed for the interval of the predetermined number of beats in the time-series data, and the predetermined number of beats can be determined arbitrarily. In order to do so, it is preferable to set it to a short interval of, for example, about 25 beats. If the electrocardiographic data to be input is for a predetermined number of beats from time to time, the entire data may be processed.
Through the frequency analysis in S25, the power spectral density of the time-series data (predetermined number of beats section) to be processed is obtained. At S26, the spectral density obtained at S25 is output through the input/output interface 11. FIG.

When frequency analysis is performed on the time-series data of the interval of the predetermined number of beats from time to time, the processes of S23 to S26 are repeatedly executed for the time-series data of the interval of the predetermined number of beats, and the power spectral density of each You can ask for

As described above, replacing the reciprocal of the square root of each RR interval of the time series data or a constant multiple thereof with the amplitude in the RR interval period generates time series data of optimized time domain representation. It is for That is, the purpose is to equalize the sum of the power spectral densities per beat over the entire time-series data, thereby equalizing the sum of the power spectral densities of the predetermined number of beats intervals of the time-series data. This optimization ensures that the power spectral density is obtained under the condition that the sum of the power spectral densities of the predetermined beat count intervals of the time series data is equal.
The optimization of time-series data in the time domain representation will be described below.

FIG. 3 is an explanatory diagram showing the relationship between the RR interval (x-axis) and amplitude (y-axis) of electrocardiographic data for one beat.

Here, assuming that the RR interval in the x-axis direction is B2-B1 and the amplitude in the y-axis direction is A, the total energy of the signal f(t) in the time domain is expressed by Equation (1).

Further, when the frequency domain signal obtained by frequency analysis of the time domain signal f(t) is denoted by F(f), the sum of the energies thereof is represented by Equation (2).

According to Pazel's theorem, Equation (3) holds.

なお、式（２）、式（３）における上線（￣）付のＦ（ｆ）は、Ｆ（ｆ）の共役複素数を表す。

Note that the overscored F(f) in equations (2) and (3) represents the complex conjugate of F(f).

From equation (3), the sum of the power spectral densities in FIG. 3 can be expressed by equation (4). Equation (5) is derived from Equation (4), and if the amplitude A conforms to Equation (5), the sum of the power spectral densities per beat over the entire time-series data can be a constant value C . This ensures that the power spectral density is obtained under the condition that the sum of the power spectral densities of the predetermined number of beats intervals of the time-series data is equal. Note that C in Expression (5) is an arbitrary constant.

Equation (5) replaces the amplitude A between each RR of the time series data with the reciprocal of the square root of the RR interval or a constant multiple thereof, so that the sum of the power spectral densities per beat is constant. This indicates that the sum of the power spectral densities of the predetermined number of beats intervals of the time-series data can be equalized.

Therefore, in the present invention, as time series data of optimized time domain expression, the reciprocal of the square root of each RR interval sequentially measured or a constant multiple thereof is used as the amplitude in the RR interval period. data so that when frequency analysis is performed on occasional predetermined beat count intervals of the time series data, the power spectral density is equal to the sum of the power spectral densities of the predetermined beat count intervals. Ensure that what is required is guaranteed. By simply comparing the power spectral densities obtained under this guarantee, it is possible to correctly evaluate changes over time in the working conditions of the autonomic nerves.

FIG. 4 is a diagram showing time series data for a single beat number interval whose time domain representation has been optimized according to the present invention.

In FIG. 4, the reciprocal of the square root of each RR interval is replaced by the amplitude during the RR interval, but as is clear from equation (5), the constant of the reciprocal of the square root of each RR interval is The double may be replaced by the amplitude during that RR interval.

FIG. 5 is an explanatory diagram showing the sum of power spectral densities for each beat in the present invention and the conventional method. Here, the RR interval of each beat is the RR interval (sec) data sample (transition in the y direction and then the x direction) shown in FIG. The sum of the power spectral densities at . The result is shown in FIG. The amplitude in the y-axis direction is sec ⁻¹ in the present invention, sec ⁻² in the 1/f fluctuation, and sec ² in the Equal and Beat functions.

From FIG. 5, according to the present invention, the sum of the power spectral densities for each beat is constant (=1), which means that the sum of the power spectral densities of the time-series data for the predetermined number of beats is equal. It can be seen that it is guaranteed that the power spectral densities of the data in those sections can be obtained under the conditions.

On the other hand, in the conventional method (Equal, 1/f fluctuation, Beat Function), the sum of the power spectral densities for each beat is not constant, and the sum of the power spectral densities of the time-series data in a predetermined number of beats interval Since the condition of equality is not guaranteed, it is not possible to correctly evaluate changes over time in the working condition of the autonomic nerves simply by comparing the power spectral densities obtained at different times.

The above is a case where the present invention is implemented as a heart rate variability analysis device. A heart rate variability analysis method comprising: a first step in which RR interval measurement means sequentially measures RR intervals from electrocardiographic data; and a time-series data generation means in the first step A second step of generating time-series data in which the reciprocal of the square root of each RR interval sequentially measured or a constant multiple thereof is the amplitude in the RR interval period; It can be implemented as a heart rate variability analysis method having a third step of frequency-analyzing the time-series data generated by the step to obtain its power spectral density.

Further, the present invention provides a computer with a first function of sequentially measuring RR intervals from electrocardiographic data, and a reciprocal of the square root of each RR interval sequentially measured by the first function or a constant multiple thereof. A second function of generating time-series data with the amplitude in the RR interval period, and a third function of frequency-analyzing the time-series data generated by the second function to obtain its power spectral density. It can also be realized as a program for realizing the function.
Although the embodiments have been described above, the present invention is not limited to the above embodiments and can be modified in various ways. For example, the following changes can be made, or these changes can be combined as appropriate.
• The electrocardiographic data to be input is stored in advance in an external memory.
- ECG data obtained at different times are subject to comparison processing.
・Processes such as RR interval measurement, time-series data generation, and frequency analysis are executed by separate CPUs or hardware.
・From the power spectral density, calculate the sum LF of predetermined low frequency components, such as 0.04 to 0.15 Hz components, and the sum HF of predetermined high frequency components, such as 0.15 to 0.40 Hz components, and autonomously calculate LF/HF and HF Output as a neural index.
A means for comparing power spectrum densities and autonomic nerve indexes respectively obtained for a plurality of predetermined beat number intervals.
・Temporarily store time-series data and other data using an external memory as appropriate.
- A display means for displaying analysis results of heart rate variability is provided.

10 Heart rate variability analysis device 11 Input/output interface 12 Memory 13 CPU 14 RR interval measuring means 15 Time-series data generating means 16 ... frequency analysis means

Claims (5)

A heart rate variability analysis device for analyzing heart rate variability from electrocardiographic data,
RR interval measuring means for sequentially measuring RR intervals in electrocardiographic data;
a time-series data generating means for generating time-series data in which the reciprocal of the square root of each RR interval sequentially measured by the RR interval measuring means or a constant multiple thereof is used as the amplitude in the RR interval period;
A heart rate variability analysis apparatus comprising frequency analysis means for frequency-analyzing the time-series data generated by the time-series data generation means to obtain a power spectral density thereof. 2. The frequency analysis means according to claim 1, wherein said frequency analysis means calculates a total sum LF of predetermined low frequency components and a total sum HF of predetermined high frequency components in the power spectral density, or additionally calculates LF/HF. Heart rate variability analyzer. 3. The heart rate variability analysis apparatus according to claim 1, wherein said frequency analysis means frequency-analyzes data in a predetermined number of beats section of the time-series data. A heart rate variability analysis method in which a computer functions as RR interval measurement means, time-series data generation means, and frequency analysis means to analyze heart rate variability from electrocardiographic data,
a first step in which the RR interval measuring means sequentially measures RR intervals in electrocardiographic data;
The time-series data generating means generates time-series data in which the reciprocal of the square root of each RR interval sequentially measured in the first step or a constant multiple thereof is the amplitude in the RR interval period. 2 steps;
A heart rate variability analysis method comprising a third step, wherein the frequency analysis means frequency-analyzes the time-series data generated in the second step to determine its power spectral density. to the computer,
a first function of sequentially measuring RR intervals in electrocardiographic data;
a second function of generating time-series data in which the reciprocal of the square root of each RR interval sequentially measured by the first function or a constant multiple thereof is the amplitude in the RR interval period;
A program for realizing a third function of frequency-analyzing the time-series data generated by the second function and determining its power spectral density.

Images