JP2010051387A - Device for detecting cvhr accompanying apnea attack or infrequent respiration attack of sleep-disordered respiration - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide technologies for accurately detecting CVHR (cyclic variation of heart rate) accompanying the sleep-disordered respiration. <P>SOLUTION: The CVHR detection device 2 computes the amplitude of a high-frequency component of the heart rate variation based on R-R interval time-serial data. The CVHR detection device 2 determines a threshold for the depth of dip as a value intrinsic to each data from the amplitude value of the high-frequency component, and specifies a dip group having the depth larger than the threshold. The CVHR detection device 2 further specifies a dip with similarity in the shape, and further specifies a dip group continuously appearing with periodicity. The CVHR detection device 2 can detect a finally specified dip group as CVHR. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a technique for detecting periodic fluctuations in heart rate associated with an apnea attack or hypopnea attack of sleep disordered breathing. In the present specification, the cyclic variation of the heart rate associated with the apnea attack or hypopnea attack of sleep disordered breathing is simply expressed as CVHR (cyclic variation of heart rate).

  Non-Patent Document 1 discloses a method for detecting sleep apnea by heart rate variability analysis. In this detection method, a power spectrum representing the intensity of the frequency component of heartbeat variability is created from the heartbeat interval time-series data. When the power in the frequency band corresponding to the ultra-low frequency region (0.003 to 0.04 Hz) is increased in this power spectrum, it is determined that sleep apnea exists.

Roche et al., “Cardiac Interbeat Interval Increment for the Identification of Obstructive Sleep Apnea”, Journal of Pacing and Clinical Electrophysiology, (USA), August 2002, Vol. 25, No. 8, pp.1192-1199.

CVHR appearing accompanying a sleep apnea attack has a frequency of 25 to 120 seconds, that is, a frequency of 0.008 to 0.04 Hz, and in this frequency band, an extremely low frequency component that is a physiological heart rate fluctuation component (0.003 to 0.04 Hz) also exist at the same time, and the two cannot be distinguished only by frequency analysis such as spectrum analysis. Further, in the spectrum analysis, even if the existence of CVHR can be estimated to some extent, the frequency of CVHR and the temporal position of each CVHR cannot be specified. Furthermore, CVHR associated with apnea and hypopnea is mediated by autonomic nerves, and the intensity of CVHR (the peak height seen in heart rate changes) is related to the level of parasympathetic activity. For this reason, even if CVHR exists, the intensity | strength of CVHR changes largely by the change of the autonomic nervous system activity by the individual difference of an autonomic-nerve function, or an individual's health state.
Therefore, in the method of Non-Patent Document 1, detection of CVHR becomes inaccurate due to individual differences in physiological ultra-low frequency components and increase / decrease due to individual state, and the specificity and sensitivity of sleep apnea detection are reduced. . Further, since the method of Non-Patent Document 1 does not directly detect CVHR, the frequency of CVHR and the temporal position of each CVHR cannot be known, and the frequency and temporal distribution of apnea and hypopnea attacks are accurately determined. I can't figure it out. Furthermore, since the method of Non-Patent Document 1 does not consider changes in the intensity of CVHR due to individual differences or individual states, CVHR can be appropriately detected when there is an autonomic nervous function disorder or decreased activity. Can not. Sleep apnea is likely to be associated with lifestyle-related diseases such as cardiovascular diseases and obesity, but these diseases generally tend to be accompanied by autonomic dysfunction, particularly parasympathetic dysfunction.

  The present specification provides a technique that can accurately detect CVHR regardless of the intensity of physiological ultra-low frequency components and the level of autonomic nerve activity, and specify the frequency and temporal position thereof.

  The technique disclosed by this specification is a detection apparatus for CVHR associated with an apnea attack or a hypopnea attack of sleep disordered breathing. The CVHR detection device includes time series data input means, peak detection means, peak height calculation means, heart rate variability index calculation means, peak height threshold determination means, peak group identification means, and analysis result output means.

  The time series data input means inputs time series data of heartbeat or pulse interval or time series data of heartbeat or pulse frequency. As time-series data of beat intervals, time-series data of beat intervals for each beat (electrocardiogram RR interval, etc.), time-series data of instantaneous beat intervals (beat period function, etc.), etc. may be adopted. . As time-series data of the beat frequency, time-series data of heart rate and pulse rate every predetermined time, time-series data of instantaneous heart rate and instantaneous pulse rate, and the like may be adopted. The time series data preferably has a resolution of about 10 ms in terms of beat interval.

  It is preferable to pre-process the time-series data input by the time-series data input unit as necessary. It is preferable to remove data fluctuations caused by arrhythmia such as extrasystole and cardiac block by preprocessing. It is also preferable to remove artifacts due to measurement and recording.

  Subsequently, it is preferable to interpolate the time series data subjected to these arithmetic processes and perform resampling. Resampling is preferably performed at a frequency of about 2 Hz. The interpolation method is not particularly limited, but step interpolation, linear interpolation, spline interpolation, or the like may be used.

  The peak detection means detects a plurality of local peaks from the time series data. The term “peak” is a concept that includes both decreasing and increasing peaks. A decrease peak (ie, dip) is detected from the time-series data of beat intervals. An increase peak is detected from the time-series data of the beat frequency. The peak detection method is not particularly limited. For example, the peak may be calculated based on the shape of the data variation of the time series data. For example, a data variation having a shape similar to a preset peak shape may be detected as a peak. Or you may detect the data fluctuation | variation which has a shape similar to the preset peak shape as a peak. In addition, for example, when detecting an increase peak, if the peak of a predetermined data fluctuation indicates a value larger than the apex of the adjacent data fluctuation, the data fluctuation may be detected as a peak. Conversely, when detecting a decrease peak, if the peak of a predetermined data fluctuation shows a smaller value than the apex of the adjacent data fluctuation, the data fluctuation may be detected as a peak.

  The peak height calculation means calculates the height of each of the plurality of local peaks detected by the peak detection means. In addition, the calculation method of peak height is not specifically limited. For example, the distance between the upper end and the lower end of the peak may be calculated as the peak height. Further, for example, a reference baseline may be determined for each peak, and the distance between the baseline and the peak apex may be calculated as the peak height. There are various ways to set the baseline and peak apexes. For example, the baseline may be a moving average value or a moving median value calculated from a distribution of data fluctuations within a predetermined time range before and after the peak. The peak apex may be a value at the center time of the peak width, for example.

  The intensity of CVHR that accompanies apnea and hypopnea varies depending on individual differences such as the level of autonomic nerve activity and the health status of individuals. For example, if there is a suppression of parasympathetic nerve activity or a functional disorder due to coronary artery disease, heart failure, diabetes or the like, the intensity of CVHR, that is, the height of the peak decreases. For this reason, if it is going to identify the fluctuation | variation used as the candidate of CVHR from peak height using the same reference value about a some test subject, specificity and a sensitivity will fall. For this purpose, in the present invention, a heart rate variability index that reflects autonomic nervous activity is calculated for each time series data, and a threshold value for peak height is determined as a value unique to each time series data from the index, A technique for detecting CVHR by specifying a peak having a height larger than the peak height threshold is adopted.

  The heart rate variability index calculating means calculates a heart rate variability index that reflects the autonomic nerve activity. The heart rate variability index includes the power or amplitude of a high frequency component (e.g., 0.15 to 0.45 Hz) of the fluctuation of the beat cycle or beat frequency, and the root mean square of the difference values of successive beat cycles or beat frequencies. It is preferable to use an index that mainly reflects the parasympathetic nerve function such as (root mean square of successive difference), but the heart rate variability index is not limited thereto. For example, a low frequency component (low frequency component, 0.04 to 0.15 Hz) of fluctuation in beat cycle or beat frequency may be used as an index because it reflects the parasympathetic nerve function particularly at night. Further, the frequency band of the high frequency component of the heartbeat fluctuation is not limited to 0.15 to 0.45 Hz. A part of the frequency band or a frequency band expanded to a higher frequency band may be used as a frequency band for calculating a high frequency component. Further, the method for calculating the power or amplitude of the high frequency component and the low frequency component is not particularly limited. For example, the average value of power and amplitude in the analysis interval may be obtained by spectrum analysis in the frequency domain, etc., and the power and amplitude may be calculated in time by wavelet analysis in the time / frequency domain and complex demodulation analysis in the time domain. It may be calculated as a function of The time length of data used for calculating the heart rate variability index is not particularly limited. For example, the heart rate variability index may be calculated from 24-hour time series data, or the heart rate variability index may be calculated from time series data at night or other time zones.

  The peak height threshold value determining means uses a heart rate variability index to determine a threshold value related to peak height for identifying a peak that is a candidate for CVHR from the peaks detected by the peak detecting means. It is determined as a unique value. As described above, the intensity of CVHR associated with apnea and hypopnea changes depending on individual differences in the level of autonomic nervous activity and the health condition of individuals, so even if CVHR exists, the expected peak height is It differs for each time series data. Therefore, the peak height threshold determining means determines the peak height threshold for each time series data from the heart rate variability index obtained from each time series data. Here, the peak height threshold is calculated as a function of the value of the heart rate variability index. This function depends on the type of the heart rate variability index, but when the value of the heart rate variability index is converted into the amplitude of the high frequency component (0.15 to 0.45 Hz) of the beat cycle, it is about 2 to 4 times the value. Is a peak height threshold candidate. However, the value of the heart rate variability index varies depending on the resolution of the time series data, the preprocessing method and accuracy regarding the removal of arrhythmia and artifacts, the method of calculating the heart rate variability index, etc., even if they are the same index. Therefore, the magnification of the peak height threshold with respect to the amplitude of the high frequency component of the beat period is not limited to the above value.

  The peak group specifying means specifies a peak group that has a peak height larger than the peak height threshold and is continuous with periodicity from among the peaks detected by the peak detecting means. . Apnea attacks and hypopnea attacks often occur periodically, and the heartbeat or pulse is periodically decreased and increased accordingly to form CVHR. As described above, in the frequency band common to or close to CVHR, there is a very low frequency component that is a physiological heartbeat fluctuation, but unlike the CVHR, the ultra low frequency component is a non-periodic fluctuation having no periodicity. It is known. Therefore, in the fluctuation due to the ultra-low frequency component, the peak directionality (increase peak or decrease peak) as in CVHR is not constant, and the appearance of the peak does not have periodicity. The peak group specifying means uses this property, and has a peak group that has a height larger than the above-mentioned peak height threshold among a plurality of local peaks and is continuous with periodicity. Is identified. As a result, the peak group reflecting CVHR is detected by being distinguished from the peak due to the ultra-low frequency component. The peak group specified by the peak group specifying means is finally determined as CVHR. The peak group specifying means is limited to the order of specifying these peak groups among the peak group having a peak height larger than the peak height threshold and the peak group continuous with periodicity. is not.

  The analysis result output means outputs an analysis result including the frequency and temporal distribution of the peak group specified by the peak group specifying means. There are various methods of outputting the peak group frequency and temporal distribution in the analysis result output means. For example, for the peak group frequency, the total number of overnight peaks or overnight CVHR, the frequency per hour, and the frequency of each time zone may be output. Regarding the temporal distribution of the peak group, the appearance time of each peak may be output, or the temporal position of the peak may be output alone or as a graph with respect to the time axis together with the electrocardiogram, beat interval, and beat frequency. Furthermore, peak height and peak width data may be output, or a time difference from peaks appearing before and after may be output. There are various types of data output by the analysis result output means. The analysis result output means may display data relating to the peak group on the screen or may output it to a printer. Alternatively, the analysis result output unit may store the data in a storage medium such as a CD-R, or may transmit the data to an external computer.

  In the CVHR detection device of the present invention, a peak group having a peak height larger than the peak height threshold is determined using a peak height threshold determined based on the heart rate variability index calculated for each time series data. Can be identified. This makes it possible to accurately detect CVHR regardless of individual differences in CVHR intensity associated with apnea or hypopnea and changes due to the individual's health condition. Also, by specifying a group of peaks that are continuous with periodicity, peaks due to fluctuations that do not have periodicity such as ultra-low frequency components can be excluded, and CVHR can be accurately detected.

  The heart rate variability index calculating means may calculate the power or amplitude of the variation (that is, high frequency component) included in at least the frequency band of 0.15 to 0.45 Hz of the heart rate variability from the time series data. The power or amplitude of the high frequency component of the heart rate variability is an index reflecting the parasympathetic nerve function.

  The heart rate variability index calculating means may calculate a root mean square of successive difference from at least successive beat periods or beat frequency difference values from time series data. The mean square of the difference values between successive beat periods or beat frequencies is an index reflecting the parasympathetic nerve function.

  Time series based on the index (the power or amplitude of the high frequency component of heart rate variability, or the root mean square of the difference value of the continuous beat period or beat frequency), the threshold unique to the time series data related to peak height. By determining for each data, CVHR can be accurately detected regardless of changes in the intensity of CVHR due to individual differences in parasympathetic nerve activity levels or changes due to individual health conditions. For example, when apnea and hypopnea occur in a person with a high autonomic nervous function and a high level of parasympathetic nerve activity at night, CVHR with a large peak height is generated along with it. Physiological heart rate variability including such is also strong. In this case, by setting the peak height threshold value to a large value based on the heartbeat variability index, the rate of erroneously detecting peaks due to fluctuations other than CVHR as CVHR can be lowered, and the specificity of CVHR detection is increased. Conversely, even if apnea or hypopnea occurs in a person with autonomic nerve dysfunction or a person whose nighttime parasympathetic nerve activity level is suppressed, the CVHR associated therewith only has a small peak height. Such a person also has a small physiological heart rate fluctuation including an ultra-low frequency component. In this case, by setting the peak height threshold to a small value based on the heart rate variability index, it is possible to reduce the detection defect rate that overlooks CVHR in which only slight changes occur in the beat interval and beat frequency, and the sensitivity of CVHR detection is improved. Get higher.

  The peak group specifying means has a peak height larger than the peak height threshold among a plurality of local peaks using a peak height threshold determined based on the heart rate variability index calculated for each time series data. It is preferable to identify a peak group that is present and to identify a peak group that is continuous with periodicity from the peak group. As a result, it is possible to remove peaks due to fluctuations that do not have periodicity, such as ultra-low frequency components, and to accurately detect CVHR.

  The peak group identification means identifies three or more consecutive peaks in which the time difference between two consecutive peaks is within a predetermined range, thereby allowing continuous peaks with periodicity. A group may be identified. For example, the predetermined range may be 25 seconds to 120 seconds. Moreover, you may specify the peak group in which the dispersion | variation in several continuous time differences is contained in a predetermined range. Further, four or more continuous peaks may be specified.

  In CVHR associated with apnea or hypopnea attacks, peaks that are similar in shape often appear continuously. The peak group specifying means has a peak that has a height larger than the peak height threshold among a plurality of local peaks, has a similar shape, and is continuous with periodicity. A group may be identified. The peak group finally specified by the peak group specifying means is determined as CVHR. For example, the peak group specifying means may determine whether or not the shapes of two adjacent peaks are similar. The similarity of peaks may be determined based on, for example, peak width and / or peak height. For example, the peak group whose peak width is included in the first predetermined range may be determined to have similarity. For example, the peak group whose peak height is included in the second predetermined range may be determined to have similarity. CVHR can be accurately detected by specifying a peak having similarity in addition to the peak height and periodicity larger than the peak height threshold.

  The peak group specifying means specifies a peak group having a peak height larger than the peak height threshold from a plurality of local peaks, and has a predetermined similar shape from the peak group. It is preferable to identify a peak group and identify a peak group that is continuous from the peak group with a predetermined periodicity. By specifying a peak group that is continuous with periodicity at the end, it is possible to obtain an accurate specification result rather than specifying in another order.

The present specification also provides a computer program for realizing the CVHR detection device. This computer program causes a computer to execute the following processes.
(1) Time-series data input processing for inputting time-series data of heartbeat or pulse interval or time-frequency data of heartbeat or pulse frequency.
(2) Peak detection processing for detecting a plurality of local peaks from time series data.
(3) Peak height calculation processing for calculating the height of each of a plurality of local peaks.
(4) Heart rate variability index calculation processing for calculating a heart rate variability index that reflects autonomic nerve activity from time series data.
(5) Peak height threshold value determination processing for determining a threshold value related to peak height as a value unique to each data from the heartbeat variability index.
(6) Peak group specifying processing for specifying a peak group having a height larger than the peak height threshold and continuous with periodicity from among a plurality of local peaks.
(7) An analysis result output process for outputting an analysis result including the peak group frequency and temporal distribution.
By using this computer program, it is possible to realize a CVHR detection device that can identify the frequency and temporal position of CVHR regardless of individual differences in autonomic nerve activity or changes due to individual health conditions.

The main features of the embodiment described below are organized. In this embodiment, time series data of the heartbeat interval is used, and a case where CVHR forms a peak (dip) in a decreasing direction will be described. Accordingly, the peak height here means the depth of the dip.
(Feature 1) peak detecting means, when a predetermined shape having an apex downwards when painted series data as a graph against time axis parabola (e.g. H = T ^2/49, where T is from the center of the parabola Time [seconds], H is a peak candidate where the height from the apex of the parabola [milliseconds] can be inscribed, and the peak candidate is present in a predetermined time before and after (for example, 10 seconds before and after) When the value of the apex of the parabola inscribed in the peak candidate is smaller, the peak candidate is specified as a peak.
(Characteristic 2) The peak height calculation means calculates a moving average of time series data for a predetermined time before and after each central time (for example, 25 seconds before and after) for each peak, and calculates the maximum value of each moving average. . The peak height calculation means calculates the height of each peak as the difference between the value at the peak center time and the average value of the maximum values of the moving averages before and after. The moving average window width (filter length) is set to a predetermined value (for example, 5 seconds).
(Characteristic 4) The peak height threshold value determining means uses a value obtained by multiplying the amplitude of the high frequency component (0.15 to 0.45 Hz) of the heart rate variability obtained from the beat interval time series data by 2.5 as the time series data. Used as a unique peak height threshold.
(Feature 5) The peak group specifying means has a peak width within a predetermined range with respect to adjacent peaks, has a peak height within a predetermined range with respect to adjacent peaks, and is adjacent A peak having a ratio between a peak width and a peak height within a predetermined range with respect to the peak to be identified is specified as a peak having similarity.
(Characteristic 6) The peak group specifying means is such that the time difference between two adjacent peaks is within a predetermined time (for example, 25 seconds to 120 seconds) for all two adjacent peaks among four consecutive peaks. When there is a variation in the magnitude of the time difference within a predetermined range, the four peaks are identified as peaks having periodicity. For example, four consecutive peaks are P1, P2, P3, P4, the time difference between P1 and P2 is T12, the time difference between P2 and P3 is T23, and the time difference between P3 and P4 is T34. In this case, when all of T12, T23, and T34 are within the predetermined time range and the variation of T12, T23, and T34 is within the predetermined range, P1, P2, P3, and P4 have periodicity. Identify as a peak.

  Embodiments will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the CVHR detection device 2 of the present embodiment. The CVHR detection apparatus 2 includes an RR interval time series data input unit 16, a dip detection unit 18, a dip depth calculation unit 20, a heart rate variability index calculation unit 22, an individual threshold value determination processing unit 24, a dip width calculation unit 26, and a dip. An interval calculation unit 28, a dip group identification unit 30, another calculation unit 32, a storage unit 34, an operation unit 36 and a display unit 38 are provided. In addition, said each part 16-32 grade | etc., Is implement | achieved when the computer mounted in the CVHR detection apparatus 2 performs a process according to a program.

  The RR interval time series data input unit 16 is connected to the communication line 14. The communication line 14 is connected to an RR interval measurement device (for example, a Holter electrocardiogram analyzer or a Holter electrocardiograph having a Holter electrocardiogram analysis function). The RR interval time series data input unit 16 inputs the RR interval measurement data measured and output by the RR interval measuring device. The chart 70 in FIGS. 7 and 8 shows an example of RR interval measurement data. In the chart 70, the RR interval repeats increasing and decreasing periodically from the middle. The dip detector 18 detects a plurality of local dip from the RR interval time series data. The dip detector 18 detects a dip group satisfying a predetermined dip shape from data such as a dip width and a dip depth. The dip detection method will be described in detail later. The dip depth calculation unit 20 calculates the depth of each dip group detected by the dip detection unit 18. A method for calculating the dip depth will be described in detail later.

  The heart rate variability index calculator 22 calculates the amplitude of the high frequency component (0.15 Hz to 0.45 Hz) from the RR interval time series data. The heart rate variability index calculation unit 22 can extract a frequency component by any of the following calculation methods. For example, the heart rate variability index calculation unit 22 may calculate the amplitude of the high frequency component by complex demodulation analysis. The heart rate variability index calculation unit 22 may calculate the amplitude of the high frequency component by fast Fourier transform or autoregressive analysis. The heart rate variability index calculation unit 22 may calculate the amplitude of the frequency component by wave bread transform or short-time Fourier transform. The heart rate variability index calculating unit 22 may calculate a root mean square of successive difference as an estimated value of the amplitude of the high frequency component.

  The individual threshold value determination processing unit 24 determines a data-specific threshold value related to the depth of a dip that is a candidate for CVHR as a data-specific threshold value from the amplitude of the high-frequency component extracted by the heartbeat variability index calculating unit 22. In this embodiment, a value 2.5 times the amplitude of the high frequency component is adopted as the data specific threshold. The dip width calculation unit 26 calculates the width of each of the plurality of local dips (that is, the length of time in which each dip appears). The dip interval calculation unit 28 calculates the interval between each two consecutive dip. The dip interval is the time from the center value of the dip width to the center value of the dip width of the adjacent dip.

The dip group identification unit 30 executes the following processes.
(1) A dip group having a dip depth larger than the data specific threshold is specified as a significant dip group from a plurality of local dip.
(2) A dip group having a predetermined similar shape is specified as a similar dip group from the significant dip groups specified in (1) above.
(3) From among the similar dip groups identified in (2) above, a dip group that is continuous with a predetermined periodicity is identified as a periodic dip group.
The periodic dip group specified in (3) is CVHR.

  In the above (1), since the data specific threshold calculated for each data is used as a criterion for determining the significance of the depth of the dip, the dip group specified in the above (1) is called a significant dip group. The dip group specified in (2) above is called a similar dip group. The dip group specified in (3) above is called a periodic dip group. The other calculation part 32 performs various calculation processes other than the above. The arithmetic processing performed by the arithmetic unit 32 will be described in detail later.

  The storage unit 34 is configured by a ROM, an EEPROM, a RAM, and the like. The storage unit 34 can store various information. For example, the storage unit 34 can store various information regarding the dip group specified by the dip group specifying unit 30. For example, the storage unit 34 may store the RR interval time series data input to the RR interval time series data input unit 16. Or the memory | storage part 34 may memorize | store the appearance time of each dip, and may memorize the appearance time zone of a dip group. Alternatively, the storage unit 34 may store data such as the width and depth of each dip. The operation unit 36 has a plurality of keys. The user can input various information to each part of the CVHR detection device 2 by operating the operation unit 36. The display unit 38 can display various information on the screen.

  The contents of the CVHR detection process executed by the computer program installed in the CVHR detection device 2 will be described. 2 to 4 show flowcharts of the CVHR detection process. The RR interval time series data input unit 16 inputs RR interval time series data via the communication line 14 (S10). The RR interval time-series data preferably has a time resolution of about 10 ms (that is, a resolution capable of identifying a difference in RR interval of 10 ms or more). Reference numeral 70 in FIGS. 7 and 8 indicates an example of RR interval time-series data.

  The RR interval time-series data input in S10 includes non-physiologic arrhythmias such as extrasystole and cardiac block, and data fluctuations due to artifacts. Therefore, the calculation unit 32 performs calculation processing to remove data fluctuations caused by non-physiologic arrhythmia and artifacts (S12). As a result, fluctuations in data caused by causes other than physiological heartbeat fluctuations and heartbeat fluctuations due to apnea and hypopnea can be eliminated.

In S14, the calculation unit 32 performs interpolation of RR interval time series data. For example, when step interpolation is performed, an interpolation function is used between each RR interval such that the function value takes a constant value equal to the value of the RR interval. Subsequently, the calculation unit 32 resamples the value of the interpolation function at a frequency of 2 Hz. Thus, RR interval time series data X (t) sampled at equal intervals is created. Subsequently, the dip detection unit 18 sets a time t at which the following (Expression 1) is satisfied with respect to all T in the range of −5 to 5 seconds on the time-series data X (t). (S16).
(Equation 1) {X (t) + T 2/49 ≧ X (t + T), T = -5,5}
(Equation 1) is the time series parabola having a vertex downwardly when data X (t) is drawn as a graph against time ^{t (H = T 2/49} , where T is from the center of the parabola time [s] , H detects a fluctuation portion in which the height [millisecond] from the apex of the parabola can be inscribed as the dip candidate existence time.

  The dip detection unit 18 specifies the dip candidate as a dip when the vertex of the parabola inscribed in the dip candidate is smaller than the value of the parabola vertex inscribed in any of the dip candidates existing for 10 seconds before and after. (S18).

Dip depth calculating unit 20, a plurality of local dip detected in S18, the calculated respective dip depth D _i. i is the ordinal number of the detected dip. Figure 4 shows a flowchart of a calculation process of dipping depth D _i. The dip depth calculation unit 20 executes the process (S50 to S56) of FIG. 4 for each dip detected in S18.

The dip depth calculation unit 20 calculates a moving average with a window width of 5 seconds for time-series data for 25 seconds before and after the center time of the dip. A time series obtained by correcting the phase shift of the obtained moving average value is defined as X _MV5 (t) (S50). X (d _i ) is calculated at the center point (center time) d _i in the time axis direction of the dip (S54). X (d _i ) is a value near the vertex of the dip. Dip depth calculating unit 20 calculates the dip depth _{D i} by the following Equation (2) (S56).
(Formula 2) D _i = {max [X _MV5 (t), t = d _i−25 , d _i ] + max [X _MV5 (t), t = d _i , d _{i + 25} ]} / 2−X (d _i )
That is, the dip depth calculating unit 20, the maximum value of the moving average X _{MV 5} (t) at 25 seconds of the previous center point _{d i} dip, the moving average of 25 seconds after the center point _{d i} X _{MV 5} (t ), And the average value of both is calculated as the baseline value. Dip depth calculating unit 20, by calculating the difference between the value near baseline and vertices to calculate the dip depth D _i.

In S22 of FIG. 2, the heart rate variability index calculation unit 22 calculates the amplitude HF _AMP of the high frequency component (0.15 to 0.45 Hz) from the RR interval time series data by fast Fourier transform. The heart rate variability index calculation unit 22 sets the threshold value DD _TH related to the dip depth unique to the data to a value 2.5 times the HF _AMP (S24). The amplitude HF _AMP of the high frequency component is calculated for each data. Therefore, DD _TH is a data-specific threshold value adapted to the data.

i-th dip dip _i, describe the depth and D _i. The dip group specifying unit 30 determines whether the dip _i is a significant dip based on whether the dip depth D _i is greater than the data specific threshold DD _TH (S25). In the case of YES here, the dip group specifying unit 30 leaves the dip _i as a significant dip (S26). The dip group left in S26 is a significant dip group. Subsequently, the dip group specifying unit 30 determines whether or not the dip _i is the last dip of the RR interval time series data (S28). In the case of YES here, the process proceeds to S30 in FIG. On the other hand, if NO in S28, the dip group specifying unit 30 specifies the next dip (S29) and returns to S25. As a result, for the next dip, the dip depth D _i is compared with the data specific threshold DD _TH .

On the other hand, if NO in S25, the dip group identification unit 30 removes dip _i (S27). Subsequently, the dip group specifying unit 30 determines whether or not the dip _i is the last dip of the RR interval time series data (S28). In the case of YES here, the process proceeds to S30 in FIG. On the other hand, if NO in S28, the dip group specifying unit 30 specifies the next dip (S29) and returns to S25.

In S30 in FIG. 3, the arithmetic unit 32 calculates the dip width _{W i} at 2/3 of the height of the _{D i} from the lowest value of the dip. Subsequently, the dip group specifying unit 30 determines whether or not all of the following (Equation 3), (Equation 4), and (Equation 5) are satisfied for each dip (S31).
(Formula 3) abs (log (D _i / D _{i + 1} ) <log (2.5)
(Formula 4) abs (log (W _i / W _{i + 1} ) <log (2.5)
(Formula 5) abs (log (W _i · D _{i + 1} / W _{i + 1} · D _i ) <log (2.5)
Here, the dip group specifying unit 30 determines from the width and depth of the dip whether or not the shapes of the continuous dip _i and dip _{i + 1} are similar. If YES in S31, the dip group identification unit 30 leaves dip _i and dip _{i + 1} (S32). The dip group remaining in S32 is a similar dip group. Subsequently, the dip group specifying unit 30 determines whether or not the dip _i is the last dip of the RR interval time series data (S34). If YES here, the process proceeds to S36. On the other hand, in the case of NO in S34, the dip group specifying unit 30 specifies the next dip (S35), and returns to S31. In S31, the dip group specifying unit 30 determines whether there is similarity for the next dip.

On the other hand, if NO in S31, the dip group identification unit 30 removes dip _i (S33). Subsequently, the dip group specifying unit 30 determines whether or not the dip _i is the last dip of the RR interval time series data (S34). If YES here, the process proceeds to S36. On the other hand, in the case of NO in S34, the dip group specifying unit 30 specifies the next dip (S35), and returns to S31.

FIG. 5 is a schematic diagram of RR interval time-series data. With reference to FIG. 5, a detailed description will be given of a determination method for determining which dip is left after the dip group specifying unit 30 finishes the process of S31. Dips _i to _{i + 3} appear continuously in time series. _Wi is the dip width of dip _i . _Di is the dip depth of dip _i . First, the dip group specifying unit 30 determines the similarity of the combination A of dip _i and dip _{i + 1} . Subsequently, the similarity of the combination B of dip _{i + 1} and dip _{i + 2} is determined. Subsequently, the similarity of the combination C of dip _{i + 2} and dip _{i + 3} is determined.

When the combination A satisfies the similarity, the dip group specifying unit 30 leaves both dip _i and dip _{i + 1} . Subsequently, when the combination B also satisfies the similarity, the dip group specifying unit 30 leaves the dip _{i + 1} and the dip _{i + 2} . At this time, the dip _{i + 1} is left in both the processes of the combinations A and B. On the other hand, when the combination B does not satisfy the similarity, only the dip _{i + 2} is removed. The dip _{i + 1} that remains once in the combination A is not removed regardless of the result of the combination B.

When the combination B does not satisfy the similarity and the combination C satisfies the similarity, the dip group specifying unit 30 leaves the dip _{i + 2} and the dip _{i + 3} . That is, dip _{i + 2} is removed in combination B but can remain in combination C.

In S36, the calculation unit 32 calculates time differences I _i , I _{i + 1} , I _{i + 2} between two adjacent dips in the four consecutive dips in the dip group remaining in S34. Time difference _{I i} is the time difference between the central time _{d i + 1} of the dip _{i + 1} which is continuous with central time _{d i} dip _i. The dip group specifying unit 30 leaves four consecutive dip groups that satisfy all of the following (Expression 6), (Expression 7), and (Expression 8) (S38).
(Expression 6) 25 <I _i , I _{i + 1} , I _{i + 2} <120
(Formula 7) (3-2I _i / S) (3-2I _{i + 1} / S) (3-2I _{i + 2} /S)>0.6
(Equation 8) S = (I _i + I _{i + 1} + I _{i + 2} ) / 3
Here, the dip group specifying unit 30 determines whether the four dip groups forming the time differences I _i , I _{i + 1} , and I _{i + 2} have periodicity based on the magnitude of the time difference and the variation in the magnitude of successive time differences. Judge whether or not. The dip group remaining in S38 is a periodic dip group. In the CVHR apparatus 2, the periodic dip group remaining in S38 is detected as CVHR.

FIG. 6 is a schematic diagram of RR interval time series data. A determination method for determining which dip group to leave in the processing of S38 by the dip group specifying unit 30 will be described in detail with reference to FIG. Dips _i to dip _{i + 7} appear continuously in time series. First, the dip group specifying unit 30 determines the periodicity of the combination A composed of I _{i to} I _{i + 2} . Subsequently, the dip group specifying unit 30 determines the periodicity of the combination B composed of I _{i + 1 to} I _{i + 3} . Similarly, the dip group specifying unit 30 determines by shifting the dip one by one in time series order, and determines the periodicity of the combination E formed of I _{i + 4 to} I _{i + 7} .

When the combination A satisfies the periodicity, the dip group specifying unit 30 leaves dip _i to dip _{i + 3} . Subsequently, when the combination B satisfies the periodicity, the dip group specifying unit 30 leaves dip _{i + 1} to dip _{i + 4} . At this time, dip _{i + 1} to dip _{i + 3} are left in both processes of the combinations A and B. On the other hand, when the combination B does not satisfy the periodicity, only the dip _{i + 4} is removed from the dip _{i + 1} to the dip _{i + 4} constituting the combination B. Dips _{i + 1} to _{i + 3} that remain once in combination A are not removed regardless of the result of combination B.

When the combination B does not satisfy the periodicity and the combination E satisfies the periodicity, the dip group specifying unit 30 leaves all the dip _{i + 4} to dip _{i + 7} . That is, dip _{i + 4} is removed in combination B but can remain in combination E. Although there is a combination (not shown) between the combination B and the combination E, dip _{i + 4} to dip _{i + 7} are left when the combination E satisfies the periodicity regardless of the determination result.

In S40, the display unit 38 displays data on the dip group remaining in S38 on the screen. FIG. 7 shows an example of data displayed by the display unit 38. The chart 70 shows RR interval time series data. The bar graph 72 indicates the appearance time of the dip group specified by the CVHR device 2. Chart 74 shows a chart of percutaneous arterial oxygen saturation (SpO ₂ ). A chart 76 shows a chart of the ventilation amount due to respiration.

The display unit 38 may display the identified dip appearance time together with the RR interval time series data. The display unit 38 may display the identified dip appearance time together with other analysis results such as percutaneous arterial oxygen saturation (SpO ₂ ). In addition, the display unit 38 displays the CVHR appearance frequency per predetermined time, the short time zone in which the dip appearance frequency is maximum (for example, 30 minutes), the dip appearance frequency therebetween, and the average value of the dip depth and dip width. In addition, standard deviation and the like may be displayed.

Here, a method for diagnosing obstructive sleep apnea syndrome (OSAS) based on percutaneous arterial oxygen saturation (SpO ₂ ) will be described. A method of diagnosing OSAS by SpO ₂ is known as a highly reliable method. This method utilizes the fact that the oxygen concentration in the blood decreases in the apnea state. That is, the more frequently the dip of a certain depth or more in the chart 74 (for example, a decrease of 3% or more) is, the higher the possibility of OSAS is. In this embodiment, the fact that an analysis result equivalent to SpO ₂ is obtained is set as an index of detection sensitivity and specificity for accurately detecting CVHR.

  The contents of the actual analysis processing performed by the CVHR detection device 2 described above will be described. 7 and 8 show analysis results of RR interval time series data of different OSAS patients. Comparing the chart 70 of the RR interval time series data of FIGS. 7 and 8, a large dip appears periodically in FIG. 7, and the CVHR is also observed by visually observing the chart 70 of the RR interval time series data. Easy to judge. On the other hand, in FIG. 8, since there are few large dips and there are variations in the dip depth, it is difficult to determine CVHR by viewing the chart 70 of the RR interval time series data.

The bar graph 72 shows the analysis result of the CVHR detection device 2 of this example. In FIG. 7, a bar graph appears in the bar graph 72 at the same time as the time when a large dip appears in the chart 70 of the RR interval time series data. Further, a bar graph appears in the bar graph 72 at the same time as the time when the dip appears in the SpO ₂ chart 74. From this, it can be said that the CVHR detection device 2 can reliably detect CVHR that can be discriminated by visually observing the chart 70 of the RR interval time series data. It can also be said that the CVHR detection device 2 can obtain an analysis result equivalent to the analysis result of SpO ₂ .

In FIG. 8, it is difficult to discriminate CVHR by viewing the chart 70 of the RR interval time series data, but a dip appears in the SpO ₂ chart 74. In the bar graph 72, the bar graph appears at the same time as the time when the dip appears in the SpO ₂ chart 74. Therefore, the CVHR detection device 2 can detect CVHR with a detection sensitivity equivalent to that of SpO ₂ even with a CVHR that cannot be determined by viewing the chart 70 of the RR interval time series data. I can say that.

The number of times per hour that SpO ₂ has decreased by a predetermined amount (that is, the number of dip in chart 74) is referred to as an oxygen desaturation index (ODI). In particular, the number of times per hour that SpO ₂ has decreased by 3% or more is referred to as 3% ODI. It is known that subjects with a 3% ODI of 15 / hr or higher are likely to have moderate or higher OSAS.

FIG. 9 shows a correlation diagram between the analysis result of the CVHR detection device 2 and the analysis result of SpO ₂ . Based on the RR interval time series data during bedtime overnight, the frequency (vertical axis) of CVHR per hour was calculated using the CVHR detector 2. The horizontal axis indicates 3% ODI per hour when SpO ₂ is measured under the same conditions. The condition of the test subjects is 62 persons including OSAS patients and non-OSAS persons. There are 36 subjects (58%) whose 3% ODI is greater than 15 / hr. One plot shows data for one subject.

The correlation coefficient of the frequency of CVHR detected by CVHR detector 2 (times / hr) and 3% ODI r is 0.90, described ratio ^{r 2} is 81%. This shows that the analysis result of the CVHR detection device 2 and the analysis result of SpO ₂ show a high correlation. The CVHR detection device 2 can accurately detect CVHR.

  Here, the process of performing the processes of S22 to 29 and specifying the significant dip group in S26 is referred to as a data adaptability test. The process of performing the processes of S30 to S35 and specifying the dip group having similarity in S32 is called similarity inspection. The process of performing the process of S36 and specifying the dip group having a periodicity in S38 is called a periodicity inspection. The CVHR detection device 2 detects a dip group that has achieved three items of inspections of data adaptability, similarity, and periodicity as CVHR. Whether or not CVHR can be accurately detected by only one or two of the three-item inspections was actually analyzed and examined.

  FIG. 10 shows a combination table 100 for each inspection item. The Full 86 indicates a combination in the case where all three items of the data adaptability test 80, the similarity test 82, and the periodicity test 84 are adopted. Fixed DT (44 ms) 88 indicates a combination in the case where the similarity test 82 and the periodic test 84 are employed. At this time, the process of S26 was performed using 44 ms as a fixed value instead of the data unique threshold. Simi off 90 indicates a combination in the case where the data adaptability test 80 and the periodicity test 84 are employed. CL off 92 indicates a combination when the data adaptability test 80 and the similarity test 82 are employed.

  ADT only 94 indicates a combination when only the data adaptability test 80 is employed. Simi only 96 indicates a combination when only the similarity test 82 is employed. CL only 98 indicates a combination when only the periodic inspection 84 is employed.

FIG. 11 shows a correlation diagram between the analysis result of the CVHR detection device 2 and the analysis result of SpO ₂ in each combination shown in FIG. The analysis conditions are the same as in FIG. From the correlation diagram of FIG. 10, an explanation rate r ² close to 80% was obtained in Full and Simi off. From this, it can be seen that the analysis result of the CVHR detection device 2 shows a high correlation with the analysis result of SpO ₂ in Full and Simi off.

FIG. 12 shows the OSAS detection capability table 120. FIG. 12 shows sensitivity 110, specificity 112, positive predictive value 114, negative predictive value 116, and concordance rate 118 for each combination shown in FIG. The sensitivity 110, the specificity 112, the positive predictive value 114, the negative predictive value 116, and the coincidence rate 118 are items for evaluating the detection sensitivity of the CVHR detection device 2 in comparison with the analysis result of SpO ₂ . Each evaluation item is calculated from the analysis result of FIG. Here, subjects whose 3% ODI is greater than 15 / hr are defined as “true OSAS patients”. A subject with a 3% ODI of 15 / hr or less is defined as a “fake OSAS patient”. In the CVHR detection device 2, a subject whose CVHR appears more frequently than 15 / hr is determined to be OSAS. The CVHR detection apparatus 2 determines that the subject whose CVHR appears 15 / hr or less is not OSAS.

  The sensitivity 110 is a ratio of the number of subjects determined to be OSAS by the CVHR detection device 2 to the true number of OSAS patients. It can be said that the higher the value of sensitivity 110, the higher the detection sensitivity. The specificity 112 is the ratio of the number of subjects who are determined not to be OSAS by the CVHR detection device 2 to the fake OSAS patients. Higher specificity indicates lower false detection.

  The positive predictive value 114 indicates the ratio of true OSAS patients included in the number of subjects determined to be OSAS by the CVHR detection device 2. The higher the positive predictive value, the higher the detection sensitivity. The negative predictive value 116 indicates the proportion of false OSAS patients included in the number of subjects determined not to be OSAS by the CVHR detection device 2. According to the result of FIG. 12, in all the evaluation items 110 to 118, high values are obtained for Full and Simi off. From the analysis results of FIGS. 11 and 12, CVHR can be accurately detected even with only two items of data adaptability and periodicity (that is, Simi off). By adding similarity analysis to data adaptability and periodicity (ie Full combination), the accuracy of CVHR detection can be further improved.

In the CVHR detection apparatus 2 of the present embodiment, a threshold value related to the dip depth is calculated as a data-specific value from the amplitude of the high frequency component of the RR interval time series data. The CVHR detection device 2 can identify a dip group having a dip depth larger than the data-specific dip depth threshold DD _TH as a significant dip group. Further, the CVHR detection device 2 can identify a similar dip group having similarity in shape from the width and depth of the dip. Furthermore, the CVHR detection device 2 can identify a periodic dip group having periodicity from the time difference between the occurrence times of successive dip. The CVHR detection device 2 can detect a dip group having data adaptability, similarity, and periodicity as CVHR. The analysis result obtained by the CVHR detection device 2 has a high correlation with the SpO ₂ analysis result, and both detection sensitivity and specificity are high even when used for OSAS detection. For this reason, the CVHR detection device 2 can accurately detect CVHR.

Specific examples of the present invention have been described in detail above, but these are merely examples and do not limit the scope of the claims. The technology described in the claims includes various modifications and changes of the specific examples illustrated above.
In addition, the technical elements described in the present specification or drawings exhibit technical usefulness alone or in various combinations, and are not limited to the combinations described in the claims at the time of filing. In addition, the technology illustrated in the present specification or the drawings achieves a plurality of objects at the same time, and has technical utility by achieving one of the objects.

The block diagram which shows the structure of a CVHR detection apparatus is shown. The flowchart of a CVHR detection process is shown. FIG. 3 is a flowchart continued from FIG. 2. FIG. The flowchart of a dip depth calculation process is shown. RR interval time series data is shown typically. RR interval time series data is shown typically. An example of the analysis result of a CVHR detection apparatus is shown. An example of the analysis result of a CVHR detection apparatus is shown. It shows a correlation diagram between the analysis result of the analysis result and SpO ₂ of CVHR detector. The combination table of the inspection item of a CVHR detection apparatus is shown. It shows a correlation diagram between the analysis result of the analysis result and SpO ₂ of CVHR detector. The OSAS detection capability of a CVHR detector is shown.

Explanation of symbols

2: CVHR detection device 16: RR interval time series data input unit 18: dip detection unit 20: dip depth calculation unit 22: heart rate variability index calculation unit 24: individual threshold value determination processing unit 26: dip width calculation unit 28: Dip interval calculation unit 30: Dip group identification unit 32: Other calculation unit 34: Storage unit 36: Operation unit 38: Display unit

Claims (8)

A device for detecting CVHR associated with an apnea attack or hypopnea attack of sleep disordered breathing,
Time-series data input means for inputting time-series data of heartbeat or pulse interval or time-series data of heartbeat or pulse frequency;
Peak detecting means for detecting a plurality of local peaks from the time series data;
Peak height calculating means for calculating the height of each of the plurality of local peaks;
A heart rate variability index calculating means for calculating a heart rate variability index reflecting autonomic nerve activity from the time series data;
Peak height threshold value determining means for determining a threshold value related to the peak height as a value unique to each time series data from the heart rate variability index;
Peak group specifying means for specifying a peak group that has a height greater than the peak height threshold value and is continuous with periodicity from among the plurality of local peaks,
A CVHR detection apparatus comprising: an analysis result output unit that outputs an analysis result including a frequency and a temporal distribution of the peak group specified by the peak group specifying unit. The CVHR detection apparatus according to claim 1, wherein the heartbeat variability index calculating means calculates at least the power or amplitude of fluctuation included in a frequency band of 0.15 to 0.45 Hz of heartbeat variability. The CVHR detection device according to claim 1, wherein the heart rate variability index calculating unit calculates a mean square of a difference value between at least successive beat periods or beat frequencies. The peak group specifying means specifies a peak group having a height greater than the peak height threshold value from the plurality of local peaks, and continues from the peak group with the periodicity. The CVHR detection device according to any one of claims 1 to 3, wherein a peak group is specified. The peak group specifying means specifies three or more continuous peaks in which a time difference between two continuous peaks is within a predetermined range, thereby continuously having the periodicity. The CVHR detection device according to claim 1, wherein a peak group is specified. The peak group specifying means has a height larger than the peak height threshold among the plurality of local peaks, has a similar shape, and is continuous with the periodicity. The CVHR detection device according to any one of claims 1 to 5, wherein a peak group is identified. The peak group specifying means specifies a peak group having a height larger than the peak height threshold from the plurality of local peaks, and has the similar shape from the peak group. The CVHR detection device according to claim 6, wherein a peak group that is continuous is identified from among the peak groups. A computer program for detecting CVHR associated with apnea or hypopnea attacks of sleep disordered breathing,
Time-series data input processing for inputting time-series data of heartbeat or pulse interval or time-frequency data of heartbeat or pulse frequency;
Peak detection processing for detecting a plurality of local peaks from the time series data;
A peak height calculation process for calculating the height of each of the plurality of local peaks;
A heart rate variability index calculating process for calculating a heart rate variability index reflecting autonomic nerve activity from the time series data;
From the heart rate variability index, a peak height threshold value determination process for determining a threshold value related to the peak height as a value specific to each time-series data;
Among the plurality of local peaks, a peak group specifying process for specifying a peak group having a height larger than the peak height threshold and having a periodicity, and
A computer program for causing a computer to execute an analysis result output process for outputting an analysis result including a peak group frequency and a temporal distribution specified by the peak group specifying means.

Images