DE102004028726A1 - Method for measurement of cardiac action in order to determine heart rate and variation of same, using motion of thorax - Google Patents

Abstract

The breathing process of a patient is examined with an appropriate method using the motion of the thorax as an indicator. The part of the signal transmitted from the heart is used for a determination of the sequence of the cardiac action during a particular period and a conclusion regarding the heart rate as well as the variation of same achieved. The result facilitates an elimination of an unwanted cardiogenic artifact from the breathing signal. The method can also be used in order to determine the heart rate and the variation of same when no other method can be applied.

Description

The The invention relates to a method for determining the time intervals of individual Heart actions from measurements that primarily have the goal, a determination of the respiratory excursion / the respiratory rate from the thorax movement of a Derive subjects.

at various diagnostics well-known in medicine Procedure the movement of the thorax as a result of breathing is recorded. Thorax movement is common in pulmonary function diagnostics Measured variables not or only limited quantifiable, but provides a sufficient Description of the respiratory rate and statements on respiratory arrest (central Apneas) associated with a disruption of thoracic respiration. These methods of registering respiration, e.g. in sleep medicine, biofeedback and respiratory training Asthmatics.

to Registration of thoracic breathing will be various known physical Measurement principles used:

1. Piezoelectric transducer

One elastic belt with a piezo strain gauge is around the Thorax of the patient placed and slightly tense. By yourself periodically changing Belt tension due to respiration becomes one of the respiration at the piezo element proportional AC voltage generated.

2. Measurement of thoracic impedance

The change of electric resistance of thorax (chest impedance) the breathing is registered. about an electrode is fed with a low-power AC voltage and over this and a second electrode of the voltage drop measured. The two electrodes are preferably on the left and right Side of the thorax glued.

3. Change the pad pressure on a pneumatic mat

This Procedure has especially in infant diagnostics Dissemination found. The pressure changes are used in a pneumatic mat on which the infant or possibly an adult Subject is lying (baby mat).

at all of these procedures for the registration of thoracic respiration is in addition the respiratory signal also a signal from the heart activity of the subject, the balistocardiogram, with recorded. In some cases, this is due to cardiac activity caused signal component considered as a disorder (cardiogenic Artifact in the apnea detection), and it will be both in the hardware as well as in the software frequency filter used to this signal component to eliminate. In other applications, this signal component forms the basis for the calculation of heart rate.

In the procedures mentioned, the frequency ranges of cardiac and respiratory rates are not clearly separated. The individual differences are too large to limit the frequency range for one or the other frequency to physiologically meaningful values. In a single human, it is assumed that the respiratory rate is about one third to one fifth of the heart rate. Hyperventilates a human, but can also be a ratio of 1: 2. In a Sten's breathing, heart rate and respiratory rate even match. If a measuring method is to be used in different individuals, and this is the rule in medical technology, the task is additionally complicated by the fact that cardiac and respiratory frequencies overlap in the physiologically possible frequency range. For example, respiratory rates of 40 to 60 min ^{-1 are} quite normal in infants. In adults, the heart rate during sleep is below 50 min- ¹ . It must therefore be determined in advance whether the procedure is used in an adult or in an infant.

The The object of the invention is therefore to provide a method by means of which the signal superimposed on the respiratory signal from the heart activity of a Any subjects can be separated.

These The object is achieved by the features mentioned in claim 1.

Out the signal portion of the heart becomes a chronological sequence of heart actions and determines the heart rate and heart rate variance. The Thus obtained result can be used to the unwanted to eliminate cardiogenic artifact in the respiratory signal. There, where no other methods for determining heart rate and heart rate variance are displayed, the method may also serve to determine these quantities.

First, the sum signal of both vital parameters is measured using one of the measuring methods described above (piezo-strain gauges, thoracic impedance, pneumatic pressure changes in a mat, use of acceleration sensors). registered. Independently of the selected measurement method, the hardware for signal acquisition must be designed in such a way that the signal component caused by the cardiac actions has approximately the same intensity as the respiratory signal. This can always be achieved by a suitable choice of the frequency response of the hardware. High-frequency interference must be eliminated by a low pass. The measurement signal is then fed to an AD converter and digitized at a sampling rate greater than 100 Hz. The further signal processing is done by software.

einer FFT (Fast Fourier Transformation) unterzogen, wobei Y_i das i-te diskrete Element der FFT des komplexen Feldes X ist. Das Leistungsspektrum wird nach der Formel

berechnet.According to the invention, the digitized measurement signal according to the formula

an FFT (Fast Fourier Transformation), where Y _{i is} the ith discrete element of the complex field X FFT. The service spectrum is according to the formula

calculated.

there there is a correlation between the sampling rate and the time interval the for the FFT used measured values. The time interval should be chosen so that it big enough is to determine the respiratory rate with sufficient accuracy. The resolution in the Frequency range should be better than 0.05 Hz. The time interval is allowed but not be chosen arbitrarily large, because breathing is really only in a few cases is periodic. A calm and steady breathing is only in deep sleep registered. In REM sleep and wakefulness, 5 to 10 are periodic breaths the rule. Then breathing becomes through movements, convulsions, apnea, a change the frequency or similar interrupted. In frequency space leads this leads to a broadening and flattening of the respiratory Maximum. Assuming a typical respiratory rate of 15 to 20 breaths per adult in an adult, should be a period of time for the FFT 10 to 20 s chosen become.

In the power spectrum, the maximum 1 _atm = max (Y _powi ) to be assigned to the respiration is determined. Since, for physiological reasons, the respiratory rate can not be arbitrarily high, it is determined as a restrictive condition that the maximum to be assigned to the respiration must be smaller than the frequency f _{atm_max} (40 min ^-1 for adults or 60 min ^-1 for babies). For filtering the respiratory signal in the frequency domain, a normal distribution according to the formula F (x) = 1 / √ 2πσ · Exp (- (x - y) 2 / 2σ 2 ) used. The half- _{width of} the normal distribution is σ = f _{atm_max} / 2. The multiplication necessary for the filtering is carried out in such a way that the 1st maximum attributable to the respiration and the maximum value of the normal distribution in the frequency coordinate agree, ie f _atm = y After a normalization of the remaining spectrum, an FFT inverse transformation is carried out in the time domain. In the period you get a signal that only contains the respiratory rate. The time signal allows a determination of the individual breaths as the distance of the successive minima (or maxima). The obtained distances between the minima (respectively maxima) are used to determine the respiratory rate.

berechnet, wobei n die Anzahl der pro Minute bestimmten Atemfrequenzen ist.Furthermore, for a plausibility test in determining the heart rate, a mean respiratory rate f _{atm_mtl is} preferably required over one minute. This will be according to the formula

where n is the number of respiratory rates per minute.

To determine the heart rate is also assumed the power spectrum. The signal generated by the heartbeat is more stable in frequency than the respiratory signal. Typical times for significant changes in heart rate are 10 to 20 s or - an average heart rate in the adult of 60 subordinated to 80 min ^-1 - 10 to 30 individual heartbeats. However, the frequency rarely changes by more than a factor of two. It would be enough to subject a time interval of 4 to 7 s to an FFT in order to record the typical heart rates. For practical reasons, in particular in order not to have to carry out the computationally intensive process of the FFT transformation several times, it is also possible to use the power spectrum that was generated for determining the respiratory frequency.

The cardiac frequency spectrum can take many different forms. Since the heartbeat in the period is more likely to be described with a delta function than with a sine, the spectrum usually has several harmonics. Especially the 1st harmonic can have a much higher intensity than the fundamental wave. In these cases, an unambiguous assignment of the maximums occurring in the spectrum to the heartbeat is problematic only according to the intensity. It is therefore assumed that the heart rate is usually higher by a factor of 3 to 5 than the respiratory rate. The position of the 1st maximum f _atm to be assigned to the respiration has already been determined. The fundamental wave of the heart rate f _hz should therefore be in an interval 3 · f atm <f hz <5 · f atm to be found. The location of this maximum is determined.

unterzogen. Die Parameter n und p werden so gewählt, dass für die Halbwertsbreite σ der Verteilung gilt: 3·fatm = σ = √n·p·(1 – p) The power spectrum is used to determine the heart rate of a filter with a Poisson distribution according to the formula

undergo. The parameters n and p are chosen so that the half width σ of the distribution holds: 3 · f atm = σ = √ n · p · (1 - p)

The multiplication necessary for the filtering is carried out such that the maximum value of the Poisson distribution coincides with the fundamental wave of the heart rate f _hz . After a normalization of the remaining spectrum, a back transformation is performed in the period.

Furthermore, the measurement signal processed in this way is analyzed in the time period. It is assumed that an arbitrary assumed time interval dt the individual heart actions. An initial value, which has been proven in practice, is in infants at dt _children = 0.33 s (= 180 beats / min) and adults in dt _adults = 0.67 s (= 90 beats / min). It is further assumed that each heart action causes a local maximum in the measurement signal. Another plausible assumption is that the heart rate changes slowly from one heart action to the next. In accordance with the assumptions made above, the local maxima are searched in the period of time considered. First, in a time interval of 0 to 2 · dt, the 1st absolute maximum I _{max is} determined. This maximum at the time coordinate t ₁ is regarded as the first cardiac action found and determined as the starting point for the further search. The time coordinate t ₁ is stored. The search for the next maximum starts at the time coordinate t ₁ + 0.7 * dt and continues until another maximum is found. Let t ₂ be the time coordinate at which the second maximum was found. t ₂ + 0.7 · dt is now the time coordinate for the beginning of the next maximum search. The process continues until the entire time period to be analyzed has been completed. As a result one obtains a vector t with n time coordinates t ₀ to t _n . The intensity of the maxima and their time coordinates must correspond to the limiting condition explained below.

bestimmt. Als Maximum wird ein Wert nur dann gewertet, wenn gilt 3·Imtl < Ik < 10·Imtl The limiting condition is determined as follows: First, for the period of time to be analyzed, the mean intensity of all measured values I _mtl according to the formula

certainly. The maximum is a value only if it applies 3 · I monthly <I k <10 · I monthly

With this method, it is possible that not all heart actions are detected. Before a heart rate is calculated from the time intervals of the time coordinates stored in the vector t, the vector t is processed as follows: All differences Dt = t _k -t _k-1 are checked. If a difference Dt is found that matches the condition 1.4 · dt <Dt <2.8 · dt suffice it can be assumed that one and only one heart action was not recorded. An additional time coordinate is added to the vector t between the time coordinates t _k-1 and t _k at t _k-1 + (t _k -t _k-1 ) / 2. Distances between two time coordinates greater than 2.8 dt are not corrected. The distances between the maxima thus obtained are used to determine the heart rate. Again, only the time intervals are included in the calculation, that of the condition 0.5 · dt <Dt <1.5 · dt where dt is the moving average of the distances between two time coordinates.

For the determination of the moving average, a frequency distribution of the differences Dt is generated. It is assumed that the heart rates in humans can be between 30 beats / min and 200 beats / min. In the time domain, the intervals between the individual heart actions are from 2 s to 0.3 s. The differences Dt found are arranged in a corresponding grid, for example of 0.1 s increments. The moving average in the period of time to be analyzed is considered to be the distance Dt _mtl in which most of the differences Dt lie.

genügt. Nur für diesen Fall wird der Anfangswert für den Abstand zwischen den einzelnen Herzaktionen dt nach der Formel dtneu = (dtalt + Dtmtl + 1/fatm_mtl)/3korrigiert.To rule out errors in the determination of the important for the calculation in the next period size Dt _mtl by sporadic disturbances, a plausibility test is performed with the average respiratory rate. The initial value for the distance between the individual heart _actions dt is only changed if Dt _{mtl of} the condition

enough. Only in this case is the initial value for the distance between the individual heart actions dt according to the formula dt New = (dt old + Dt monthly + 1 / f atm_mtl ) / 3 corrected.

The result is a continuous series of time coordinates t _k , which correspond with high reliability to the heart actions of the examined person.

If Cardiac actions always at registration of chest respiration to be detected as a disruption considered in the measurement procedure, with the knowledge of the dates, to which this disorder occurs, appropriate action (for example, a smoothing) be targeted, which targeted this disorder eliminated.

Conversely, it is possible to carry out a calculation of the heart rate at equidistant intervals, eg per second, with knowledge of the time coordinates t _k .

To be able to make a statement about the heart rate variance, it is customary to consider time intervals of> 3 min. The classical way to determine the heart rate variance from the R-wave of the ECG is to determine the time between two consecutive R-waves Dt _R (beat-to-beat distance) and a function for the time dependence in the form dt R = Dt R (T) from the values Dt _R (eg described in "Special Report" Circulation 93 (5) 1996) This function is subjected to an FFT and evaluated in the frequency domain The function dt _R = Dt _R (t) no longer contains the heart rate itself The frequencies in the spectrum are assigned to various physiological processes (for example described by Shimoni, Guilleminault, Sasanabe, Hirota, Maekawa, Kobayashi in Sleep 19 (5) 1996.) The best studied is the respiratory sinus arrhythmia of the heart rate.

By the algorithm described above for detecting the heart actions and thus also the beat-to-beat intervals, it would also be possible here to form a function of the form dt _R = Dt _R (t). However, it is simpler to sum up (mapping) and evaluate the already existing power spectra over 3 to 5 minutes.

The mean distance of the heart _actions Dt _mtl is formed over the considered period of 3 to 5 min. The associated frequency is F _mtl = 1 / 2π · Dt _mtl . The spectrum is evaluated in the frequency interval F _mtl ± 2 · f _atm . The two largest maxima with the intensities I ₁ and I _{2 are} searched in this interval, where I ₁ > I ₂ If the maximum I ₂ <0.5 * I ₁ , it is assumed that the sinus arrhythmia is only slightly pronounced and the heart rate is described by a frequency maximum. The value for the heart rate variance is assigned the width of the maximum at 0.5 · I ₁ . If a clear double maximum for the heart rate is found by the sinus arrhythmia, the value of the heart rate variance is assigned the sum of the respective half widths at 0.5 * I ₁ and 0.5 * I ₂ .

With the described procedure it is possible to have a time dependence dF of the heart rate variance in the form dF = DF (t) to determine. If the temporal resolution of 3 to 5 minutes is insufficient, the algorithm can also be applied to shorter time intervals and with a corresponding overlap.

differences in the heart rate variance depending from the time are in particular for the determination of REM sleep phases of importance. In REM sleep the heart rate variance is significantly increased. The measurement of this size is so an indirect possibility To determine REM sleep phases.

Claims (5)

Method for determining the time sequence of cardiac actions for the determination of the heart rate and the heart rate variance from the superposition of the measured excursion results by a balistocardiogram, in which - first the thorax movement as a result of breathing is registered with a diagnostic method known in medicine; - the obtained measured values are digitized; - then decomposed the measurement signal into 10 to 20 s sections and each of these sections is subjected to an FFT; - The obtained spectrum with a normal distribution, which is adapted to the respiratory rate, filtered back into the period transformed back and in the period of determination of the current and a mean respiratory rate is performed; - Subsequently, the same spectrum a second time with a Poisson distribution in which the position of the maximum is derived from the mean respiratory rate, filtered transformed back into the period; - The signal is analyzed in the period with a search algorithm for the heart actions; - The search algorithm is used repeatedly and itteratively in consideration of the time Heart attacks found recently and using the relationship that the heart rate is usually three to five times the respiratory rate, is adapted to the actually existing physiological conditions; - Thereafter, from the cardiac actions found, taking into account limiting criteria, the current and a mean heart rate is determined; - a mapping of the spectra over a period of 3 to 5 min in the frequency space is performed; - Then in the frequency space, the intensity maximum is determined, which is to be assigned to the heart actions; and finally the heart rate variance is determined from the line width of the intensity maximum. Method according to claim 1, characterized in that that the time sequence of cardiac actions determined by the procedure to correct the cardiogenic artifact in the thorax motion measurements as a result of breathing. Method according to claim 1, characterized in that that the determination of the heart rate takes place in real time. Method according to claim 1, characterized in that that in sleep diagnostics by determining the heart rate variance Dedicated REM sleep phases can be. Method according to claim 1, characterized in that that the algorithms described by controllers or CPU with low computing power, especially on a PDA (Personal Digital Assistant) become.