DK176955B1 - Highly Efficient HRV Detection - Google Patents

Description

, DK 176955 B1

Highly efficient HRV detection

TECHNICAL FIELD

The invention relates to a method of determining heart-rate variability (HRV) of an individual, such as a human. In one embodiment the invention provides relatively high accuracy as well as relatively low computational requirements which may translate into simpler processing and thereby may enable relatively cost effective measurement devices.

BACKGROUND ART

The heart-rate variability has previously been identified as a potential (early) indicator of the onset of tissue rejection of a transplanted heart. The HRV is commonly determined based on the measurement obtain via an electrocardiograph as an electrocardiogram.

The mechanical events of the heart are preceded and initiated by the electrochemical activity of the heart (i.e., the propagation of the action potential). The electrocardiograph is arranged to transform the electrochemical activity of the heart into a form visible to the human eye. This visual representation is known as the electrocardiogram (ECG). During the recording of an ECG, electrodes are attached to the body surface. The electrodes are specially treated to allow the charge carriers within the electrodes (electrons) to communicate with the charge carriers within the body (ions) via electrochemical exchange. Attaching electrodes to the body surface allows the voltage changes within the body to be recorded after adequate amplification of the signal. Commonly a galvanometer within the electrocardiograph is used as a recording device. Galvanometers record potential differences between two electrodes. The ECG is merely the recording of differences in voltage between two electrodes on the body surface as a function of time, and is usually recorded on a strip chart. When the heart is at rest, diastole, the cardiac cells are polarized and no charge movement is taking place. Consequently, the galvanometers of the ECG do not record any deflection. However, when the heart begins to propagate an action potential, the galvanometer will deflect since an electrode underneath 2 DK 176955 B1 which depolarization has occurred will record a potential difference from a region on the body under which the heart has not yet depolarized.

A complete heart cycle is known as a heartbeat. On an ECG, a normal heartbeat has a distinctive signal. Initially, the galvanometer notes a relatively short duration rounded positive deflection {known as the P wave), which is caused by atrial depolarization. Subsequent to this, there is a small but sharp negative deflection (known as the Q wave). Next, there is a very large and sharp positive deflection (known as the R wave), after which there is a sharp and large negative deflection (known as the S wave). When the Q, R, and S waves are taken together, they are known as the QRS complex. The QRS complex is caused by ventricular depolarization. Subsequent to the QRS complex, is a relatively long duration rounded positive deflection (known as the T wave), which is caused by ventricular repolarization.

The ECG, in practice, uses many sets of electrodes. But these electrodes are so arranged on the surface of the body such that the signals received will have the similar shape as that just described. Well-known bipolar pairs of electrodes are typically located on a patient's right arm (RA), left arm (LA), right leg (RL) (commonly used as a reference), and left leg (LL). Unipolar electrodes referenced properly are referred to as V leads and are positioned anatomically on a patient's chest according to an established convention (labeled as follows as Leads V1-V6). In heart monitoring and diagnosis, the voltage differential appearing between two such electrodes or between one electrode and the average of a group of other electrodes represents a particular perspective of the heart's electrical activity and is generally referred to as the ECG. Particular combinations of electrodes are called leads. For example, the leads which may be employed in a "gold standard" 12-lead electrocardiogram system are: Lead l=(LA-RA) Lead ll=(LL-RA) Lead lll=(LL-LA) Lead aVR=RA-(LA+LL)/2 Lead aVL=LA-(RA+LL)/2 Lead aVF=LL-(LA+RA)/2 Lead V1=V1-(LA+RA+LL)/3 Lead V2=V2-(LA+RA+LL)/3 Lead V3=V3-(LA+RA+LL)/3 Lead V4=V4-(LA+RA+LL)/3 Lead V5=V5-(LA+RA+LL)/3 Lead V6=V6-(LA+RA+LL)/3 3 DK 176955 B1

Thus, although the term "lead" would appear to indicate a physical wire, in electrocardiography the term actually means the electrical signal taken from a certain electrode arrangement as illustrated above.

Over the years, health care professionals have built up a body of knowledge wherein they have learned to correlate variations in the ECG with different diseases and heart defects. Formally, this process of correlating is known as "electrocardiography." Electrocardiography, as practiced by human cardiologists, is primarily a visually-oriented art in that the human cardiologists visually inspects a waveform tracing of electrocardiographic measurements taken over time, and on the basis of the morphological (i.e., shape) changes of the waveforms making up the waveform over time the human cardiologist makes a diagnosis of heart function.

Apart from human interpretation of the shape of ECG’s several objective measures has been defined to enable more automatic diagnostics or at least assist in diagnostics by providing objective measures which derivable from the ECG by machines computation. One such measure is Heart Rate Variability (HRV) which is a measure of the variation in the time from R peak to the next R peak in a patient (also referred to as the R-R interval). It has been reported that that heart rate variability can be decreased with severe coronary artery disease, congestive heart failure, aging and diabetic neuropathy. Accordingly, accurate determination of the HRV may provide an important diagnostic tool.

DISCLOSURE OF INVENTION

However, the present inventors have realized that the HRV may be indicative of a long range of ailments and/or in many case functions as a good monitoring measure. One example of such use of HRV is in the provision of beta-blockers. Here the inventors have found HRV to be a good indicator of insufficient, sufficient and over medicating of the patient.

In many instances useful HRV detection requires a temporal resolution of about 10ms or less. Accordingly, in one embodiment the method and systems of the invention are arranged to determine the HRV with a resolution of 10ms or less, such as 1 ms or less, such as 500ps or less, such as 250ps or less, DK 176955 B1 4 such as 100ps or less, such as 50ps or less, such as 10ps or less. At the same time the ECG signal may be severely influenced by interference signals due to muscle movements, net hum and noise in general. Most prior art measurement of HRV with sufficient accuracy therefore require the patient to remain still in a clinical environment where electrical noise source are minimized. Obviously, such requirements are not easily compatible with everyday measurement or even continuous monitoring.

Traditionally, the temporal distance between two heartbeats is determined as outlined in Fig. 1 .The band-pass filtered ECG signal is additionally filtered in a Matched Filter, which is normally adapted to an estimate of an average QRS complex. Previously, before the dissemination of digital signal processing, the amplitude spectrum was approximated by an average QRS amplitude spectrum. Nowadays, the approximation may be carried out in the time domain, and thus a better result may be obtained. The purpose of the Matched filter is commonly to eliminate as much disturbing noise as possible.

After the Matched Filter follows a pdetector, which traditionally detects the timing of the peak of the R wave. In some embodiments of peak detectors the filtered ECG signal is differentiated, and the temporal position of the zero crossing of the differentiated signal is determined to determine the position of the peak. This is in one embodiment preferable when the peak detector is implemented in analog electronics. For normal heart rate measurement in clinical use, where one merely wish to monitor hear rate (HR) with low precision, the traditional systems are normally sufficient.

As will be discussed further below, the noise contribution due to muscle contraction and mains hum are not readily filtered out without significantly distorting the R peak normally smoothing out the R peak and thereby reducing the accuracy with which the peak position may be determined. At the same time the noise contributions will also distort the R peak and interfere with the determination of the peak position. Accordingly, the traditional measurement methods are not suitable for accurate determination of HRV in the presence of noise present during out-of-clinic measurements.

5 DK 176955 B1

The inventors has surprisingly found that it is possible to achieve at high resolution HRV measurement by a combination of filtering the ECG signal comprising a QRS signal so the QRS is distorted in order to reduce noise sufficiently and applying a mathematical function, a locater function, for determining the position of the R peak. The locater function is designed in such a way that while the relative position of the R peak may be shifted due to the distortion, this shift is stationary within the desired accuracy. In other words, while the frequency spectrum of the QRS complex may be distorted due to the filtering the locater function only, at least within the desired accuracy, depends on parts of the frequency spectrum where the fourier transform (FT) of the QRS complex is either not distorted or distorted in a stationary fashion. Accordingly, in one embodiment the invention relates to a method of determining HRV of from an individual comprising a. Sampling an ECG signal, r(t), from said individual to obtain a sampled ECG signal, said sampled ECG signal comprising one or more QRS complex b. filtering at least part of said ECG signal and/or said sampled ECG signal to obtain a filtered ECG signal having one or more filtered QRS complex c. applying a locater function to determine a temporal location of one or more QRS complexes of the filtered ECG signal d. applying said temporal location(s) in a calculation of the HRV of the individual wherein said locater function comprises a calculation of the gravity center of the filtered QRS complex or part thereof, such as the R-peak

In one embodiment R(co) is the Fourier transform of a single QRS complex of said one or more QRS complex and Rf(m) is the Fourier transform of a single filtered QRS complex of said one or more filtered QRS complex wherein said one or more filtered QRS complex is distorted relative to said one or more QRS complex.

6 DK 176955 B1

In the context of the present text distorted refers to shape of RKco) differs from the shape (i.e. apart from a scaling) of R{ta)by more than 10%, such as by more than 20%, such as by more than 30%, such as by more than 40%, such as by more than 50%, such as by more than 75%, such as by more than 95%.

In one embodiment this deviation may be calculated in the absence of noise sources.

In one embodiment the temporal location of the QRS complex correspond to the temporal location of the R wave also referred to as R peak. In one embodiment other parts of the QRS complex are utilized to locate the temporal location of the QRS complex, such as the entire QRS complex. The temporal location a QRS complex may be utilized to calculate the HRV in a plethora of ways as will be apperant to the skilled person. In one embodiment a series of temporal locations are recorded and this time series is analysed spectrally such as via one or more Fourier transforms. In one embodiment this is combined with statistical methods such as correlation via auto-regressive modeling. In one embodiment the temporal location is utilized to determine the temporal distance to the previous or next QRS complex. The temporal distance is then applied in the calculation of the HRV. In one embodiment the method comprises having stored a series of temporal locations and/or a statistical values relating to a series of temporal locations and applying said stored value(s) in calculation of the HRV.

As will be discussed further below, the inventors has discovered that a locator function comprising the calculation of the gravity center of the QRS complex or parts thereof may be particularly effective in determining the HRV with high precision. Accordingly, in one embodiment the invention relates to a method of determining HRV of from an individual comprising e. Sampling an ECG signal, r(t), from said individual to obtain a sampled ECG signal, said sampled ECG signal comprising one or more QRS complex DK 176955 B1 s 7 f. filtering at least part of said ECG signal and/or said sampled ECG signal to obtain a filtered ECG signal having one or more filtered QRS complex g. applying a locater function to determine a temporal location of one or more QRS complexes of the filtered ECG signal h. applying said temporal location(s) in a calculation of the HRV of the individual where R(to) is the Fourier transform of a single QRS complex of said one or more QRS complex and Rf(æ) is the Fourier transform of a single filtered QRS complex of said one or more filtered QRS complex wherein said locater function comprises a calculation of the gravity center.

In one embodiment the invention relates to a system for measuring a HRV of an individual comprising an antenna unit suitable for receiving an ECG signal, a signal processer unit and an interface wherein said signal processer unit is arranged to perform any of the methods of the invention.

In one embodiment the invention relates to the use of the method and/or devices set fourth. In one embodiment the invention relates to a diagnostic aid arranged to perform the methods of determining an HRV according to the invention. In another embodiment the invention relates to a method of monitoring a subject comprising measuring a HRV from said subject via a method and/or system for determining an HRV according to the invention.

BRIEF DESCRIPTION OF DRAWINGS

The invention will be explained more fully below in connection with a preferred embodiment and with reference to the drawings in which: FIG. 1 shows prior arts system for measurement of heart rate (HR), FIG. 2 shows example of centre of gravity determination for R wave, FIG, 3 shows relative error (percent of sample interval) at the trig point for common prior arts system (A) and the a method according to the 8 DK 176955 B1 g invention involving the calculation of the gravity center (B) with sliding sampling pattern and additive white noise, and FIG. 4 shows an illustration of the reduction of mains hum with a 4th order FIR low-pass filter.

Fig. 5 shows the spectral contents of a QRS complex of a healthy athletic young man.

The figures are schematic and may be simplified for clarity

Further scope of applicability of the present invention will become apparent from the detailed description given hereinafter. However, it should be understood that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from this detailed description.

DETAILS OF THE INVENTION

The invention is defined by the features of the independent claim(s). Preferred embodiments are defined in the dependent claims. Any reference numerals in the claims are intended to be non-limiting for their scope.

Some preferred embodiments have been shown in the foregoing, but it should be stressed that the invention is not limited to these, but may be embodied in other ways within the subject-matter defined in the following claims. For example,

Commonly, in a traditionally digitally implemented HR system no interpolation is performed between samples (of the ECG signal) and sampling rates are kept within the norm for human inspection of ECG, which leaves a relatively poor resolution for HRV detection. Instead some prior art HRV systems attempt to achieve a higher temporal resolution by increasing sampling rate and/or interpolation. Apart from being laborious for the detection equipment this approach will in most cases not alleviate the problem of noise. However, DK 176955 B1 s g the inventor has found that in one embodiment interpolation is preferably incorporated into the locator function as the involvement of multiple weighted points (i.e. interpolation) may provide improved accuracy and stability to the locater function.

One particular suitable locater function has been identified by the inventor to be a locator function comprising calculation of the gravity center of the QRS complex. The suitability will be discussed further below in relation to implementation and noise. In one embodiment the ECG signal is filtered and a peak detector is applied to detect the R peak similarly to the prior art systems discussed above. In one embodiment only a section {i.e. a set of sample) around the estimated location of the R peak is used in the locater function.

Therefore in one embodiment the method comprises applying peak detector to said ECG signal, said sampled ECG signal, said filtered ECG signal and/or or intermediates thereof, which allows the locater function to ignore substantially all of the filtered ECG signal not comprising one or more filtered QRS complexes.

In one embodiment negative values of the ECG are ignored so that only the R peak itself is used in the locater function. In one embodiment the negative values are utilized as well, i.e. by applying by taking the absolute value of the samples, and thereby more or the entire QRS complex is applied in the locater function. The absolute value may in one embodiment be taken of the absolute value of the ECG signal, the samples ECG signal, the filtered ECG signal and/or intermediates thereof.

The advantage of only taking a section of the signal at or around the R peak is that noise contribution in the time where there is no QRS complex may be relatively high relative to the measured signal. Furthermore, in one embodiment the gravity center provide a more reproducible determination of the R peak location as more of the peak is considered and therefore the method is less sensitive to shape changes of the R peak between heart beats. Furthermore, in one embodiment one advantage of the invention is that the accuracy of the determination of the temporal location of the QRS complex is less dependent upon the sample frequency relative to a peak detection and 10 DK 176955 B1 therefore less samples may be required for the filtered ECG signal. Less samples may in one embodiment translate to simplified implementation and/or less power consumption of the processing. The latter may be important for prolonged used of a portable system. In one embodiment the HRV measurement is implemented with out a multiplier function as this function consumes a relatively high amount of power compared to summation and subtraction.

In the remainder of the text the locater function comprising gravity will used as an exemplary locator function. However, it should be kept in mind that other functions sharing the properties discussed above may be applied without departing from the scope of the invention.

In one embodiment of calculating the gravity center as part of the locater function the the QRST or QRS complexes are regarded as a periodical repetition of an a-periodic function r(t) £ 0, of duration Tr. For a single period of r(t), the centre of gravity of the function tp on the time axis is determined by Γ t-r(t)dt t = - 1 1

Ip I. I

[ r(t)dt

R

We presume that a well-defined centre of gravity exists. It may be found that tp =—[jR(co)] forco = 0, 2.1 δω

Where R(co) is the Fourier transformation of r(t).

From formula 2.1 we see that it is the low frequencies that determine the position of the centre of gravity. This is important when r(t) is sampled with the sampling frequency fs = 1/Ts. As time limited functions, i.e. functions r(t), for which it applies that r(t)=0 for |t| > TR, have spectra that spread all over the frequency axis and thus have to be band limited for sampling and noise removal, formula 2.1 means that this does not affect the determination of tP considerably. The fact that the locater function only depends on a relatively narrow spectral window (in this case near DC) allows for easy but very 11 DK 176955 B1 restrictive filtering and therefore further noise reduction. Accordingly, in one embodiment the locater function is substantially independent of values of Rf(to) for frequencies higher than 1Hz, such as higher than 5 Hz, such as higher than 10Hz,such as higher than 20 Hz, such as higher than 30 Hz, such as higher than 40Hz, such as higher than 50Hz, such as higher than 60 Hz, such as higher 70Hz, such as higher than 80 Hz, such as higher than 90 Hz, such as higher than 100 Hz, such as higher than 110 Hz, such as higher than 120 Hz, such as higher than 130 Hz. In one embodiment the locater function depends substantially only on a relatively narrow frequency interval such as less than 100Hz, such as less than 50 hertz, such as less than 30 Hz, such as less than 20 Hz, such as less than 10Hz. In one embodiment the filtering comprises an adaptive filter to allow for further noise reduction. In one embodiment this filer comprises feed back to allow further refinement of the filter values applied and hence its frequency response.

However, it does require that the distortion of the QRS complex discussed above is stationary within the frequency regime where the locater function is dependent. So in one embodiment the filtering of the method of the invention is arranged to only introduce substantially stationary distortion in this frequency interval. Exemplarily, the distortion due to filtering is in one embodiment where the locater function depends substantially only of low frequencies, such as the gravity center, stationary at least for frequencies equal to or below 100Hz, such as equal to or below 90Hz, such as equal to or below 80Hz, such as equal to or below 70Hz, such as equal to or below 60Hz, such as equal to or below 50Hz, such as equal to or below 40Hz, such as equal to or below 20Hz, such as equal to or below 10Hz, such as equal to or below 5Hz, such as equal to or below 1Hz.

The centre of gravity may thus with good precision in one embodiment be determined by ,P = V'r(n) 3.1 Σ»/">

Where r(n)=r(t) for simplicity and t = nTs.

DK 176955 B1 s 12

Formula 3.1 can relatively efficiently be implemented in a typical signal processor. In one embodiment is has been found that apart from the updating of a sample counter, only 1 multiplication and 2 accumulations per sample must be carried out.

As discussed above the collected noise in connection with ECG signals stems from several independent sources comprising one or more of: • Noise stemming from other muscle groups.

• Mains hum.

• Base-line offset.

• Base-line migration.

The centre of gravity determination for interpolation has been found preferable in some embodiment, apart from being cost-effective, it is generally precise in connection with noiseless signals. However, it may appear to be noise sensitive under influence by different types of noise, as no average is performed. However, for the determination of HRV with high precision the inventor has surprisingly found this not to be the case. In the following the noise sensitivity is analysed for the different noise components. As the centre of gravity determination is linear, and the single noise sources are independent, the total uncertainty can easily be determined from the centres of gravity of the single source.

As discussed above, in one embodiment method comprises a peak determination of the R wave where filtering and a peak detector is applied similarly to that of the prior art. This is in one embodiment followed by a centre of gravity determination for the R wave, and only for this, by only considering samples of ECG signal within a set spacing from the position provided by the peak detector. The principle is illustrated in Fig. 2.

Noise stemming from other muscle groups

At least for the sake of modelling the noise w(n) from the muscle groups other than the heart itself from which the electrodes also pick up signals can be regarded as Gaussian (i.e. Poisson distributed with large parameter), stationary white noise with a mean value 0 and variance ow2. The noise 13 DK 176955 B1 g naturally reduces the accuracy in the calculation of the centre of gravity, and this must be taken into consideration in the practical implementation. If signal samples with a small amplitude of r(n) "far” from the centre of gravity are included, they will contribute relatively much to the final result according to formula 3.1. The final noise contribution tw to tp will be '~ΣΙ«»>+Γ<η)) ^ 12 " Σ>> with the variance ΣΝ° 22 ^-- 2.2 ς;»

Figure 3 shows the error (percent of sample interval) of peak determination for prior arts method with peak detection (A) and the method according to the invention (B) with a sliding sampling pattern and additive white noise. This illustrates how a sliding sampling pattern combined with white noise affects the determination of the trig point, i.e. the temporal location of the R peak and/or QRS complex as determined by the applied method. In this case the R wave has been modelled by a Hanning impulse with amplitude 2 width of 90 samples. The noise is Gaussian white noise with a spreading of 0.001. The determination of the gravity center takes place with a 10 times lower sampling frequency than that applied by the peak detection, as a maximum of a total of 7 samples is included in the determination.

The prior arts method (A in Fig. 3), which consists of time localization of the sample with a peak value, shows a linear relation between the relative deviation of the sampling pattern and the percentage error. It is observed that the relatively low noise level creates relatively large spreading at the correct temporal location of the R peak.

DK 176955 B1 s 14

The method according to the invention {B in Fig. 3) shows a faint dependence on the sampling pattern. The error stemming from white noise, as seen by the variance around the otherwise horizontal (B) or sloping (A) lines, is in this case substantially the same as for the prior arts method.

The R wave may roughly be approximated by the main loop and the two largest lateral loops of a sine function, i.e. a approximately a square function in the frequency domain. The upper limiting frequency is commonly app. 35 Hz. In prior arts systems HRV is regarded as a additional feature to the general ECG, however, as the inventors have found narrowing filtration to about 35Hz from commonly used 125 Hz will often not affect the QRS complex radically but contribute to a reduction in noise. However, as can be seen from Fig. 5 the spectrum of a QRS complex may contain significant spectral content above 35Hz and even above 60Hz. In one embodiment is possible in a HRV system (i.e. an ECG system where a normal ECG graph is not required) to sample much more slowly than the 250 Hz necessary if all frequencies in the general ECG is required. In one embodiment such a reduction in sample rate facilitates easier and cheaper implementation.

Furthermore, in the present example where the gravity center is applied as part of the locater function after a peak detector has identified the presence of the QRS complex, frequencies about 0 Hz influence have the most significant influence on the HRV determination, cf. Eq. 2.1, so in one embodiment an even slower even slower sampling rate may be sufficient. Accordingly, in one embodiment the sampled ECG signal comprises samples representing 400 samples per second or less, such as 300 samples per second or less, such as 200 samples per second or less, such as 150 samples per second or less, such as 100 samples per second or less, such as 75 samples per second or less, such as 50 samples per second or less, such as 35 samples per second or less, such as 25 samples per second or less.

Having a low number of samples in the sampled ECG signal may in one embodiment allow for lower computation speeds and/or simplified computational design in the implementation of the method. However, in one embodiment it is advantageous to initially have a faster A/D conversion such 15 DK 176955 B1 as better results of digital filtration and/or easier implementation of the A/D converter. Accordingly, in one embodiment the A/D converter over samples relative to the sampled ECG signal by a factor of 2 or more, such as by a factor of 4 or more, such as by a factor of 4 or more, such as by a factor of 8 or more, such as by a factor of 16 or more, such as by a factor of 32 or more, such as by a factor of 64 or more, such as by a factor of 128 or more. In one embodiment followed by a filter suitable for preventing aliasing effects in the sampled ECG signal.

Usually the low-pass filtration is carried out with a matched filter resembling an average QRS complex. In a pure HRV system, it may in one embodiment be advantageous to apply a substantially sine shaped filter in the A/D converter, i.e, a square filter in the frequency domain. The variance ow2of the white noise can thus be reduced by low-pass filtration of the total ECG signal as a bandwidth reduction with a factor γ also reduces the variance of the noise by a factor y.The reduction of the bandwidth also affects the QRS complex to a certain extent, but as it is the frequencies around 0 that are crucial, a very drastic reduction of the band width is necessary before this affects tp noticeably. In all circumstances, the change that distortion imposed on R(u)) is the same for all complexes, so that the change does not affect the peak to peak time (i.e. the HRV) or can relatively easily be compensated for.

This is not a feature of the prior arts method. Accordingly, the present invention allows noise reduction by bandlimiting below 125Hz and even further noise reduction than band limiting below about 35 herz. Accordingly, in one embodiment the filtering of the method of the invention comprises eliminating frequencies ranging from an upper frequency to a frequency less than 125 Hz, such to a frequency less than 100Hz, such as to a frequency less than 75Hz, such as to a frequency less than 50Hz, such as to a frequency less than 45Hz, such as to a frequency less than 40Hz, such as to a frequency less than 35Hz, such as to a frequency less than 30Hz such as to a frequency less than 25Hz, such as to a frequency less than 20Hz, such as 16 DK 176955 B1 to a frequency less than 15Hz, such as to a frequency less than 10Hz. In one embodiment said upper frequency is125 Hz or more, such as 500 Hz or more.

Also HRV based on peak value determination benefits from low-pass filtration, but if the limiting frequency in the low-pass filter is set below app. 35 Hz, the top of the R wave will flatten out and prevent accurate peak detection.

Mains hum

Sinusoidal signals stemming from the mains grid can be found almost everywhere. In order to reduce the common mode hum problem prior arts ECG systems often apply a reference electrode located in a place on the patient where there is very little QRS complex detectable (often the right leg).

In one embodiment a system according to the invention where the system is battery operated and relatively short wires, preferably running close to the body, are used to connect the electrodes mains hum may still influence the HRV measurement. In the example in figure 4, the hum is so large that it can clearly be felt by the patient.

In one embodiment where interference from mains hum turn out to be a problem, a simple system for destructive interference could be implemented.

It is estimated that such a system would require app. 20 cycles per sample in a signal processor. Alternatively, a simple FIR or MR notch filter can be implemented as the hum frequency if often well known and fairly stable. The destructive intereference approach s in one embodiment preferable for applications where the full ECG may be utilized, as it affects the shape of the QRS complex to a lesser degree. Therefore in one embodiment the filtering of the method according to the invention comprises applying a mains hum filter suitable for eliminating noise contribution from the general net supply, such as a filter with substantially a zero point at or around the frequency of the general net supply.

In figure 4 the effect of a 4th order FIR low-pass filter with the neutral point in 50 or 100 Hz can be seen. An extremely satisfactory elimination of the hum has been achieved without affecting the QRS complex noticeably, The QRS

17 DK 176955 B1 complex has been simulated as the product of a cosine and a Hanning function. The filter output has been scaled and offset for the sake of the plot.

Base-line offset and migration

The ECG signal has been high-pass filtered before the sampling, and thus the filtered signal does not contain any DC component. As the shape and amplitude of the QRS complex may not be exactly constant, this can cause temporal local shifts of the base-line, which again may affect the centre of gravity determination. Accordingly, in one embodiment the said filtering in the method of the invention comprises a base line filter suitable for substantially eliminating base line wander. In one embodiment this filter is implemented as an analog filter prior to A/D conversion.

Considerable simulations with different amplitude offset showed that in one embodiment the error introduced by a change of offset of 50 % of the amplitude of the R wave can create a maximum error of app. 20 % of the sample time interval at positive offsets and somewhat larger for negative offsets. As expected, there were a linear relation between offset and error.

In one embodiment the method further comprises means for eliminating errors stemming from base-line offset and migration. In one embodiment this is implemented by supervising the content of the lowest frequencies with subsequent elimination of the HRV from unreliable sections, i.e. were the base line change is too high. In one embodiment it is without importance if, for example, 10 % of the measurements are eliminated in this fashion.

Alternatively, in pure HRV systems the lower limiting frequency can be raised to 10-20 Hz. While this may affect the determination of the slope as 0Hz according to Eq. 2.1 the alteration may in some embodiment be considered stationary and therefore does not affect the HRV determination according to the method of the invention. SO that in one embodiment the base line filter comprises a high pass filter with a cut off at a frequency less than 40 Hz, such as less than 30Hz, such as less than 20 Hz, such as less than 10 Hz, such as less than 5 Hz, such as less than 1Hz. In one embodiment this filter is implemented as an analog filter, i.e. prior to digital to analog (D/A) conversion 18 DK 176955 B1 as such as filter is relatively easily implemented in this way and as it provides simplified operation of the (D/A conversion).

In implementing the system of the invention the signal from the subject must be picked up by the antenna unit in order to be processed. In principle this could be any suitable detector for detecting a ECG signal. In one embodiment the antenna unit is an antenna, such as an antenna integrated into a necklace so that it may be convenient to carry around. In one embodiment the antenna unit comprises two or more electrodes located on at least one electrode patch as electrodes are fairly effective in detecting body potentials. While the electrode may in one embodiment be placed on a single patch it may be preferable to utilize a larger spacing between the electrodes to improve the detectability of changes in the body potential originating from a heart beat. To simplify the use of the system the antenna unit is in one embodiment formed by less than 5 separated electrodes, such as by less than 4 separated electrodes, such as less than 3 separated electrodes, such as by 2 separated electrodes. In one embodiment one or more of the electrodes are invasive electrodes such as needle electrodes. Invasive electrode may in one embodiment be more comfortable for prolonged used and may in some embodiment provide improved signal relative to the non-invasive detector units. In one embodiment the antenna unit is connected to the processor unit wirelessly to provide a more flexible system. In one embodiment considerations such as power consumption and/or simplicity and/or cost and/or noise interference may lead to connecting the electrodes via a wire, in one embodiment preferable short and close to the body to reduce noise pickup. In one embodiment the processor unit is integrated with at least part of said antenna unit such as located on or integrated with one of the electrodes. The processor unit is in one embodiment arranged to perform one or more of the filtering operations and HRV determinations described above.

The processor unit may in principle be divided into separate subunits such as one or more of analog filtering, digital filtern, HRV determination, storage, interface and other sub-functions. However, in one embodiment the processor unit comprises all sub-functions.

19 9 DK 176955 B1

In one embodiment the system further comprises a local status unit such as a beeper or a light emitter. Such a local status unit may provide a convenient indicator for a subject carrying the system also called the user and/or clinical personal or others monitoring the user. In one embodiment the local status unit is arranged to indicate one or more of “poor signal”, “low battery”, “fault”, “seek help", “update required", “alarm", “all clear”, “please connect", “initialization required”. Such indication may be critical for the subject carrying the system so that action may be taken to e.g. maintain constant monitoring or seeking immediate help. The indicator may also provide comfort in knowing that the system indicates “Ok".

In one embodiment the interface of the system comprises a wireless interface, such as WiFi, Bluetooth, IR, UMTS, GSM, GPRS. In principle the interface may be wired instead or a combination thereof. Such an interface may serve one or more of several functions. It may be applied by user and/or clinical personal or others monitoring the user to provide status reports, upgrade firmware, during initialization or other similar functions. In one embodiment data may transferred locally, i.e. to a computer, and from there transferred to an external location such as a clinical setting for analysis and/or alert.

However, the interface may also be implemented to communicate directly, e.g. via cell phone technologies so that for example an abnormal situation in the HRV may be reported immediately. In one embodiment the system further comprises means for localisation such as a GPS chip or GPS tracker so that the users position may be tracked in the event of an event that requires immediate attention. As will be apparent to the skilled person from these examples there exist a vast number of design possibilities and functionalities that may be incorporated into the system so that the information relating to the HRV may be relayed to the user and/or supervising (e.g. health care) personnel. Accordingly, in one embodiment the system further comprises a reporting function where data stored by the signal processor unit is uploaded via said interface. In one embodiment the system further comprises a full ECG collection function comprising filtering the ECG signal according to clinical standards and storing the full signal. In this way a full ECG signal may be recorded, i.e. not just the HRV, where filter setting etc. have been set to abide 20 DK 176955 B1 by clinical requirements. Such a function may be useful for a plethora of applications. In one example an abnormal HRV is present but only for a brief period. If the system is capable of collecting a full ECG at that time this recoding may provide more insight into the cause of the abnormal HRV. In another example (which may be combined with the first) a full ECG is collected a regular intervals for diagnostic and/or monitoring purposes.

In one embodiment the system comprises an initialization function comprising one or more of programming the signal processor unit, receiving data via said interface, reporting date via said interface. Such initialization may comprise one or more of several functions. Firstly, it may comprise of programming the system for the type of user. It may comprise programming the system with data that has been found to be clinically relevant and it may comprise programming the system with communication functionalities such as those described above e g. when to report, what indications to give etc. In one embodiment the initialization comprises assisting the user in correct setup, e g. placing of electrodes to obtain a sufficient signal, i.e. in one embodiment the initialization function comprises measuring ECG data and reporting said data, measuring ECG data and reporting quality of data, measuring ECG data and reporting quality of data, measuring ECG data and recommended changes to antenna unit. The initialization is in one embodiment performed when the system is wired or wireless communication with a computer which in one embodiment guides the user. In one embodiment the initialization comprises adjusting the filter functions of the processing unit and the like to adapt the system to the particulars of the users. In one embodiment the initialization function comprises programming the system, such as setting the filtering, setting triggering values e.g. for the peak detector etc., behaviour of the local status unit, setting measurement behaviour (e.g. when to record a full ECG), setting reporting behaviour, setting variables for the individual, setting other measurement setting.

In one embodiment the invention relates to a diagnostic aid comprising a system according to the invention suitable for providing and/or monitoring the HRV of a subject and in one embodiment is further comprises comparison 21 DK 176955 B1 with at least one reference value so as to indicate the status of the HRV of the user, i.e. the subject being measured by the system.

In one embodiment the invention relates to a method of monitoring a subject comprising measuring a HRV from said subject via a method and/or system according to the invention. Similarly to the diagnostic aid one embodiment of the method comprises relating said HRV to one or more references values.

In one embodiment the method further comprises utilising said HRV to determine a recommend an action comprising one or more of a. administer a medication b. ask the subject to change behaviour, such as rest c. contact help.

The method may also be applied to utilized the measured HRV to let the system determine when to perform an action to a. administer medication via an automated release in communication with the system b. report

c. Record full ECG

In one embodiment the method further comprises logging the development in the HRV, such as based on said relation to at least one reference value. In one embodiment the reference value(s) is an individual value such as determined in a clinical setting or by the system. The latter may in one embodiment be determined via the initialization and/or as a result of a monitoring period. The value(s) may in one embodiment also be adjusted during use. In one embodiment said reference values are applied to determine one or more states of the subject relating to one or more selected from the group of stress, labour, glucose level, rejection of a transplanted heart, infection or over and/or under dosage of beta-blockers, antidepressants, sedatives.

X

22 DK 176955 B1

In one embodiment the subject or user is a mamma! and in one embodiment non-mammal. In one embodiment a mammal selected from the group of human, canine, cat, cattle, horse.

Claims (9)

A method of determining HRV from an individual comprising a. Sampling an ECG signal, r (t), from the individual to obtain a sampled ECG signal, which sampled ECG signal comprises one or more QRS complexes , b. filtering at least a portion of said ECG signal and / or the sampled ECG signal to obtain a filtered ECG signal having one or more filtered QRS complexes, c. using a locator function for determining a time-dependent position of one or more QRS complexes of the filtered ECG signal, d. using the time position or positions in a calculation of HRV for the individual, wherein the locator function comprises a calculation of the center of gravity of the filtered QRS complex or a part thereof, such as the R peak value. A method according to any one of the preceding claims, wherein the sampled ECG signal comprises samples representing 400 samples per second. second or less, such as 300 samples per second. per second or less, such as 200 samples per second. second or less, such as 150 samples per second. second or less, such as 100 samples per second. second or less, such as 75 samples per second. second or less, such as 50 samples per second. second or less, such as 35 samples per second. second or less, such as 25 samples per second. second or less. A method according to any one of the preceding claims, further comprising taking the absolute value of the ECG signal, the sampled ECG signal, the filtered ECG signal and / or intermediate values thereof, A method according to any one of the preceding claims, wherein the filtering substantially removes frequencies ranging from an upper frequency to a frequency less than 125 Hz, such as to a frequency less than 100 Hz, such as to a frequency less than 75 Hz, such as for a frequency less than 50 Hz, such as for a frequency less than 45 Hz, such as for a frequency less than 40 Hz, such as for a frequency less than 35 Hz, such as for a frequency less than 30 Hz, DK 176955 B1 such as for a frequency less than 25 Hz, such as for a frequency less than 20 Hz, such as for a frequency less than 15 Hz, such as for a frequency less than 10 Hz. The method of claim 5, wherein the upper frequency is 125 Hz or higher, such as 500 Hz or higher. A method according to any one of the preceding claims, wherein the filtering comprises using a baseline filter suitable for substantially removing baseline migration. The method of claim 0, wherein the baseline filter comprises a high pass filter having a cutoff at a frequency less than 40 Hz, such as less than 30 Hz, such as less than 20 Hz, such as less than 10 Hz, such as less than 5 Hz, such as less than 1 Hz. A method according to any one of the preceding claims, wherein the filtering comprises using a mains hum filter suitable for removing noise contributions from the supply network. A method according to any one of the preceding claims, wherein HRV is determined with a resolution of 10 ms or less, such as 1 ms or less, such as 500 ps or less, such as 250 ps or less, such as 100 ps or less, such as 50 ps or less, such as 10 ps or less. A method according to any one of the preceding claims, wherein the method further comprises applying a peak value detector to said ECG signal, the sampled ECG signal, the filtered ECG signal and / or intermediate values thereof, enabling the locator function to ignore substantially all of the filtered ECG signal that does not include one or more filtered QRS complexes.